Upload folder using huggingface_hub
Browse filesThis view is limited to 50 files because it contains too many changes.
See raw diff
- README.md +64 -91
- checkpoint-1080/README.md +202 -0
- checkpoint-1080/adapter_config.json +29 -0
- checkpoint-1080/adapter_model.safetensors +3 -0
- checkpoint-1080/optimizer.pt +3 -0
- checkpoint-1080/rng_state.pth +3 -0
- checkpoint-1080/scheduler.pt +3 -0
- checkpoint-1080/special_tokens_map.json +24 -0
- checkpoint-1080/tokenizer.json +0 -0
- checkpoint-1080/tokenizer.model +3 -0
- checkpoint-1080/tokenizer_config.json +49 -0
- checkpoint-1080/trainer_state.json +2217 -0
- checkpoint-1080/training_args.bin +3 -0
- checkpoint-1170/README.md +202 -0
- checkpoint-1170/adapter_config.json +29 -0
- checkpoint-1170/adapter_model.safetensors +3 -0
- checkpoint-1170/optimizer.pt +3 -0
- checkpoint-1170/rng_state.pth +3 -0
- checkpoint-1170/scheduler.pt +3 -0
- checkpoint-1170/special_tokens_map.json +24 -0
- checkpoint-1170/tokenizer.json +0 -0
- checkpoint-1170/tokenizer.model +3 -0
- checkpoint-1170/tokenizer_config.json +49 -0
- checkpoint-1170/trainer_state.json +2400 -0
- checkpoint-1170/training_args.bin +3 -0
- checkpoint-1260/README.md +202 -0
- checkpoint-1260/adapter_config.json +29 -0
- checkpoint-1260/adapter_model.safetensors +3 -0
- checkpoint-1260/optimizer.pt +3 -0
- checkpoint-1260/rng_state.pth +3 -0
- checkpoint-1260/scheduler.pt +3 -0
- checkpoint-1260/special_tokens_map.json +24 -0
- checkpoint-1260/tokenizer.json +0 -0
- checkpoint-1260/tokenizer.model +3 -0
- checkpoint-1260/tokenizer_config.json +49 -0
- checkpoint-1260/trainer_state.json +2583 -0
- checkpoint-1260/training_args.bin +3 -0
- checkpoint-1350/README.md +202 -0
- checkpoint-1350/adapter_config.json +29 -0
- checkpoint-1350/adapter_model.safetensors +3 -0
- checkpoint-1350/optimizer.pt +3 -0
- checkpoint-1350/rng_state.pth +3 -0
- checkpoint-1350/scheduler.pt +3 -0
- checkpoint-1350/special_tokens_map.json +24 -0
- checkpoint-1350/tokenizer.json +0 -0
- checkpoint-1350/tokenizer.model +3 -0
- checkpoint-1350/tokenizer_config.json +49 -0
- checkpoint-1350/trainer_state.json +2766 -0
- checkpoint-1350/training_args.bin +3 -0
- checkpoint-1440/README.md +202 -0
README.md
CHANGED
@@ -1,118 +1,91 @@
|
|
1 |
---
|
2 |
-
|
|
|
3 |
tags:
|
4 |
-
-
|
5 |
-
-
|
6 |
-
- generated_from_trainer
|
7 |
base_model: Weni/WeniGPT-Agents-Mistral-1.0.6-SFT-merged
|
8 |
model-index:
|
9 |
-
- name: WeniGPT-Agents-Mistral-1.0.6-SFT-1.0.7-DPO
|
10 |
results: []
|
|
|
11 |
---
|
12 |
|
13 |
-
|
14 |
-
should probably proofread and complete it, then remove this comment. -->
|
15 |
|
16 |
-
|
|
|
17 |
|
18 |
-
This model is a fine-tuned version of [Weni/WeniGPT-Agents-Mistral-1.0.6-SFT-merged](https://huggingface.co/Weni/WeniGPT-Agents-Mistral-1.0.6-SFT-merged) on an unknown dataset.
|
19 |
It achieves the following results on the evaluation set:
|
20 |
-
-
|
21 |
-
- Rewards/chosen: 1.3984
|
22 |
-
- Rewards/rejected: -6.4179
|
23 |
-
- Rewards/accuracies: 0.9643
|
24 |
-
- Rewards/margins: 7.8163
|
25 |
-
- Logps/rejected: -264.5786
|
26 |
-
- Logps/chosen: -189.8816
|
27 |
-
- Logits/rejected: -1.8496
|
28 |
-
- Logits/chosen: -1.8101
|
29 |
|
30 |
-
##
|
31 |
|
32 |
-
|
33 |
|
34 |
-
##
|
35 |
|
36 |
-
|
37 |
|
38 |
-
|
|
|
|
|
|
|
|
|
39 |
|
40 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
41 |
|
42 |
-
|
|
|
|
|
|
|
43 |
|
44 |
### Training hyperparameters
|
45 |
|
46 |
The following hyperparameters were used during training:
|
47 |
- learning_rate: 5e-06
|
48 |
-
-
|
49 |
-
-
|
50 |
-
-
|
51 |
-
-
|
52 |
-
-
|
53 |
-
-
|
54 |
-
-
|
55 |
-
-
|
|
|
|
|
56 |
|
57 |
### Training results
|
58 |
|
59 |
-
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
60 |
-
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
61 |
-
| 0.6781 | 0.12 | 30 | 0.6762 | 0.0504 | 0.0157 | 0.75 | 0.0347 | -243.1332 | -194.3750 | -1.8308 | -1.7951 |
|
62 |
-
| 0.5918 | 0.24 | 60 | 0.5998 | 0.2476 | 0.0383 | 0.7857 | 0.2093 | -243.0578 | -193.7174 | -1.8333 | -1.7975 |
|
63 |
-
| 0.4932 | 0.37 | 90 | 0.5072 | 0.5622 | 0.0680 | 0.8214 | 0.4942 | -242.9590 | -192.6691 | -1.8364 | -1.8004 |
|
64 |
-
| 0.4391 | 0.49 | 120 | 0.4336 | 0.9734 | 0.1121 | 0.7857 | 0.8613 | -242.8120 | -191.2982 | -1.8413 | -1.8051 |
|
65 |
-
| 0.3208 | 0.61 | 150 | 0.3933 | 1.3961 | 0.0824 | 0.7857 | 1.3137 | -242.9110 | -189.8893 | -1.8492 | -1.8130 |
|
66 |
-
| 0.3215 | 0.73 | 180 | 0.3756 | 1.8483 | 0.0151 | 0.7857 | 1.8332 | -243.1354 | -188.3820 | -1.8562 | -1.8194 |
|
67 |
-
| 0.0817 | 0.86 | 210 | 0.3835 | 2.3139 | -0.1849 | 0.7857 | 2.4989 | -243.8021 | -186.8299 | -1.8641 | -1.8266 |
|
68 |
-
| 0.137 | 0.98 | 240 | 0.4132 | 2.5979 | -0.5021 | 0.75 | 3.1001 | -244.8594 | -185.8831 | -1.8722 | -1.8343 |
|
69 |
-
| 0.0997 | 1.1 | 270 | 0.4657 | 2.7384 | -1.0053 | 0.75 | 3.7438 | -246.5367 | -185.4148 | -1.8816 | -1.8430 |
|
70 |
-
| 0.0432 | 1.22 | 300 | 0.5011 | 2.7041 | -1.4771 | 0.75 | 4.1812 | -248.1093 | -185.5293 | -1.8884 | -1.8495 |
|
71 |
-
| 0.1819 | 1.35 | 330 | 0.4785 | 2.7004 | -1.8249 | 0.75 | 4.5253 | -249.2688 | -185.5418 | -1.8878 | -1.8487 |
|
72 |
-
| 0.0169 | 1.47 | 360 | 0.4872 | 2.6643 | -2.1577 | 0.75 | 4.8220 | -250.3781 | -185.6619 | -1.8907 | -1.8510 |
|
73 |
-
| 0.235 | 1.59 | 390 | 0.4886 | 2.6565 | -2.3834 | 0.75 | 5.0399 | -251.1302 | -185.6880 | -1.8930 | -1.8532 |
|
74 |
-
| 0.7551 | 1.71 | 420 | 0.4380 | 2.7229 | -2.3468 | 0.75 | 5.0697 | -251.0082 | -185.4665 | -1.8921 | -1.8527 |
|
75 |
-
| 0.134 | 1.84 | 450 | 0.4383 | 2.6666 | -2.5566 | 0.75 | 5.2232 | -251.7077 | -185.6543 | -1.8925 | -1.8531 |
|
76 |
-
| 0.0662 | 1.96 | 480 | 0.4448 | 2.5586 | -2.9192 | 0.75 | 5.4778 | -252.9164 | -186.0143 | -1.8964 | -1.8569 |
|
77 |
-
| 0.1093 | 2.08 | 510 | 0.4262 | 2.5211 | -3.0726 | 0.75 | 5.5937 | -253.4277 | -186.1394 | -1.8955 | -1.8561 |
|
78 |
-
| 0.1557 | 2.2 | 540 | 0.4264 | 2.3694 | -3.4198 | 0.75 | 5.7892 | -254.5848 | -186.6449 | -1.8965 | -1.8566 |
|
79 |
-
| 0.0962 | 2.33 | 570 | 0.4182 | 2.2640 | -3.7076 | 0.75 | 5.9716 | -255.5444 | -186.9964 | -1.8978 | -1.8582 |
|
80 |
-
| 0.0437 | 2.45 | 600 | 0.3824 | 2.2618 | -3.7757 | 0.75 | 6.0375 | -255.7713 | -187.0037 | -1.8933 | -1.8534 |
|
81 |
-
| 0.0278 | 2.57 | 630 | 0.3571 | 2.3503 | -3.7557 | 0.8571 | 6.1060 | -255.7046 | -186.7086 | -1.8932 | -1.8536 |
|
82 |
-
| 0.2399 | 2.69 | 660 | 0.3313 | 2.3025 | -3.9256 | 0.8571 | 6.2281 | -256.2710 | -186.8678 | -1.8909 | -1.8512 |
|
83 |
-
| 0.039 | 2.82 | 690 | 0.3131 | 2.2138 | -4.1650 | 0.8929 | 6.3789 | -257.0691 | -187.1635 | -1.8906 | -1.8510 |
|
84 |
-
| 0.3389 | 2.94 | 720 | 0.2763 | 2.2605 | -4.2160 | 0.8929 | 6.4765 | -257.2390 | -187.0079 | -1.8873 | -1.8480 |
|
85 |
-
| 0.0154 | 3.06 | 750 | 0.2704 | 2.2526 | -4.3017 | 0.8929 | 6.5544 | -257.5247 | -187.0342 | -1.8862 | -1.8470 |
|
86 |
-
| 0.021 | 3.18 | 780 | 0.2422 | 2.2548 | -4.3438 | 0.8929 | 6.5986 | -257.6650 | -187.0270 | -1.8838 | -1.8448 |
|
87 |
-
| 0.0614 | 3.31 | 810 | 0.2144 | 2.2331 | -4.4495 | 0.8929 | 6.6826 | -258.0172 | -187.0992 | -1.8805 | -1.8417 |
|
88 |
-
| 0.0529 | 3.43 | 840 | 0.2121 | 2.1562 | -4.6740 | 0.8929 | 6.8302 | -258.7657 | -187.3555 | -1.8809 | -1.8423 |
|
89 |
-
| 0.001 | 3.55 | 870 | 0.2092 | 2.1034 | -4.8454 | 0.8929 | 6.9487 | -259.3368 | -187.5317 | -1.8799 | -1.8410 |
|
90 |
-
| 0.0284 | 3.67 | 900 | 0.2006 | 1.9814 | -5.1388 | 0.8929 | 7.1202 | -260.3150 | -187.9384 | -1.8760 | -1.8366 |
|
91 |
-
| 0.0744 | 3.8 | 930 | 0.1813 | 1.9437 | -5.2351 | 0.8929 | 7.1788 | -260.6358 | -188.0639 | -1.8733 | -1.8339 |
|
92 |
-
| 0.091 | 3.92 | 960 | 0.1722 | 1.8333 | -5.4335 | 0.8929 | 7.2668 | -261.2973 | -188.4319 | -1.8707 | -1.8313 |
|
93 |
-
| 0.3504 | 4.04 | 990 | 0.1487 | 1.8678 | -5.3589 | 0.9286 | 7.2268 | -261.0488 | -188.3168 | -1.8672 | -1.8279 |
|
94 |
-
| 0.0071 | 4.16 | 1020 | 0.1403 | 1.7989 | -5.5185 | 0.9286 | 7.3173 | -261.5805 | -188.5468 | -1.8637 | -1.8243 |
|
95 |
-
| 0.0131 | 4.29 | 1050 | 0.1312 | 1.8050 | -5.5495 | 0.9286 | 7.3545 | -261.6841 | -188.5262 | -1.8616 | -1.8222 |
|
96 |
-
| 0.0868 | 4.41 | 1080 | 0.1210 | 1.7626 | -5.6284 | 0.9286 | 7.3911 | -261.9471 | -188.6675 | -1.8587 | -1.8195 |
|
97 |
-
| 0.0041 | 4.53 | 1110 | 0.1206 | 1.6865 | -5.7780 | 0.9286 | 7.4645 | -262.4456 | -188.9213 | -1.8566 | -1.8173 |
|
98 |
-
| 0.0107 | 4.65 | 1140 | 0.1178 | 1.6370 | -5.8895 | 0.9643 | 7.5266 | -262.8174 | -189.0862 | -1.8563 | -1.8171 |
|
99 |
-
| 0.0084 | 4.78 | 1170 | 0.1123 | 1.6107 | -5.9365 | 0.9643 | 7.5471 | -262.9738 | -189.1741 | -1.8552 | -1.8159 |
|
100 |
-
| 0.0049 | 4.9 | 1200 | 0.1083 | 1.5710 | -6.0495 | 0.9643 | 7.6206 | -263.3507 | -189.3061 | -1.8545 | -1.8151 |
|
101 |
-
| 0.0746 | 5.02 | 1230 | 0.1034 | 1.5328 | -6.1286 | 0.9643 | 7.6614 | -263.6144 | -189.4336 | -1.8535 | -1.8140 |
|
102 |
-
| 0.0091 | 5.14 | 1260 | 0.1031 | 1.4764 | -6.2562 | 0.9643 | 7.7327 | -264.0397 | -189.6215 | -1.8531 | -1.8136 |
|
103 |
-
| 0.0526 | 5.27 | 1290 | 0.0997 | 1.4526 | -6.3037 | 0.9643 | 7.7564 | -264.1981 | -189.7009 | -1.8528 | -1.8133 |
|
104 |
-
| 0.0316 | 5.39 | 1320 | 0.0965 | 1.4471 | -6.3114 | 0.9643 | 7.7585 | -264.2236 | -189.7192 | -1.8517 | -1.8124 |
|
105 |
-
| 0.0249 | 5.51 | 1350 | 0.0950 | 1.4370 | -6.3384 | 0.9643 | 7.7755 | -264.3138 | -189.7529 | -1.8509 | -1.8115 |
|
106 |
-
| 0.2078 | 5.63 | 1380 | 0.0937 | 1.4141 | -6.3790 | 0.9643 | 7.7931 | -264.4489 | -189.8293 | -1.8504 | -1.8111 |
|
107 |
-
| 0.013 | 5.76 | 1410 | 0.0926 | 1.4237 | -6.3666 | 0.9643 | 7.7902 | -264.4076 | -189.7974 | -1.8498 | -1.8103 |
|
108 |
-
| 0.0194 | 5.88 | 1440 | 0.0923 | 1.3984 | -6.4179 | 0.9643 | 7.8163 | -264.5786 | -189.8816 | -1.8496 | -1.8101 |
|
109 |
-
| 0.0111 | 6.0 | 1470 | 0.0919 | 1.3959 | -6.4219 | 0.9643 | 7.8179 | -264.5919 | -189.8898 | -1.8495 | -1.8100 |
|
110 |
-
|
111 |
-
|
112 |
### Framework versions
|
113 |
|
114 |
-
-
|
115 |
-
-
|
116 |
-
-
|
117 |
-
-
|
118 |
-
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
+
license: mit
|
3 |
+
library_name: "trl"
|
4 |
tags:
|
5 |
+
- DPO
|
6 |
+
- WeniGPT
|
|
|
7 |
base_model: Weni/WeniGPT-Agents-Mistral-1.0.6-SFT-merged
|
8 |
model-index:
|
9 |
+
- name: Weni/WeniGPT-Agents-Mistral-1.0.6-SFT-1.0.7-DPO
|
10 |
results: []
|
11 |
+
language: ['pt']
|
12 |
---
|
13 |
|
14 |
+
# Weni/WeniGPT-Agents-Mistral-1.0.6-SFT-1.0.7-DPO
|
|
|
15 |
|
16 |
+
This model is a fine-tuned version of [Weni/WeniGPT-Agents-Mistral-1.0.6-SFT-merged] on the dataset Weni/wenigpt-agent-dpo-1.0.0 with the DPO trainer. It is part of the WeniGPT project for [Weni](https://weni.ai/).
|
17 |
+
Description: Experiment on DPO with other hyperparameters and best SFT model of WeniGPT
|
18 |
|
|
|
19 |
It achieves the following results on the evaluation set:
|
20 |
+
{'eval_loss': 0.09226309508085251, 'eval_runtime': 26.2286, 'eval_samples_per_second': 1.068, 'eval_steps_per_second': 1.068, 'eval_rewards/chosen': 1.398409128189087, 'eval_rewards/rejected': -6.417877197265625, 'eval_rewards/accuracies': 0.9642857313156128, 'eval_rewards/margins': 7.816287040710449, 'eval_logps/rejected': -264.57855224609375, 'eval_logps/chosen': -189.88156127929688, 'eval_logits/rejected': -1.8495510816574097, 'eval_logits/chosen': -1.810078740119934, 'epoch': 6.0}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
|
22 |
+
## Intended uses & limitations
|
23 |
|
24 |
+
This model has not been trained to avoid specific intructions.
|
25 |
|
26 |
+
## Training procedure
|
27 |
|
28 |
+
Finetuning was done on the model Weni/WeniGPT-Agents-Mistral-1.0.6-SFT-merged with the following prompt:
|
29 |
|
30 |
+
```
|
31 |
+
---------------------
|
32 |
+
System_prompt:
|
33 |
+
Agora você se chama {name}, você é {occupation} e seu objetivo é {chatbot_goal}. O adjetivo que mais define a sua personalidade é {adjective} e você se comporta da seguinte forma:
|
34 |
+
{instructions_formatted}
|
35 |
|
36 |
+
{context_statement}
|
37 |
+
|
38 |
+
Lista de requisitos:
|
39 |
+
- Responda de forma natural, mas nunca fale sobre um assunto fora do contexto.
|
40 |
+
- Nunca traga informações do seu próprio conhecimento.
|
41 |
+
- Repito é crucial que você responda usando apenas informações do contexto.
|
42 |
+
- Nunca mencione o contexto fornecido.
|
43 |
+
- Nunca mencione a pergunta fornecida.
|
44 |
+
- Gere a resposta mais útil possível para a pergunta usando informações do conexto acima.
|
45 |
+
- Nunca elabore sobre o porque e como você fez a tarefa, apenas responda.
|
46 |
|
47 |
+
|
48 |
+
---------------------
|
49 |
+
|
50 |
+
```
|
51 |
|
52 |
### Training hyperparameters
|
53 |
|
54 |
The following hyperparameters were used during training:
|
55 |
- learning_rate: 5e-06
|
56 |
+
- per_device_train_batch_size: 1
|
57 |
+
- per_device_eval_batch_size: 1
|
58 |
+
- gradient_accumulation_steps: 1
|
59 |
+
- num_gpus: 1
|
60 |
+
- total_train_batch_size: 1
|
61 |
+
- optimizer: AdamW
|
62 |
+
- lr_scheduler_type: cosine
|
63 |
+
- num_steps: 1470
|
64 |
+
- quantization_type: bitsandbytes
|
65 |
+
- LoRA: ("\n - bits: 4\n - use_exllama: True\n - device_map: auto\n - use_cache: False\n - lora_r: 8\n - lora_alpha: 16\n - lora_dropout: 0.05\n - bias: none\n - target_modules: ['v_proj', 'q_proj']\n - task_type: CAUSAL_LM",)
|
66 |
|
67 |
### Training results
|
68 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
69 |
### Framework versions
|
70 |
|
71 |
+
- transformers==4.38.2
|
72 |
+
- datasets==2.18.0
|
73 |
+
- peft==0.10.0
|
74 |
+
- safetensors==0.4.2
|
75 |
+
- evaluate==0.4.1
|
76 |
+
- bitsandbytes==0.43
|
77 |
+
- huggingface_hub==0.22.2
|
78 |
+
- seqeval==1.2.2
|
79 |
+
- optimum==1.18.1
|
80 |
+
- auto-gptq==0.7.1
|
81 |
+
- gpustat==1.1.1
|
82 |
+
- deepspeed==0.14.0
|
83 |
+
- wandb==0.16.6
|
84 |
+
- trl==0.8.1
|
85 |
+
- accelerate==0.29.2
|
86 |
+
- coloredlogs==15.0.1
|
87 |
+
- traitlets==5.14.2
|
88 |
+
- autoawq@https://github.com/casper-hansen/AutoAWQ/releases/download/v0.2.4/autoawq-0.2.4+cu118-cp310-cp310-linux_x86_64.whl
|
89 |
+
|
90 |
+
### Hardware
|
91 |
+
- Cloud provided: runpod.io
|
checkpoint-1080/README.md
ADDED
@@ -0,0 +1,202 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
library_name: peft
|
3 |
+
base_model: Weni/WeniGPT-Agents-Mistral-1.0.6-SFT-merged
|
4 |
+
---
|
5 |
+
|
6 |
+
# Model Card for Model ID
|
7 |
+
|
8 |
+
<!-- Provide a quick summary of what the model is/does. -->
|
9 |
+
|
10 |
+
|
11 |
+
|
12 |
+
## Model Details
|
13 |
+
|
14 |
+
### Model Description
|
15 |
+
|
16 |
+
<!-- Provide a longer summary of what this model is. -->
|
17 |
+
|
18 |
+
|
19 |
+
|
20 |
+
- **Developed by:** [More Information Needed]
|
21 |
+
- **Funded by [optional]:** [More Information Needed]
|
22 |
+
- **Shared by [optional]:** [More Information Needed]
|
23 |
+
- **Model type:** [More Information Needed]
|
24 |
+
- **Language(s) (NLP):** [More Information Needed]
|
25 |
+
- **License:** [More Information Needed]
|
26 |
+
- **Finetuned from model [optional]:** [More Information Needed]
|
27 |
+
|
28 |
+
### Model Sources [optional]
|
29 |
+
|
30 |
+
<!-- Provide the basic links for the model. -->
|
31 |
+
|
32 |
+
- **Repository:** [More Information Needed]
|
33 |
+
- **Paper [optional]:** [More Information Needed]
|
34 |
+
- **Demo [optional]:** [More Information Needed]
|
35 |
+
|
36 |
+
## Uses
|
37 |
+
|
38 |
+
<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
|
39 |
+
|
40 |
+
### Direct Use
|
41 |
+
|
42 |
+
<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
|
43 |
+
|
44 |
+
[More Information Needed]
|
45 |
+
|
46 |
+
### Downstream Use [optional]
|
47 |
+
|
48 |
+
<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
|
49 |
+
|
50 |
+
[More Information Needed]
|
51 |
+
|
52 |
+
### Out-of-Scope Use
|
53 |
+
|
54 |
+
<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
|
55 |
+
|
56 |
+
[More Information Needed]
|
57 |
+
|
58 |
+
## Bias, Risks, and Limitations
|
59 |
+
|
60 |
+
<!-- This section is meant to convey both technical and sociotechnical limitations. -->
|
61 |
+
|
62 |
+
[More Information Needed]
|
63 |
+
|
64 |
+
### Recommendations
|
65 |
+
|
66 |
+
<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
|
67 |
+
|
68 |
+
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
|
69 |
+
|
70 |
+
## How to Get Started with the Model
|
71 |
+
|
72 |
+
Use the code below to get started with the model.
|
73 |
+
|
74 |
+
[More Information Needed]
|
75 |
+
|
76 |
+
## Training Details
|
77 |
+
|
78 |
+
### Training Data
|
79 |
+
|
80 |
+
<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
|
81 |
+
|
82 |
+
[More Information Needed]
|
83 |
+
|
84 |
+
### Training Procedure
|
85 |
+
|
86 |
+
<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
|
87 |
+
|
88 |
+
#### Preprocessing [optional]
|
89 |
+
|
90 |
+
[More Information Needed]
|
91 |
+
|
92 |
+
|
93 |
+
#### Training Hyperparameters
|
94 |
+
|
95 |
+
- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
|
96 |
+
|
97 |
+
#### Speeds, Sizes, Times [optional]
|
98 |
+
|
99 |
+
<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
|
100 |
+
|
101 |
+
[More Information Needed]
|
102 |
+
|
103 |
+
## Evaluation
|
104 |
+
|
105 |
+
<!-- This section describes the evaluation protocols and provides the results. -->
|
106 |
+
|
107 |
+
### Testing Data, Factors & Metrics
|
108 |
+
|
109 |
+
#### Testing Data
|
110 |
+
|
111 |
+
<!-- This should link to a Dataset Card if possible. -->
|
112 |
+
|
113 |
+
[More Information Needed]
|
114 |
+
|
115 |
+
#### Factors
|
116 |
+
|
117 |
+
<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
|
118 |
+
|
119 |
+
[More Information Needed]
|
120 |
+
|
121 |
+
#### Metrics
|
122 |
+
|
123 |
+
<!-- These are the evaluation metrics being used, ideally with a description of why. -->
|
124 |
+
|
125 |
+
[More Information Needed]
|
126 |
+
|
127 |
+
### Results
|
128 |
+
|
129 |
+
[More Information Needed]
|
130 |
+
|
131 |
+
#### Summary
|
132 |
+
|
133 |
+
|
134 |
+
|
135 |
+
## Model Examination [optional]
|
136 |
+
|
137 |
+
<!-- Relevant interpretability work for the model goes here -->
|
138 |
+
|
139 |
+
[More Information Needed]
|
140 |
+
|
141 |
+
## Environmental Impact
|
142 |
+
|
143 |
+
<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
|
144 |
+
|
145 |
+
Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
|
146 |
+
|
147 |
+
- **Hardware Type:** [More Information Needed]
|
148 |
+
- **Hours used:** [More Information Needed]
|
149 |
+
- **Cloud Provider:** [More Information Needed]
|
150 |
+
- **Compute Region:** [More Information Needed]
|
151 |
+
- **Carbon Emitted:** [More Information Needed]
|
152 |
+
|
153 |
+
## Technical Specifications [optional]
|
154 |
+
|
155 |
+
### Model Architecture and Objective
|
156 |
+
|
157 |
+
[More Information Needed]
|
158 |
+
|
159 |
+
### Compute Infrastructure
|
160 |
+
|
161 |
+
[More Information Needed]
|
162 |
+
|
163 |
+
#### Hardware
|
164 |
+
|
165 |
+
[More Information Needed]
|
166 |
+
|
167 |
+
#### Software
|
168 |
+
|
169 |
+
[More Information Needed]
|
170 |
+
|
171 |
+
## Citation [optional]
|
172 |
+
|
173 |
+
<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
|
174 |
+
|
175 |
+
**BibTeX:**
|
176 |
+
|
177 |
+
[More Information Needed]
|
178 |
+
|
179 |
+
**APA:**
|
180 |
+
|
181 |
+
[More Information Needed]
|
182 |
+
|
183 |
+
## Glossary [optional]
|
184 |
+
|
185 |
+
<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
|
186 |
+
|
187 |
+
[More Information Needed]
|
188 |
+
|
189 |
+
## More Information [optional]
|
190 |
+
|
191 |
+
[More Information Needed]
|
192 |
+
|
193 |
+
## Model Card Authors [optional]
|
194 |
+
|
195 |
+
[More Information Needed]
|
196 |
+
|
197 |
+
## Model Card Contact
|
198 |
+
|
199 |
+
[More Information Needed]
|
200 |
+
### Framework versions
|
201 |
+
|
202 |
+
- PEFT 0.10.0
|
checkpoint-1080/adapter_config.json
ADDED
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"alpha_pattern": {},
|
3 |
+
"auto_mapping": null,
|
4 |
+
"base_model_name_or_path": "Weni/WeniGPT-Agents-Mistral-1.0.6-SFT-merged",
|
5 |
+
"bias": "none",
|
6 |
+
"fan_in_fan_out": false,
|
7 |
+
"inference_mode": true,
|
8 |
+
"init_lora_weights": true,
|
9 |
+
"layer_replication": null,
|
10 |
+
"layers_pattern": null,
|
11 |
+
"layers_to_transform": null,
|
12 |
+
"loftq_config": {},
|
13 |
+
"lora_alpha": 16,
|
14 |
+
"lora_dropout": 0.05,
|
15 |
+
"megatron_config": null,
|
16 |
+
"megatron_core": "megatron.core",
|
17 |
+
"modules_to_save": null,
|
18 |
+
"peft_type": "LORA",
|
19 |
+
"r": 8,
|
20 |
+
"rank_pattern": {},
|
21 |
+
"revision": null,
|
22 |
+
"target_modules": [
|
23 |
+
"q_proj",
|
24 |
+
"v_proj"
|
25 |
+
],
|
26 |
+
"task_type": "CAUSAL_LM",
|
27 |
+
"use_dora": false,
|
28 |
+
"use_rslora": false
|
29 |
+
}
|
checkpoint-1080/adapter_model.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:21aaa9616e5325d274fa6a81cb06b3d8b840db9bc6f26f301d7e1fe8fa2af29d
|
3 |
+
size 13648432
|
checkpoint-1080/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:93532f526081b2e0e3c21860a061df2fecfd596ea04599542c1557e9f373e547
|
3 |
+
size 27370618
|
checkpoint-1080/rng_state.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3aff5959553392abd501b3f79410659c7c7572b63703f17ccfc9d3476abd5dff
|
3 |
+
size 14244
|
checkpoint-1080/scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3d4c28c6560137d660a07fd31b83b2cf8e7cead26f5c85a5b8cdd3c2f1d28898
|
3 |
+
size 1064
|
checkpoint-1080/special_tokens_map.json
ADDED
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"bos_token": {
|
3 |
+
"content": "<s>",
|
4 |
+
"lstrip": false,
|
5 |
+
"normalized": false,
|
6 |
+
"rstrip": false,
|
7 |
+
"single_word": false
|
8 |
+
},
|
9 |
+
"eos_token": {
|
10 |
+
"content": "</s>",
|
11 |
+
"lstrip": false,
|
12 |
+
"normalized": false,
|
13 |
+
"rstrip": false,
|
14 |
+
"single_word": false
|
15 |
+
},
|
16 |
+
"pad_token": "<unk>",
|
17 |
+
"unk_token": {
|
18 |
+
"content": "<unk>",
|
19 |
+
"lstrip": false,
|
20 |
+
"normalized": false,
|
21 |
+
"rstrip": false,
|
22 |
+
"single_word": false
|
23 |
+
}
|
24 |
+
}
|
checkpoint-1080/tokenizer.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-1080/tokenizer.model
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:dadfd56d766715c61d2ef780a525ab43b8e6da4de6865bda3d95fdef5e134055
|
3 |
+
size 493443
|
checkpoint-1080/tokenizer_config.json
ADDED
@@ -0,0 +1,49 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"add_bos_token": true,
|
3 |
+
"add_eos_token": false,
|
4 |
+
"added_tokens_decoder": {
|
5 |
+
"0": {
|
6 |
+
"content": "<unk>",
|
7 |
+
"lstrip": false,
|
8 |
+
"normalized": false,
|
9 |
+
"rstrip": false,
|
10 |
+
"single_word": false,
|
11 |
+
"special": true
|
12 |
+
},
|
13 |
+
"1": {
|
14 |
+
"content": "<s>",
|
15 |
+
"lstrip": false,
|
16 |
+
"normalized": false,
|
17 |
+
"rstrip": false,
|
18 |
+
"single_word": false,
|
19 |
+
"special": true
|
20 |
+
},
|
21 |
+
"2": {
|
22 |
+
"content": "</s>",
|
23 |
+
"lstrip": false,
|
24 |
+
"normalized": false,
|
25 |
+
"rstrip": false,
|
26 |
+
"single_word": false,
|
27 |
+
"special": true
|
28 |
+
}
|
29 |
+
},
|
30 |
+
"additional_special_tokens": [],
|
31 |
+
"bos_token": "<s>",
|
32 |
+
"chat_template": "{% for message in messages %}{% if message['role'] == 'user' %}{{ bos_token + '[INST] ' + message['content'] + ' [/INST]' }}{% elif message['role'] == 'system' %}{{ '<<SYS>>\\n' + message['content'] + '\\n<</SYS>>\\n\\n' }}{% elif message['role'] == 'assistant' %}{{ ' ' + message['content'] + ' ' + eos_token }}{% endif %}{% endfor %}",
|
33 |
+
"clean_up_tokenization_spaces": false,
|
34 |
+
"eos_token": "</s>",
|
35 |
+
"legacy": true,
|
36 |
+
"max_lenght": 8192,
|
37 |
+
"max_length": 8192,
|
38 |
+
"model_max_length": 1000000000000000019884624838656,
|
39 |
+
"pad_token": "<unk>",
|
40 |
+
"padding": true,
|
41 |
+
"sp_model_kwargs": {},
|
42 |
+
"spaces_between_special_tokens": false,
|
43 |
+
"stride": 0,
|
44 |
+
"tokenizer_class": "LlamaTokenizer",
|
45 |
+
"truncation_side": "right",
|
46 |
+
"truncation_strategy": "longest_first",
|
47 |
+
"unk_token": "<unk>",
|
48 |
+
"use_default_system_prompt": false
|
49 |
+
}
|
checkpoint-1080/trainer_state.json
ADDED
@@ -0,0 +1,2217 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": 0.1210150271654129,
|
3 |
+
"best_model_checkpoint": "./mistral/20-04-24-Weni-WeniGPT-Agents-Mistral-1.0.6-SFT-1.0.7-DPO_Experiment on DPO with other hyperparameters and best SFT model of WeniGPT-2_max_steps-1470_batch_1_2024-04-20_ppid_9/checkpoint-1080",
|
4 |
+
"epoch": 4.408163265306122,
|
5 |
+
"eval_steps": 30,
|
6 |
+
"global_step": 1080,
|
7 |
+
"is_hyper_param_search": false,
|
8 |
+
"is_local_process_zero": true,
|
9 |
+
"is_world_process_zero": true,
|
10 |
+
"log_history": [
|
11 |
+
{
|
12 |
+
"epoch": 0.04,
|
13 |
+
"grad_norm": 83.44550323486328,
|
14 |
+
"learning_rate": 6.666666666666667e-07,
|
15 |
+
"logits/chosen": -1.7825796604156494,
|
16 |
+
"logits/rejected": -1.7886196374893188,
|
17 |
+
"logps/chosen": -169.71481323242188,
|
18 |
+
"logps/rejected": -167.6764373779297,
|
19 |
+
"loss": 0.6937,
|
20 |
+
"rewards/accuracies": 0.20000000298023224,
|
21 |
+
"rewards/chosen": -0.0002815247280523181,
|
22 |
+
"rewards/margins": -0.0010146332206204534,
|
23 |
+
"rewards/rejected": 0.0007331084925681353,
|
24 |
+
"step": 10
|
25 |
+
},
|
26 |
+
{
|
27 |
+
"epoch": 0.08,
|
28 |
+
"grad_norm": 87.56982421875,
|
29 |
+
"learning_rate": 1.777777777777778e-06,
|
30 |
+
"logits/chosen": -1.623265027999878,
|
31 |
+
"logits/rejected": -1.6968040466308594,
|
32 |
+
"logps/chosen": -128.7312469482422,
|
33 |
+
"logps/rejected": -184.20187377929688,
|
34 |
+
"loss": 0.6912,
|
35 |
+
"rewards/accuracies": 0.5,
|
36 |
+
"rewards/chosen": 0.0030528828501701355,
|
37 |
+
"rewards/margins": 0.004004458896815777,
|
38 |
+
"rewards/rejected": -0.0009515761630609632,
|
39 |
+
"step": 20
|
40 |
+
},
|
41 |
+
{
|
42 |
+
"epoch": 0.12,
|
43 |
+
"grad_norm": 113.4741439819336,
|
44 |
+
"learning_rate": 2.888888888888889e-06,
|
45 |
+
"logits/chosen": -1.7152118682861328,
|
46 |
+
"logits/rejected": -1.758681058883667,
|
47 |
+
"logps/chosen": -146.91494750976562,
|
48 |
+
"logps/rejected": -213.5966339111328,
|
49 |
+
"loss": 0.6781,
|
50 |
+
"rewards/accuracies": 1.0,
|
51 |
+
"rewards/chosen": 0.03251449763774872,
|
52 |
+
"rewards/margins": 0.030697176232933998,
|
53 |
+
"rewards/rejected": 0.0018173219868913293,
|
54 |
+
"step": 30
|
55 |
+
},
|
56 |
+
{
|
57 |
+
"epoch": 0.12,
|
58 |
+
"eval_logits/chosen": -1.795080304145813,
|
59 |
+
"eval_logits/rejected": -1.8307874202728271,
|
60 |
+
"eval_logps/chosen": -194.3749542236328,
|
61 |
+
"eval_logps/rejected": -243.1332244873047,
|
62 |
+
"eval_loss": 0.6762310266494751,
|
63 |
+
"eval_rewards/accuracies": 0.75,
|
64 |
+
"eval_rewards/chosen": 0.05039369314908981,
|
65 |
+
"eval_rewards/margins": 0.03467999026179314,
|
66 |
+
"eval_rewards/rejected": 0.01571369729936123,
|
67 |
+
"eval_runtime": 26.2365,
|
68 |
+
"eval_samples_per_second": 1.067,
|
69 |
+
"eval_steps_per_second": 1.067,
|
70 |
+
"step": 30
|
71 |
+
},
|
72 |
+
{
|
73 |
+
"epoch": 0.16,
|
74 |
+
"grad_norm": 50.695308685302734,
|
75 |
+
"learning_rate": 4.000000000000001e-06,
|
76 |
+
"logits/chosen": -1.9357620477676392,
|
77 |
+
"logits/rejected": -1.9375683069229126,
|
78 |
+
"logps/chosen": -183.8218231201172,
|
79 |
+
"logps/rejected": -115.80632019042969,
|
80 |
+
"loss": 0.6768,
|
81 |
+
"rewards/accuracies": 0.699999988079071,
|
82 |
+
"rewards/chosen": 0.056698836386203766,
|
83 |
+
"rewards/margins": 0.03413978964090347,
|
84 |
+
"rewards/rejected": 0.02255905233323574,
|
85 |
+
"step": 40
|
86 |
+
},
|
87 |
+
{
|
88 |
+
"epoch": 0.2,
|
89 |
+
"grad_norm": 71.6834716796875,
|
90 |
+
"learning_rate": 4.996491228070176e-06,
|
91 |
+
"logits/chosen": -1.7773185968399048,
|
92 |
+
"logits/rejected": -1.8331743478775024,
|
93 |
+
"logps/chosen": -167.78533935546875,
|
94 |
+
"logps/rejected": -227.0582733154297,
|
95 |
+
"loss": 0.6205,
|
96 |
+
"rewards/accuracies": 0.8999999761581421,
|
97 |
+
"rewards/chosen": 0.1449195146560669,
|
98 |
+
"rewards/margins": 0.15578031539916992,
|
99 |
+
"rewards/rejected": -0.010860783979296684,
|
100 |
+
"step": 50
|
101 |
+
},
|
102 |
+
{
|
103 |
+
"epoch": 0.24,
|
104 |
+
"grad_norm": 93.83502197265625,
|
105 |
+
"learning_rate": 4.96140350877193e-06,
|
106 |
+
"logits/chosen": -1.7739536762237549,
|
107 |
+
"logits/rejected": -1.8414790630340576,
|
108 |
+
"logps/chosen": -153.57481384277344,
|
109 |
+
"logps/rejected": -194.3485870361328,
|
110 |
+
"loss": 0.5918,
|
111 |
+
"rewards/accuracies": 0.8999999761581421,
|
112 |
+
"rewards/chosen": 0.2464807778596878,
|
113 |
+
"rewards/margins": 0.22958576679229736,
|
114 |
+
"rewards/rejected": 0.016895027831196785,
|
115 |
+
"step": 60
|
116 |
+
},
|
117 |
+
{
|
118 |
+
"epoch": 0.24,
|
119 |
+
"eval_logits/chosen": -1.7974774837493896,
|
120 |
+
"eval_logits/rejected": -1.8332849740982056,
|
121 |
+
"eval_logps/chosen": -193.71743774414062,
|
122 |
+
"eval_logps/rejected": -243.05783081054688,
|
123 |
+
"eval_loss": 0.5998325943946838,
|
124 |
+
"eval_rewards/accuracies": 0.7857142686843872,
|
125 |
+
"eval_rewards/chosen": 0.24764502048492432,
|
126 |
+
"eval_rewards/margins": 0.20930658280849457,
|
127 |
+
"eval_rewards/rejected": 0.03833846375346184,
|
128 |
+
"eval_runtime": 26.2349,
|
129 |
+
"eval_samples_per_second": 1.067,
|
130 |
+
"eval_steps_per_second": 1.067,
|
131 |
+
"step": 60
|
132 |
+
},
|
133 |
+
{
|
134 |
+
"epoch": 0.29,
|
135 |
+
"grad_norm": 109.71602630615234,
|
136 |
+
"learning_rate": 4.926315789473685e-06,
|
137 |
+
"logits/chosen": -1.8950660228729248,
|
138 |
+
"logits/rejected": -1.9148075580596924,
|
139 |
+
"logps/chosen": -169.89028930664062,
|
140 |
+
"logps/rejected": -177.87301635742188,
|
141 |
+
"loss": 0.5992,
|
142 |
+
"rewards/accuracies": 0.800000011920929,
|
143 |
+
"rewards/chosen": 0.15271134674549103,
|
144 |
+
"rewards/margins": 0.21031668782234192,
|
145 |
+
"rewards/rejected": -0.0576053187251091,
|
146 |
+
"step": 70
|
147 |
+
},
|
148 |
+
{
|
149 |
+
"epoch": 0.33,
|
150 |
+
"grad_norm": 90.3881607055664,
|
151 |
+
"learning_rate": 4.8912280701754386e-06,
|
152 |
+
"logits/chosen": -1.798600196838379,
|
153 |
+
"logits/rejected": -1.855521559715271,
|
154 |
+
"logps/chosen": -160.3217315673828,
|
155 |
+
"logps/rejected": -247.6388397216797,
|
156 |
+
"loss": 0.5551,
|
157 |
+
"rewards/accuracies": 0.699999988079071,
|
158 |
+
"rewards/chosen": 0.29141485691070557,
|
159 |
+
"rewards/margins": 0.3369868993759155,
|
160 |
+
"rewards/rejected": -0.04557197913527489,
|
161 |
+
"step": 80
|
162 |
+
},
|
163 |
+
{
|
164 |
+
"epoch": 0.37,
|
165 |
+
"grad_norm": 147.27755737304688,
|
166 |
+
"learning_rate": 4.856140350877193e-06,
|
167 |
+
"logits/chosen": -1.8193047046661377,
|
168 |
+
"logits/rejected": -1.8773219585418701,
|
169 |
+
"logps/chosen": -155.37123107910156,
|
170 |
+
"logps/rejected": -245.6790008544922,
|
171 |
+
"loss": 0.4932,
|
172 |
+
"rewards/accuracies": 0.699999988079071,
|
173 |
+
"rewards/chosen": 0.5351110696792603,
|
174 |
+
"rewards/margins": 0.5927553176879883,
|
175 |
+
"rewards/rejected": -0.05764435604214668,
|
176 |
+
"step": 90
|
177 |
+
},
|
178 |
+
{
|
179 |
+
"epoch": 0.37,
|
180 |
+
"eval_logits/chosen": -1.8004387617111206,
|
181 |
+
"eval_logits/rejected": -1.836446762084961,
|
182 |
+
"eval_logps/chosen": -192.66908264160156,
|
183 |
+
"eval_logps/rejected": -242.95899963378906,
|
184 |
+
"eval_loss": 0.5072147846221924,
|
185 |
+
"eval_rewards/accuracies": 0.8214285969734192,
|
186 |
+
"eval_rewards/chosen": 0.5621528029441833,
|
187 |
+
"eval_rewards/margins": 0.4941706657409668,
|
188 |
+
"eval_rewards/rejected": 0.06798211485147476,
|
189 |
+
"eval_runtime": 26.2422,
|
190 |
+
"eval_samples_per_second": 1.067,
|
191 |
+
"eval_steps_per_second": 1.067,
|
192 |
+
"step": 90
|
193 |
+
},
|
194 |
+
{
|
195 |
+
"epoch": 0.41,
|
196 |
+
"grad_norm": 62.75928497314453,
|
197 |
+
"learning_rate": 4.821052631578948e-06,
|
198 |
+
"logits/chosen": -1.7835988998413086,
|
199 |
+
"logits/rejected": -1.8885362148284912,
|
200 |
+
"logps/chosen": -109.94624328613281,
|
201 |
+
"logps/rejected": -232.5241241455078,
|
202 |
+
"loss": 0.4252,
|
203 |
+
"rewards/accuracies": 1.0,
|
204 |
+
"rewards/chosen": 0.5394761562347412,
|
205 |
+
"rewards/margins": 0.787738025188446,
|
206 |
+
"rewards/rejected": -0.248261958360672,
|
207 |
+
"step": 100
|
208 |
+
},
|
209 |
+
{
|
210 |
+
"epoch": 0.45,
|
211 |
+
"grad_norm": 84.42039489746094,
|
212 |
+
"learning_rate": 4.785964912280702e-06,
|
213 |
+
"logits/chosen": -1.7081111669540405,
|
214 |
+
"logits/rejected": -1.7244077920913696,
|
215 |
+
"logps/chosen": -92.51998138427734,
|
216 |
+
"logps/rejected": -135.5617218017578,
|
217 |
+
"loss": 0.5077,
|
218 |
+
"rewards/accuracies": 0.800000011920929,
|
219 |
+
"rewards/chosen": 0.6028972864151001,
|
220 |
+
"rewards/margins": 0.511993408203125,
|
221 |
+
"rewards/rejected": 0.09090389311313629,
|
222 |
+
"step": 110
|
223 |
+
},
|
224 |
+
{
|
225 |
+
"epoch": 0.49,
|
226 |
+
"grad_norm": 26.61499786376953,
|
227 |
+
"learning_rate": 4.754385964912281e-06,
|
228 |
+
"logits/chosen": -1.9250166416168213,
|
229 |
+
"logits/rejected": -1.9426968097686768,
|
230 |
+
"logps/chosen": -152.0164337158203,
|
231 |
+
"logps/rejected": -154.2359161376953,
|
232 |
+
"loss": 0.4391,
|
233 |
+
"rewards/accuracies": 0.8999999761581421,
|
234 |
+
"rewards/chosen": 0.7938548922538757,
|
235 |
+
"rewards/margins": 0.8368040919303894,
|
236 |
+
"rewards/rejected": -0.042949117720127106,
|
237 |
+
"step": 120
|
238 |
+
},
|
239 |
+
{
|
240 |
+
"epoch": 0.49,
|
241 |
+
"eval_logits/chosen": -1.8051280975341797,
|
242 |
+
"eval_logits/rejected": -1.8413046598434448,
|
243 |
+
"eval_logps/chosen": -191.2982177734375,
|
244 |
+
"eval_logps/rejected": -242.81198120117188,
|
245 |
+
"eval_loss": 0.4335751235485077,
|
246 |
+
"eval_rewards/accuracies": 0.7857142686843872,
|
247 |
+
"eval_rewards/chosen": 0.9734136462211609,
|
248 |
+
"eval_rewards/margins": 0.8613229990005493,
|
249 |
+
"eval_rewards/rejected": 0.11209066957235336,
|
250 |
+
"eval_runtime": 26.2345,
|
251 |
+
"eval_samples_per_second": 1.067,
|
252 |
+
"eval_steps_per_second": 1.067,
|
253 |
+
"step": 120
|
254 |
+
},
|
255 |
+
{
|
256 |
+
"epoch": 0.53,
|
257 |
+
"grad_norm": 90.6368637084961,
|
258 |
+
"learning_rate": 4.7192982456140355e-06,
|
259 |
+
"logits/chosen": -1.865144968032837,
|
260 |
+
"logits/rejected": -1.9209120273590088,
|
261 |
+
"logps/chosen": -149.78990173339844,
|
262 |
+
"logps/rejected": -293.8749084472656,
|
263 |
+
"loss": 0.4133,
|
264 |
+
"rewards/accuracies": 0.800000011920929,
|
265 |
+
"rewards/chosen": 0.4554510712623596,
|
266 |
+
"rewards/margins": 0.9285024404525757,
|
267 |
+
"rewards/rejected": -0.47305139899253845,
|
268 |
+
"step": 130
|
269 |
+
},
|
270 |
+
{
|
271 |
+
"epoch": 0.57,
|
272 |
+
"grad_norm": 6.8597941398620605,
|
273 |
+
"learning_rate": 4.68421052631579e-06,
|
274 |
+
"logits/chosen": -1.8089487552642822,
|
275 |
+
"logits/rejected": -1.8955650329589844,
|
276 |
+
"logps/chosen": -100.26250457763672,
|
277 |
+
"logps/rejected": -209.1724090576172,
|
278 |
+
"loss": 0.3414,
|
279 |
+
"rewards/accuracies": 1.0,
|
280 |
+
"rewards/chosen": 0.9389357566833496,
|
281 |
+
"rewards/margins": 1.521246314048767,
|
282 |
+
"rewards/rejected": -0.5823107361793518,
|
283 |
+
"step": 140
|
284 |
+
},
|
285 |
+
{
|
286 |
+
"epoch": 0.61,
|
287 |
+
"grad_norm": 79.9189453125,
|
288 |
+
"learning_rate": 4.652631578947368e-06,
|
289 |
+
"logits/chosen": -1.760371446609497,
|
290 |
+
"logits/rejected": -1.8096182346343994,
|
291 |
+
"logps/chosen": -193.62579345703125,
|
292 |
+
"logps/rejected": -166.35800170898438,
|
293 |
+
"loss": 0.3208,
|
294 |
+
"rewards/accuracies": 0.8999999761581421,
|
295 |
+
"rewards/chosen": 1.2600924968719482,
|
296 |
+
"rewards/margins": 1.5845091342926025,
|
297 |
+
"rewards/rejected": -0.32441645860671997,
|
298 |
+
"step": 150
|
299 |
+
},
|
300 |
+
{
|
301 |
+
"epoch": 0.61,
|
302 |
+
"eval_logits/chosen": -1.8129692077636719,
|
303 |
+
"eval_logits/rejected": -1.849239706993103,
|
304 |
+
"eval_logps/chosen": -189.88929748535156,
|
305 |
+
"eval_logps/rejected": -242.91104125976562,
|
306 |
+
"eval_loss": 0.39326220750808716,
|
307 |
+
"eval_rewards/accuracies": 0.7857142686843872,
|
308 |
+
"eval_rewards/chosen": 1.3960883617401123,
|
309 |
+
"eval_rewards/margins": 1.3137177228927612,
|
310 |
+
"eval_rewards/rejected": 0.08237060159444809,
|
311 |
+
"eval_runtime": 26.3372,
|
312 |
+
"eval_samples_per_second": 1.063,
|
313 |
+
"eval_steps_per_second": 1.063,
|
314 |
+
"step": 150
|
315 |
+
},
|
316 |
+
{
|
317 |
+
"epoch": 0.65,
|
318 |
+
"grad_norm": 3.151426315307617,
|
319 |
+
"learning_rate": 4.617543859649123e-06,
|
320 |
+
"logits/chosen": -1.7185112237930298,
|
321 |
+
"logits/rejected": -1.749427080154419,
|
322 |
+
"logps/chosen": -178.8656463623047,
|
323 |
+
"logps/rejected": -201.32858276367188,
|
324 |
+
"loss": 0.4418,
|
325 |
+
"rewards/accuracies": 0.699999988079071,
|
326 |
+
"rewards/chosen": 1.3884060382843018,
|
327 |
+
"rewards/margins": 1.4542597532272339,
|
328 |
+
"rewards/rejected": -0.06585375964641571,
|
329 |
+
"step": 160
|
330 |
+
},
|
331 |
+
{
|
332 |
+
"epoch": 0.69,
|
333 |
+
"grad_norm": 37.79099655151367,
|
334 |
+
"learning_rate": 4.582456140350878e-06,
|
335 |
+
"logits/chosen": -1.730787992477417,
|
336 |
+
"logits/rejected": -1.8099721670150757,
|
337 |
+
"logps/chosen": -158.50277709960938,
|
338 |
+
"logps/rejected": -185.48129272460938,
|
339 |
+
"loss": 0.3912,
|
340 |
+
"rewards/accuracies": 0.800000011920929,
|
341 |
+
"rewards/chosen": 1.4253065586090088,
|
342 |
+
"rewards/margins": 1.2418479919433594,
|
343 |
+
"rewards/rejected": 0.18345877528190613,
|
344 |
+
"step": 170
|
345 |
+
},
|
346 |
+
{
|
347 |
+
"epoch": 0.73,
|
348 |
+
"grad_norm": 18.49593734741211,
|
349 |
+
"learning_rate": 4.547368421052632e-06,
|
350 |
+
"logits/chosen": -1.842390775680542,
|
351 |
+
"logits/rejected": -1.8834030628204346,
|
352 |
+
"logps/chosen": -125.84483337402344,
|
353 |
+
"logps/rejected": -182.78854370117188,
|
354 |
+
"loss": 0.3215,
|
355 |
+
"rewards/accuracies": 0.8999999761581421,
|
356 |
+
"rewards/chosen": 1.5408955812454224,
|
357 |
+
"rewards/margins": 1.4677695035934448,
|
358 |
+
"rewards/rejected": 0.07312598079442978,
|
359 |
+
"step": 180
|
360 |
+
},
|
361 |
+
{
|
362 |
+
"epoch": 0.73,
|
363 |
+
"eval_logits/chosen": -1.8194406032562256,
|
364 |
+
"eval_logits/rejected": -1.8561830520629883,
|
365 |
+
"eval_logps/chosen": -188.3820037841797,
|
366 |
+
"eval_logps/rejected": -243.13536071777344,
|
367 |
+
"eval_loss": 0.3756321966648102,
|
368 |
+
"eval_rewards/accuracies": 0.7857142686843872,
|
369 |
+
"eval_rewards/chosen": 1.8482797145843506,
|
370 |
+
"eval_rewards/margins": 1.8331985473632812,
|
371 |
+
"eval_rewards/rejected": 0.015081183984875679,
|
372 |
+
"eval_runtime": 26.2381,
|
373 |
+
"eval_samples_per_second": 1.067,
|
374 |
+
"eval_steps_per_second": 1.067,
|
375 |
+
"step": 180
|
376 |
+
},
|
377 |
+
{
|
378 |
+
"epoch": 0.78,
|
379 |
+
"grad_norm": 106.56973266601562,
|
380 |
+
"learning_rate": 4.512280701754386e-06,
|
381 |
+
"logits/chosen": -1.8318649530410767,
|
382 |
+
"logits/rejected": -1.8742272853851318,
|
383 |
+
"logps/chosen": -197.66619873046875,
|
384 |
+
"logps/rejected": -247.83560180664062,
|
385 |
+
"loss": 0.3303,
|
386 |
+
"rewards/accuracies": 0.800000011920929,
|
387 |
+
"rewards/chosen": 2.4889655113220215,
|
388 |
+
"rewards/margins": 2.536695957183838,
|
389 |
+
"rewards/rejected": -0.047730110585689545,
|
390 |
+
"step": 190
|
391 |
+
},
|
392 |
+
{
|
393 |
+
"epoch": 0.82,
|
394 |
+
"grad_norm": 1.9826362133026123,
|
395 |
+
"learning_rate": 4.47719298245614e-06,
|
396 |
+
"logits/chosen": -1.7579128742218018,
|
397 |
+
"logits/rejected": -1.7890180349349976,
|
398 |
+
"logps/chosen": -129.37191772460938,
|
399 |
+
"logps/rejected": -166.67636108398438,
|
400 |
+
"loss": 0.3292,
|
401 |
+
"rewards/accuracies": 0.8999999761581421,
|
402 |
+
"rewards/chosen": 2.5646984577178955,
|
403 |
+
"rewards/margins": 2.6955490112304688,
|
404 |
+
"rewards/rejected": -0.13085035979747772,
|
405 |
+
"step": 200
|
406 |
+
},
|
407 |
+
{
|
408 |
+
"epoch": 0.86,
|
409 |
+
"grad_norm": 6.777231693267822,
|
410 |
+
"learning_rate": 4.442105263157896e-06,
|
411 |
+
"logits/chosen": -1.7335002422332764,
|
412 |
+
"logits/rejected": -1.8272641897201538,
|
413 |
+
"logps/chosen": -114.22731018066406,
|
414 |
+
"logps/rejected": -215.8419647216797,
|
415 |
+
"loss": 0.0817,
|
416 |
+
"rewards/accuracies": 1.0,
|
417 |
+
"rewards/chosen": 3.278646469116211,
|
418 |
+
"rewards/margins": 4.009958267211914,
|
419 |
+
"rewards/rejected": -0.7313117980957031,
|
420 |
+
"step": 210
|
421 |
+
},
|
422 |
+
{
|
423 |
+
"epoch": 0.86,
|
424 |
+
"eval_logits/chosen": -1.826608419418335,
|
425 |
+
"eval_logits/rejected": -1.8640671968460083,
|
426 |
+
"eval_logps/chosen": -186.82992553710938,
|
427 |
+
"eval_logps/rejected": -243.80209350585938,
|
428 |
+
"eval_loss": 0.3834719657897949,
|
429 |
+
"eval_rewards/accuracies": 0.7857142686843872,
|
430 |
+
"eval_rewards/chosen": 2.3139071464538574,
|
431 |
+
"eval_rewards/margins": 2.4988515377044678,
|
432 |
+
"eval_rewards/rejected": -0.1849442571401596,
|
433 |
+
"eval_runtime": 26.2309,
|
434 |
+
"eval_samples_per_second": 1.067,
|
435 |
+
"eval_steps_per_second": 1.067,
|
436 |
+
"step": 210
|
437 |
+
},
|
438 |
+
{
|
439 |
+
"epoch": 0.9,
|
440 |
+
"grad_norm": 1.0230931043624878,
|
441 |
+
"learning_rate": 4.40701754385965e-06,
|
442 |
+
"logits/chosen": -1.7561031579971313,
|
443 |
+
"logits/rejected": -1.8601219654083252,
|
444 |
+
"logps/chosen": -134.11428833007812,
|
445 |
+
"logps/rejected": -273.5272216796875,
|
446 |
+
"loss": 0.1796,
|
447 |
+
"rewards/accuracies": 0.8999999761581421,
|
448 |
+
"rewards/chosen": 1.9340410232543945,
|
449 |
+
"rewards/margins": 3.8275578022003174,
|
450 |
+
"rewards/rejected": -1.8935168981552124,
|
451 |
+
"step": 220
|
452 |
+
},
|
453 |
+
{
|
454 |
+
"epoch": 0.94,
|
455 |
+
"grad_norm": 54.65635299682617,
|
456 |
+
"learning_rate": 4.371929824561404e-06,
|
457 |
+
"logits/chosen": -1.8512550592422485,
|
458 |
+
"logits/rejected": -1.900559663772583,
|
459 |
+
"logps/chosen": -210.9802703857422,
|
460 |
+
"logps/rejected": -240.60269165039062,
|
461 |
+
"loss": 0.5004,
|
462 |
+
"rewards/accuracies": 0.800000011920929,
|
463 |
+
"rewards/chosen": 1.5826597213745117,
|
464 |
+
"rewards/margins": 2.925461530685425,
|
465 |
+
"rewards/rejected": -1.342801809310913,
|
466 |
+
"step": 230
|
467 |
+
},
|
468 |
+
{
|
469 |
+
"epoch": 0.98,
|
470 |
+
"grad_norm": 18.797134399414062,
|
471 |
+
"learning_rate": 4.336842105263158e-06,
|
472 |
+
"logits/chosen": -1.7781873941421509,
|
473 |
+
"logits/rejected": -1.8780953884124756,
|
474 |
+
"logps/chosen": -136.18722534179688,
|
475 |
+
"logps/rejected": -257.26495361328125,
|
476 |
+
"loss": 0.137,
|
477 |
+
"rewards/accuracies": 0.8999999761581421,
|
478 |
+
"rewards/chosen": 1.9192733764648438,
|
479 |
+
"rewards/margins": 3.8388113975524902,
|
480 |
+
"rewards/rejected": -1.9195384979248047,
|
481 |
+
"step": 240
|
482 |
+
},
|
483 |
+
{
|
484 |
+
"epoch": 0.98,
|
485 |
+
"eval_logits/chosen": -1.83432936668396,
|
486 |
+
"eval_logits/rejected": -1.8722076416015625,
|
487 |
+
"eval_logps/chosen": -185.8831329345703,
|
488 |
+
"eval_logps/rejected": -244.8594207763672,
|
489 |
+
"eval_loss": 0.41323208808898926,
|
490 |
+
"eval_rewards/accuracies": 0.75,
|
491 |
+
"eval_rewards/chosen": 2.5979421138763428,
|
492 |
+
"eval_rewards/margins": 3.1000888347625732,
|
493 |
+
"eval_rewards/rejected": -0.5021467208862305,
|
494 |
+
"eval_runtime": 26.2323,
|
495 |
+
"eval_samples_per_second": 1.067,
|
496 |
+
"eval_steps_per_second": 1.067,
|
497 |
+
"step": 240
|
498 |
+
},
|
499 |
+
{
|
500 |
+
"epoch": 1.02,
|
501 |
+
"grad_norm": 132.51947021484375,
|
502 |
+
"learning_rate": 4.301754385964912e-06,
|
503 |
+
"logits/chosen": -1.8787914514541626,
|
504 |
+
"logits/rejected": -1.9520515203475952,
|
505 |
+
"logps/chosen": -186.46717834472656,
|
506 |
+
"logps/rejected": -250.008056640625,
|
507 |
+
"loss": 0.3531,
|
508 |
+
"rewards/accuracies": 0.800000011920929,
|
509 |
+
"rewards/chosen": 2.192800283432007,
|
510 |
+
"rewards/margins": 3.485269069671631,
|
511 |
+
"rewards/rejected": -1.2924686670303345,
|
512 |
+
"step": 250
|
513 |
+
},
|
514 |
+
{
|
515 |
+
"epoch": 1.06,
|
516 |
+
"grad_norm": 6.529983043670654,
|
517 |
+
"learning_rate": 4.266666666666668e-06,
|
518 |
+
"logits/chosen": -1.7029941082000732,
|
519 |
+
"logits/rejected": -1.7871618270874023,
|
520 |
+
"logps/chosen": -128.5960235595703,
|
521 |
+
"logps/rejected": -209.27603149414062,
|
522 |
+
"loss": 0.0883,
|
523 |
+
"rewards/accuracies": 0.8999999761581421,
|
524 |
+
"rewards/chosen": 2.7405991554260254,
|
525 |
+
"rewards/margins": 5.148273468017578,
|
526 |
+
"rewards/rejected": -2.4076738357543945,
|
527 |
+
"step": 260
|
528 |
+
},
|
529 |
+
{
|
530 |
+
"epoch": 1.1,
|
531 |
+
"grad_norm": 0.27644646167755127,
|
532 |
+
"learning_rate": 4.2315789473684215e-06,
|
533 |
+
"logits/chosen": -1.7499496936798096,
|
534 |
+
"logits/rejected": -1.8194109201431274,
|
535 |
+
"logps/chosen": -119.34013366699219,
|
536 |
+
"logps/rejected": -230.69229125976562,
|
537 |
+
"loss": 0.0997,
|
538 |
+
"rewards/accuracies": 0.8999999761581421,
|
539 |
+
"rewards/chosen": 4.146358013153076,
|
540 |
+
"rewards/margins": 5.72900390625,
|
541 |
+
"rewards/rejected": -1.5826464891433716,
|
542 |
+
"step": 270
|
543 |
+
},
|
544 |
+
{
|
545 |
+
"epoch": 1.1,
|
546 |
+
"eval_logits/chosen": -1.842972755432129,
|
547 |
+
"eval_logits/rejected": -1.8816190958023071,
|
548 |
+
"eval_logps/chosen": -185.4148406982422,
|
549 |
+
"eval_logps/rejected": -246.53672790527344,
|
550 |
+
"eval_loss": 0.46570485830307007,
|
551 |
+
"eval_rewards/accuracies": 0.75,
|
552 |
+
"eval_rewards/chosen": 2.7384250164031982,
|
553 |
+
"eval_rewards/margins": 3.743762969970703,
|
554 |
+
"eval_rewards/rejected": -1.0053375959396362,
|
555 |
+
"eval_runtime": 26.2348,
|
556 |
+
"eval_samples_per_second": 1.067,
|
557 |
+
"eval_steps_per_second": 1.067,
|
558 |
+
"step": 270
|
559 |
+
},
|
560 |
+
{
|
561 |
+
"epoch": 1.14,
|
562 |
+
"grad_norm": 31.16436195373535,
|
563 |
+
"learning_rate": 4.196491228070176e-06,
|
564 |
+
"logits/chosen": -1.8344509601593018,
|
565 |
+
"logits/rejected": -1.919553518295288,
|
566 |
+
"logps/chosen": -131.51156616210938,
|
567 |
+
"logps/rejected": -246.3345184326172,
|
568 |
+
"loss": 0.138,
|
569 |
+
"rewards/accuracies": 1.0,
|
570 |
+
"rewards/chosen": 2.982485771179199,
|
571 |
+
"rewards/margins": 6.261662483215332,
|
572 |
+
"rewards/rejected": -3.2791759967803955,
|
573 |
+
"step": 280
|
574 |
+
},
|
575 |
+
{
|
576 |
+
"epoch": 1.18,
|
577 |
+
"grad_norm": 0.08183232694864273,
|
578 |
+
"learning_rate": 4.16140350877193e-06,
|
579 |
+
"logits/chosen": -1.847378134727478,
|
580 |
+
"logits/rejected": -1.8774774074554443,
|
581 |
+
"logps/chosen": -127.52336120605469,
|
582 |
+
"logps/rejected": -157.3633270263672,
|
583 |
+
"loss": 0.3883,
|
584 |
+
"rewards/accuracies": 0.800000011920929,
|
585 |
+
"rewards/chosen": 0.7564169764518738,
|
586 |
+
"rewards/margins": 2.541962146759033,
|
587 |
+
"rewards/rejected": -1.7855453491210938,
|
588 |
+
"step": 290
|
589 |
+
},
|
590 |
+
{
|
591 |
+
"epoch": 1.22,
|
592 |
+
"grad_norm": 10.61989974975586,
|
593 |
+
"learning_rate": 4.126315789473685e-06,
|
594 |
+
"logits/chosen": -1.7770633697509766,
|
595 |
+
"logits/rejected": -1.8337452411651611,
|
596 |
+
"logps/chosen": -112.7281265258789,
|
597 |
+
"logps/rejected": -195.87020874023438,
|
598 |
+
"loss": 0.0432,
|
599 |
+
"rewards/accuracies": 1.0,
|
600 |
+
"rewards/chosen": 3.176725387573242,
|
601 |
+
"rewards/margins": 6.761993408203125,
|
602 |
+
"rewards/rejected": -3.5852675437927246,
|
603 |
+
"step": 300
|
604 |
+
},
|
605 |
+
{
|
606 |
+
"epoch": 1.22,
|
607 |
+
"eval_logits/chosen": -1.849483609199524,
|
608 |
+
"eval_logits/rejected": -1.888432502746582,
|
609 |
+
"eval_logps/chosen": -185.52931213378906,
|
610 |
+
"eval_logps/rejected": -248.10926818847656,
|
611 |
+
"eval_loss": 0.501145601272583,
|
612 |
+
"eval_rewards/accuracies": 0.75,
|
613 |
+
"eval_rewards/chosen": 2.704090118408203,
|
614 |
+
"eval_rewards/margins": 4.181191921234131,
|
615 |
+
"eval_rewards/rejected": -1.477101445198059,
|
616 |
+
"eval_runtime": 26.2437,
|
617 |
+
"eval_samples_per_second": 1.067,
|
618 |
+
"eval_steps_per_second": 1.067,
|
619 |
+
"step": 300
|
620 |
+
},
|
621 |
+
{
|
622 |
+
"epoch": 1.27,
|
623 |
+
"grad_norm": 327.02899169921875,
|
624 |
+
"learning_rate": 4.091228070175439e-06,
|
625 |
+
"logits/chosen": -1.8802036046981812,
|
626 |
+
"logits/rejected": -1.9435373544692993,
|
627 |
+
"logps/chosen": -164.2046661376953,
|
628 |
+
"logps/rejected": -269.22235107421875,
|
629 |
+
"loss": 0.4088,
|
630 |
+
"rewards/accuracies": 0.800000011920929,
|
631 |
+
"rewards/chosen": 3.3527073860168457,
|
632 |
+
"rewards/margins": 5.9436140060424805,
|
633 |
+
"rewards/rejected": -2.5909066200256348,
|
634 |
+
"step": 310
|
635 |
+
},
|
636 |
+
{
|
637 |
+
"epoch": 1.31,
|
638 |
+
"grad_norm": 6.565308094024658,
|
639 |
+
"learning_rate": 4.056140350877193e-06,
|
640 |
+
"logits/chosen": -1.9250032901763916,
|
641 |
+
"logits/rejected": -1.9860813617706299,
|
642 |
+
"logps/chosen": -197.4870147705078,
|
643 |
+
"logps/rejected": -207.25009155273438,
|
644 |
+
"loss": 0.4969,
|
645 |
+
"rewards/accuracies": 0.699999988079071,
|
646 |
+
"rewards/chosen": 0.32615354657173157,
|
647 |
+
"rewards/margins": 2.2872400283813477,
|
648 |
+
"rewards/rejected": -1.9610865116119385,
|
649 |
+
"step": 320
|
650 |
+
},
|
651 |
+
{
|
652 |
+
"epoch": 1.35,
|
653 |
+
"grad_norm": 33.34074020385742,
|
654 |
+
"learning_rate": 4.021052631578948e-06,
|
655 |
+
"logits/chosen": -1.8470687866210938,
|
656 |
+
"logits/rejected": -1.8675216436386108,
|
657 |
+
"logps/chosen": -117.5759506225586,
|
658 |
+
"logps/rejected": -151.83091735839844,
|
659 |
+
"loss": 0.1819,
|
660 |
+
"rewards/accuracies": 0.8999999761581421,
|
661 |
+
"rewards/chosen": 2.878262758255005,
|
662 |
+
"rewards/margins": 5.000906467437744,
|
663 |
+
"rewards/rejected": -2.12264347076416,
|
664 |
+
"step": 330
|
665 |
+
},
|
666 |
+
{
|
667 |
+
"epoch": 1.35,
|
668 |
+
"eval_logits/chosen": -1.8486573696136475,
|
669 |
+
"eval_logits/rejected": -1.8877869844436646,
|
670 |
+
"eval_logps/chosen": -185.54177856445312,
|
671 |
+
"eval_logps/rejected": -249.2687530517578,
|
672 |
+
"eval_loss": 0.478495329618454,
|
673 |
+
"eval_rewards/accuracies": 0.75,
|
674 |
+
"eval_rewards/chosen": 2.7003517150878906,
|
675 |
+
"eval_rewards/margins": 4.525294780731201,
|
676 |
+
"eval_rewards/rejected": -1.8249431848526,
|
677 |
+
"eval_runtime": 26.2362,
|
678 |
+
"eval_samples_per_second": 1.067,
|
679 |
+
"eval_steps_per_second": 1.067,
|
680 |
+
"step": 330
|
681 |
+
},
|
682 |
+
{
|
683 |
+
"epoch": 1.39,
|
684 |
+
"grad_norm": 0.6777727603912354,
|
685 |
+
"learning_rate": 3.985964912280702e-06,
|
686 |
+
"logits/chosen": -1.8584426641464233,
|
687 |
+
"logits/rejected": -1.884598970413208,
|
688 |
+
"logps/chosen": -78.69309997558594,
|
689 |
+
"logps/rejected": -126.11468505859375,
|
690 |
+
"loss": 0.3154,
|
691 |
+
"rewards/accuracies": 0.8999999761581421,
|
692 |
+
"rewards/chosen": 2.5596954822540283,
|
693 |
+
"rewards/margins": 4.089261054992676,
|
694 |
+
"rewards/rejected": -1.5295660495758057,
|
695 |
+
"step": 340
|
696 |
+
},
|
697 |
+
{
|
698 |
+
"epoch": 1.43,
|
699 |
+
"grad_norm": 45.252723693847656,
|
700 |
+
"learning_rate": 3.950877192982457e-06,
|
701 |
+
"logits/chosen": -1.8422210216522217,
|
702 |
+
"logits/rejected": -1.8741470575332642,
|
703 |
+
"logps/chosen": -198.17526245117188,
|
704 |
+
"logps/rejected": -179.26759338378906,
|
705 |
+
"loss": 0.2128,
|
706 |
+
"rewards/accuracies": 0.8999999761581421,
|
707 |
+
"rewards/chosen": 2.886838912963867,
|
708 |
+
"rewards/margins": 4.228543758392334,
|
709 |
+
"rewards/rejected": -1.3417048454284668,
|
710 |
+
"step": 350
|
711 |
+
},
|
712 |
+
{
|
713 |
+
"epoch": 1.47,
|
714 |
+
"grad_norm": 0.0007244591251946986,
|
715 |
+
"learning_rate": 3.9157894736842104e-06,
|
716 |
+
"logits/chosen": -1.8951257467269897,
|
717 |
+
"logits/rejected": -1.9684120416641235,
|
718 |
+
"logps/chosen": -123.7894515991211,
|
719 |
+
"logps/rejected": -278.1185607910156,
|
720 |
+
"loss": 0.0169,
|
721 |
+
"rewards/accuracies": 1.0,
|
722 |
+
"rewards/chosen": 3.9708969593048096,
|
723 |
+
"rewards/margins": 9.946398735046387,
|
724 |
+
"rewards/rejected": -5.975502014160156,
|
725 |
+
"step": 360
|
726 |
+
},
|
727 |
+
{
|
728 |
+
"epoch": 1.47,
|
729 |
+
"eval_logits/chosen": -1.8510137796401978,
|
730 |
+
"eval_logits/rejected": -1.8906936645507812,
|
731 |
+
"eval_logps/chosen": -185.66189575195312,
|
732 |
+
"eval_logps/rejected": -250.37806701660156,
|
733 |
+
"eval_loss": 0.4872412383556366,
|
734 |
+
"eval_rewards/accuracies": 0.75,
|
735 |
+
"eval_rewards/chosen": 2.664308547973633,
|
736 |
+
"eval_rewards/margins": 4.822035789489746,
|
737 |
+
"eval_rewards/rejected": -2.157727003097534,
|
738 |
+
"eval_runtime": 26.2354,
|
739 |
+
"eval_samples_per_second": 1.067,
|
740 |
+
"eval_steps_per_second": 1.067,
|
741 |
+
"step": 360
|
742 |
+
},
|
743 |
+
{
|
744 |
+
"epoch": 1.51,
|
745 |
+
"grad_norm": 0.001088384771719575,
|
746 |
+
"learning_rate": 3.880701754385965e-06,
|
747 |
+
"logits/chosen": -1.859140157699585,
|
748 |
+
"logits/rejected": -1.903721809387207,
|
749 |
+
"logps/chosen": -171.89047241210938,
|
750 |
+
"logps/rejected": -257.6295471191406,
|
751 |
+
"loss": 0.237,
|
752 |
+
"rewards/accuracies": 0.800000011920929,
|
753 |
+
"rewards/chosen": 2.257797956466675,
|
754 |
+
"rewards/margins": 7.423349857330322,
|
755 |
+
"rewards/rejected": -5.165551662445068,
|
756 |
+
"step": 370
|
757 |
+
},
|
758 |
+
{
|
759 |
+
"epoch": 1.55,
|
760 |
+
"grad_norm": 13.049817085266113,
|
761 |
+
"learning_rate": 3.84561403508772e-06,
|
762 |
+
"logits/chosen": -1.850102186203003,
|
763 |
+
"logits/rejected": -1.922501564025879,
|
764 |
+
"logps/chosen": -145.50692749023438,
|
765 |
+
"logps/rejected": -221.7799530029297,
|
766 |
+
"loss": 0.2906,
|
767 |
+
"rewards/accuracies": 0.8999999761581421,
|
768 |
+
"rewards/chosen": 1.9433931112289429,
|
769 |
+
"rewards/margins": 6.384174346923828,
|
770 |
+
"rewards/rejected": -4.440781593322754,
|
771 |
+
"step": 380
|
772 |
+
},
|
773 |
+
{
|
774 |
+
"epoch": 1.59,
|
775 |
+
"grad_norm": 0.006105802953243256,
|
776 |
+
"learning_rate": 3.810526315789474e-06,
|
777 |
+
"logits/chosen": -1.9277305603027344,
|
778 |
+
"logits/rejected": -1.9736566543579102,
|
779 |
+
"logps/chosen": -140.36146545410156,
|
780 |
+
"logps/rejected": -188.66836547851562,
|
781 |
+
"loss": 0.235,
|
782 |
+
"rewards/accuracies": 0.8999999761581421,
|
783 |
+
"rewards/chosen": 2.6781504154205322,
|
784 |
+
"rewards/margins": 5.825950622558594,
|
785 |
+
"rewards/rejected": -3.1477997303009033,
|
786 |
+
"step": 390
|
787 |
+
},
|
788 |
+
{
|
789 |
+
"epoch": 1.59,
|
790 |
+
"eval_logits/chosen": -1.8531960248947144,
|
791 |
+
"eval_logits/rejected": -1.8930160999298096,
|
792 |
+
"eval_logps/chosen": -185.68800354003906,
|
793 |
+
"eval_logps/rejected": -251.1302032470703,
|
794 |
+
"eval_loss": 0.48860952258110046,
|
795 |
+
"eval_rewards/accuracies": 0.75,
|
796 |
+
"eval_rewards/chosen": 2.656480550765991,
|
797 |
+
"eval_rewards/margins": 5.039853572845459,
|
798 |
+
"eval_rewards/rejected": -2.3833730220794678,
|
799 |
+
"eval_runtime": 26.2308,
|
800 |
+
"eval_samples_per_second": 1.067,
|
801 |
+
"eval_steps_per_second": 1.067,
|
802 |
+
"step": 390
|
803 |
+
},
|
804 |
+
{
|
805 |
+
"epoch": 1.63,
|
806 |
+
"grad_norm": 0.00030716226319782436,
|
807 |
+
"learning_rate": 3.7754385964912284e-06,
|
808 |
+
"logits/chosen": -1.8817369937896729,
|
809 |
+
"logits/rejected": -1.9124549627304077,
|
810 |
+
"logps/chosen": -171.40383911132812,
|
811 |
+
"logps/rejected": -198.42446899414062,
|
812 |
+
"loss": 0.4322,
|
813 |
+
"rewards/accuracies": 0.800000011920929,
|
814 |
+
"rewards/chosen": 1.219795823097229,
|
815 |
+
"rewards/margins": 6.402298927307129,
|
816 |
+
"rewards/rejected": -5.182502269744873,
|
817 |
+
"step": 400
|
818 |
+
},
|
819 |
+
{
|
820 |
+
"epoch": 1.67,
|
821 |
+
"grad_norm": 0.0007249619229696691,
|
822 |
+
"learning_rate": 3.7403508771929827e-06,
|
823 |
+
"logits/chosen": -1.8018410205841064,
|
824 |
+
"logits/rejected": -1.869511365890503,
|
825 |
+
"logps/chosen": -168.9734649658203,
|
826 |
+
"logps/rejected": -223.9024658203125,
|
827 |
+
"loss": 0.9332,
|
828 |
+
"rewards/accuracies": 0.699999988079071,
|
829 |
+
"rewards/chosen": 2.475252151489258,
|
830 |
+
"rewards/margins": 5.069212913513184,
|
831 |
+
"rewards/rejected": -2.593961238861084,
|
832 |
+
"step": 410
|
833 |
+
},
|
834 |
+
{
|
835 |
+
"epoch": 1.71,
|
836 |
+
"grad_norm": 46.79059982299805,
|
837 |
+
"learning_rate": 3.7052631578947374e-06,
|
838 |
+
"logits/chosen": -1.774857759475708,
|
839 |
+
"logits/rejected": -1.8125171661376953,
|
840 |
+
"logps/chosen": -140.12551879882812,
|
841 |
+
"logps/rejected": -208.7825927734375,
|
842 |
+
"loss": 0.7551,
|
843 |
+
"rewards/accuracies": 0.800000011920929,
|
844 |
+
"rewards/chosen": 2.244128704071045,
|
845 |
+
"rewards/margins": 6.2417426109313965,
|
846 |
+
"rewards/rejected": -3.997614622116089,
|
847 |
+
"step": 420
|
848 |
+
},
|
849 |
+
{
|
850 |
+
"epoch": 1.71,
|
851 |
+
"eval_logits/chosen": -1.8527201414108276,
|
852 |
+
"eval_logits/rejected": -1.892095923423767,
|
853 |
+
"eval_logps/chosen": -185.46646118164062,
|
854 |
+
"eval_logps/rejected": -251.00816345214844,
|
855 |
+
"eval_loss": 0.438032329082489,
|
856 |
+
"eval_rewards/accuracies": 0.75,
|
857 |
+
"eval_rewards/chosen": 2.7229368686676025,
|
858 |
+
"eval_rewards/margins": 5.069703102111816,
|
859 |
+
"eval_rewards/rejected": -2.346766233444214,
|
860 |
+
"eval_runtime": 26.2319,
|
861 |
+
"eval_samples_per_second": 1.067,
|
862 |
+
"eval_steps_per_second": 1.067,
|
863 |
+
"step": 420
|
864 |
+
},
|
865 |
+
{
|
866 |
+
"epoch": 1.76,
|
867 |
+
"grad_norm": 0.1679692417383194,
|
868 |
+
"learning_rate": 3.6701754385964917e-06,
|
869 |
+
"logits/chosen": -1.9766556024551392,
|
870 |
+
"logits/rejected": -1.967247724533081,
|
871 |
+
"logps/chosen": -181.25978088378906,
|
872 |
+
"logps/rejected": -174.0313262939453,
|
873 |
+
"loss": 0.5746,
|
874 |
+
"rewards/accuracies": 0.800000011920929,
|
875 |
+
"rewards/chosen": 2.037475347518921,
|
876 |
+
"rewards/margins": 4.661454200744629,
|
877 |
+
"rewards/rejected": -2.623978853225708,
|
878 |
+
"step": 430
|
879 |
+
},
|
880 |
+
{
|
881 |
+
"epoch": 1.8,
|
882 |
+
"grad_norm": 0.028290700167417526,
|
883 |
+
"learning_rate": 3.635087719298246e-06,
|
884 |
+
"logits/chosen": -1.7324810028076172,
|
885 |
+
"logits/rejected": -1.8381952047348022,
|
886 |
+
"logps/chosen": -118.8620376586914,
|
887 |
+
"logps/rejected": -249.5945587158203,
|
888 |
+
"loss": 0.1919,
|
889 |
+
"rewards/accuracies": 0.8999999761581421,
|
890 |
+
"rewards/chosen": 2.84242582321167,
|
891 |
+
"rewards/margins": 8.061222076416016,
|
892 |
+
"rewards/rejected": -5.218796730041504,
|
893 |
+
"step": 440
|
894 |
+
},
|
895 |
+
{
|
896 |
+
"epoch": 1.84,
|
897 |
+
"grad_norm": 3.435687303543091,
|
898 |
+
"learning_rate": 3.6000000000000003e-06,
|
899 |
+
"logits/chosen": -1.788468599319458,
|
900 |
+
"logits/rejected": -1.8786054849624634,
|
901 |
+
"logps/chosen": -127.3366928100586,
|
902 |
+
"logps/rejected": -223.4063262939453,
|
903 |
+
"loss": 0.134,
|
904 |
+
"rewards/accuracies": 1.0,
|
905 |
+
"rewards/chosen": 0.9258162379264832,
|
906 |
+
"rewards/margins": 6.352460861206055,
|
907 |
+
"rewards/rejected": -5.426644802093506,
|
908 |
+
"step": 450
|
909 |
+
},
|
910 |
+
{
|
911 |
+
"epoch": 1.84,
|
912 |
+
"eval_logits/chosen": -1.8530861139297485,
|
913 |
+
"eval_logits/rejected": -1.8925215005874634,
|
914 |
+
"eval_logps/chosen": -185.6543426513672,
|
915 |
+
"eval_logps/rejected": -251.7077178955078,
|
916 |
+
"eval_loss": 0.4383006691932678,
|
917 |
+
"eval_rewards/accuracies": 0.75,
|
918 |
+
"eval_rewards/chosen": 2.6665782928466797,
|
919 |
+
"eval_rewards/margins": 5.223209857940674,
|
920 |
+
"eval_rewards/rejected": -2.556631565093994,
|
921 |
+
"eval_runtime": 26.2389,
|
922 |
+
"eval_samples_per_second": 1.067,
|
923 |
+
"eval_steps_per_second": 1.067,
|
924 |
+
"step": 450
|
925 |
+
},
|
926 |
+
{
|
927 |
+
"epoch": 1.88,
|
928 |
+
"grad_norm": 0.2778886556625366,
|
929 |
+
"learning_rate": 3.5649122807017545e-06,
|
930 |
+
"logits/chosen": -1.8211870193481445,
|
931 |
+
"logits/rejected": -1.8562345504760742,
|
932 |
+
"logps/chosen": -90.34883117675781,
|
933 |
+
"logps/rejected": -171.23648071289062,
|
934 |
+
"loss": 0.1284,
|
935 |
+
"rewards/accuracies": 1.0,
|
936 |
+
"rewards/chosen": 3.756679058074951,
|
937 |
+
"rewards/margins": 7.474773406982422,
|
938 |
+
"rewards/rejected": -3.7180938720703125,
|
939 |
+
"step": 460
|
940 |
+
},
|
941 |
+
{
|
942 |
+
"epoch": 1.92,
|
943 |
+
"grad_norm": 0.00020280799071770161,
|
944 |
+
"learning_rate": 3.5298245614035093e-06,
|
945 |
+
"logits/chosen": -1.7977367639541626,
|
946 |
+
"logits/rejected": -1.8836923837661743,
|
947 |
+
"logps/chosen": -180.67724609375,
|
948 |
+
"logps/rejected": -215.39608764648438,
|
949 |
+
"loss": 0.5069,
|
950 |
+
"rewards/accuracies": 0.800000011920929,
|
951 |
+
"rewards/chosen": 2.3001179695129395,
|
952 |
+
"rewards/margins": 7.598956108093262,
|
953 |
+
"rewards/rejected": -5.298838138580322,
|
954 |
+
"step": 470
|
955 |
+
},
|
956 |
+
{
|
957 |
+
"epoch": 1.96,
|
958 |
+
"grad_norm": 0.0038736851420253515,
|
959 |
+
"learning_rate": 3.4947368421052635e-06,
|
960 |
+
"logits/chosen": -1.8841644525527954,
|
961 |
+
"logits/rejected": -1.986575722694397,
|
962 |
+
"logps/chosen": -189.42210388183594,
|
963 |
+
"logps/rejected": -368.7510070800781,
|
964 |
+
"loss": 0.0662,
|
965 |
+
"rewards/accuracies": 1.0,
|
966 |
+
"rewards/chosen": 2.894408702850342,
|
967 |
+
"rewards/margins": 9.953689575195312,
|
968 |
+
"rewards/rejected": -7.0592803955078125,
|
969 |
+
"step": 480
|
970 |
+
},
|
971 |
+
{
|
972 |
+
"epoch": 1.96,
|
973 |
+
"eval_logits/chosen": -1.8568991422653198,
|
974 |
+
"eval_logits/rejected": -1.896437406539917,
|
975 |
+
"eval_logps/chosen": -186.01429748535156,
|
976 |
+
"eval_logps/rejected": -252.9164276123047,
|
977 |
+
"eval_loss": 0.44481489062309265,
|
978 |
+
"eval_rewards/accuracies": 0.75,
|
979 |
+
"eval_rewards/chosen": 2.55859112739563,
|
980 |
+
"eval_rewards/margins": 5.477838039398193,
|
981 |
+
"eval_rewards/rejected": -2.9192464351654053,
|
982 |
+
"eval_runtime": 26.2387,
|
983 |
+
"eval_samples_per_second": 1.067,
|
984 |
+
"eval_steps_per_second": 1.067,
|
985 |
+
"step": 480
|
986 |
+
},
|
987 |
+
{
|
988 |
+
"epoch": 2.0,
|
989 |
+
"grad_norm": 0.00020024628611281514,
|
990 |
+
"learning_rate": 3.459649122807018e-06,
|
991 |
+
"logits/chosen": -1.7868010997772217,
|
992 |
+
"logits/rejected": -1.8775966167449951,
|
993 |
+
"logps/chosen": -168.01686096191406,
|
994 |
+
"logps/rejected": -268.81329345703125,
|
995 |
+
"loss": 0.2127,
|
996 |
+
"rewards/accuracies": 0.8999999761581421,
|
997 |
+
"rewards/chosen": 4.091753959655762,
|
998 |
+
"rewards/margins": 10.252196311950684,
|
999 |
+
"rewards/rejected": -6.160443305969238,
|
1000 |
+
"step": 490
|
1001 |
+
},
|
1002 |
+
{
|
1003 |
+
"epoch": 2.04,
|
1004 |
+
"grad_norm": 2.933464288711548,
|
1005 |
+
"learning_rate": 3.424561403508772e-06,
|
1006 |
+
"logits/chosen": -1.8647997379302979,
|
1007 |
+
"logits/rejected": -1.8972787857055664,
|
1008 |
+
"logps/chosen": -206.39712524414062,
|
1009 |
+
"logps/rejected": -255.3307647705078,
|
1010 |
+
"loss": 0.1551,
|
1011 |
+
"rewards/accuracies": 0.8999999761581421,
|
1012 |
+
"rewards/chosen": 3.2640254497528076,
|
1013 |
+
"rewards/margins": 8.8242826461792,
|
1014 |
+
"rewards/rejected": -5.560257434844971,
|
1015 |
+
"step": 500
|
1016 |
+
},
|
1017 |
+
{
|
1018 |
+
"epoch": 2.08,
|
1019 |
+
"grad_norm": 0.088715098798275,
|
1020 |
+
"learning_rate": 3.3894736842105264e-06,
|
1021 |
+
"logits/chosen": -1.8153291940689087,
|
1022 |
+
"logits/rejected": -1.907467246055603,
|
1023 |
+
"logps/chosen": -114.2393569946289,
|
1024 |
+
"logps/rejected": -245.9196319580078,
|
1025 |
+
"loss": 0.1093,
|
1026 |
+
"rewards/accuracies": 1.0,
|
1027 |
+
"rewards/chosen": 1.2050167322158813,
|
1028 |
+
"rewards/margins": 7.797829627990723,
|
1029 |
+
"rewards/rejected": -6.592813014984131,
|
1030 |
+
"step": 510
|
1031 |
+
},
|
1032 |
+
{
|
1033 |
+
"epoch": 2.08,
|
1034 |
+
"eval_logits/chosen": -1.8561475276947021,
|
1035 |
+
"eval_logits/rejected": -1.8955259323120117,
|
1036 |
+
"eval_logps/chosen": -186.13938903808594,
|
1037 |
+
"eval_logps/rejected": -253.4276580810547,
|
1038 |
+
"eval_loss": 0.4262312054634094,
|
1039 |
+
"eval_rewards/accuracies": 0.75,
|
1040 |
+
"eval_rewards/chosen": 2.521066904067993,
|
1041 |
+
"eval_rewards/margins": 5.59368371963501,
|
1042 |
+
"eval_rewards/rejected": -3.0726168155670166,
|
1043 |
+
"eval_runtime": 26.2357,
|
1044 |
+
"eval_samples_per_second": 1.067,
|
1045 |
+
"eval_steps_per_second": 1.067,
|
1046 |
+
"step": 510
|
1047 |
+
},
|
1048 |
+
{
|
1049 |
+
"epoch": 2.12,
|
1050 |
+
"grad_norm": 0.0037095127627253532,
|
1051 |
+
"learning_rate": 3.354385964912281e-06,
|
1052 |
+
"logits/chosen": -1.904370665550232,
|
1053 |
+
"logits/rejected": -1.9216587543487549,
|
1054 |
+
"logps/chosen": -173.86373901367188,
|
1055 |
+
"logps/rejected": -173.6374053955078,
|
1056 |
+
"loss": 0.1557,
|
1057 |
+
"rewards/accuracies": 0.8999999761581421,
|
1058 |
+
"rewards/chosen": 2.4744720458984375,
|
1059 |
+
"rewards/margins": 5.560853004455566,
|
1060 |
+
"rewards/rejected": -3.0863804817199707,
|
1061 |
+
"step": 520
|
1062 |
+
},
|
1063 |
+
{
|
1064 |
+
"epoch": 2.16,
|
1065 |
+
"grad_norm": 1.6992926532566344e-07,
|
1066 |
+
"learning_rate": 3.3192982456140354e-06,
|
1067 |
+
"logits/chosen": -1.7665284872055054,
|
1068 |
+
"logits/rejected": -1.859400987625122,
|
1069 |
+
"logps/chosen": -85.06950378417969,
|
1070 |
+
"logps/rejected": -232.3406524658203,
|
1071 |
+
"loss": 0.0056,
|
1072 |
+
"rewards/accuracies": 1.0,
|
1073 |
+
"rewards/chosen": 3.268359661102295,
|
1074 |
+
"rewards/margins": 9.698001861572266,
|
1075 |
+
"rewards/rejected": -6.4296417236328125,
|
1076 |
+
"step": 530
|
1077 |
+
},
|
1078 |
+
{
|
1079 |
+
"epoch": 2.2,
|
1080 |
+
"grad_norm": 0.00015174441796261817,
|
1081 |
+
"learning_rate": 3.2842105263157897e-06,
|
1082 |
+
"logits/chosen": -1.7254387140274048,
|
1083 |
+
"logits/rejected": -1.8300654888153076,
|
1084 |
+
"logps/chosen": -126.21636962890625,
|
1085 |
+
"logps/rejected": -236.91250610351562,
|
1086 |
+
"loss": 0.1557,
|
1087 |
+
"rewards/accuracies": 0.8999999761581421,
|
1088 |
+
"rewards/chosen": 2.9851126670837402,
|
1089 |
+
"rewards/margins": 9.359209060668945,
|
1090 |
+
"rewards/rejected": -6.374096870422363,
|
1091 |
+
"step": 540
|
1092 |
+
},
|
1093 |
+
{
|
1094 |
+
"epoch": 2.2,
|
1095 |
+
"eval_logits/chosen": -1.85663902759552,
|
1096 |
+
"eval_logits/rejected": -1.8965102434158325,
|
1097 |
+
"eval_logps/chosen": -186.6448974609375,
|
1098 |
+
"eval_logps/rejected": -254.58482360839844,
|
1099 |
+
"eval_loss": 0.4263899326324463,
|
1100 |
+
"eval_rewards/accuracies": 0.75,
|
1101 |
+
"eval_rewards/chosen": 2.369408369064331,
|
1102 |
+
"eval_rewards/margins": 5.789170265197754,
|
1103 |
+
"eval_rewards/rejected": -3.4197616577148438,
|
1104 |
+
"eval_runtime": 26.2541,
|
1105 |
+
"eval_samples_per_second": 1.066,
|
1106 |
+
"eval_steps_per_second": 1.066,
|
1107 |
+
"step": 540
|
1108 |
+
},
|
1109 |
+
{
|
1110 |
+
"epoch": 2.24,
|
1111 |
+
"grad_norm": 7.969775676727295,
|
1112 |
+
"learning_rate": 3.249122807017544e-06,
|
1113 |
+
"logits/chosen": -1.8786284923553467,
|
1114 |
+
"logits/rejected": -1.9631255865097046,
|
1115 |
+
"logps/chosen": -186.79092407226562,
|
1116 |
+
"logps/rejected": -295.82598876953125,
|
1117 |
+
"loss": 0.0625,
|
1118 |
+
"rewards/accuracies": 1.0,
|
1119 |
+
"rewards/chosen": 0.5664806365966797,
|
1120 |
+
"rewards/margins": 6.786992073059082,
|
1121 |
+
"rewards/rejected": -6.220510959625244,
|
1122 |
+
"step": 550
|
1123 |
+
},
|
1124 |
+
{
|
1125 |
+
"epoch": 2.29,
|
1126 |
+
"grad_norm": 41.887001037597656,
|
1127 |
+
"learning_rate": 3.2140350877192982e-06,
|
1128 |
+
"logits/chosen": -1.8324702978134155,
|
1129 |
+
"logits/rejected": -1.8570420742034912,
|
1130 |
+
"logps/chosen": -198.67893981933594,
|
1131 |
+
"logps/rejected": -203.76016235351562,
|
1132 |
+
"loss": 0.389,
|
1133 |
+
"rewards/accuracies": 0.8999999761581421,
|
1134 |
+
"rewards/chosen": 2.470458984375,
|
1135 |
+
"rewards/margins": 8.410517692565918,
|
1136 |
+
"rewards/rejected": -5.940059185028076,
|
1137 |
+
"step": 560
|
1138 |
+
},
|
1139 |
+
{
|
1140 |
+
"epoch": 2.33,
|
1141 |
+
"grad_norm": 0.022604621946811676,
|
1142 |
+
"learning_rate": 3.178947368421053e-06,
|
1143 |
+
"logits/chosen": -1.7625017166137695,
|
1144 |
+
"logits/rejected": -1.823395013809204,
|
1145 |
+
"logps/chosen": -145.4344024658203,
|
1146 |
+
"logps/rejected": -202.81698608398438,
|
1147 |
+
"loss": 0.0962,
|
1148 |
+
"rewards/accuracies": 0.8999999761581421,
|
1149 |
+
"rewards/chosen": 3.2409331798553467,
|
1150 |
+
"rewards/margins": 7.634140968322754,
|
1151 |
+
"rewards/rejected": -4.39320707321167,
|
1152 |
+
"step": 570
|
1153 |
+
},
|
1154 |
+
{
|
1155 |
+
"epoch": 2.33,
|
1156 |
+
"eval_logits/chosen": -1.8582419157028198,
|
1157 |
+
"eval_logits/rejected": -1.897825837135315,
|
1158 |
+
"eval_logps/chosen": -186.9964141845703,
|
1159 |
+
"eval_logps/rejected": -255.54441833496094,
|
1160 |
+
"eval_loss": 0.41822776198387146,
|
1161 |
+
"eval_rewards/accuracies": 0.75,
|
1162 |
+
"eval_rewards/chosen": 2.2639553546905518,
|
1163 |
+
"eval_rewards/margins": 5.971593379974365,
|
1164 |
+
"eval_rewards/rejected": -3.70763897895813,
|
1165 |
+
"eval_runtime": 26.2352,
|
1166 |
+
"eval_samples_per_second": 1.067,
|
1167 |
+
"eval_steps_per_second": 1.067,
|
1168 |
+
"step": 570
|
1169 |
+
},
|
1170 |
+
{
|
1171 |
+
"epoch": 2.37,
|
1172 |
+
"grad_norm": 0.0053796968422830105,
|
1173 |
+
"learning_rate": 3.1438596491228072e-06,
|
1174 |
+
"logits/chosen": -1.8564494848251343,
|
1175 |
+
"logits/rejected": -1.9009406566619873,
|
1176 |
+
"logps/chosen": -153.97903442382812,
|
1177 |
+
"logps/rejected": -182.94284057617188,
|
1178 |
+
"loss": 0.5733,
|
1179 |
+
"rewards/accuracies": 0.800000011920929,
|
1180 |
+
"rewards/chosen": 2.0778138637542725,
|
1181 |
+
"rewards/margins": 5.466890335083008,
|
1182 |
+
"rewards/rejected": -3.389075756072998,
|
1183 |
+
"step": 580
|
1184 |
+
},
|
1185 |
+
{
|
1186 |
+
"epoch": 2.41,
|
1187 |
+
"grad_norm": 5.523962020874023,
|
1188 |
+
"learning_rate": 3.1087719298245615e-06,
|
1189 |
+
"logits/chosen": -1.8511816263198853,
|
1190 |
+
"logits/rejected": -1.853109359741211,
|
1191 |
+
"logps/chosen": -182.23892211914062,
|
1192 |
+
"logps/rejected": -120.5928955078125,
|
1193 |
+
"loss": 0.4958,
|
1194 |
+
"rewards/accuracies": 0.800000011920929,
|
1195 |
+
"rewards/chosen": 3.659731388092041,
|
1196 |
+
"rewards/margins": 4.932890892028809,
|
1197 |
+
"rewards/rejected": -1.2731596231460571,
|
1198 |
+
"step": 590
|
1199 |
+
},
|
1200 |
+
{
|
1201 |
+
"epoch": 2.45,
|
1202 |
+
"grad_norm": 0.0013734139502048492,
|
1203 |
+
"learning_rate": 3.0736842105263158e-06,
|
1204 |
+
"logits/chosen": -1.7618497610092163,
|
1205 |
+
"logits/rejected": -1.8786863088607788,
|
1206 |
+
"logps/chosen": -114.43098449707031,
|
1207 |
+
"logps/rejected": -229.5573272705078,
|
1208 |
+
"loss": 0.0437,
|
1209 |
+
"rewards/accuracies": 1.0,
|
1210 |
+
"rewards/chosen": 2.8845677375793457,
|
1211 |
+
"rewards/margins": 9.519918441772461,
|
1212 |
+
"rewards/rejected": -6.635350227355957,
|
1213 |
+
"step": 600
|
1214 |
+
},
|
1215 |
+
{
|
1216 |
+
"epoch": 2.45,
|
1217 |
+
"eval_logits/chosen": -1.8533600568771362,
|
1218 |
+
"eval_logits/rejected": -1.8932533264160156,
|
1219 |
+
"eval_logps/chosen": -187.0037384033203,
|
1220 |
+
"eval_logps/rejected": -255.77125549316406,
|
1221 |
+
"eval_loss": 0.3824370801448822,
|
1222 |
+
"eval_rewards/accuracies": 0.75,
|
1223 |
+
"eval_rewards/chosen": 2.2617576122283936,
|
1224 |
+
"eval_rewards/margins": 6.037452220916748,
|
1225 |
+
"eval_rewards/rejected": -3.7756943702697754,
|
1226 |
+
"eval_runtime": 26.2326,
|
1227 |
+
"eval_samples_per_second": 1.067,
|
1228 |
+
"eval_steps_per_second": 1.067,
|
1229 |
+
"step": 600
|
1230 |
+
},
|
1231 |
+
{
|
1232 |
+
"epoch": 2.49,
|
1233 |
+
"grad_norm": 3.847404956817627,
|
1234 |
+
"learning_rate": 3.03859649122807e-06,
|
1235 |
+
"logits/chosen": -1.9108350276947021,
|
1236 |
+
"logits/rejected": -1.9136813879013062,
|
1237 |
+
"logps/chosen": -147.76666259765625,
|
1238 |
+
"logps/rejected": -151.54275512695312,
|
1239 |
+
"loss": 0.1939,
|
1240 |
+
"rewards/accuracies": 0.8999999761581421,
|
1241 |
+
"rewards/chosen": 2.7343037128448486,
|
1242 |
+
"rewards/margins": 7.082615852355957,
|
1243 |
+
"rewards/rejected": -4.348311424255371,
|
1244 |
+
"step": 610
|
1245 |
+
},
|
1246 |
+
{
|
1247 |
+
"epoch": 2.53,
|
1248 |
+
"grad_norm": 6.025547172328061e-09,
|
1249 |
+
"learning_rate": 3.003508771929825e-06,
|
1250 |
+
"logits/chosen": -1.9030263423919678,
|
1251 |
+
"logits/rejected": -1.964647889137268,
|
1252 |
+
"logps/chosen": -157.2318115234375,
|
1253 |
+
"logps/rejected": -335.1673583984375,
|
1254 |
+
"loss": 0.1472,
|
1255 |
+
"rewards/accuracies": 0.8999999761581421,
|
1256 |
+
"rewards/chosen": 0.17058677971363068,
|
1257 |
+
"rewards/margins": 10.464014053344727,
|
1258 |
+
"rewards/rejected": -10.293428421020508,
|
1259 |
+
"step": 620
|
1260 |
+
},
|
1261 |
+
{
|
1262 |
+
"epoch": 2.57,
|
1263 |
+
"grad_norm": 0.7363545298576355,
|
1264 |
+
"learning_rate": 2.9684210526315795e-06,
|
1265 |
+
"logits/chosen": -1.8167072534561157,
|
1266 |
+
"logits/rejected": -1.9096784591674805,
|
1267 |
+
"logps/chosen": -128.2840118408203,
|
1268 |
+
"logps/rejected": -278.52862548828125,
|
1269 |
+
"loss": 0.0278,
|
1270 |
+
"rewards/accuracies": 1.0,
|
1271 |
+
"rewards/chosen": 2.6698997020721436,
|
1272 |
+
"rewards/margins": 10.468381881713867,
|
1273 |
+
"rewards/rejected": -7.798482418060303,
|
1274 |
+
"step": 630
|
1275 |
+
},
|
1276 |
+
{
|
1277 |
+
"epoch": 2.57,
|
1278 |
+
"eval_logits/chosen": -1.8536275625228882,
|
1279 |
+
"eval_logits/rejected": -1.8932437896728516,
|
1280 |
+
"eval_logps/chosen": -186.7086181640625,
|
1281 |
+
"eval_logps/rejected": -255.70458984375,
|
1282 |
+
"eval_loss": 0.3570733964443207,
|
1283 |
+
"eval_rewards/accuracies": 0.8571428656578064,
|
1284 |
+
"eval_rewards/chosen": 2.3502955436706543,
|
1285 |
+
"eval_rewards/margins": 6.105983257293701,
|
1286 |
+
"eval_rewards/rejected": -3.755687713623047,
|
1287 |
+
"eval_runtime": 26.2519,
|
1288 |
+
"eval_samples_per_second": 1.067,
|
1289 |
+
"eval_steps_per_second": 1.067,
|
1290 |
+
"step": 630
|
1291 |
+
},
|
1292 |
+
{
|
1293 |
+
"epoch": 2.61,
|
1294 |
+
"grad_norm": 19.793140411376953,
|
1295 |
+
"learning_rate": 2.9333333333333338e-06,
|
1296 |
+
"logits/chosen": -1.9250261783599854,
|
1297 |
+
"logits/rejected": -1.969919204711914,
|
1298 |
+
"logps/chosen": -169.236572265625,
|
1299 |
+
"logps/rejected": -216.4037322998047,
|
1300 |
+
"loss": 0.3144,
|
1301 |
+
"rewards/accuracies": 0.800000011920929,
|
1302 |
+
"rewards/chosen": 1.5923322439193726,
|
1303 |
+
"rewards/margins": 8.081656455993652,
|
1304 |
+
"rewards/rejected": -6.48932409286499,
|
1305 |
+
"step": 640
|
1306 |
+
},
|
1307 |
+
{
|
1308 |
+
"epoch": 2.65,
|
1309 |
+
"grad_norm": 0.0,
|
1310 |
+
"learning_rate": 2.898245614035088e-06,
|
1311 |
+
"logits/chosen": -1.840488076210022,
|
1312 |
+
"logits/rejected": -1.9099540710449219,
|
1313 |
+
"logps/chosen": -188.36514282226562,
|
1314 |
+
"logps/rejected": -266.2427978515625,
|
1315 |
+
"loss": 0.0455,
|
1316 |
+
"rewards/accuracies": 1.0,
|
1317 |
+
"rewards/chosen": 3.294522762298584,
|
1318 |
+
"rewards/margins": 9.861984252929688,
|
1319 |
+
"rewards/rejected": -6.5674614906311035,
|
1320 |
+
"step": 650
|
1321 |
+
},
|
1322 |
+
{
|
1323 |
+
"epoch": 2.69,
|
1324 |
+
"grad_norm": 350.0583801269531,
|
1325 |
+
"learning_rate": 2.8631578947368423e-06,
|
1326 |
+
"logits/chosen": -1.875605583190918,
|
1327 |
+
"logits/rejected": -1.9353086948394775,
|
1328 |
+
"logps/chosen": -128.51104736328125,
|
1329 |
+
"logps/rejected": -191.3477325439453,
|
1330 |
+
"loss": 0.2399,
|
1331 |
+
"rewards/accuracies": 0.8999999761581421,
|
1332 |
+
"rewards/chosen": 3.1921534538269043,
|
1333 |
+
"rewards/margins": 7.449667453765869,
|
1334 |
+
"rewards/rejected": -4.257513999938965,
|
1335 |
+
"step": 660
|
1336 |
+
},
|
1337 |
+
{
|
1338 |
+
"epoch": 2.69,
|
1339 |
+
"eval_logits/chosen": -1.851164698600769,
|
1340 |
+
"eval_logits/rejected": -1.8909271955490112,
|
1341 |
+
"eval_logps/chosen": -186.86781311035156,
|
1342 |
+
"eval_logps/rejected": -256.2709655761719,
|
1343 |
+
"eval_loss": 0.3312711715698242,
|
1344 |
+
"eval_rewards/accuracies": 0.8571428656578064,
|
1345 |
+
"eval_rewards/chosen": 2.3025403022766113,
|
1346 |
+
"eval_rewards/margins": 6.228146553039551,
|
1347 |
+
"eval_rewards/rejected": -3.9256062507629395,
|
1348 |
+
"eval_runtime": 26.2337,
|
1349 |
+
"eval_samples_per_second": 1.067,
|
1350 |
+
"eval_steps_per_second": 1.067,
|
1351 |
+
"step": 660
|
1352 |
+
},
|
1353 |
+
{
|
1354 |
+
"epoch": 2.73,
|
1355 |
+
"grad_norm": 0.0,
|
1356 |
+
"learning_rate": 2.828070175438597e-06,
|
1357 |
+
"logits/chosen": -1.7718509435653687,
|
1358 |
+
"logits/rejected": -1.8662782907485962,
|
1359 |
+
"logps/chosen": -100.0200424194336,
|
1360 |
+
"logps/rejected": -176.11337280273438,
|
1361 |
+
"loss": 0.0636,
|
1362 |
+
"rewards/accuracies": 1.0,
|
1363 |
+
"rewards/chosen": 2.2194771766662598,
|
1364 |
+
"rewards/margins": 7.908571720123291,
|
1365 |
+
"rewards/rejected": -5.689094543457031,
|
1366 |
+
"step": 670
|
1367 |
+
},
|
1368 |
+
{
|
1369 |
+
"epoch": 2.78,
|
1370 |
+
"grad_norm": 9.849269554251805e-05,
|
1371 |
+
"learning_rate": 2.7929824561403513e-06,
|
1372 |
+
"logits/chosen": -1.8379380702972412,
|
1373 |
+
"logits/rejected": -1.9032844305038452,
|
1374 |
+
"logps/chosen": -147.157958984375,
|
1375 |
+
"logps/rejected": -311.342529296875,
|
1376 |
+
"loss": 0.0081,
|
1377 |
+
"rewards/accuracies": 1.0,
|
1378 |
+
"rewards/chosen": 1.7090831995010376,
|
1379 |
+
"rewards/margins": 9.63591480255127,
|
1380 |
+
"rewards/rejected": -7.926831245422363,
|
1381 |
+
"step": 680
|
1382 |
+
},
|
1383 |
+
{
|
1384 |
+
"epoch": 2.82,
|
1385 |
+
"grad_norm": 5.133792877197266,
|
1386 |
+
"learning_rate": 2.7578947368421056e-06,
|
1387 |
+
"logits/chosen": -1.8350670337677002,
|
1388 |
+
"logits/rejected": -1.9039970636367798,
|
1389 |
+
"logps/chosen": -160.75314331054688,
|
1390 |
+
"logps/rejected": -247.68008422851562,
|
1391 |
+
"loss": 0.039,
|
1392 |
+
"rewards/accuracies": 1.0,
|
1393 |
+
"rewards/chosen": 4.146931171417236,
|
1394 |
+
"rewards/margins": 11.529654502868652,
|
1395 |
+
"rewards/rejected": -7.382723808288574,
|
1396 |
+
"step": 690
|
1397 |
+
},
|
1398 |
+
{
|
1399 |
+
"epoch": 2.82,
|
1400 |
+
"eval_logits/chosen": -1.8510338068008423,
|
1401 |
+
"eval_logits/rejected": -1.8906004428863525,
|
1402 |
+
"eval_logps/chosen": -187.16348266601562,
|
1403 |
+
"eval_logps/rejected": -257.069091796875,
|
1404 |
+
"eval_loss": 0.3131271302700043,
|
1405 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1406 |
+
"eval_rewards/chosen": 2.213836193084717,
|
1407 |
+
"eval_rewards/margins": 6.378880977630615,
|
1408 |
+
"eval_rewards/rejected": -4.165044784545898,
|
1409 |
+
"eval_runtime": 26.2333,
|
1410 |
+
"eval_samples_per_second": 1.067,
|
1411 |
+
"eval_steps_per_second": 1.067,
|
1412 |
+
"step": 690
|
1413 |
+
},
|
1414 |
+
{
|
1415 |
+
"epoch": 2.86,
|
1416 |
+
"grad_norm": 0.001812128466553986,
|
1417 |
+
"learning_rate": 2.72280701754386e-06,
|
1418 |
+
"logits/chosen": -1.888392448425293,
|
1419 |
+
"logits/rejected": -1.9376449584960938,
|
1420 |
+
"logps/chosen": -123.4749755859375,
|
1421 |
+
"logps/rejected": -221.9653778076172,
|
1422 |
+
"loss": 0.1581,
|
1423 |
+
"rewards/accuracies": 0.8999999761581421,
|
1424 |
+
"rewards/chosen": 3.1250205039978027,
|
1425 |
+
"rewards/margins": 9.35824966430664,
|
1426 |
+
"rewards/rejected": -6.23322868347168,
|
1427 |
+
"step": 700
|
1428 |
+
},
|
1429 |
+
{
|
1430 |
+
"epoch": 2.9,
|
1431 |
+
"grad_norm": 0.0007696045213378966,
|
1432 |
+
"learning_rate": 2.687719298245614e-06,
|
1433 |
+
"logits/chosen": -1.7869523763656616,
|
1434 |
+
"logits/rejected": -1.9070037603378296,
|
1435 |
+
"logps/chosen": -116.8504409790039,
|
1436 |
+
"logps/rejected": -300.51806640625,
|
1437 |
+
"loss": 0.1832,
|
1438 |
+
"rewards/accuracies": 0.8999999761581421,
|
1439 |
+
"rewards/chosen": 2.3490161895751953,
|
1440 |
+
"rewards/margins": 10.493500709533691,
|
1441 |
+
"rewards/rejected": -8.144485473632812,
|
1442 |
+
"step": 710
|
1443 |
+
},
|
1444 |
+
{
|
1445 |
+
"epoch": 2.94,
|
1446 |
+
"grad_norm": 83.29313659667969,
|
1447 |
+
"learning_rate": 2.652631578947369e-06,
|
1448 |
+
"logits/chosen": -1.8838169574737549,
|
1449 |
+
"logits/rejected": -1.9118610620498657,
|
1450 |
+
"logps/chosen": -104.75813293457031,
|
1451 |
+
"logps/rejected": -167.18580627441406,
|
1452 |
+
"loss": 0.3389,
|
1453 |
+
"rewards/accuracies": 0.8999999761581421,
|
1454 |
+
"rewards/chosen": 2.7628464698791504,
|
1455 |
+
"rewards/margins": 6.317173957824707,
|
1456 |
+
"rewards/rejected": -3.5543277263641357,
|
1457 |
+
"step": 720
|
1458 |
+
},
|
1459 |
+
{
|
1460 |
+
"epoch": 2.94,
|
1461 |
+
"eval_logits/chosen": -1.8480366468429565,
|
1462 |
+
"eval_logits/rejected": -1.8872734308242798,
|
1463 |
+
"eval_logps/chosen": -187.00791931152344,
|
1464 |
+
"eval_logps/rejected": -257.2390441894531,
|
1465 |
+
"eval_loss": 0.27630648016929626,
|
1466 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1467 |
+
"eval_rewards/chosen": 2.2605104446411133,
|
1468 |
+
"eval_rewards/margins": 6.476548194885254,
|
1469 |
+
"eval_rewards/rejected": -4.216037750244141,
|
1470 |
+
"eval_runtime": 26.23,
|
1471 |
+
"eval_samples_per_second": 1.067,
|
1472 |
+
"eval_steps_per_second": 1.067,
|
1473 |
+
"step": 720
|
1474 |
+
},
|
1475 |
+
{
|
1476 |
+
"epoch": 2.98,
|
1477 |
+
"grad_norm": 0.030682619661092758,
|
1478 |
+
"learning_rate": 2.617543859649123e-06,
|
1479 |
+
"logits/chosen": -1.8667634725570679,
|
1480 |
+
"logits/rejected": -1.8890842199325562,
|
1481 |
+
"logps/chosen": -197.29690551757812,
|
1482 |
+
"logps/rejected": -219.12332153320312,
|
1483 |
+
"loss": 0.617,
|
1484 |
+
"rewards/accuracies": 0.800000011920929,
|
1485 |
+
"rewards/chosen": -0.587865948677063,
|
1486 |
+
"rewards/margins": 5.496161937713623,
|
1487 |
+
"rewards/rejected": -6.0840277671813965,
|
1488 |
+
"step": 730
|
1489 |
+
},
|
1490 |
+
{
|
1491 |
+
"epoch": 3.02,
|
1492 |
+
"grad_norm": 0.005717798601835966,
|
1493 |
+
"learning_rate": 2.5824561403508774e-06,
|
1494 |
+
"logits/chosen": -1.8484325408935547,
|
1495 |
+
"logits/rejected": -1.8692123889923096,
|
1496 |
+
"logps/chosen": -102.5673599243164,
|
1497 |
+
"logps/rejected": -141.8745880126953,
|
1498 |
+
"loss": 0.044,
|
1499 |
+
"rewards/accuracies": 1.0,
|
1500 |
+
"rewards/chosen": 2.209364175796509,
|
1501 |
+
"rewards/margins": 8.334315299987793,
|
1502 |
+
"rewards/rejected": -6.124951362609863,
|
1503 |
+
"step": 740
|
1504 |
+
},
|
1505 |
+
{
|
1506 |
+
"epoch": 3.06,
|
1507 |
+
"grad_norm": 0.0006890746881254017,
|
1508 |
+
"learning_rate": 2.5473684210526317e-06,
|
1509 |
+
"logits/chosen": -1.865378737449646,
|
1510 |
+
"logits/rejected": -1.9292535781860352,
|
1511 |
+
"logps/chosen": -103.72840881347656,
|
1512 |
+
"logps/rejected": -203.67276000976562,
|
1513 |
+
"loss": 0.0154,
|
1514 |
+
"rewards/accuracies": 1.0,
|
1515 |
+
"rewards/chosen": 3.9861044883728027,
|
1516 |
+
"rewards/margins": 8.630514144897461,
|
1517 |
+
"rewards/rejected": -4.644408702850342,
|
1518 |
+
"step": 750
|
1519 |
+
},
|
1520 |
+
{
|
1521 |
+
"epoch": 3.06,
|
1522 |
+
"eval_logits/chosen": -1.846961259841919,
|
1523 |
+
"eval_logits/rejected": -1.886168122291565,
|
1524 |
+
"eval_logps/chosen": -187.0342254638672,
|
1525 |
+
"eval_logps/rejected": -257.52471923828125,
|
1526 |
+
"eval_loss": 0.2703961730003357,
|
1527 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1528 |
+
"eval_rewards/chosen": 2.252613067626953,
|
1529 |
+
"eval_rewards/margins": 6.554351806640625,
|
1530 |
+
"eval_rewards/rejected": -4.301738739013672,
|
1531 |
+
"eval_runtime": 26.236,
|
1532 |
+
"eval_samples_per_second": 1.067,
|
1533 |
+
"eval_steps_per_second": 1.067,
|
1534 |
+
"step": 750
|
1535 |
+
},
|
1536 |
+
{
|
1537 |
+
"epoch": 3.1,
|
1538 |
+
"grad_norm": 0.0,
|
1539 |
+
"learning_rate": 2.512280701754386e-06,
|
1540 |
+
"logits/chosen": -1.921618103981018,
|
1541 |
+
"logits/rejected": -1.9239141941070557,
|
1542 |
+
"logps/chosen": -248.826416015625,
|
1543 |
+
"logps/rejected": -182.67323303222656,
|
1544 |
+
"loss": 0.5385,
|
1545 |
+
"rewards/accuracies": 0.699999988079071,
|
1546 |
+
"rewards/chosen": -0.7813287377357483,
|
1547 |
+
"rewards/margins": 4.614243507385254,
|
1548 |
+
"rewards/rejected": -5.395572185516357,
|
1549 |
+
"step": 760
|
1550 |
+
},
|
1551 |
+
{
|
1552 |
+
"epoch": 3.14,
|
1553 |
+
"grad_norm": 20.82038688659668,
|
1554 |
+
"learning_rate": 2.4771929824561407e-06,
|
1555 |
+
"logits/chosen": -1.8723421096801758,
|
1556 |
+
"logits/rejected": -1.9096221923828125,
|
1557 |
+
"logps/chosen": -144.3488311767578,
|
1558 |
+
"logps/rejected": -200.52149963378906,
|
1559 |
+
"loss": 0.0473,
|
1560 |
+
"rewards/accuracies": 1.0,
|
1561 |
+
"rewards/chosen": 0.6265076994895935,
|
1562 |
+
"rewards/margins": 5.623400688171387,
|
1563 |
+
"rewards/rejected": -4.996891975402832,
|
1564 |
+
"step": 770
|
1565 |
+
},
|
1566 |
+
{
|
1567 |
+
"epoch": 3.18,
|
1568 |
+
"grad_norm": 0.0,
|
1569 |
+
"learning_rate": 2.442105263157895e-06,
|
1570 |
+
"logits/chosen": -1.967357873916626,
|
1571 |
+
"logits/rejected": -1.995941162109375,
|
1572 |
+
"logps/chosen": -169.40643310546875,
|
1573 |
+
"logps/rejected": -225.83267211914062,
|
1574 |
+
"loss": 0.021,
|
1575 |
+
"rewards/accuracies": 1.0,
|
1576 |
+
"rewards/chosen": 3.36498761177063,
|
1577 |
+
"rewards/margins": 8.846434593200684,
|
1578 |
+
"rewards/rejected": -5.481447219848633,
|
1579 |
+
"step": 780
|
1580 |
+
},
|
1581 |
+
{
|
1582 |
+
"epoch": 3.18,
|
1583 |
+
"eval_logits/chosen": -1.84483802318573,
|
1584 |
+
"eval_logits/rejected": -1.8837740421295166,
|
1585 |
+
"eval_logps/chosen": -187.02699279785156,
|
1586 |
+
"eval_logps/rejected": -257.6650085449219,
|
1587 |
+
"eval_loss": 0.24223719537258148,
|
1588 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1589 |
+
"eval_rewards/chosen": 2.254775285720825,
|
1590 |
+
"eval_rewards/margins": 6.598587989807129,
|
1591 |
+
"eval_rewards/rejected": -4.343813419342041,
|
1592 |
+
"eval_runtime": 26.2524,
|
1593 |
+
"eval_samples_per_second": 1.067,
|
1594 |
+
"eval_steps_per_second": 1.067,
|
1595 |
+
"step": 780
|
1596 |
+
},
|
1597 |
+
{
|
1598 |
+
"epoch": 3.22,
|
1599 |
+
"grad_norm": 0.3396748900413513,
|
1600 |
+
"learning_rate": 2.4070175438596493e-06,
|
1601 |
+
"logits/chosen": -1.795336127281189,
|
1602 |
+
"logits/rejected": -1.9104626178741455,
|
1603 |
+
"logps/chosen": -127.96372985839844,
|
1604 |
+
"logps/rejected": -321.76751708984375,
|
1605 |
+
"loss": 0.0057,
|
1606 |
+
"rewards/accuracies": 1.0,
|
1607 |
+
"rewards/chosen": 1.3614065647125244,
|
1608 |
+
"rewards/margins": 11.285585403442383,
|
1609 |
+
"rewards/rejected": -9.924179077148438,
|
1610 |
+
"step": 790
|
1611 |
+
},
|
1612 |
+
{
|
1613 |
+
"epoch": 3.27,
|
1614 |
+
"grad_norm": 89.25263214111328,
|
1615 |
+
"learning_rate": 2.3719298245614036e-06,
|
1616 |
+
"logits/chosen": -1.7844104766845703,
|
1617 |
+
"logits/rejected": -1.835338830947876,
|
1618 |
+
"logps/chosen": -138.7672119140625,
|
1619 |
+
"logps/rejected": -212.5260772705078,
|
1620 |
+
"loss": 0.099,
|
1621 |
+
"rewards/accuracies": 0.8999999761581421,
|
1622 |
+
"rewards/chosen": 3.200477123260498,
|
1623 |
+
"rewards/margins": 9.856096267700195,
|
1624 |
+
"rewards/rejected": -6.655619144439697,
|
1625 |
+
"step": 800
|
1626 |
+
},
|
1627 |
+
{
|
1628 |
+
"epoch": 3.31,
|
1629 |
+
"grad_norm": 0.0014283270575106144,
|
1630 |
+
"learning_rate": 2.3368421052631583e-06,
|
1631 |
+
"logits/chosen": -1.8158670663833618,
|
1632 |
+
"logits/rejected": -1.896251916885376,
|
1633 |
+
"logps/chosen": -144.15419006347656,
|
1634 |
+
"logps/rejected": -283.3548583984375,
|
1635 |
+
"loss": 0.0614,
|
1636 |
+
"rewards/accuracies": 1.0,
|
1637 |
+
"rewards/chosen": 3.173565149307251,
|
1638 |
+
"rewards/margins": 12.303560256958008,
|
1639 |
+
"rewards/rejected": -9.129995346069336,
|
1640 |
+
"step": 810
|
1641 |
+
},
|
1642 |
+
{
|
1643 |
+
"epoch": 3.31,
|
1644 |
+
"eval_logits/chosen": -1.8417214155197144,
|
1645 |
+
"eval_logits/rejected": -1.8804857730865479,
|
1646 |
+
"eval_logps/chosen": -187.0991668701172,
|
1647 |
+
"eval_logps/rejected": -258.0171813964844,
|
1648 |
+
"eval_loss": 0.2144124060869217,
|
1649 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1650 |
+
"eval_rewards/chosen": 2.2331290245056152,
|
1651 |
+
"eval_rewards/margins": 6.682602882385254,
|
1652 |
+
"eval_rewards/rejected": -4.449473857879639,
|
1653 |
+
"eval_runtime": 26.2397,
|
1654 |
+
"eval_samples_per_second": 1.067,
|
1655 |
+
"eval_steps_per_second": 1.067,
|
1656 |
+
"step": 810
|
1657 |
+
},
|
1658 |
+
{
|
1659 |
+
"epoch": 3.35,
|
1660 |
+
"grad_norm": 8.28421070764307e-06,
|
1661 |
+
"learning_rate": 2.3017543859649126e-06,
|
1662 |
+
"logits/chosen": -1.7904294729232788,
|
1663 |
+
"logits/rejected": -1.7954041957855225,
|
1664 |
+
"logps/chosen": -136.8679962158203,
|
1665 |
+
"logps/rejected": -173.52316284179688,
|
1666 |
+
"loss": 0.0305,
|
1667 |
+
"rewards/accuracies": 1.0,
|
1668 |
+
"rewards/chosen": 3.9618313312530518,
|
1669 |
+
"rewards/margins": 9.469245910644531,
|
1670 |
+
"rewards/rejected": -5.507414817810059,
|
1671 |
+
"step": 820
|
1672 |
+
},
|
1673 |
+
{
|
1674 |
+
"epoch": 3.39,
|
1675 |
+
"grad_norm": 0.00032607687171548605,
|
1676 |
+
"learning_rate": 2.266666666666667e-06,
|
1677 |
+
"logits/chosen": -1.7678247690200806,
|
1678 |
+
"logits/rejected": -1.8045036792755127,
|
1679 |
+
"logps/chosen": -97.9904556274414,
|
1680 |
+
"logps/rejected": -158.86538696289062,
|
1681 |
+
"loss": 0.0114,
|
1682 |
+
"rewards/accuracies": 1.0,
|
1683 |
+
"rewards/chosen": 3.4605960845947266,
|
1684 |
+
"rewards/margins": 9.087323188781738,
|
1685 |
+
"rewards/rejected": -5.626727104187012,
|
1686 |
+
"step": 830
|
1687 |
+
},
|
1688 |
+
{
|
1689 |
+
"epoch": 3.43,
|
1690 |
+
"grad_norm": 9.237614631652832,
|
1691 |
+
"learning_rate": 2.231578947368421e-06,
|
1692 |
+
"logits/chosen": -1.8179492950439453,
|
1693 |
+
"logits/rejected": -1.8643207550048828,
|
1694 |
+
"logps/chosen": -158.73011779785156,
|
1695 |
+
"logps/rejected": -175.10726928710938,
|
1696 |
+
"loss": 0.0529,
|
1697 |
+
"rewards/accuracies": 1.0,
|
1698 |
+
"rewards/chosen": 1.9895429611206055,
|
1699 |
+
"rewards/margins": 7.000767707824707,
|
1700 |
+
"rewards/rejected": -5.011223793029785,
|
1701 |
+
"step": 840
|
1702 |
+
},
|
1703 |
+
{
|
1704 |
+
"epoch": 3.43,
|
1705 |
+
"eval_logits/chosen": -1.842317819595337,
|
1706 |
+
"eval_logits/rejected": -1.8808995485305786,
|
1707 |
+
"eval_logps/chosen": -187.3555145263672,
|
1708 |
+
"eval_logps/rejected": -258.76568603515625,
|
1709 |
+
"eval_loss": 0.2120712250471115,
|
1710 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1711 |
+
"eval_rewards/chosen": 2.1562187671661377,
|
1712 |
+
"eval_rewards/margins": 6.830246925354004,
|
1713 |
+
"eval_rewards/rejected": -4.674027919769287,
|
1714 |
+
"eval_runtime": 26.238,
|
1715 |
+
"eval_samples_per_second": 1.067,
|
1716 |
+
"eval_steps_per_second": 1.067,
|
1717 |
+
"step": 840
|
1718 |
+
},
|
1719 |
+
{
|
1720 |
+
"epoch": 3.47,
|
1721 |
+
"grad_norm": 5.9564328694250435e-05,
|
1722 |
+
"learning_rate": 2.1964912280701754e-06,
|
1723 |
+
"logits/chosen": -1.782867431640625,
|
1724 |
+
"logits/rejected": -1.8579784631729126,
|
1725 |
+
"logps/chosen": -105.4168472290039,
|
1726 |
+
"logps/rejected": -259.2267761230469,
|
1727 |
+
"loss": 0.258,
|
1728 |
+
"rewards/accuracies": 0.8999999761581421,
|
1729 |
+
"rewards/chosen": 4.022895336151123,
|
1730 |
+
"rewards/margins": 12.175846099853516,
|
1731 |
+
"rewards/rejected": -8.152950286865234,
|
1732 |
+
"step": 850
|
1733 |
+
},
|
1734 |
+
{
|
1735 |
+
"epoch": 3.51,
|
1736 |
+
"grad_norm": 1.3973094610264525e-05,
|
1737 |
+
"learning_rate": 2.16140350877193e-06,
|
1738 |
+
"logits/chosen": -1.7417545318603516,
|
1739 |
+
"logits/rejected": -1.8234918117523193,
|
1740 |
+
"logps/chosen": -115.1645736694336,
|
1741 |
+
"logps/rejected": -248.0410614013672,
|
1742 |
+
"loss": 0.039,
|
1743 |
+
"rewards/accuracies": 1.0,
|
1744 |
+
"rewards/chosen": 3.356029987335205,
|
1745 |
+
"rewards/margins": 12.46558952331543,
|
1746 |
+
"rewards/rejected": -9.109560012817383,
|
1747 |
+
"step": 860
|
1748 |
+
},
|
1749 |
+
{
|
1750 |
+
"epoch": 3.55,
|
1751 |
+
"grad_norm": 0.005180311389267445,
|
1752 |
+
"learning_rate": 2.1263157894736844e-06,
|
1753 |
+
"logits/chosen": -1.8123457431793213,
|
1754 |
+
"logits/rejected": -1.9022102355957031,
|
1755 |
+
"logps/chosen": -128.5855255126953,
|
1756 |
+
"logps/rejected": -246.685791015625,
|
1757 |
+
"loss": 0.001,
|
1758 |
+
"rewards/accuracies": 1.0,
|
1759 |
+
"rewards/chosen": 2.4856154918670654,
|
1760 |
+
"rewards/margins": 12.966238021850586,
|
1761 |
+
"rewards/rejected": -10.480623245239258,
|
1762 |
+
"step": 870
|
1763 |
+
},
|
1764 |
+
{
|
1765 |
+
"epoch": 3.55,
|
1766 |
+
"eval_logits/chosen": -1.8409830331802368,
|
1767 |
+
"eval_logits/rejected": -1.87991464138031,
|
1768 |
+
"eval_logps/chosen": -187.5316925048828,
|
1769 |
+
"eval_logps/rejected": -259.3367614746094,
|
1770 |
+
"eval_loss": 0.2091565877199173,
|
1771 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1772 |
+
"eval_rewards/chosen": 2.1033718585968018,
|
1773 |
+
"eval_rewards/margins": 6.94872522354126,
|
1774 |
+
"eval_rewards/rejected": -4.845353126525879,
|
1775 |
+
"eval_runtime": 26.2322,
|
1776 |
+
"eval_samples_per_second": 1.067,
|
1777 |
+
"eval_steps_per_second": 1.067,
|
1778 |
+
"step": 870
|
1779 |
+
},
|
1780 |
+
{
|
1781 |
+
"epoch": 3.59,
|
1782 |
+
"grad_norm": 0.3039734661579132,
|
1783 |
+
"learning_rate": 2.0912280701754387e-06,
|
1784 |
+
"logits/chosen": -1.7311649322509766,
|
1785 |
+
"logits/rejected": -1.8156330585479736,
|
1786 |
+
"logps/chosen": -146.54473876953125,
|
1787 |
+
"logps/rejected": -300.264892578125,
|
1788 |
+
"loss": 0.0179,
|
1789 |
+
"rewards/accuracies": 1.0,
|
1790 |
+
"rewards/chosen": 3.358633041381836,
|
1791 |
+
"rewards/margins": 11.333723068237305,
|
1792 |
+
"rewards/rejected": -7.975089073181152,
|
1793 |
+
"step": 880
|
1794 |
+
},
|
1795 |
+
{
|
1796 |
+
"epoch": 3.63,
|
1797 |
+
"grad_norm": 0.8097418546676636,
|
1798 |
+
"learning_rate": 2.0561403508771934e-06,
|
1799 |
+
"logits/chosen": -1.8806817531585693,
|
1800 |
+
"logits/rejected": -1.922952651977539,
|
1801 |
+
"logps/chosen": -126.9049301147461,
|
1802 |
+
"logps/rejected": -184.9642333984375,
|
1803 |
+
"loss": 0.0541,
|
1804 |
+
"rewards/accuracies": 1.0,
|
1805 |
+
"rewards/chosen": 1.0604941844940186,
|
1806 |
+
"rewards/margins": 6.54372501373291,
|
1807 |
+
"rewards/rejected": -5.4832305908203125,
|
1808 |
+
"step": 890
|
1809 |
+
},
|
1810 |
+
{
|
1811 |
+
"epoch": 3.67,
|
1812 |
+
"grad_norm": 0.025914475321769714,
|
1813 |
+
"learning_rate": 2.0210526315789477e-06,
|
1814 |
+
"logits/chosen": -1.8227866888046265,
|
1815 |
+
"logits/rejected": -1.9303566217422485,
|
1816 |
+
"logps/chosen": -167.85279846191406,
|
1817 |
+
"logps/rejected": -311.27655029296875,
|
1818 |
+
"loss": 0.0284,
|
1819 |
+
"rewards/accuracies": 1.0,
|
1820 |
+
"rewards/chosen": 2.1594152450561523,
|
1821 |
+
"rewards/margins": 9.647980690002441,
|
1822 |
+
"rewards/rejected": -7.488564968109131,
|
1823 |
+
"step": 900
|
1824 |
+
},
|
1825 |
+
{
|
1826 |
+
"epoch": 3.67,
|
1827 |
+
"eval_logits/chosen": -1.8366106748580933,
|
1828 |
+
"eval_logits/rejected": -1.8760128021240234,
|
1829 |
+
"eval_logps/chosen": -187.9383544921875,
|
1830 |
+
"eval_logps/rejected": -260.31500244140625,
|
1831 |
+
"eval_loss": 0.20062340795993805,
|
1832 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1833 |
+
"eval_rewards/chosen": 1.9813728332519531,
|
1834 |
+
"eval_rewards/margins": 7.1201934814453125,
|
1835 |
+
"eval_rewards/rejected": -5.138820171356201,
|
1836 |
+
"eval_runtime": 26.2437,
|
1837 |
+
"eval_samples_per_second": 1.067,
|
1838 |
+
"eval_steps_per_second": 1.067,
|
1839 |
+
"step": 900
|
1840 |
+
},
|
1841 |
+
{
|
1842 |
+
"epoch": 3.71,
|
1843 |
+
"grad_norm": 0.02055520936846733,
|
1844 |
+
"learning_rate": 1.985964912280702e-06,
|
1845 |
+
"logits/chosen": -1.8666080236434937,
|
1846 |
+
"logits/rejected": -1.9068281650543213,
|
1847 |
+
"logps/chosen": -171.9093475341797,
|
1848 |
+
"logps/rejected": -251.93643188476562,
|
1849 |
+
"loss": 0.2622,
|
1850 |
+
"rewards/accuracies": 0.800000011920929,
|
1851 |
+
"rewards/chosen": 0.18651972711086273,
|
1852 |
+
"rewards/margins": 8.190787315368652,
|
1853 |
+
"rewards/rejected": -8.004267692565918,
|
1854 |
+
"step": 910
|
1855 |
+
},
|
1856 |
+
{
|
1857 |
+
"epoch": 3.76,
|
1858 |
+
"grad_norm": 197.38717651367188,
|
1859 |
+
"learning_rate": 1.9508771929824562e-06,
|
1860 |
+
"logits/chosen": -1.8426361083984375,
|
1861 |
+
"logits/rejected": -1.9008222818374634,
|
1862 |
+
"logps/chosen": -150.75900268554688,
|
1863 |
+
"logps/rejected": -251.9907989501953,
|
1864 |
+
"loss": 0.0599,
|
1865 |
+
"rewards/accuracies": 1.0,
|
1866 |
+
"rewards/chosen": 0.7380223870277405,
|
1867 |
+
"rewards/margins": 9.459989547729492,
|
1868 |
+
"rewards/rejected": -8.721967697143555,
|
1869 |
+
"step": 920
|
1870 |
+
},
|
1871 |
+
{
|
1872 |
+
"epoch": 3.8,
|
1873 |
+
"grad_norm": 0.0,
|
1874 |
+
"learning_rate": 1.9157894736842105e-06,
|
1875 |
+
"logits/chosen": -1.8063271045684814,
|
1876 |
+
"logits/rejected": -1.8660167455673218,
|
1877 |
+
"logps/chosen": -198.50723266601562,
|
1878 |
+
"logps/rejected": -222.7967529296875,
|
1879 |
+
"loss": 0.0744,
|
1880 |
+
"rewards/accuracies": 0.8999999761581421,
|
1881 |
+
"rewards/chosen": 1.713536262512207,
|
1882 |
+
"rewards/margins": 9.678624153137207,
|
1883 |
+
"rewards/rejected": -7.965087890625,
|
1884 |
+
"step": 930
|
1885 |
+
},
|
1886 |
+
{
|
1887 |
+
"epoch": 3.8,
|
1888 |
+
"eval_logits/chosen": -1.833928108215332,
|
1889 |
+
"eval_logits/rejected": -1.873318076133728,
|
1890 |
+
"eval_logps/chosen": -188.06385803222656,
|
1891 |
+
"eval_logps/rejected": -260.6358337402344,
|
1892 |
+
"eval_loss": 0.1813022941350937,
|
1893 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1894 |
+
"eval_rewards/chosen": 1.943717122077942,
|
1895 |
+
"eval_rewards/margins": 7.178780555725098,
|
1896 |
+
"eval_rewards/rejected": -5.2350640296936035,
|
1897 |
+
"eval_runtime": 26.2336,
|
1898 |
+
"eval_samples_per_second": 1.067,
|
1899 |
+
"eval_steps_per_second": 1.067,
|
1900 |
+
"step": 930
|
1901 |
+
},
|
1902 |
+
{
|
1903 |
+
"epoch": 3.84,
|
1904 |
+
"grad_norm": 4.8926853679631677e-08,
|
1905 |
+
"learning_rate": 1.8807017543859652e-06,
|
1906 |
+
"logits/chosen": -1.8202307224273682,
|
1907 |
+
"logits/rejected": -1.9166450500488281,
|
1908 |
+
"logps/chosen": -112.94303131103516,
|
1909 |
+
"logps/rejected": -244.4494171142578,
|
1910 |
+
"loss": 0.3721,
|
1911 |
+
"rewards/accuracies": 0.8999999761581421,
|
1912 |
+
"rewards/chosen": 2.255749225616455,
|
1913 |
+
"rewards/margins": 10.898755073547363,
|
1914 |
+
"rewards/rejected": -8.643006324768066,
|
1915 |
+
"step": 940
|
1916 |
+
},
|
1917 |
+
{
|
1918 |
+
"epoch": 3.88,
|
1919 |
+
"grad_norm": 5.456500053405762,
|
1920 |
+
"learning_rate": 1.8456140350877193e-06,
|
1921 |
+
"logits/chosen": -1.9108766317367554,
|
1922 |
+
"logits/rejected": -1.9368479251861572,
|
1923 |
+
"logps/chosen": -132.4052734375,
|
1924 |
+
"logps/rejected": -179.675537109375,
|
1925 |
+
"loss": 0.0257,
|
1926 |
+
"rewards/accuracies": 1.0,
|
1927 |
+
"rewards/chosen": 1.554168701171875,
|
1928 |
+
"rewards/margins": 6.65416955947876,
|
1929 |
+
"rewards/rejected": -5.100000858306885,
|
1930 |
+
"step": 950
|
1931 |
+
},
|
1932 |
+
{
|
1933 |
+
"epoch": 3.92,
|
1934 |
+
"grad_norm": 9.124670028686523,
|
1935 |
+
"learning_rate": 1.810526315789474e-06,
|
1936 |
+
"logits/chosen": -1.7729440927505493,
|
1937 |
+
"logits/rejected": -1.8252880573272705,
|
1938 |
+
"logps/chosen": -139.13653564453125,
|
1939 |
+
"logps/rejected": -193.011962890625,
|
1940 |
+
"loss": 0.091,
|
1941 |
+
"rewards/accuracies": 1.0,
|
1942 |
+
"rewards/chosen": 3.6216557025909424,
|
1943 |
+
"rewards/margins": 8.979948043823242,
|
1944 |
+
"rewards/rejected": -5.3582916259765625,
|
1945 |
+
"step": 960
|
1946 |
+
},
|
1947 |
+
{
|
1948 |
+
"epoch": 3.92,
|
1949 |
+
"eval_logits/chosen": -1.8312675952911377,
|
1950 |
+
"eval_logits/rejected": -1.8707492351531982,
|
1951 |
+
"eval_logps/chosen": -188.43190002441406,
|
1952 |
+
"eval_logps/rejected": -261.29730224609375,
|
1953 |
+
"eval_loss": 0.17215047776699066,
|
1954 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1955 |
+
"eval_rewards/chosen": 1.8333097696304321,
|
1956 |
+
"eval_rewards/margins": 7.266810894012451,
|
1957 |
+
"eval_rewards/rejected": -5.433501243591309,
|
1958 |
+
"eval_runtime": 26.2333,
|
1959 |
+
"eval_samples_per_second": 1.067,
|
1960 |
+
"eval_steps_per_second": 1.067,
|
1961 |
+
"step": 960
|
1962 |
+
},
|
1963 |
+
{
|
1964 |
+
"epoch": 3.96,
|
1965 |
+
"grad_norm": 1.8921099353974569e-06,
|
1966 |
+
"learning_rate": 1.7754385964912283e-06,
|
1967 |
+
"logits/chosen": -1.8464349508285522,
|
1968 |
+
"logits/rejected": -1.8927011489868164,
|
1969 |
+
"logps/chosen": -172.9676971435547,
|
1970 |
+
"logps/rejected": -224.15780639648438,
|
1971 |
+
"loss": 0.0795,
|
1972 |
+
"rewards/accuracies": 1.0,
|
1973 |
+
"rewards/chosen": 2.649458169937134,
|
1974 |
+
"rewards/margins": 9.076958656311035,
|
1975 |
+
"rewards/rejected": -6.4275007247924805,
|
1976 |
+
"step": 970
|
1977 |
+
},
|
1978 |
+
{
|
1979 |
+
"epoch": 4.0,
|
1980 |
+
"grad_norm": 318.0891418457031,
|
1981 |
+
"learning_rate": 1.7403508771929826e-06,
|
1982 |
+
"logits/chosen": -1.8083856105804443,
|
1983 |
+
"logits/rejected": -1.8901259899139404,
|
1984 |
+
"logps/chosen": -209.17422485351562,
|
1985 |
+
"logps/rejected": -284.4140625,
|
1986 |
+
"loss": 0.0583,
|
1987 |
+
"rewards/accuracies": 1.0,
|
1988 |
+
"rewards/chosen": 2.1195216178894043,
|
1989 |
+
"rewards/margins": 12.128747940063477,
|
1990 |
+
"rewards/rejected": -10.00922679901123,
|
1991 |
+
"step": 980
|
1992 |
+
},
|
1993 |
+
{
|
1994 |
+
"epoch": 4.04,
|
1995 |
+
"grad_norm": 13.0223388671875,
|
1996 |
+
"learning_rate": 1.705263157894737e-06,
|
1997 |
+
"logits/chosen": -1.847835898399353,
|
1998 |
+
"logits/rejected": -1.868665337562561,
|
1999 |
+
"logps/chosen": -149.8837127685547,
|
2000 |
+
"logps/rejected": -161.1732940673828,
|
2001 |
+
"loss": 0.3504,
|
2002 |
+
"rewards/accuracies": 0.8999999761581421,
|
2003 |
+
"rewards/chosen": 1.0685189962387085,
|
2004 |
+
"rewards/margins": 6.018086910247803,
|
2005 |
+
"rewards/rejected": -4.9495673179626465,
|
2006 |
+
"step": 990
|
2007 |
+
},
|
2008 |
+
{
|
2009 |
+
"epoch": 4.04,
|
2010 |
+
"eval_logits/chosen": -1.8279262781143188,
|
2011 |
+
"eval_logits/rejected": -1.8672196865081787,
|
2012 |
+
"eval_logps/chosen": -188.3168487548828,
|
2013 |
+
"eval_logps/rejected": -261.04876708984375,
|
2014 |
+
"eval_loss": 0.14871637523174286,
|
2015 |
+
"eval_rewards/accuracies": 0.9285714030265808,
|
2016 |
+
"eval_rewards/chosen": 1.8678233623504639,
|
2017 |
+
"eval_rewards/margins": 7.22676944732666,
|
2018 |
+
"eval_rewards/rejected": -5.358946800231934,
|
2019 |
+
"eval_runtime": 26.2331,
|
2020 |
+
"eval_samples_per_second": 1.067,
|
2021 |
+
"eval_steps_per_second": 1.067,
|
2022 |
+
"step": 990
|
2023 |
+
},
|
2024 |
+
{
|
2025 |
+
"epoch": 4.08,
|
2026 |
+
"grad_norm": 0.6881242394447327,
|
2027 |
+
"learning_rate": 1.6701754385964914e-06,
|
2028 |
+
"logits/chosen": -1.9393212795257568,
|
2029 |
+
"logits/rejected": -1.9376901388168335,
|
2030 |
+
"logps/chosen": -155.8722381591797,
|
2031 |
+
"logps/rejected": -191.2816925048828,
|
2032 |
+
"loss": 0.0102,
|
2033 |
+
"rewards/accuracies": 1.0,
|
2034 |
+
"rewards/chosen": 2.3404548168182373,
|
2035 |
+
"rewards/margins": 7.390922546386719,
|
2036 |
+
"rewards/rejected": -5.0504679679870605,
|
2037 |
+
"step": 1000
|
2038 |
+
},
|
2039 |
+
{
|
2040 |
+
"epoch": 4.12,
|
2041 |
+
"grad_norm": 0.0015760493697598577,
|
2042 |
+
"learning_rate": 1.6350877192982456e-06,
|
2043 |
+
"logits/chosen": -1.7732092142105103,
|
2044 |
+
"logits/rejected": -1.8348764181137085,
|
2045 |
+
"logps/chosen": -204.78118896484375,
|
2046 |
+
"logps/rejected": -285.19171142578125,
|
2047 |
+
"loss": 0.0728,
|
2048 |
+
"rewards/accuracies": 0.8999999761581421,
|
2049 |
+
"rewards/chosen": 1.1964905261993408,
|
2050 |
+
"rewards/margins": 11.535087585449219,
|
2051 |
+
"rewards/rejected": -10.338597297668457,
|
2052 |
+
"step": 1010
|
2053 |
+
},
|
2054 |
+
{
|
2055 |
+
"epoch": 4.16,
|
2056 |
+
"grad_norm": 5.320696345734177e-07,
|
2057 |
+
"learning_rate": 1.6000000000000001e-06,
|
2058 |
+
"logits/chosen": -1.7163060903549194,
|
2059 |
+
"logits/rejected": -1.8586156368255615,
|
2060 |
+
"logps/chosen": -123.50263977050781,
|
2061 |
+
"logps/rejected": -320.79693603515625,
|
2062 |
+
"loss": 0.0071,
|
2063 |
+
"rewards/accuracies": 1.0,
|
2064 |
+
"rewards/chosen": 0.7505568265914917,
|
2065 |
+
"rewards/margins": 10.981633186340332,
|
2066 |
+
"rewards/rejected": -10.231074333190918,
|
2067 |
+
"step": 1020
|
2068 |
+
},
|
2069 |
+
{
|
2070 |
+
"epoch": 4.16,
|
2071 |
+
"eval_logits/chosen": -1.824275016784668,
|
2072 |
+
"eval_logits/rejected": -1.8636810779571533,
|
2073 |
+
"eval_logps/chosen": -188.54676818847656,
|
2074 |
+
"eval_logps/rejected": -261.5804748535156,
|
2075 |
+
"eval_loss": 0.14029009640216827,
|
2076 |
+
"eval_rewards/accuracies": 0.9285714030265808,
|
2077 |
+
"eval_rewards/chosen": 1.7988510131835938,
|
2078 |
+
"eval_rewards/margins": 7.317309856414795,
|
2079 |
+
"eval_rewards/rejected": -5.518458843231201,
|
2080 |
+
"eval_runtime": 26.2364,
|
2081 |
+
"eval_samples_per_second": 1.067,
|
2082 |
+
"eval_steps_per_second": 1.067,
|
2083 |
+
"step": 1020
|
2084 |
+
},
|
2085 |
+
{
|
2086 |
+
"epoch": 4.2,
|
2087 |
+
"grad_norm": 0.002488874364644289,
|
2088 |
+
"learning_rate": 1.5649122807017544e-06,
|
2089 |
+
"logits/chosen": -1.7597904205322266,
|
2090 |
+
"logits/rejected": -1.8473920822143555,
|
2091 |
+
"logps/chosen": -165.49359130859375,
|
2092 |
+
"logps/rejected": -273.9131774902344,
|
2093 |
+
"loss": 0.0071,
|
2094 |
+
"rewards/accuracies": 1.0,
|
2095 |
+
"rewards/chosen": 1.4215660095214844,
|
2096 |
+
"rewards/margins": 10.477560043334961,
|
2097 |
+
"rewards/rejected": -9.055994033813477,
|
2098 |
+
"step": 1030
|
2099 |
+
},
|
2100 |
+
{
|
2101 |
+
"epoch": 4.24,
|
2102 |
+
"grad_norm": 0.0017074132338166237,
|
2103 |
+
"learning_rate": 1.529824561403509e-06,
|
2104 |
+
"logits/chosen": -1.7763557434082031,
|
2105 |
+
"logits/rejected": -1.8936008214950562,
|
2106 |
+
"logps/chosen": -133.30006408691406,
|
2107 |
+
"logps/rejected": -336.57427978515625,
|
2108 |
+
"loss": 0.018,
|
2109 |
+
"rewards/accuracies": 1.0,
|
2110 |
+
"rewards/chosen": 4.363164901733398,
|
2111 |
+
"rewards/margins": 14.957908630371094,
|
2112 |
+
"rewards/rejected": -10.594743728637695,
|
2113 |
+
"step": 1040
|
2114 |
+
},
|
2115 |
+
{
|
2116 |
+
"epoch": 4.29,
|
2117 |
+
"grad_norm": 66.57980346679688,
|
2118 |
+
"learning_rate": 1.4947368421052632e-06,
|
2119 |
+
"logits/chosen": -1.835628867149353,
|
2120 |
+
"logits/rejected": -1.8976163864135742,
|
2121 |
+
"logps/chosen": -177.20309448242188,
|
2122 |
+
"logps/rejected": -287.8599548339844,
|
2123 |
+
"loss": 0.0131,
|
2124 |
+
"rewards/accuracies": 1.0,
|
2125 |
+
"rewards/chosen": 2.568208694458008,
|
2126 |
+
"rewards/margins": 11.586195945739746,
|
2127 |
+
"rewards/rejected": -9.017987251281738,
|
2128 |
+
"step": 1050
|
2129 |
+
},
|
2130 |
+
{
|
2131 |
+
"epoch": 4.29,
|
2132 |
+
"eval_logits/chosen": -1.8221967220306396,
|
2133 |
+
"eval_logits/rejected": -1.8615671396255493,
|
2134 |
+
"eval_logps/chosen": -188.5262451171875,
|
2135 |
+
"eval_logps/rejected": -261.6840515136719,
|
2136 |
+
"eval_loss": 0.13124893605709076,
|
2137 |
+
"eval_rewards/accuracies": 0.9285714030265808,
|
2138 |
+
"eval_rewards/chosen": 1.805002212524414,
|
2139 |
+
"eval_rewards/margins": 7.354526042938232,
|
2140 |
+
"eval_rewards/rejected": -5.549524784088135,
|
2141 |
+
"eval_runtime": 26.2328,
|
2142 |
+
"eval_samples_per_second": 1.067,
|
2143 |
+
"eval_steps_per_second": 1.067,
|
2144 |
+
"step": 1050
|
2145 |
+
},
|
2146 |
+
{
|
2147 |
+
"epoch": 4.33,
|
2148 |
+
"grad_norm": 0.027719292789697647,
|
2149 |
+
"learning_rate": 1.4596491228070175e-06,
|
2150 |
+
"logits/chosen": -1.7560068368911743,
|
2151 |
+
"logits/rejected": -1.8672195672988892,
|
2152 |
+
"logps/chosen": -189.530029296875,
|
2153 |
+
"logps/rejected": -330.6167907714844,
|
2154 |
+
"loss": 0.003,
|
2155 |
+
"rewards/accuracies": 1.0,
|
2156 |
+
"rewards/chosen": 1.0810093879699707,
|
2157 |
+
"rewards/margins": 9.271745681762695,
|
2158 |
+
"rewards/rejected": -8.19073486328125,
|
2159 |
+
"step": 1060
|
2160 |
+
},
|
2161 |
+
{
|
2162 |
+
"epoch": 4.37,
|
2163 |
+
"grad_norm": 0.0,
|
2164 |
+
"learning_rate": 1.4245614035087722e-06,
|
2165 |
+
"logits/chosen": -1.8792110681533813,
|
2166 |
+
"logits/rejected": -1.9164355993270874,
|
2167 |
+
"logps/chosen": -212.42788696289062,
|
2168 |
+
"logps/rejected": -238.2472686767578,
|
2169 |
+
"loss": 0.1334,
|
2170 |
+
"rewards/accuracies": 0.8999999761581421,
|
2171 |
+
"rewards/chosen": 1.0076087713241577,
|
2172 |
+
"rewards/margins": 9.041505813598633,
|
2173 |
+
"rewards/rejected": -8.033895492553711,
|
2174 |
+
"step": 1070
|
2175 |
+
},
|
2176 |
+
{
|
2177 |
+
"epoch": 4.41,
|
2178 |
+
"grad_norm": 3.609489340306027e-06,
|
2179 |
+
"learning_rate": 1.3894736842105263e-06,
|
2180 |
+
"logits/chosen": -1.8213157653808594,
|
2181 |
+
"logits/rejected": -1.870356798171997,
|
2182 |
+
"logps/chosen": -80.74689483642578,
|
2183 |
+
"logps/rejected": -158.96109008789062,
|
2184 |
+
"loss": 0.0868,
|
2185 |
+
"rewards/accuracies": 1.0,
|
2186 |
+
"rewards/chosen": 2.509101629257202,
|
2187 |
+
"rewards/margins": 7.621730804443359,
|
2188 |
+
"rewards/rejected": -5.11262845993042,
|
2189 |
+
"step": 1080
|
2190 |
+
},
|
2191 |
+
{
|
2192 |
+
"epoch": 4.41,
|
2193 |
+
"eval_logits/chosen": -1.81948721408844,
|
2194 |
+
"eval_logits/rejected": -1.858702540397644,
|
2195 |
+
"eval_logps/chosen": -188.66748046875,
|
2196 |
+
"eval_logps/rejected": -261.94708251953125,
|
2197 |
+
"eval_loss": 0.1210150271654129,
|
2198 |
+
"eval_rewards/accuracies": 0.9285714030265808,
|
2199 |
+
"eval_rewards/chosen": 1.76263427734375,
|
2200 |
+
"eval_rewards/margins": 7.39107608795166,
|
2201 |
+
"eval_rewards/rejected": -5.62844181060791,
|
2202 |
+
"eval_runtime": 26.2393,
|
2203 |
+
"eval_samples_per_second": 1.067,
|
2204 |
+
"eval_steps_per_second": 1.067,
|
2205 |
+
"step": 1080
|
2206 |
+
}
|
2207 |
+
],
|
2208 |
+
"logging_steps": 10,
|
2209 |
+
"max_steps": 1470,
|
2210 |
+
"num_input_tokens_seen": 0,
|
2211 |
+
"num_train_epochs": 6,
|
2212 |
+
"save_steps": 90,
|
2213 |
+
"total_flos": 0.0,
|
2214 |
+
"train_batch_size": 1,
|
2215 |
+
"trial_name": null,
|
2216 |
+
"trial_params": null
|
2217 |
+
}
|
checkpoint-1080/training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:41693834eff4b96ebb423cc3037dde0e25267c10f91f2e959260eee7bddf1ddb
|
3 |
+
size 5304
|
checkpoint-1170/README.md
ADDED
@@ -0,0 +1,202 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
library_name: peft
|
3 |
+
base_model: Weni/WeniGPT-Agents-Mistral-1.0.6-SFT-merged
|
4 |
+
---
|
5 |
+
|
6 |
+
# Model Card for Model ID
|
7 |
+
|
8 |
+
<!-- Provide a quick summary of what the model is/does. -->
|
9 |
+
|
10 |
+
|
11 |
+
|
12 |
+
## Model Details
|
13 |
+
|
14 |
+
### Model Description
|
15 |
+
|
16 |
+
<!-- Provide a longer summary of what this model is. -->
|
17 |
+
|
18 |
+
|
19 |
+
|
20 |
+
- **Developed by:** [More Information Needed]
|
21 |
+
- **Funded by [optional]:** [More Information Needed]
|
22 |
+
- **Shared by [optional]:** [More Information Needed]
|
23 |
+
- **Model type:** [More Information Needed]
|
24 |
+
- **Language(s) (NLP):** [More Information Needed]
|
25 |
+
- **License:** [More Information Needed]
|
26 |
+
- **Finetuned from model [optional]:** [More Information Needed]
|
27 |
+
|
28 |
+
### Model Sources [optional]
|
29 |
+
|
30 |
+
<!-- Provide the basic links for the model. -->
|
31 |
+
|
32 |
+
- **Repository:** [More Information Needed]
|
33 |
+
- **Paper [optional]:** [More Information Needed]
|
34 |
+
- **Demo [optional]:** [More Information Needed]
|
35 |
+
|
36 |
+
## Uses
|
37 |
+
|
38 |
+
<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
|
39 |
+
|
40 |
+
### Direct Use
|
41 |
+
|
42 |
+
<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
|
43 |
+
|
44 |
+
[More Information Needed]
|
45 |
+
|
46 |
+
### Downstream Use [optional]
|
47 |
+
|
48 |
+
<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
|
49 |
+
|
50 |
+
[More Information Needed]
|
51 |
+
|
52 |
+
### Out-of-Scope Use
|
53 |
+
|
54 |
+
<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
|
55 |
+
|
56 |
+
[More Information Needed]
|
57 |
+
|
58 |
+
## Bias, Risks, and Limitations
|
59 |
+
|
60 |
+
<!-- This section is meant to convey both technical and sociotechnical limitations. -->
|
61 |
+
|
62 |
+
[More Information Needed]
|
63 |
+
|
64 |
+
### Recommendations
|
65 |
+
|
66 |
+
<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
|
67 |
+
|
68 |
+
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
|
69 |
+
|
70 |
+
## How to Get Started with the Model
|
71 |
+
|
72 |
+
Use the code below to get started with the model.
|
73 |
+
|
74 |
+
[More Information Needed]
|
75 |
+
|
76 |
+
## Training Details
|
77 |
+
|
78 |
+
### Training Data
|
79 |
+
|
80 |
+
<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
|
81 |
+
|
82 |
+
[More Information Needed]
|
83 |
+
|
84 |
+
### Training Procedure
|
85 |
+
|
86 |
+
<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
|
87 |
+
|
88 |
+
#### Preprocessing [optional]
|
89 |
+
|
90 |
+
[More Information Needed]
|
91 |
+
|
92 |
+
|
93 |
+
#### Training Hyperparameters
|
94 |
+
|
95 |
+
- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
|
96 |
+
|
97 |
+
#### Speeds, Sizes, Times [optional]
|
98 |
+
|
99 |
+
<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
|
100 |
+
|
101 |
+
[More Information Needed]
|
102 |
+
|
103 |
+
## Evaluation
|
104 |
+
|
105 |
+
<!-- This section describes the evaluation protocols and provides the results. -->
|
106 |
+
|
107 |
+
### Testing Data, Factors & Metrics
|
108 |
+
|
109 |
+
#### Testing Data
|
110 |
+
|
111 |
+
<!-- This should link to a Dataset Card if possible. -->
|
112 |
+
|
113 |
+
[More Information Needed]
|
114 |
+
|
115 |
+
#### Factors
|
116 |
+
|
117 |
+
<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
|
118 |
+
|
119 |
+
[More Information Needed]
|
120 |
+
|
121 |
+
#### Metrics
|
122 |
+
|
123 |
+
<!-- These are the evaluation metrics being used, ideally with a description of why. -->
|
124 |
+
|
125 |
+
[More Information Needed]
|
126 |
+
|
127 |
+
### Results
|
128 |
+
|
129 |
+
[More Information Needed]
|
130 |
+
|
131 |
+
#### Summary
|
132 |
+
|
133 |
+
|
134 |
+
|
135 |
+
## Model Examination [optional]
|
136 |
+
|
137 |
+
<!-- Relevant interpretability work for the model goes here -->
|
138 |
+
|
139 |
+
[More Information Needed]
|
140 |
+
|
141 |
+
## Environmental Impact
|
142 |
+
|
143 |
+
<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
|
144 |
+
|
145 |
+
Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
|
146 |
+
|
147 |
+
- **Hardware Type:** [More Information Needed]
|
148 |
+
- **Hours used:** [More Information Needed]
|
149 |
+
- **Cloud Provider:** [More Information Needed]
|
150 |
+
- **Compute Region:** [More Information Needed]
|
151 |
+
- **Carbon Emitted:** [More Information Needed]
|
152 |
+
|
153 |
+
## Technical Specifications [optional]
|
154 |
+
|
155 |
+
### Model Architecture and Objective
|
156 |
+
|
157 |
+
[More Information Needed]
|
158 |
+
|
159 |
+
### Compute Infrastructure
|
160 |
+
|
161 |
+
[More Information Needed]
|
162 |
+
|
163 |
+
#### Hardware
|
164 |
+
|
165 |
+
[More Information Needed]
|
166 |
+
|
167 |
+
#### Software
|
168 |
+
|
169 |
+
[More Information Needed]
|
170 |
+
|
171 |
+
## Citation [optional]
|
172 |
+
|
173 |
+
<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
|
174 |
+
|
175 |
+
**BibTeX:**
|
176 |
+
|
177 |
+
[More Information Needed]
|
178 |
+
|
179 |
+
**APA:**
|
180 |
+
|
181 |
+
[More Information Needed]
|
182 |
+
|
183 |
+
## Glossary [optional]
|
184 |
+
|
185 |
+
<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
|
186 |
+
|
187 |
+
[More Information Needed]
|
188 |
+
|
189 |
+
## More Information [optional]
|
190 |
+
|
191 |
+
[More Information Needed]
|
192 |
+
|
193 |
+
## Model Card Authors [optional]
|
194 |
+
|
195 |
+
[More Information Needed]
|
196 |
+
|
197 |
+
## Model Card Contact
|
198 |
+
|
199 |
+
[More Information Needed]
|
200 |
+
### Framework versions
|
201 |
+
|
202 |
+
- PEFT 0.10.0
|
checkpoint-1170/adapter_config.json
ADDED
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"alpha_pattern": {},
|
3 |
+
"auto_mapping": null,
|
4 |
+
"base_model_name_or_path": "Weni/WeniGPT-Agents-Mistral-1.0.6-SFT-merged",
|
5 |
+
"bias": "none",
|
6 |
+
"fan_in_fan_out": false,
|
7 |
+
"inference_mode": true,
|
8 |
+
"init_lora_weights": true,
|
9 |
+
"layer_replication": null,
|
10 |
+
"layers_pattern": null,
|
11 |
+
"layers_to_transform": null,
|
12 |
+
"loftq_config": {},
|
13 |
+
"lora_alpha": 16,
|
14 |
+
"lora_dropout": 0.05,
|
15 |
+
"megatron_config": null,
|
16 |
+
"megatron_core": "megatron.core",
|
17 |
+
"modules_to_save": null,
|
18 |
+
"peft_type": "LORA",
|
19 |
+
"r": 8,
|
20 |
+
"rank_pattern": {},
|
21 |
+
"revision": null,
|
22 |
+
"target_modules": [
|
23 |
+
"q_proj",
|
24 |
+
"v_proj"
|
25 |
+
],
|
26 |
+
"task_type": "CAUSAL_LM",
|
27 |
+
"use_dora": false,
|
28 |
+
"use_rslora": false
|
29 |
+
}
|
checkpoint-1170/adapter_model.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b22e04f497eece462e7ecc3596e659c0b1f1de14397e1ae6b93eea4f348668e9
|
3 |
+
size 13648432
|
checkpoint-1170/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3467918e27991f309664976d65b8b0480c79633d309fe717dfb1d0facefe86eb
|
3 |
+
size 27370618
|
checkpoint-1170/rng_state.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a446f0425374fa045ac6c683a19a8525794cbf14fa7b3d3378d950e0a4a18370
|
3 |
+
size 14244
|
checkpoint-1170/scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:2953bffe864c21970454953655761421a814ade23dec50ad6ba515056da82252
|
3 |
+
size 1064
|
checkpoint-1170/special_tokens_map.json
ADDED
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"bos_token": {
|
3 |
+
"content": "<s>",
|
4 |
+
"lstrip": false,
|
5 |
+
"normalized": false,
|
6 |
+
"rstrip": false,
|
7 |
+
"single_word": false
|
8 |
+
},
|
9 |
+
"eos_token": {
|
10 |
+
"content": "</s>",
|
11 |
+
"lstrip": false,
|
12 |
+
"normalized": false,
|
13 |
+
"rstrip": false,
|
14 |
+
"single_word": false
|
15 |
+
},
|
16 |
+
"pad_token": "<unk>",
|
17 |
+
"unk_token": {
|
18 |
+
"content": "<unk>",
|
19 |
+
"lstrip": false,
|
20 |
+
"normalized": false,
|
21 |
+
"rstrip": false,
|
22 |
+
"single_word": false
|
23 |
+
}
|
24 |
+
}
|
checkpoint-1170/tokenizer.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-1170/tokenizer.model
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:dadfd56d766715c61d2ef780a525ab43b8e6da4de6865bda3d95fdef5e134055
|
3 |
+
size 493443
|
checkpoint-1170/tokenizer_config.json
ADDED
@@ -0,0 +1,49 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"add_bos_token": true,
|
3 |
+
"add_eos_token": false,
|
4 |
+
"added_tokens_decoder": {
|
5 |
+
"0": {
|
6 |
+
"content": "<unk>",
|
7 |
+
"lstrip": false,
|
8 |
+
"normalized": false,
|
9 |
+
"rstrip": false,
|
10 |
+
"single_word": false,
|
11 |
+
"special": true
|
12 |
+
},
|
13 |
+
"1": {
|
14 |
+
"content": "<s>",
|
15 |
+
"lstrip": false,
|
16 |
+
"normalized": false,
|
17 |
+
"rstrip": false,
|
18 |
+
"single_word": false,
|
19 |
+
"special": true
|
20 |
+
},
|
21 |
+
"2": {
|
22 |
+
"content": "</s>",
|
23 |
+
"lstrip": false,
|
24 |
+
"normalized": false,
|
25 |
+
"rstrip": false,
|
26 |
+
"single_word": false,
|
27 |
+
"special": true
|
28 |
+
}
|
29 |
+
},
|
30 |
+
"additional_special_tokens": [],
|
31 |
+
"bos_token": "<s>",
|
32 |
+
"chat_template": "{% for message in messages %}{% if message['role'] == 'user' %}{{ bos_token + '[INST] ' + message['content'] + ' [/INST]' }}{% elif message['role'] == 'system' %}{{ '<<SYS>>\\n' + message['content'] + '\\n<</SYS>>\\n\\n' }}{% elif message['role'] == 'assistant' %}{{ ' ' + message['content'] + ' ' + eos_token }}{% endif %}{% endfor %}",
|
33 |
+
"clean_up_tokenization_spaces": false,
|
34 |
+
"eos_token": "</s>",
|
35 |
+
"legacy": true,
|
36 |
+
"max_lenght": 8192,
|
37 |
+
"max_length": 8192,
|
38 |
+
"model_max_length": 1000000000000000019884624838656,
|
39 |
+
"pad_token": "<unk>",
|
40 |
+
"padding": true,
|
41 |
+
"sp_model_kwargs": {},
|
42 |
+
"spaces_between_special_tokens": false,
|
43 |
+
"stride": 0,
|
44 |
+
"tokenizer_class": "LlamaTokenizer",
|
45 |
+
"truncation_side": "right",
|
46 |
+
"truncation_strategy": "longest_first",
|
47 |
+
"unk_token": "<unk>",
|
48 |
+
"use_default_system_prompt": false
|
49 |
+
}
|
checkpoint-1170/trainer_state.json
ADDED
@@ -0,0 +1,2400 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": 0.11231832951307297,
|
3 |
+
"best_model_checkpoint": "./mistral/20-04-24-Weni-WeniGPT-Agents-Mistral-1.0.6-SFT-1.0.7-DPO_Experiment on DPO with other hyperparameters and best SFT model of WeniGPT-2_max_steps-1470_batch_1_2024-04-20_ppid_9/checkpoint-1170",
|
4 |
+
"epoch": 4.775510204081632,
|
5 |
+
"eval_steps": 30,
|
6 |
+
"global_step": 1170,
|
7 |
+
"is_hyper_param_search": false,
|
8 |
+
"is_local_process_zero": true,
|
9 |
+
"is_world_process_zero": true,
|
10 |
+
"log_history": [
|
11 |
+
{
|
12 |
+
"epoch": 0.04,
|
13 |
+
"grad_norm": 83.44550323486328,
|
14 |
+
"learning_rate": 6.666666666666667e-07,
|
15 |
+
"logits/chosen": -1.7825796604156494,
|
16 |
+
"logits/rejected": -1.7886196374893188,
|
17 |
+
"logps/chosen": -169.71481323242188,
|
18 |
+
"logps/rejected": -167.6764373779297,
|
19 |
+
"loss": 0.6937,
|
20 |
+
"rewards/accuracies": 0.20000000298023224,
|
21 |
+
"rewards/chosen": -0.0002815247280523181,
|
22 |
+
"rewards/margins": -0.0010146332206204534,
|
23 |
+
"rewards/rejected": 0.0007331084925681353,
|
24 |
+
"step": 10
|
25 |
+
},
|
26 |
+
{
|
27 |
+
"epoch": 0.08,
|
28 |
+
"grad_norm": 87.56982421875,
|
29 |
+
"learning_rate": 1.777777777777778e-06,
|
30 |
+
"logits/chosen": -1.623265027999878,
|
31 |
+
"logits/rejected": -1.6968040466308594,
|
32 |
+
"logps/chosen": -128.7312469482422,
|
33 |
+
"logps/rejected": -184.20187377929688,
|
34 |
+
"loss": 0.6912,
|
35 |
+
"rewards/accuracies": 0.5,
|
36 |
+
"rewards/chosen": 0.0030528828501701355,
|
37 |
+
"rewards/margins": 0.004004458896815777,
|
38 |
+
"rewards/rejected": -0.0009515761630609632,
|
39 |
+
"step": 20
|
40 |
+
},
|
41 |
+
{
|
42 |
+
"epoch": 0.12,
|
43 |
+
"grad_norm": 113.4741439819336,
|
44 |
+
"learning_rate": 2.888888888888889e-06,
|
45 |
+
"logits/chosen": -1.7152118682861328,
|
46 |
+
"logits/rejected": -1.758681058883667,
|
47 |
+
"logps/chosen": -146.91494750976562,
|
48 |
+
"logps/rejected": -213.5966339111328,
|
49 |
+
"loss": 0.6781,
|
50 |
+
"rewards/accuracies": 1.0,
|
51 |
+
"rewards/chosen": 0.03251449763774872,
|
52 |
+
"rewards/margins": 0.030697176232933998,
|
53 |
+
"rewards/rejected": 0.0018173219868913293,
|
54 |
+
"step": 30
|
55 |
+
},
|
56 |
+
{
|
57 |
+
"epoch": 0.12,
|
58 |
+
"eval_logits/chosen": -1.795080304145813,
|
59 |
+
"eval_logits/rejected": -1.8307874202728271,
|
60 |
+
"eval_logps/chosen": -194.3749542236328,
|
61 |
+
"eval_logps/rejected": -243.1332244873047,
|
62 |
+
"eval_loss": 0.6762310266494751,
|
63 |
+
"eval_rewards/accuracies": 0.75,
|
64 |
+
"eval_rewards/chosen": 0.05039369314908981,
|
65 |
+
"eval_rewards/margins": 0.03467999026179314,
|
66 |
+
"eval_rewards/rejected": 0.01571369729936123,
|
67 |
+
"eval_runtime": 26.2365,
|
68 |
+
"eval_samples_per_second": 1.067,
|
69 |
+
"eval_steps_per_second": 1.067,
|
70 |
+
"step": 30
|
71 |
+
},
|
72 |
+
{
|
73 |
+
"epoch": 0.16,
|
74 |
+
"grad_norm": 50.695308685302734,
|
75 |
+
"learning_rate": 4.000000000000001e-06,
|
76 |
+
"logits/chosen": -1.9357620477676392,
|
77 |
+
"logits/rejected": -1.9375683069229126,
|
78 |
+
"logps/chosen": -183.8218231201172,
|
79 |
+
"logps/rejected": -115.80632019042969,
|
80 |
+
"loss": 0.6768,
|
81 |
+
"rewards/accuracies": 0.699999988079071,
|
82 |
+
"rewards/chosen": 0.056698836386203766,
|
83 |
+
"rewards/margins": 0.03413978964090347,
|
84 |
+
"rewards/rejected": 0.02255905233323574,
|
85 |
+
"step": 40
|
86 |
+
},
|
87 |
+
{
|
88 |
+
"epoch": 0.2,
|
89 |
+
"grad_norm": 71.6834716796875,
|
90 |
+
"learning_rate": 4.996491228070176e-06,
|
91 |
+
"logits/chosen": -1.7773185968399048,
|
92 |
+
"logits/rejected": -1.8331743478775024,
|
93 |
+
"logps/chosen": -167.78533935546875,
|
94 |
+
"logps/rejected": -227.0582733154297,
|
95 |
+
"loss": 0.6205,
|
96 |
+
"rewards/accuracies": 0.8999999761581421,
|
97 |
+
"rewards/chosen": 0.1449195146560669,
|
98 |
+
"rewards/margins": 0.15578031539916992,
|
99 |
+
"rewards/rejected": -0.010860783979296684,
|
100 |
+
"step": 50
|
101 |
+
},
|
102 |
+
{
|
103 |
+
"epoch": 0.24,
|
104 |
+
"grad_norm": 93.83502197265625,
|
105 |
+
"learning_rate": 4.96140350877193e-06,
|
106 |
+
"logits/chosen": -1.7739536762237549,
|
107 |
+
"logits/rejected": -1.8414790630340576,
|
108 |
+
"logps/chosen": -153.57481384277344,
|
109 |
+
"logps/rejected": -194.3485870361328,
|
110 |
+
"loss": 0.5918,
|
111 |
+
"rewards/accuracies": 0.8999999761581421,
|
112 |
+
"rewards/chosen": 0.2464807778596878,
|
113 |
+
"rewards/margins": 0.22958576679229736,
|
114 |
+
"rewards/rejected": 0.016895027831196785,
|
115 |
+
"step": 60
|
116 |
+
},
|
117 |
+
{
|
118 |
+
"epoch": 0.24,
|
119 |
+
"eval_logits/chosen": -1.7974774837493896,
|
120 |
+
"eval_logits/rejected": -1.8332849740982056,
|
121 |
+
"eval_logps/chosen": -193.71743774414062,
|
122 |
+
"eval_logps/rejected": -243.05783081054688,
|
123 |
+
"eval_loss": 0.5998325943946838,
|
124 |
+
"eval_rewards/accuracies": 0.7857142686843872,
|
125 |
+
"eval_rewards/chosen": 0.24764502048492432,
|
126 |
+
"eval_rewards/margins": 0.20930658280849457,
|
127 |
+
"eval_rewards/rejected": 0.03833846375346184,
|
128 |
+
"eval_runtime": 26.2349,
|
129 |
+
"eval_samples_per_second": 1.067,
|
130 |
+
"eval_steps_per_second": 1.067,
|
131 |
+
"step": 60
|
132 |
+
},
|
133 |
+
{
|
134 |
+
"epoch": 0.29,
|
135 |
+
"grad_norm": 109.71602630615234,
|
136 |
+
"learning_rate": 4.926315789473685e-06,
|
137 |
+
"logits/chosen": -1.8950660228729248,
|
138 |
+
"logits/rejected": -1.9148075580596924,
|
139 |
+
"logps/chosen": -169.89028930664062,
|
140 |
+
"logps/rejected": -177.87301635742188,
|
141 |
+
"loss": 0.5992,
|
142 |
+
"rewards/accuracies": 0.800000011920929,
|
143 |
+
"rewards/chosen": 0.15271134674549103,
|
144 |
+
"rewards/margins": 0.21031668782234192,
|
145 |
+
"rewards/rejected": -0.0576053187251091,
|
146 |
+
"step": 70
|
147 |
+
},
|
148 |
+
{
|
149 |
+
"epoch": 0.33,
|
150 |
+
"grad_norm": 90.3881607055664,
|
151 |
+
"learning_rate": 4.8912280701754386e-06,
|
152 |
+
"logits/chosen": -1.798600196838379,
|
153 |
+
"logits/rejected": -1.855521559715271,
|
154 |
+
"logps/chosen": -160.3217315673828,
|
155 |
+
"logps/rejected": -247.6388397216797,
|
156 |
+
"loss": 0.5551,
|
157 |
+
"rewards/accuracies": 0.699999988079071,
|
158 |
+
"rewards/chosen": 0.29141485691070557,
|
159 |
+
"rewards/margins": 0.3369868993759155,
|
160 |
+
"rewards/rejected": -0.04557197913527489,
|
161 |
+
"step": 80
|
162 |
+
},
|
163 |
+
{
|
164 |
+
"epoch": 0.37,
|
165 |
+
"grad_norm": 147.27755737304688,
|
166 |
+
"learning_rate": 4.856140350877193e-06,
|
167 |
+
"logits/chosen": -1.8193047046661377,
|
168 |
+
"logits/rejected": -1.8773219585418701,
|
169 |
+
"logps/chosen": -155.37123107910156,
|
170 |
+
"logps/rejected": -245.6790008544922,
|
171 |
+
"loss": 0.4932,
|
172 |
+
"rewards/accuracies": 0.699999988079071,
|
173 |
+
"rewards/chosen": 0.5351110696792603,
|
174 |
+
"rewards/margins": 0.5927553176879883,
|
175 |
+
"rewards/rejected": -0.05764435604214668,
|
176 |
+
"step": 90
|
177 |
+
},
|
178 |
+
{
|
179 |
+
"epoch": 0.37,
|
180 |
+
"eval_logits/chosen": -1.8004387617111206,
|
181 |
+
"eval_logits/rejected": -1.836446762084961,
|
182 |
+
"eval_logps/chosen": -192.66908264160156,
|
183 |
+
"eval_logps/rejected": -242.95899963378906,
|
184 |
+
"eval_loss": 0.5072147846221924,
|
185 |
+
"eval_rewards/accuracies": 0.8214285969734192,
|
186 |
+
"eval_rewards/chosen": 0.5621528029441833,
|
187 |
+
"eval_rewards/margins": 0.4941706657409668,
|
188 |
+
"eval_rewards/rejected": 0.06798211485147476,
|
189 |
+
"eval_runtime": 26.2422,
|
190 |
+
"eval_samples_per_second": 1.067,
|
191 |
+
"eval_steps_per_second": 1.067,
|
192 |
+
"step": 90
|
193 |
+
},
|
194 |
+
{
|
195 |
+
"epoch": 0.41,
|
196 |
+
"grad_norm": 62.75928497314453,
|
197 |
+
"learning_rate": 4.821052631578948e-06,
|
198 |
+
"logits/chosen": -1.7835988998413086,
|
199 |
+
"logits/rejected": -1.8885362148284912,
|
200 |
+
"logps/chosen": -109.94624328613281,
|
201 |
+
"logps/rejected": -232.5241241455078,
|
202 |
+
"loss": 0.4252,
|
203 |
+
"rewards/accuracies": 1.0,
|
204 |
+
"rewards/chosen": 0.5394761562347412,
|
205 |
+
"rewards/margins": 0.787738025188446,
|
206 |
+
"rewards/rejected": -0.248261958360672,
|
207 |
+
"step": 100
|
208 |
+
},
|
209 |
+
{
|
210 |
+
"epoch": 0.45,
|
211 |
+
"grad_norm": 84.42039489746094,
|
212 |
+
"learning_rate": 4.785964912280702e-06,
|
213 |
+
"logits/chosen": -1.7081111669540405,
|
214 |
+
"logits/rejected": -1.7244077920913696,
|
215 |
+
"logps/chosen": -92.51998138427734,
|
216 |
+
"logps/rejected": -135.5617218017578,
|
217 |
+
"loss": 0.5077,
|
218 |
+
"rewards/accuracies": 0.800000011920929,
|
219 |
+
"rewards/chosen": 0.6028972864151001,
|
220 |
+
"rewards/margins": 0.511993408203125,
|
221 |
+
"rewards/rejected": 0.09090389311313629,
|
222 |
+
"step": 110
|
223 |
+
},
|
224 |
+
{
|
225 |
+
"epoch": 0.49,
|
226 |
+
"grad_norm": 26.61499786376953,
|
227 |
+
"learning_rate": 4.754385964912281e-06,
|
228 |
+
"logits/chosen": -1.9250166416168213,
|
229 |
+
"logits/rejected": -1.9426968097686768,
|
230 |
+
"logps/chosen": -152.0164337158203,
|
231 |
+
"logps/rejected": -154.2359161376953,
|
232 |
+
"loss": 0.4391,
|
233 |
+
"rewards/accuracies": 0.8999999761581421,
|
234 |
+
"rewards/chosen": 0.7938548922538757,
|
235 |
+
"rewards/margins": 0.8368040919303894,
|
236 |
+
"rewards/rejected": -0.042949117720127106,
|
237 |
+
"step": 120
|
238 |
+
},
|
239 |
+
{
|
240 |
+
"epoch": 0.49,
|
241 |
+
"eval_logits/chosen": -1.8051280975341797,
|
242 |
+
"eval_logits/rejected": -1.8413046598434448,
|
243 |
+
"eval_logps/chosen": -191.2982177734375,
|
244 |
+
"eval_logps/rejected": -242.81198120117188,
|
245 |
+
"eval_loss": 0.4335751235485077,
|
246 |
+
"eval_rewards/accuracies": 0.7857142686843872,
|
247 |
+
"eval_rewards/chosen": 0.9734136462211609,
|
248 |
+
"eval_rewards/margins": 0.8613229990005493,
|
249 |
+
"eval_rewards/rejected": 0.11209066957235336,
|
250 |
+
"eval_runtime": 26.2345,
|
251 |
+
"eval_samples_per_second": 1.067,
|
252 |
+
"eval_steps_per_second": 1.067,
|
253 |
+
"step": 120
|
254 |
+
},
|
255 |
+
{
|
256 |
+
"epoch": 0.53,
|
257 |
+
"grad_norm": 90.6368637084961,
|
258 |
+
"learning_rate": 4.7192982456140355e-06,
|
259 |
+
"logits/chosen": -1.865144968032837,
|
260 |
+
"logits/rejected": -1.9209120273590088,
|
261 |
+
"logps/chosen": -149.78990173339844,
|
262 |
+
"logps/rejected": -293.8749084472656,
|
263 |
+
"loss": 0.4133,
|
264 |
+
"rewards/accuracies": 0.800000011920929,
|
265 |
+
"rewards/chosen": 0.4554510712623596,
|
266 |
+
"rewards/margins": 0.9285024404525757,
|
267 |
+
"rewards/rejected": -0.47305139899253845,
|
268 |
+
"step": 130
|
269 |
+
},
|
270 |
+
{
|
271 |
+
"epoch": 0.57,
|
272 |
+
"grad_norm": 6.8597941398620605,
|
273 |
+
"learning_rate": 4.68421052631579e-06,
|
274 |
+
"logits/chosen": -1.8089487552642822,
|
275 |
+
"logits/rejected": -1.8955650329589844,
|
276 |
+
"logps/chosen": -100.26250457763672,
|
277 |
+
"logps/rejected": -209.1724090576172,
|
278 |
+
"loss": 0.3414,
|
279 |
+
"rewards/accuracies": 1.0,
|
280 |
+
"rewards/chosen": 0.9389357566833496,
|
281 |
+
"rewards/margins": 1.521246314048767,
|
282 |
+
"rewards/rejected": -0.5823107361793518,
|
283 |
+
"step": 140
|
284 |
+
},
|
285 |
+
{
|
286 |
+
"epoch": 0.61,
|
287 |
+
"grad_norm": 79.9189453125,
|
288 |
+
"learning_rate": 4.652631578947368e-06,
|
289 |
+
"logits/chosen": -1.760371446609497,
|
290 |
+
"logits/rejected": -1.8096182346343994,
|
291 |
+
"logps/chosen": -193.62579345703125,
|
292 |
+
"logps/rejected": -166.35800170898438,
|
293 |
+
"loss": 0.3208,
|
294 |
+
"rewards/accuracies": 0.8999999761581421,
|
295 |
+
"rewards/chosen": 1.2600924968719482,
|
296 |
+
"rewards/margins": 1.5845091342926025,
|
297 |
+
"rewards/rejected": -0.32441645860671997,
|
298 |
+
"step": 150
|
299 |
+
},
|
300 |
+
{
|
301 |
+
"epoch": 0.61,
|
302 |
+
"eval_logits/chosen": -1.8129692077636719,
|
303 |
+
"eval_logits/rejected": -1.849239706993103,
|
304 |
+
"eval_logps/chosen": -189.88929748535156,
|
305 |
+
"eval_logps/rejected": -242.91104125976562,
|
306 |
+
"eval_loss": 0.39326220750808716,
|
307 |
+
"eval_rewards/accuracies": 0.7857142686843872,
|
308 |
+
"eval_rewards/chosen": 1.3960883617401123,
|
309 |
+
"eval_rewards/margins": 1.3137177228927612,
|
310 |
+
"eval_rewards/rejected": 0.08237060159444809,
|
311 |
+
"eval_runtime": 26.3372,
|
312 |
+
"eval_samples_per_second": 1.063,
|
313 |
+
"eval_steps_per_second": 1.063,
|
314 |
+
"step": 150
|
315 |
+
},
|
316 |
+
{
|
317 |
+
"epoch": 0.65,
|
318 |
+
"grad_norm": 3.151426315307617,
|
319 |
+
"learning_rate": 4.617543859649123e-06,
|
320 |
+
"logits/chosen": -1.7185112237930298,
|
321 |
+
"logits/rejected": -1.749427080154419,
|
322 |
+
"logps/chosen": -178.8656463623047,
|
323 |
+
"logps/rejected": -201.32858276367188,
|
324 |
+
"loss": 0.4418,
|
325 |
+
"rewards/accuracies": 0.699999988079071,
|
326 |
+
"rewards/chosen": 1.3884060382843018,
|
327 |
+
"rewards/margins": 1.4542597532272339,
|
328 |
+
"rewards/rejected": -0.06585375964641571,
|
329 |
+
"step": 160
|
330 |
+
},
|
331 |
+
{
|
332 |
+
"epoch": 0.69,
|
333 |
+
"grad_norm": 37.79099655151367,
|
334 |
+
"learning_rate": 4.582456140350878e-06,
|
335 |
+
"logits/chosen": -1.730787992477417,
|
336 |
+
"logits/rejected": -1.8099721670150757,
|
337 |
+
"logps/chosen": -158.50277709960938,
|
338 |
+
"logps/rejected": -185.48129272460938,
|
339 |
+
"loss": 0.3912,
|
340 |
+
"rewards/accuracies": 0.800000011920929,
|
341 |
+
"rewards/chosen": 1.4253065586090088,
|
342 |
+
"rewards/margins": 1.2418479919433594,
|
343 |
+
"rewards/rejected": 0.18345877528190613,
|
344 |
+
"step": 170
|
345 |
+
},
|
346 |
+
{
|
347 |
+
"epoch": 0.73,
|
348 |
+
"grad_norm": 18.49593734741211,
|
349 |
+
"learning_rate": 4.547368421052632e-06,
|
350 |
+
"logits/chosen": -1.842390775680542,
|
351 |
+
"logits/rejected": -1.8834030628204346,
|
352 |
+
"logps/chosen": -125.84483337402344,
|
353 |
+
"logps/rejected": -182.78854370117188,
|
354 |
+
"loss": 0.3215,
|
355 |
+
"rewards/accuracies": 0.8999999761581421,
|
356 |
+
"rewards/chosen": 1.5408955812454224,
|
357 |
+
"rewards/margins": 1.4677695035934448,
|
358 |
+
"rewards/rejected": 0.07312598079442978,
|
359 |
+
"step": 180
|
360 |
+
},
|
361 |
+
{
|
362 |
+
"epoch": 0.73,
|
363 |
+
"eval_logits/chosen": -1.8194406032562256,
|
364 |
+
"eval_logits/rejected": -1.8561830520629883,
|
365 |
+
"eval_logps/chosen": -188.3820037841797,
|
366 |
+
"eval_logps/rejected": -243.13536071777344,
|
367 |
+
"eval_loss": 0.3756321966648102,
|
368 |
+
"eval_rewards/accuracies": 0.7857142686843872,
|
369 |
+
"eval_rewards/chosen": 1.8482797145843506,
|
370 |
+
"eval_rewards/margins": 1.8331985473632812,
|
371 |
+
"eval_rewards/rejected": 0.015081183984875679,
|
372 |
+
"eval_runtime": 26.2381,
|
373 |
+
"eval_samples_per_second": 1.067,
|
374 |
+
"eval_steps_per_second": 1.067,
|
375 |
+
"step": 180
|
376 |
+
},
|
377 |
+
{
|
378 |
+
"epoch": 0.78,
|
379 |
+
"grad_norm": 106.56973266601562,
|
380 |
+
"learning_rate": 4.512280701754386e-06,
|
381 |
+
"logits/chosen": -1.8318649530410767,
|
382 |
+
"logits/rejected": -1.8742272853851318,
|
383 |
+
"logps/chosen": -197.66619873046875,
|
384 |
+
"logps/rejected": -247.83560180664062,
|
385 |
+
"loss": 0.3303,
|
386 |
+
"rewards/accuracies": 0.800000011920929,
|
387 |
+
"rewards/chosen": 2.4889655113220215,
|
388 |
+
"rewards/margins": 2.536695957183838,
|
389 |
+
"rewards/rejected": -0.047730110585689545,
|
390 |
+
"step": 190
|
391 |
+
},
|
392 |
+
{
|
393 |
+
"epoch": 0.82,
|
394 |
+
"grad_norm": 1.9826362133026123,
|
395 |
+
"learning_rate": 4.47719298245614e-06,
|
396 |
+
"logits/chosen": -1.7579128742218018,
|
397 |
+
"logits/rejected": -1.7890180349349976,
|
398 |
+
"logps/chosen": -129.37191772460938,
|
399 |
+
"logps/rejected": -166.67636108398438,
|
400 |
+
"loss": 0.3292,
|
401 |
+
"rewards/accuracies": 0.8999999761581421,
|
402 |
+
"rewards/chosen": 2.5646984577178955,
|
403 |
+
"rewards/margins": 2.6955490112304688,
|
404 |
+
"rewards/rejected": -0.13085035979747772,
|
405 |
+
"step": 200
|
406 |
+
},
|
407 |
+
{
|
408 |
+
"epoch": 0.86,
|
409 |
+
"grad_norm": 6.777231693267822,
|
410 |
+
"learning_rate": 4.442105263157896e-06,
|
411 |
+
"logits/chosen": -1.7335002422332764,
|
412 |
+
"logits/rejected": -1.8272641897201538,
|
413 |
+
"logps/chosen": -114.22731018066406,
|
414 |
+
"logps/rejected": -215.8419647216797,
|
415 |
+
"loss": 0.0817,
|
416 |
+
"rewards/accuracies": 1.0,
|
417 |
+
"rewards/chosen": 3.278646469116211,
|
418 |
+
"rewards/margins": 4.009958267211914,
|
419 |
+
"rewards/rejected": -0.7313117980957031,
|
420 |
+
"step": 210
|
421 |
+
},
|
422 |
+
{
|
423 |
+
"epoch": 0.86,
|
424 |
+
"eval_logits/chosen": -1.826608419418335,
|
425 |
+
"eval_logits/rejected": -1.8640671968460083,
|
426 |
+
"eval_logps/chosen": -186.82992553710938,
|
427 |
+
"eval_logps/rejected": -243.80209350585938,
|
428 |
+
"eval_loss": 0.3834719657897949,
|
429 |
+
"eval_rewards/accuracies": 0.7857142686843872,
|
430 |
+
"eval_rewards/chosen": 2.3139071464538574,
|
431 |
+
"eval_rewards/margins": 2.4988515377044678,
|
432 |
+
"eval_rewards/rejected": -0.1849442571401596,
|
433 |
+
"eval_runtime": 26.2309,
|
434 |
+
"eval_samples_per_second": 1.067,
|
435 |
+
"eval_steps_per_second": 1.067,
|
436 |
+
"step": 210
|
437 |
+
},
|
438 |
+
{
|
439 |
+
"epoch": 0.9,
|
440 |
+
"grad_norm": 1.0230931043624878,
|
441 |
+
"learning_rate": 4.40701754385965e-06,
|
442 |
+
"logits/chosen": -1.7561031579971313,
|
443 |
+
"logits/rejected": -1.8601219654083252,
|
444 |
+
"logps/chosen": -134.11428833007812,
|
445 |
+
"logps/rejected": -273.5272216796875,
|
446 |
+
"loss": 0.1796,
|
447 |
+
"rewards/accuracies": 0.8999999761581421,
|
448 |
+
"rewards/chosen": 1.9340410232543945,
|
449 |
+
"rewards/margins": 3.8275578022003174,
|
450 |
+
"rewards/rejected": -1.8935168981552124,
|
451 |
+
"step": 220
|
452 |
+
},
|
453 |
+
{
|
454 |
+
"epoch": 0.94,
|
455 |
+
"grad_norm": 54.65635299682617,
|
456 |
+
"learning_rate": 4.371929824561404e-06,
|
457 |
+
"logits/chosen": -1.8512550592422485,
|
458 |
+
"logits/rejected": -1.900559663772583,
|
459 |
+
"logps/chosen": -210.9802703857422,
|
460 |
+
"logps/rejected": -240.60269165039062,
|
461 |
+
"loss": 0.5004,
|
462 |
+
"rewards/accuracies": 0.800000011920929,
|
463 |
+
"rewards/chosen": 1.5826597213745117,
|
464 |
+
"rewards/margins": 2.925461530685425,
|
465 |
+
"rewards/rejected": -1.342801809310913,
|
466 |
+
"step": 230
|
467 |
+
},
|
468 |
+
{
|
469 |
+
"epoch": 0.98,
|
470 |
+
"grad_norm": 18.797134399414062,
|
471 |
+
"learning_rate": 4.336842105263158e-06,
|
472 |
+
"logits/chosen": -1.7781873941421509,
|
473 |
+
"logits/rejected": -1.8780953884124756,
|
474 |
+
"logps/chosen": -136.18722534179688,
|
475 |
+
"logps/rejected": -257.26495361328125,
|
476 |
+
"loss": 0.137,
|
477 |
+
"rewards/accuracies": 0.8999999761581421,
|
478 |
+
"rewards/chosen": 1.9192733764648438,
|
479 |
+
"rewards/margins": 3.8388113975524902,
|
480 |
+
"rewards/rejected": -1.9195384979248047,
|
481 |
+
"step": 240
|
482 |
+
},
|
483 |
+
{
|
484 |
+
"epoch": 0.98,
|
485 |
+
"eval_logits/chosen": -1.83432936668396,
|
486 |
+
"eval_logits/rejected": -1.8722076416015625,
|
487 |
+
"eval_logps/chosen": -185.8831329345703,
|
488 |
+
"eval_logps/rejected": -244.8594207763672,
|
489 |
+
"eval_loss": 0.41323208808898926,
|
490 |
+
"eval_rewards/accuracies": 0.75,
|
491 |
+
"eval_rewards/chosen": 2.5979421138763428,
|
492 |
+
"eval_rewards/margins": 3.1000888347625732,
|
493 |
+
"eval_rewards/rejected": -0.5021467208862305,
|
494 |
+
"eval_runtime": 26.2323,
|
495 |
+
"eval_samples_per_second": 1.067,
|
496 |
+
"eval_steps_per_second": 1.067,
|
497 |
+
"step": 240
|
498 |
+
},
|
499 |
+
{
|
500 |
+
"epoch": 1.02,
|
501 |
+
"grad_norm": 132.51947021484375,
|
502 |
+
"learning_rate": 4.301754385964912e-06,
|
503 |
+
"logits/chosen": -1.8787914514541626,
|
504 |
+
"logits/rejected": -1.9520515203475952,
|
505 |
+
"logps/chosen": -186.46717834472656,
|
506 |
+
"logps/rejected": -250.008056640625,
|
507 |
+
"loss": 0.3531,
|
508 |
+
"rewards/accuracies": 0.800000011920929,
|
509 |
+
"rewards/chosen": 2.192800283432007,
|
510 |
+
"rewards/margins": 3.485269069671631,
|
511 |
+
"rewards/rejected": -1.2924686670303345,
|
512 |
+
"step": 250
|
513 |
+
},
|
514 |
+
{
|
515 |
+
"epoch": 1.06,
|
516 |
+
"grad_norm": 6.529983043670654,
|
517 |
+
"learning_rate": 4.266666666666668e-06,
|
518 |
+
"logits/chosen": -1.7029941082000732,
|
519 |
+
"logits/rejected": -1.7871618270874023,
|
520 |
+
"logps/chosen": -128.5960235595703,
|
521 |
+
"logps/rejected": -209.27603149414062,
|
522 |
+
"loss": 0.0883,
|
523 |
+
"rewards/accuracies": 0.8999999761581421,
|
524 |
+
"rewards/chosen": 2.7405991554260254,
|
525 |
+
"rewards/margins": 5.148273468017578,
|
526 |
+
"rewards/rejected": -2.4076738357543945,
|
527 |
+
"step": 260
|
528 |
+
},
|
529 |
+
{
|
530 |
+
"epoch": 1.1,
|
531 |
+
"grad_norm": 0.27644646167755127,
|
532 |
+
"learning_rate": 4.2315789473684215e-06,
|
533 |
+
"logits/chosen": -1.7499496936798096,
|
534 |
+
"logits/rejected": -1.8194109201431274,
|
535 |
+
"logps/chosen": -119.34013366699219,
|
536 |
+
"logps/rejected": -230.69229125976562,
|
537 |
+
"loss": 0.0997,
|
538 |
+
"rewards/accuracies": 0.8999999761581421,
|
539 |
+
"rewards/chosen": 4.146358013153076,
|
540 |
+
"rewards/margins": 5.72900390625,
|
541 |
+
"rewards/rejected": -1.5826464891433716,
|
542 |
+
"step": 270
|
543 |
+
},
|
544 |
+
{
|
545 |
+
"epoch": 1.1,
|
546 |
+
"eval_logits/chosen": -1.842972755432129,
|
547 |
+
"eval_logits/rejected": -1.8816190958023071,
|
548 |
+
"eval_logps/chosen": -185.4148406982422,
|
549 |
+
"eval_logps/rejected": -246.53672790527344,
|
550 |
+
"eval_loss": 0.46570485830307007,
|
551 |
+
"eval_rewards/accuracies": 0.75,
|
552 |
+
"eval_rewards/chosen": 2.7384250164031982,
|
553 |
+
"eval_rewards/margins": 3.743762969970703,
|
554 |
+
"eval_rewards/rejected": -1.0053375959396362,
|
555 |
+
"eval_runtime": 26.2348,
|
556 |
+
"eval_samples_per_second": 1.067,
|
557 |
+
"eval_steps_per_second": 1.067,
|
558 |
+
"step": 270
|
559 |
+
},
|
560 |
+
{
|
561 |
+
"epoch": 1.14,
|
562 |
+
"grad_norm": 31.16436195373535,
|
563 |
+
"learning_rate": 4.196491228070176e-06,
|
564 |
+
"logits/chosen": -1.8344509601593018,
|
565 |
+
"logits/rejected": -1.919553518295288,
|
566 |
+
"logps/chosen": -131.51156616210938,
|
567 |
+
"logps/rejected": -246.3345184326172,
|
568 |
+
"loss": 0.138,
|
569 |
+
"rewards/accuracies": 1.0,
|
570 |
+
"rewards/chosen": 2.982485771179199,
|
571 |
+
"rewards/margins": 6.261662483215332,
|
572 |
+
"rewards/rejected": -3.2791759967803955,
|
573 |
+
"step": 280
|
574 |
+
},
|
575 |
+
{
|
576 |
+
"epoch": 1.18,
|
577 |
+
"grad_norm": 0.08183232694864273,
|
578 |
+
"learning_rate": 4.16140350877193e-06,
|
579 |
+
"logits/chosen": -1.847378134727478,
|
580 |
+
"logits/rejected": -1.8774774074554443,
|
581 |
+
"logps/chosen": -127.52336120605469,
|
582 |
+
"logps/rejected": -157.3633270263672,
|
583 |
+
"loss": 0.3883,
|
584 |
+
"rewards/accuracies": 0.800000011920929,
|
585 |
+
"rewards/chosen": 0.7564169764518738,
|
586 |
+
"rewards/margins": 2.541962146759033,
|
587 |
+
"rewards/rejected": -1.7855453491210938,
|
588 |
+
"step": 290
|
589 |
+
},
|
590 |
+
{
|
591 |
+
"epoch": 1.22,
|
592 |
+
"grad_norm": 10.61989974975586,
|
593 |
+
"learning_rate": 4.126315789473685e-06,
|
594 |
+
"logits/chosen": -1.7770633697509766,
|
595 |
+
"logits/rejected": -1.8337452411651611,
|
596 |
+
"logps/chosen": -112.7281265258789,
|
597 |
+
"logps/rejected": -195.87020874023438,
|
598 |
+
"loss": 0.0432,
|
599 |
+
"rewards/accuracies": 1.0,
|
600 |
+
"rewards/chosen": 3.176725387573242,
|
601 |
+
"rewards/margins": 6.761993408203125,
|
602 |
+
"rewards/rejected": -3.5852675437927246,
|
603 |
+
"step": 300
|
604 |
+
},
|
605 |
+
{
|
606 |
+
"epoch": 1.22,
|
607 |
+
"eval_logits/chosen": -1.849483609199524,
|
608 |
+
"eval_logits/rejected": -1.888432502746582,
|
609 |
+
"eval_logps/chosen": -185.52931213378906,
|
610 |
+
"eval_logps/rejected": -248.10926818847656,
|
611 |
+
"eval_loss": 0.501145601272583,
|
612 |
+
"eval_rewards/accuracies": 0.75,
|
613 |
+
"eval_rewards/chosen": 2.704090118408203,
|
614 |
+
"eval_rewards/margins": 4.181191921234131,
|
615 |
+
"eval_rewards/rejected": -1.477101445198059,
|
616 |
+
"eval_runtime": 26.2437,
|
617 |
+
"eval_samples_per_second": 1.067,
|
618 |
+
"eval_steps_per_second": 1.067,
|
619 |
+
"step": 300
|
620 |
+
},
|
621 |
+
{
|
622 |
+
"epoch": 1.27,
|
623 |
+
"grad_norm": 327.02899169921875,
|
624 |
+
"learning_rate": 4.091228070175439e-06,
|
625 |
+
"logits/chosen": -1.8802036046981812,
|
626 |
+
"logits/rejected": -1.9435373544692993,
|
627 |
+
"logps/chosen": -164.2046661376953,
|
628 |
+
"logps/rejected": -269.22235107421875,
|
629 |
+
"loss": 0.4088,
|
630 |
+
"rewards/accuracies": 0.800000011920929,
|
631 |
+
"rewards/chosen": 3.3527073860168457,
|
632 |
+
"rewards/margins": 5.9436140060424805,
|
633 |
+
"rewards/rejected": -2.5909066200256348,
|
634 |
+
"step": 310
|
635 |
+
},
|
636 |
+
{
|
637 |
+
"epoch": 1.31,
|
638 |
+
"grad_norm": 6.565308094024658,
|
639 |
+
"learning_rate": 4.056140350877193e-06,
|
640 |
+
"logits/chosen": -1.9250032901763916,
|
641 |
+
"logits/rejected": -1.9860813617706299,
|
642 |
+
"logps/chosen": -197.4870147705078,
|
643 |
+
"logps/rejected": -207.25009155273438,
|
644 |
+
"loss": 0.4969,
|
645 |
+
"rewards/accuracies": 0.699999988079071,
|
646 |
+
"rewards/chosen": 0.32615354657173157,
|
647 |
+
"rewards/margins": 2.2872400283813477,
|
648 |
+
"rewards/rejected": -1.9610865116119385,
|
649 |
+
"step": 320
|
650 |
+
},
|
651 |
+
{
|
652 |
+
"epoch": 1.35,
|
653 |
+
"grad_norm": 33.34074020385742,
|
654 |
+
"learning_rate": 4.021052631578948e-06,
|
655 |
+
"logits/chosen": -1.8470687866210938,
|
656 |
+
"logits/rejected": -1.8675216436386108,
|
657 |
+
"logps/chosen": -117.5759506225586,
|
658 |
+
"logps/rejected": -151.83091735839844,
|
659 |
+
"loss": 0.1819,
|
660 |
+
"rewards/accuracies": 0.8999999761581421,
|
661 |
+
"rewards/chosen": 2.878262758255005,
|
662 |
+
"rewards/margins": 5.000906467437744,
|
663 |
+
"rewards/rejected": -2.12264347076416,
|
664 |
+
"step": 330
|
665 |
+
},
|
666 |
+
{
|
667 |
+
"epoch": 1.35,
|
668 |
+
"eval_logits/chosen": -1.8486573696136475,
|
669 |
+
"eval_logits/rejected": -1.8877869844436646,
|
670 |
+
"eval_logps/chosen": -185.54177856445312,
|
671 |
+
"eval_logps/rejected": -249.2687530517578,
|
672 |
+
"eval_loss": 0.478495329618454,
|
673 |
+
"eval_rewards/accuracies": 0.75,
|
674 |
+
"eval_rewards/chosen": 2.7003517150878906,
|
675 |
+
"eval_rewards/margins": 4.525294780731201,
|
676 |
+
"eval_rewards/rejected": -1.8249431848526,
|
677 |
+
"eval_runtime": 26.2362,
|
678 |
+
"eval_samples_per_second": 1.067,
|
679 |
+
"eval_steps_per_second": 1.067,
|
680 |
+
"step": 330
|
681 |
+
},
|
682 |
+
{
|
683 |
+
"epoch": 1.39,
|
684 |
+
"grad_norm": 0.6777727603912354,
|
685 |
+
"learning_rate": 3.985964912280702e-06,
|
686 |
+
"logits/chosen": -1.8584426641464233,
|
687 |
+
"logits/rejected": -1.884598970413208,
|
688 |
+
"logps/chosen": -78.69309997558594,
|
689 |
+
"logps/rejected": -126.11468505859375,
|
690 |
+
"loss": 0.3154,
|
691 |
+
"rewards/accuracies": 0.8999999761581421,
|
692 |
+
"rewards/chosen": 2.5596954822540283,
|
693 |
+
"rewards/margins": 4.089261054992676,
|
694 |
+
"rewards/rejected": -1.5295660495758057,
|
695 |
+
"step": 340
|
696 |
+
},
|
697 |
+
{
|
698 |
+
"epoch": 1.43,
|
699 |
+
"grad_norm": 45.252723693847656,
|
700 |
+
"learning_rate": 3.950877192982457e-06,
|
701 |
+
"logits/chosen": -1.8422210216522217,
|
702 |
+
"logits/rejected": -1.8741470575332642,
|
703 |
+
"logps/chosen": -198.17526245117188,
|
704 |
+
"logps/rejected": -179.26759338378906,
|
705 |
+
"loss": 0.2128,
|
706 |
+
"rewards/accuracies": 0.8999999761581421,
|
707 |
+
"rewards/chosen": 2.886838912963867,
|
708 |
+
"rewards/margins": 4.228543758392334,
|
709 |
+
"rewards/rejected": -1.3417048454284668,
|
710 |
+
"step": 350
|
711 |
+
},
|
712 |
+
{
|
713 |
+
"epoch": 1.47,
|
714 |
+
"grad_norm": 0.0007244591251946986,
|
715 |
+
"learning_rate": 3.9157894736842104e-06,
|
716 |
+
"logits/chosen": -1.8951257467269897,
|
717 |
+
"logits/rejected": -1.9684120416641235,
|
718 |
+
"logps/chosen": -123.7894515991211,
|
719 |
+
"logps/rejected": -278.1185607910156,
|
720 |
+
"loss": 0.0169,
|
721 |
+
"rewards/accuracies": 1.0,
|
722 |
+
"rewards/chosen": 3.9708969593048096,
|
723 |
+
"rewards/margins": 9.946398735046387,
|
724 |
+
"rewards/rejected": -5.975502014160156,
|
725 |
+
"step": 360
|
726 |
+
},
|
727 |
+
{
|
728 |
+
"epoch": 1.47,
|
729 |
+
"eval_logits/chosen": -1.8510137796401978,
|
730 |
+
"eval_logits/rejected": -1.8906936645507812,
|
731 |
+
"eval_logps/chosen": -185.66189575195312,
|
732 |
+
"eval_logps/rejected": -250.37806701660156,
|
733 |
+
"eval_loss": 0.4872412383556366,
|
734 |
+
"eval_rewards/accuracies": 0.75,
|
735 |
+
"eval_rewards/chosen": 2.664308547973633,
|
736 |
+
"eval_rewards/margins": 4.822035789489746,
|
737 |
+
"eval_rewards/rejected": -2.157727003097534,
|
738 |
+
"eval_runtime": 26.2354,
|
739 |
+
"eval_samples_per_second": 1.067,
|
740 |
+
"eval_steps_per_second": 1.067,
|
741 |
+
"step": 360
|
742 |
+
},
|
743 |
+
{
|
744 |
+
"epoch": 1.51,
|
745 |
+
"grad_norm": 0.001088384771719575,
|
746 |
+
"learning_rate": 3.880701754385965e-06,
|
747 |
+
"logits/chosen": -1.859140157699585,
|
748 |
+
"logits/rejected": -1.903721809387207,
|
749 |
+
"logps/chosen": -171.89047241210938,
|
750 |
+
"logps/rejected": -257.6295471191406,
|
751 |
+
"loss": 0.237,
|
752 |
+
"rewards/accuracies": 0.800000011920929,
|
753 |
+
"rewards/chosen": 2.257797956466675,
|
754 |
+
"rewards/margins": 7.423349857330322,
|
755 |
+
"rewards/rejected": -5.165551662445068,
|
756 |
+
"step": 370
|
757 |
+
},
|
758 |
+
{
|
759 |
+
"epoch": 1.55,
|
760 |
+
"grad_norm": 13.049817085266113,
|
761 |
+
"learning_rate": 3.84561403508772e-06,
|
762 |
+
"logits/chosen": -1.850102186203003,
|
763 |
+
"logits/rejected": -1.922501564025879,
|
764 |
+
"logps/chosen": -145.50692749023438,
|
765 |
+
"logps/rejected": -221.7799530029297,
|
766 |
+
"loss": 0.2906,
|
767 |
+
"rewards/accuracies": 0.8999999761581421,
|
768 |
+
"rewards/chosen": 1.9433931112289429,
|
769 |
+
"rewards/margins": 6.384174346923828,
|
770 |
+
"rewards/rejected": -4.440781593322754,
|
771 |
+
"step": 380
|
772 |
+
},
|
773 |
+
{
|
774 |
+
"epoch": 1.59,
|
775 |
+
"grad_norm": 0.006105802953243256,
|
776 |
+
"learning_rate": 3.810526315789474e-06,
|
777 |
+
"logits/chosen": -1.9277305603027344,
|
778 |
+
"logits/rejected": -1.9736566543579102,
|
779 |
+
"logps/chosen": -140.36146545410156,
|
780 |
+
"logps/rejected": -188.66836547851562,
|
781 |
+
"loss": 0.235,
|
782 |
+
"rewards/accuracies": 0.8999999761581421,
|
783 |
+
"rewards/chosen": 2.6781504154205322,
|
784 |
+
"rewards/margins": 5.825950622558594,
|
785 |
+
"rewards/rejected": -3.1477997303009033,
|
786 |
+
"step": 390
|
787 |
+
},
|
788 |
+
{
|
789 |
+
"epoch": 1.59,
|
790 |
+
"eval_logits/chosen": -1.8531960248947144,
|
791 |
+
"eval_logits/rejected": -1.8930160999298096,
|
792 |
+
"eval_logps/chosen": -185.68800354003906,
|
793 |
+
"eval_logps/rejected": -251.1302032470703,
|
794 |
+
"eval_loss": 0.48860952258110046,
|
795 |
+
"eval_rewards/accuracies": 0.75,
|
796 |
+
"eval_rewards/chosen": 2.656480550765991,
|
797 |
+
"eval_rewards/margins": 5.039853572845459,
|
798 |
+
"eval_rewards/rejected": -2.3833730220794678,
|
799 |
+
"eval_runtime": 26.2308,
|
800 |
+
"eval_samples_per_second": 1.067,
|
801 |
+
"eval_steps_per_second": 1.067,
|
802 |
+
"step": 390
|
803 |
+
},
|
804 |
+
{
|
805 |
+
"epoch": 1.63,
|
806 |
+
"grad_norm": 0.00030716226319782436,
|
807 |
+
"learning_rate": 3.7754385964912284e-06,
|
808 |
+
"logits/chosen": -1.8817369937896729,
|
809 |
+
"logits/rejected": -1.9124549627304077,
|
810 |
+
"logps/chosen": -171.40383911132812,
|
811 |
+
"logps/rejected": -198.42446899414062,
|
812 |
+
"loss": 0.4322,
|
813 |
+
"rewards/accuracies": 0.800000011920929,
|
814 |
+
"rewards/chosen": 1.219795823097229,
|
815 |
+
"rewards/margins": 6.402298927307129,
|
816 |
+
"rewards/rejected": -5.182502269744873,
|
817 |
+
"step": 400
|
818 |
+
},
|
819 |
+
{
|
820 |
+
"epoch": 1.67,
|
821 |
+
"grad_norm": 0.0007249619229696691,
|
822 |
+
"learning_rate": 3.7403508771929827e-06,
|
823 |
+
"logits/chosen": -1.8018410205841064,
|
824 |
+
"logits/rejected": -1.869511365890503,
|
825 |
+
"logps/chosen": -168.9734649658203,
|
826 |
+
"logps/rejected": -223.9024658203125,
|
827 |
+
"loss": 0.9332,
|
828 |
+
"rewards/accuracies": 0.699999988079071,
|
829 |
+
"rewards/chosen": 2.475252151489258,
|
830 |
+
"rewards/margins": 5.069212913513184,
|
831 |
+
"rewards/rejected": -2.593961238861084,
|
832 |
+
"step": 410
|
833 |
+
},
|
834 |
+
{
|
835 |
+
"epoch": 1.71,
|
836 |
+
"grad_norm": 46.79059982299805,
|
837 |
+
"learning_rate": 3.7052631578947374e-06,
|
838 |
+
"logits/chosen": -1.774857759475708,
|
839 |
+
"logits/rejected": -1.8125171661376953,
|
840 |
+
"logps/chosen": -140.12551879882812,
|
841 |
+
"logps/rejected": -208.7825927734375,
|
842 |
+
"loss": 0.7551,
|
843 |
+
"rewards/accuracies": 0.800000011920929,
|
844 |
+
"rewards/chosen": 2.244128704071045,
|
845 |
+
"rewards/margins": 6.2417426109313965,
|
846 |
+
"rewards/rejected": -3.997614622116089,
|
847 |
+
"step": 420
|
848 |
+
},
|
849 |
+
{
|
850 |
+
"epoch": 1.71,
|
851 |
+
"eval_logits/chosen": -1.8527201414108276,
|
852 |
+
"eval_logits/rejected": -1.892095923423767,
|
853 |
+
"eval_logps/chosen": -185.46646118164062,
|
854 |
+
"eval_logps/rejected": -251.00816345214844,
|
855 |
+
"eval_loss": 0.438032329082489,
|
856 |
+
"eval_rewards/accuracies": 0.75,
|
857 |
+
"eval_rewards/chosen": 2.7229368686676025,
|
858 |
+
"eval_rewards/margins": 5.069703102111816,
|
859 |
+
"eval_rewards/rejected": -2.346766233444214,
|
860 |
+
"eval_runtime": 26.2319,
|
861 |
+
"eval_samples_per_second": 1.067,
|
862 |
+
"eval_steps_per_second": 1.067,
|
863 |
+
"step": 420
|
864 |
+
},
|
865 |
+
{
|
866 |
+
"epoch": 1.76,
|
867 |
+
"grad_norm": 0.1679692417383194,
|
868 |
+
"learning_rate": 3.6701754385964917e-06,
|
869 |
+
"logits/chosen": -1.9766556024551392,
|
870 |
+
"logits/rejected": -1.967247724533081,
|
871 |
+
"logps/chosen": -181.25978088378906,
|
872 |
+
"logps/rejected": -174.0313262939453,
|
873 |
+
"loss": 0.5746,
|
874 |
+
"rewards/accuracies": 0.800000011920929,
|
875 |
+
"rewards/chosen": 2.037475347518921,
|
876 |
+
"rewards/margins": 4.661454200744629,
|
877 |
+
"rewards/rejected": -2.623978853225708,
|
878 |
+
"step": 430
|
879 |
+
},
|
880 |
+
{
|
881 |
+
"epoch": 1.8,
|
882 |
+
"grad_norm": 0.028290700167417526,
|
883 |
+
"learning_rate": 3.635087719298246e-06,
|
884 |
+
"logits/chosen": -1.7324810028076172,
|
885 |
+
"logits/rejected": -1.8381952047348022,
|
886 |
+
"logps/chosen": -118.8620376586914,
|
887 |
+
"logps/rejected": -249.5945587158203,
|
888 |
+
"loss": 0.1919,
|
889 |
+
"rewards/accuracies": 0.8999999761581421,
|
890 |
+
"rewards/chosen": 2.84242582321167,
|
891 |
+
"rewards/margins": 8.061222076416016,
|
892 |
+
"rewards/rejected": -5.218796730041504,
|
893 |
+
"step": 440
|
894 |
+
},
|
895 |
+
{
|
896 |
+
"epoch": 1.84,
|
897 |
+
"grad_norm": 3.435687303543091,
|
898 |
+
"learning_rate": 3.6000000000000003e-06,
|
899 |
+
"logits/chosen": -1.788468599319458,
|
900 |
+
"logits/rejected": -1.8786054849624634,
|
901 |
+
"logps/chosen": -127.3366928100586,
|
902 |
+
"logps/rejected": -223.4063262939453,
|
903 |
+
"loss": 0.134,
|
904 |
+
"rewards/accuracies": 1.0,
|
905 |
+
"rewards/chosen": 0.9258162379264832,
|
906 |
+
"rewards/margins": 6.352460861206055,
|
907 |
+
"rewards/rejected": -5.426644802093506,
|
908 |
+
"step": 450
|
909 |
+
},
|
910 |
+
{
|
911 |
+
"epoch": 1.84,
|
912 |
+
"eval_logits/chosen": -1.8530861139297485,
|
913 |
+
"eval_logits/rejected": -1.8925215005874634,
|
914 |
+
"eval_logps/chosen": -185.6543426513672,
|
915 |
+
"eval_logps/rejected": -251.7077178955078,
|
916 |
+
"eval_loss": 0.4383006691932678,
|
917 |
+
"eval_rewards/accuracies": 0.75,
|
918 |
+
"eval_rewards/chosen": 2.6665782928466797,
|
919 |
+
"eval_rewards/margins": 5.223209857940674,
|
920 |
+
"eval_rewards/rejected": -2.556631565093994,
|
921 |
+
"eval_runtime": 26.2389,
|
922 |
+
"eval_samples_per_second": 1.067,
|
923 |
+
"eval_steps_per_second": 1.067,
|
924 |
+
"step": 450
|
925 |
+
},
|
926 |
+
{
|
927 |
+
"epoch": 1.88,
|
928 |
+
"grad_norm": 0.2778886556625366,
|
929 |
+
"learning_rate": 3.5649122807017545e-06,
|
930 |
+
"logits/chosen": -1.8211870193481445,
|
931 |
+
"logits/rejected": -1.8562345504760742,
|
932 |
+
"logps/chosen": -90.34883117675781,
|
933 |
+
"logps/rejected": -171.23648071289062,
|
934 |
+
"loss": 0.1284,
|
935 |
+
"rewards/accuracies": 1.0,
|
936 |
+
"rewards/chosen": 3.756679058074951,
|
937 |
+
"rewards/margins": 7.474773406982422,
|
938 |
+
"rewards/rejected": -3.7180938720703125,
|
939 |
+
"step": 460
|
940 |
+
},
|
941 |
+
{
|
942 |
+
"epoch": 1.92,
|
943 |
+
"grad_norm": 0.00020280799071770161,
|
944 |
+
"learning_rate": 3.5298245614035093e-06,
|
945 |
+
"logits/chosen": -1.7977367639541626,
|
946 |
+
"logits/rejected": -1.8836923837661743,
|
947 |
+
"logps/chosen": -180.67724609375,
|
948 |
+
"logps/rejected": -215.39608764648438,
|
949 |
+
"loss": 0.5069,
|
950 |
+
"rewards/accuracies": 0.800000011920929,
|
951 |
+
"rewards/chosen": 2.3001179695129395,
|
952 |
+
"rewards/margins": 7.598956108093262,
|
953 |
+
"rewards/rejected": -5.298838138580322,
|
954 |
+
"step": 470
|
955 |
+
},
|
956 |
+
{
|
957 |
+
"epoch": 1.96,
|
958 |
+
"grad_norm": 0.0038736851420253515,
|
959 |
+
"learning_rate": 3.4947368421052635e-06,
|
960 |
+
"logits/chosen": -1.8841644525527954,
|
961 |
+
"logits/rejected": -1.986575722694397,
|
962 |
+
"logps/chosen": -189.42210388183594,
|
963 |
+
"logps/rejected": -368.7510070800781,
|
964 |
+
"loss": 0.0662,
|
965 |
+
"rewards/accuracies": 1.0,
|
966 |
+
"rewards/chosen": 2.894408702850342,
|
967 |
+
"rewards/margins": 9.953689575195312,
|
968 |
+
"rewards/rejected": -7.0592803955078125,
|
969 |
+
"step": 480
|
970 |
+
},
|
971 |
+
{
|
972 |
+
"epoch": 1.96,
|
973 |
+
"eval_logits/chosen": -1.8568991422653198,
|
974 |
+
"eval_logits/rejected": -1.896437406539917,
|
975 |
+
"eval_logps/chosen": -186.01429748535156,
|
976 |
+
"eval_logps/rejected": -252.9164276123047,
|
977 |
+
"eval_loss": 0.44481489062309265,
|
978 |
+
"eval_rewards/accuracies": 0.75,
|
979 |
+
"eval_rewards/chosen": 2.55859112739563,
|
980 |
+
"eval_rewards/margins": 5.477838039398193,
|
981 |
+
"eval_rewards/rejected": -2.9192464351654053,
|
982 |
+
"eval_runtime": 26.2387,
|
983 |
+
"eval_samples_per_second": 1.067,
|
984 |
+
"eval_steps_per_second": 1.067,
|
985 |
+
"step": 480
|
986 |
+
},
|
987 |
+
{
|
988 |
+
"epoch": 2.0,
|
989 |
+
"grad_norm": 0.00020024628611281514,
|
990 |
+
"learning_rate": 3.459649122807018e-06,
|
991 |
+
"logits/chosen": -1.7868010997772217,
|
992 |
+
"logits/rejected": -1.8775966167449951,
|
993 |
+
"logps/chosen": -168.01686096191406,
|
994 |
+
"logps/rejected": -268.81329345703125,
|
995 |
+
"loss": 0.2127,
|
996 |
+
"rewards/accuracies": 0.8999999761581421,
|
997 |
+
"rewards/chosen": 4.091753959655762,
|
998 |
+
"rewards/margins": 10.252196311950684,
|
999 |
+
"rewards/rejected": -6.160443305969238,
|
1000 |
+
"step": 490
|
1001 |
+
},
|
1002 |
+
{
|
1003 |
+
"epoch": 2.04,
|
1004 |
+
"grad_norm": 2.933464288711548,
|
1005 |
+
"learning_rate": 3.424561403508772e-06,
|
1006 |
+
"logits/chosen": -1.8647997379302979,
|
1007 |
+
"logits/rejected": -1.8972787857055664,
|
1008 |
+
"logps/chosen": -206.39712524414062,
|
1009 |
+
"logps/rejected": -255.3307647705078,
|
1010 |
+
"loss": 0.1551,
|
1011 |
+
"rewards/accuracies": 0.8999999761581421,
|
1012 |
+
"rewards/chosen": 3.2640254497528076,
|
1013 |
+
"rewards/margins": 8.8242826461792,
|
1014 |
+
"rewards/rejected": -5.560257434844971,
|
1015 |
+
"step": 500
|
1016 |
+
},
|
1017 |
+
{
|
1018 |
+
"epoch": 2.08,
|
1019 |
+
"grad_norm": 0.088715098798275,
|
1020 |
+
"learning_rate": 3.3894736842105264e-06,
|
1021 |
+
"logits/chosen": -1.8153291940689087,
|
1022 |
+
"logits/rejected": -1.907467246055603,
|
1023 |
+
"logps/chosen": -114.2393569946289,
|
1024 |
+
"logps/rejected": -245.9196319580078,
|
1025 |
+
"loss": 0.1093,
|
1026 |
+
"rewards/accuracies": 1.0,
|
1027 |
+
"rewards/chosen": 1.2050167322158813,
|
1028 |
+
"rewards/margins": 7.797829627990723,
|
1029 |
+
"rewards/rejected": -6.592813014984131,
|
1030 |
+
"step": 510
|
1031 |
+
},
|
1032 |
+
{
|
1033 |
+
"epoch": 2.08,
|
1034 |
+
"eval_logits/chosen": -1.8561475276947021,
|
1035 |
+
"eval_logits/rejected": -1.8955259323120117,
|
1036 |
+
"eval_logps/chosen": -186.13938903808594,
|
1037 |
+
"eval_logps/rejected": -253.4276580810547,
|
1038 |
+
"eval_loss": 0.4262312054634094,
|
1039 |
+
"eval_rewards/accuracies": 0.75,
|
1040 |
+
"eval_rewards/chosen": 2.521066904067993,
|
1041 |
+
"eval_rewards/margins": 5.59368371963501,
|
1042 |
+
"eval_rewards/rejected": -3.0726168155670166,
|
1043 |
+
"eval_runtime": 26.2357,
|
1044 |
+
"eval_samples_per_second": 1.067,
|
1045 |
+
"eval_steps_per_second": 1.067,
|
1046 |
+
"step": 510
|
1047 |
+
},
|
1048 |
+
{
|
1049 |
+
"epoch": 2.12,
|
1050 |
+
"grad_norm": 0.0037095127627253532,
|
1051 |
+
"learning_rate": 3.354385964912281e-06,
|
1052 |
+
"logits/chosen": -1.904370665550232,
|
1053 |
+
"logits/rejected": -1.9216587543487549,
|
1054 |
+
"logps/chosen": -173.86373901367188,
|
1055 |
+
"logps/rejected": -173.6374053955078,
|
1056 |
+
"loss": 0.1557,
|
1057 |
+
"rewards/accuracies": 0.8999999761581421,
|
1058 |
+
"rewards/chosen": 2.4744720458984375,
|
1059 |
+
"rewards/margins": 5.560853004455566,
|
1060 |
+
"rewards/rejected": -3.0863804817199707,
|
1061 |
+
"step": 520
|
1062 |
+
},
|
1063 |
+
{
|
1064 |
+
"epoch": 2.16,
|
1065 |
+
"grad_norm": 1.6992926532566344e-07,
|
1066 |
+
"learning_rate": 3.3192982456140354e-06,
|
1067 |
+
"logits/chosen": -1.7665284872055054,
|
1068 |
+
"logits/rejected": -1.859400987625122,
|
1069 |
+
"logps/chosen": -85.06950378417969,
|
1070 |
+
"logps/rejected": -232.3406524658203,
|
1071 |
+
"loss": 0.0056,
|
1072 |
+
"rewards/accuracies": 1.0,
|
1073 |
+
"rewards/chosen": 3.268359661102295,
|
1074 |
+
"rewards/margins": 9.698001861572266,
|
1075 |
+
"rewards/rejected": -6.4296417236328125,
|
1076 |
+
"step": 530
|
1077 |
+
},
|
1078 |
+
{
|
1079 |
+
"epoch": 2.2,
|
1080 |
+
"grad_norm": 0.00015174441796261817,
|
1081 |
+
"learning_rate": 3.2842105263157897e-06,
|
1082 |
+
"logits/chosen": -1.7254387140274048,
|
1083 |
+
"logits/rejected": -1.8300654888153076,
|
1084 |
+
"logps/chosen": -126.21636962890625,
|
1085 |
+
"logps/rejected": -236.91250610351562,
|
1086 |
+
"loss": 0.1557,
|
1087 |
+
"rewards/accuracies": 0.8999999761581421,
|
1088 |
+
"rewards/chosen": 2.9851126670837402,
|
1089 |
+
"rewards/margins": 9.359209060668945,
|
1090 |
+
"rewards/rejected": -6.374096870422363,
|
1091 |
+
"step": 540
|
1092 |
+
},
|
1093 |
+
{
|
1094 |
+
"epoch": 2.2,
|
1095 |
+
"eval_logits/chosen": -1.85663902759552,
|
1096 |
+
"eval_logits/rejected": -1.8965102434158325,
|
1097 |
+
"eval_logps/chosen": -186.6448974609375,
|
1098 |
+
"eval_logps/rejected": -254.58482360839844,
|
1099 |
+
"eval_loss": 0.4263899326324463,
|
1100 |
+
"eval_rewards/accuracies": 0.75,
|
1101 |
+
"eval_rewards/chosen": 2.369408369064331,
|
1102 |
+
"eval_rewards/margins": 5.789170265197754,
|
1103 |
+
"eval_rewards/rejected": -3.4197616577148438,
|
1104 |
+
"eval_runtime": 26.2541,
|
1105 |
+
"eval_samples_per_second": 1.066,
|
1106 |
+
"eval_steps_per_second": 1.066,
|
1107 |
+
"step": 540
|
1108 |
+
},
|
1109 |
+
{
|
1110 |
+
"epoch": 2.24,
|
1111 |
+
"grad_norm": 7.969775676727295,
|
1112 |
+
"learning_rate": 3.249122807017544e-06,
|
1113 |
+
"logits/chosen": -1.8786284923553467,
|
1114 |
+
"logits/rejected": -1.9631255865097046,
|
1115 |
+
"logps/chosen": -186.79092407226562,
|
1116 |
+
"logps/rejected": -295.82598876953125,
|
1117 |
+
"loss": 0.0625,
|
1118 |
+
"rewards/accuracies": 1.0,
|
1119 |
+
"rewards/chosen": 0.5664806365966797,
|
1120 |
+
"rewards/margins": 6.786992073059082,
|
1121 |
+
"rewards/rejected": -6.220510959625244,
|
1122 |
+
"step": 550
|
1123 |
+
},
|
1124 |
+
{
|
1125 |
+
"epoch": 2.29,
|
1126 |
+
"grad_norm": 41.887001037597656,
|
1127 |
+
"learning_rate": 3.2140350877192982e-06,
|
1128 |
+
"logits/chosen": -1.8324702978134155,
|
1129 |
+
"logits/rejected": -1.8570420742034912,
|
1130 |
+
"logps/chosen": -198.67893981933594,
|
1131 |
+
"logps/rejected": -203.76016235351562,
|
1132 |
+
"loss": 0.389,
|
1133 |
+
"rewards/accuracies": 0.8999999761581421,
|
1134 |
+
"rewards/chosen": 2.470458984375,
|
1135 |
+
"rewards/margins": 8.410517692565918,
|
1136 |
+
"rewards/rejected": -5.940059185028076,
|
1137 |
+
"step": 560
|
1138 |
+
},
|
1139 |
+
{
|
1140 |
+
"epoch": 2.33,
|
1141 |
+
"grad_norm": 0.022604621946811676,
|
1142 |
+
"learning_rate": 3.178947368421053e-06,
|
1143 |
+
"logits/chosen": -1.7625017166137695,
|
1144 |
+
"logits/rejected": -1.823395013809204,
|
1145 |
+
"logps/chosen": -145.4344024658203,
|
1146 |
+
"logps/rejected": -202.81698608398438,
|
1147 |
+
"loss": 0.0962,
|
1148 |
+
"rewards/accuracies": 0.8999999761581421,
|
1149 |
+
"rewards/chosen": 3.2409331798553467,
|
1150 |
+
"rewards/margins": 7.634140968322754,
|
1151 |
+
"rewards/rejected": -4.39320707321167,
|
1152 |
+
"step": 570
|
1153 |
+
},
|
1154 |
+
{
|
1155 |
+
"epoch": 2.33,
|
1156 |
+
"eval_logits/chosen": -1.8582419157028198,
|
1157 |
+
"eval_logits/rejected": -1.897825837135315,
|
1158 |
+
"eval_logps/chosen": -186.9964141845703,
|
1159 |
+
"eval_logps/rejected": -255.54441833496094,
|
1160 |
+
"eval_loss": 0.41822776198387146,
|
1161 |
+
"eval_rewards/accuracies": 0.75,
|
1162 |
+
"eval_rewards/chosen": 2.2639553546905518,
|
1163 |
+
"eval_rewards/margins": 5.971593379974365,
|
1164 |
+
"eval_rewards/rejected": -3.70763897895813,
|
1165 |
+
"eval_runtime": 26.2352,
|
1166 |
+
"eval_samples_per_second": 1.067,
|
1167 |
+
"eval_steps_per_second": 1.067,
|
1168 |
+
"step": 570
|
1169 |
+
},
|
1170 |
+
{
|
1171 |
+
"epoch": 2.37,
|
1172 |
+
"grad_norm": 0.0053796968422830105,
|
1173 |
+
"learning_rate": 3.1438596491228072e-06,
|
1174 |
+
"logits/chosen": -1.8564494848251343,
|
1175 |
+
"logits/rejected": -1.9009406566619873,
|
1176 |
+
"logps/chosen": -153.97903442382812,
|
1177 |
+
"logps/rejected": -182.94284057617188,
|
1178 |
+
"loss": 0.5733,
|
1179 |
+
"rewards/accuracies": 0.800000011920929,
|
1180 |
+
"rewards/chosen": 2.0778138637542725,
|
1181 |
+
"rewards/margins": 5.466890335083008,
|
1182 |
+
"rewards/rejected": -3.389075756072998,
|
1183 |
+
"step": 580
|
1184 |
+
},
|
1185 |
+
{
|
1186 |
+
"epoch": 2.41,
|
1187 |
+
"grad_norm": 5.523962020874023,
|
1188 |
+
"learning_rate": 3.1087719298245615e-06,
|
1189 |
+
"logits/chosen": -1.8511816263198853,
|
1190 |
+
"logits/rejected": -1.853109359741211,
|
1191 |
+
"logps/chosen": -182.23892211914062,
|
1192 |
+
"logps/rejected": -120.5928955078125,
|
1193 |
+
"loss": 0.4958,
|
1194 |
+
"rewards/accuracies": 0.800000011920929,
|
1195 |
+
"rewards/chosen": 3.659731388092041,
|
1196 |
+
"rewards/margins": 4.932890892028809,
|
1197 |
+
"rewards/rejected": -1.2731596231460571,
|
1198 |
+
"step": 590
|
1199 |
+
},
|
1200 |
+
{
|
1201 |
+
"epoch": 2.45,
|
1202 |
+
"grad_norm": 0.0013734139502048492,
|
1203 |
+
"learning_rate": 3.0736842105263158e-06,
|
1204 |
+
"logits/chosen": -1.7618497610092163,
|
1205 |
+
"logits/rejected": -1.8786863088607788,
|
1206 |
+
"logps/chosen": -114.43098449707031,
|
1207 |
+
"logps/rejected": -229.5573272705078,
|
1208 |
+
"loss": 0.0437,
|
1209 |
+
"rewards/accuracies": 1.0,
|
1210 |
+
"rewards/chosen": 2.8845677375793457,
|
1211 |
+
"rewards/margins": 9.519918441772461,
|
1212 |
+
"rewards/rejected": -6.635350227355957,
|
1213 |
+
"step": 600
|
1214 |
+
},
|
1215 |
+
{
|
1216 |
+
"epoch": 2.45,
|
1217 |
+
"eval_logits/chosen": -1.8533600568771362,
|
1218 |
+
"eval_logits/rejected": -1.8932533264160156,
|
1219 |
+
"eval_logps/chosen": -187.0037384033203,
|
1220 |
+
"eval_logps/rejected": -255.77125549316406,
|
1221 |
+
"eval_loss": 0.3824370801448822,
|
1222 |
+
"eval_rewards/accuracies": 0.75,
|
1223 |
+
"eval_rewards/chosen": 2.2617576122283936,
|
1224 |
+
"eval_rewards/margins": 6.037452220916748,
|
1225 |
+
"eval_rewards/rejected": -3.7756943702697754,
|
1226 |
+
"eval_runtime": 26.2326,
|
1227 |
+
"eval_samples_per_second": 1.067,
|
1228 |
+
"eval_steps_per_second": 1.067,
|
1229 |
+
"step": 600
|
1230 |
+
},
|
1231 |
+
{
|
1232 |
+
"epoch": 2.49,
|
1233 |
+
"grad_norm": 3.847404956817627,
|
1234 |
+
"learning_rate": 3.03859649122807e-06,
|
1235 |
+
"logits/chosen": -1.9108350276947021,
|
1236 |
+
"logits/rejected": -1.9136813879013062,
|
1237 |
+
"logps/chosen": -147.76666259765625,
|
1238 |
+
"logps/rejected": -151.54275512695312,
|
1239 |
+
"loss": 0.1939,
|
1240 |
+
"rewards/accuracies": 0.8999999761581421,
|
1241 |
+
"rewards/chosen": 2.7343037128448486,
|
1242 |
+
"rewards/margins": 7.082615852355957,
|
1243 |
+
"rewards/rejected": -4.348311424255371,
|
1244 |
+
"step": 610
|
1245 |
+
},
|
1246 |
+
{
|
1247 |
+
"epoch": 2.53,
|
1248 |
+
"grad_norm": 6.025547172328061e-09,
|
1249 |
+
"learning_rate": 3.003508771929825e-06,
|
1250 |
+
"logits/chosen": -1.9030263423919678,
|
1251 |
+
"logits/rejected": -1.964647889137268,
|
1252 |
+
"logps/chosen": -157.2318115234375,
|
1253 |
+
"logps/rejected": -335.1673583984375,
|
1254 |
+
"loss": 0.1472,
|
1255 |
+
"rewards/accuracies": 0.8999999761581421,
|
1256 |
+
"rewards/chosen": 0.17058677971363068,
|
1257 |
+
"rewards/margins": 10.464014053344727,
|
1258 |
+
"rewards/rejected": -10.293428421020508,
|
1259 |
+
"step": 620
|
1260 |
+
},
|
1261 |
+
{
|
1262 |
+
"epoch": 2.57,
|
1263 |
+
"grad_norm": 0.7363545298576355,
|
1264 |
+
"learning_rate": 2.9684210526315795e-06,
|
1265 |
+
"logits/chosen": -1.8167072534561157,
|
1266 |
+
"logits/rejected": -1.9096784591674805,
|
1267 |
+
"logps/chosen": -128.2840118408203,
|
1268 |
+
"logps/rejected": -278.52862548828125,
|
1269 |
+
"loss": 0.0278,
|
1270 |
+
"rewards/accuracies": 1.0,
|
1271 |
+
"rewards/chosen": 2.6698997020721436,
|
1272 |
+
"rewards/margins": 10.468381881713867,
|
1273 |
+
"rewards/rejected": -7.798482418060303,
|
1274 |
+
"step": 630
|
1275 |
+
},
|
1276 |
+
{
|
1277 |
+
"epoch": 2.57,
|
1278 |
+
"eval_logits/chosen": -1.8536275625228882,
|
1279 |
+
"eval_logits/rejected": -1.8932437896728516,
|
1280 |
+
"eval_logps/chosen": -186.7086181640625,
|
1281 |
+
"eval_logps/rejected": -255.70458984375,
|
1282 |
+
"eval_loss": 0.3570733964443207,
|
1283 |
+
"eval_rewards/accuracies": 0.8571428656578064,
|
1284 |
+
"eval_rewards/chosen": 2.3502955436706543,
|
1285 |
+
"eval_rewards/margins": 6.105983257293701,
|
1286 |
+
"eval_rewards/rejected": -3.755687713623047,
|
1287 |
+
"eval_runtime": 26.2519,
|
1288 |
+
"eval_samples_per_second": 1.067,
|
1289 |
+
"eval_steps_per_second": 1.067,
|
1290 |
+
"step": 630
|
1291 |
+
},
|
1292 |
+
{
|
1293 |
+
"epoch": 2.61,
|
1294 |
+
"grad_norm": 19.793140411376953,
|
1295 |
+
"learning_rate": 2.9333333333333338e-06,
|
1296 |
+
"logits/chosen": -1.9250261783599854,
|
1297 |
+
"logits/rejected": -1.969919204711914,
|
1298 |
+
"logps/chosen": -169.236572265625,
|
1299 |
+
"logps/rejected": -216.4037322998047,
|
1300 |
+
"loss": 0.3144,
|
1301 |
+
"rewards/accuracies": 0.800000011920929,
|
1302 |
+
"rewards/chosen": 1.5923322439193726,
|
1303 |
+
"rewards/margins": 8.081656455993652,
|
1304 |
+
"rewards/rejected": -6.48932409286499,
|
1305 |
+
"step": 640
|
1306 |
+
},
|
1307 |
+
{
|
1308 |
+
"epoch": 2.65,
|
1309 |
+
"grad_norm": 0.0,
|
1310 |
+
"learning_rate": 2.898245614035088e-06,
|
1311 |
+
"logits/chosen": -1.840488076210022,
|
1312 |
+
"logits/rejected": -1.9099540710449219,
|
1313 |
+
"logps/chosen": -188.36514282226562,
|
1314 |
+
"logps/rejected": -266.2427978515625,
|
1315 |
+
"loss": 0.0455,
|
1316 |
+
"rewards/accuracies": 1.0,
|
1317 |
+
"rewards/chosen": 3.294522762298584,
|
1318 |
+
"rewards/margins": 9.861984252929688,
|
1319 |
+
"rewards/rejected": -6.5674614906311035,
|
1320 |
+
"step": 650
|
1321 |
+
},
|
1322 |
+
{
|
1323 |
+
"epoch": 2.69,
|
1324 |
+
"grad_norm": 350.0583801269531,
|
1325 |
+
"learning_rate": 2.8631578947368423e-06,
|
1326 |
+
"logits/chosen": -1.875605583190918,
|
1327 |
+
"logits/rejected": -1.9353086948394775,
|
1328 |
+
"logps/chosen": -128.51104736328125,
|
1329 |
+
"logps/rejected": -191.3477325439453,
|
1330 |
+
"loss": 0.2399,
|
1331 |
+
"rewards/accuracies": 0.8999999761581421,
|
1332 |
+
"rewards/chosen": 3.1921534538269043,
|
1333 |
+
"rewards/margins": 7.449667453765869,
|
1334 |
+
"rewards/rejected": -4.257513999938965,
|
1335 |
+
"step": 660
|
1336 |
+
},
|
1337 |
+
{
|
1338 |
+
"epoch": 2.69,
|
1339 |
+
"eval_logits/chosen": -1.851164698600769,
|
1340 |
+
"eval_logits/rejected": -1.8909271955490112,
|
1341 |
+
"eval_logps/chosen": -186.86781311035156,
|
1342 |
+
"eval_logps/rejected": -256.2709655761719,
|
1343 |
+
"eval_loss": 0.3312711715698242,
|
1344 |
+
"eval_rewards/accuracies": 0.8571428656578064,
|
1345 |
+
"eval_rewards/chosen": 2.3025403022766113,
|
1346 |
+
"eval_rewards/margins": 6.228146553039551,
|
1347 |
+
"eval_rewards/rejected": -3.9256062507629395,
|
1348 |
+
"eval_runtime": 26.2337,
|
1349 |
+
"eval_samples_per_second": 1.067,
|
1350 |
+
"eval_steps_per_second": 1.067,
|
1351 |
+
"step": 660
|
1352 |
+
},
|
1353 |
+
{
|
1354 |
+
"epoch": 2.73,
|
1355 |
+
"grad_norm": 0.0,
|
1356 |
+
"learning_rate": 2.828070175438597e-06,
|
1357 |
+
"logits/chosen": -1.7718509435653687,
|
1358 |
+
"logits/rejected": -1.8662782907485962,
|
1359 |
+
"logps/chosen": -100.0200424194336,
|
1360 |
+
"logps/rejected": -176.11337280273438,
|
1361 |
+
"loss": 0.0636,
|
1362 |
+
"rewards/accuracies": 1.0,
|
1363 |
+
"rewards/chosen": 2.2194771766662598,
|
1364 |
+
"rewards/margins": 7.908571720123291,
|
1365 |
+
"rewards/rejected": -5.689094543457031,
|
1366 |
+
"step": 670
|
1367 |
+
},
|
1368 |
+
{
|
1369 |
+
"epoch": 2.78,
|
1370 |
+
"grad_norm": 9.849269554251805e-05,
|
1371 |
+
"learning_rate": 2.7929824561403513e-06,
|
1372 |
+
"logits/chosen": -1.8379380702972412,
|
1373 |
+
"logits/rejected": -1.9032844305038452,
|
1374 |
+
"logps/chosen": -147.157958984375,
|
1375 |
+
"logps/rejected": -311.342529296875,
|
1376 |
+
"loss": 0.0081,
|
1377 |
+
"rewards/accuracies": 1.0,
|
1378 |
+
"rewards/chosen": 1.7090831995010376,
|
1379 |
+
"rewards/margins": 9.63591480255127,
|
1380 |
+
"rewards/rejected": -7.926831245422363,
|
1381 |
+
"step": 680
|
1382 |
+
},
|
1383 |
+
{
|
1384 |
+
"epoch": 2.82,
|
1385 |
+
"grad_norm": 5.133792877197266,
|
1386 |
+
"learning_rate": 2.7578947368421056e-06,
|
1387 |
+
"logits/chosen": -1.8350670337677002,
|
1388 |
+
"logits/rejected": -1.9039970636367798,
|
1389 |
+
"logps/chosen": -160.75314331054688,
|
1390 |
+
"logps/rejected": -247.68008422851562,
|
1391 |
+
"loss": 0.039,
|
1392 |
+
"rewards/accuracies": 1.0,
|
1393 |
+
"rewards/chosen": 4.146931171417236,
|
1394 |
+
"rewards/margins": 11.529654502868652,
|
1395 |
+
"rewards/rejected": -7.382723808288574,
|
1396 |
+
"step": 690
|
1397 |
+
},
|
1398 |
+
{
|
1399 |
+
"epoch": 2.82,
|
1400 |
+
"eval_logits/chosen": -1.8510338068008423,
|
1401 |
+
"eval_logits/rejected": -1.8906004428863525,
|
1402 |
+
"eval_logps/chosen": -187.16348266601562,
|
1403 |
+
"eval_logps/rejected": -257.069091796875,
|
1404 |
+
"eval_loss": 0.3131271302700043,
|
1405 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1406 |
+
"eval_rewards/chosen": 2.213836193084717,
|
1407 |
+
"eval_rewards/margins": 6.378880977630615,
|
1408 |
+
"eval_rewards/rejected": -4.165044784545898,
|
1409 |
+
"eval_runtime": 26.2333,
|
1410 |
+
"eval_samples_per_second": 1.067,
|
1411 |
+
"eval_steps_per_second": 1.067,
|
1412 |
+
"step": 690
|
1413 |
+
},
|
1414 |
+
{
|
1415 |
+
"epoch": 2.86,
|
1416 |
+
"grad_norm": 0.001812128466553986,
|
1417 |
+
"learning_rate": 2.72280701754386e-06,
|
1418 |
+
"logits/chosen": -1.888392448425293,
|
1419 |
+
"logits/rejected": -1.9376449584960938,
|
1420 |
+
"logps/chosen": -123.4749755859375,
|
1421 |
+
"logps/rejected": -221.9653778076172,
|
1422 |
+
"loss": 0.1581,
|
1423 |
+
"rewards/accuracies": 0.8999999761581421,
|
1424 |
+
"rewards/chosen": 3.1250205039978027,
|
1425 |
+
"rewards/margins": 9.35824966430664,
|
1426 |
+
"rewards/rejected": -6.23322868347168,
|
1427 |
+
"step": 700
|
1428 |
+
},
|
1429 |
+
{
|
1430 |
+
"epoch": 2.9,
|
1431 |
+
"grad_norm": 0.0007696045213378966,
|
1432 |
+
"learning_rate": 2.687719298245614e-06,
|
1433 |
+
"logits/chosen": -1.7869523763656616,
|
1434 |
+
"logits/rejected": -1.9070037603378296,
|
1435 |
+
"logps/chosen": -116.8504409790039,
|
1436 |
+
"logps/rejected": -300.51806640625,
|
1437 |
+
"loss": 0.1832,
|
1438 |
+
"rewards/accuracies": 0.8999999761581421,
|
1439 |
+
"rewards/chosen": 2.3490161895751953,
|
1440 |
+
"rewards/margins": 10.493500709533691,
|
1441 |
+
"rewards/rejected": -8.144485473632812,
|
1442 |
+
"step": 710
|
1443 |
+
},
|
1444 |
+
{
|
1445 |
+
"epoch": 2.94,
|
1446 |
+
"grad_norm": 83.29313659667969,
|
1447 |
+
"learning_rate": 2.652631578947369e-06,
|
1448 |
+
"logits/chosen": -1.8838169574737549,
|
1449 |
+
"logits/rejected": -1.9118610620498657,
|
1450 |
+
"logps/chosen": -104.75813293457031,
|
1451 |
+
"logps/rejected": -167.18580627441406,
|
1452 |
+
"loss": 0.3389,
|
1453 |
+
"rewards/accuracies": 0.8999999761581421,
|
1454 |
+
"rewards/chosen": 2.7628464698791504,
|
1455 |
+
"rewards/margins": 6.317173957824707,
|
1456 |
+
"rewards/rejected": -3.5543277263641357,
|
1457 |
+
"step": 720
|
1458 |
+
},
|
1459 |
+
{
|
1460 |
+
"epoch": 2.94,
|
1461 |
+
"eval_logits/chosen": -1.8480366468429565,
|
1462 |
+
"eval_logits/rejected": -1.8872734308242798,
|
1463 |
+
"eval_logps/chosen": -187.00791931152344,
|
1464 |
+
"eval_logps/rejected": -257.2390441894531,
|
1465 |
+
"eval_loss": 0.27630648016929626,
|
1466 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1467 |
+
"eval_rewards/chosen": 2.2605104446411133,
|
1468 |
+
"eval_rewards/margins": 6.476548194885254,
|
1469 |
+
"eval_rewards/rejected": -4.216037750244141,
|
1470 |
+
"eval_runtime": 26.23,
|
1471 |
+
"eval_samples_per_second": 1.067,
|
1472 |
+
"eval_steps_per_second": 1.067,
|
1473 |
+
"step": 720
|
1474 |
+
},
|
1475 |
+
{
|
1476 |
+
"epoch": 2.98,
|
1477 |
+
"grad_norm": 0.030682619661092758,
|
1478 |
+
"learning_rate": 2.617543859649123e-06,
|
1479 |
+
"logits/chosen": -1.8667634725570679,
|
1480 |
+
"logits/rejected": -1.8890842199325562,
|
1481 |
+
"logps/chosen": -197.29690551757812,
|
1482 |
+
"logps/rejected": -219.12332153320312,
|
1483 |
+
"loss": 0.617,
|
1484 |
+
"rewards/accuracies": 0.800000011920929,
|
1485 |
+
"rewards/chosen": -0.587865948677063,
|
1486 |
+
"rewards/margins": 5.496161937713623,
|
1487 |
+
"rewards/rejected": -6.0840277671813965,
|
1488 |
+
"step": 730
|
1489 |
+
},
|
1490 |
+
{
|
1491 |
+
"epoch": 3.02,
|
1492 |
+
"grad_norm": 0.005717798601835966,
|
1493 |
+
"learning_rate": 2.5824561403508774e-06,
|
1494 |
+
"logits/chosen": -1.8484325408935547,
|
1495 |
+
"logits/rejected": -1.8692123889923096,
|
1496 |
+
"logps/chosen": -102.5673599243164,
|
1497 |
+
"logps/rejected": -141.8745880126953,
|
1498 |
+
"loss": 0.044,
|
1499 |
+
"rewards/accuracies": 1.0,
|
1500 |
+
"rewards/chosen": 2.209364175796509,
|
1501 |
+
"rewards/margins": 8.334315299987793,
|
1502 |
+
"rewards/rejected": -6.124951362609863,
|
1503 |
+
"step": 740
|
1504 |
+
},
|
1505 |
+
{
|
1506 |
+
"epoch": 3.06,
|
1507 |
+
"grad_norm": 0.0006890746881254017,
|
1508 |
+
"learning_rate": 2.5473684210526317e-06,
|
1509 |
+
"logits/chosen": -1.865378737449646,
|
1510 |
+
"logits/rejected": -1.9292535781860352,
|
1511 |
+
"logps/chosen": -103.72840881347656,
|
1512 |
+
"logps/rejected": -203.67276000976562,
|
1513 |
+
"loss": 0.0154,
|
1514 |
+
"rewards/accuracies": 1.0,
|
1515 |
+
"rewards/chosen": 3.9861044883728027,
|
1516 |
+
"rewards/margins": 8.630514144897461,
|
1517 |
+
"rewards/rejected": -4.644408702850342,
|
1518 |
+
"step": 750
|
1519 |
+
},
|
1520 |
+
{
|
1521 |
+
"epoch": 3.06,
|
1522 |
+
"eval_logits/chosen": -1.846961259841919,
|
1523 |
+
"eval_logits/rejected": -1.886168122291565,
|
1524 |
+
"eval_logps/chosen": -187.0342254638672,
|
1525 |
+
"eval_logps/rejected": -257.52471923828125,
|
1526 |
+
"eval_loss": 0.2703961730003357,
|
1527 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1528 |
+
"eval_rewards/chosen": 2.252613067626953,
|
1529 |
+
"eval_rewards/margins": 6.554351806640625,
|
1530 |
+
"eval_rewards/rejected": -4.301738739013672,
|
1531 |
+
"eval_runtime": 26.236,
|
1532 |
+
"eval_samples_per_second": 1.067,
|
1533 |
+
"eval_steps_per_second": 1.067,
|
1534 |
+
"step": 750
|
1535 |
+
},
|
1536 |
+
{
|
1537 |
+
"epoch": 3.1,
|
1538 |
+
"grad_norm": 0.0,
|
1539 |
+
"learning_rate": 2.512280701754386e-06,
|
1540 |
+
"logits/chosen": -1.921618103981018,
|
1541 |
+
"logits/rejected": -1.9239141941070557,
|
1542 |
+
"logps/chosen": -248.826416015625,
|
1543 |
+
"logps/rejected": -182.67323303222656,
|
1544 |
+
"loss": 0.5385,
|
1545 |
+
"rewards/accuracies": 0.699999988079071,
|
1546 |
+
"rewards/chosen": -0.7813287377357483,
|
1547 |
+
"rewards/margins": 4.614243507385254,
|
1548 |
+
"rewards/rejected": -5.395572185516357,
|
1549 |
+
"step": 760
|
1550 |
+
},
|
1551 |
+
{
|
1552 |
+
"epoch": 3.14,
|
1553 |
+
"grad_norm": 20.82038688659668,
|
1554 |
+
"learning_rate": 2.4771929824561407e-06,
|
1555 |
+
"logits/chosen": -1.8723421096801758,
|
1556 |
+
"logits/rejected": -1.9096221923828125,
|
1557 |
+
"logps/chosen": -144.3488311767578,
|
1558 |
+
"logps/rejected": -200.52149963378906,
|
1559 |
+
"loss": 0.0473,
|
1560 |
+
"rewards/accuracies": 1.0,
|
1561 |
+
"rewards/chosen": 0.6265076994895935,
|
1562 |
+
"rewards/margins": 5.623400688171387,
|
1563 |
+
"rewards/rejected": -4.996891975402832,
|
1564 |
+
"step": 770
|
1565 |
+
},
|
1566 |
+
{
|
1567 |
+
"epoch": 3.18,
|
1568 |
+
"grad_norm": 0.0,
|
1569 |
+
"learning_rate": 2.442105263157895e-06,
|
1570 |
+
"logits/chosen": -1.967357873916626,
|
1571 |
+
"logits/rejected": -1.995941162109375,
|
1572 |
+
"logps/chosen": -169.40643310546875,
|
1573 |
+
"logps/rejected": -225.83267211914062,
|
1574 |
+
"loss": 0.021,
|
1575 |
+
"rewards/accuracies": 1.0,
|
1576 |
+
"rewards/chosen": 3.36498761177063,
|
1577 |
+
"rewards/margins": 8.846434593200684,
|
1578 |
+
"rewards/rejected": -5.481447219848633,
|
1579 |
+
"step": 780
|
1580 |
+
},
|
1581 |
+
{
|
1582 |
+
"epoch": 3.18,
|
1583 |
+
"eval_logits/chosen": -1.84483802318573,
|
1584 |
+
"eval_logits/rejected": -1.8837740421295166,
|
1585 |
+
"eval_logps/chosen": -187.02699279785156,
|
1586 |
+
"eval_logps/rejected": -257.6650085449219,
|
1587 |
+
"eval_loss": 0.24223719537258148,
|
1588 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1589 |
+
"eval_rewards/chosen": 2.254775285720825,
|
1590 |
+
"eval_rewards/margins": 6.598587989807129,
|
1591 |
+
"eval_rewards/rejected": -4.343813419342041,
|
1592 |
+
"eval_runtime": 26.2524,
|
1593 |
+
"eval_samples_per_second": 1.067,
|
1594 |
+
"eval_steps_per_second": 1.067,
|
1595 |
+
"step": 780
|
1596 |
+
},
|
1597 |
+
{
|
1598 |
+
"epoch": 3.22,
|
1599 |
+
"grad_norm": 0.3396748900413513,
|
1600 |
+
"learning_rate": 2.4070175438596493e-06,
|
1601 |
+
"logits/chosen": -1.795336127281189,
|
1602 |
+
"logits/rejected": -1.9104626178741455,
|
1603 |
+
"logps/chosen": -127.96372985839844,
|
1604 |
+
"logps/rejected": -321.76751708984375,
|
1605 |
+
"loss": 0.0057,
|
1606 |
+
"rewards/accuracies": 1.0,
|
1607 |
+
"rewards/chosen": 1.3614065647125244,
|
1608 |
+
"rewards/margins": 11.285585403442383,
|
1609 |
+
"rewards/rejected": -9.924179077148438,
|
1610 |
+
"step": 790
|
1611 |
+
},
|
1612 |
+
{
|
1613 |
+
"epoch": 3.27,
|
1614 |
+
"grad_norm": 89.25263214111328,
|
1615 |
+
"learning_rate": 2.3719298245614036e-06,
|
1616 |
+
"logits/chosen": -1.7844104766845703,
|
1617 |
+
"logits/rejected": -1.835338830947876,
|
1618 |
+
"logps/chosen": -138.7672119140625,
|
1619 |
+
"logps/rejected": -212.5260772705078,
|
1620 |
+
"loss": 0.099,
|
1621 |
+
"rewards/accuracies": 0.8999999761581421,
|
1622 |
+
"rewards/chosen": 3.200477123260498,
|
1623 |
+
"rewards/margins": 9.856096267700195,
|
1624 |
+
"rewards/rejected": -6.655619144439697,
|
1625 |
+
"step": 800
|
1626 |
+
},
|
1627 |
+
{
|
1628 |
+
"epoch": 3.31,
|
1629 |
+
"grad_norm": 0.0014283270575106144,
|
1630 |
+
"learning_rate": 2.3368421052631583e-06,
|
1631 |
+
"logits/chosen": -1.8158670663833618,
|
1632 |
+
"logits/rejected": -1.896251916885376,
|
1633 |
+
"logps/chosen": -144.15419006347656,
|
1634 |
+
"logps/rejected": -283.3548583984375,
|
1635 |
+
"loss": 0.0614,
|
1636 |
+
"rewards/accuracies": 1.0,
|
1637 |
+
"rewards/chosen": 3.173565149307251,
|
1638 |
+
"rewards/margins": 12.303560256958008,
|
1639 |
+
"rewards/rejected": -9.129995346069336,
|
1640 |
+
"step": 810
|
1641 |
+
},
|
1642 |
+
{
|
1643 |
+
"epoch": 3.31,
|
1644 |
+
"eval_logits/chosen": -1.8417214155197144,
|
1645 |
+
"eval_logits/rejected": -1.8804857730865479,
|
1646 |
+
"eval_logps/chosen": -187.0991668701172,
|
1647 |
+
"eval_logps/rejected": -258.0171813964844,
|
1648 |
+
"eval_loss": 0.2144124060869217,
|
1649 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1650 |
+
"eval_rewards/chosen": 2.2331290245056152,
|
1651 |
+
"eval_rewards/margins": 6.682602882385254,
|
1652 |
+
"eval_rewards/rejected": -4.449473857879639,
|
1653 |
+
"eval_runtime": 26.2397,
|
1654 |
+
"eval_samples_per_second": 1.067,
|
1655 |
+
"eval_steps_per_second": 1.067,
|
1656 |
+
"step": 810
|
1657 |
+
},
|
1658 |
+
{
|
1659 |
+
"epoch": 3.35,
|
1660 |
+
"grad_norm": 8.28421070764307e-06,
|
1661 |
+
"learning_rate": 2.3017543859649126e-06,
|
1662 |
+
"logits/chosen": -1.7904294729232788,
|
1663 |
+
"logits/rejected": -1.7954041957855225,
|
1664 |
+
"logps/chosen": -136.8679962158203,
|
1665 |
+
"logps/rejected": -173.52316284179688,
|
1666 |
+
"loss": 0.0305,
|
1667 |
+
"rewards/accuracies": 1.0,
|
1668 |
+
"rewards/chosen": 3.9618313312530518,
|
1669 |
+
"rewards/margins": 9.469245910644531,
|
1670 |
+
"rewards/rejected": -5.507414817810059,
|
1671 |
+
"step": 820
|
1672 |
+
},
|
1673 |
+
{
|
1674 |
+
"epoch": 3.39,
|
1675 |
+
"grad_norm": 0.00032607687171548605,
|
1676 |
+
"learning_rate": 2.266666666666667e-06,
|
1677 |
+
"logits/chosen": -1.7678247690200806,
|
1678 |
+
"logits/rejected": -1.8045036792755127,
|
1679 |
+
"logps/chosen": -97.9904556274414,
|
1680 |
+
"logps/rejected": -158.86538696289062,
|
1681 |
+
"loss": 0.0114,
|
1682 |
+
"rewards/accuracies": 1.0,
|
1683 |
+
"rewards/chosen": 3.4605960845947266,
|
1684 |
+
"rewards/margins": 9.087323188781738,
|
1685 |
+
"rewards/rejected": -5.626727104187012,
|
1686 |
+
"step": 830
|
1687 |
+
},
|
1688 |
+
{
|
1689 |
+
"epoch": 3.43,
|
1690 |
+
"grad_norm": 9.237614631652832,
|
1691 |
+
"learning_rate": 2.231578947368421e-06,
|
1692 |
+
"logits/chosen": -1.8179492950439453,
|
1693 |
+
"logits/rejected": -1.8643207550048828,
|
1694 |
+
"logps/chosen": -158.73011779785156,
|
1695 |
+
"logps/rejected": -175.10726928710938,
|
1696 |
+
"loss": 0.0529,
|
1697 |
+
"rewards/accuracies": 1.0,
|
1698 |
+
"rewards/chosen": 1.9895429611206055,
|
1699 |
+
"rewards/margins": 7.000767707824707,
|
1700 |
+
"rewards/rejected": -5.011223793029785,
|
1701 |
+
"step": 840
|
1702 |
+
},
|
1703 |
+
{
|
1704 |
+
"epoch": 3.43,
|
1705 |
+
"eval_logits/chosen": -1.842317819595337,
|
1706 |
+
"eval_logits/rejected": -1.8808995485305786,
|
1707 |
+
"eval_logps/chosen": -187.3555145263672,
|
1708 |
+
"eval_logps/rejected": -258.76568603515625,
|
1709 |
+
"eval_loss": 0.2120712250471115,
|
1710 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1711 |
+
"eval_rewards/chosen": 2.1562187671661377,
|
1712 |
+
"eval_rewards/margins": 6.830246925354004,
|
1713 |
+
"eval_rewards/rejected": -4.674027919769287,
|
1714 |
+
"eval_runtime": 26.238,
|
1715 |
+
"eval_samples_per_second": 1.067,
|
1716 |
+
"eval_steps_per_second": 1.067,
|
1717 |
+
"step": 840
|
1718 |
+
},
|
1719 |
+
{
|
1720 |
+
"epoch": 3.47,
|
1721 |
+
"grad_norm": 5.9564328694250435e-05,
|
1722 |
+
"learning_rate": 2.1964912280701754e-06,
|
1723 |
+
"logits/chosen": -1.782867431640625,
|
1724 |
+
"logits/rejected": -1.8579784631729126,
|
1725 |
+
"logps/chosen": -105.4168472290039,
|
1726 |
+
"logps/rejected": -259.2267761230469,
|
1727 |
+
"loss": 0.258,
|
1728 |
+
"rewards/accuracies": 0.8999999761581421,
|
1729 |
+
"rewards/chosen": 4.022895336151123,
|
1730 |
+
"rewards/margins": 12.175846099853516,
|
1731 |
+
"rewards/rejected": -8.152950286865234,
|
1732 |
+
"step": 850
|
1733 |
+
},
|
1734 |
+
{
|
1735 |
+
"epoch": 3.51,
|
1736 |
+
"grad_norm": 1.3973094610264525e-05,
|
1737 |
+
"learning_rate": 2.16140350877193e-06,
|
1738 |
+
"logits/chosen": -1.7417545318603516,
|
1739 |
+
"logits/rejected": -1.8234918117523193,
|
1740 |
+
"logps/chosen": -115.1645736694336,
|
1741 |
+
"logps/rejected": -248.0410614013672,
|
1742 |
+
"loss": 0.039,
|
1743 |
+
"rewards/accuracies": 1.0,
|
1744 |
+
"rewards/chosen": 3.356029987335205,
|
1745 |
+
"rewards/margins": 12.46558952331543,
|
1746 |
+
"rewards/rejected": -9.109560012817383,
|
1747 |
+
"step": 860
|
1748 |
+
},
|
1749 |
+
{
|
1750 |
+
"epoch": 3.55,
|
1751 |
+
"grad_norm": 0.005180311389267445,
|
1752 |
+
"learning_rate": 2.1263157894736844e-06,
|
1753 |
+
"logits/chosen": -1.8123457431793213,
|
1754 |
+
"logits/rejected": -1.9022102355957031,
|
1755 |
+
"logps/chosen": -128.5855255126953,
|
1756 |
+
"logps/rejected": -246.685791015625,
|
1757 |
+
"loss": 0.001,
|
1758 |
+
"rewards/accuracies": 1.0,
|
1759 |
+
"rewards/chosen": 2.4856154918670654,
|
1760 |
+
"rewards/margins": 12.966238021850586,
|
1761 |
+
"rewards/rejected": -10.480623245239258,
|
1762 |
+
"step": 870
|
1763 |
+
},
|
1764 |
+
{
|
1765 |
+
"epoch": 3.55,
|
1766 |
+
"eval_logits/chosen": -1.8409830331802368,
|
1767 |
+
"eval_logits/rejected": -1.87991464138031,
|
1768 |
+
"eval_logps/chosen": -187.5316925048828,
|
1769 |
+
"eval_logps/rejected": -259.3367614746094,
|
1770 |
+
"eval_loss": 0.2091565877199173,
|
1771 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1772 |
+
"eval_rewards/chosen": 2.1033718585968018,
|
1773 |
+
"eval_rewards/margins": 6.94872522354126,
|
1774 |
+
"eval_rewards/rejected": -4.845353126525879,
|
1775 |
+
"eval_runtime": 26.2322,
|
1776 |
+
"eval_samples_per_second": 1.067,
|
1777 |
+
"eval_steps_per_second": 1.067,
|
1778 |
+
"step": 870
|
1779 |
+
},
|
1780 |
+
{
|
1781 |
+
"epoch": 3.59,
|
1782 |
+
"grad_norm": 0.3039734661579132,
|
1783 |
+
"learning_rate": 2.0912280701754387e-06,
|
1784 |
+
"logits/chosen": -1.7311649322509766,
|
1785 |
+
"logits/rejected": -1.8156330585479736,
|
1786 |
+
"logps/chosen": -146.54473876953125,
|
1787 |
+
"logps/rejected": -300.264892578125,
|
1788 |
+
"loss": 0.0179,
|
1789 |
+
"rewards/accuracies": 1.0,
|
1790 |
+
"rewards/chosen": 3.358633041381836,
|
1791 |
+
"rewards/margins": 11.333723068237305,
|
1792 |
+
"rewards/rejected": -7.975089073181152,
|
1793 |
+
"step": 880
|
1794 |
+
},
|
1795 |
+
{
|
1796 |
+
"epoch": 3.63,
|
1797 |
+
"grad_norm": 0.8097418546676636,
|
1798 |
+
"learning_rate": 2.0561403508771934e-06,
|
1799 |
+
"logits/chosen": -1.8806817531585693,
|
1800 |
+
"logits/rejected": -1.922952651977539,
|
1801 |
+
"logps/chosen": -126.9049301147461,
|
1802 |
+
"logps/rejected": -184.9642333984375,
|
1803 |
+
"loss": 0.0541,
|
1804 |
+
"rewards/accuracies": 1.0,
|
1805 |
+
"rewards/chosen": 1.0604941844940186,
|
1806 |
+
"rewards/margins": 6.54372501373291,
|
1807 |
+
"rewards/rejected": -5.4832305908203125,
|
1808 |
+
"step": 890
|
1809 |
+
},
|
1810 |
+
{
|
1811 |
+
"epoch": 3.67,
|
1812 |
+
"grad_norm": 0.025914475321769714,
|
1813 |
+
"learning_rate": 2.0210526315789477e-06,
|
1814 |
+
"logits/chosen": -1.8227866888046265,
|
1815 |
+
"logits/rejected": -1.9303566217422485,
|
1816 |
+
"logps/chosen": -167.85279846191406,
|
1817 |
+
"logps/rejected": -311.27655029296875,
|
1818 |
+
"loss": 0.0284,
|
1819 |
+
"rewards/accuracies": 1.0,
|
1820 |
+
"rewards/chosen": 2.1594152450561523,
|
1821 |
+
"rewards/margins": 9.647980690002441,
|
1822 |
+
"rewards/rejected": -7.488564968109131,
|
1823 |
+
"step": 900
|
1824 |
+
},
|
1825 |
+
{
|
1826 |
+
"epoch": 3.67,
|
1827 |
+
"eval_logits/chosen": -1.8366106748580933,
|
1828 |
+
"eval_logits/rejected": -1.8760128021240234,
|
1829 |
+
"eval_logps/chosen": -187.9383544921875,
|
1830 |
+
"eval_logps/rejected": -260.31500244140625,
|
1831 |
+
"eval_loss": 0.20062340795993805,
|
1832 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1833 |
+
"eval_rewards/chosen": 1.9813728332519531,
|
1834 |
+
"eval_rewards/margins": 7.1201934814453125,
|
1835 |
+
"eval_rewards/rejected": -5.138820171356201,
|
1836 |
+
"eval_runtime": 26.2437,
|
1837 |
+
"eval_samples_per_second": 1.067,
|
1838 |
+
"eval_steps_per_second": 1.067,
|
1839 |
+
"step": 900
|
1840 |
+
},
|
1841 |
+
{
|
1842 |
+
"epoch": 3.71,
|
1843 |
+
"grad_norm": 0.02055520936846733,
|
1844 |
+
"learning_rate": 1.985964912280702e-06,
|
1845 |
+
"logits/chosen": -1.8666080236434937,
|
1846 |
+
"logits/rejected": -1.9068281650543213,
|
1847 |
+
"logps/chosen": -171.9093475341797,
|
1848 |
+
"logps/rejected": -251.93643188476562,
|
1849 |
+
"loss": 0.2622,
|
1850 |
+
"rewards/accuracies": 0.800000011920929,
|
1851 |
+
"rewards/chosen": 0.18651972711086273,
|
1852 |
+
"rewards/margins": 8.190787315368652,
|
1853 |
+
"rewards/rejected": -8.004267692565918,
|
1854 |
+
"step": 910
|
1855 |
+
},
|
1856 |
+
{
|
1857 |
+
"epoch": 3.76,
|
1858 |
+
"grad_norm": 197.38717651367188,
|
1859 |
+
"learning_rate": 1.9508771929824562e-06,
|
1860 |
+
"logits/chosen": -1.8426361083984375,
|
1861 |
+
"logits/rejected": -1.9008222818374634,
|
1862 |
+
"logps/chosen": -150.75900268554688,
|
1863 |
+
"logps/rejected": -251.9907989501953,
|
1864 |
+
"loss": 0.0599,
|
1865 |
+
"rewards/accuracies": 1.0,
|
1866 |
+
"rewards/chosen": 0.7380223870277405,
|
1867 |
+
"rewards/margins": 9.459989547729492,
|
1868 |
+
"rewards/rejected": -8.721967697143555,
|
1869 |
+
"step": 920
|
1870 |
+
},
|
1871 |
+
{
|
1872 |
+
"epoch": 3.8,
|
1873 |
+
"grad_norm": 0.0,
|
1874 |
+
"learning_rate": 1.9157894736842105e-06,
|
1875 |
+
"logits/chosen": -1.8063271045684814,
|
1876 |
+
"logits/rejected": -1.8660167455673218,
|
1877 |
+
"logps/chosen": -198.50723266601562,
|
1878 |
+
"logps/rejected": -222.7967529296875,
|
1879 |
+
"loss": 0.0744,
|
1880 |
+
"rewards/accuracies": 0.8999999761581421,
|
1881 |
+
"rewards/chosen": 1.713536262512207,
|
1882 |
+
"rewards/margins": 9.678624153137207,
|
1883 |
+
"rewards/rejected": -7.965087890625,
|
1884 |
+
"step": 930
|
1885 |
+
},
|
1886 |
+
{
|
1887 |
+
"epoch": 3.8,
|
1888 |
+
"eval_logits/chosen": -1.833928108215332,
|
1889 |
+
"eval_logits/rejected": -1.873318076133728,
|
1890 |
+
"eval_logps/chosen": -188.06385803222656,
|
1891 |
+
"eval_logps/rejected": -260.6358337402344,
|
1892 |
+
"eval_loss": 0.1813022941350937,
|
1893 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1894 |
+
"eval_rewards/chosen": 1.943717122077942,
|
1895 |
+
"eval_rewards/margins": 7.178780555725098,
|
1896 |
+
"eval_rewards/rejected": -5.2350640296936035,
|
1897 |
+
"eval_runtime": 26.2336,
|
1898 |
+
"eval_samples_per_second": 1.067,
|
1899 |
+
"eval_steps_per_second": 1.067,
|
1900 |
+
"step": 930
|
1901 |
+
},
|
1902 |
+
{
|
1903 |
+
"epoch": 3.84,
|
1904 |
+
"grad_norm": 4.8926853679631677e-08,
|
1905 |
+
"learning_rate": 1.8807017543859652e-06,
|
1906 |
+
"logits/chosen": -1.8202307224273682,
|
1907 |
+
"logits/rejected": -1.9166450500488281,
|
1908 |
+
"logps/chosen": -112.94303131103516,
|
1909 |
+
"logps/rejected": -244.4494171142578,
|
1910 |
+
"loss": 0.3721,
|
1911 |
+
"rewards/accuracies": 0.8999999761581421,
|
1912 |
+
"rewards/chosen": 2.255749225616455,
|
1913 |
+
"rewards/margins": 10.898755073547363,
|
1914 |
+
"rewards/rejected": -8.643006324768066,
|
1915 |
+
"step": 940
|
1916 |
+
},
|
1917 |
+
{
|
1918 |
+
"epoch": 3.88,
|
1919 |
+
"grad_norm": 5.456500053405762,
|
1920 |
+
"learning_rate": 1.8456140350877193e-06,
|
1921 |
+
"logits/chosen": -1.9108766317367554,
|
1922 |
+
"logits/rejected": -1.9368479251861572,
|
1923 |
+
"logps/chosen": -132.4052734375,
|
1924 |
+
"logps/rejected": -179.675537109375,
|
1925 |
+
"loss": 0.0257,
|
1926 |
+
"rewards/accuracies": 1.0,
|
1927 |
+
"rewards/chosen": 1.554168701171875,
|
1928 |
+
"rewards/margins": 6.65416955947876,
|
1929 |
+
"rewards/rejected": -5.100000858306885,
|
1930 |
+
"step": 950
|
1931 |
+
},
|
1932 |
+
{
|
1933 |
+
"epoch": 3.92,
|
1934 |
+
"grad_norm": 9.124670028686523,
|
1935 |
+
"learning_rate": 1.810526315789474e-06,
|
1936 |
+
"logits/chosen": -1.7729440927505493,
|
1937 |
+
"logits/rejected": -1.8252880573272705,
|
1938 |
+
"logps/chosen": -139.13653564453125,
|
1939 |
+
"logps/rejected": -193.011962890625,
|
1940 |
+
"loss": 0.091,
|
1941 |
+
"rewards/accuracies": 1.0,
|
1942 |
+
"rewards/chosen": 3.6216557025909424,
|
1943 |
+
"rewards/margins": 8.979948043823242,
|
1944 |
+
"rewards/rejected": -5.3582916259765625,
|
1945 |
+
"step": 960
|
1946 |
+
},
|
1947 |
+
{
|
1948 |
+
"epoch": 3.92,
|
1949 |
+
"eval_logits/chosen": -1.8312675952911377,
|
1950 |
+
"eval_logits/rejected": -1.8707492351531982,
|
1951 |
+
"eval_logps/chosen": -188.43190002441406,
|
1952 |
+
"eval_logps/rejected": -261.29730224609375,
|
1953 |
+
"eval_loss": 0.17215047776699066,
|
1954 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1955 |
+
"eval_rewards/chosen": 1.8333097696304321,
|
1956 |
+
"eval_rewards/margins": 7.266810894012451,
|
1957 |
+
"eval_rewards/rejected": -5.433501243591309,
|
1958 |
+
"eval_runtime": 26.2333,
|
1959 |
+
"eval_samples_per_second": 1.067,
|
1960 |
+
"eval_steps_per_second": 1.067,
|
1961 |
+
"step": 960
|
1962 |
+
},
|
1963 |
+
{
|
1964 |
+
"epoch": 3.96,
|
1965 |
+
"grad_norm": 1.8921099353974569e-06,
|
1966 |
+
"learning_rate": 1.7754385964912283e-06,
|
1967 |
+
"logits/chosen": -1.8464349508285522,
|
1968 |
+
"logits/rejected": -1.8927011489868164,
|
1969 |
+
"logps/chosen": -172.9676971435547,
|
1970 |
+
"logps/rejected": -224.15780639648438,
|
1971 |
+
"loss": 0.0795,
|
1972 |
+
"rewards/accuracies": 1.0,
|
1973 |
+
"rewards/chosen": 2.649458169937134,
|
1974 |
+
"rewards/margins": 9.076958656311035,
|
1975 |
+
"rewards/rejected": -6.4275007247924805,
|
1976 |
+
"step": 970
|
1977 |
+
},
|
1978 |
+
{
|
1979 |
+
"epoch": 4.0,
|
1980 |
+
"grad_norm": 318.0891418457031,
|
1981 |
+
"learning_rate": 1.7403508771929826e-06,
|
1982 |
+
"logits/chosen": -1.8083856105804443,
|
1983 |
+
"logits/rejected": -1.8901259899139404,
|
1984 |
+
"logps/chosen": -209.17422485351562,
|
1985 |
+
"logps/rejected": -284.4140625,
|
1986 |
+
"loss": 0.0583,
|
1987 |
+
"rewards/accuracies": 1.0,
|
1988 |
+
"rewards/chosen": 2.1195216178894043,
|
1989 |
+
"rewards/margins": 12.128747940063477,
|
1990 |
+
"rewards/rejected": -10.00922679901123,
|
1991 |
+
"step": 980
|
1992 |
+
},
|
1993 |
+
{
|
1994 |
+
"epoch": 4.04,
|
1995 |
+
"grad_norm": 13.0223388671875,
|
1996 |
+
"learning_rate": 1.705263157894737e-06,
|
1997 |
+
"logits/chosen": -1.847835898399353,
|
1998 |
+
"logits/rejected": -1.868665337562561,
|
1999 |
+
"logps/chosen": -149.8837127685547,
|
2000 |
+
"logps/rejected": -161.1732940673828,
|
2001 |
+
"loss": 0.3504,
|
2002 |
+
"rewards/accuracies": 0.8999999761581421,
|
2003 |
+
"rewards/chosen": 1.0685189962387085,
|
2004 |
+
"rewards/margins": 6.018086910247803,
|
2005 |
+
"rewards/rejected": -4.9495673179626465,
|
2006 |
+
"step": 990
|
2007 |
+
},
|
2008 |
+
{
|
2009 |
+
"epoch": 4.04,
|
2010 |
+
"eval_logits/chosen": -1.8279262781143188,
|
2011 |
+
"eval_logits/rejected": -1.8672196865081787,
|
2012 |
+
"eval_logps/chosen": -188.3168487548828,
|
2013 |
+
"eval_logps/rejected": -261.04876708984375,
|
2014 |
+
"eval_loss": 0.14871637523174286,
|
2015 |
+
"eval_rewards/accuracies": 0.9285714030265808,
|
2016 |
+
"eval_rewards/chosen": 1.8678233623504639,
|
2017 |
+
"eval_rewards/margins": 7.22676944732666,
|
2018 |
+
"eval_rewards/rejected": -5.358946800231934,
|
2019 |
+
"eval_runtime": 26.2331,
|
2020 |
+
"eval_samples_per_second": 1.067,
|
2021 |
+
"eval_steps_per_second": 1.067,
|
2022 |
+
"step": 990
|
2023 |
+
},
|
2024 |
+
{
|
2025 |
+
"epoch": 4.08,
|
2026 |
+
"grad_norm": 0.6881242394447327,
|
2027 |
+
"learning_rate": 1.6701754385964914e-06,
|
2028 |
+
"logits/chosen": -1.9393212795257568,
|
2029 |
+
"logits/rejected": -1.9376901388168335,
|
2030 |
+
"logps/chosen": -155.8722381591797,
|
2031 |
+
"logps/rejected": -191.2816925048828,
|
2032 |
+
"loss": 0.0102,
|
2033 |
+
"rewards/accuracies": 1.0,
|
2034 |
+
"rewards/chosen": 2.3404548168182373,
|
2035 |
+
"rewards/margins": 7.390922546386719,
|
2036 |
+
"rewards/rejected": -5.0504679679870605,
|
2037 |
+
"step": 1000
|
2038 |
+
},
|
2039 |
+
{
|
2040 |
+
"epoch": 4.12,
|
2041 |
+
"grad_norm": 0.0015760493697598577,
|
2042 |
+
"learning_rate": 1.6350877192982456e-06,
|
2043 |
+
"logits/chosen": -1.7732092142105103,
|
2044 |
+
"logits/rejected": -1.8348764181137085,
|
2045 |
+
"logps/chosen": -204.78118896484375,
|
2046 |
+
"logps/rejected": -285.19171142578125,
|
2047 |
+
"loss": 0.0728,
|
2048 |
+
"rewards/accuracies": 0.8999999761581421,
|
2049 |
+
"rewards/chosen": 1.1964905261993408,
|
2050 |
+
"rewards/margins": 11.535087585449219,
|
2051 |
+
"rewards/rejected": -10.338597297668457,
|
2052 |
+
"step": 1010
|
2053 |
+
},
|
2054 |
+
{
|
2055 |
+
"epoch": 4.16,
|
2056 |
+
"grad_norm": 5.320696345734177e-07,
|
2057 |
+
"learning_rate": 1.6000000000000001e-06,
|
2058 |
+
"logits/chosen": -1.7163060903549194,
|
2059 |
+
"logits/rejected": -1.8586156368255615,
|
2060 |
+
"logps/chosen": -123.50263977050781,
|
2061 |
+
"logps/rejected": -320.79693603515625,
|
2062 |
+
"loss": 0.0071,
|
2063 |
+
"rewards/accuracies": 1.0,
|
2064 |
+
"rewards/chosen": 0.7505568265914917,
|
2065 |
+
"rewards/margins": 10.981633186340332,
|
2066 |
+
"rewards/rejected": -10.231074333190918,
|
2067 |
+
"step": 1020
|
2068 |
+
},
|
2069 |
+
{
|
2070 |
+
"epoch": 4.16,
|
2071 |
+
"eval_logits/chosen": -1.824275016784668,
|
2072 |
+
"eval_logits/rejected": -1.8636810779571533,
|
2073 |
+
"eval_logps/chosen": -188.54676818847656,
|
2074 |
+
"eval_logps/rejected": -261.5804748535156,
|
2075 |
+
"eval_loss": 0.14029009640216827,
|
2076 |
+
"eval_rewards/accuracies": 0.9285714030265808,
|
2077 |
+
"eval_rewards/chosen": 1.7988510131835938,
|
2078 |
+
"eval_rewards/margins": 7.317309856414795,
|
2079 |
+
"eval_rewards/rejected": -5.518458843231201,
|
2080 |
+
"eval_runtime": 26.2364,
|
2081 |
+
"eval_samples_per_second": 1.067,
|
2082 |
+
"eval_steps_per_second": 1.067,
|
2083 |
+
"step": 1020
|
2084 |
+
},
|
2085 |
+
{
|
2086 |
+
"epoch": 4.2,
|
2087 |
+
"grad_norm": 0.002488874364644289,
|
2088 |
+
"learning_rate": 1.5649122807017544e-06,
|
2089 |
+
"logits/chosen": -1.7597904205322266,
|
2090 |
+
"logits/rejected": -1.8473920822143555,
|
2091 |
+
"logps/chosen": -165.49359130859375,
|
2092 |
+
"logps/rejected": -273.9131774902344,
|
2093 |
+
"loss": 0.0071,
|
2094 |
+
"rewards/accuracies": 1.0,
|
2095 |
+
"rewards/chosen": 1.4215660095214844,
|
2096 |
+
"rewards/margins": 10.477560043334961,
|
2097 |
+
"rewards/rejected": -9.055994033813477,
|
2098 |
+
"step": 1030
|
2099 |
+
},
|
2100 |
+
{
|
2101 |
+
"epoch": 4.24,
|
2102 |
+
"grad_norm": 0.0017074132338166237,
|
2103 |
+
"learning_rate": 1.529824561403509e-06,
|
2104 |
+
"logits/chosen": -1.7763557434082031,
|
2105 |
+
"logits/rejected": -1.8936008214950562,
|
2106 |
+
"logps/chosen": -133.30006408691406,
|
2107 |
+
"logps/rejected": -336.57427978515625,
|
2108 |
+
"loss": 0.018,
|
2109 |
+
"rewards/accuracies": 1.0,
|
2110 |
+
"rewards/chosen": 4.363164901733398,
|
2111 |
+
"rewards/margins": 14.957908630371094,
|
2112 |
+
"rewards/rejected": -10.594743728637695,
|
2113 |
+
"step": 1040
|
2114 |
+
},
|
2115 |
+
{
|
2116 |
+
"epoch": 4.29,
|
2117 |
+
"grad_norm": 66.57980346679688,
|
2118 |
+
"learning_rate": 1.4947368421052632e-06,
|
2119 |
+
"logits/chosen": -1.835628867149353,
|
2120 |
+
"logits/rejected": -1.8976163864135742,
|
2121 |
+
"logps/chosen": -177.20309448242188,
|
2122 |
+
"logps/rejected": -287.8599548339844,
|
2123 |
+
"loss": 0.0131,
|
2124 |
+
"rewards/accuracies": 1.0,
|
2125 |
+
"rewards/chosen": 2.568208694458008,
|
2126 |
+
"rewards/margins": 11.586195945739746,
|
2127 |
+
"rewards/rejected": -9.017987251281738,
|
2128 |
+
"step": 1050
|
2129 |
+
},
|
2130 |
+
{
|
2131 |
+
"epoch": 4.29,
|
2132 |
+
"eval_logits/chosen": -1.8221967220306396,
|
2133 |
+
"eval_logits/rejected": -1.8615671396255493,
|
2134 |
+
"eval_logps/chosen": -188.5262451171875,
|
2135 |
+
"eval_logps/rejected": -261.6840515136719,
|
2136 |
+
"eval_loss": 0.13124893605709076,
|
2137 |
+
"eval_rewards/accuracies": 0.9285714030265808,
|
2138 |
+
"eval_rewards/chosen": 1.805002212524414,
|
2139 |
+
"eval_rewards/margins": 7.354526042938232,
|
2140 |
+
"eval_rewards/rejected": -5.549524784088135,
|
2141 |
+
"eval_runtime": 26.2328,
|
2142 |
+
"eval_samples_per_second": 1.067,
|
2143 |
+
"eval_steps_per_second": 1.067,
|
2144 |
+
"step": 1050
|
2145 |
+
},
|
2146 |
+
{
|
2147 |
+
"epoch": 4.33,
|
2148 |
+
"grad_norm": 0.027719292789697647,
|
2149 |
+
"learning_rate": 1.4596491228070175e-06,
|
2150 |
+
"logits/chosen": -1.7560068368911743,
|
2151 |
+
"logits/rejected": -1.8672195672988892,
|
2152 |
+
"logps/chosen": -189.530029296875,
|
2153 |
+
"logps/rejected": -330.6167907714844,
|
2154 |
+
"loss": 0.003,
|
2155 |
+
"rewards/accuracies": 1.0,
|
2156 |
+
"rewards/chosen": 1.0810093879699707,
|
2157 |
+
"rewards/margins": 9.271745681762695,
|
2158 |
+
"rewards/rejected": -8.19073486328125,
|
2159 |
+
"step": 1060
|
2160 |
+
},
|
2161 |
+
{
|
2162 |
+
"epoch": 4.37,
|
2163 |
+
"grad_norm": 0.0,
|
2164 |
+
"learning_rate": 1.4245614035087722e-06,
|
2165 |
+
"logits/chosen": -1.8792110681533813,
|
2166 |
+
"logits/rejected": -1.9164355993270874,
|
2167 |
+
"logps/chosen": -212.42788696289062,
|
2168 |
+
"logps/rejected": -238.2472686767578,
|
2169 |
+
"loss": 0.1334,
|
2170 |
+
"rewards/accuracies": 0.8999999761581421,
|
2171 |
+
"rewards/chosen": 1.0076087713241577,
|
2172 |
+
"rewards/margins": 9.041505813598633,
|
2173 |
+
"rewards/rejected": -8.033895492553711,
|
2174 |
+
"step": 1070
|
2175 |
+
},
|
2176 |
+
{
|
2177 |
+
"epoch": 4.41,
|
2178 |
+
"grad_norm": 3.609489340306027e-06,
|
2179 |
+
"learning_rate": 1.3894736842105263e-06,
|
2180 |
+
"logits/chosen": -1.8213157653808594,
|
2181 |
+
"logits/rejected": -1.870356798171997,
|
2182 |
+
"logps/chosen": -80.74689483642578,
|
2183 |
+
"logps/rejected": -158.96109008789062,
|
2184 |
+
"loss": 0.0868,
|
2185 |
+
"rewards/accuracies": 1.0,
|
2186 |
+
"rewards/chosen": 2.509101629257202,
|
2187 |
+
"rewards/margins": 7.621730804443359,
|
2188 |
+
"rewards/rejected": -5.11262845993042,
|
2189 |
+
"step": 1080
|
2190 |
+
},
|
2191 |
+
{
|
2192 |
+
"epoch": 4.41,
|
2193 |
+
"eval_logits/chosen": -1.81948721408844,
|
2194 |
+
"eval_logits/rejected": -1.858702540397644,
|
2195 |
+
"eval_logps/chosen": -188.66748046875,
|
2196 |
+
"eval_logps/rejected": -261.94708251953125,
|
2197 |
+
"eval_loss": 0.1210150271654129,
|
2198 |
+
"eval_rewards/accuracies": 0.9285714030265808,
|
2199 |
+
"eval_rewards/chosen": 1.76263427734375,
|
2200 |
+
"eval_rewards/margins": 7.39107608795166,
|
2201 |
+
"eval_rewards/rejected": -5.62844181060791,
|
2202 |
+
"eval_runtime": 26.2393,
|
2203 |
+
"eval_samples_per_second": 1.067,
|
2204 |
+
"eval_steps_per_second": 1.067,
|
2205 |
+
"step": 1080
|
2206 |
+
},
|
2207 |
+
{
|
2208 |
+
"epoch": 4.45,
|
2209 |
+
"grad_norm": 0.7020970582962036,
|
2210 |
+
"learning_rate": 1.354385964912281e-06,
|
2211 |
+
"logits/chosen": -1.8757814168930054,
|
2212 |
+
"logits/rejected": -1.9657201766967773,
|
2213 |
+
"logps/chosen": -117.080810546875,
|
2214 |
+
"logps/rejected": -241.55252075195312,
|
2215 |
+
"loss": 0.0046,
|
2216 |
+
"rewards/accuracies": 1.0,
|
2217 |
+
"rewards/chosen": 1.6267999410629272,
|
2218 |
+
"rewards/margins": 10.32630729675293,
|
2219 |
+
"rewards/rejected": -8.699506759643555,
|
2220 |
+
"step": 1090
|
2221 |
+
},
|
2222 |
+
{
|
2223 |
+
"epoch": 4.49,
|
2224 |
+
"grad_norm": 0.00016405929636675864,
|
2225 |
+
"learning_rate": 1.3192982456140353e-06,
|
2226 |
+
"logits/chosen": -1.84481680393219,
|
2227 |
+
"logits/rejected": -1.8962256908416748,
|
2228 |
+
"logps/chosen": -176.97457885742188,
|
2229 |
+
"logps/rejected": -248.0132598876953,
|
2230 |
+
"loss": 0.0261,
|
2231 |
+
"rewards/accuracies": 1.0,
|
2232 |
+
"rewards/chosen": 1.7715377807617188,
|
2233 |
+
"rewards/margins": 11.10531234741211,
|
2234 |
+
"rewards/rejected": -9.333776473999023,
|
2235 |
+
"step": 1100
|
2236 |
+
},
|
2237 |
+
{
|
2238 |
+
"epoch": 4.53,
|
2239 |
+
"grad_norm": 0.0024285567924380302,
|
2240 |
+
"learning_rate": 1.2842105263157895e-06,
|
2241 |
+
"logits/chosen": -1.9240976572036743,
|
2242 |
+
"logits/rejected": -1.9567447900772095,
|
2243 |
+
"logps/chosen": -148.41506958007812,
|
2244 |
+
"logps/rejected": -262.0478820800781,
|
2245 |
+
"loss": 0.0041,
|
2246 |
+
"rewards/accuracies": 1.0,
|
2247 |
+
"rewards/chosen": 2.377047538757324,
|
2248 |
+
"rewards/margins": 12.98823070526123,
|
2249 |
+
"rewards/rejected": -10.611183166503906,
|
2250 |
+
"step": 1110
|
2251 |
+
},
|
2252 |
+
{
|
2253 |
+
"epoch": 4.53,
|
2254 |
+
"eval_logits/chosen": -1.8172996044158936,
|
2255 |
+
"eval_logits/rejected": -1.8566371202468872,
|
2256 |
+
"eval_logps/chosen": -188.92127990722656,
|
2257 |
+
"eval_logps/rejected": -262.44561767578125,
|
2258 |
+
"eval_loss": 0.1205989271402359,
|
2259 |
+
"eval_rewards/accuracies": 0.9285714030265808,
|
2260 |
+
"eval_rewards/chosen": 1.686492681503296,
|
2261 |
+
"eval_rewards/margins": 7.464495658874512,
|
2262 |
+
"eval_rewards/rejected": -5.778003215789795,
|
2263 |
+
"eval_runtime": 26.2482,
|
2264 |
+
"eval_samples_per_second": 1.067,
|
2265 |
+
"eval_steps_per_second": 1.067,
|
2266 |
+
"step": 1110
|
2267 |
+
},
|
2268 |
+
{
|
2269 |
+
"epoch": 4.57,
|
2270 |
+
"grad_norm": 26.255502700805664,
|
2271 |
+
"learning_rate": 1.249122807017544e-06,
|
2272 |
+
"logits/chosen": -1.7752002477645874,
|
2273 |
+
"logits/rejected": -1.8067470788955688,
|
2274 |
+
"logps/chosen": -199.05264282226562,
|
2275 |
+
"logps/rejected": -198.15122985839844,
|
2276 |
+
"loss": 0.0289,
|
2277 |
+
"rewards/accuracies": 1.0,
|
2278 |
+
"rewards/chosen": 2.024845600128174,
|
2279 |
+
"rewards/margins": 10.319567680358887,
|
2280 |
+
"rewards/rejected": -8.294721603393555,
|
2281 |
+
"step": 1120
|
2282 |
+
},
|
2283 |
+
{
|
2284 |
+
"epoch": 4.61,
|
2285 |
+
"grad_norm": 170.23622131347656,
|
2286 |
+
"learning_rate": 1.2140350877192983e-06,
|
2287 |
+
"logits/chosen": -1.7735440731048584,
|
2288 |
+
"logits/rejected": -1.8578115701675415,
|
2289 |
+
"logps/chosen": -140.52511596679688,
|
2290 |
+
"logps/rejected": -257.21124267578125,
|
2291 |
+
"loss": 0.0478,
|
2292 |
+
"rewards/accuracies": 1.0,
|
2293 |
+
"rewards/chosen": 3.585864305496216,
|
2294 |
+
"rewards/margins": 12.626684188842773,
|
2295 |
+
"rewards/rejected": -9.04081916809082,
|
2296 |
+
"step": 1130
|
2297 |
+
},
|
2298 |
+
{
|
2299 |
+
"epoch": 4.65,
|
2300 |
+
"grad_norm": 0.09534689784049988,
|
2301 |
+
"learning_rate": 1.1789473684210526e-06,
|
2302 |
+
"logits/chosen": -1.8365780115127563,
|
2303 |
+
"logits/rejected": -1.846933364868164,
|
2304 |
+
"logps/chosen": -90.59309387207031,
|
2305 |
+
"logps/rejected": -117.09541320800781,
|
2306 |
+
"loss": 0.0107,
|
2307 |
+
"rewards/accuracies": 1.0,
|
2308 |
+
"rewards/chosen": 4.559220314025879,
|
2309 |
+
"rewards/margins": 10.236574172973633,
|
2310 |
+
"rewards/rejected": -5.677353382110596,
|
2311 |
+
"step": 1140
|
2312 |
+
},
|
2313 |
+
{
|
2314 |
+
"epoch": 4.65,
|
2315 |
+
"eval_logits/chosen": -1.8170740604400635,
|
2316 |
+
"eval_logits/rejected": -1.8563201427459717,
|
2317 |
+
"eval_logps/chosen": -189.08616638183594,
|
2318 |
+
"eval_logps/rejected": -262.8173828125,
|
2319 |
+
"eval_loss": 0.1178145483136177,
|
2320 |
+
"eval_rewards/accuracies": 0.9642857313156128,
|
2321 |
+
"eval_rewards/chosen": 1.6370302438735962,
|
2322 |
+
"eval_rewards/margins": 7.526562213897705,
|
2323 |
+
"eval_rewards/rejected": -5.88953161239624,
|
2324 |
+
"eval_runtime": 26.2301,
|
2325 |
+
"eval_samples_per_second": 1.067,
|
2326 |
+
"eval_steps_per_second": 1.067,
|
2327 |
+
"step": 1140
|
2328 |
+
},
|
2329 |
+
{
|
2330 |
+
"epoch": 4.69,
|
2331 |
+
"grad_norm": 0.0,
|
2332 |
+
"learning_rate": 1.143859649122807e-06,
|
2333 |
+
"logits/chosen": -1.7817316055297852,
|
2334 |
+
"logits/rejected": -1.8392887115478516,
|
2335 |
+
"logps/chosen": -162.9131317138672,
|
2336 |
+
"logps/rejected": -241.8101043701172,
|
2337 |
+
"loss": 0.0639,
|
2338 |
+
"rewards/accuracies": 1.0,
|
2339 |
+
"rewards/chosen": 1.5432535409927368,
|
2340 |
+
"rewards/margins": 9.659387588500977,
|
2341 |
+
"rewards/rejected": -8.116134643554688,
|
2342 |
+
"step": 1150
|
2343 |
+
},
|
2344 |
+
{
|
2345 |
+
"epoch": 4.73,
|
2346 |
+
"grad_norm": 1.92129647731781,
|
2347 |
+
"learning_rate": 1.1087719298245616e-06,
|
2348 |
+
"logits/chosen": -1.856563925743103,
|
2349 |
+
"logits/rejected": -1.8815357685089111,
|
2350 |
+
"logps/chosen": -140.67437744140625,
|
2351 |
+
"logps/rejected": -180.8864288330078,
|
2352 |
+
"loss": 0.2,
|
2353 |
+
"rewards/accuracies": 0.8999999761581421,
|
2354 |
+
"rewards/chosen": 1.501416563987732,
|
2355 |
+
"rewards/margins": 8.230588912963867,
|
2356 |
+
"rewards/rejected": -6.7291717529296875,
|
2357 |
+
"step": 1160
|
2358 |
+
},
|
2359 |
+
{
|
2360 |
+
"epoch": 4.78,
|
2361 |
+
"grad_norm": 4.147223472595215,
|
2362 |
+
"learning_rate": 1.0736842105263159e-06,
|
2363 |
+
"logits/chosen": -1.7213938236236572,
|
2364 |
+
"logits/rejected": -1.7904603481292725,
|
2365 |
+
"logps/chosen": -125.3969497680664,
|
2366 |
+
"logps/rejected": -262.72662353515625,
|
2367 |
+
"loss": 0.0084,
|
2368 |
+
"rewards/accuracies": 1.0,
|
2369 |
+
"rewards/chosen": 3.621525287628174,
|
2370 |
+
"rewards/margins": 12.910409927368164,
|
2371 |
+
"rewards/rejected": -9.288885116577148,
|
2372 |
+
"step": 1170
|
2373 |
+
},
|
2374 |
+
{
|
2375 |
+
"epoch": 4.78,
|
2376 |
+
"eval_logits/chosen": -1.815858244895935,
|
2377 |
+
"eval_logits/rejected": -1.8551596403121948,
|
2378 |
+
"eval_logps/chosen": -189.174072265625,
|
2379 |
+
"eval_logps/rejected": -262.9738464355469,
|
2380 |
+
"eval_loss": 0.11231832951307297,
|
2381 |
+
"eval_rewards/accuracies": 0.9642857313156128,
|
2382 |
+
"eval_rewards/chosen": 1.610654592514038,
|
2383 |
+
"eval_rewards/margins": 7.547122478485107,
|
2384 |
+
"eval_rewards/rejected": -5.936468601226807,
|
2385 |
+
"eval_runtime": 26.2436,
|
2386 |
+
"eval_samples_per_second": 1.067,
|
2387 |
+
"eval_steps_per_second": 1.067,
|
2388 |
+
"step": 1170
|
2389 |
+
}
|
2390 |
+
],
|
2391 |
+
"logging_steps": 10,
|
2392 |
+
"max_steps": 1470,
|
2393 |
+
"num_input_tokens_seen": 0,
|
2394 |
+
"num_train_epochs": 6,
|
2395 |
+
"save_steps": 90,
|
2396 |
+
"total_flos": 0.0,
|
2397 |
+
"train_batch_size": 1,
|
2398 |
+
"trial_name": null,
|
2399 |
+
"trial_params": null
|
2400 |
+
}
|
checkpoint-1170/training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:41693834eff4b96ebb423cc3037dde0e25267c10f91f2e959260eee7bddf1ddb
|
3 |
+
size 5304
|
checkpoint-1260/README.md
ADDED
@@ -0,0 +1,202 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
library_name: peft
|
3 |
+
base_model: Weni/WeniGPT-Agents-Mistral-1.0.6-SFT-merged
|
4 |
+
---
|
5 |
+
|
6 |
+
# Model Card for Model ID
|
7 |
+
|
8 |
+
<!-- Provide a quick summary of what the model is/does. -->
|
9 |
+
|
10 |
+
|
11 |
+
|
12 |
+
## Model Details
|
13 |
+
|
14 |
+
### Model Description
|
15 |
+
|
16 |
+
<!-- Provide a longer summary of what this model is. -->
|
17 |
+
|
18 |
+
|
19 |
+
|
20 |
+
- **Developed by:** [More Information Needed]
|
21 |
+
- **Funded by [optional]:** [More Information Needed]
|
22 |
+
- **Shared by [optional]:** [More Information Needed]
|
23 |
+
- **Model type:** [More Information Needed]
|
24 |
+
- **Language(s) (NLP):** [More Information Needed]
|
25 |
+
- **License:** [More Information Needed]
|
26 |
+
- **Finetuned from model [optional]:** [More Information Needed]
|
27 |
+
|
28 |
+
### Model Sources [optional]
|
29 |
+
|
30 |
+
<!-- Provide the basic links for the model. -->
|
31 |
+
|
32 |
+
- **Repository:** [More Information Needed]
|
33 |
+
- **Paper [optional]:** [More Information Needed]
|
34 |
+
- **Demo [optional]:** [More Information Needed]
|
35 |
+
|
36 |
+
## Uses
|
37 |
+
|
38 |
+
<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
|
39 |
+
|
40 |
+
### Direct Use
|
41 |
+
|
42 |
+
<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
|
43 |
+
|
44 |
+
[More Information Needed]
|
45 |
+
|
46 |
+
### Downstream Use [optional]
|
47 |
+
|
48 |
+
<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
|
49 |
+
|
50 |
+
[More Information Needed]
|
51 |
+
|
52 |
+
### Out-of-Scope Use
|
53 |
+
|
54 |
+
<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
|
55 |
+
|
56 |
+
[More Information Needed]
|
57 |
+
|
58 |
+
## Bias, Risks, and Limitations
|
59 |
+
|
60 |
+
<!-- This section is meant to convey both technical and sociotechnical limitations. -->
|
61 |
+
|
62 |
+
[More Information Needed]
|
63 |
+
|
64 |
+
### Recommendations
|
65 |
+
|
66 |
+
<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
|
67 |
+
|
68 |
+
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
|
69 |
+
|
70 |
+
## How to Get Started with the Model
|
71 |
+
|
72 |
+
Use the code below to get started with the model.
|
73 |
+
|
74 |
+
[More Information Needed]
|
75 |
+
|
76 |
+
## Training Details
|
77 |
+
|
78 |
+
### Training Data
|
79 |
+
|
80 |
+
<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
|
81 |
+
|
82 |
+
[More Information Needed]
|
83 |
+
|
84 |
+
### Training Procedure
|
85 |
+
|
86 |
+
<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
|
87 |
+
|
88 |
+
#### Preprocessing [optional]
|
89 |
+
|
90 |
+
[More Information Needed]
|
91 |
+
|
92 |
+
|
93 |
+
#### Training Hyperparameters
|
94 |
+
|
95 |
+
- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
|
96 |
+
|
97 |
+
#### Speeds, Sizes, Times [optional]
|
98 |
+
|
99 |
+
<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
|
100 |
+
|
101 |
+
[More Information Needed]
|
102 |
+
|
103 |
+
## Evaluation
|
104 |
+
|
105 |
+
<!-- This section describes the evaluation protocols and provides the results. -->
|
106 |
+
|
107 |
+
### Testing Data, Factors & Metrics
|
108 |
+
|
109 |
+
#### Testing Data
|
110 |
+
|
111 |
+
<!-- This should link to a Dataset Card if possible. -->
|
112 |
+
|
113 |
+
[More Information Needed]
|
114 |
+
|
115 |
+
#### Factors
|
116 |
+
|
117 |
+
<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
|
118 |
+
|
119 |
+
[More Information Needed]
|
120 |
+
|
121 |
+
#### Metrics
|
122 |
+
|
123 |
+
<!-- These are the evaluation metrics being used, ideally with a description of why. -->
|
124 |
+
|
125 |
+
[More Information Needed]
|
126 |
+
|
127 |
+
### Results
|
128 |
+
|
129 |
+
[More Information Needed]
|
130 |
+
|
131 |
+
#### Summary
|
132 |
+
|
133 |
+
|
134 |
+
|
135 |
+
## Model Examination [optional]
|
136 |
+
|
137 |
+
<!-- Relevant interpretability work for the model goes here -->
|
138 |
+
|
139 |
+
[More Information Needed]
|
140 |
+
|
141 |
+
## Environmental Impact
|
142 |
+
|
143 |
+
<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
|
144 |
+
|
145 |
+
Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
|
146 |
+
|
147 |
+
- **Hardware Type:** [More Information Needed]
|
148 |
+
- **Hours used:** [More Information Needed]
|
149 |
+
- **Cloud Provider:** [More Information Needed]
|
150 |
+
- **Compute Region:** [More Information Needed]
|
151 |
+
- **Carbon Emitted:** [More Information Needed]
|
152 |
+
|
153 |
+
## Technical Specifications [optional]
|
154 |
+
|
155 |
+
### Model Architecture and Objective
|
156 |
+
|
157 |
+
[More Information Needed]
|
158 |
+
|
159 |
+
### Compute Infrastructure
|
160 |
+
|
161 |
+
[More Information Needed]
|
162 |
+
|
163 |
+
#### Hardware
|
164 |
+
|
165 |
+
[More Information Needed]
|
166 |
+
|
167 |
+
#### Software
|
168 |
+
|
169 |
+
[More Information Needed]
|
170 |
+
|
171 |
+
## Citation [optional]
|
172 |
+
|
173 |
+
<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
|
174 |
+
|
175 |
+
**BibTeX:**
|
176 |
+
|
177 |
+
[More Information Needed]
|
178 |
+
|
179 |
+
**APA:**
|
180 |
+
|
181 |
+
[More Information Needed]
|
182 |
+
|
183 |
+
## Glossary [optional]
|
184 |
+
|
185 |
+
<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
|
186 |
+
|
187 |
+
[More Information Needed]
|
188 |
+
|
189 |
+
## More Information [optional]
|
190 |
+
|
191 |
+
[More Information Needed]
|
192 |
+
|
193 |
+
## Model Card Authors [optional]
|
194 |
+
|
195 |
+
[More Information Needed]
|
196 |
+
|
197 |
+
## Model Card Contact
|
198 |
+
|
199 |
+
[More Information Needed]
|
200 |
+
### Framework versions
|
201 |
+
|
202 |
+
- PEFT 0.10.0
|
checkpoint-1260/adapter_config.json
ADDED
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"alpha_pattern": {},
|
3 |
+
"auto_mapping": null,
|
4 |
+
"base_model_name_or_path": "Weni/WeniGPT-Agents-Mistral-1.0.6-SFT-merged",
|
5 |
+
"bias": "none",
|
6 |
+
"fan_in_fan_out": false,
|
7 |
+
"inference_mode": true,
|
8 |
+
"init_lora_weights": true,
|
9 |
+
"layer_replication": null,
|
10 |
+
"layers_pattern": null,
|
11 |
+
"layers_to_transform": null,
|
12 |
+
"loftq_config": {},
|
13 |
+
"lora_alpha": 16,
|
14 |
+
"lora_dropout": 0.05,
|
15 |
+
"megatron_config": null,
|
16 |
+
"megatron_core": "megatron.core",
|
17 |
+
"modules_to_save": null,
|
18 |
+
"peft_type": "LORA",
|
19 |
+
"r": 8,
|
20 |
+
"rank_pattern": {},
|
21 |
+
"revision": null,
|
22 |
+
"target_modules": [
|
23 |
+
"q_proj",
|
24 |
+
"v_proj"
|
25 |
+
],
|
26 |
+
"task_type": "CAUSAL_LM",
|
27 |
+
"use_dora": false,
|
28 |
+
"use_rslora": false
|
29 |
+
}
|
checkpoint-1260/adapter_model.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:66bd037b3e4bc3756ef67880ba16855485d46283c7ccc9c3bc4bd21cbeadd670
|
3 |
+
size 13648432
|
checkpoint-1260/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:903823ee69190a5c49c8a9a4047a9274964cf5112b925f64f270bfdd5743e3a4
|
3 |
+
size 27370618
|
checkpoint-1260/rng_state.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:618ef55f6c5423457249c94a2577645800716eebc494bf2745faa205a24bad51
|
3 |
+
size 14244
|
checkpoint-1260/scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:606b186f14fe40f661006ac9490703065c74d82486d61757dded794261b66804
|
3 |
+
size 1064
|
checkpoint-1260/special_tokens_map.json
ADDED
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"bos_token": {
|
3 |
+
"content": "<s>",
|
4 |
+
"lstrip": false,
|
5 |
+
"normalized": false,
|
6 |
+
"rstrip": false,
|
7 |
+
"single_word": false
|
8 |
+
},
|
9 |
+
"eos_token": {
|
10 |
+
"content": "</s>",
|
11 |
+
"lstrip": false,
|
12 |
+
"normalized": false,
|
13 |
+
"rstrip": false,
|
14 |
+
"single_word": false
|
15 |
+
},
|
16 |
+
"pad_token": "<unk>",
|
17 |
+
"unk_token": {
|
18 |
+
"content": "<unk>",
|
19 |
+
"lstrip": false,
|
20 |
+
"normalized": false,
|
21 |
+
"rstrip": false,
|
22 |
+
"single_word": false
|
23 |
+
}
|
24 |
+
}
|
checkpoint-1260/tokenizer.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-1260/tokenizer.model
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:dadfd56d766715c61d2ef780a525ab43b8e6da4de6865bda3d95fdef5e134055
|
3 |
+
size 493443
|
checkpoint-1260/tokenizer_config.json
ADDED
@@ -0,0 +1,49 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"add_bos_token": true,
|
3 |
+
"add_eos_token": false,
|
4 |
+
"added_tokens_decoder": {
|
5 |
+
"0": {
|
6 |
+
"content": "<unk>",
|
7 |
+
"lstrip": false,
|
8 |
+
"normalized": false,
|
9 |
+
"rstrip": false,
|
10 |
+
"single_word": false,
|
11 |
+
"special": true
|
12 |
+
},
|
13 |
+
"1": {
|
14 |
+
"content": "<s>",
|
15 |
+
"lstrip": false,
|
16 |
+
"normalized": false,
|
17 |
+
"rstrip": false,
|
18 |
+
"single_word": false,
|
19 |
+
"special": true
|
20 |
+
},
|
21 |
+
"2": {
|
22 |
+
"content": "</s>",
|
23 |
+
"lstrip": false,
|
24 |
+
"normalized": false,
|
25 |
+
"rstrip": false,
|
26 |
+
"single_word": false,
|
27 |
+
"special": true
|
28 |
+
}
|
29 |
+
},
|
30 |
+
"additional_special_tokens": [],
|
31 |
+
"bos_token": "<s>",
|
32 |
+
"chat_template": "{% for message in messages %}{% if message['role'] == 'user' %}{{ bos_token + '[INST] ' + message['content'] + ' [/INST]' }}{% elif message['role'] == 'system' %}{{ '<<SYS>>\\n' + message['content'] + '\\n<</SYS>>\\n\\n' }}{% elif message['role'] == 'assistant' %}{{ ' ' + message['content'] + ' ' + eos_token }}{% endif %}{% endfor %}",
|
33 |
+
"clean_up_tokenization_spaces": false,
|
34 |
+
"eos_token": "</s>",
|
35 |
+
"legacy": true,
|
36 |
+
"max_lenght": 8192,
|
37 |
+
"max_length": 8192,
|
38 |
+
"model_max_length": 1000000000000000019884624838656,
|
39 |
+
"pad_token": "<unk>",
|
40 |
+
"padding": true,
|
41 |
+
"sp_model_kwargs": {},
|
42 |
+
"spaces_between_special_tokens": false,
|
43 |
+
"stride": 0,
|
44 |
+
"tokenizer_class": "LlamaTokenizer",
|
45 |
+
"truncation_side": "right",
|
46 |
+
"truncation_strategy": "longest_first",
|
47 |
+
"unk_token": "<unk>",
|
48 |
+
"use_default_system_prompt": false
|
49 |
+
}
|
checkpoint-1260/trainer_state.json
ADDED
@@ -0,0 +1,2583 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": 0.10307549685239792,
|
3 |
+
"best_model_checkpoint": "./mistral/20-04-24-Weni-WeniGPT-Agents-Mistral-1.0.6-SFT-1.0.7-DPO_Experiment on DPO with other hyperparameters and best SFT model of WeniGPT-2_max_steps-1470_batch_1_2024-04-20_ppid_9/checkpoint-1260",
|
4 |
+
"epoch": 5.142857142857143,
|
5 |
+
"eval_steps": 30,
|
6 |
+
"global_step": 1260,
|
7 |
+
"is_hyper_param_search": false,
|
8 |
+
"is_local_process_zero": true,
|
9 |
+
"is_world_process_zero": true,
|
10 |
+
"log_history": [
|
11 |
+
{
|
12 |
+
"epoch": 0.04,
|
13 |
+
"grad_norm": 83.44550323486328,
|
14 |
+
"learning_rate": 6.666666666666667e-07,
|
15 |
+
"logits/chosen": -1.7825796604156494,
|
16 |
+
"logits/rejected": -1.7886196374893188,
|
17 |
+
"logps/chosen": -169.71481323242188,
|
18 |
+
"logps/rejected": -167.6764373779297,
|
19 |
+
"loss": 0.6937,
|
20 |
+
"rewards/accuracies": 0.20000000298023224,
|
21 |
+
"rewards/chosen": -0.0002815247280523181,
|
22 |
+
"rewards/margins": -0.0010146332206204534,
|
23 |
+
"rewards/rejected": 0.0007331084925681353,
|
24 |
+
"step": 10
|
25 |
+
},
|
26 |
+
{
|
27 |
+
"epoch": 0.08,
|
28 |
+
"grad_norm": 87.56982421875,
|
29 |
+
"learning_rate": 1.777777777777778e-06,
|
30 |
+
"logits/chosen": -1.623265027999878,
|
31 |
+
"logits/rejected": -1.6968040466308594,
|
32 |
+
"logps/chosen": -128.7312469482422,
|
33 |
+
"logps/rejected": -184.20187377929688,
|
34 |
+
"loss": 0.6912,
|
35 |
+
"rewards/accuracies": 0.5,
|
36 |
+
"rewards/chosen": 0.0030528828501701355,
|
37 |
+
"rewards/margins": 0.004004458896815777,
|
38 |
+
"rewards/rejected": -0.0009515761630609632,
|
39 |
+
"step": 20
|
40 |
+
},
|
41 |
+
{
|
42 |
+
"epoch": 0.12,
|
43 |
+
"grad_norm": 113.4741439819336,
|
44 |
+
"learning_rate": 2.888888888888889e-06,
|
45 |
+
"logits/chosen": -1.7152118682861328,
|
46 |
+
"logits/rejected": -1.758681058883667,
|
47 |
+
"logps/chosen": -146.91494750976562,
|
48 |
+
"logps/rejected": -213.5966339111328,
|
49 |
+
"loss": 0.6781,
|
50 |
+
"rewards/accuracies": 1.0,
|
51 |
+
"rewards/chosen": 0.03251449763774872,
|
52 |
+
"rewards/margins": 0.030697176232933998,
|
53 |
+
"rewards/rejected": 0.0018173219868913293,
|
54 |
+
"step": 30
|
55 |
+
},
|
56 |
+
{
|
57 |
+
"epoch": 0.12,
|
58 |
+
"eval_logits/chosen": -1.795080304145813,
|
59 |
+
"eval_logits/rejected": -1.8307874202728271,
|
60 |
+
"eval_logps/chosen": -194.3749542236328,
|
61 |
+
"eval_logps/rejected": -243.1332244873047,
|
62 |
+
"eval_loss": 0.6762310266494751,
|
63 |
+
"eval_rewards/accuracies": 0.75,
|
64 |
+
"eval_rewards/chosen": 0.05039369314908981,
|
65 |
+
"eval_rewards/margins": 0.03467999026179314,
|
66 |
+
"eval_rewards/rejected": 0.01571369729936123,
|
67 |
+
"eval_runtime": 26.2365,
|
68 |
+
"eval_samples_per_second": 1.067,
|
69 |
+
"eval_steps_per_second": 1.067,
|
70 |
+
"step": 30
|
71 |
+
},
|
72 |
+
{
|
73 |
+
"epoch": 0.16,
|
74 |
+
"grad_norm": 50.695308685302734,
|
75 |
+
"learning_rate": 4.000000000000001e-06,
|
76 |
+
"logits/chosen": -1.9357620477676392,
|
77 |
+
"logits/rejected": -1.9375683069229126,
|
78 |
+
"logps/chosen": -183.8218231201172,
|
79 |
+
"logps/rejected": -115.80632019042969,
|
80 |
+
"loss": 0.6768,
|
81 |
+
"rewards/accuracies": 0.699999988079071,
|
82 |
+
"rewards/chosen": 0.056698836386203766,
|
83 |
+
"rewards/margins": 0.03413978964090347,
|
84 |
+
"rewards/rejected": 0.02255905233323574,
|
85 |
+
"step": 40
|
86 |
+
},
|
87 |
+
{
|
88 |
+
"epoch": 0.2,
|
89 |
+
"grad_norm": 71.6834716796875,
|
90 |
+
"learning_rate": 4.996491228070176e-06,
|
91 |
+
"logits/chosen": -1.7773185968399048,
|
92 |
+
"logits/rejected": -1.8331743478775024,
|
93 |
+
"logps/chosen": -167.78533935546875,
|
94 |
+
"logps/rejected": -227.0582733154297,
|
95 |
+
"loss": 0.6205,
|
96 |
+
"rewards/accuracies": 0.8999999761581421,
|
97 |
+
"rewards/chosen": 0.1449195146560669,
|
98 |
+
"rewards/margins": 0.15578031539916992,
|
99 |
+
"rewards/rejected": -0.010860783979296684,
|
100 |
+
"step": 50
|
101 |
+
},
|
102 |
+
{
|
103 |
+
"epoch": 0.24,
|
104 |
+
"grad_norm": 93.83502197265625,
|
105 |
+
"learning_rate": 4.96140350877193e-06,
|
106 |
+
"logits/chosen": -1.7739536762237549,
|
107 |
+
"logits/rejected": -1.8414790630340576,
|
108 |
+
"logps/chosen": -153.57481384277344,
|
109 |
+
"logps/rejected": -194.3485870361328,
|
110 |
+
"loss": 0.5918,
|
111 |
+
"rewards/accuracies": 0.8999999761581421,
|
112 |
+
"rewards/chosen": 0.2464807778596878,
|
113 |
+
"rewards/margins": 0.22958576679229736,
|
114 |
+
"rewards/rejected": 0.016895027831196785,
|
115 |
+
"step": 60
|
116 |
+
},
|
117 |
+
{
|
118 |
+
"epoch": 0.24,
|
119 |
+
"eval_logits/chosen": -1.7974774837493896,
|
120 |
+
"eval_logits/rejected": -1.8332849740982056,
|
121 |
+
"eval_logps/chosen": -193.71743774414062,
|
122 |
+
"eval_logps/rejected": -243.05783081054688,
|
123 |
+
"eval_loss": 0.5998325943946838,
|
124 |
+
"eval_rewards/accuracies": 0.7857142686843872,
|
125 |
+
"eval_rewards/chosen": 0.24764502048492432,
|
126 |
+
"eval_rewards/margins": 0.20930658280849457,
|
127 |
+
"eval_rewards/rejected": 0.03833846375346184,
|
128 |
+
"eval_runtime": 26.2349,
|
129 |
+
"eval_samples_per_second": 1.067,
|
130 |
+
"eval_steps_per_second": 1.067,
|
131 |
+
"step": 60
|
132 |
+
},
|
133 |
+
{
|
134 |
+
"epoch": 0.29,
|
135 |
+
"grad_norm": 109.71602630615234,
|
136 |
+
"learning_rate": 4.926315789473685e-06,
|
137 |
+
"logits/chosen": -1.8950660228729248,
|
138 |
+
"logits/rejected": -1.9148075580596924,
|
139 |
+
"logps/chosen": -169.89028930664062,
|
140 |
+
"logps/rejected": -177.87301635742188,
|
141 |
+
"loss": 0.5992,
|
142 |
+
"rewards/accuracies": 0.800000011920929,
|
143 |
+
"rewards/chosen": 0.15271134674549103,
|
144 |
+
"rewards/margins": 0.21031668782234192,
|
145 |
+
"rewards/rejected": -0.0576053187251091,
|
146 |
+
"step": 70
|
147 |
+
},
|
148 |
+
{
|
149 |
+
"epoch": 0.33,
|
150 |
+
"grad_norm": 90.3881607055664,
|
151 |
+
"learning_rate": 4.8912280701754386e-06,
|
152 |
+
"logits/chosen": -1.798600196838379,
|
153 |
+
"logits/rejected": -1.855521559715271,
|
154 |
+
"logps/chosen": -160.3217315673828,
|
155 |
+
"logps/rejected": -247.6388397216797,
|
156 |
+
"loss": 0.5551,
|
157 |
+
"rewards/accuracies": 0.699999988079071,
|
158 |
+
"rewards/chosen": 0.29141485691070557,
|
159 |
+
"rewards/margins": 0.3369868993759155,
|
160 |
+
"rewards/rejected": -0.04557197913527489,
|
161 |
+
"step": 80
|
162 |
+
},
|
163 |
+
{
|
164 |
+
"epoch": 0.37,
|
165 |
+
"grad_norm": 147.27755737304688,
|
166 |
+
"learning_rate": 4.856140350877193e-06,
|
167 |
+
"logits/chosen": -1.8193047046661377,
|
168 |
+
"logits/rejected": -1.8773219585418701,
|
169 |
+
"logps/chosen": -155.37123107910156,
|
170 |
+
"logps/rejected": -245.6790008544922,
|
171 |
+
"loss": 0.4932,
|
172 |
+
"rewards/accuracies": 0.699999988079071,
|
173 |
+
"rewards/chosen": 0.5351110696792603,
|
174 |
+
"rewards/margins": 0.5927553176879883,
|
175 |
+
"rewards/rejected": -0.05764435604214668,
|
176 |
+
"step": 90
|
177 |
+
},
|
178 |
+
{
|
179 |
+
"epoch": 0.37,
|
180 |
+
"eval_logits/chosen": -1.8004387617111206,
|
181 |
+
"eval_logits/rejected": -1.836446762084961,
|
182 |
+
"eval_logps/chosen": -192.66908264160156,
|
183 |
+
"eval_logps/rejected": -242.95899963378906,
|
184 |
+
"eval_loss": 0.5072147846221924,
|
185 |
+
"eval_rewards/accuracies": 0.8214285969734192,
|
186 |
+
"eval_rewards/chosen": 0.5621528029441833,
|
187 |
+
"eval_rewards/margins": 0.4941706657409668,
|
188 |
+
"eval_rewards/rejected": 0.06798211485147476,
|
189 |
+
"eval_runtime": 26.2422,
|
190 |
+
"eval_samples_per_second": 1.067,
|
191 |
+
"eval_steps_per_second": 1.067,
|
192 |
+
"step": 90
|
193 |
+
},
|
194 |
+
{
|
195 |
+
"epoch": 0.41,
|
196 |
+
"grad_norm": 62.75928497314453,
|
197 |
+
"learning_rate": 4.821052631578948e-06,
|
198 |
+
"logits/chosen": -1.7835988998413086,
|
199 |
+
"logits/rejected": -1.8885362148284912,
|
200 |
+
"logps/chosen": -109.94624328613281,
|
201 |
+
"logps/rejected": -232.5241241455078,
|
202 |
+
"loss": 0.4252,
|
203 |
+
"rewards/accuracies": 1.0,
|
204 |
+
"rewards/chosen": 0.5394761562347412,
|
205 |
+
"rewards/margins": 0.787738025188446,
|
206 |
+
"rewards/rejected": -0.248261958360672,
|
207 |
+
"step": 100
|
208 |
+
},
|
209 |
+
{
|
210 |
+
"epoch": 0.45,
|
211 |
+
"grad_norm": 84.42039489746094,
|
212 |
+
"learning_rate": 4.785964912280702e-06,
|
213 |
+
"logits/chosen": -1.7081111669540405,
|
214 |
+
"logits/rejected": -1.7244077920913696,
|
215 |
+
"logps/chosen": -92.51998138427734,
|
216 |
+
"logps/rejected": -135.5617218017578,
|
217 |
+
"loss": 0.5077,
|
218 |
+
"rewards/accuracies": 0.800000011920929,
|
219 |
+
"rewards/chosen": 0.6028972864151001,
|
220 |
+
"rewards/margins": 0.511993408203125,
|
221 |
+
"rewards/rejected": 0.09090389311313629,
|
222 |
+
"step": 110
|
223 |
+
},
|
224 |
+
{
|
225 |
+
"epoch": 0.49,
|
226 |
+
"grad_norm": 26.61499786376953,
|
227 |
+
"learning_rate": 4.754385964912281e-06,
|
228 |
+
"logits/chosen": -1.9250166416168213,
|
229 |
+
"logits/rejected": -1.9426968097686768,
|
230 |
+
"logps/chosen": -152.0164337158203,
|
231 |
+
"logps/rejected": -154.2359161376953,
|
232 |
+
"loss": 0.4391,
|
233 |
+
"rewards/accuracies": 0.8999999761581421,
|
234 |
+
"rewards/chosen": 0.7938548922538757,
|
235 |
+
"rewards/margins": 0.8368040919303894,
|
236 |
+
"rewards/rejected": -0.042949117720127106,
|
237 |
+
"step": 120
|
238 |
+
},
|
239 |
+
{
|
240 |
+
"epoch": 0.49,
|
241 |
+
"eval_logits/chosen": -1.8051280975341797,
|
242 |
+
"eval_logits/rejected": -1.8413046598434448,
|
243 |
+
"eval_logps/chosen": -191.2982177734375,
|
244 |
+
"eval_logps/rejected": -242.81198120117188,
|
245 |
+
"eval_loss": 0.4335751235485077,
|
246 |
+
"eval_rewards/accuracies": 0.7857142686843872,
|
247 |
+
"eval_rewards/chosen": 0.9734136462211609,
|
248 |
+
"eval_rewards/margins": 0.8613229990005493,
|
249 |
+
"eval_rewards/rejected": 0.11209066957235336,
|
250 |
+
"eval_runtime": 26.2345,
|
251 |
+
"eval_samples_per_second": 1.067,
|
252 |
+
"eval_steps_per_second": 1.067,
|
253 |
+
"step": 120
|
254 |
+
},
|
255 |
+
{
|
256 |
+
"epoch": 0.53,
|
257 |
+
"grad_norm": 90.6368637084961,
|
258 |
+
"learning_rate": 4.7192982456140355e-06,
|
259 |
+
"logits/chosen": -1.865144968032837,
|
260 |
+
"logits/rejected": -1.9209120273590088,
|
261 |
+
"logps/chosen": -149.78990173339844,
|
262 |
+
"logps/rejected": -293.8749084472656,
|
263 |
+
"loss": 0.4133,
|
264 |
+
"rewards/accuracies": 0.800000011920929,
|
265 |
+
"rewards/chosen": 0.4554510712623596,
|
266 |
+
"rewards/margins": 0.9285024404525757,
|
267 |
+
"rewards/rejected": -0.47305139899253845,
|
268 |
+
"step": 130
|
269 |
+
},
|
270 |
+
{
|
271 |
+
"epoch": 0.57,
|
272 |
+
"grad_norm": 6.8597941398620605,
|
273 |
+
"learning_rate": 4.68421052631579e-06,
|
274 |
+
"logits/chosen": -1.8089487552642822,
|
275 |
+
"logits/rejected": -1.8955650329589844,
|
276 |
+
"logps/chosen": -100.26250457763672,
|
277 |
+
"logps/rejected": -209.1724090576172,
|
278 |
+
"loss": 0.3414,
|
279 |
+
"rewards/accuracies": 1.0,
|
280 |
+
"rewards/chosen": 0.9389357566833496,
|
281 |
+
"rewards/margins": 1.521246314048767,
|
282 |
+
"rewards/rejected": -0.5823107361793518,
|
283 |
+
"step": 140
|
284 |
+
},
|
285 |
+
{
|
286 |
+
"epoch": 0.61,
|
287 |
+
"grad_norm": 79.9189453125,
|
288 |
+
"learning_rate": 4.652631578947368e-06,
|
289 |
+
"logits/chosen": -1.760371446609497,
|
290 |
+
"logits/rejected": -1.8096182346343994,
|
291 |
+
"logps/chosen": -193.62579345703125,
|
292 |
+
"logps/rejected": -166.35800170898438,
|
293 |
+
"loss": 0.3208,
|
294 |
+
"rewards/accuracies": 0.8999999761581421,
|
295 |
+
"rewards/chosen": 1.2600924968719482,
|
296 |
+
"rewards/margins": 1.5845091342926025,
|
297 |
+
"rewards/rejected": -0.32441645860671997,
|
298 |
+
"step": 150
|
299 |
+
},
|
300 |
+
{
|
301 |
+
"epoch": 0.61,
|
302 |
+
"eval_logits/chosen": -1.8129692077636719,
|
303 |
+
"eval_logits/rejected": -1.849239706993103,
|
304 |
+
"eval_logps/chosen": -189.88929748535156,
|
305 |
+
"eval_logps/rejected": -242.91104125976562,
|
306 |
+
"eval_loss": 0.39326220750808716,
|
307 |
+
"eval_rewards/accuracies": 0.7857142686843872,
|
308 |
+
"eval_rewards/chosen": 1.3960883617401123,
|
309 |
+
"eval_rewards/margins": 1.3137177228927612,
|
310 |
+
"eval_rewards/rejected": 0.08237060159444809,
|
311 |
+
"eval_runtime": 26.3372,
|
312 |
+
"eval_samples_per_second": 1.063,
|
313 |
+
"eval_steps_per_second": 1.063,
|
314 |
+
"step": 150
|
315 |
+
},
|
316 |
+
{
|
317 |
+
"epoch": 0.65,
|
318 |
+
"grad_norm": 3.151426315307617,
|
319 |
+
"learning_rate": 4.617543859649123e-06,
|
320 |
+
"logits/chosen": -1.7185112237930298,
|
321 |
+
"logits/rejected": -1.749427080154419,
|
322 |
+
"logps/chosen": -178.8656463623047,
|
323 |
+
"logps/rejected": -201.32858276367188,
|
324 |
+
"loss": 0.4418,
|
325 |
+
"rewards/accuracies": 0.699999988079071,
|
326 |
+
"rewards/chosen": 1.3884060382843018,
|
327 |
+
"rewards/margins": 1.4542597532272339,
|
328 |
+
"rewards/rejected": -0.06585375964641571,
|
329 |
+
"step": 160
|
330 |
+
},
|
331 |
+
{
|
332 |
+
"epoch": 0.69,
|
333 |
+
"grad_norm": 37.79099655151367,
|
334 |
+
"learning_rate": 4.582456140350878e-06,
|
335 |
+
"logits/chosen": -1.730787992477417,
|
336 |
+
"logits/rejected": -1.8099721670150757,
|
337 |
+
"logps/chosen": -158.50277709960938,
|
338 |
+
"logps/rejected": -185.48129272460938,
|
339 |
+
"loss": 0.3912,
|
340 |
+
"rewards/accuracies": 0.800000011920929,
|
341 |
+
"rewards/chosen": 1.4253065586090088,
|
342 |
+
"rewards/margins": 1.2418479919433594,
|
343 |
+
"rewards/rejected": 0.18345877528190613,
|
344 |
+
"step": 170
|
345 |
+
},
|
346 |
+
{
|
347 |
+
"epoch": 0.73,
|
348 |
+
"grad_norm": 18.49593734741211,
|
349 |
+
"learning_rate": 4.547368421052632e-06,
|
350 |
+
"logits/chosen": -1.842390775680542,
|
351 |
+
"logits/rejected": -1.8834030628204346,
|
352 |
+
"logps/chosen": -125.84483337402344,
|
353 |
+
"logps/rejected": -182.78854370117188,
|
354 |
+
"loss": 0.3215,
|
355 |
+
"rewards/accuracies": 0.8999999761581421,
|
356 |
+
"rewards/chosen": 1.5408955812454224,
|
357 |
+
"rewards/margins": 1.4677695035934448,
|
358 |
+
"rewards/rejected": 0.07312598079442978,
|
359 |
+
"step": 180
|
360 |
+
},
|
361 |
+
{
|
362 |
+
"epoch": 0.73,
|
363 |
+
"eval_logits/chosen": -1.8194406032562256,
|
364 |
+
"eval_logits/rejected": -1.8561830520629883,
|
365 |
+
"eval_logps/chosen": -188.3820037841797,
|
366 |
+
"eval_logps/rejected": -243.13536071777344,
|
367 |
+
"eval_loss": 0.3756321966648102,
|
368 |
+
"eval_rewards/accuracies": 0.7857142686843872,
|
369 |
+
"eval_rewards/chosen": 1.8482797145843506,
|
370 |
+
"eval_rewards/margins": 1.8331985473632812,
|
371 |
+
"eval_rewards/rejected": 0.015081183984875679,
|
372 |
+
"eval_runtime": 26.2381,
|
373 |
+
"eval_samples_per_second": 1.067,
|
374 |
+
"eval_steps_per_second": 1.067,
|
375 |
+
"step": 180
|
376 |
+
},
|
377 |
+
{
|
378 |
+
"epoch": 0.78,
|
379 |
+
"grad_norm": 106.56973266601562,
|
380 |
+
"learning_rate": 4.512280701754386e-06,
|
381 |
+
"logits/chosen": -1.8318649530410767,
|
382 |
+
"logits/rejected": -1.8742272853851318,
|
383 |
+
"logps/chosen": -197.66619873046875,
|
384 |
+
"logps/rejected": -247.83560180664062,
|
385 |
+
"loss": 0.3303,
|
386 |
+
"rewards/accuracies": 0.800000011920929,
|
387 |
+
"rewards/chosen": 2.4889655113220215,
|
388 |
+
"rewards/margins": 2.536695957183838,
|
389 |
+
"rewards/rejected": -0.047730110585689545,
|
390 |
+
"step": 190
|
391 |
+
},
|
392 |
+
{
|
393 |
+
"epoch": 0.82,
|
394 |
+
"grad_norm": 1.9826362133026123,
|
395 |
+
"learning_rate": 4.47719298245614e-06,
|
396 |
+
"logits/chosen": -1.7579128742218018,
|
397 |
+
"logits/rejected": -1.7890180349349976,
|
398 |
+
"logps/chosen": -129.37191772460938,
|
399 |
+
"logps/rejected": -166.67636108398438,
|
400 |
+
"loss": 0.3292,
|
401 |
+
"rewards/accuracies": 0.8999999761581421,
|
402 |
+
"rewards/chosen": 2.5646984577178955,
|
403 |
+
"rewards/margins": 2.6955490112304688,
|
404 |
+
"rewards/rejected": -0.13085035979747772,
|
405 |
+
"step": 200
|
406 |
+
},
|
407 |
+
{
|
408 |
+
"epoch": 0.86,
|
409 |
+
"grad_norm": 6.777231693267822,
|
410 |
+
"learning_rate": 4.442105263157896e-06,
|
411 |
+
"logits/chosen": -1.7335002422332764,
|
412 |
+
"logits/rejected": -1.8272641897201538,
|
413 |
+
"logps/chosen": -114.22731018066406,
|
414 |
+
"logps/rejected": -215.8419647216797,
|
415 |
+
"loss": 0.0817,
|
416 |
+
"rewards/accuracies": 1.0,
|
417 |
+
"rewards/chosen": 3.278646469116211,
|
418 |
+
"rewards/margins": 4.009958267211914,
|
419 |
+
"rewards/rejected": -0.7313117980957031,
|
420 |
+
"step": 210
|
421 |
+
},
|
422 |
+
{
|
423 |
+
"epoch": 0.86,
|
424 |
+
"eval_logits/chosen": -1.826608419418335,
|
425 |
+
"eval_logits/rejected": -1.8640671968460083,
|
426 |
+
"eval_logps/chosen": -186.82992553710938,
|
427 |
+
"eval_logps/rejected": -243.80209350585938,
|
428 |
+
"eval_loss": 0.3834719657897949,
|
429 |
+
"eval_rewards/accuracies": 0.7857142686843872,
|
430 |
+
"eval_rewards/chosen": 2.3139071464538574,
|
431 |
+
"eval_rewards/margins": 2.4988515377044678,
|
432 |
+
"eval_rewards/rejected": -0.1849442571401596,
|
433 |
+
"eval_runtime": 26.2309,
|
434 |
+
"eval_samples_per_second": 1.067,
|
435 |
+
"eval_steps_per_second": 1.067,
|
436 |
+
"step": 210
|
437 |
+
},
|
438 |
+
{
|
439 |
+
"epoch": 0.9,
|
440 |
+
"grad_norm": 1.0230931043624878,
|
441 |
+
"learning_rate": 4.40701754385965e-06,
|
442 |
+
"logits/chosen": -1.7561031579971313,
|
443 |
+
"logits/rejected": -1.8601219654083252,
|
444 |
+
"logps/chosen": -134.11428833007812,
|
445 |
+
"logps/rejected": -273.5272216796875,
|
446 |
+
"loss": 0.1796,
|
447 |
+
"rewards/accuracies": 0.8999999761581421,
|
448 |
+
"rewards/chosen": 1.9340410232543945,
|
449 |
+
"rewards/margins": 3.8275578022003174,
|
450 |
+
"rewards/rejected": -1.8935168981552124,
|
451 |
+
"step": 220
|
452 |
+
},
|
453 |
+
{
|
454 |
+
"epoch": 0.94,
|
455 |
+
"grad_norm": 54.65635299682617,
|
456 |
+
"learning_rate": 4.371929824561404e-06,
|
457 |
+
"logits/chosen": -1.8512550592422485,
|
458 |
+
"logits/rejected": -1.900559663772583,
|
459 |
+
"logps/chosen": -210.9802703857422,
|
460 |
+
"logps/rejected": -240.60269165039062,
|
461 |
+
"loss": 0.5004,
|
462 |
+
"rewards/accuracies": 0.800000011920929,
|
463 |
+
"rewards/chosen": 1.5826597213745117,
|
464 |
+
"rewards/margins": 2.925461530685425,
|
465 |
+
"rewards/rejected": -1.342801809310913,
|
466 |
+
"step": 230
|
467 |
+
},
|
468 |
+
{
|
469 |
+
"epoch": 0.98,
|
470 |
+
"grad_norm": 18.797134399414062,
|
471 |
+
"learning_rate": 4.336842105263158e-06,
|
472 |
+
"logits/chosen": -1.7781873941421509,
|
473 |
+
"logits/rejected": -1.8780953884124756,
|
474 |
+
"logps/chosen": -136.18722534179688,
|
475 |
+
"logps/rejected": -257.26495361328125,
|
476 |
+
"loss": 0.137,
|
477 |
+
"rewards/accuracies": 0.8999999761581421,
|
478 |
+
"rewards/chosen": 1.9192733764648438,
|
479 |
+
"rewards/margins": 3.8388113975524902,
|
480 |
+
"rewards/rejected": -1.9195384979248047,
|
481 |
+
"step": 240
|
482 |
+
},
|
483 |
+
{
|
484 |
+
"epoch": 0.98,
|
485 |
+
"eval_logits/chosen": -1.83432936668396,
|
486 |
+
"eval_logits/rejected": -1.8722076416015625,
|
487 |
+
"eval_logps/chosen": -185.8831329345703,
|
488 |
+
"eval_logps/rejected": -244.8594207763672,
|
489 |
+
"eval_loss": 0.41323208808898926,
|
490 |
+
"eval_rewards/accuracies": 0.75,
|
491 |
+
"eval_rewards/chosen": 2.5979421138763428,
|
492 |
+
"eval_rewards/margins": 3.1000888347625732,
|
493 |
+
"eval_rewards/rejected": -0.5021467208862305,
|
494 |
+
"eval_runtime": 26.2323,
|
495 |
+
"eval_samples_per_second": 1.067,
|
496 |
+
"eval_steps_per_second": 1.067,
|
497 |
+
"step": 240
|
498 |
+
},
|
499 |
+
{
|
500 |
+
"epoch": 1.02,
|
501 |
+
"grad_norm": 132.51947021484375,
|
502 |
+
"learning_rate": 4.301754385964912e-06,
|
503 |
+
"logits/chosen": -1.8787914514541626,
|
504 |
+
"logits/rejected": -1.9520515203475952,
|
505 |
+
"logps/chosen": -186.46717834472656,
|
506 |
+
"logps/rejected": -250.008056640625,
|
507 |
+
"loss": 0.3531,
|
508 |
+
"rewards/accuracies": 0.800000011920929,
|
509 |
+
"rewards/chosen": 2.192800283432007,
|
510 |
+
"rewards/margins": 3.485269069671631,
|
511 |
+
"rewards/rejected": -1.2924686670303345,
|
512 |
+
"step": 250
|
513 |
+
},
|
514 |
+
{
|
515 |
+
"epoch": 1.06,
|
516 |
+
"grad_norm": 6.529983043670654,
|
517 |
+
"learning_rate": 4.266666666666668e-06,
|
518 |
+
"logits/chosen": -1.7029941082000732,
|
519 |
+
"logits/rejected": -1.7871618270874023,
|
520 |
+
"logps/chosen": -128.5960235595703,
|
521 |
+
"logps/rejected": -209.27603149414062,
|
522 |
+
"loss": 0.0883,
|
523 |
+
"rewards/accuracies": 0.8999999761581421,
|
524 |
+
"rewards/chosen": 2.7405991554260254,
|
525 |
+
"rewards/margins": 5.148273468017578,
|
526 |
+
"rewards/rejected": -2.4076738357543945,
|
527 |
+
"step": 260
|
528 |
+
},
|
529 |
+
{
|
530 |
+
"epoch": 1.1,
|
531 |
+
"grad_norm": 0.27644646167755127,
|
532 |
+
"learning_rate": 4.2315789473684215e-06,
|
533 |
+
"logits/chosen": -1.7499496936798096,
|
534 |
+
"logits/rejected": -1.8194109201431274,
|
535 |
+
"logps/chosen": -119.34013366699219,
|
536 |
+
"logps/rejected": -230.69229125976562,
|
537 |
+
"loss": 0.0997,
|
538 |
+
"rewards/accuracies": 0.8999999761581421,
|
539 |
+
"rewards/chosen": 4.146358013153076,
|
540 |
+
"rewards/margins": 5.72900390625,
|
541 |
+
"rewards/rejected": -1.5826464891433716,
|
542 |
+
"step": 270
|
543 |
+
},
|
544 |
+
{
|
545 |
+
"epoch": 1.1,
|
546 |
+
"eval_logits/chosen": -1.842972755432129,
|
547 |
+
"eval_logits/rejected": -1.8816190958023071,
|
548 |
+
"eval_logps/chosen": -185.4148406982422,
|
549 |
+
"eval_logps/rejected": -246.53672790527344,
|
550 |
+
"eval_loss": 0.46570485830307007,
|
551 |
+
"eval_rewards/accuracies": 0.75,
|
552 |
+
"eval_rewards/chosen": 2.7384250164031982,
|
553 |
+
"eval_rewards/margins": 3.743762969970703,
|
554 |
+
"eval_rewards/rejected": -1.0053375959396362,
|
555 |
+
"eval_runtime": 26.2348,
|
556 |
+
"eval_samples_per_second": 1.067,
|
557 |
+
"eval_steps_per_second": 1.067,
|
558 |
+
"step": 270
|
559 |
+
},
|
560 |
+
{
|
561 |
+
"epoch": 1.14,
|
562 |
+
"grad_norm": 31.16436195373535,
|
563 |
+
"learning_rate": 4.196491228070176e-06,
|
564 |
+
"logits/chosen": -1.8344509601593018,
|
565 |
+
"logits/rejected": -1.919553518295288,
|
566 |
+
"logps/chosen": -131.51156616210938,
|
567 |
+
"logps/rejected": -246.3345184326172,
|
568 |
+
"loss": 0.138,
|
569 |
+
"rewards/accuracies": 1.0,
|
570 |
+
"rewards/chosen": 2.982485771179199,
|
571 |
+
"rewards/margins": 6.261662483215332,
|
572 |
+
"rewards/rejected": -3.2791759967803955,
|
573 |
+
"step": 280
|
574 |
+
},
|
575 |
+
{
|
576 |
+
"epoch": 1.18,
|
577 |
+
"grad_norm": 0.08183232694864273,
|
578 |
+
"learning_rate": 4.16140350877193e-06,
|
579 |
+
"logits/chosen": -1.847378134727478,
|
580 |
+
"logits/rejected": -1.8774774074554443,
|
581 |
+
"logps/chosen": -127.52336120605469,
|
582 |
+
"logps/rejected": -157.3633270263672,
|
583 |
+
"loss": 0.3883,
|
584 |
+
"rewards/accuracies": 0.800000011920929,
|
585 |
+
"rewards/chosen": 0.7564169764518738,
|
586 |
+
"rewards/margins": 2.541962146759033,
|
587 |
+
"rewards/rejected": -1.7855453491210938,
|
588 |
+
"step": 290
|
589 |
+
},
|
590 |
+
{
|
591 |
+
"epoch": 1.22,
|
592 |
+
"grad_norm": 10.61989974975586,
|
593 |
+
"learning_rate": 4.126315789473685e-06,
|
594 |
+
"logits/chosen": -1.7770633697509766,
|
595 |
+
"logits/rejected": -1.8337452411651611,
|
596 |
+
"logps/chosen": -112.7281265258789,
|
597 |
+
"logps/rejected": -195.87020874023438,
|
598 |
+
"loss": 0.0432,
|
599 |
+
"rewards/accuracies": 1.0,
|
600 |
+
"rewards/chosen": 3.176725387573242,
|
601 |
+
"rewards/margins": 6.761993408203125,
|
602 |
+
"rewards/rejected": -3.5852675437927246,
|
603 |
+
"step": 300
|
604 |
+
},
|
605 |
+
{
|
606 |
+
"epoch": 1.22,
|
607 |
+
"eval_logits/chosen": -1.849483609199524,
|
608 |
+
"eval_logits/rejected": -1.888432502746582,
|
609 |
+
"eval_logps/chosen": -185.52931213378906,
|
610 |
+
"eval_logps/rejected": -248.10926818847656,
|
611 |
+
"eval_loss": 0.501145601272583,
|
612 |
+
"eval_rewards/accuracies": 0.75,
|
613 |
+
"eval_rewards/chosen": 2.704090118408203,
|
614 |
+
"eval_rewards/margins": 4.181191921234131,
|
615 |
+
"eval_rewards/rejected": -1.477101445198059,
|
616 |
+
"eval_runtime": 26.2437,
|
617 |
+
"eval_samples_per_second": 1.067,
|
618 |
+
"eval_steps_per_second": 1.067,
|
619 |
+
"step": 300
|
620 |
+
},
|
621 |
+
{
|
622 |
+
"epoch": 1.27,
|
623 |
+
"grad_norm": 327.02899169921875,
|
624 |
+
"learning_rate": 4.091228070175439e-06,
|
625 |
+
"logits/chosen": -1.8802036046981812,
|
626 |
+
"logits/rejected": -1.9435373544692993,
|
627 |
+
"logps/chosen": -164.2046661376953,
|
628 |
+
"logps/rejected": -269.22235107421875,
|
629 |
+
"loss": 0.4088,
|
630 |
+
"rewards/accuracies": 0.800000011920929,
|
631 |
+
"rewards/chosen": 3.3527073860168457,
|
632 |
+
"rewards/margins": 5.9436140060424805,
|
633 |
+
"rewards/rejected": -2.5909066200256348,
|
634 |
+
"step": 310
|
635 |
+
},
|
636 |
+
{
|
637 |
+
"epoch": 1.31,
|
638 |
+
"grad_norm": 6.565308094024658,
|
639 |
+
"learning_rate": 4.056140350877193e-06,
|
640 |
+
"logits/chosen": -1.9250032901763916,
|
641 |
+
"logits/rejected": -1.9860813617706299,
|
642 |
+
"logps/chosen": -197.4870147705078,
|
643 |
+
"logps/rejected": -207.25009155273438,
|
644 |
+
"loss": 0.4969,
|
645 |
+
"rewards/accuracies": 0.699999988079071,
|
646 |
+
"rewards/chosen": 0.32615354657173157,
|
647 |
+
"rewards/margins": 2.2872400283813477,
|
648 |
+
"rewards/rejected": -1.9610865116119385,
|
649 |
+
"step": 320
|
650 |
+
},
|
651 |
+
{
|
652 |
+
"epoch": 1.35,
|
653 |
+
"grad_norm": 33.34074020385742,
|
654 |
+
"learning_rate": 4.021052631578948e-06,
|
655 |
+
"logits/chosen": -1.8470687866210938,
|
656 |
+
"logits/rejected": -1.8675216436386108,
|
657 |
+
"logps/chosen": -117.5759506225586,
|
658 |
+
"logps/rejected": -151.83091735839844,
|
659 |
+
"loss": 0.1819,
|
660 |
+
"rewards/accuracies": 0.8999999761581421,
|
661 |
+
"rewards/chosen": 2.878262758255005,
|
662 |
+
"rewards/margins": 5.000906467437744,
|
663 |
+
"rewards/rejected": -2.12264347076416,
|
664 |
+
"step": 330
|
665 |
+
},
|
666 |
+
{
|
667 |
+
"epoch": 1.35,
|
668 |
+
"eval_logits/chosen": -1.8486573696136475,
|
669 |
+
"eval_logits/rejected": -1.8877869844436646,
|
670 |
+
"eval_logps/chosen": -185.54177856445312,
|
671 |
+
"eval_logps/rejected": -249.2687530517578,
|
672 |
+
"eval_loss": 0.478495329618454,
|
673 |
+
"eval_rewards/accuracies": 0.75,
|
674 |
+
"eval_rewards/chosen": 2.7003517150878906,
|
675 |
+
"eval_rewards/margins": 4.525294780731201,
|
676 |
+
"eval_rewards/rejected": -1.8249431848526,
|
677 |
+
"eval_runtime": 26.2362,
|
678 |
+
"eval_samples_per_second": 1.067,
|
679 |
+
"eval_steps_per_second": 1.067,
|
680 |
+
"step": 330
|
681 |
+
},
|
682 |
+
{
|
683 |
+
"epoch": 1.39,
|
684 |
+
"grad_norm": 0.6777727603912354,
|
685 |
+
"learning_rate": 3.985964912280702e-06,
|
686 |
+
"logits/chosen": -1.8584426641464233,
|
687 |
+
"logits/rejected": -1.884598970413208,
|
688 |
+
"logps/chosen": -78.69309997558594,
|
689 |
+
"logps/rejected": -126.11468505859375,
|
690 |
+
"loss": 0.3154,
|
691 |
+
"rewards/accuracies": 0.8999999761581421,
|
692 |
+
"rewards/chosen": 2.5596954822540283,
|
693 |
+
"rewards/margins": 4.089261054992676,
|
694 |
+
"rewards/rejected": -1.5295660495758057,
|
695 |
+
"step": 340
|
696 |
+
},
|
697 |
+
{
|
698 |
+
"epoch": 1.43,
|
699 |
+
"grad_norm": 45.252723693847656,
|
700 |
+
"learning_rate": 3.950877192982457e-06,
|
701 |
+
"logits/chosen": -1.8422210216522217,
|
702 |
+
"logits/rejected": -1.8741470575332642,
|
703 |
+
"logps/chosen": -198.17526245117188,
|
704 |
+
"logps/rejected": -179.26759338378906,
|
705 |
+
"loss": 0.2128,
|
706 |
+
"rewards/accuracies": 0.8999999761581421,
|
707 |
+
"rewards/chosen": 2.886838912963867,
|
708 |
+
"rewards/margins": 4.228543758392334,
|
709 |
+
"rewards/rejected": -1.3417048454284668,
|
710 |
+
"step": 350
|
711 |
+
},
|
712 |
+
{
|
713 |
+
"epoch": 1.47,
|
714 |
+
"grad_norm": 0.0007244591251946986,
|
715 |
+
"learning_rate": 3.9157894736842104e-06,
|
716 |
+
"logits/chosen": -1.8951257467269897,
|
717 |
+
"logits/rejected": -1.9684120416641235,
|
718 |
+
"logps/chosen": -123.7894515991211,
|
719 |
+
"logps/rejected": -278.1185607910156,
|
720 |
+
"loss": 0.0169,
|
721 |
+
"rewards/accuracies": 1.0,
|
722 |
+
"rewards/chosen": 3.9708969593048096,
|
723 |
+
"rewards/margins": 9.946398735046387,
|
724 |
+
"rewards/rejected": -5.975502014160156,
|
725 |
+
"step": 360
|
726 |
+
},
|
727 |
+
{
|
728 |
+
"epoch": 1.47,
|
729 |
+
"eval_logits/chosen": -1.8510137796401978,
|
730 |
+
"eval_logits/rejected": -1.8906936645507812,
|
731 |
+
"eval_logps/chosen": -185.66189575195312,
|
732 |
+
"eval_logps/rejected": -250.37806701660156,
|
733 |
+
"eval_loss": 0.4872412383556366,
|
734 |
+
"eval_rewards/accuracies": 0.75,
|
735 |
+
"eval_rewards/chosen": 2.664308547973633,
|
736 |
+
"eval_rewards/margins": 4.822035789489746,
|
737 |
+
"eval_rewards/rejected": -2.157727003097534,
|
738 |
+
"eval_runtime": 26.2354,
|
739 |
+
"eval_samples_per_second": 1.067,
|
740 |
+
"eval_steps_per_second": 1.067,
|
741 |
+
"step": 360
|
742 |
+
},
|
743 |
+
{
|
744 |
+
"epoch": 1.51,
|
745 |
+
"grad_norm": 0.001088384771719575,
|
746 |
+
"learning_rate": 3.880701754385965e-06,
|
747 |
+
"logits/chosen": -1.859140157699585,
|
748 |
+
"logits/rejected": -1.903721809387207,
|
749 |
+
"logps/chosen": -171.89047241210938,
|
750 |
+
"logps/rejected": -257.6295471191406,
|
751 |
+
"loss": 0.237,
|
752 |
+
"rewards/accuracies": 0.800000011920929,
|
753 |
+
"rewards/chosen": 2.257797956466675,
|
754 |
+
"rewards/margins": 7.423349857330322,
|
755 |
+
"rewards/rejected": -5.165551662445068,
|
756 |
+
"step": 370
|
757 |
+
},
|
758 |
+
{
|
759 |
+
"epoch": 1.55,
|
760 |
+
"grad_norm": 13.049817085266113,
|
761 |
+
"learning_rate": 3.84561403508772e-06,
|
762 |
+
"logits/chosen": -1.850102186203003,
|
763 |
+
"logits/rejected": -1.922501564025879,
|
764 |
+
"logps/chosen": -145.50692749023438,
|
765 |
+
"logps/rejected": -221.7799530029297,
|
766 |
+
"loss": 0.2906,
|
767 |
+
"rewards/accuracies": 0.8999999761581421,
|
768 |
+
"rewards/chosen": 1.9433931112289429,
|
769 |
+
"rewards/margins": 6.384174346923828,
|
770 |
+
"rewards/rejected": -4.440781593322754,
|
771 |
+
"step": 380
|
772 |
+
},
|
773 |
+
{
|
774 |
+
"epoch": 1.59,
|
775 |
+
"grad_norm": 0.006105802953243256,
|
776 |
+
"learning_rate": 3.810526315789474e-06,
|
777 |
+
"logits/chosen": -1.9277305603027344,
|
778 |
+
"logits/rejected": -1.9736566543579102,
|
779 |
+
"logps/chosen": -140.36146545410156,
|
780 |
+
"logps/rejected": -188.66836547851562,
|
781 |
+
"loss": 0.235,
|
782 |
+
"rewards/accuracies": 0.8999999761581421,
|
783 |
+
"rewards/chosen": 2.6781504154205322,
|
784 |
+
"rewards/margins": 5.825950622558594,
|
785 |
+
"rewards/rejected": -3.1477997303009033,
|
786 |
+
"step": 390
|
787 |
+
},
|
788 |
+
{
|
789 |
+
"epoch": 1.59,
|
790 |
+
"eval_logits/chosen": -1.8531960248947144,
|
791 |
+
"eval_logits/rejected": -1.8930160999298096,
|
792 |
+
"eval_logps/chosen": -185.68800354003906,
|
793 |
+
"eval_logps/rejected": -251.1302032470703,
|
794 |
+
"eval_loss": 0.48860952258110046,
|
795 |
+
"eval_rewards/accuracies": 0.75,
|
796 |
+
"eval_rewards/chosen": 2.656480550765991,
|
797 |
+
"eval_rewards/margins": 5.039853572845459,
|
798 |
+
"eval_rewards/rejected": -2.3833730220794678,
|
799 |
+
"eval_runtime": 26.2308,
|
800 |
+
"eval_samples_per_second": 1.067,
|
801 |
+
"eval_steps_per_second": 1.067,
|
802 |
+
"step": 390
|
803 |
+
},
|
804 |
+
{
|
805 |
+
"epoch": 1.63,
|
806 |
+
"grad_norm": 0.00030716226319782436,
|
807 |
+
"learning_rate": 3.7754385964912284e-06,
|
808 |
+
"logits/chosen": -1.8817369937896729,
|
809 |
+
"logits/rejected": -1.9124549627304077,
|
810 |
+
"logps/chosen": -171.40383911132812,
|
811 |
+
"logps/rejected": -198.42446899414062,
|
812 |
+
"loss": 0.4322,
|
813 |
+
"rewards/accuracies": 0.800000011920929,
|
814 |
+
"rewards/chosen": 1.219795823097229,
|
815 |
+
"rewards/margins": 6.402298927307129,
|
816 |
+
"rewards/rejected": -5.182502269744873,
|
817 |
+
"step": 400
|
818 |
+
},
|
819 |
+
{
|
820 |
+
"epoch": 1.67,
|
821 |
+
"grad_norm": 0.0007249619229696691,
|
822 |
+
"learning_rate": 3.7403508771929827e-06,
|
823 |
+
"logits/chosen": -1.8018410205841064,
|
824 |
+
"logits/rejected": -1.869511365890503,
|
825 |
+
"logps/chosen": -168.9734649658203,
|
826 |
+
"logps/rejected": -223.9024658203125,
|
827 |
+
"loss": 0.9332,
|
828 |
+
"rewards/accuracies": 0.699999988079071,
|
829 |
+
"rewards/chosen": 2.475252151489258,
|
830 |
+
"rewards/margins": 5.069212913513184,
|
831 |
+
"rewards/rejected": -2.593961238861084,
|
832 |
+
"step": 410
|
833 |
+
},
|
834 |
+
{
|
835 |
+
"epoch": 1.71,
|
836 |
+
"grad_norm": 46.79059982299805,
|
837 |
+
"learning_rate": 3.7052631578947374e-06,
|
838 |
+
"logits/chosen": -1.774857759475708,
|
839 |
+
"logits/rejected": -1.8125171661376953,
|
840 |
+
"logps/chosen": -140.12551879882812,
|
841 |
+
"logps/rejected": -208.7825927734375,
|
842 |
+
"loss": 0.7551,
|
843 |
+
"rewards/accuracies": 0.800000011920929,
|
844 |
+
"rewards/chosen": 2.244128704071045,
|
845 |
+
"rewards/margins": 6.2417426109313965,
|
846 |
+
"rewards/rejected": -3.997614622116089,
|
847 |
+
"step": 420
|
848 |
+
},
|
849 |
+
{
|
850 |
+
"epoch": 1.71,
|
851 |
+
"eval_logits/chosen": -1.8527201414108276,
|
852 |
+
"eval_logits/rejected": -1.892095923423767,
|
853 |
+
"eval_logps/chosen": -185.46646118164062,
|
854 |
+
"eval_logps/rejected": -251.00816345214844,
|
855 |
+
"eval_loss": 0.438032329082489,
|
856 |
+
"eval_rewards/accuracies": 0.75,
|
857 |
+
"eval_rewards/chosen": 2.7229368686676025,
|
858 |
+
"eval_rewards/margins": 5.069703102111816,
|
859 |
+
"eval_rewards/rejected": -2.346766233444214,
|
860 |
+
"eval_runtime": 26.2319,
|
861 |
+
"eval_samples_per_second": 1.067,
|
862 |
+
"eval_steps_per_second": 1.067,
|
863 |
+
"step": 420
|
864 |
+
},
|
865 |
+
{
|
866 |
+
"epoch": 1.76,
|
867 |
+
"grad_norm": 0.1679692417383194,
|
868 |
+
"learning_rate": 3.6701754385964917e-06,
|
869 |
+
"logits/chosen": -1.9766556024551392,
|
870 |
+
"logits/rejected": -1.967247724533081,
|
871 |
+
"logps/chosen": -181.25978088378906,
|
872 |
+
"logps/rejected": -174.0313262939453,
|
873 |
+
"loss": 0.5746,
|
874 |
+
"rewards/accuracies": 0.800000011920929,
|
875 |
+
"rewards/chosen": 2.037475347518921,
|
876 |
+
"rewards/margins": 4.661454200744629,
|
877 |
+
"rewards/rejected": -2.623978853225708,
|
878 |
+
"step": 430
|
879 |
+
},
|
880 |
+
{
|
881 |
+
"epoch": 1.8,
|
882 |
+
"grad_norm": 0.028290700167417526,
|
883 |
+
"learning_rate": 3.635087719298246e-06,
|
884 |
+
"logits/chosen": -1.7324810028076172,
|
885 |
+
"logits/rejected": -1.8381952047348022,
|
886 |
+
"logps/chosen": -118.8620376586914,
|
887 |
+
"logps/rejected": -249.5945587158203,
|
888 |
+
"loss": 0.1919,
|
889 |
+
"rewards/accuracies": 0.8999999761581421,
|
890 |
+
"rewards/chosen": 2.84242582321167,
|
891 |
+
"rewards/margins": 8.061222076416016,
|
892 |
+
"rewards/rejected": -5.218796730041504,
|
893 |
+
"step": 440
|
894 |
+
},
|
895 |
+
{
|
896 |
+
"epoch": 1.84,
|
897 |
+
"grad_norm": 3.435687303543091,
|
898 |
+
"learning_rate": 3.6000000000000003e-06,
|
899 |
+
"logits/chosen": -1.788468599319458,
|
900 |
+
"logits/rejected": -1.8786054849624634,
|
901 |
+
"logps/chosen": -127.3366928100586,
|
902 |
+
"logps/rejected": -223.4063262939453,
|
903 |
+
"loss": 0.134,
|
904 |
+
"rewards/accuracies": 1.0,
|
905 |
+
"rewards/chosen": 0.9258162379264832,
|
906 |
+
"rewards/margins": 6.352460861206055,
|
907 |
+
"rewards/rejected": -5.426644802093506,
|
908 |
+
"step": 450
|
909 |
+
},
|
910 |
+
{
|
911 |
+
"epoch": 1.84,
|
912 |
+
"eval_logits/chosen": -1.8530861139297485,
|
913 |
+
"eval_logits/rejected": -1.8925215005874634,
|
914 |
+
"eval_logps/chosen": -185.6543426513672,
|
915 |
+
"eval_logps/rejected": -251.7077178955078,
|
916 |
+
"eval_loss": 0.4383006691932678,
|
917 |
+
"eval_rewards/accuracies": 0.75,
|
918 |
+
"eval_rewards/chosen": 2.6665782928466797,
|
919 |
+
"eval_rewards/margins": 5.223209857940674,
|
920 |
+
"eval_rewards/rejected": -2.556631565093994,
|
921 |
+
"eval_runtime": 26.2389,
|
922 |
+
"eval_samples_per_second": 1.067,
|
923 |
+
"eval_steps_per_second": 1.067,
|
924 |
+
"step": 450
|
925 |
+
},
|
926 |
+
{
|
927 |
+
"epoch": 1.88,
|
928 |
+
"grad_norm": 0.2778886556625366,
|
929 |
+
"learning_rate": 3.5649122807017545e-06,
|
930 |
+
"logits/chosen": -1.8211870193481445,
|
931 |
+
"logits/rejected": -1.8562345504760742,
|
932 |
+
"logps/chosen": -90.34883117675781,
|
933 |
+
"logps/rejected": -171.23648071289062,
|
934 |
+
"loss": 0.1284,
|
935 |
+
"rewards/accuracies": 1.0,
|
936 |
+
"rewards/chosen": 3.756679058074951,
|
937 |
+
"rewards/margins": 7.474773406982422,
|
938 |
+
"rewards/rejected": -3.7180938720703125,
|
939 |
+
"step": 460
|
940 |
+
},
|
941 |
+
{
|
942 |
+
"epoch": 1.92,
|
943 |
+
"grad_norm": 0.00020280799071770161,
|
944 |
+
"learning_rate": 3.5298245614035093e-06,
|
945 |
+
"logits/chosen": -1.7977367639541626,
|
946 |
+
"logits/rejected": -1.8836923837661743,
|
947 |
+
"logps/chosen": -180.67724609375,
|
948 |
+
"logps/rejected": -215.39608764648438,
|
949 |
+
"loss": 0.5069,
|
950 |
+
"rewards/accuracies": 0.800000011920929,
|
951 |
+
"rewards/chosen": 2.3001179695129395,
|
952 |
+
"rewards/margins": 7.598956108093262,
|
953 |
+
"rewards/rejected": -5.298838138580322,
|
954 |
+
"step": 470
|
955 |
+
},
|
956 |
+
{
|
957 |
+
"epoch": 1.96,
|
958 |
+
"grad_norm": 0.0038736851420253515,
|
959 |
+
"learning_rate": 3.4947368421052635e-06,
|
960 |
+
"logits/chosen": -1.8841644525527954,
|
961 |
+
"logits/rejected": -1.986575722694397,
|
962 |
+
"logps/chosen": -189.42210388183594,
|
963 |
+
"logps/rejected": -368.7510070800781,
|
964 |
+
"loss": 0.0662,
|
965 |
+
"rewards/accuracies": 1.0,
|
966 |
+
"rewards/chosen": 2.894408702850342,
|
967 |
+
"rewards/margins": 9.953689575195312,
|
968 |
+
"rewards/rejected": -7.0592803955078125,
|
969 |
+
"step": 480
|
970 |
+
},
|
971 |
+
{
|
972 |
+
"epoch": 1.96,
|
973 |
+
"eval_logits/chosen": -1.8568991422653198,
|
974 |
+
"eval_logits/rejected": -1.896437406539917,
|
975 |
+
"eval_logps/chosen": -186.01429748535156,
|
976 |
+
"eval_logps/rejected": -252.9164276123047,
|
977 |
+
"eval_loss": 0.44481489062309265,
|
978 |
+
"eval_rewards/accuracies": 0.75,
|
979 |
+
"eval_rewards/chosen": 2.55859112739563,
|
980 |
+
"eval_rewards/margins": 5.477838039398193,
|
981 |
+
"eval_rewards/rejected": -2.9192464351654053,
|
982 |
+
"eval_runtime": 26.2387,
|
983 |
+
"eval_samples_per_second": 1.067,
|
984 |
+
"eval_steps_per_second": 1.067,
|
985 |
+
"step": 480
|
986 |
+
},
|
987 |
+
{
|
988 |
+
"epoch": 2.0,
|
989 |
+
"grad_norm": 0.00020024628611281514,
|
990 |
+
"learning_rate": 3.459649122807018e-06,
|
991 |
+
"logits/chosen": -1.7868010997772217,
|
992 |
+
"logits/rejected": -1.8775966167449951,
|
993 |
+
"logps/chosen": -168.01686096191406,
|
994 |
+
"logps/rejected": -268.81329345703125,
|
995 |
+
"loss": 0.2127,
|
996 |
+
"rewards/accuracies": 0.8999999761581421,
|
997 |
+
"rewards/chosen": 4.091753959655762,
|
998 |
+
"rewards/margins": 10.252196311950684,
|
999 |
+
"rewards/rejected": -6.160443305969238,
|
1000 |
+
"step": 490
|
1001 |
+
},
|
1002 |
+
{
|
1003 |
+
"epoch": 2.04,
|
1004 |
+
"grad_norm": 2.933464288711548,
|
1005 |
+
"learning_rate": 3.424561403508772e-06,
|
1006 |
+
"logits/chosen": -1.8647997379302979,
|
1007 |
+
"logits/rejected": -1.8972787857055664,
|
1008 |
+
"logps/chosen": -206.39712524414062,
|
1009 |
+
"logps/rejected": -255.3307647705078,
|
1010 |
+
"loss": 0.1551,
|
1011 |
+
"rewards/accuracies": 0.8999999761581421,
|
1012 |
+
"rewards/chosen": 3.2640254497528076,
|
1013 |
+
"rewards/margins": 8.8242826461792,
|
1014 |
+
"rewards/rejected": -5.560257434844971,
|
1015 |
+
"step": 500
|
1016 |
+
},
|
1017 |
+
{
|
1018 |
+
"epoch": 2.08,
|
1019 |
+
"grad_norm": 0.088715098798275,
|
1020 |
+
"learning_rate": 3.3894736842105264e-06,
|
1021 |
+
"logits/chosen": -1.8153291940689087,
|
1022 |
+
"logits/rejected": -1.907467246055603,
|
1023 |
+
"logps/chosen": -114.2393569946289,
|
1024 |
+
"logps/rejected": -245.9196319580078,
|
1025 |
+
"loss": 0.1093,
|
1026 |
+
"rewards/accuracies": 1.0,
|
1027 |
+
"rewards/chosen": 1.2050167322158813,
|
1028 |
+
"rewards/margins": 7.797829627990723,
|
1029 |
+
"rewards/rejected": -6.592813014984131,
|
1030 |
+
"step": 510
|
1031 |
+
},
|
1032 |
+
{
|
1033 |
+
"epoch": 2.08,
|
1034 |
+
"eval_logits/chosen": -1.8561475276947021,
|
1035 |
+
"eval_logits/rejected": -1.8955259323120117,
|
1036 |
+
"eval_logps/chosen": -186.13938903808594,
|
1037 |
+
"eval_logps/rejected": -253.4276580810547,
|
1038 |
+
"eval_loss": 0.4262312054634094,
|
1039 |
+
"eval_rewards/accuracies": 0.75,
|
1040 |
+
"eval_rewards/chosen": 2.521066904067993,
|
1041 |
+
"eval_rewards/margins": 5.59368371963501,
|
1042 |
+
"eval_rewards/rejected": -3.0726168155670166,
|
1043 |
+
"eval_runtime": 26.2357,
|
1044 |
+
"eval_samples_per_second": 1.067,
|
1045 |
+
"eval_steps_per_second": 1.067,
|
1046 |
+
"step": 510
|
1047 |
+
},
|
1048 |
+
{
|
1049 |
+
"epoch": 2.12,
|
1050 |
+
"grad_norm": 0.0037095127627253532,
|
1051 |
+
"learning_rate": 3.354385964912281e-06,
|
1052 |
+
"logits/chosen": -1.904370665550232,
|
1053 |
+
"logits/rejected": -1.9216587543487549,
|
1054 |
+
"logps/chosen": -173.86373901367188,
|
1055 |
+
"logps/rejected": -173.6374053955078,
|
1056 |
+
"loss": 0.1557,
|
1057 |
+
"rewards/accuracies": 0.8999999761581421,
|
1058 |
+
"rewards/chosen": 2.4744720458984375,
|
1059 |
+
"rewards/margins": 5.560853004455566,
|
1060 |
+
"rewards/rejected": -3.0863804817199707,
|
1061 |
+
"step": 520
|
1062 |
+
},
|
1063 |
+
{
|
1064 |
+
"epoch": 2.16,
|
1065 |
+
"grad_norm": 1.6992926532566344e-07,
|
1066 |
+
"learning_rate": 3.3192982456140354e-06,
|
1067 |
+
"logits/chosen": -1.7665284872055054,
|
1068 |
+
"logits/rejected": -1.859400987625122,
|
1069 |
+
"logps/chosen": -85.06950378417969,
|
1070 |
+
"logps/rejected": -232.3406524658203,
|
1071 |
+
"loss": 0.0056,
|
1072 |
+
"rewards/accuracies": 1.0,
|
1073 |
+
"rewards/chosen": 3.268359661102295,
|
1074 |
+
"rewards/margins": 9.698001861572266,
|
1075 |
+
"rewards/rejected": -6.4296417236328125,
|
1076 |
+
"step": 530
|
1077 |
+
},
|
1078 |
+
{
|
1079 |
+
"epoch": 2.2,
|
1080 |
+
"grad_norm": 0.00015174441796261817,
|
1081 |
+
"learning_rate": 3.2842105263157897e-06,
|
1082 |
+
"logits/chosen": -1.7254387140274048,
|
1083 |
+
"logits/rejected": -1.8300654888153076,
|
1084 |
+
"logps/chosen": -126.21636962890625,
|
1085 |
+
"logps/rejected": -236.91250610351562,
|
1086 |
+
"loss": 0.1557,
|
1087 |
+
"rewards/accuracies": 0.8999999761581421,
|
1088 |
+
"rewards/chosen": 2.9851126670837402,
|
1089 |
+
"rewards/margins": 9.359209060668945,
|
1090 |
+
"rewards/rejected": -6.374096870422363,
|
1091 |
+
"step": 540
|
1092 |
+
},
|
1093 |
+
{
|
1094 |
+
"epoch": 2.2,
|
1095 |
+
"eval_logits/chosen": -1.85663902759552,
|
1096 |
+
"eval_logits/rejected": -1.8965102434158325,
|
1097 |
+
"eval_logps/chosen": -186.6448974609375,
|
1098 |
+
"eval_logps/rejected": -254.58482360839844,
|
1099 |
+
"eval_loss": 0.4263899326324463,
|
1100 |
+
"eval_rewards/accuracies": 0.75,
|
1101 |
+
"eval_rewards/chosen": 2.369408369064331,
|
1102 |
+
"eval_rewards/margins": 5.789170265197754,
|
1103 |
+
"eval_rewards/rejected": -3.4197616577148438,
|
1104 |
+
"eval_runtime": 26.2541,
|
1105 |
+
"eval_samples_per_second": 1.066,
|
1106 |
+
"eval_steps_per_second": 1.066,
|
1107 |
+
"step": 540
|
1108 |
+
},
|
1109 |
+
{
|
1110 |
+
"epoch": 2.24,
|
1111 |
+
"grad_norm": 7.969775676727295,
|
1112 |
+
"learning_rate": 3.249122807017544e-06,
|
1113 |
+
"logits/chosen": -1.8786284923553467,
|
1114 |
+
"logits/rejected": -1.9631255865097046,
|
1115 |
+
"logps/chosen": -186.79092407226562,
|
1116 |
+
"logps/rejected": -295.82598876953125,
|
1117 |
+
"loss": 0.0625,
|
1118 |
+
"rewards/accuracies": 1.0,
|
1119 |
+
"rewards/chosen": 0.5664806365966797,
|
1120 |
+
"rewards/margins": 6.786992073059082,
|
1121 |
+
"rewards/rejected": -6.220510959625244,
|
1122 |
+
"step": 550
|
1123 |
+
},
|
1124 |
+
{
|
1125 |
+
"epoch": 2.29,
|
1126 |
+
"grad_norm": 41.887001037597656,
|
1127 |
+
"learning_rate": 3.2140350877192982e-06,
|
1128 |
+
"logits/chosen": -1.8324702978134155,
|
1129 |
+
"logits/rejected": -1.8570420742034912,
|
1130 |
+
"logps/chosen": -198.67893981933594,
|
1131 |
+
"logps/rejected": -203.76016235351562,
|
1132 |
+
"loss": 0.389,
|
1133 |
+
"rewards/accuracies": 0.8999999761581421,
|
1134 |
+
"rewards/chosen": 2.470458984375,
|
1135 |
+
"rewards/margins": 8.410517692565918,
|
1136 |
+
"rewards/rejected": -5.940059185028076,
|
1137 |
+
"step": 560
|
1138 |
+
},
|
1139 |
+
{
|
1140 |
+
"epoch": 2.33,
|
1141 |
+
"grad_norm": 0.022604621946811676,
|
1142 |
+
"learning_rate": 3.178947368421053e-06,
|
1143 |
+
"logits/chosen": -1.7625017166137695,
|
1144 |
+
"logits/rejected": -1.823395013809204,
|
1145 |
+
"logps/chosen": -145.4344024658203,
|
1146 |
+
"logps/rejected": -202.81698608398438,
|
1147 |
+
"loss": 0.0962,
|
1148 |
+
"rewards/accuracies": 0.8999999761581421,
|
1149 |
+
"rewards/chosen": 3.2409331798553467,
|
1150 |
+
"rewards/margins": 7.634140968322754,
|
1151 |
+
"rewards/rejected": -4.39320707321167,
|
1152 |
+
"step": 570
|
1153 |
+
},
|
1154 |
+
{
|
1155 |
+
"epoch": 2.33,
|
1156 |
+
"eval_logits/chosen": -1.8582419157028198,
|
1157 |
+
"eval_logits/rejected": -1.897825837135315,
|
1158 |
+
"eval_logps/chosen": -186.9964141845703,
|
1159 |
+
"eval_logps/rejected": -255.54441833496094,
|
1160 |
+
"eval_loss": 0.41822776198387146,
|
1161 |
+
"eval_rewards/accuracies": 0.75,
|
1162 |
+
"eval_rewards/chosen": 2.2639553546905518,
|
1163 |
+
"eval_rewards/margins": 5.971593379974365,
|
1164 |
+
"eval_rewards/rejected": -3.70763897895813,
|
1165 |
+
"eval_runtime": 26.2352,
|
1166 |
+
"eval_samples_per_second": 1.067,
|
1167 |
+
"eval_steps_per_second": 1.067,
|
1168 |
+
"step": 570
|
1169 |
+
},
|
1170 |
+
{
|
1171 |
+
"epoch": 2.37,
|
1172 |
+
"grad_norm": 0.0053796968422830105,
|
1173 |
+
"learning_rate": 3.1438596491228072e-06,
|
1174 |
+
"logits/chosen": -1.8564494848251343,
|
1175 |
+
"logits/rejected": -1.9009406566619873,
|
1176 |
+
"logps/chosen": -153.97903442382812,
|
1177 |
+
"logps/rejected": -182.94284057617188,
|
1178 |
+
"loss": 0.5733,
|
1179 |
+
"rewards/accuracies": 0.800000011920929,
|
1180 |
+
"rewards/chosen": 2.0778138637542725,
|
1181 |
+
"rewards/margins": 5.466890335083008,
|
1182 |
+
"rewards/rejected": -3.389075756072998,
|
1183 |
+
"step": 580
|
1184 |
+
},
|
1185 |
+
{
|
1186 |
+
"epoch": 2.41,
|
1187 |
+
"grad_norm": 5.523962020874023,
|
1188 |
+
"learning_rate": 3.1087719298245615e-06,
|
1189 |
+
"logits/chosen": -1.8511816263198853,
|
1190 |
+
"logits/rejected": -1.853109359741211,
|
1191 |
+
"logps/chosen": -182.23892211914062,
|
1192 |
+
"logps/rejected": -120.5928955078125,
|
1193 |
+
"loss": 0.4958,
|
1194 |
+
"rewards/accuracies": 0.800000011920929,
|
1195 |
+
"rewards/chosen": 3.659731388092041,
|
1196 |
+
"rewards/margins": 4.932890892028809,
|
1197 |
+
"rewards/rejected": -1.2731596231460571,
|
1198 |
+
"step": 590
|
1199 |
+
},
|
1200 |
+
{
|
1201 |
+
"epoch": 2.45,
|
1202 |
+
"grad_norm": 0.0013734139502048492,
|
1203 |
+
"learning_rate": 3.0736842105263158e-06,
|
1204 |
+
"logits/chosen": -1.7618497610092163,
|
1205 |
+
"logits/rejected": -1.8786863088607788,
|
1206 |
+
"logps/chosen": -114.43098449707031,
|
1207 |
+
"logps/rejected": -229.5573272705078,
|
1208 |
+
"loss": 0.0437,
|
1209 |
+
"rewards/accuracies": 1.0,
|
1210 |
+
"rewards/chosen": 2.8845677375793457,
|
1211 |
+
"rewards/margins": 9.519918441772461,
|
1212 |
+
"rewards/rejected": -6.635350227355957,
|
1213 |
+
"step": 600
|
1214 |
+
},
|
1215 |
+
{
|
1216 |
+
"epoch": 2.45,
|
1217 |
+
"eval_logits/chosen": -1.8533600568771362,
|
1218 |
+
"eval_logits/rejected": -1.8932533264160156,
|
1219 |
+
"eval_logps/chosen": -187.0037384033203,
|
1220 |
+
"eval_logps/rejected": -255.77125549316406,
|
1221 |
+
"eval_loss": 0.3824370801448822,
|
1222 |
+
"eval_rewards/accuracies": 0.75,
|
1223 |
+
"eval_rewards/chosen": 2.2617576122283936,
|
1224 |
+
"eval_rewards/margins": 6.037452220916748,
|
1225 |
+
"eval_rewards/rejected": -3.7756943702697754,
|
1226 |
+
"eval_runtime": 26.2326,
|
1227 |
+
"eval_samples_per_second": 1.067,
|
1228 |
+
"eval_steps_per_second": 1.067,
|
1229 |
+
"step": 600
|
1230 |
+
},
|
1231 |
+
{
|
1232 |
+
"epoch": 2.49,
|
1233 |
+
"grad_norm": 3.847404956817627,
|
1234 |
+
"learning_rate": 3.03859649122807e-06,
|
1235 |
+
"logits/chosen": -1.9108350276947021,
|
1236 |
+
"logits/rejected": -1.9136813879013062,
|
1237 |
+
"logps/chosen": -147.76666259765625,
|
1238 |
+
"logps/rejected": -151.54275512695312,
|
1239 |
+
"loss": 0.1939,
|
1240 |
+
"rewards/accuracies": 0.8999999761581421,
|
1241 |
+
"rewards/chosen": 2.7343037128448486,
|
1242 |
+
"rewards/margins": 7.082615852355957,
|
1243 |
+
"rewards/rejected": -4.348311424255371,
|
1244 |
+
"step": 610
|
1245 |
+
},
|
1246 |
+
{
|
1247 |
+
"epoch": 2.53,
|
1248 |
+
"grad_norm": 6.025547172328061e-09,
|
1249 |
+
"learning_rate": 3.003508771929825e-06,
|
1250 |
+
"logits/chosen": -1.9030263423919678,
|
1251 |
+
"logits/rejected": -1.964647889137268,
|
1252 |
+
"logps/chosen": -157.2318115234375,
|
1253 |
+
"logps/rejected": -335.1673583984375,
|
1254 |
+
"loss": 0.1472,
|
1255 |
+
"rewards/accuracies": 0.8999999761581421,
|
1256 |
+
"rewards/chosen": 0.17058677971363068,
|
1257 |
+
"rewards/margins": 10.464014053344727,
|
1258 |
+
"rewards/rejected": -10.293428421020508,
|
1259 |
+
"step": 620
|
1260 |
+
},
|
1261 |
+
{
|
1262 |
+
"epoch": 2.57,
|
1263 |
+
"grad_norm": 0.7363545298576355,
|
1264 |
+
"learning_rate": 2.9684210526315795e-06,
|
1265 |
+
"logits/chosen": -1.8167072534561157,
|
1266 |
+
"logits/rejected": -1.9096784591674805,
|
1267 |
+
"logps/chosen": -128.2840118408203,
|
1268 |
+
"logps/rejected": -278.52862548828125,
|
1269 |
+
"loss": 0.0278,
|
1270 |
+
"rewards/accuracies": 1.0,
|
1271 |
+
"rewards/chosen": 2.6698997020721436,
|
1272 |
+
"rewards/margins": 10.468381881713867,
|
1273 |
+
"rewards/rejected": -7.798482418060303,
|
1274 |
+
"step": 630
|
1275 |
+
},
|
1276 |
+
{
|
1277 |
+
"epoch": 2.57,
|
1278 |
+
"eval_logits/chosen": -1.8536275625228882,
|
1279 |
+
"eval_logits/rejected": -1.8932437896728516,
|
1280 |
+
"eval_logps/chosen": -186.7086181640625,
|
1281 |
+
"eval_logps/rejected": -255.70458984375,
|
1282 |
+
"eval_loss": 0.3570733964443207,
|
1283 |
+
"eval_rewards/accuracies": 0.8571428656578064,
|
1284 |
+
"eval_rewards/chosen": 2.3502955436706543,
|
1285 |
+
"eval_rewards/margins": 6.105983257293701,
|
1286 |
+
"eval_rewards/rejected": -3.755687713623047,
|
1287 |
+
"eval_runtime": 26.2519,
|
1288 |
+
"eval_samples_per_second": 1.067,
|
1289 |
+
"eval_steps_per_second": 1.067,
|
1290 |
+
"step": 630
|
1291 |
+
},
|
1292 |
+
{
|
1293 |
+
"epoch": 2.61,
|
1294 |
+
"grad_norm": 19.793140411376953,
|
1295 |
+
"learning_rate": 2.9333333333333338e-06,
|
1296 |
+
"logits/chosen": -1.9250261783599854,
|
1297 |
+
"logits/rejected": -1.969919204711914,
|
1298 |
+
"logps/chosen": -169.236572265625,
|
1299 |
+
"logps/rejected": -216.4037322998047,
|
1300 |
+
"loss": 0.3144,
|
1301 |
+
"rewards/accuracies": 0.800000011920929,
|
1302 |
+
"rewards/chosen": 1.5923322439193726,
|
1303 |
+
"rewards/margins": 8.081656455993652,
|
1304 |
+
"rewards/rejected": -6.48932409286499,
|
1305 |
+
"step": 640
|
1306 |
+
},
|
1307 |
+
{
|
1308 |
+
"epoch": 2.65,
|
1309 |
+
"grad_norm": 0.0,
|
1310 |
+
"learning_rate": 2.898245614035088e-06,
|
1311 |
+
"logits/chosen": -1.840488076210022,
|
1312 |
+
"logits/rejected": -1.9099540710449219,
|
1313 |
+
"logps/chosen": -188.36514282226562,
|
1314 |
+
"logps/rejected": -266.2427978515625,
|
1315 |
+
"loss": 0.0455,
|
1316 |
+
"rewards/accuracies": 1.0,
|
1317 |
+
"rewards/chosen": 3.294522762298584,
|
1318 |
+
"rewards/margins": 9.861984252929688,
|
1319 |
+
"rewards/rejected": -6.5674614906311035,
|
1320 |
+
"step": 650
|
1321 |
+
},
|
1322 |
+
{
|
1323 |
+
"epoch": 2.69,
|
1324 |
+
"grad_norm": 350.0583801269531,
|
1325 |
+
"learning_rate": 2.8631578947368423e-06,
|
1326 |
+
"logits/chosen": -1.875605583190918,
|
1327 |
+
"logits/rejected": -1.9353086948394775,
|
1328 |
+
"logps/chosen": -128.51104736328125,
|
1329 |
+
"logps/rejected": -191.3477325439453,
|
1330 |
+
"loss": 0.2399,
|
1331 |
+
"rewards/accuracies": 0.8999999761581421,
|
1332 |
+
"rewards/chosen": 3.1921534538269043,
|
1333 |
+
"rewards/margins": 7.449667453765869,
|
1334 |
+
"rewards/rejected": -4.257513999938965,
|
1335 |
+
"step": 660
|
1336 |
+
},
|
1337 |
+
{
|
1338 |
+
"epoch": 2.69,
|
1339 |
+
"eval_logits/chosen": -1.851164698600769,
|
1340 |
+
"eval_logits/rejected": -1.8909271955490112,
|
1341 |
+
"eval_logps/chosen": -186.86781311035156,
|
1342 |
+
"eval_logps/rejected": -256.2709655761719,
|
1343 |
+
"eval_loss": 0.3312711715698242,
|
1344 |
+
"eval_rewards/accuracies": 0.8571428656578064,
|
1345 |
+
"eval_rewards/chosen": 2.3025403022766113,
|
1346 |
+
"eval_rewards/margins": 6.228146553039551,
|
1347 |
+
"eval_rewards/rejected": -3.9256062507629395,
|
1348 |
+
"eval_runtime": 26.2337,
|
1349 |
+
"eval_samples_per_second": 1.067,
|
1350 |
+
"eval_steps_per_second": 1.067,
|
1351 |
+
"step": 660
|
1352 |
+
},
|
1353 |
+
{
|
1354 |
+
"epoch": 2.73,
|
1355 |
+
"grad_norm": 0.0,
|
1356 |
+
"learning_rate": 2.828070175438597e-06,
|
1357 |
+
"logits/chosen": -1.7718509435653687,
|
1358 |
+
"logits/rejected": -1.8662782907485962,
|
1359 |
+
"logps/chosen": -100.0200424194336,
|
1360 |
+
"logps/rejected": -176.11337280273438,
|
1361 |
+
"loss": 0.0636,
|
1362 |
+
"rewards/accuracies": 1.0,
|
1363 |
+
"rewards/chosen": 2.2194771766662598,
|
1364 |
+
"rewards/margins": 7.908571720123291,
|
1365 |
+
"rewards/rejected": -5.689094543457031,
|
1366 |
+
"step": 670
|
1367 |
+
},
|
1368 |
+
{
|
1369 |
+
"epoch": 2.78,
|
1370 |
+
"grad_norm": 9.849269554251805e-05,
|
1371 |
+
"learning_rate": 2.7929824561403513e-06,
|
1372 |
+
"logits/chosen": -1.8379380702972412,
|
1373 |
+
"logits/rejected": -1.9032844305038452,
|
1374 |
+
"logps/chosen": -147.157958984375,
|
1375 |
+
"logps/rejected": -311.342529296875,
|
1376 |
+
"loss": 0.0081,
|
1377 |
+
"rewards/accuracies": 1.0,
|
1378 |
+
"rewards/chosen": 1.7090831995010376,
|
1379 |
+
"rewards/margins": 9.63591480255127,
|
1380 |
+
"rewards/rejected": -7.926831245422363,
|
1381 |
+
"step": 680
|
1382 |
+
},
|
1383 |
+
{
|
1384 |
+
"epoch": 2.82,
|
1385 |
+
"grad_norm": 5.133792877197266,
|
1386 |
+
"learning_rate": 2.7578947368421056e-06,
|
1387 |
+
"logits/chosen": -1.8350670337677002,
|
1388 |
+
"logits/rejected": -1.9039970636367798,
|
1389 |
+
"logps/chosen": -160.75314331054688,
|
1390 |
+
"logps/rejected": -247.68008422851562,
|
1391 |
+
"loss": 0.039,
|
1392 |
+
"rewards/accuracies": 1.0,
|
1393 |
+
"rewards/chosen": 4.146931171417236,
|
1394 |
+
"rewards/margins": 11.529654502868652,
|
1395 |
+
"rewards/rejected": -7.382723808288574,
|
1396 |
+
"step": 690
|
1397 |
+
},
|
1398 |
+
{
|
1399 |
+
"epoch": 2.82,
|
1400 |
+
"eval_logits/chosen": -1.8510338068008423,
|
1401 |
+
"eval_logits/rejected": -1.8906004428863525,
|
1402 |
+
"eval_logps/chosen": -187.16348266601562,
|
1403 |
+
"eval_logps/rejected": -257.069091796875,
|
1404 |
+
"eval_loss": 0.3131271302700043,
|
1405 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1406 |
+
"eval_rewards/chosen": 2.213836193084717,
|
1407 |
+
"eval_rewards/margins": 6.378880977630615,
|
1408 |
+
"eval_rewards/rejected": -4.165044784545898,
|
1409 |
+
"eval_runtime": 26.2333,
|
1410 |
+
"eval_samples_per_second": 1.067,
|
1411 |
+
"eval_steps_per_second": 1.067,
|
1412 |
+
"step": 690
|
1413 |
+
},
|
1414 |
+
{
|
1415 |
+
"epoch": 2.86,
|
1416 |
+
"grad_norm": 0.001812128466553986,
|
1417 |
+
"learning_rate": 2.72280701754386e-06,
|
1418 |
+
"logits/chosen": -1.888392448425293,
|
1419 |
+
"logits/rejected": -1.9376449584960938,
|
1420 |
+
"logps/chosen": -123.4749755859375,
|
1421 |
+
"logps/rejected": -221.9653778076172,
|
1422 |
+
"loss": 0.1581,
|
1423 |
+
"rewards/accuracies": 0.8999999761581421,
|
1424 |
+
"rewards/chosen": 3.1250205039978027,
|
1425 |
+
"rewards/margins": 9.35824966430664,
|
1426 |
+
"rewards/rejected": -6.23322868347168,
|
1427 |
+
"step": 700
|
1428 |
+
},
|
1429 |
+
{
|
1430 |
+
"epoch": 2.9,
|
1431 |
+
"grad_norm": 0.0007696045213378966,
|
1432 |
+
"learning_rate": 2.687719298245614e-06,
|
1433 |
+
"logits/chosen": -1.7869523763656616,
|
1434 |
+
"logits/rejected": -1.9070037603378296,
|
1435 |
+
"logps/chosen": -116.8504409790039,
|
1436 |
+
"logps/rejected": -300.51806640625,
|
1437 |
+
"loss": 0.1832,
|
1438 |
+
"rewards/accuracies": 0.8999999761581421,
|
1439 |
+
"rewards/chosen": 2.3490161895751953,
|
1440 |
+
"rewards/margins": 10.493500709533691,
|
1441 |
+
"rewards/rejected": -8.144485473632812,
|
1442 |
+
"step": 710
|
1443 |
+
},
|
1444 |
+
{
|
1445 |
+
"epoch": 2.94,
|
1446 |
+
"grad_norm": 83.29313659667969,
|
1447 |
+
"learning_rate": 2.652631578947369e-06,
|
1448 |
+
"logits/chosen": -1.8838169574737549,
|
1449 |
+
"logits/rejected": -1.9118610620498657,
|
1450 |
+
"logps/chosen": -104.75813293457031,
|
1451 |
+
"logps/rejected": -167.18580627441406,
|
1452 |
+
"loss": 0.3389,
|
1453 |
+
"rewards/accuracies": 0.8999999761581421,
|
1454 |
+
"rewards/chosen": 2.7628464698791504,
|
1455 |
+
"rewards/margins": 6.317173957824707,
|
1456 |
+
"rewards/rejected": -3.5543277263641357,
|
1457 |
+
"step": 720
|
1458 |
+
},
|
1459 |
+
{
|
1460 |
+
"epoch": 2.94,
|
1461 |
+
"eval_logits/chosen": -1.8480366468429565,
|
1462 |
+
"eval_logits/rejected": -1.8872734308242798,
|
1463 |
+
"eval_logps/chosen": -187.00791931152344,
|
1464 |
+
"eval_logps/rejected": -257.2390441894531,
|
1465 |
+
"eval_loss": 0.27630648016929626,
|
1466 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1467 |
+
"eval_rewards/chosen": 2.2605104446411133,
|
1468 |
+
"eval_rewards/margins": 6.476548194885254,
|
1469 |
+
"eval_rewards/rejected": -4.216037750244141,
|
1470 |
+
"eval_runtime": 26.23,
|
1471 |
+
"eval_samples_per_second": 1.067,
|
1472 |
+
"eval_steps_per_second": 1.067,
|
1473 |
+
"step": 720
|
1474 |
+
},
|
1475 |
+
{
|
1476 |
+
"epoch": 2.98,
|
1477 |
+
"grad_norm": 0.030682619661092758,
|
1478 |
+
"learning_rate": 2.617543859649123e-06,
|
1479 |
+
"logits/chosen": -1.8667634725570679,
|
1480 |
+
"logits/rejected": -1.8890842199325562,
|
1481 |
+
"logps/chosen": -197.29690551757812,
|
1482 |
+
"logps/rejected": -219.12332153320312,
|
1483 |
+
"loss": 0.617,
|
1484 |
+
"rewards/accuracies": 0.800000011920929,
|
1485 |
+
"rewards/chosen": -0.587865948677063,
|
1486 |
+
"rewards/margins": 5.496161937713623,
|
1487 |
+
"rewards/rejected": -6.0840277671813965,
|
1488 |
+
"step": 730
|
1489 |
+
},
|
1490 |
+
{
|
1491 |
+
"epoch": 3.02,
|
1492 |
+
"grad_norm": 0.005717798601835966,
|
1493 |
+
"learning_rate": 2.5824561403508774e-06,
|
1494 |
+
"logits/chosen": -1.8484325408935547,
|
1495 |
+
"logits/rejected": -1.8692123889923096,
|
1496 |
+
"logps/chosen": -102.5673599243164,
|
1497 |
+
"logps/rejected": -141.8745880126953,
|
1498 |
+
"loss": 0.044,
|
1499 |
+
"rewards/accuracies": 1.0,
|
1500 |
+
"rewards/chosen": 2.209364175796509,
|
1501 |
+
"rewards/margins": 8.334315299987793,
|
1502 |
+
"rewards/rejected": -6.124951362609863,
|
1503 |
+
"step": 740
|
1504 |
+
},
|
1505 |
+
{
|
1506 |
+
"epoch": 3.06,
|
1507 |
+
"grad_norm": 0.0006890746881254017,
|
1508 |
+
"learning_rate": 2.5473684210526317e-06,
|
1509 |
+
"logits/chosen": -1.865378737449646,
|
1510 |
+
"logits/rejected": -1.9292535781860352,
|
1511 |
+
"logps/chosen": -103.72840881347656,
|
1512 |
+
"logps/rejected": -203.67276000976562,
|
1513 |
+
"loss": 0.0154,
|
1514 |
+
"rewards/accuracies": 1.0,
|
1515 |
+
"rewards/chosen": 3.9861044883728027,
|
1516 |
+
"rewards/margins": 8.630514144897461,
|
1517 |
+
"rewards/rejected": -4.644408702850342,
|
1518 |
+
"step": 750
|
1519 |
+
},
|
1520 |
+
{
|
1521 |
+
"epoch": 3.06,
|
1522 |
+
"eval_logits/chosen": -1.846961259841919,
|
1523 |
+
"eval_logits/rejected": -1.886168122291565,
|
1524 |
+
"eval_logps/chosen": -187.0342254638672,
|
1525 |
+
"eval_logps/rejected": -257.52471923828125,
|
1526 |
+
"eval_loss": 0.2703961730003357,
|
1527 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1528 |
+
"eval_rewards/chosen": 2.252613067626953,
|
1529 |
+
"eval_rewards/margins": 6.554351806640625,
|
1530 |
+
"eval_rewards/rejected": -4.301738739013672,
|
1531 |
+
"eval_runtime": 26.236,
|
1532 |
+
"eval_samples_per_second": 1.067,
|
1533 |
+
"eval_steps_per_second": 1.067,
|
1534 |
+
"step": 750
|
1535 |
+
},
|
1536 |
+
{
|
1537 |
+
"epoch": 3.1,
|
1538 |
+
"grad_norm": 0.0,
|
1539 |
+
"learning_rate": 2.512280701754386e-06,
|
1540 |
+
"logits/chosen": -1.921618103981018,
|
1541 |
+
"logits/rejected": -1.9239141941070557,
|
1542 |
+
"logps/chosen": -248.826416015625,
|
1543 |
+
"logps/rejected": -182.67323303222656,
|
1544 |
+
"loss": 0.5385,
|
1545 |
+
"rewards/accuracies": 0.699999988079071,
|
1546 |
+
"rewards/chosen": -0.7813287377357483,
|
1547 |
+
"rewards/margins": 4.614243507385254,
|
1548 |
+
"rewards/rejected": -5.395572185516357,
|
1549 |
+
"step": 760
|
1550 |
+
},
|
1551 |
+
{
|
1552 |
+
"epoch": 3.14,
|
1553 |
+
"grad_norm": 20.82038688659668,
|
1554 |
+
"learning_rate": 2.4771929824561407e-06,
|
1555 |
+
"logits/chosen": -1.8723421096801758,
|
1556 |
+
"logits/rejected": -1.9096221923828125,
|
1557 |
+
"logps/chosen": -144.3488311767578,
|
1558 |
+
"logps/rejected": -200.52149963378906,
|
1559 |
+
"loss": 0.0473,
|
1560 |
+
"rewards/accuracies": 1.0,
|
1561 |
+
"rewards/chosen": 0.6265076994895935,
|
1562 |
+
"rewards/margins": 5.623400688171387,
|
1563 |
+
"rewards/rejected": -4.996891975402832,
|
1564 |
+
"step": 770
|
1565 |
+
},
|
1566 |
+
{
|
1567 |
+
"epoch": 3.18,
|
1568 |
+
"grad_norm": 0.0,
|
1569 |
+
"learning_rate": 2.442105263157895e-06,
|
1570 |
+
"logits/chosen": -1.967357873916626,
|
1571 |
+
"logits/rejected": -1.995941162109375,
|
1572 |
+
"logps/chosen": -169.40643310546875,
|
1573 |
+
"logps/rejected": -225.83267211914062,
|
1574 |
+
"loss": 0.021,
|
1575 |
+
"rewards/accuracies": 1.0,
|
1576 |
+
"rewards/chosen": 3.36498761177063,
|
1577 |
+
"rewards/margins": 8.846434593200684,
|
1578 |
+
"rewards/rejected": -5.481447219848633,
|
1579 |
+
"step": 780
|
1580 |
+
},
|
1581 |
+
{
|
1582 |
+
"epoch": 3.18,
|
1583 |
+
"eval_logits/chosen": -1.84483802318573,
|
1584 |
+
"eval_logits/rejected": -1.8837740421295166,
|
1585 |
+
"eval_logps/chosen": -187.02699279785156,
|
1586 |
+
"eval_logps/rejected": -257.6650085449219,
|
1587 |
+
"eval_loss": 0.24223719537258148,
|
1588 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1589 |
+
"eval_rewards/chosen": 2.254775285720825,
|
1590 |
+
"eval_rewards/margins": 6.598587989807129,
|
1591 |
+
"eval_rewards/rejected": -4.343813419342041,
|
1592 |
+
"eval_runtime": 26.2524,
|
1593 |
+
"eval_samples_per_second": 1.067,
|
1594 |
+
"eval_steps_per_second": 1.067,
|
1595 |
+
"step": 780
|
1596 |
+
},
|
1597 |
+
{
|
1598 |
+
"epoch": 3.22,
|
1599 |
+
"grad_norm": 0.3396748900413513,
|
1600 |
+
"learning_rate": 2.4070175438596493e-06,
|
1601 |
+
"logits/chosen": -1.795336127281189,
|
1602 |
+
"logits/rejected": -1.9104626178741455,
|
1603 |
+
"logps/chosen": -127.96372985839844,
|
1604 |
+
"logps/rejected": -321.76751708984375,
|
1605 |
+
"loss": 0.0057,
|
1606 |
+
"rewards/accuracies": 1.0,
|
1607 |
+
"rewards/chosen": 1.3614065647125244,
|
1608 |
+
"rewards/margins": 11.285585403442383,
|
1609 |
+
"rewards/rejected": -9.924179077148438,
|
1610 |
+
"step": 790
|
1611 |
+
},
|
1612 |
+
{
|
1613 |
+
"epoch": 3.27,
|
1614 |
+
"grad_norm": 89.25263214111328,
|
1615 |
+
"learning_rate": 2.3719298245614036e-06,
|
1616 |
+
"logits/chosen": -1.7844104766845703,
|
1617 |
+
"logits/rejected": -1.835338830947876,
|
1618 |
+
"logps/chosen": -138.7672119140625,
|
1619 |
+
"logps/rejected": -212.5260772705078,
|
1620 |
+
"loss": 0.099,
|
1621 |
+
"rewards/accuracies": 0.8999999761581421,
|
1622 |
+
"rewards/chosen": 3.200477123260498,
|
1623 |
+
"rewards/margins": 9.856096267700195,
|
1624 |
+
"rewards/rejected": -6.655619144439697,
|
1625 |
+
"step": 800
|
1626 |
+
},
|
1627 |
+
{
|
1628 |
+
"epoch": 3.31,
|
1629 |
+
"grad_norm": 0.0014283270575106144,
|
1630 |
+
"learning_rate": 2.3368421052631583e-06,
|
1631 |
+
"logits/chosen": -1.8158670663833618,
|
1632 |
+
"logits/rejected": -1.896251916885376,
|
1633 |
+
"logps/chosen": -144.15419006347656,
|
1634 |
+
"logps/rejected": -283.3548583984375,
|
1635 |
+
"loss": 0.0614,
|
1636 |
+
"rewards/accuracies": 1.0,
|
1637 |
+
"rewards/chosen": 3.173565149307251,
|
1638 |
+
"rewards/margins": 12.303560256958008,
|
1639 |
+
"rewards/rejected": -9.129995346069336,
|
1640 |
+
"step": 810
|
1641 |
+
},
|
1642 |
+
{
|
1643 |
+
"epoch": 3.31,
|
1644 |
+
"eval_logits/chosen": -1.8417214155197144,
|
1645 |
+
"eval_logits/rejected": -1.8804857730865479,
|
1646 |
+
"eval_logps/chosen": -187.0991668701172,
|
1647 |
+
"eval_logps/rejected": -258.0171813964844,
|
1648 |
+
"eval_loss": 0.2144124060869217,
|
1649 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1650 |
+
"eval_rewards/chosen": 2.2331290245056152,
|
1651 |
+
"eval_rewards/margins": 6.682602882385254,
|
1652 |
+
"eval_rewards/rejected": -4.449473857879639,
|
1653 |
+
"eval_runtime": 26.2397,
|
1654 |
+
"eval_samples_per_second": 1.067,
|
1655 |
+
"eval_steps_per_second": 1.067,
|
1656 |
+
"step": 810
|
1657 |
+
},
|
1658 |
+
{
|
1659 |
+
"epoch": 3.35,
|
1660 |
+
"grad_norm": 8.28421070764307e-06,
|
1661 |
+
"learning_rate": 2.3017543859649126e-06,
|
1662 |
+
"logits/chosen": -1.7904294729232788,
|
1663 |
+
"logits/rejected": -1.7954041957855225,
|
1664 |
+
"logps/chosen": -136.8679962158203,
|
1665 |
+
"logps/rejected": -173.52316284179688,
|
1666 |
+
"loss": 0.0305,
|
1667 |
+
"rewards/accuracies": 1.0,
|
1668 |
+
"rewards/chosen": 3.9618313312530518,
|
1669 |
+
"rewards/margins": 9.469245910644531,
|
1670 |
+
"rewards/rejected": -5.507414817810059,
|
1671 |
+
"step": 820
|
1672 |
+
},
|
1673 |
+
{
|
1674 |
+
"epoch": 3.39,
|
1675 |
+
"grad_norm": 0.00032607687171548605,
|
1676 |
+
"learning_rate": 2.266666666666667e-06,
|
1677 |
+
"logits/chosen": -1.7678247690200806,
|
1678 |
+
"logits/rejected": -1.8045036792755127,
|
1679 |
+
"logps/chosen": -97.9904556274414,
|
1680 |
+
"logps/rejected": -158.86538696289062,
|
1681 |
+
"loss": 0.0114,
|
1682 |
+
"rewards/accuracies": 1.0,
|
1683 |
+
"rewards/chosen": 3.4605960845947266,
|
1684 |
+
"rewards/margins": 9.087323188781738,
|
1685 |
+
"rewards/rejected": -5.626727104187012,
|
1686 |
+
"step": 830
|
1687 |
+
},
|
1688 |
+
{
|
1689 |
+
"epoch": 3.43,
|
1690 |
+
"grad_norm": 9.237614631652832,
|
1691 |
+
"learning_rate": 2.231578947368421e-06,
|
1692 |
+
"logits/chosen": -1.8179492950439453,
|
1693 |
+
"logits/rejected": -1.8643207550048828,
|
1694 |
+
"logps/chosen": -158.73011779785156,
|
1695 |
+
"logps/rejected": -175.10726928710938,
|
1696 |
+
"loss": 0.0529,
|
1697 |
+
"rewards/accuracies": 1.0,
|
1698 |
+
"rewards/chosen": 1.9895429611206055,
|
1699 |
+
"rewards/margins": 7.000767707824707,
|
1700 |
+
"rewards/rejected": -5.011223793029785,
|
1701 |
+
"step": 840
|
1702 |
+
},
|
1703 |
+
{
|
1704 |
+
"epoch": 3.43,
|
1705 |
+
"eval_logits/chosen": -1.842317819595337,
|
1706 |
+
"eval_logits/rejected": -1.8808995485305786,
|
1707 |
+
"eval_logps/chosen": -187.3555145263672,
|
1708 |
+
"eval_logps/rejected": -258.76568603515625,
|
1709 |
+
"eval_loss": 0.2120712250471115,
|
1710 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1711 |
+
"eval_rewards/chosen": 2.1562187671661377,
|
1712 |
+
"eval_rewards/margins": 6.830246925354004,
|
1713 |
+
"eval_rewards/rejected": -4.674027919769287,
|
1714 |
+
"eval_runtime": 26.238,
|
1715 |
+
"eval_samples_per_second": 1.067,
|
1716 |
+
"eval_steps_per_second": 1.067,
|
1717 |
+
"step": 840
|
1718 |
+
},
|
1719 |
+
{
|
1720 |
+
"epoch": 3.47,
|
1721 |
+
"grad_norm": 5.9564328694250435e-05,
|
1722 |
+
"learning_rate": 2.1964912280701754e-06,
|
1723 |
+
"logits/chosen": -1.782867431640625,
|
1724 |
+
"logits/rejected": -1.8579784631729126,
|
1725 |
+
"logps/chosen": -105.4168472290039,
|
1726 |
+
"logps/rejected": -259.2267761230469,
|
1727 |
+
"loss": 0.258,
|
1728 |
+
"rewards/accuracies": 0.8999999761581421,
|
1729 |
+
"rewards/chosen": 4.022895336151123,
|
1730 |
+
"rewards/margins": 12.175846099853516,
|
1731 |
+
"rewards/rejected": -8.152950286865234,
|
1732 |
+
"step": 850
|
1733 |
+
},
|
1734 |
+
{
|
1735 |
+
"epoch": 3.51,
|
1736 |
+
"grad_norm": 1.3973094610264525e-05,
|
1737 |
+
"learning_rate": 2.16140350877193e-06,
|
1738 |
+
"logits/chosen": -1.7417545318603516,
|
1739 |
+
"logits/rejected": -1.8234918117523193,
|
1740 |
+
"logps/chosen": -115.1645736694336,
|
1741 |
+
"logps/rejected": -248.0410614013672,
|
1742 |
+
"loss": 0.039,
|
1743 |
+
"rewards/accuracies": 1.0,
|
1744 |
+
"rewards/chosen": 3.356029987335205,
|
1745 |
+
"rewards/margins": 12.46558952331543,
|
1746 |
+
"rewards/rejected": -9.109560012817383,
|
1747 |
+
"step": 860
|
1748 |
+
},
|
1749 |
+
{
|
1750 |
+
"epoch": 3.55,
|
1751 |
+
"grad_norm": 0.005180311389267445,
|
1752 |
+
"learning_rate": 2.1263157894736844e-06,
|
1753 |
+
"logits/chosen": -1.8123457431793213,
|
1754 |
+
"logits/rejected": -1.9022102355957031,
|
1755 |
+
"logps/chosen": -128.5855255126953,
|
1756 |
+
"logps/rejected": -246.685791015625,
|
1757 |
+
"loss": 0.001,
|
1758 |
+
"rewards/accuracies": 1.0,
|
1759 |
+
"rewards/chosen": 2.4856154918670654,
|
1760 |
+
"rewards/margins": 12.966238021850586,
|
1761 |
+
"rewards/rejected": -10.480623245239258,
|
1762 |
+
"step": 870
|
1763 |
+
},
|
1764 |
+
{
|
1765 |
+
"epoch": 3.55,
|
1766 |
+
"eval_logits/chosen": -1.8409830331802368,
|
1767 |
+
"eval_logits/rejected": -1.87991464138031,
|
1768 |
+
"eval_logps/chosen": -187.5316925048828,
|
1769 |
+
"eval_logps/rejected": -259.3367614746094,
|
1770 |
+
"eval_loss": 0.2091565877199173,
|
1771 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1772 |
+
"eval_rewards/chosen": 2.1033718585968018,
|
1773 |
+
"eval_rewards/margins": 6.94872522354126,
|
1774 |
+
"eval_rewards/rejected": -4.845353126525879,
|
1775 |
+
"eval_runtime": 26.2322,
|
1776 |
+
"eval_samples_per_second": 1.067,
|
1777 |
+
"eval_steps_per_second": 1.067,
|
1778 |
+
"step": 870
|
1779 |
+
},
|
1780 |
+
{
|
1781 |
+
"epoch": 3.59,
|
1782 |
+
"grad_norm": 0.3039734661579132,
|
1783 |
+
"learning_rate": 2.0912280701754387e-06,
|
1784 |
+
"logits/chosen": -1.7311649322509766,
|
1785 |
+
"logits/rejected": -1.8156330585479736,
|
1786 |
+
"logps/chosen": -146.54473876953125,
|
1787 |
+
"logps/rejected": -300.264892578125,
|
1788 |
+
"loss": 0.0179,
|
1789 |
+
"rewards/accuracies": 1.0,
|
1790 |
+
"rewards/chosen": 3.358633041381836,
|
1791 |
+
"rewards/margins": 11.333723068237305,
|
1792 |
+
"rewards/rejected": -7.975089073181152,
|
1793 |
+
"step": 880
|
1794 |
+
},
|
1795 |
+
{
|
1796 |
+
"epoch": 3.63,
|
1797 |
+
"grad_norm": 0.8097418546676636,
|
1798 |
+
"learning_rate": 2.0561403508771934e-06,
|
1799 |
+
"logits/chosen": -1.8806817531585693,
|
1800 |
+
"logits/rejected": -1.922952651977539,
|
1801 |
+
"logps/chosen": -126.9049301147461,
|
1802 |
+
"logps/rejected": -184.9642333984375,
|
1803 |
+
"loss": 0.0541,
|
1804 |
+
"rewards/accuracies": 1.0,
|
1805 |
+
"rewards/chosen": 1.0604941844940186,
|
1806 |
+
"rewards/margins": 6.54372501373291,
|
1807 |
+
"rewards/rejected": -5.4832305908203125,
|
1808 |
+
"step": 890
|
1809 |
+
},
|
1810 |
+
{
|
1811 |
+
"epoch": 3.67,
|
1812 |
+
"grad_norm": 0.025914475321769714,
|
1813 |
+
"learning_rate": 2.0210526315789477e-06,
|
1814 |
+
"logits/chosen": -1.8227866888046265,
|
1815 |
+
"logits/rejected": -1.9303566217422485,
|
1816 |
+
"logps/chosen": -167.85279846191406,
|
1817 |
+
"logps/rejected": -311.27655029296875,
|
1818 |
+
"loss": 0.0284,
|
1819 |
+
"rewards/accuracies": 1.0,
|
1820 |
+
"rewards/chosen": 2.1594152450561523,
|
1821 |
+
"rewards/margins": 9.647980690002441,
|
1822 |
+
"rewards/rejected": -7.488564968109131,
|
1823 |
+
"step": 900
|
1824 |
+
},
|
1825 |
+
{
|
1826 |
+
"epoch": 3.67,
|
1827 |
+
"eval_logits/chosen": -1.8366106748580933,
|
1828 |
+
"eval_logits/rejected": -1.8760128021240234,
|
1829 |
+
"eval_logps/chosen": -187.9383544921875,
|
1830 |
+
"eval_logps/rejected": -260.31500244140625,
|
1831 |
+
"eval_loss": 0.20062340795993805,
|
1832 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1833 |
+
"eval_rewards/chosen": 1.9813728332519531,
|
1834 |
+
"eval_rewards/margins": 7.1201934814453125,
|
1835 |
+
"eval_rewards/rejected": -5.138820171356201,
|
1836 |
+
"eval_runtime": 26.2437,
|
1837 |
+
"eval_samples_per_second": 1.067,
|
1838 |
+
"eval_steps_per_second": 1.067,
|
1839 |
+
"step": 900
|
1840 |
+
},
|
1841 |
+
{
|
1842 |
+
"epoch": 3.71,
|
1843 |
+
"grad_norm": 0.02055520936846733,
|
1844 |
+
"learning_rate": 1.985964912280702e-06,
|
1845 |
+
"logits/chosen": -1.8666080236434937,
|
1846 |
+
"logits/rejected": -1.9068281650543213,
|
1847 |
+
"logps/chosen": -171.9093475341797,
|
1848 |
+
"logps/rejected": -251.93643188476562,
|
1849 |
+
"loss": 0.2622,
|
1850 |
+
"rewards/accuracies": 0.800000011920929,
|
1851 |
+
"rewards/chosen": 0.18651972711086273,
|
1852 |
+
"rewards/margins": 8.190787315368652,
|
1853 |
+
"rewards/rejected": -8.004267692565918,
|
1854 |
+
"step": 910
|
1855 |
+
},
|
1856 |
+
{
|
1857 |
+
"epoch": 3.76,
|
1858 |
+
"grad_norm": 197.38717651367188,
|
1859 |
+
"learning_rate": 1.9508771929824562e-06,
|
1860 |
+
"logits/chosen": -1.8426361083984375,
|
1861 |
+
"logits/rejected": -1.9008222818374634,
|
1862 |
+
"logps/chosen": -150.75900268554688,
|
1863 |
+
"logps/rejected": -251.9907989501953,
|
1864 |
+
"loss": 0.0599,
|
1865 |
+
"rewards/accuracies": 1.0,
|
1866 |
+
"rewards/chosen": 0.7380223870277405,
|
1867 |
+
"rewards/margins": 9.459989547729492,
|
1868 |
+
"rewards/rejected": -8.721967697143555,
|
1869 |
+
"step": 920
|
1870 |
+
},
|
1871 |
+
{
|
1872 |
+
"epoch": 3.8,
|
1873 |
+
"grad_norm": 0.0,
|
1874 |
+
"learning_rate": 1.9157894736842105e-06,
|
1875 |
+
"logits/chosen": -1.8063271045684814,
|
1876 |
+
"logits/rejected": -1.8660167455673218,
|
1877 |
+
"logps/chosen": -198.50723266601562,
|
1878 |
+
"logps/rejected": -222.7967529296875,
|
1879 |
+
"loss": 0.0744,
|
1880 |
+
"rewards/accuracies": 0.8999999761581421,
|
1881 |
+
"rewards/chosen": 1.713536262512207,
|
1882 |
+
"rewards/margins": 9.678624153137207,
|
1883 |
+
"rewards/rejected": -7.965087890625,
|
1884 |
+
"step": 930
|
1885 |
+
},
|
1886 |
+
{
|
1887 |
+
"epoch": 3.8,
|
1888 |
+
"eval_logits/chosen": -1.833928108215332,
|
1889 |
+
"eval_logits/rejected": -1.873318076133728,
|
1890 |
+
"eval_logps/chosen": -188.06385803222656,
|
1891 |
+
"eval_logps/rejected": -260.6358337402344,
|
1892 |
+
"eval_loss": 0.1813022941350937,
|
1893 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1894 |
+
"eval_rewards/chosen": 1.943717122077942,
|
1895 |
+
"eval_rewards/margins": 7.178780555725098,
|
1896 |
+
"eval_rewards/rejected": -5.2350640296936035,
|
1897 |
+
"eval_runtime": 26.2336,
|
1898 |
+
"eval_samples_per_second": 1.067,
|
1899 |
+
"eval_steps_per_second": 1.067,
|
1900 |
+
"step": 930
|
1901 |
+
},
|
1902 |
+
{
|
1903 |
+
"epoch": 3.84,
|
1904 |
+
"grad_norm": 4.8926853679631677e-08,
|
1905 |
+
"learning_rate": 1.8807017543859652e-06,
|
1906 |
+
"logits/chosen": -1.8202307224273682,
|
1907 |
+
"logits/rejected": -1.9166450500488281,
|
1908 |
+
"logps/chosen": -112.94303131103516,
|
1909 |
+
"logps/rejected": -244.4494171142578,
|
1910 |
+
"loss": 0.3721,
|
1911 |
+
"rewards/accuracies": 0.8999999761581421,
|
1912 |
+
"rewards/chosen": 2.255749225616455,
|
1913 |
+
"rewards/margins": 10.898755073547363,
|
1914 |
+
"rewards/rejected": -8.643006324768066,
|
1915 |
+
"step": 940
|
1916 |
+
},
|
1917 |
+
{
|
1918 |
+
"epoch": 3.88,
|
1919 |
+
"grad_norm": 5.456500053405762,
|
1920 |
+
"learning_rate": 1.8456140350877193e-06,
|
1921 |
+
"logits/chosen": -1.9108766317367554,
|
1922 |
+
"logits/rejected": -1.9368479251861572,
|
1923 |
+
"logps/chosen": -132.4052734375,
|
1924 |
+
"logps/rejected": -179.675537109375,
|
1925 |
+
"loss": 0.0257,
|
1926 |
+
"rewards/accuracies": 1.0,
|
1927 |
+
"rewards/chosen": 1.554168701171875,
|
1928 |
+
"rewards/margins": 6.65416955947876,
|
1929 |
+
"rewards/rejected": -5.100000858306885,
|
1930 |
+
"step": 950
|
1931 |
+
},
|
1932 |
+
{
|
1933 |
+
"epoch": 3.92,
|
1934 |
+
"grad_norm": 9.124670028686523,
|
1935 |
+
"learning_rate": 1.810526315789474e-06,
|
1936 |
+
"logits/chosen": -1.7729440927505493,
|
1937 |
+
"logits/rejected": -1.8252880573272705,
|
1938 |
+
"logps/chosen": -139.13653564453125,
|
1939 |
+
"logps/rejected": -193.011962890625,
|
1940 |
+
"loss": 0.091,
|
1941 |
+
"rewards/accuracies": 1.0,
|
1942 |
+
"rewards/chosen": 3.6216557025909424,
|
1943 |
+
"rewards/margins": 8.979948043823242,
|
1944 |
+
"rewards/rejected": -5.3582916259765625,
|
1945 |
+
"step": 960
|
1946 |
+
},
|
1947 |
+
{
|
1948 |
+
"epoch": 3.92,
|
1949 |
+
"eval_logits/chosen": -1.8312675952911377,
|
1950 |
+
"eval_logits/rejected": -1.8707492351531982,
|
1951 |
+
"eval_logps/chosen": -188.43190002441406,
|
1952 |
+
"eval_logps/rejected": -261.29730224609375,
|
1953 |
+
"eval_loss": 0.17215047776699066,
|
1954 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1955 |
+
"eval_rewards/chosen": 1.8333097696304321,
|
1956 |
+
"eval_rewards/margins": 7.266810894012451,
|
1957 |
+
"eval_rewards/rejected": -5.433501243591309,
|
1958 |
+
"eval_runtime": 26.2333,
|
1959 |
+
"eval_samples_per_second": 1.067,
|
1960 |
+
"eval_steps_per_second": 1.067,
|
1961 |
+
"step": 960
|
1962 |
+
},
|
1963 |
+
{
|
1964 |
+
"epoch": 3.96,
|
1965 |
+
"grad_norm": 1.8921099353974569e-06,
|
1966 |
+
"learning_rate": 1.7754385964912283e-06,
|
1967 |
+
"logits/chosen": -1.8464349508285522,
|
1968 |
+
"logits/rejected": -1.8927011489868164,
|
1969 |
+
"logps/chosen": -172.9676971435547,
|
1970 |
+
"logps/rejected": -224.15780639648438,
|
1971 |
+
"loss": 0.0795,
|
1972 |
+
"rewards/accuracies": 1.0,
|
1973 |
+
"rewards/chosen": 2.649458169937134,
|
1974 |
+
"rewards/margins": 9.076958656311035,
|
1975 |
+
"rewards/rejected": -6.4275007247924805,
|
1976 |
+
"step": 970
|
1977 |
+
},
|
1978 |
+
{
|
1979 |
+
"epoch": 4.0,
|
1980 |
+
"grad_norm": 318.0891418457031,
|
1981 |
+
"learning_rate": 1.7403508771929826e-06,
|
1982 |
+
"logits/chosen": -1.8083856105804443,
|
1983 |
+
"logits/rejected": -1.8901259899139404,
|
1984 |
+
"logps/chosen": -209.17422485351562,
|
1985 |
+
"logps/rejected": -284.4140625,
|
1986 |
+
"loss": 0.0583,
|
1987 |
+
"rewards/accuracies": 1.0,
|
1988 |
+
"rewards/chosen": 2.1195216178894043,
|
1989 |
+
"rewards/margins": 12.128747940063477,
|
1990 |
+
"rewards/rejected": -10.00922679901123,
|
1991 |
+
"step": 980
|
1992 |
+
},
|
1993 |
+
{
|
1994 |
+
"epoch": 4.04,
|
1995 |
+
"grad_norm": 13.0223388671875,
|
1996 |
+
"learning_rate": 1.705263157894737e-06,
|
1997 |
+
"logits/chosen": -1.847835898399353,
|
1998 |
+
"logits/rejected": -1.868665337562561,
|
1999 |
+
"logps/chosen": -149.8837127685547,
|
2000 |
+
"logps/rejected": -161.1732940673828,
|
2001 |
+
"loss": 0.3504,
|
2002 |
+
"rewards/accuracies": 0.8999999761581421,
|
2003 |
+
"rewards/chosen": 1.0685189962387085,
|
2004 |
+
"rewards/margins": 6.018086910247803,
|
2005 |
+
"rewards/rejected": -4.9495673179626465,
|
2006 |
+
"step": 990
|
2007 |
+
},
|
2008 |
+
{
|
2009 |
+
"epoch": 4.04,
|
2010 |
+
"eval_logits/chosen": -1.8279262781143188,
|
2011 |
+
"eval_logits/rejected": -1.8672196865081787,
|
2012 |
+
"eval_logps/chosen": -188.3168487548828,
|
2013 |
+
"eval_logps/rejected": -261.04876708984375,
|
2014 |
+
"eval_loss": 0.14871637523174286,
|
2015 |
+
"eval_rewards/accuracies": 0.9285714030265808,
|
2016 |
+
"eval_rewards/chosen": 1.8678233623504639,
|
2017 |
+
"eval_rewards/margins": 7.22676944732666,
|
2018 |
+
"eval_rewards/rejected": -5.358946800231934,
|
2019 |
+
"eval_runtime": 26.2331,
|
2020 |
+
"eval_samples_per_second": 1.067,
|
2021 |
+
"eval_steps_per_second": 1.067,
|
2022 |
+
"step": 990
|
2023 |
+
},
|
2024 |
+
{
|
2025 |
+
"epoch": 4.08,
|
2026 |
+
"grad_norm": 0.6881242394447327,
|
2027 |
+
"learning_rate": 1.6701754385964914e-06,
|
2028 |
+
"logits/chosen": -1.9393212795257568,
|
2029 |
+
"logits/rejected": -1.9376901388168335,
|
2030 |
+
"logps/chosen": -155.8722381591797,
|
2031 |
+
"logps/rejected": -191.2816925048828,
|
2032 |
+
"loss": 0.0102,
|
2033 |
+
"rewards/accuracies": 1.0,
|
2034 |
+
"rewards/chosen": 2.3404548168182373,
|
2035 |
+
"rewards/margins": 7.390922546386719,
|
2036 |
+
"rewards/rejected": -5.0504679679870605,
|
2037 |
+
"step": 1000
|
2038 |
+
},
|
2039 |
+
{
|
2040 |
+
"epoch": 4.12,
|
2041 |
+
"grad_norm": 0.0015760493697598577,
|
2042 |
+
"learning_rate": 1.6350877192982456e-06,
|
2043 |
+
"logits/chosen": -1.7732092142105103,
|
2044 |
+
"logits/rejected": -1.8348764181137085,
|
2045 |
+
"logps/chosen": -204.78118896484375,
|
2046 |
+
"logps/rejected": -285.19171142578125,
|
2047 |
+
"loss": 0.0728,
|
2048 |
+
"rewards/accuracies": 0.8999999761581421,
|
2049 |
+
"rewards/chosen": 1.1964905261993408,
|
2050 |
+
"rewards/margins": 11.535087585449219,
|
2051 |
+
"rewards/rejected": -10.338597297668457,
|
2052 |
+
"step": 1010
|
2053 |
+
},
|
2054 |
+
{
|
2055 |
+
"epoch": 4.16,
|
2056 |
+
"grad_norm": 5.320696345734177e-07,
|
2057 |
+
"learning_rate": 1.6000000000000001e-06,
|
2058 |
+
"logits/chosen": -1.7163060903549194,
|
2059 |
+
"logits/rejected": -1.8586156368255615,
|
2060 |
+
"logps/chosen": -123.50263977050781,
|
2061 |
+
"logps/rejected": -320.79693603515625,
|
2062 |
+
"loss": 0.0071,
|
2063 |
+
"rewards/accuracies": 1.0,
|
2064 |
+
"rewards/chosen": 0.7505568265914917,
|
2065 |
+
"rewards/margins": 10.981633186340332,
|
2066 |
+
"rewards/rejected": -10.231074333190918,
|
2067 |
+
"step": 1020
|
2068 |
+
},
|
2069 |
+
{
|
2070 |
+
"epoch": 4.16,
|
2071 |
+
"eval_logits/chosen": -1.824275016784668,
|
2072 |
+
"eval_logits/rejected": -1.8636810779571533,
|
2073 |
+
"eval_logps/chosen": -188.54676818847656,
|
2074 |
+
"eval_logps/rejected": -261.5804748535156,
|
2075 |
+
"eval_loss": 0.14029009640216827,
|
2076 |
+
"eval_rewards/accuracies": 0.9285714030265808,
|
2077 |
+
"eval_rewards/chosen": 1.7988510131835938,
|
2078 |
+
"eval_rewards/margins": 7.317309856414795,
|
2079 |
+
"eval_rewards/rejected": -5.518458843231201,
|
2080 |
+
"eval_runtime": 26.2364,
|
2081 |
+
"eval_samples_per_second": 1.067,
|
2082 |
+
"eval_steps_per_second": 1.067,
|
2083 |
+
"step": 1020
|
2084 |
+
},
|
2085 |
+
{
|
2086 |
+
"epoch": 4.2,
|
2087 |
+
"grad_norm": 0.002488874364644289,
|
2088 |
+
"learning_rate": 1.5649122807017544e-06,
|
2089 |
+
"logits/chosen": -1.7597904205322266,
|
2090 |
+
"logits/rejected": -1.8473920822143555,
|
2091 |
+
"logps/chosen": -165.49359130859375,
|
2092 |
+
"logps/rejected": -273.9131774902344,
|
2093 |
+
"loss": 0.0071,
|
2094 |
+
"rewards/accuracies": 1.0,
|
2095 |
+
"rewards/chosen": 1.4215660095214844,
|
2096 |
+
"rewards/margins": 10.477560043334961,
|
2097 |
+
"rewards/rejected": -9.055994033813477,
|
2098 |
+
"step": 1030
|
2099 |
+
},
|
2100 |
+
{
|
2101 |
+
"epoch": 4.24,
|
2102 |
+
"grad_norm": 0.0017074132338166237,
|
2103 |
+
"learning_rate": 1.529824561403509e-06,
|
2104 |
+
"logits/chosen": -1.7763557434082031,
|
2105 |
+
"logits/rejected": -1.8936008214950562,
|
2106 |
+
"logps/chosen": -133.30006408691406,
|
2107 |
+
"logps/rejected": -336.57427978515625,
|
2108 |
+
"loss": 0.018,
|
2109 |
+
"rewards/accuracies": 1.0,
|
2110 |
+
"rewards/chosen": 4.363164901733398,
|
2111 |
+
"rewards/margins": 14.957908630371094,
|
2112 |
+
"rewards/rejected": -10.594743728637695,
|
2113 |
+
"step": 1040
|
2114 |
+
},
|
2115 |
+
{
|
2116 |
+
"epoch": 4.29,
|
2117 |
+
"grad_norm": 66.57980346679688,
|
2118 |
+
"learning_rate": 1.4947368421052632e-06,
|
2119 |
+
"logits/chosen": -1.835628867149353,
|
2120 |
+
"logits/rejected": -1.8976163864135742,
|
2121 |
+
"logps/chosen": -177.20309448242188,
|
2122 |
+
"logps/rejected": -287.8599548339844,
|
2123 |
+
"loss": 0.0131,
|
2124 |
+
"rewards/accuracies": 1.0,
|
2125 |
+
"rewards/chosen": 2.568208694458008,
|
2126 |
+
"rewards/margins": 11.586195945739746,
|
2127 |
+
"rewards/rejected": -9.017987251281738,
|
2128 |
+
"step": 1050
|
2129 |
+
},
|
2130 |
+
{
|
2131 |
+
"epoch": 4.29,
|
2132 |
+
"eval_logits/chosen": -1.8221967220306396,
|
2133 |
+
"eval_logits/rejected": -1.8615671396255493,
|
2134 |
+
"eval_logps/chosen": -188.5262451171875,
|
2135 |
+
"eval_logps/rejected": -261.6840515136719,
|
2136 |
+
"eval_loss": 0.13124893605709076,
|
2137 |
+
"eval_rewards/accuracies": 0.9285714030265808,
|
2138 |
+
"eval_rewards/chosen": 1.805002212524414,
|
2139 |
+
"eval_rewards/margins": 7.354526042938232,
|
2140 |
+
"eval_rewards/rejected": -5.549524784088135,
|
2141 |
+
"eval_runtime": 26.2328,
|
2142 |
+
"eval_samples_per_second": 1.067,
|
2143 |
+
"eval_steps_per_second": 1.067,
|
2144 |
+
"step": 1050
|
2145 |
+
},
|
2146 |
+
{
|
2147 |
+
"epoch": 4.33,
|
2148 |
+
"grad_norm": 0.027719292789697647,
|
2149 |
+
"learning_rate": 1.4596491228070175e-06,
|
2150 |
+
"logits/chosen": -1.7560068368911743,
|
2151 |
+
"logits/rejected": -1.8672195672988892,
|
2152 |
+
"logps/chosen": -189.530029296875,
|
2153 |
+
"logps/rejected": -330.6167907714844,
|
2154 |
+
"loss": 0.003,
|
2155 |
+
"rewards/accuracies": 1.0,
|
2156 |
+
"rewards/chosen": 1.0810093879699707,
|
2157 |
+
"rewards/margins": 9.271745681762695,
|
2158 |
+
"rewards/rejected": -8.19073486328125,
|
2159 |
+
"step": 1060
|
2160 |
+
},
|
2161 |
+
{
|
2162 |
+
"epoch": 4.37,
|
2163 |
+
"grad_norm": 0.0,
|
2164 |
+
"learning_rate": 1.4245614035087722e-06,
|
2165 |
+
"logits/chosen": -1.8792110681533813,
|
2166 |
+
"logits/rejected": -1.9164355993270874,
|
2167 |
+
"logps/chosen": -212.42788696289062,
|
2168 |
+
"logps/rejected": -238.2472686767578,
|
2169 |
+
"loss": 0.1334,
|
2170 |
+
"rewards/accuracies": 0.8999999761581421,
|
2171 |
+
"rewards/chosen": 1.0076087713241577,
|
2172 |
+
"rewards/margins": 9.041505813598633,
|
2173 |
+
"rewards/rejected": -8.033895492553711,
|
2174 |
+
"step": 1070
|
2175 |
+
},
|
2176 |
+
{
|
2177 |
+
"epoch": 4.41,
|
2178 |
+
"grad_norm": 3.609489340306027e-06,
|
2179 |
+
"learning_rate": 1.3894736842105263e-06,
|
2180 |
+
"logits/chosen": -1.8213157653808594,
|
2181 |
+
"logits/rejected": -1.870356798171997,
|
2182 |
+
"logps/chosen": -80.74689483642578,
|
2183 |
+
"logps/rejected": -158.96109008789062,
|
2184 |
+
"loss": 0.0868,
|
2185 |
+
"rewards/accuracies": 1.0,
|
2186 |
+
"rewards/chosen": 2.509101629257202,
|
2187 |
+
"rewards/margins": 7.621730804443359,
|
2188 |
+
"rewards/rejected": -5.11262845993042,
|
2189 |
+
"step": 1080
|
2190 |
+
},
|
2191 |
+
{
|
2192 |
+
"epoch": 4.41,
|
2193 |
+
"eval_logits/chosen": -1.81948721408844,
|
2194 |
+
"eval_logits/rejected": -1.858702540397644,
|
2195 |
+
"eval_logps/chosen": -188.66748046875,
|
2196 |
+
"eval_logps/rejected": -261.94708251953125,
|
2197 |
+
"eval_loss": 0.1210150271654129,
|
2198 |
+
"eval_rewards/accuracies": 0.9285714030265808,
|
2199 |
+
"eval_rewards/chosen": 1.76263427734375,
|
2200 |
+
"eval_rewards/margins": 7.39107608795166,
|
2201 |
+
"eval_rewards/rejected": -5.62844181060791,
|
2202 |
+
"eval_runtime": 26.2393,
|
2203 |
+
"eval_samples_per_second": 1.067,
|
2204 |
+
"eval_steps_per_second": 1.067,
|
2205 |
+
"step": 1080
|
2206 |
+
},
|
2207 |
+
{
|
2208 |
+
"epoch": 4.45,
|
2209 |
+
"grad_norm": 0.7020970582962036,
|
2210 |
+
"learning_rate": 1.354385964912281e-06,
|
2211 |
+
"logits/chosen": -1.8757814168930054,
|
2212 |
+
"logits/rejected": -1.9657201766967773,
|
2213 |
+
"logps/chosen": -117.080810546875,
|
2214 |
+
"logps/rejected": -241.55252075195312,
|
2215 |
+
"loss": 0.0046,
|
2216 |
+
"rewards/accuracies": 1.0,
|
2217 |
+
"rewards/chosen": 1.6267999410629272,
|
2218 |
+
"rewards/margins": 10.32630729675293,
|
2219 |
+
"rewards/rejected": -8.699506759643555,
|
2220 |
+
"step": 1090
|
2221 |
+
},
|
2222 |
+
{
|
2223 |
+
"epoch": 4.49,
|
2224 |
+
"grad_norm": 0.00016405929636675864,
|
2225 |
+
"learning_rate": 1.3192982456140353e-06,
|
2226 |
+
"logits/chosen": -1.84481680393219,
|
2227 |
+
"logits/rejected": -1.8962256908416748,
|
2228 |
+
"logps/chosen": -176.97457885742188,
|
2229 |
+
"logps/rejected": -248.0132598876953,
|
2230 |
+
"loss": 0.0261,
|
2231 |
+
"rewards/accuracies": 1.0,
|
2232 |
+
"rewards/chosen": 1.7715377807617188,
|
2233 |
+
"rewards/margins": 11.10531234741211,
|
2234 |
+
"rewards/rejected": -9.333776473999023,
|
2235 |
+
"step": 1100
|
2236 |
+
},
|
2237 |
+
{
|
2238 |
+
"epoch": 4.53,
|
2239 |
+
"grad_norm": 0.0024285567924380302,
|
2240 |
+
"learning_rate": 1.2842105263157895e-06,
|
2241 |
+
"logits/chosen": -1.9240976572036743,
|
2242 |
+
"logits/rejected": -1.9567447900772095,
|
2243 |
+
"logps/chosen": -148.41506958007812,
|
2244 |
+
"logps/rejected": -262.0478820800781,
|
2245 |
+
"loss": 0.0041,
|
2246 |
+
"rewards/accuracies": 1.0,
|
2247 |
+
"rewards/chosen": 2.377047538757324,
|
2248 |
+
"rewards/margins": 12.98823070526123,
|
2249 |
+
"rewards/rejected": -10.611183166503906,
|
2250 |
+
"step": 1110
|
2251 |
+
},
|
2252 |
+
{
|
2253 |
+
"epoch": 4.53,
|
2254 |
+
"eval_logits/chosen": -1.8172996044158936,
|
2255 |
+
"eval_logits/rejected": -1.8566371202468872,
|
2256 |
+
"eval_logps/chosen": -188.92127990722656,
|
2257 |
+
"eval_logps/rejected": -262.44561767578125,
|
2258 |
+
"eval_loss": 0.1205989271402359,
|
2259 |
+
"eval_rewards/accuracies": 0.9285714030265808,
|
2260 |
+
"eval_rewards/chosen": 1.686492681503296,
|
2261 |
+
"eval_rewards/margins": 7.464495658874512,
|
2262 |
+
"eval_rewards/rejected": -5.778003215789795,
|
2263 |
+
"eval_runtime": 26.2482,
|
2264 |
+
"eval_samples_per_second": 1.067,
|
2265 |
+
"eval_steps_per_second": 1.067,
|
2266 |
+
"step": 1110
|
2267 |
+
},
|
2268 |
+
{
|
2269 |
+
"epoch": 4.57,
|
2270 |
+
"grad_norm": 26.255502700805664,
|
2271 |
+
"learning_rate": 1.249122807017544e-06,
|
2272 |
+
"logits/chosen": -1.7752002477645874,
|
2273 |
+
"logits/rejected": -1.8067470788955688,
|
2274 |
+
"logps/chosen": -199.05264282226562,
|
2275 |
+
"logps/rejected": -198.15122985839844,
|
2276 |
+
"loss": 0.0289,
|
2277 |
+
"rewards/accuracies": 1.0,
|
2278 |
+
"rewards/chosen": 2.024845600128174,
|
2279 |
+
"rewards/margins": 10.319567680358887,
|
2280 |
+
"rewards/rejected": -8.294721603393555,
|
2281 |
+
"step": 1120
|
2282 |
+
},
|
2283 |
+
{
|
2284 |
+
"epoch": 4.61,
|
2285 |
+
"grad_norm": 170.23622131347656,
|
2286 |
+
"learning_rate": 1.2140350877192983e-06,
|
2287 |
+
"logits/chosen": -1.7735440731048584,
|
2288 |
+
"logits/rejected": -1.8578115701675415,
|
2289 |
+
"logps/chosen": -140.52511596679688,
|
2290 |
+
"logps/rejected": -257.21124267578125,
|
2291 |
+
"loss": 0.0478,
|
2292 |
+
"rewards/accuracies": 1.0,
|
2293 |
+
"rewards/chosen": 3.585864305496216,
|
2294 |
+
"rewards/margins": 12.626684188842773,
|
2295 |
+
"rewards/rejected": -9.04081916809082,
|
2296 |
+
"step": 1130
|
2297 |
+
},
|
2298 |
+
{
|
2299 |
+
"epoch": 4.65,
|
2300 |
+
"grad_norm": 0.09534689784049988,
|
2301 |
+
"learning_rate": 1.1789473684210526e-06,
|
2302 |
+
"logits/chosen": -1.8365780115127563,
|
2303 |
+
"logits/rejected": -1.846933364868164,
|
2304 |
+
"logps/chosen": -90.59309387207031,
|
2305 |
+
"logps/rejected": -117.09541320800781,
|
2306 |
+
"loss": 0.0107,
|
2307 |
+
"rewards/accuracies": 1.0,
|
2308 |
+
"rewards/chosen": 4.559220314025879,
|
2309 |
+
"rewards/margins": 10.236574172973633,
|
2310 |
+
"rewards/rejected": -5.677353382110596,
|
2311 |
+
"step": 1140
|
2312 |
+
},
|
2313 |
+
{
|
2314 |
+
"epoch": 4.65,
|
2315 |
+
"eval_logits/chosen": -1.8170740604400635,
|
2316 |
+
"eval_logits/rejected": -1.8563201427459717,
|
2317 |
+
"eval_logps/chosen": -189.08616638183594,
|
2318 |
+
"eval_logps/rejected": -262.8173828125,
|
2319 |
+
"eval_loss": 0.1178145483136177,
|
2320 |
+
"eval_rewards/accuracies": 0.9642857313156128,
|
2321 |
+
"eval_rewards/chosen": 1.6370302438735962,
|
2322 |
+
"eval_rewards/margins": 7.526562213897705,
|
2323 |
+
"eval_rewards/rejected": -5.88953161239624,
|
2324 |
+
"eval_runtime": 26.2301,
|
2325 |
+
"eval_samples_per_second": 1.067,
|
2326 |
+
"eval_steps_per_second": 1.067,
|
2327 |
+
"step": 1140
|
2328 |
+
},
|
2329 |
+
{
|
2330 |
+
"epoch": 4.69,
|
2331 |
+
"grad_norm": 0.0,
|
2332 |
+
"learning_rate": 1.143859649122807e-06,
|
2333 |
+
"logits/chosen": -1.7817316055297852,
|
2334 |
+
"logits/rejected": -1.8392887115478516,
|
2335 |
+
"logps/chosen": -162.9131317138672,
|
2336 |
+
"logps/rejected": -241.8101043701172,
|
2337 |
+
"loss": 0.0639,
|
2338 |
+
"rewards/accuracies": 1.0,
|
2339 |
+
"rewards/chosen": 1.5432535409927368,
|
2340 |
+
"rewards/margins": 9.659387588500977,
|
2341 |
+
"rewards/rejected": -8.116134643554688,
|
2342 |
+
"step": 1150
|
2343 |
+
},
|
2344 |
+
{
|
2345 |
+
"epoch": 4.73,
|
2346 |
+
"grad_norm": 1.92129647731781,
|
2347 |
+
"learning_rate": 1.1087719298245616e-06,
|
2348 |
+
"logits/chosen": -1.856563925743103,
|
2349 |
+
"logits/rejected": -1.8815357685089111,
|
2350 |
+
"logps/chosen": -140.67437744140625,
|
2351 |
+
"logps/rejected": -180.8864288330078,
|
2352 |
+
"loss": 0.2,
|
2353 |
+
"rewards/accuracies": 0.8999999761581421,
|
2354 |
+
"rewards/chosen": 1.501416563987732,
|
2355 |
+
"rewards/margins": 8.230588912963867,
|
2356 |
+
"rewards/rejected": -6.7291717529296875,
|
2357 |
+
"step": 1160
|
2358 |
+
},
|
2359 |
+
{
|
2360 |
+
"epoch": 4.78,
|
2361 |
+
"grad_norm": 4.147223472595215,
|
2362 |
+
"learning_rate": 1.0736842105263159e-06,
|
2363 |
+
"logits/chosen": -1.7213938236236572,
|
2364 |
+
"logits/rejected": -1.7904603481292725,
|
2365 |
+
"logps/chosen": -125.3969497680664,
|
2366 |
+
"logps/rejected": -262.72662353515625,
|
2367 |
+
"loss": 0.0084,
|
2368 |
+
"rewards/accuracies": 1.0,
|
2369 |
+
"rewards/chosen": 3.621525287628174,
|
2370 |
+
"rewards/margins": 12.910409927368164,
|
2371 |
+
"rewards/rejected": -9.288885116577148,
|
2372 |
+
"step": 1170
|
2373 |
+
},
|
2374 |
+
{
|
2375 |
+
"epoch": 4.78,
|
2376 |
+
"eval_logits/chosen": -1.815858244895935,
|
2377 |
+
"eval_logits/rejected": -1.8551596403121948,
|
2378 |
+
"eval_logps/chosen": -189.174072265625,
|
2379 |
+
"eval_logps/rejected": -262.9738464355469,
|
2380 |
+
"eval_loss": 0.11231832951307297,
|
2381 |
+
"eval_rewards/accuracies": 0.9642857313156128,
|
2382 |
+
"eval_rewards/chosen": 1.610654592514038,
|
2383 |
+
"eval_rewards/margins": 7.547122478485107,
|
2384 |
+
"eval_rewards/rejected": -5.936468601226807,
|
2385 |
+
"eval_runtime": 26.2436,
|
2386 |
+
"eval_samples_per_second": 1.067,
|
2387 |
+
"eval_steps_per_second": 1.067,
|
2388 |
+
"step": 1170
|
2389 |
+
},
|
2390 |
+
{
|
2391 |
+
"epoch": 4.82,
|
2392 |
+
"grad_norm": 8.42303293779878e-09,
|
2393 |
+
"learning_rate": 1.0385964912280702e-06,
|
2394 |
+
"logits/chosen": -1.810766577720642,
|
2395 |
+
"logits/rejected": -1.8864467144012451,
|
2396 |
+
"logps/chosen": -138.02499389648438,
|
2397 |
+
"logps/rejected": -274.3287048339844,
|
2398 |
+
"loss": 0.0587,
|
2399 |
+
"rewards/accuracies": 1.0,
|
2400 |
+
"rewards/chosen": 2.259162187576294,
|
2401 |
+
"rewards/margins": 12.501955032348633,
|
2402 |
+
"rewards/rejected": -10.242792129516602,
|
2403 |
+
"step": 1180
|
2404 |
+
},
|
2405 |
+
{
|
2406 |
+
"epoch": 4.86,
|
2407 |
+
"grad_norm": 1.6972613821053528e-06,
|
2408 |
+
"learning_rate": 1.0035087719298247e-06,
|
2409 |
+
"logits/chosen": -1.8087619543075562,
|
2410 |
+
"logits/rejected": -1.864861249923706,
|
2411 |
+
"logps/chosen": -106.66157531738281,
|
2412 |
+
"logps/rejected": -172.39846801757812,
|
2413 |
+
"loss": 0.0167,
|
2414 |
+
"rewards/accuracies": 1.0,
|
2415 |
+
"rewards/chosen": 1.4979627132415771,
|
2416 |
+
"rewards/margins": 9.330595016479492,
|
2417 |
+
"rewards/rejected": -7.832631587982178,
|
2418 |
+
"step": 1190
|
2419 |
+
},
|
2420 |
+
{
|
2421 |
+
"epoch": 4.9,
|
2422 |
+
"grad_norm": 0.07996608316898346,
|
2423 |
+
"learning_rate": 9.68421052631579e-07,
|
2424 |
+
"logits/chosen": -1.6860878467559814,
|
2425 |
+
"logits/rejected": -1.7491111755371094,
|
2426 |
+
"logps/chosen": -86.7617416381836,
|
2427 |
+
"logps/rejected": -172.6136016845703,
|
2428 |
+
"loss": 0.0049,
|
2429 |
+
"rewards/accuracies": 1.0,
|
2430 |
+
"rewards/chosen": 3.904423952102661,
|
2431 |
+
"rewards/margins": 10.647790908813477,
|
2432 |
+
"rewards/rejected": -6.743367671966553,
|
2433 |
+
"step": 1200
|
2434 |
+
},
|
2435 |
+
{
|
2436 |
+
"epoch": 4.9,
|
2437 |
+
"eval_logits/chosen": -1.8151248693466187,
|
2438 |
+
"eval_logits/rejected": -1.8545377254486084,
|
2439 |
+
"eval_logps/chosen": -189.30612182617188,
|
2440 |
+
"eval_logps/rejected": -263.3506774902344,
|
2441 |
+
"eval_loss": 0.10828591883182526,
|
2442 |
+
"eval_rewards/accuracies": 0.9642857313156128,
|
2443 |
+
"eval_rewards/chosen": 1.5710465908050537,
|
2444 |
+
"eval_rewards/margins": 7.62056303024292,
|
2445 |
+
"eval_rewards/rejected": -6.049516201019287,
|
2446 |
+
"eval_runtime": 26.2353,
|
2447 |
+
"eval_samples_per_second": 1.067,
|
2448 |
+
"eval_steps_per_second": 1.067,
|
2449 |
+
"step": 1200
|
2450 |
+
},
|
2451 |
+
{
|
2452 |
+
"epoch": 4.94,
|
2453 |
+
"grad_norm": 0.0006144978106021881,
|
2454 |
+
"learning_rate": 9.333333333333334e-07,
|
2455 |
+
"logits/chosen": -1.7948280572891235,
|
2456 |
+
"logits/rejected": -1.8062083721160889,
|
2457 |
+
"logps/chosen": -201.0863037109375,
|
2458 |
+
"logps/rejected": -189.9558563232422,
|
2459 |
+
"loss": 0.0226,
|
2460 |
+
"rewards/accuracies": 1.0,
|
2461 |
+
"rewards/chosen": 2.141186237335205,
|
2462 |
+
"rewards/margins": 9.003381729125977,
|
2463 |
+
"rewards/rejected": -6.862195014953613,
|
2464 |
+
"step": 1210
|
2465 |
+
},
|
2466 |
+
{
|
2467 |
+
"epoch": 4.98,
|
2468 |
+
"grad_norm": 0.004125670529901981,
|
2469 |
+
"learning_rate": 8.982456140350877e-07,
|
2470 |
+
"logits/chosen": -1.7143564224243164,
|
2471 |
+
"logits/rejected": -1.755270004272461,
|
2472 |
+
"logps/chosen": -129.1698760986328,
|
2473 |
+
"logps/rejected": -167.7386016845703,
|
2474 |
+
"loss": 0.0439,
|
2475 |
+
"rewards/accuracies": 1.0,
|
2476 |
+
"rewards/chosen": 0.8997966647148132,
|
2477 |
+
"rewards/margins": 8.175169944763184,
|
2478 |
+
"rewards/rejected": -7.275373935699463,
|
2479 |
+
"step": 1220
|
2480 |
+
},
|
2481 |
+
{
|
2482 |
+
"epoch": 5.02,
|
2483 |
+
"grad_norm": 0.000474505708552897,
|
2484 |
+
"learning_rate": 8.631578947368421e-07,
|
2485 |
+
"logits/chosen": -1.8216488361358643,
|
2486 |
+
"logits/rejected": -1.880822777748108,
|
2487 |
+
"logps/chosen": -124.86578369140625,
|
2488 |
+
"logps/rejected": -239.4408721923828,
|
2489 |
+
"loss": 0.0746,
|
2490 |
+
"rewards/accuracies": 0.8999999761581421,
|
2491 |
+
"rewards/chosen": 1.0542094707489014,
|
2492 |
+
"rewards/margins": 9.776211738586426,
|
2493 |
+
"rewards/rejected": -8.722002029418945,
|
2494 |
+
"step": 1230
|
2495 |
+
},
|
2496 |
+
{
|
2497 |
+
"epoch": 5.02,
|
2498 |
+
"eval_logits/chosen": -1.8140125274658203,
|
2499 |
+
"eval_logits/rejected": -1.8534783124923706,
|
2500 |
+
"eval_logps/chosen": -189.43362426757812,
|
2501 |
+
"eval_logps/rejected": -263.6144104003906,
|
2502 |
+
"eval_loss": 0.10336750000715256,
|
2503 |
+
"eval_rewards/accuracies": 0.9642857313156128,
|
2504 |
+
"eval_rewards/chosen": 1.5327917337417603,
|
2505 |
+
"eval_rewards/margins": 7.6614274978637695,
|
2506 |
+
"eval_rewards/rejected": -6.128636837005615,
|
2507 |
+
"eval_runtime": 26.2322,
|
2508 |
+
"eval_samples_per_second": 1.067,
|
2509 |
+
"eval_steps_per_second": 1.067,
|
2510 |
+
"step": 1230
|
2511 |
+
},
|
2512 |
+
{
|
2513 |
+
"epoch": 5.06,
|
2514 |
+
"grad_norm": 0.029490573331713676,
|
2515 |
+
"learning_rate": 8.280701754385966e-07,
|
2516 |
+
"logits/chosen": -1.778794288635254,
|
2517 |
+
"logits/rejected": -1.876511573791504,
|
2518 |
+
"logps/chosen": -113.80903625488281,
|
2519 |
+
"logps/rejected": -301.50201416015625,
|
2520 |
+
"loss": 0.0002,
|
2521 |
+
"rewards/accuracies": 1.0,
|
2522 |
+
"rewards/chosen": 2.4956843852996826,
|
2523 |
+
"rewards/margins": 14.169008255004883,
|
2524 |
+
"rewards/rejected": -11.673322677612305,
|
2525 |
+
"step": 1240
|
2526 |
+
},
|
2527 |
+
{
|
2528 |
+
"epoch": 5.1,
|
2529 |
+
"grad_norm": 0.8206026554107666,
|
2530 |
+
"learning_rate": 7.92982456140351e-07,
|
2531 |
+
"logits/chosen": -1.7988386154174805,
|
2532 |
+
"logits/rejected": -1.8411651849746704,
|
2533 |
+
"logps/chosen": -80.52401733398438,
|
2534 |
+
"logps/rejected": -122.70835876464844,
|
2535 |
+
"loss": 0.0049,
|
2536 |
+
"rewards/accuracies": 1.0,
|
2537 |
+
"rewards/chosen": 3.3624215126037598,
|
2538 |
+
"rewards/margins": 10.361730575561523,
|
2539 |
+
"rewards/rejected": -6.999309539794922,
|
2540 |
+
"step": 1250
|
2541 |
+
},
|
2542 |
+
{
|
2543 |
+
"epoch": 5.14,
|
2544 |
+
"grad_norm": 0.004326160531491041,
|
2545 |
+
"learning_rate": 7.578947368421054e-07,
|
2546 |
+
"logits/chosen": -1.735650658607483,
|
2547 |
+
"logits/rejected": -1.7782208919525146,
|
2548 |
+
"logps/chosen": -143.6359100341797,
|
2549 |
+
"logps/rejected": -203.25271606445312,
|
2550 |
+
"loss": 0.0091,
|
2551 |
+
"rewards/accuracies": 1.0,
|
2552 |
+
"rewards/chosen": 2.7012696266174316,
|
2553 |
+
"rewards/margins": 9.915846824645996,
|
2554 |
+
"rewards/rejected": -7.214578151702881,
|
2555 |
+
"step": 1260
|
2556 |
+
},
|
2557 |
+
{
|
2558 |
+
"epoch": 5.14,
|
2559 |
+
"eval_logits/chosen": -1.8135929107666016,
|
2560 |
+
"eval_logits/rejected": -1.8531488180160522,
|
2561 |
+
"eval_logps/chosen": -189.6214599609375,
|
2562 |
+
"eval_logps/rejected": -264.03973388671875,
|
2563 |
+
"eval_loss": 0.10307549685239792,
|
2564 |
+
"eval_rewards/accuracies": 0.9642857313156128,
|
2565 |
+
"eval_rewards/chosen": 1.4764469861984253,
|
2566 |
+
"eval_rewards/margins": 7.73268461227417,
|
2567 |
+
"eval_rewards/rejected": -6.256237506866455,
|
2568 |
+
"eval_runtime": 26.2397,
|
2569 |
+
"eval_samples_per_second": 1.067,
|
2570 |
+
"eval_steps_per_second": 1.067,
|
2571 |
+
"step": 1260
|
2572 |
+
}
|
2573 |
+
],
|
2574 |
+
"logging_steps": 10,
|
2575 |
+
"max_steps": 1470,
|
2576 |
+
"num_input_tokens_seen": 0,
|
2577 |
+
"num_train_epochs": 6,
|
2578 |
+
"save_steps": 90,
|
2579 |
+
"total_flos": 0.0,
|
2580 |
+
"train_batch_size": 1,
|
2581 |
+
"trial_name": null,
|
2582 |
+
"trial_params": null
|
2583 |
+
}
|
checkpoint-1260/training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:41693834eff4b96ebb423cc3037dde0e25267c10f91f2e959260eee7bddf1ddb
|
3 |
+
size 5304
|
checkpoint-1350/README.md
ADDED
@@ -0,0 +1,202 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
library_name: peft
|
3 |
+
base_model: Weni/WeniGPT-Agents-Mistral-1.0.6-SFT-merged
|
4 |
+
---
|
5 |
+
|
6 |
+
# Model Card for Model ID
|
7 |
+
|
8 |
+
<!-- Provide a quick summary of what the model is/does. -->
|
9 |
+
|
10 |
+
|
11 |
+
|
12 |
+
## Model Details
|
13 |
+
|
14 |
+
### Model Description
|
15 |
+
|
16 |
+
<!-- Provide a longer summary of what this model is. -->
|
17 |
+
|
18 |
+
|
19 |
+
|
20 |
+
- **Developed by:** [More Information Needed]
|
21 |
+
- **Funded by [optional]:** [More Information Needed]
|
22 |
+
- **Shared by [optional]:** [More Information Needed]
|
23 |
+
- **Model type:** [More Information Needed]
|
24 |
+
- **Language(s) (NLP):** [More Information Needed]
|
25 |
+
- **License:** [More Information Needed]
|
26 |
+
- **Finetuned from model [optional]:** [More Information Needed]
|
27 |
+
|
28 |
+
### Model Sources [optional]
|
29 |
+
|
30 |
+
<!-- Provide the basic links for the model. -->
|
31 |
+
|
32 |
+
- **Repository:** [More Information Needed]
|
33 |
+
- **Paper [optional]:** [More Information Needed]
|
34 |
+
- **Demo [optional]:** [More Information Needed]
|
35 |
+
|
36 |
+
## Uses
|
37 |
+
|
38 |
+
<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
|
39 |
+
|
40 |
+
### Direct Use
|
41 |
+
|
42 |
+
<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
|
43 |
+
|
44 |
+
[More Information Needed]
|
45 |
+
|
46 |
+
### Downstream Use [optional]
|
47 |
+
|
48 |
+
<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
|
49 |
+
|
50 |
+
[More Information Needed]
|
51 |
+
|
52 |
+
### Out-of-Scope Use
|
53 |
+
|
54 |
+
<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
|
55 |
+
|
56 |
+
[More Information Needed]
|
57 |
+
|
58 |
+
## Bias, Risks, and Limitations
|
59 |
+
|
60 |
+
<!-- This section is meant to convey both technical and sociotechnical limitations. -->
|
61 |
+
|
62 |
+
[More Information Needed]
|
63 |
+
|
64 |
+
### Recommendations
|
65 |
+
|
66 |
+
<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
|
67 |
+
|
68 |
+
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
|
69 |
+
|
70 |
+
## How to Get Started with the Model
|
71 |
+
|
72 |
+
Use the code below to get started with the model.
|
73 |
+
|
74 |
+
[More Information Needed]
|
75 |
+
|
76 |
+
## Training Details
|
77 |
+
|
78 |
+
### Training Data
|
79 |
+
|
80 |
+
<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
|
81 |
+
|
82 |
+
[More Information Needed]
|
83 |
+
|
84 |
+
### Training Procedure
|
85 |
+
|
86 |
+
<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
|
87 |
+
|
88 |
+
#### Preprocessing [optional]
|
89 |
+
|
90 |
+
[More Information Needed]
|
91 |
+
|
92 |
+
|
93 |
+
#### Training Hyperparameters
|
94 |
+
|
95 |
+
- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
|
96 |
+
|
97 |
+
#### Speeds, Sizes, Times [optional]
|
98 |
+
|
99 |
+
<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
|
100 |
+
|
101 |
+
[More Information Needed]
|
102 |
+
|
103 |
+
## Evaluation
|
104 |
+
|
105 |
+
<!-- This section describes the evaluation protocols and provides the results. -->
|
106 |
+
|
107 |
+
### Testing Data, Factors & Metrics
|
108 |
+
|
109 |
+
#### Testing Data
|
110 |
+
|
111 |
+
<!-- This should link to a Dataset Card if possible. -->
|
112 |
+
|
113 |
+
[More Information Needed]
|
114 |
+
|
115 |
+
#### Factors
|
116 |
+
|
117 |
+
<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
|
118 |
+
|
119 |
+
[More Information Needed]
|
120 |
+
|
121 |
+
#### Metrics
|
122 |
+
|
123 |
+
<!-- These are the evaluation metrics being used, ideally with a description of why. -->
|
124 |
+
|
125 |
+
[More Information Needed]
|
126 |
+
|
127 |
+
### Results
|
128 |
+
|
129 |
+
[More Information Needed]
|
130 |
+
|
131 |
+
#### Summary
|
132 |
+
|
133 |
+
|
134 |
+
|
135 |
+
## Model Examination [optional]
|
136 |
+
|
137 |
+
<!-- Relevant interpretability work for the model goes here -->
|
138 |
+
|
139 |
+
[More Information Needed]
|
140 |
+
|
141 |
+
## Environmental Impact
|
142 |
+
|
143 |
+
<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
|
144 |
+
|
145 |
+
Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
|
146 |
+
|
147 |
+
- **Hardware Type:** [More Information Needed]
|
148 |
+
- **Hours used:** [More Information Needed]
|
149 |
+
- **Cloud Provider:** [More Information Needed]
|
150 |
+
- **Compute Region:** [More Information Needed]
|
151 |
+
- **Carbon Emitted:** [More Information Needed]
|
152 |
+
|
153 |
+
## Technical Specifications [optional]
|
154 |
+
|
155 |
+
### Model Architecture and Objective
|
156 |
+
|
157 |
+
[More Information Needed]
|
158 |
+
|
159 |
+
### Compute Infrastructure
|
160 |
+
|
161 |
+
[More Information Needed]
|
162 |
+
|
163 |
+
#### Hardware
|
164 |
+
|
165 |
+
[More Information Needed]
|
166 |
+
|
167 |
+
#### Software
|
168 |
+
|
169 |
+
[More Information Needed]
|
170 |
+
|
171 |
+
## Citation [optional]
|
172 |
+
|
173 |
+
<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
|
174 |
+
|
175 |
+
**BibTeX:**
|
176 |
+
|
177 |
+
[More Information Needed]
|
178 |
+
|
179 |
+
**APA:**
|
180 |
+
|
181 |
+
[More Information Needed]
|
182 |
+
|
183 |
+
## Glossary [optional]
|
184 |
+
|
185 |
+
<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
|
186 |
+
|
187 |
+
[More Information Needed]
|
188 |
+
|
189 |
+
## More Information [optional]
|
190 |
+
|
191 |
+
[More Information Needed]
|
192 |
+
|
193 |
+
## Model Card Authors [optional]
|
194 |
+
|
195 |
+
[More Information Needed]
|
196 |
+
|
197 |
+
## Model Card Contact
|
198 |
+
|
199 |
+
[More Information Needed]
|
200 |
+
### Framework versions
|
201 |
+
|
202 |
+
- PEFT 0.10.0
|
checkpoint-1350/adapter_config.json
ADDED
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"alpha_pattern": {},
|
3 |
+
"auto_mapping": null,
|
4 |
+
"base_model_name_or_path": "Weni/WeniGPT-Agents-Mistral-1.0.6-SFT-merged",
|
5 |
+
"bias": "none",
|
6 |
+
"fan_in_fan_out": false,
|
7 |
+
"inference_mode": true,
|
8 |
+
"init_lora_weights": true,
|
9 |
+
"layer_replication": null,
|
10 |
+
"layers_pattern": null,
|
11 |
+
"layers_to_transform": null,
|
12 |
+
"loftq_config": {},
|
13 |
+
"lora_alpha": 16,
|
14 |
+
"lora_dropout": 0.05,
|
15 |
+
"megatron_config": null,
|
16 |
+
"megatron_core": "megatron.core",
|
17 |
+
"modules_to_save": null,
|
18 |
+
"peft_type": "LORA",
|
19 |
+
"r": 8,
|
20 |
+
"rank_pattern": {},
|
21 |
+
"revision": null,
|
22 |
+
"target_modules": [
|
23 |
+
"q_proj",
|
24 |
+
"v_proj"
|
25 |
+
],
|
26 |
+
"task_type": "CAUSAL_LM",
|
27 |
+
"use_dora": false,
|
28 |
+
"use_rslora": false
|
29 |
+
}
|
checkpoint-1350/adapter_model.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d964c5e5c43d94ec40c7ed379be88dce7d0a44afa351d029c8d8ad44a8a167b1
|
3 |
+
size 13648432
|
checkpoint-1350/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:6971b6b49f8306c7b241b1505ad348ae024baac98dd1411010307b6cfcf21a0d
|
3 |
+
size 27370618
|
checkpoint-1350/rng_state.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:592c43fcd4f224a713b94e9b0547d56fff2d04c9a3f50aaae922858b09cf4f47
|
3 |
+
size 14244
|
checkpoint-1350/scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:136b6be5c9ba018c605107bc92ab6ea71244aa566e0465b79392769629761a4f
|
3 |
+
size 1064
|
checkpoint-1350/special_tokens_map.json
ADDED
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"bos_token": {
|
3 |
+
"content": "<s>",
|
4 |
+
"lstrip": false,
|
5 |
+
"normalized": false,
|
6 |
+
"rstrip": false,
|
7 |
+
"single_word": false
|
8 |
+
},
|
9 |
+
"eos_token": {
|
10 |
+
"content": "</s>",
|
11 |
+
"lstrip": false,
|
12 |
+
"normalized": false,
|
13 |
+
"rstrip": false,
|
14 |
+
"single_word": false
|
15 |
+
},
|
16 |
+
"pad_token": "<unk>",
|
17 |
+
"unk_token": {
|
18 |
+
"content": "<unk>",
|
19 |
+
"lstrip": false,
|
20 |
+
"normalized": false,
|
21 |
+
"rstrip": false,
|
22 |
+
"single_word": false
|
23 |
+
}
|
24 |
+
}
|
checkpoint-1350/tokenizer.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-1350/tokenizer.model
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:dadfd56d766715c61d2ef780a525ab43b8e6da4de6865bda3d95fdef5e134055
|
3 |
+
size 493443
|
checkpoint-1350/tokenizer_config.json
ADDED
@@ -0,0 +1,49 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"add_bos_token": true,
|
3 |
+
"add_eos_token": false,
|
4 |
+
"added_tokens_decoder": {
|
5 |
+
"0": {
|
6 |
+
"content": "<unk>",
|
7 |
+
"lstrip": false,
|
8 |
+
"normalized": false,
|
9 |
+
"rstrip": false,
|
10 |
+
"single_word": false,
|
11 |
+
"special": true
|
12 |
+
},
|
13 |
+
"1": {
|
14 |
+
"content": "<s>",
|
15 |
+
"lstrip": false,
|
16 |
+
"normalized": false,
|
17 |
+
"rstrip": false,
|
18 |
+
"single_word": false,
|
19 |
+
"special": true
|
20 |
+
},
|
21 |
+
"2": {
|
22 |
+
"content": "</s>",
|
23 |
+
"lstrip": false,
|
24 |
+
"normalized": false,
|
25 |
+
"rstrip": false,
|
26 |
+
"single_word": false,
|
27 |
+
"special": true
|
28 |
+
}
|
29 |
+
},
|
30 |
+
"additional_special_tokens": [],
|
31 |
+
"bos_token": "<s>",
|
32 |
+
"chat_template": "{% for message in messages %}{% if message['role'] == 'user' %}{{ bos_token + '[INST] ' + message['content'] + ' [/INST]' }}{% elif message['role'] == 'system' %}{{ '<<SYS>>\\n' + message['content'] + '\\n<</SYS>>\\n\\n' }}{% elif message['role'] == 'assistant' %}{{ ' ' + message['content'] + ' ' + eos_token }}{% endif %}{% endfor %}",
|
33 |
+
"clean_up_tokenization_spaces": false,
|
34 |
+
"eos_token": "</s>",
|
35 |
+
"legacy": true,
|
36 |
+
"max_lenght": 8192,
|
37 |
+
"max_length": 8192,
|
38 |
+
"model_max_length": 1000000000000000019884624838656,
|
39 |
+
"pad_token": "<unk>",
|
40 |
+
"padding": true,
|
41 |
+
"sp_model_kwargs": {},
|
42 |
+
"spaces_between_special_tokens": false,
|
43 |
+
"stride": 0,
|
44 |
+
"tokenizer_class": "LlamaTokenizer",
|
45 |
+
"truncation_side": "right",
|
46 |
+
"truncation_strategy": "longest_first",
|
47 |
+
"unk_token": "<unk>",
|
48 |
+
"use_default_system_prompt": false
|
49 |
+
}
|
checkpoint-1350/trainer_state.json
ADDED
@@ -0,0 +1,2766 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": 0.09495646506547928,
|
3 |
+
"best_model_checkpoint": "./mistral/20-04-24-Weni-WeniGPT-Agents-Mistral-1.0.6-SFT-1.0.7-DPO_Experiment on DPO with other hyperparameters and best SFT model of WeniGPT-2_max_steps-1470_batch_1_2024-04-20_ppid_9/checkpoint-1350",
|
4 |
+
"epoch": 5.510204081632653,
|
5 |
+
"eval_steps": 30,
|
6 |
+
"global_step": 1350,
|
7 |
+
"is_hyper_param_search": false,
|
8 |
+
"is_local_process_zero": true,
|
9 |
+
"is_world_process_zero": true,
|
10 |
+
"log_history": [
|
11 |
+
{
|
12 |
+
"epoch": 0.04,
|
13 |
+
"grad_norm": 83.44550323486328,
|
14 |
+
"learning_rate": 6.666666666666667e-07,
|
15 |
+
"logits/chosen": -1.7825796604156494,
|
16 |
+
"logits/rejected": -1.7886196374893188,
|
17 |
+
"logps/chosen": -169.71481323242188,
|
18 |
+
"logps/rejected": -167.6764373779297,
|
19 |
+
"loss": 0.6937,
|
20 |
+
"rewards/accuracies": 0.20000000298023224,
|
21 |
+
"rewards/chosen": -0.0002815247280523181,
|
22 |
+
"rewards/margins": -0.0010146332206204534,
|
23 |
+
"rewards/rejected": 0.0007331084925681353,
|
24 |
+
"step": 10
|
25 |
+
},
|
26 |
+
{
|
27 |
+
"epoch": 0.08,
|
28 |
+
"grad_norm": 87.56982421875,
|
29 |
+
"learning_rate": 1.777777777777778e-06,
|
30 |
+
"logits/chosen": -1.623265027999878,
|
31 |
+
"logits/rejected": -1.6968040466308594,
|
32 |
+
"logps/chosen": -128.7312469482422,
|
33 |
+
"logps/rejected": -184.20187377929688,
|
34 |
+
"loss": 0.6912,
|
35 |
+
"rewards/accuracies": 0.5,
|
36 |
+
"rewards/chosen": 0.0030528828501701355,
|
37 |
+
"rewards/margins": 0.004004458896815777,
|
38 |
+
"rewards/rejected": -0.0009515761630609632,
|
39 |
+
"step": 20
|
40 |
+
},
|
41 |
+
{
|
42 |
+
"epoch": 0.12,
|
43 |
+
"grad_norm": 113.4741439819336,
|
44 |
+
"learning_rate": 2.888888888888889e-06,
|
45 |
+
"logits/chosen": -1.7152118682861328,
|
46 |
+
"logits/rejected": -1.758681058883667,
|
47 |
+
"logps/chosen": -146.91494750976562,
|
48 |
+
"logps/rejected": -213.5966339111328,
|
49 |
+
"loss": 0.6781,
|
50 |
+
"rewards/accuracies": 1.0,
|
51 |
+
"rewards/chosen": 0.03251449763774872,
|
52 |
+
"rewards/margins": 0.030697176232933998,
|
53 |
+
"rewards/rejected": 0.0018173219868913293,
|
54 |
+
"step": 30
|
55 |
+
},
|
56 |
+
{
|
57 |
+
"epoch": 0.12,
|
58 |
+
"eval_logits/chosen": -1.795080304145813,
|
59 |
+
"eval_logits/rejected": -1.8307874202728271,
|
60 |
+
"eval_logps/chosen": -194.3749542236328,
|
61 |
+
"eval_logps/rejected": -243.1332244873047,
|
62 |
+
"eval_loss": 0.6762310266494751,
|
63 |
+
"eval_rewards/accuracies": 0.75,
|
64 |
+
"eval_rewards/chosen": 0.05039369314908981,
|
65 |
+
"eval_rewards/margins": 0.03467999026179314,
|
66 |
+
"eval_rewards/rejected": 0.01571369729936123,
|
67 |
+
"eval_runtime": 26.2365,
|
68 |
+
"eval_samples_per_second": 1.067,
|
69 |
+
"eval_steps_per_second": 1.067,
|
70 |
+
"step": 30
|
71 |
+
},
|
72 |
+
{
|
73 |
+
"epoch": 0.16,
|
74 |
+
"grad_norm": 50.695308685302734,
|
75 |
+
"learning_rate": 4.000000000000001e-06,
|
76 |
+
"logits/chosen": -1.9357620477676392,
|
77 |
+
"logits/rejected": -1.9375683069229126,
|
78 |
+
"logps/chosen": -183.8218231201172,
|
79 |
+
"logps/rejected": -115.80632019042969,
|
80 |
+
"loss": 0.6768,
|
81 |
+
"rewards/accuracies": 0.699999988079071,
|
82 |
+
"rewards/chosen": 0.056698836386203766,
|
83 |
+
"rewards/margins": 0.03413978964090347,
|
84 |
+
"rewards/rejected": 0.02255905233323574,
|
85 |
+
"step": 40
|
86 |
+
},
|
87 |
+
{
|
88 |
+
"epoch": 0.2,
|
89 |
+
"grad_norm": 71.6834716796875,
|
90 |
+
"learning_rate": 4.996491228070176e-06,
|
91 |
+
"logits/chosen": -1.7773185968399048,
|
92 |
+
"logits/rejected": -1.8331743478775024,
|
93 |
+
"logps/chosen": -167.78533935546875,
|
94 |
+
"logps/rejected": -227.0582733154297,
|
95 |
+
"loss": 0.6205,
|
96 |
+
"rewards/accuracies": 0.8999999761581421,
|
97 |
+
"rewards/chosen": 0.1449195146560669,
|
98 |
+
"rewards/margins": 0.15578031539916992,
|
99 |
+
"rewards/rejected": -0.010860783979296684,
|
100 |
+
"step": 50
|
101 |
+
},
|
102 |
+
{
|
103 |
+
"epoch": 0.24,
|
104 |
+
"grad_norm": 93.83502197265625,
|
105 |
+
"learning_rate": 4.96140350877193e-06,
|
106 |
+
"logits/chosen": -1.7739536762237549,
|
107 |
+
"logits/rejected": -1.8414790630340576,
|
108 |
+
"logps/chosen": -153.57481384277344,
|
109 |
+
"logps/rejected": -194.3485870361328,
|
110 |
+
"loss": 0.5918,
|
111 |
+
"rewards/accuracies": 0.8999999761581421,
|
112 |
+
"rewards/chosen": 0.2464807778596878,
|
113 |
+
"rewards/margins": 0.22958576679229736,
|
114 |
+
"rewards/rejected": 0.016895027831196785,
|
115 |
+
"step": 60
|
116 |
+
},
|
117 |
+
{
|
118 |
+
"epoch": 0.24,
|
119 |
+
"eval_logits/chosen": -1.7974774837493896,
|
120 |
+
"eval_logits/rejected": -1.8332849740982056,
|
121 |
+
"eval_logps/chosen": -193.71743774414062,
|
122 |
+
"eval_logps/rejected": -243.05783081054688,
|
123 |
+
"eval_loss": 0.5998325943946838,
|
124 |
+
"eval_rewards/accuracies": 0.7857142686843872,
|
125 |
+
"eval_rewards/chosen": 0.24764502048492432,
|
126 |
+
"eval_rewards/margins": 0.20930658280849457,
|
127 |
+
"eval_rewards/rejected": 0.03833846375346184,
|
128 |
+
"eval_runtime": 26.2349,
|
129 |
+
"eval_samples_per_second": 1.067,
|
130 |
+
"eval_steps_per_second": 1.067,
|
131 |
+
"step": 60
|
132 |
+
},
|
133 |
+
{
|
134 |
+
"epoch": 0.29,
|
135 |
+
"grad_norm": 109.71602630615234,
|
136 |
+
"learning_rate": 4.926315789473685e-06,
|
137 |
+
"logits/chosen": -1.8950660228729248,
|
138 |
+
"logits/rejected": -1.9148075580596924,
|
139 |
+
"logps/chosen": -169.89028930664062,
|
140 |
+
"logps/rejected": -177.87301635742188,
|
141 |
+
"loss": 0.5992,
|
142 |
+
"rewards/accuracies": 0.800000011920929,
|
143 |
+
"rewards/chosen": 0.15271134674549103,
|
144 |
+
"rewards/margins": 0.21031668782234192,
|
145 |
+
"rewards/rejected": -0.0576053187251091,
|
146 |
+
"step": 70
|
147 |
+
},
|
148 |
+
{
|
149 |
+
"epoch": 0.33,
|
150 |
+
"grad_norm": 90.3881607055664,
|
151 |
+
"learning_rate": 4.8912280701754386e-06,
|
152 |
+
"logits/chosen": -1.798600196838379,
|
153 |
+
"logits/rejected": -1.855521559715271,
|
154 |
+
"logps/chosen": -160.3217315673828,
|
155 |
+
"logps/rejected": -247.6388397216797,
|
156 |
+
"loss": 0.5551,
|
157 |
+
"rewards/accuracies": 0.699999988079071,
|
158 |
+
"rewards/chosen": 0.29141485691070557,
|
159 |
+
"rewards/margins": 0.3369868993759155,
|
160 |
+
"rewards/rejected": -0.04557197913527489,
|
161 |
+
"step": 80
|
162 |
+
},
|
163 |
+
{
|
164 |
+
"epoch": 0.37,
|
165 |
+
"grad_norm": 147.27755737304688,
|
166 |
+
"learning_rate": 4.856140350877193e-06,
|
167 |
+
"logits/chosen": -1.8193047046661377,
|
168 |
+
"logits/rejected": -1.8773219585418701,
|
169 |
+
"logps/chosen": -155.37123107910156,
|
170 |
+
"logps/rejected": -245.6790008544922,
|
171 |
+
"loss": 0.4932,
|
172 |
+
"rewards/accuracies": 0.699999988079071,
|
173 |
+
"rewards/chosen": 0.5351110696792603,
|
174 |
+
"rewards/margins": 0.5927553176879883,
|
175 |
+
"rewards/rejected": -0.05764435604214668,
|
176 |
+
"step": 90
|
177 |
+
},
|
178 |
+
{
|
179 |
+
"epoch": 0.37,
|
180 |
+
"eval_logits/chosen": -1.8004387617111206,
|
181 |
+
"eval_logits/rejected": -1.836446762084961,
|
182 |
+
"eval_logps/chosen": -192.66908264160156,
|
183 |
+
"eval_logps/rejected": -242.95899963378906,
|
184 |
+
"eval_loss": 0.5072147846221924,
|
185 |
+
"eval_rewards/accuracies": 0.8214285969734192,
|
186 |
+
"eval_rewards/chosen": 0.5621528029441833,
|
187 |
+
"eval_rewards/margins": 0.4941706657409668,
|
188 |
+
"eval_rewards/rejected": 0.06798211485147476,
|
189 |
+
"eval_runtime": 26.2422,
|
190 |
+
"eval_samples_per_second": 1.067,
|
191 |
+
"eval_steps_per_second": 1.067,
|
192 |
+
"step": 90
|
193 |
+
},
|
194 |
+
{
|
195 |
+
"epoch": 0.41,
|
196 |
+
"grad_norm": 62.75928497314453,
|
197 |
+
"learning_rate": 4.821052631578948e-06,
|
198 |
+
"logits/chosen": -1.7835988998413086,
|
199 |
+
"logits/rejected": -1.8885362148284912,
|
200 |
+
"logps/chosen": -109.94624328613281,
|
201 |
+
"logps/rejected": -232.5241241455078,
|
202 |
+
"loss": 0.4252,
|
203 |
+
"rewards/accuracies": 1.0,
|
204 |
+
"rewards/chosen": 0.5394761562347412,
|
205 |
+
"rewards/margins": 0.787738025188446,
|
206 |
+
"rewards/rejected": -0.248261958360672,
|
207 |
+
"step": 100
|
208 |
+
},
|
209 |
+
{
|
210 |
+
"epoch": 0.45,
|
211 |
+
"grad_norm": 84.42039489746094,
|
212 |
+
"learning_rate": 4.785964912280702e-06,
|
213 |
+
"logits/chosen": -1.7081111669540405,
|
214 |
+
"logits/rejected": -1.7244077920913696,
|
215 |
+
"logps/chosen": -92.51998138427734,
|
216 |
+
"logps/rejected": -135.5617218017578,
|
217 |
+
"loss": 0.5077,
|
218 |
+
"rewards/accuracies": 0.800000011920929,
|
219 |
+
"rewards/chosen": 0.6028972864151001,
|
220 |
+
"rewards/margins": 0.511993408203125,
|
221 |
+
"rewards/rejected": 0.09090389311313629,
|
222 |
+
"step": 110
|
223 |
+
},
|
224 |
+
{
|
225 |
+
"epoch": 0.49,
|
226 |
+
"grad_norm": 26.61499786376953,
|
227 |
+
"learning_rate": 4.754385964912281e-06,
|
228 |
+
"logits/chosen": -1.9250166416168213,
|
229 |
+
"logits/rejected": -1.9426968097686768,
|
230 |
+
"logps/chosen": -152.0164337158203,
|
231 |
+
"logps/rejected": -154.2359161376953,
|
232 |
+
"loss": 0.4391,
|
233 |
+
"rewards/accuracies": 0.8999999761581421,
|
234 |
+
"rewards/chosen": 0.7938548922538757,
|
235 |
+
"rewards/margins": 0.8368040919303894,
|
236 |
+
"rewards/rejected": -0.042949117720127106,
|
237 |
+
"step": 120
|
238 |
+
},
|
239 |
+
{
|
240 |
+
"epoch": 0.49,
|
241 |
+
"eval_logits/chosen": -1.8051280975341797,
|
242 |
+
"eval_logits/rejected": -1.8413046598434448,
|
243 |
+
"eval_logps/chosen": -191.2982177734375,
|
244 |
+
"eval_logps/rejected": -242.81198120117188,
|
245 |
+
"eval_loss": 0.4335751235485077,
|
246 |
+
"eval_rewards/accuracies": 0.7857142686843872,
|
247 |
+
"eval_rewards/chosen": 0.9734136462211609,
|
248 |
+
"eval_rewards/margins": 0.8613229990005493,
|
249 |
+
"eval_rewards/rejected": 0.11209066957235336,
|
250 |
+
"eval_runtime": 26.2345,
|
251 |
+
"eval_samples_per_second": 1.067,
|
252 |
+
"eval_steps_per_second": 1.067,
|
253 |
+
"step": 120
|
254 |
+
},
|
255 |
+
{
|
256 |
+
"epoch": 0.53,
|
257 |
+
"grad_norm": 90.6368637084961,
|
258 |
+
"learning_rate": 4.7192982456140355e-06,
|
259 |
+
"logits/chosen": -1.865144968032837,
|
260 |
+
"logits/rejected": -1.9209120273590088,
|
261 |
+
"logps/chosen": -149.78990173339844,
|
262 |
+
"logps/rejected": -293.8749084472656,
|
263 |
+
"loss": 0.4133,
|
264 |
+
"rewards/accuracies": 0.800000011920929,
|
265 |
+
"rewards/chosen": 0.4554510712623596,
|
266 |
+
"rewards/margins": 0.9285024404525757,
|
267 |
+
"rewards/rejected": -0.47305139899253845,
|
268 |
+
"step": 130
|
269 |
+
},
|
270 |
+
{
|
271 |
+
"epoch": 0.57,
|
272 |
+
"grad_norm": 6.8597941398620605,
|
273 |
+
"learning_rate": 4.68421052631579e-06,
|
274 |
+
"logits/chosen": -1.8089487552642822,
|
275 |
+
"logits/rejected": -1.8955650329589844,
|
276 |
+
"logps/chosen": -100.26250457763672,
|
277 |
+
"logps/rejected": -209.1724090576172,
|
278 |
+
"loss": 0.3414,
|
279 |
+
"rewards/accuracies": 1.0,
|
280 |
+
"rewards/chosen": 0.9389357566833496,
|
281 |
+
"rewards/margins": 1.521246314048767,
|
282 |
+
"rewards/rejected": -0.5823107361793518,
|
283 |
+
"step": 140
|
284 |
+
},
|
285 |
+
{
|
286 |
+
"epoch": 0.61,
|
287 |
+
"grad_norm": 79.9189453125,
|
288 |
+
"learning_rate": 4.652631578947368e-06,
|
289 |
+
"logits/chosen": -1.760371446609497,
|
290 |
+
"logits/rejected": -1.8096182346343994,
|
291 |
+
"logps/chosen": -193.62579345703125,
|
292 |
+
"logps/rejected": -166.35800170898438,
|
293 |
+
"loss": 0.3208,
|
294 |
+
"rewards/accuracies": 0.8999999761581421,
|
295 |
+
"rewards/chosen": 1.2600924968719482,
|
296 |
+
"rewards/margins": 1.5845091342926025,
|
297 |
+
"rewards/rejected": -0.32441645860671997,
|
298 |
+
"step": 150
|
299 |
+
},
|
300 |
+
{
|
301 |
+
"epoch": 0.61,
|
302 |
+
"eval_logits/chosen": -1.8129692077636719,
|
303 |
+
"eval_logits/rejected": -1.849239706993103,
|
304 |
+
"eval_logps/chosen": -189.88929748535156,
|
305 |
+
"eval_logps/rejected": -242.91104125976562,
|
306 |
+
"eval_loss": 0.39326220750808716,
|
307 |
+
"eval_rewards/accuracies": 0.7857142686843872,
|
308 |
+
"eval_rewards/chosen": 1.3960883617401123,
|
309 |
+
"eval_rewards/margins": 1.3137177228927612,
|
310 |
+
"eval_rewards/rejected": 0.08237060159444809,
|
311 |
+
"eval_runtime": 26.3372,
|
312 |
+
"eval_samples_per_second": 1.063,
|
313 |
+
"eval_steps_per_second": 1.063,
|
314 |
+
"step": 150
|
315 |
+
},
|
316 |
+
{
|
317 |
+
"epoch": 0.65,
|
318 |
+
"grad_norm": 3.151426315307617,
|
319 |
+
"learning_rate": 4.617543859649123e-06,
|
320 |
+
"logits/chosen": -1.7185112237930298,
|
321 |
+
"logits/rejected": -1.749427080154419,
|
322 |
+
"logps/chosen": -178.8656463623047,
|
323 |
+
"logps/rejected": -201.32858276367188,
|
324 |
+
"loss": 0.4418,
|
325 |
+
"rewards/accuracies": 0.699999988079071,
|
326 |
+
"rewards/chosen": 1.3884060382843018,
|
327 |
+
"rewards/margins": 1.4542597532272339,
|
328 |
+
"rewards/rejected": -0.06585375964641571,
|
329 |
+
"step": 160
|
330 |
+
},
|
331 |
+
{
|
332 |
+
"epoch": 0.69,
|
333 |
+
"grad_norm": 37.79099655151367,
|
334 |
+
"learning_rate": 4.582456140350878e-06,
|
335 |
+
"logits/chosen": -1.730787992477417,
|
336 |
+
"logits/rejected": -1.8099721670150757,
|
337 |
+
"logps/chosen": -158.50277709960938,
|
338 |
+
"logps/rejected": -185.48129272460938,
|
339 |
+
"loss": 0.3912,
|
340 |
+
"rewards/accuracies": 0.800000011920929,
|
341 |
+
"rewards/chosen": 1.4253065586090088,
|
342 |
+
"rewards/margins": 1.2418479919433594,
|
343 |
+
"rewards/rejected": 0.18345877528190613,
|
344 |
+
"step": 170
|
345 |
+
},
|
346 |
+
{
|
347 |
+
"epoch": 0.73,
|
348 |
+
"grad_norm": 18.49593734741211,
|
349 |
+
"learning_rate": 4.547368421052632e-06,
|
350 |
+
"logits/chosen": -1.842390775680542,
|
351 |
+
"logits/rejected": -1.8834030628204346,
|
352 |
+
"logps/chosen": -125.84483337402344,
|
353 |
+
"logps/rejected": -182.78854370117188,
|
354 |
+
"loss": 0.3215,
|
355 |
+
"rewards/accuracies": 0.8999999761581421,
|
356 |
+
"rewards/chosen": 1.5408955812454224,
|
357 |
+
"rewards/margins": 1.4677695035934448,
|
358 |
+
"rewards/rejected": 0.07312598079442978,
|
359 |
+
"step": 180
|
360 |
+
},
|
361 |
+
{
|
362 |
+
"epoch": 0.73,
|
363 |
+
"eval_logits/chosen": -1.8194406032562256,
|
364 |
+
"eval_logits/rejected": -1.8561830520629883,
|
365 |
+
"eval_logps/chosen": -188.3820037841797,
|
366 |
+
"eval_logps/rejected": -243.13536071777344,
|
367 |
+
"eval_loss": 0.3756321966648102,
|
368 |
+
"eval_rewards/accuracies": 0.7857142686843872,
|
369 |
+
"eval_rewards/chosen": 1.8482797145843506,
|
370 |
+
"eval_rewards/margins": 1.8331985473632812,
|
371 |
+
"eval_rewards/rejected": 0.015081183984875679,
|
372 |
+
"eval_runtime": 26.2381,
|
373 |
+
"eval_samples_per_second": 1.067,
|
374 |
+
"eval_steps_per_second": 1.067,
|
375 |
+
"step": 180
|
376 |
+
},
|
377 |
+
{
|
378 |
+
"epoch": 0.78,
|
379 |
+
"grad_norm": 106.56973266601562,
|
380 |
+
"learning_rate": 4.512280701754386e-06,
|
381 |
+
"logits/chosen": -1.8318649530410767,
|
382 |
+
"logits/rejected": -1.8742272853851318,
|
383 |
+
"logps/chosen": -197.66619873046875,
|
384 |
+
"logps/rejected": -247.83560180664062,
|
385 |
+
"loss": 0.3303,
|
386 |
+
"rewards/accuracies": 0.800000011920929,
|
387 |
+
"rewards/chosen": 2.4889655113220215,
|
388 |
+
"rewards/margins": 2.536695957183838,
|
389 |
+
"rewards/rejected": -0.047730110585689545,
|
390 |
+
"step": 190
|
391 |
+
},
|
392 |
+
{
|
393 |
+
"epoch": 0.82,
|
394 |
+
"grad_norm": 1.9826362133026123,
|
395 |
+
"learning_rate": 4.47719298245614e-06,
|
396 |
+
"logits/chosen": -1.7579128742218018,
|
397 |
+
"logits/rejected": -1.7890180349349976,
|
398 |
+
"logps/chosen": -129.37191772460938,
|
399 |
+
"logps/rejected": -166.67636108398438,
|
400 |
+
"loss": 0.3292,
|
401 |
+
"rewards/accuracies": 0.8999999761581421,
|
402 |
+
"rewards/chosen": 2.5646984577178955,
|
403 |
+
"rewards/margins": 2.6955490112304688,
|
404 |
+
"rewards/rejected": -0.13085035979747772,
|
405 |
+
"step": 200
|
406 |
+
},
|
407 |
+
{
|
408 |
+
"epoch": 0.86,
|
409 |
+
"grad_norm": 6.777231693267822,
|
410 |
+
"learning_rate": 4.442105263157896e-06,
|
411 |
+
"logits/chosen": -1.7335002422332764,
|
412 |
+
"logits/rejected": -1.8272641897201538,
|
413 |
+
"logps/chosen": -114.22731018066406,
|
414 |
+
"logps/rejected": -215.8419647216797,
|
415 |
+
"loss": 0.0817,
|
416 |
+
"rewards/accuracies": 1.0,
|
417 |
+
"rewards/chosen": 3.278646469116211,
|
418 |
+
"rewards/margins": 4.009958267211914,
|
419 |
+
"rewards/rejected": -0.7313117980957031,
|
420 |
+
"step": 210
|
421 |
+
},
|
422 |
+
{
|
423 |
+
"epoch": 0.86,
|
424 |
+
"eval_logits/chosen": -1.826608419418335,
|
425 |
+
"eval_logits/rejected": -1.8640671968460083,
|
426 |
+
"eval_logps/chosen": -186.82992553710938,
|
427 |
+
"eval_logps/rejected": -243.80209350585938,
|
428 |
+
"eval_loss": 0.3834719657897949,
|
429 |
+
"eval_rewards/accuracies": 0.7857142686843872,
|
430 |
+
"eval_rewards/chosen": 2.3139071464538574,
|
431 |
+
"eval_rewards/margins": 2.4988515377044678,
|
432 |
+
"eval_rewards/rejected": -0.1849442571401596,
|
433 |
+
"eval_runtime": 26.2309,
|
434 |
+
"eval_samples_per_second": 1.067,
|
435 |
+
"eval_steps_per_second": 1.067,
|
436 |
+
"step": 210
|
437 |
+
},
|
438 |
+
{
|
439 |
+
"epoch": 0.9,
|
440 |
+
"grad_norm": 1.0230931043624878,
|
441 |
+
"learning_rate": 4.40701754385965e-06,
|
442 |
+
"logits/chosen": -1.7561031579971313,
|
443 |
+
"logits/rejected": -1.8601219654083252,
|
444 |
+
"logps/chosen": -134.11428833007812,
|
445 |
+
"logps/rejected": -273.5272216796875,
|
446 |
+
"loss": 0.1796,
|
447 |
+
"rewards/accuracies": 0.8999999761581421,
|
448 |
+
"rewards/chosen": 1.9340410232543945,
|
449 |
+
"rewards/margins": 3.8275578022003174,
|
450 |
+
"rewards/rejected": -1.8935168981552124,
|
451 |
+
"step": 220
|
452 |
+
},
|
453 |
+
{
|
454 |
+
"epoch": 0.94,
|
455 |
+
"grad_norm": 54.65635299682617,
|
456 |
+
"learning_rate": 4.371929824561404e-06,
|
457 |
+
"logits/chosen": -1.8512550592422485,
|
458 |
+
"logits/rejected": -1.900559663772583,
|
459 |
+
"logps/chosen": -210.9802703857422,
|
460 |
+
"logps/rejected": -240.60269165039062,
|
461 |
+
"loss": 0.5004,
|
462 |
+
"rewards/accuracies": 0.800000011920929,
|
463 |
+
"rewards/chosen": 1.5826597213745117,
|
464 |
+
"rewards/margins": 2.925461530685425,
|
465 |
+
"rewards/rejected": -1.342801809310913,
|
466 |
+
"step": 230
|
467 |
+
},
|
468 |
+
{
|
469 |
+
"epoch": 0.98,
|
470 |
+
"grad_norm": 18.797134399414062,
|
471 |
+
"learning_rate": 4.336842105263158e-06,
|
472 |
+
"logits/chosen": -1.7781873941421509,
|
473 |
+
"logits/rejected": -1.8780953884124756,
|
474 |
+
"logps/chosen": -136.18722534179688,
|
475 |
+
"logps/rejected": -257.26495361328125,
|
476 |
+
"loss": 0.137,
|
477 |
+
"rewards/accuracies": 0.8999999761581421,
|
478 |
+
"rewards/chosen": 1.9192733764648438,
|
479 |
+
"rewards/margins": 3.8388113975524902,
|
480 |
+
"rewards/rejected": -1.9195384979248047,
|
481 |
+
"step": 240
|
482 |
+
},
|
483 |
+
{
|
484 |
+
"epoch": 0.98,
|
485 |
+
"eval_logits/chosen": -1.83432936668396,
|
486 |
+
"eval_logits/rejected": -1.8722076416015625,
|
487 |
+
"eval_logps/chosen": -185.8831329345703,
|
488 |
+
"eval_logps/rejected": -244.8594207763672,
|
489 |
+
"eval_loss": 0.41323208808898926,
|
490 |
+
"eval_rewards/accuracies": 0.75,
|
491 |
+
"eval_rewards/chosen": 2.5979421138763428,
|
492 |
+
"eval_rewards/margins": 3.1000888347625732,
|
493 |
+
"eval_rewards/rejected": -0.5021467208862305,
|
494 |
+
"eval_runtime": 26.2323,
|
495 |
+
"eval_samples_per_second": 1.067,
|
496 |
+
"eval_steps_per_second": 1.067,
|
497 |
+
"step": 240
|
498 |
+
},
|
499 |
+
{
|
500 |
+
"epoch": 1.02,
|
501 |
+
"grad_norm": 132.51947021484375,
|
502 |
+
"learning_rate": 4.301754385964912e-06,
|
503 |
+
"logits/chosen": -1.8787914514541626,
|
504 |
+
"logits/rejected": -1.9520515203475952,
|
505 |
+
"logps/chosen": -186.46717834472656,
|
506 |
+
"logps/rejected": -250.008056640625,
|
507 |
+
"loss": 0.3531,
|
508 |
+
"rewards/accuracies": 0.800000011920929,
|
509 |
+
"rewards/chosen": 2.192800283432007,
|
510 |
+
"rewards/margins": 3.485269069671631,
|
511 |
+
"rewards/rejected": -1.2924686670303345,
|
512 |
+
"step": 250
|
513 |
+
},
|
514 |
+
{
|
515 |
+
"epoch": 1.06,
|
516 |
+
"grad_norm": 6.529983043670654,
|
517 |
+
"learning_rate": 4.266666666666668e-06,
|
518 |
+
"logits/chosen": -1.7029941082000732,
|
519 |
+
"logits/rejected": -1.7871618270874023,
|
520 |
+
"logps/chosen": -128.5960235595703,
|
521 |
+
"logps/rejected": -209.27603149414062,
|
522 |
+
"loss": 0.0883,
|
523 |
+
"rewards/accuracies": 0.8999999761581421,
|
524 |
+
"rewards/chosen": 2.7405991554260254,
|
525 |
+
"rewards/margins": 5.148273468017578,
|
526 |
+
"rewards/rejected": -2.4076738357543945,
|
527 |
+
"step": 260
|
528 |
+
},
|
529 |
+
{
|
530 |
+
"epoch": 1.1,
|
531 |
+
"grad_norm": 0.27644646167755127,
|
532 |
+
"learning_rate": 4.2315789473684215e-06,
|
533 |
+
"logits/chosen": -1.7499496936798096,
|
534 |
+
"logits/rejected": -1.8194109201431274,
|
535 |
+
"logps/chosen": -119.34013366699219,
|
536 |
+
"logps/rejected": -230.69229125976562,
|
537 |
+
"loss": 0.0997,
|
538 |
+
"rewards/accuracies": 0.8999999761581421,
|
539 |
+
"rewards/chosen": 4.146358013153076,
|
540 |
+
"rewards/margins": 5.72900390625,
|
541 |
+
"rewards/rejected": -1.5826464891433716,
|
542 |
+
"step": 270
|
543 |
+
},
|
544 |
+
{
|
545 |
+
"epoch": 1.1,
|
546 |
+
"eval_logits/chosen": -1.842972755432129,
|
547 |
+
"eval_logits/rejected": -1.8816190958023071,
|
548 |
+
"eval_logps/chosen": -185.4148406982422,
|
549 |
+
"eval_logps/rejected": -246.53672790527344,
|
550 |
+
"eval_loss": 0.46570485830307007,
|
551 |
+
"eval_rewards/accuracies": 0.75,
|
552 |
+
"eval_rewards/chosen": 2.7384250164031982,
|
553 |
+
"eval_rewards/margins": 3.743762969970703,
|
554 |
+
"eval_rewards/rejected": -1.0053375959396362,
|
555 |
+
"eval_runtime": 26.2348,
|
556 |
+
"eval_samples_per_second": 1.067,
|
557 |
+
"eval_steps_per_second": 1.067,
|
558 |
+
"step": 270
|
559 |
+
},
|
560 |
+
{
|
561 |
+
"epoch": 1.14,
|
562 |
+
"grad_norm": 31.16436195373535,
|
563 |
+
"learning_rate": 4.196491228070176e-06,
|
564 |
+
"logits/chosen": -1.8344509601593018,
|
565 |
+
"logits/rejected": -1.919553518295288,
|
566 |
+
"logps/chosen": -131.51156616210938,
|
567 |
+
"logps/rejected": -246.3345184326172,
|
568 |
+
"loss": 0.138,
|
569 |
+
"rewards/accuracies": 1.0,
|
570 |
+
"rewards/chosen": 2.982485771179199,
|
571 |
+
"rewards/margins": 6.261662483215332,
|
572 |
+
"rewards/rejected": -3.2791759967803955,
|
573 |
+
"step": 280
|
574 |
+
},
|
575 |
+
{
|
576 |
+
"epoch": 1.18,
|
577 |
+
"grad_norm": 0.08183232694864273,
|
578 |
+
"learning_rate": 4.16140350877193e-06,
|
579 |
+
"logits/chosen": -1.847378134727478,
|
580 |
+
"logits/rejected": -1.8774774074554443,
|
581 |
+
"logps/chosen": -127.52336120605469,
|
582 |
+
"logps/rejected": -157.3633270263672,
|
583 |
+
"loss": 0.3883,
|
584 |
+
"rewards/accuracies": 0.800000011920929,
|
585 |
+
"rewards/chosen": 0.7564169764518738,
|
586 |
+
"rewards/margins": 2.541962146759033,
|
587 |
+
"rewards/rejected": -1.7855453491210938,
|
588 |
+
"step": 290
|
589 |
+
},
|
590 |
+
{
|
591 |
+
"epoch": 1.22,
|
592 |
+
"grad_norm": 10.61989974975586,
|
593 |
+
"learning_rate": 4.126315789473685e-06,
|
594 |
+
"logits/chosen": -1.7770633697509766,
|
595 |
+
"logits/rejected": -1.8337452411651611,
|
596 |
+
"logps/chosen": -112.7281265258789,
|
597 |
+
"logps/rejected": -195.87020874023438,
|
598 |
+
"loss": 0.0432,
|
599 |
+
"rewards/accuracies": 1.0,
|
600 |
+
"rewards/chosen": 3.176725387573242,
|
601 |
+
"rewards/margins": 6.761993408203125,
|
602 |
+
"rewards/rejected": -3.5852675437927246,
|
603 |
+
"step": 300
|
604 |
+
},
|
605 |
+
{
|
606 |
+
"epoch": 1.22,
|
607 |
+
"eval_logits/chosen": -1.849483609199524,
|
608 |
+
"eval_logits/rejected": -1.888432502746582,
|
609 |
+
"eval_logps/chosen": -185.52931213378906,
|
610 |
+
"eval_logps/rejected": -248.10926818847656,
|
611 |
+
"eval_loss": 0.501145601272583,
|
612 |
+
"eval_rewards/accuracies": 0.75,
|
613 |
+
"eval_rewards/chosen": 2.704090118408203,
|
614 |
+
"eval_rewards/margins": 4.181191921234131,
|
615 |
+
"eval_rewards/rejected": -1.477101445198059,
|
616 |
+
"eval_runtime": 26.2437,
|
617 |
+
"eval_samples_per_second": 1.067,
|
618 |
+
"eval_steps_per_second": 1.067,
|
619 |
+
"step": 300
|
620 |
+
},
|
621 |
+
{
|
622 |
+
"epoch": 1.27,
|
623 |
+
"grad_norm": 327.02899169921875,
|
624 |
+
"learning_rate": 4.091228070175439e-06,
|
625 |
+
"logits/chosen": -1.8802036046981812,
|
626 |
+
"logits/rejected": -1.9435373544692993,
|
627 |
+
"logps/chosen": -164.2046661376953,
|
628 |
+
"logps/rejected": -269.22235107421875,
|
629 |
+
"loss": 0.4088,
|
630 |
+
"rewards/accuracies": 0.800000011920929,
|
631 |
+
"rewards/chosen": 3.3527073860168457,
|
632 |
+
"rewards/margins": 5.9436140060424805,
|
633 |
+
"rewards/rejected": -2.5909066200256348,
|
634 |
+
"step": 310
|
635 |
+
},
|
636 |
+
{
|
637 |
+
"epoch": 1.31,
|
638 |
+
"grad_norm": 6.565308094024658,
|
639 |
+
"learning_rate": 4.056140350877193e-06,
|
640 |
+
"logits/chosen": -1.9250032901763916,
|
641 |
+
"logits/rejected": -1.9860813617706299,
|
642 |
+
"logps/chosen": -197.4870147705078,
|
643 |
+
"logps/rejected": -207.25009155273438,
|
644 |
+
"loss": 0.4969,
|
645 |
+
"rewards/accuracies": 0.699999988079071,
|
646 |
+
"rewards/chosen": 0.32615354657173157,
|
647 |
+
"rewards/margins": 2.2872400283813477,
|
648 |
+
"rewards/rejected": -1.9610865116119385,
|
649 |
+
"step": 320
|
650 |
+
},
|
651 |
+
{
|
652 |
+
"epoch": 1.35,
|
653 |
+
"grad_norm": 33.34074020385742,
|
654 |
+
"learning_rate": 4.021052631578948e-06,
|
655 |
+
"logits/chosen": -1.8470687866210938,
|
656 |
+
"logits/rejected": -1.8675216436386108,
|
657 |
+
"logps/chosen": -117.5759506225586,
|
658 |
+
"logps/rejected": -151.83091735839844,
|
659 |
+
"loss": 0.1819,
|
660 |
+
"rewards/accuracies": 0.8999999761581421,
|
661 |
+
"rewards/chosen": 2.878262758255005,
|
662 |
+
"rewards/margins": 5.000906467437744,
|
663 |
+
"rewards/rejected": -2.12264347076416,
|
664 |
+
"step": 330
|
665 |
+
},
|
666 |
+
{
|
667 |
+
"epoch": 1.35,
|
668 |
+
"eval_logits/chosen": -1.8486573696136475,
|
669 |
+
"eval_logits/rejected": -1.8877869844436646,
|
670 |
+
"eval_logps/chosen": -185.54177856445312,
|
671 |
+
"eval_logps/rejected": -249.2687530517578,
|
672 |
+
"eval_loss": 0.478495329618454,
|
673 |
+
"eval_rewards/accuracies": 0.75,
|
674 |
+
"eval_rewards/chosen": 2.7003517150878906,
|
675 |
+
"eval_rewards/margins": 4.525294780731201,
|
676 |
+
"eval_rewards/rejected": -1.8249431848526,
|
677 |
+
"eval_runtime": 26.2362,
|
678 |
+
"eval_samples_per_second": 1.067,
|
679 |
+
"eval_steps_per_second": 1.067,
|
680 |
+
"step": 330
|
681 |
+
},
|
682 |
+
{
|
683 |
+
"epoch": 1.39,
|
684 |
+
"grad_norm": 0.6777727603912354,
|
685 |
+
"learning_rate": 3.985964912280702e-06,
|
686 |
+
"logits/chosen": -1.8584426641464233,
|
687 |
+
"logits/rejected": -1.884598970413208,
|
688 |
+
"logps/chosen": -78.69309997558594,
|
689 |
+
"logps/rejected": -126.11468505859375,
|
690 |
+
"loss": 0.3154,
|
691 |
+
"rewards/accuracies": 0.8999999761581421,
|
692 |
+
"rewards/chosen": 2.5596954822540283,
|
693 |
+
"rewards/margins": 4.089261054992676,
|
694 |
+
"rewards/rejected": -1.5295660495758057,
|
695 |
+
"step": 340
|
696 |
+
},
|
697 |
+
{
|
698 |
+
"epoch": 1.43,
|
699 |
+
"grad_norm": 45.252723693847656,
|
700 |
+
"learning_rate": 3.950877192982457e-06,
|
701 |
+
"logits/chosen": -1.8422210216522217,
|
702 |
+
"logits/rejected": -1.8741470575332642,
|
703 |
+
"logps/chosen": -198.17526245117188,
|
704 |
+
"logps/rejected": -179.26759338378906,
|
705 |
+
"loss": 0.2128,
|
706 |
+
"rewards/accuracies": 0.8999999761581421,
|
707 |
+
"rewards/chosen": 2.886838912963867,
|
708 |
+
"rewards/margins": 4.228543758392334,
|
709 |
+
"rewards/rejected": -1.3417048454284668,
|
710 |
+
"step": 350
|
711 |
+
},
|
712 |
+
{
|
713 |
+
"epoch": 1.47,
|
714 |
+
"grad_norm": 0.0007244591251946986,
|
715 |
+
"learning_rate": 3.9157894736842104e-06,
|
716 |
+
"logits/chosen": -1.8951257467269897,
|
717 |
+
"logits/rejected": -1.9684120416641235,
|
718 |
+
"logps/chosen": -123.7894515991211,
|
719 |
+
"logps/rejected": -278.1185607910156,
|
720 |
+
"loss": 0.0169,
|
721 |
+
"rewards/accuracies": 1.0,
|
722 |
+
"rewards/chosen": 3.9708969593048096,
|
723 |
+
"rewards/margins": 9.946398735046387,
|
724 |
+
"rewards/rejected": -5.975502014160156,
|
725 |
+
"step": 360
|
726 |
+
},
|
727 |
+
{
|
728 |
+
"epoch": 1.47,
|
729 |
+
"eval_logits/chosen": -1.8510137796401978,
|
730 |
+
"eval_logits/rejected": -1.8906936645507812,
|
731 |
+
"eval_logps/chosen": -185.66189575195312,
|
732 |
+
"eval_logps/rejected": -250.37806701660156,
|
733 |
+
"eval_loss": 0.4872412383556366,
|
734 |
+
"eval_rewards/accuracies": 0.75,
|
735 |
+
"eval_rewards/chosen": 2.664308547973633,
|
736 |
+
"eval_rewards/margins": 4.822035789489746,
|
737 |
+
"eval_rewards/rejected": -2.157727003097534,
|
738 |
+
"eval_runtime": 26.2354,
|
739 |
+
"eval_samples_per_second": 1.067,
|
740 |
+
"eval_steps_per_second": 1.067,
|
741 |
+
"step": 360
|
742 |
+
},
|
743 |
+
{
|
744 |
+
"epoch": 1.51,
|
745 |
+
"grad_norm": 0.001088384771719575,
|
746 |
+
"learning_rate": 3.880701754385965e-06,
|
747 |
+
"logits/chosen": -1.859140157699585,
|
748 |
+
"logits/rejected": -1.903721809387207,
|
749 |
+
"logps/chosen": -171.89047241210938,
|
750 |
+
"logps/rejected": -257.6295471191406,
|
751 |
+
"loss": 0.237,
|
752 |
+
"rewards/accuracies": 0.800000011920929,
|
753 |
+
"rewards/chosen": 2.257797956466675,
|
754 |
+
"rewards/margins": 7.423349857330322,
|
755 |
+
"rewards/rejected": -5.165551662445068,
|
756 |
+
"step": 370
|
757 |
+
},
|
758 |
+
{
|
759 |
+
"epoch": 1.55,
|
760 |
+
"grad_norm": 13.049817085266113,
|
761 |
+
"learning_rate": 3.84561403508772e-06,
|
762 |
+
"logits/chosen": -1.850102186203003,
|
763 |
+
"logits/rejected": -1.922501564025879,
|
764 |
+
"logps/chosen": -145.50692749023438,
|
765 |
+
"logps/rejected": -221.7799530029297,
|
766 |
+
"loss": 0.2906,
|
767 |
+
"rewards/accuracies": 0.8999999761581421,
|
768 |
+
"rewards/chosen": 1.9433931112289429,
|
769 |
+
"rewards/margins": 6.384174346923828,
|
770 |
+
"rewards/rejected": -4.440781593322754,
|
771 |
+
"step": 380
|
772 |
+
},
|
773 |
+
{
|
774 |
+
"epoch": 1.59,
|
775 |
+
"grad_norm": 0.006105802953243256,
|
776 |
+
"learning_rate": 3.810526315789474e-06,
|
777 |
+
"logits/chosen": -1.9277305603027344,
|
778 |
+
"logits/rejected": -1.9736566543579102,
|
779 |
+
"logps/chosen": -140.36146545410156,
|
780 |
+
"logps/rejected": -188.66836547851562,
|
781 |
+
"loss": 0.235,
|
782 |
+
"rewards/accuracies": 0.8999999761581421,
|
783 |
+
"rewards/chosen": 2.6781504154205322,
|
784 |
+
"rewards/margins": 5.825950622558594,
|
785 |
+
"rewards/rejected": -3.1477997303009033,
|
786 |
+
"step": 390
|
787 |
+
},
|
788 |
+
{
|
789 |
+
"epoch": 1.59,
|
790 |
+
"eval_logits/chosen": -1.8531960248947144,
|
791 |
+
"eval_logits/rejected": -1.8930160999298096,
|
792 |
+
"eval_logps/chosen": -185.68800354003906,
|
793 |
+
"eval_logps/rejected": -251.1302032470703,
|
794 |
+
"eval_loss": 0.48860952258110046,
|
795 |
+
"eval_rewards/accuracies": 0.75,
|
796 |
+
"eval_rewards/chosen": 2.656480550765991,
|
797 |
+
"eval_rewards/margins": 5.039853572845459,
|
798 |
+
"eval_rewards/rejected": -2.3833730220794678,
|
799 |
+
"eval_runtime": 26.2308,
|
800 |
+
"eval_samples_per_second": 1.067,
|
801 |
+
"eval_steps_per_second": 1.067,
|
802 |
+
"step": 390
|
803 |
+
},
|
804 |
+
{
|
805 |
+
"epoch": 1.63,
|
806 |
+
"grad_norm": 0.00030716226319782436,
|
807 |
+
"learning_rate": 3.7754385964912284e-06,
|
808 |
+
"logits/chosen": -1.8817369937896729,
|
809 |
+
"logits/rejected": -1.9124549627304077,
|
810 |
+
"logps/chosen": -171.40383911132812,
|
811 |
+
"logps/rejected": -198.42446899414062,
|
812 |
+
"loss": 0.4322,
|
813 |
+
"rewards/accuracies": 0.800000011920929,
|
814 |
+
"rewards/chosen": 1.219795823097229,
|
815 |
+
"rewards/margins": 6.402298927307129,
|
816 |
+
"rewards/rejected": -5.182502269744873,
|
817 |
+
"step": 400
|
818 |
+
},
|
819 |
+
{
|
820 |
+
"epoch": 1.67,
|
821 |
+
"grad_norm": 0.0007249619229696691,
|
822 |
+
"learning_rate": 3.7403508771929827e-06,
|
823 |
+
"logits/chosen": -1.8018410205841064,
|
824 |
+
"logits/rejected": -1.869511365890503,
|
825 |
+
"logps/chosen": -168.9734649658203,
|
826 |
+
"logps/rejected": -223.9024658203125,
|
827 |
+
"loss": 0.9332,
|
828 |
+
"rewards/accuracies": 0.699999988079071,
|
829 |
+
"rewards/chosen": 2.475252151489258,
|
830 |
+
"rewards/margins": 5.069212913513184,
|
831 |
+
"rewards/rejected": -2.593961238861084,
|
832 |
+
"step": 410
|
833 |
+
},
|
834 |
+
{
|
835 |
+
"epoch": 1.71,
|
836 |
+
"grad_norm": 46.79059982299805,
|
837 |
+
"learning_rate": 3.7052631578947374e-06,
|
838 |
+
"logits/chosen": -1.774857759475708,
|
839 |
+
"logits/rejected": -1.8125171661376953,
|
840 |
+
"logps/chosen": -140.12551879882812,
|
841 |
+
"logps/rejected": -208.7825927734375,
|
842 |
+
"loss": 0.7551,
|
843 |
+
"rewards/accuracies": 0.800000011920929,
|
844 |
+
"rewards/chosen": 2.244128704071045,
|
845 |
+
"rewards/margins": 6.2417426109313965,
|
846 |
+
"rewards/rejected": -3.997614622116089,
|
847 |
+
"step": 420
|
848 |
+
},
|
849 |
+
{
|
850 |
+
"epoch": 1.71,
|
851 |
+
"eval_logits/chosen": -1.8527201414108276,
|
852 |
+
"eval_logits/rejected": -1.892095923423767,
|
853 |
+
"eval_logps/chosen": -185.46646118164062,
|
854 |
+
"eval_logps/rejected": -251.00816345214844,
|
855 |
+
"eval_loss": 0.438032329082489,
|
856 |
+
"eval_rewards/accuracies": 0.75,
|
857 |
+
"eval_rewards/chosen": 2.7229368686676025,
|
858 |
+
"eval_rewards/margins": 5.069703102111816,
|
859 |
+
"eval_rewards/rejected": -2.346766233444214,
|
860 |
+
"eval_runtime": 26.2319,
|
861 |
+
"eval_samples_per_second": 1.067,
|
862 |
+
"eval_steps_per_second": 1.067,
|
863 |
+
"step": 420
|
864 |
+
},
|
865 |
+
{
|
866 |
+
"epoch": 1.76,
|
867 |
+
"grad_norm": 0.1679692417383194,
|
868 |
+
"learning_rate": 3.6701754385964917e-06,
|
869 |
+
"logits/chosen": -1.9766556024551392,
|
870 |
+
"logits/rejected": -1.967247724533081,
|
871 |
+
"logps/chosen": -181.25978088378906,
|
872 |
+
"logps/rejected": -174.0313262939453,
|
873 |
+
"loss": 0.5746,
|
874 |
+
"rewards/accuracies": 0.800000011920929,
|
875 |
+
"rewards/chosen": 2.037475347518921,
|
876 |
+
"rewards/margins": 4.661454200744629,
|
877 |
+
"rewards/rejected": -2.623978853225708,
|
878 |
+
"step": 430
|
879 |
+
},
|
880 |
+
{
|
881 |
+
"epoch": 1.8,
|
882 |
+
"grad_norm": 0.028290700167417526,
|
883 |
+
"learning_rate": 3.635087719298246e-06,
|
884 |
+
"logits/chosen": -1.7324810028076172,
|
885 |
+
"logits/rejected": -1.8381952047348022,
|
886 |
+
"logps/chosen": -118.8620376586914,
|
887 |
+
"logps/rejected": -249.5945587158203,
|
888 |
+
"loss": 0.1919,
|
889 |
+
"rewards/accuracies": 0.8999999761581421,
|
890 |
+
"rewards/chosen": 2.84242582321167,
|
891 |
+
"rewards/margins": 8.061222076416016,
|
892 |
+
"rewards/rejected": -5.218796730041504,
|
893 |
+
"step": 440
|
894 |
+
},
|
895 |
+
{
|
896 |
+
"epoch": 1.84,
|
897 |
+
"grad_norm": 3.435687303543091,
|
898 |
+
"learning_rate": 3.6000000000000003e-06,
|
899 |
+
"logits/chosen": -1.788468599319458,
|
900 |
+
"logits/rejected": -1.8786054849624634,
|
901 |
+
"logps/chosen": -127.3366928100586,
|
902 |
+
"logps/rejected": -223.4063262939453,
|
903 |
+
"loss": 0.134,
|
904 |
+
"rewards/accuracies": 1.0,
|
905 |
+
"rewards/chosen": 0.9258162379264832,
|
906 |
+
"rewards/margins": 6.352460861206055,
|
907 |
+
"rewards/rejected": -5.426644802093506,
|
908 |
+
"step": 450
|
909 |
+
},
|
910 |
+
{
|
911 |
+
"epoch": 1.84,
|
912 |
+
"eval_logits/chosen": -1.8530861139297485,
|
913 |
+
"eval_logits/rejected": -1.8925215005874634,
|
914 |
+
"eval_logps/chosen": -185.6543426513672,
|
915 |
+
"eval_logps/rejected": -251.7077178955078,
|
916 |
+
"eval_loss": 0.4383006691932678,
|
917 |
+
"eval_rewards/accuracies": 0.75,
|
918 |
+
"eval_rewards/chosen": 2.6665782928466797,
|
919 |
+
"eval_rewards/margins": 5.223209857940674,
|
920 |
+
"eval_rewards/rejected": -2.556631565093994,
|
921 |
+
"eval_runtime": 26.2389,
|
922 |
+
"eval_samples_per_second": 1.067,
|
923 |
+
"eval_steps_per_second": 1.067,
|
924 |
+
"step": 450
|
925 |
+
},
|
926 |
+
{
|
927 |
+
"epoch": 1.88,
|
928 |
+
"grad_norm": 0.2778886556625366,
|
929 |
+
"learning_rate": 3.5649122807017545e-06,
|
930 |
+
"logits/chosen": -1.8211870193481445,
|
931 |
+
"logits/rejected": -1.8562345504760742,
|
932 |
+
"logps/chosen": -90.34883117675781,
|
933 |
+
"logps/rejected": -171.23648071289062,
|
934 |
+
"loss": 0.1284,
|
935 |
+
"rewards/accuracies": 1.0,
|
936 |
+
"rewards/chosen": 3.756679058074951,
|
937 |
+
"rewards/margins": 7.474773406982422,
|
938 |
+
"rewards/rejected": -3.7180938720703125,
|
939 |
+
"step": 460
|
940 |
+
},
|
941 |
+
{
|
942 |
+
"epoch": 1.92,
|
943 |
+
"grad_norm": 0.00020280799071770161,
|
944 |
+
"learning_rate": 3.5298245614035093e-06,
|
945 |
+
"logits/chosen": -1.7977367639541626,
|
946 |
+
"logits/rejected": -1.8836923837661743,
|
947 |
+
"logps/chosen": -180.67724609375,
|
948 |
+
"logps/rejected": -215.39608764648438,
|
949 |
+
"loss": 0.5069,
|
950 |
+
"rewards/accuracies": 0.800000011920929,
|
951 |
+
"rewards/chosen": 2.3001179695129395,
|
952 |
+
"rewards/margins": 7.598956108093262,
|
953 |
+
"rewards/rejected": -5.298838138580322,
|
954 |
+
"step": 470
|
955 |
+
},
|
956 |
+
{
|
957 |
+
"epoch": 1.96,
|
958 |
+
"grad_norm": 0.0038736851420253515,
|
959 |
+
"learning_rate": 3.4947368421052635e-06,
|
960 |
+
"logits/chosen": -1.8841644525527954,
|
961 |
+
"logits/rejected": -1.986575722694397,
|
962 |
+
"logps/chosen": -189.42210388183594,
|
963 |
+
"logps/rejected": -368.7510070800781,
|
964 |
+
"loss": 0.0662,
|
965 |
+
"rewards/accuracies": 1.0,
|
966 |
+
"rewards/chosen": 2.894408702850342,
|
967 |
+
"rewards/margins": 9.953689575195312,
|
968 |
+
"rewards/rejected": -7.0592803955078125,
|
969 |
+
"step": 480
|
970 |
+
},
|
971 |
+
{
|
972 |
+
"epoch": 1.96,
|
973 |
+
"eval_logits/chosen": -1.8568991422653198,
|
974 |
+
"eval_logits/rejected": -1.896437406539917,
|
975 |
+
"eval_logps/chosen": -186.01429748535156,
|
976 |
+
"eval_logps/rejected": -252.9164276123047,
|
977 |
+
"eval_loss": 0.44481489062309265,
|
978 |
+
"eval_rewards/accuracies": 0.75,
|
979 |
+
"eval_rewards/chosen": 2.55859112739563,
|
980 |
+
"eval_rewards/margins": 5.477838039398193,
|
981 |
+
"eval_rewards/rejected": -2.9192464351654053,
|
982 |
+
"eval_runtime": 26.2387,
|
983 |
+
"eval_samples_per_second": 1.067,
|
984 |
+
"eval_steps_per_second": 1.067,
|
985 |
+
"step": 480
|
986 |
+
},
|
987 |
+
{
|
988 |
+
"epoch": 2.0,
|
989 |
+
"grad_norm": 0.00020024628611281514,
|
990 |
+
"learning_rate": 3.459649122807018e-06,
|
991 |
+
"logits/chosen": -1.7868010997772217,
|
992 |
+
"logits/rejected": -1.8775966167449951,
|
993 |
+
"logps/chosen": -168.01686096191406,
|
994 |
+
"logps/rejected": -268.81329345703125,
|
995 |
+
"loss": 0.2127,
|
996 |
+
"rewards/accuracies": 0.8999999761581421,
|
997 |
+
"rewards/chosen": 4.091753959655762,
|
998 |
+
"rewards/margins": 10.252196311950684,
|
999 |
+
"rewards/rejected": -6.160443305969238,
|
1000 |
+
"step": 490
|
1001 |
+
},
|
1002 |
+
{
|
1003 |
+
"epoch": 2.04,
|
1004 |
+
"grad_norm": 2.933464288711548,
|
1005 |
+
"learning_rate": 3.424561403508772e-06,
|
1006 |
+
"logits/chosen": -1.8647997379302979,
|
1007 |
+
"logits/rejected": -1.8972787857055664,
|
1008 |
+
"logps/chosen": -206.39712524414062,
|
1009 |
+
"logps/rejected": -255.3307647705078,
|
1010 |
+
"loss": 0.1551,
|
1011 |
+
"rewards/accuracies": 0.8999999761581421,
|
1012 |
+
"rewards/chosen": 3.2640254497528076,
|
1013 |
+
"rewards/margins": 8.8242826461792,
|
1014 |
+
"rewards/rejected": -5.560257434844971,
|
1015 |
+
"step": 500
|
1016 |
+
},
|
1017 |
+
{
|
1018 |
+
"epoch": 2.08,
|
1019 |
+
"grad_norm": 0.088715098798275,
|
1020 |
+
"learning_rate": 3.3894736842105264e-06,
|
1021 |
+
"logits/chosen": -1.8153291940689087,
|
1022 |
+
"logits/rejected": -1.907467246055603,
|
1023 |
+
"logps/chosen": -114.2393569946289,
|
1024 |
+
"logps/rejected": -245.9196319580078,
|
1025 |
+
"loss": 0.1093,
|
1026 |
+
"rewards/accuracies": 1.0,
|
1027 |
+
"rewards/chosen": 1.2050167322158813,
|
1028 |
+
"rewards/margins": 7.797829627990723,
|
1029 |
+
"rewards/rejected": -6.592813014984131,
|
1030 |
+
"step": 510
|
1031 |
+
},
|
1032 |
+
{
|
1033 |
+
"epoch": 2.08,
|
1034 |
+
"eval_logits/chosen": -1.8561475276947021,
|
1035 |
+
"eval_logits/rejected": -1.8955259323120117,
|
1036 |
+
"eval_logps/chosen": -186.13938903808594,
|
1037 |
+
"eval_logps/rejected": -253.4276580810547,
|
1038 |
+
"eval_loss": 0.4262312054634094,
|
1039 |
+
"eval_rewards/accuracies": 0.75,
|
1040 |
+
"eval_rewards/chosen": 2.521066904067993,
|
1041 |
+
"eval_rewards/margins": 5.59368371963501,
|
1042 |
+
"eval_rewards/rejected": -3.0726168155670166,
|
1043 |
+
"eval_runtime": 26.2357,
|
1044 |
+
"eval_samples_per_second": 1.067,
|
1045 |
+
"eval_steps_per_second": 1.067,
|
1046 |
+
"step": 510
|
1047 |
+
},
|
1048 |
+
{
|
1049 |
+
"epoch": 2.12,
|
1050 |
+
"grad_norm": 0.0037095127627253532,
|
1051 |
+
"learning_rate": 3.354385964912281e-06,
|
1052 |
+
"logits/chosen": -1.904370665550232,
|
1053 |
+
"logits/rejected": -1.9216587543487549,
|
1054 |
+
"logps/chosen": -173.86373901367188,
|
1055 |
+
"logps/rejected": -173.6374053955078,
|
1056 |
+
"loss": 0.1557,
|
1057 |
+
"rewards/accuracies": 0.8999999761581421,
|
1058 |
+
"rewards/chosen": 2.4744720458984375,
|
1059 |
+
"rewards/margins": 5.560853004455566,
|
1060 |
+
"rewards/rejected": -3.0863804817199707,
|
1061 |
+
"step": 520
|
1062 |
+
},
|
1063 |
+
{
|
1064 |
+
"epoch": 2.16,
|
1065 |
+
"grad_norm": 1.6992926532566344e-07,
|
1066 |
+
"learning_rate": 3.3192982456140354e-06,
|
1067 |
+
"logits/chosen": -1.7665284872055054,
|
1068 |
+
"logits/rejected": -1.859400987625122,
|
1069 |
+
"logps/chosen": -85.06950378417969,
|
1070 |
+
"logps/rejected": -232.3406524658203,
|
1071 |
+
"loss": 0.0056,
|
1072 |
+
"rewards/accuracies": 1.0,
|
1073 |
+
"rewards/chosen": 3.268359661102295,
|
1074 |
+
"rewards/margins": 9.698001861572266,
|
1075 |
+
"rewards/rejected": -6.4296417236328125,
|
1076 |
+
"step": 530
|
1077 |
+
},
|
1078 |
+
{
|
1079 |
+
"epoch": 2.2,
|
1080 |
+
"grad_norm": 0.00015174441796261817,
|
1081 |
+
"learning_rate": 3.2842105263157897e-06,
|
1082 |
+
"logits/chosen": -1.7254387140274048,
|
1083 |
+
"logits/rejected": -1.8300654888153076,
|
1084 |
+
"logps/chosen": -126.21636962890625,
|
1085 |
+
"logps/rejected": -236.91250610351562,
|
1086 |
+
"loss": 0.1557,
|
1087 |
+
"rewards/accuracies": 0.8999999761581421,
|
1088 |
+
"rewards/chosen": 2.9851126670837402,
|
1089 |
+
"rewards/margins": 9.359209060668945,
|
1090 |
+
"rewards/rejected": -6.374096870422363,
|
1091 |
+
"step": 540
|
1092 |
+
},
|
1093 |
+
{
|
1094 |
+
"epoch": 2.2,
|
1095 |
+
"eval_logits/chosen": -1.85663902759552,
|
1096 |
+
"eval_logits/rejected": -1.8965102434158325,
|
1097 |
+
"eval_logps/chosen": -186.6448974609375,
|
1098 |
+
"eval_logps/rejected": -254.58482360839844,
|
1099 |
+
"eval_loss": 0.4263899326324463,
|
1100 |
+
"eval_rewards/accuracies": 0.75,
|
1101 |
+
"eval_rewards/chosen": 2.369408369064331,
|
1102 |
+
"eval_rewards/margins": 5.789170265197754,
|
1103 |
+
"eval_rewards/rejected": -3.4197616577148438,
|
1104 |
+
"eval_runtime": 26.2541,
|
1105 |
+
"eval_samples_per_second": 1.066,
|
1106 |
+
"eval_steps_per_second": 1.066,
|
1107 |
+
"step": 540
|
1108 |
+
},
|
1109 |
+
{
|
1110 |
+
"epoch": 2.24,
|
1111 |
+
"grad_norm": 7.969775676727295,
|
1112 |
+
"learning_rate": 3.249122807017544e-06,
|
1113 |
+
"logits/chosen": -1.8786284923553467,
|
1114 |
+
"logits/rejected": -1.9631255865097046,
|
1115 |
+
"logps/chosen": -186.79092407226562,
|
1116 |
+
"logps/rejected": -295.82598876953125,
|
1117 |
+
"loss": 0.0625,
|
1118 |
+
"rewards/accuracies": 1.0,
|
1119 |
+
"rewards/chosen": 0.5664806365966797,
|
1120 |
+
"rewards/margins": 6.786992073059082,
|
1121 |
+
"rewards/rejected": -6.220510959625244,
|
1122 |
+
"step": 550
|
1123 |
+
},
|
1124 |
+
{
|
1125 |
+
"epoch": 2.29,
|
1126 |
+
"grad_norm": 41.887001037597656,
|
1127 |
+
"learning_rate": 3.2140350877192982e-06,
|
1128 |
+
"logits/chosen": -1.8324702978134155,
|
1129 |
+
"logits/rejected": -1.8570420742034912,
|
1130 |
+
"logps/chosen": -198.67893981933594,
|
1131 |
+
"logps/rejected": -203.76016235351562,
|
1132 |
+
"loss": 0.389,
|
1133 |
+
"rewards/accuracies": 0.8999999761581421,
|
1134 |
+
"rewards/chosen": 2.470458984375,
|
1135 |
+
"rewards/margins": 8.410517692565918,
|
1136 |
+
"rewards/rejected": -5.940059185028076,
|
1137 |
+
"step": 560
|
1138 |
+
},
|
1139 |
+
{
|
1140 |
+
"epoch": 2.33,
|
1141 |
+
"grad_norm": 0.022604621946811676,
|
1142 |
+
"learning_rate": 3.178947368421053e-06,
|
1143 |
+
"logits/chosen": -1.7625017166137695,
|
1144 |
+
"logits/rejected": -1.823395013809204,
|
1145 |
+
"logps/chosen": -145.4344024658203,
|
1146 |
+
"logps/rejected": -202.81698608398438,
|
1147 |
+
"loss": 0.0962,
|
1148 |
+
"rewards/accuracies": 0.8999999761581421,
|
1149 |
+
"rewards/chosen": 3.2409331798553467,
|
1150 |
+
"rewards/margins": 7.634140968322754,
|
1151 |
+
"rewards/rejected": -4.39320707321167,
|
1152 |
+
"step": 570
|
1153 |
+
},
|
1154 |
+
{
|
1155 |
+
"epoch": 2.33,
|
1156 |
+
"eval_logits/chosen": -1.8582419157028198,
|
1157 |
+
"eval_logits/rejected": -1.897825837135315,
|
1158 |
+
"eval_logps/chosen": -186.9964141845703,
|
1159 |
+
"eval_logps/rejected": -255.54441833496094,
|
1160 |
+
"eval_loss": 0.41822776198387146,
|
1161 |
+
"eval_rewards/accuracies": 0.75,
|
1162 |
+
"eval_rewards/chosen": 2.2639553546905518,
|
1163 |
+
"eval_rewards/margins": 5.971593379974365,
|
1164 |
+
"eval_rewards/rejected": -3.70763897895813,
|
1165 |
+
"eval_runtime": 26.2352,
|
1166 |
+
"eval_samples_per_second": 1.067,
|
1167 |
+
"eval_steps_per_second": 1.067,
|
1168 |
+
"step": 570
|
1169 |
+
},
|
1170 |
+
{
|
1171 |
+
"epoch": 2.37,
|
1172 |
+
"grad_norm": 0.0053796968422830105,
|
1173 |
+
"learning_rate": 3.1438596491228072e-06,
|
1174 |
+
"logits/chosen": -1.8564494848251343,
|
1175 |
+
"logits/rejected": -1.9009406566619873,
|
1176 |
+
"logps/chosen": -153.97903442382812,
|
1177 |
+
"logps/rejected": -182.94284057617188,
|
1178 |
+
"loss": 0.5733,
|
1179 |
+
"rewards/accuracies": 0.800000011920929,
|
1180 |
+
"rewards/chosen": 2.0778138637542725,
|
1181 |
+
"rewards/margins": 5.466890335083008,
|
1182 |
+
"rewards/rejected": -3.389075756072998,
|
1183 |
+
"step": 580
|
1184 |
+
},
|
1185 |
+
{
|
1186 |
+
"epoch": 2.41,
|
1187 |
+
"grad_norm": 5.523962020874023,
|
1188 |
+
"learning_rate": 3.1087719298245615e-06,
|
1189 |
+
"logits/chosen": -1.8511816263198853,
|
1190 |
+
"logits/rejected": -1.853109359741211,
|
1191 |
+
"logps/chosen": -182.23892211914062,
|
1192 |
+
"logps/rejected": -120.5928955078125,
|
1193 |
+
"loss": 0.4958,
|
1194 |
+
"rewards/accuracies": 0.800000011920929,
|
1195 |
+
"rewards/chosen": 3.659731388092041,
|
1196 |
+
"rewards/margins": 4.932890892028809,
|
1197 |
+
"rewards/rejected": -1.2731596231460571,
|
1198 |
+
"step": 590
|
1199 |
+
},
|
1200 |
+
{
|
1201 |
+
"epoch": 2.45,
|
1202 |
+
"grad_norm": 0.0013734139502048492,
|
1203 |
+
"learning_rate": 3.0736842105263158e-06,
|
1204 |
+
"logits/chosen": -1.7618497610092163,
|
1205 |
+
"logits/rejected": -1.8786863088607788,
|
1206 |
+
"logps/chosen": -114.43098449707031,
|
1207 |
+
"logps/rejected": -229.5573272705078,
|
1208 |
+
"loss": 0.0437,
|
1209 |
+
"rewards/accuracies": 1.0,
|
1210 |
+
"rewards/chosen": 2.8845677375793457,
|
1211 |
+
"rewards/margins": 9.519918441772461,
|
1212 |
+
"rewards/rejected": -6.635350227355957,
|
1213 |
+
"step": 600
|
1214 |
+
},
|
1215 |
+
{
|
1216 |
+
"epoch": 2.45,
|
1217 |
+
"eval_logits/chosen": -1.8533600568771362,
|
1218 |
+
"eval_logits/rejected": -1.8932533264160156,
|
1219 |
+
"eval_logps/chosen": -187.0037384033203,
|
1220 |
+
"eval_logps/rejected": -255.77125549316406,
|
1221 |
+
"eval_loss": 0.3824370801448822,
|
1222 |
+
"eval_rewards/accuracies": 0.75,
|
1223 |
+
"eval_rewards/chosen": 2.2617576122283936,
|
1224 |
+
"eval_rewards/margins": 6.037452220916748,
|
1225 |
+
"eval_rewards/rejected": -3.7756943702697754,
|
1226 |
+
"eval_runtime": 26.2326,
|
1227 |
+
"eval_samples_per_second": 1.067,
|
1228 |
+
"eval_steps_per_second": 1.067,
|
1229 |
+
"step": 600
|
1230 |
+
},
|
1231 |
+
{
|
1232 |
+
"epoch": 2.49,
|
1233 |
+
"grad_norm": 3.847404956817627,
|
1234 |
+
"learning_rate": 3.03859649122807e-06,
|
1235 |
+
"logits/chosen": -1.9108350276947021,
|
1236 |
+
"logits/rejected": -1.9136813879013062,
|
1237 |
+
"logps/chosen": -147.76666259765625,
|
1238 |
+
"logps/rejected": -151.54275512695312,
|
1239 |
+
"loss": 0.1939,
|
1240 |
+
"rewards/accuracies": 0.8999999761581421,
|
1241 |
+
"rewards/chosen": 2.7343037128448486,
|
1242 |
+
"rewards/margins": 7.082615852355957,
|
1243 |
+
"rewards/rejected": -4.348311424255371,
|
1244 |
+
"step": 610
|
1245 |
+
},
|
1246 |
+
{
|
1247 |
+
"epoch": 2.53,
|
1248 |
+
"grad_norm": 6.025547172328061e-09,
|
1249 |
+
"learning_rate": 3.003508771929825e-06,
|
1250 |
+
"logits/chosen": -1.9030263423919678,
|
1251 |
+
"logits/rejected": -1.964647889137268,
|
1252 |
+
"logps/chosen": -157.2318115234375,
|
1253 |
+
"logps/rejected": -335.1673583984375,
|
1254 |
+
"loss": 0.1472,
|
1255 |
+
"rewards/accuracies": 0.8999999761581421,
|
1256 |
+
"rewards/chosen": 0.17058677971363068,
|
1257 |
+
"rewards/margins": 10.464014053344727,
|
1258 |
+
"rewards/rejected": -10.293428421020508,
|
1259 |
+
"step": 620
|
1260 |
+
},
|
1261 |
+
{
|
1262 |
+
"epoch": 2.57,
|
1263 |
+
"grad_norm": 0.7363545298576355,
|
1264 |
+
"learning_rate": 2.9684210526315795e-06,
|
1265 |
+
"logits/chosen": -1.8167072534561157,
|
1266 |
+
"logits/rejected": -1.9096784591674805,
|
1267 |
+
"logps/chosen": -128.2840118408203,
|
1268 |
+
"logps/rejected": -278.52862548828125,
|
1269 |
+
"loss": 0.0278,
|
1270 |
+
"rewards/accuracies": 1.0,
|
1271 |
+
"rewards/chosen": 2.6698997020721436,
|
1272 |
+
"rewards/margins": 10.468381881713867,
|
1273 |
+
"rewards/rejected": -7.798482418060303,
|
1274 |
+
"step": 630
|
1275 |
+
},
|
1276 |
+
{
|
1277 |
+
"epoch": 2.57,
|
1278 |
+
"eval_logits/chosen": -1.8536275625228882,
|
1279 |
+
"eval_logits/rejected": -1.8932437896728516,
|
1280 |
+
"eval_logps/chosen": -186.7086181640625,
|
1281 |
+
"eval_logps/rejected": -255.70458984375,
|
1282 |
+
"eval_loss": 0.3570733964443207,
|
1283 |
+
"eval_rewards/accuracies": 0.8571428656578064,
|
1284 |
+
"eval_rewards/chosen": 2.3502955436706543,
|
1285 |
+
"eval_rewards/margins": 6.105983257293701,
|
1286 |
+
"eval_rewards/rejected": -3.755687713623047,
|
1287 |
+
"eval_runtime": 26.2519,
|
1288 |
+
"eval_samples_per_second": 1.067,
|
1289 |
+
"eval_steps_per_second": 1.067,
|
1290 |
+
"step": 630
|
1291 |
+
},
|
1292 |
+
{
|
1293 |
+
"epoch": 2.61,
|
1294 |
+
"grad_norm": 19.793140411376953,
|
1295 |
+
"learning_rate": 2.9333333333333338e-06,
|
1296 |
+
"logits/chosen": -1.9250261783599854,
|
1297 |
+
"logits/rejected": -1.969919204711914,
|
1298 |
+
"logps/chosen": -169.236572265625,
|
1299 |
+
"logps/rejected": -216.4037322998047,
|
1300 |
+
"loss": 0.3144,
|
1301 |
+
"rewards/accuracies": 0.800000011920929,
|
1302 |
+
"rewards/chosen": 1.5923322439193726,
|
1303 |
+
"rewards/margins": 8.081656455993652,
|
1304 |
+
"rewards/rejected": -6.48932409286499,
|
1305 |
+
"step": 640
|
1306 |
+
},
|
1307 |
+
{
|
1308 |
+
"epoch": 2.65,
|
1309 |
+
"grad_norm": 0.0,
|
1310 |
+
"learning_rate": 2.898245614035088e-06,
|
1311 |
+
"logits/chosen": -1.840488076210022,
|
1312 |
+
"logits/rejected": -1.9099540710449219,
|
1313 |
+
"logps/chosen": -188.36514282226562,
|
1314 |
+
"logps/rejected": -266.2427978515625,
|
1315 |
+
"loss": 0.0455,
|
1316 |
+
"rewards/accuracies": 1.0,
|
1317 |
+
"rewards/chosen": 3.294522762298584,
|
1318 |
+
"rewards/margins": 9.861984252929688,
|
1319 |
+
"rewards/rejected": -6.5674614906311035,
|
1320 |
+
"step": 650
|
1321 |
+
},
|
1322 |
+
{
|
1323 |
+
"epoch": 2.69,
|
1324 |
+
"grad_norm": 350.0583801269531,
|
1325 |
+
"learning_rate": 2.8631578947368423e-06,
|
1326 |
+
"logits/chosen": -1.875605583190918,
|
1327 |
+
"logits/rejected": -1.9353086948394775,
|
1328 |
+
"logps/chosen": -128.51104736328125,
|
1329 |
+
"logps/rejected": -191.3477325439453,
|
1330 |
+
"loss": 0.2399,
|
1331 |
+
"rewards/accuracies": 0.8999999761581421,
|
1332 |
+
"rewards/chosen": 3.1921534538269043,
|
1333 |
+
"rewards/margins": 7.449667453765869,
|
1334 |
+
"rewards/rejected": -4.257513999938965,
|
1335 |
+
"step": 660
|
1336 |
+
},
|
1337 |
+
{
|
1338 |
+
"epoch": 2.69,
|
1339 |
+
"eval_logits/chosen": -1.851164698600769,
|
1340 |
+
"eval_logits/rejected": -1.8909271955490112,
|
1341 |
+
"eval_logps/chosen": -186.86781311035156,
|
1342 |
+
"eval_logps/rejected": -256.2709655761719,
|
1343 |
+
"eval_loss": 0.3312711715698242,
|
1344 |
+
"eval_rewards/accuracies": 0.8571428656578064,
|
1345 |
+
"eval_rewards/chosen": 2.3025403022766113,
|
1346 |
+
"eval_rewards/margins": 6.228146553039551,
|
1347 |
+
"eval_rewards/rejected": -3.9256062507629395,
|
1348 |
+
"eval_runtime": 26.2337,
|
1349 |
+
"eval_samples_per_second": 1.067,
|
1350 |
+
"eval_steps_per_second": 1.067,
|
1351 |
+
"step": 660
|
1352 |
+
},
|
1353 |
+
{
|
1354 |
+
"epoch": 2.73,
|
1355 |
+
"grad_norm": 0.0,
|
1356 |
+
"learning_rate": 2.828070175438597e-06,
|
1357 |
+
"logits/chosen": -1.7718509435653687,
|
1358 |
+
"logits/rejected": -1.8662782907485962,
|
1359 |
+
"logps/chosen": -100.0200424194336,
|
1360 |
+
"logps/rejected": -176.11337280273438,
|
1361 |
+
"loss": 0.0636,
|
1362 |
+
"rewards/accuracies": 1.0,
|
1363 |
+
"rewards/chosen": 2.2194771766662598,
|
1364 |
+
"rewards/margins": 7.908571720123291,
|
1365 |
+
"rewards/rejected": -5.689094543457031,
|
1366 |
+
"step": 670
|
1367 |
+
},
|
1368 |
+
{
|
1369 |
+
"epoch": 2.78,
|
1370 |
+
"grad_norm": 9.849269554251805e-05,
|
1371 |
+
"learning_rate": 2.7929824561403513e-06,
|
1372 |
+
"logits/chosen": -1.8379380702972412,
|
1373 |
+
"logits/rejected": -1.9032844305038452,
|
1374 |
+
"logps/chosen": -147.157958984375,
|
1375 |
+
"logps/rejected": -311.342529296875,
|
1376 |
+
"loss": 0.0081,
|
1377 |
+
"rewards/accuracies": 1.0,
|
1378 |
+
"rewards/chosen": 1.7090831995010376,
|
1379 |
+
"rewards/margins": 9.63591480255127,
|
1380 |
+
"rewards/rejected": -7.926831245422363,
|
1381 |
+
"step": 680
|
1382 |
+
},
|
1383 |
+
{
|
1384 |
+
"epoch": 2.82,
|
1385 |
+
"grad_norm": 5.133792877197266,
|
1386 |
+
"learning_rate": 2.7578947368421056e-06,
|
1387 |
+
"logits/chosen": -1.8350670337677002,
|
1388 |
+
"logits/rejected": -1.9039970636367798,
|
1389 |
+
"logps/chosen": -160.75314331054688,
|
1390 |
+
"logps/rejected": -247.68008422851562,
|
1391 |
+
"loss": 0.039,
|
1392 |
+
"rewards/accuracies": 1.0,
|
1393 |
+
"rewards/chosen": 4.146931171417236,
|
1394 |
+
"rewards/margins": 11.529654502868652,
|
1395 |
+
"rewards/rejected": -7.382723808288574,
|
1396 |
+
"step": 690
|
1397 |
+
},
|
1398 |
+
{
|
1399 |
+
"epoch": 2.82,
|
1400 |
+
"eval_logits/chosen": -1.8510338068008423,
|
1401 |
+
"eval_logits/rejected": -1.8906004428863525,
|
1402 |
+
"eval_logps/chosen": -187.16348266601562,
|
1403 |
+
"eval_logps/rejected": -257.069091796875,
|
1404 |
+
"eval_loss": 0.3131271302700043,
|
1405 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1406 |
+
"eval_rewards/chosen": 2.213836193084717,
|
1407 |
+
"eval_rewards/margins": 6.378880977630615,
|
1408 |
+
"eval_rewards/rejected": -4.165044784545898,
|
1409 |
+
"eval_runtime": 26.2333,
|
1410 |
+
"eval_samples_per_second": 1.067,
|
1411 |
+
"eval_steps_per_second": 1.067,
|
1412 |
+
"step": 690
|
1413 |
+
},
|
1414 |
+
{
|
1415 |
+
"epoch": 2.86,
|
1416 |
+
"grad_norm": 0.001812128466553986,
|
1417 |
+
"learning_rate": 2.72280701754386e-06,
|
1418 |
+
"logits/chosen": -1.888392448425293,
|
1419 |
+
"logits/rejected": -1.9376449584960938,
|
1420 |
+
"logps/chosen": -123.4749755859375,
|
1421 |
+
"logps/rejected": -221.9653778076172,
|
1422 |
+
"loss": 0.1581,
|
1423 |
+
"rewards/accuracies": 0.8999999761581421,
|
1424 |
+
"rewards/chosen": 3.1250205039978027,
|
1425 |
+
"rewards/margins": 9.35824966430664,
|
1426 |
+
"rewards/rejected": -6.23322868347168,
|
1427 |
+
"step": 700
|
1428 |
+
},
|
1429 |
+
{
|
1430 |
+
"epoch": 2.9,
|
1431 |
+
"grad_norm": 0.0007696045213378966,
|
1432 |
+
"learning_rate": 2.687719298245614e-06,
|
1433 |
+
"logits/chosen": -1.7869523763656616,
|
1434 |
+
"logits/rejected": -1.9070037603378296,
|
1435 |
+
"logps/chosen": -116.8504409790039,
|
1436 |
+
"logps/rejected": -300.51806640625,
|
1437 |
+
"loss": 0.1832,
|
1438 |
+
"rewards/accuracies": 0.8999999761581421,
|
1439 |
+
"rewards/chosen": 2.3490161895751953,
|
1440 |
+
"rewards/margins": 10.493500709533691,
|
1441 |
+
"rewards/rejected": -8.144485473632812,
|
1442 |
+
"step": 710
|
1443 |
+
},
|
1444 |
+
{
|
1445 |
+
"epoch": 2.94,
|
1446 |
+
"grad_norm": 83.29313659667969,
|
1447 |
+
"learning_rate": 2.652631578947369e-06,
|
1448 |
+
"logits/chosen": -1.8838169574737549,
|
1449 |
+
"logits/rejected": -1.9118610620498657,
|
1450 |
+
"logps/chosen": -104.75813293457031,
|
1451 |
+
"logps/rejected": -167.18580627441406,
|
1452 |
+
"loss": 0.3389,
|
1453 |
+
"rewards/accuracies": 0.8999999761581421,
|
1454 |
+
"rewards/chosen": 2.7628464698791504,
|
1455 |
+
"rewards/margins": 6.317173957824707,
|
1456 |
+
"rewards/rejected": -3.5543277263641357,
|
1457 |
+
"step": 720
|
1458 |
+
},
|
1459 |
+
{
|
1460 |
+
"epoch": 2.94,
|
1461 |
+
"eval_logits/chosen": -1.8480366468429565,
|
1462 |
+
"eval_logits/rejected": -1.8872734308242798,
|
1463 |
+
"eval_logps/chosen": -187.00791931152344,
|
1464 |
+
"eval_logps/rejected": -257.2390441894531,
|
1465 |
+
"eval_loss": 0.27630648016929626,
|
1466 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1467 |
+
"eval_rewards/chosen": 2.2605104446411133,
|
1468 |
+
"eval_rewards/margins": 6.476548194885254,
|
1469 |
+
"eval_rewards/rejected": -4.216037750244141,
|
1470 |
+
"eval_runtime": 26.23,
|
1471 |
+
"eval_samples_per_second": 1.067,
|
1472 |
+
"eval_steps_per_second": 1.067,
|
1473 |
+
"step": 720
|
1474 |
+
},
|
1475 |
+
{
|
1476 |
+
"epoch": 2.98,
|
1477 |
+
"grad_norm": 0.030682619661092758,
|
1478 |
+
"learning_rate": 2.617543859649123e-06,
|
1479 |
+
"logits/chosen": -1.8667634725570679,
|
1480 |
+
"logits/rejected": -1.8890842199325562,
|
1481 |
+
"logps/chosen": -197.29690551757812,
|
1482 |
+
"logps/rejected": -219.12332153320312,
|
1483 |
+
"loss": 0.617,
|
1484 |
+
"rewards/accuracies": 0.800000011920929,
|
1485 |
+
"rewards/chosen": -0.587865948677063,
|
1486 |
+
"rewards/margins": 5.496161937713623,
|
1487 |
+
"rewards/rejected": -6.0840277671813965,
|
1488 |
+
"step": 730
|
1489 |
+
},
|
1490 |
+
{
|
1491 |
+
"epoch": 3.02,
|
1492 |
+
"grad_norm": 0.005717798601835966,
|
1493 |
+
"learning_rate": 2.5824561403508774e-06,
|
1494 |
+
"logits/chosen": -1.8484325408935547,
|
1495 |
+
"logits/rejected": -1.8692123889923096,
|
1496 |
+
"logps/chosen": -102.5673599243164,
|
1497 |
+
"logps/rejected": -141.8745880126953,
|
1498 |
+
"loss": 0.044,
|
1499 |
+
"rewards/accuracies": 1.0,
|
1500 |
+
"rewards/chosen": 2.209364175796509,
|
1501 |
+
"rewards/margins": 8.334315299987793,
|
1502 |
+
"rewards/rejected": -6.124951362609863,
|
1503 |
+
"step": 740
|
1504 |
+
},
|
1505 |
+
{
|
1506 |
+
"epoch": 3.06,
|
1507 |
+
"grad_norm": 0.0006890746881254017,
|
1508 |
+
"learning_rate": 2.5473684210526317e-06,
|
1509 |
+
"logits/chosen": -1.865378737449646,
|
1510 |
+
"logits/rejected": -1.9292535781860352,
|
1511 |
+
"logps/chosen": -103.72840881347656,
|
1512 |
+
"logps/rejected": -203.67276000976562,
|
1513 |
+
"loss": 0.0154,
|
1514 |
+
"rewards/accuracies": 1.0,
|
1515 |
+
"rewards/chosen": 3.9861044883728027,
|
1516 |
+
"rewards/margins": 8.630514144897461,
|
1517 |
+
"rewards/rejected": -4.644408702850342,
|
1518 |
+
"step": 750
|
1519 |
+
},
|
1520 |
+
{
|
1521 |
+
"epoch": 3.06,
|
1522 |
+
"eval_logits/chosen": -1.846961259841919,
|
1523 |
+
"eval_logits/rejected": -1.886168122291565,
|
1524 |
+
"eval_logps/chosen": -187.0342254638672,
|
1525 |
+
"eval_logps/rejected": -257.52471923828125,
|
1526 |
+
"eval_loss": 0.2703961730003357,
|
1527 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1528 |
+
"eval_rewards/chosen": 2.252613067626953,
|
1529 |
+
"eval_rewards/margins": 6.554351806640625,
|
1530 |
+
"eval_rewards/rejected": -4.301738739013672,
|
1531 |
+
"eval_runtime": 26.236,
|
1532 |
+
"eval_samples_per_second": 1.067,
|
1533 |
+
"eval_steps_per_second": 1.067,
|
1534 |
+
"step": 750
|
1535 |
+
},
|
1536 |
+
{
|
1537 |
+
"epoch": 3.1,
|
1538 |
+
"grad_norm": 0.0,
|
1539 |
+
"learning_rate": 2.512280701754386e-06,
|
1540 |
+
"logits/chosen": -1.921618103981018,
|
1541 |
+
"logits/rejected": -1.9239141941070557,
|
1542 |
+
"logps/chosen": -248.826416015625,
|
1543 |
+
"logps/rejected": -182.67323303222656,
|
1544 |
+
"loss": 0.5385,
|
1545 |
+
"rewards/accuracies": 0.699999988079071,
|
1546 |
+
"rewards/chosen": -0.7813287377357483,
|
1547 |
+
"rewards/margins": 4.614243507385254,
|
1548 |
+
"rewards/rejected": -5.395572185516357,
|
1549 |
+
"step": 760
|
1550 |
+
},
|
1551 |
+
{
|
1552 |
+
"epoch": 3.14,
|
1553 |
+
"grad_norm": 20.82038688659668,
|
1554 |
+
"learning_rate": 2.4771929824561407e-06,
|
1555 |
+
"logits/chosen": -1.8723421096801758,
|
1556 |
+
"logits/rejected": -1.9096221923828125,
|
1557 |
+
"logps/chosen": -144.3488311767578,
|
1558 |
+
"logps/rejected": -200.52149963378906,
|
1559 |
+
"loss": 0.0473,
|
1560 |
+
"rewards/accuracies": 1.0,
|
1561 |
+
"rewards/chosen": 0.6265076994895935,
|
1562 |
+
"rewards/margins": 5.623400688171387,
|
1563 |
+
"rewards/rejected": -4.996891975402832,
|
1564 |
+
"step": 770
|
1565 |
+
},
|
1566 |
+
{
|
1567 |
+
"epoch": 3.18,
|
1568 |
+
"grad_norm": 0.0,
|
1569 |
+
"learning_rate": 2.442105263157895e-06,
|
1570 |
+
"logits/chosen": -1.967357873916626,
|
1571 |
+
"logits/rejected": -1.995941162109375,
|
1572 |
+
"logps/chosen": -169.40643310546875,
|
1573 |
+
"logps/rejected": -225.83267211914062,
|
1574 |
+
"loss": 0.021,
|
1575 |
+
"rewards/accuracies": 1.0,
|
1576 |
+
"rewards/chosen": 3.36498761177063,
|
1577 |
+
"rewards/margins": 8.846434593200684,
|
1578 |
+
"rewards/rejected": -5.481447219848633,
|
1579 |
+
"step": 780
|
1580 |
+
},
|
1581 |
+
{
|
1582 |
+
"epoch": 3.18,
|
1583 |
+
"eval_logits/chosen": -1.84483802318573,
|
1584 |
+
"eval_logits/rejected": -1.8837740421295166,
|
1585 |
+
"eval_logps/chosen": -187.02699279785156,
|
1586 |
+
"eval_logps/rejected": -257.6650085449219,
|
1587 |
+
"eval_loss": 0.24223719537258148,
|
1588 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1589 |
+
"eval_rewards/chosen": 2.254775285720825,
|
1590 |
+
"eval_rewards/margins": 6.598587989807129,
|
1591 |
+
"eval_rewards/rejected": -4.343813419342041,
|
1592 |
+
"eval_runtime": 26.2524,
|
1593 |
+
"eval_samples_per_second": 1.067,
|
1594 |
+
"eval_steps_per_second": 1.067,
|
1595 |
+
"step": 780
|
1596 |
+
},
|
1597 |
+
{
|
1598 |
+
"epoch": 3.22,
|
1599 |
+
"grad_norm": 0.3396748900413513,
|
1600 |
+
"learning_rate": 2.4070175438596493e-06,
|
1601 |
+
"logits/chosen": -1.795336127281189,
|
1602 |
+
"logits/rejected": -1.9104626178741455,
|
1603 |
+
"logps/chosen": -127.96372985839844,
|
1604 |
+
"logps/rejected": -321.76751708984375,
|
1605 |
+
"loss": 0.0057,
|
1606 |
+
"rewards/accuracies": 1.0,
|
1607 |
+
"rewards/chosen": 1.3614065647125244,
|
1608 |
+
"rewards/margins": 11.285585403442383,
|
1609 |
+
"rewards/rejected": -9.924179077148438,
|
1610 |
+
"step": 790
|
1611 |
+
},
|
1612 |
+
{
|
1613 |
+
"epoch": 3.27,
|
1614 |
+
"grad_norm": 89.25263214111328,
|
1615 |
+
"learning_rate": 2.3719298245614036e-06,
|
1616 |
+
"logits/chosen": -1.7844104766845703,
|
1617 |
+
"logits/rejected": -1.835338830947876,
|
1618 |
+
"logps/chosen": -138.7672119140625,
|
1619 |
+
"logps/rejected": -212.5260772705078,
|
1620 |
+
"loss": 0.099,
|
1621 |
+
"rewards/accuracies": 0.8999999761581421,
|
1622 |
+
"rewards/chosen": 3.200477123260498,
|
1623 |
+
"rewards/margins": 9.856096267700195,
|
1624 |
+
"rewards/rejected": -6.655619144439697,
|
1625 |
+
"step": 800
|
1626 |
+
},
|
1627 |
+
{
|
1628 |
+
"epoch": 3.31,
|
1629 |
+
"grad_norm": 0.0014283270575106144,
|
1630 |
+
"learning_rate": 2.3368421052631583e-06,
|
1631 |
+
"logits/chosen": -1.8158670663833618,
|
1632 |
+
"logits/rejected": -1.896251916885376,
|
1633 |
+
"logps/chosen": -144.15419006347656,
|
1634 |
+
"logps/rejected": -283.3548583984375,
|
1635 |
+
"loss": 0.0614,
|
1636 |
+
"rewards/accuracies": 1.0,
|
1637 |
+
"rewards/chosen": 3.173565149307251,
|
1638 |
+
"rewards/margins": 12.303560256958008,
|
1639 |
+
"rewards/rejected": -9.129995346069336,
|
1640 |
+
"step": 810
|
1641 |
+
},
|
1642 |
+
{
|
1643 |
+
"epoch": 3.31,
|
1644 |
+
"eval_logits/chosen": -1.8417214155197144,
|
1645 |
+
"eval_logits/rejected": -1.8804857730865479,
|
1646 |
+
"eval_logps/chosen": -187.0991668701172,
|
1647 |
+
"eval_logps/rejected": -258.0171813964844,
|
1648 |
+
"eval_loss": 0.2144124060869217,
|
1649 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1650 |
+
"eval_rewards/chosen": 2.2331290245056152,
|
1651 |
+
"eval_rewards/margins": 6.682602882385254,
|
1652 |
+
"eval_rewards/rejected": -4.449473857879639,
|
1653 |
+
"eval_runtime": 26.2397,
|
1654 |
+
"eval_samples_per_second": 1.067,
|
1655 |
+
"eval_steps_per_second": 1.067,
|
1656 |
+
"step": 810
|
1657 |
+
},
|
1658 |
+
{
|
1659 |
+
"epoch": 3.35,
|
1660 |
+
"grad_norm": 8.28421070764307e-06,
|
1661 |
+
"learning_rate": 2.3017543859649126e-06,
|
1662 |
+
"logits/chosen": -1.7904294729232788,
|
1663 |
+
"logits/rejected": -1.7954041957855225,
|
1664 |
+
"logps/chosen": -136.8679962158203,
|
1665 |
+
"logps/rejected": -173.52316284179688,
|
1666 |
+
"loss": 0.0305,
|
1667 |
+
"rewards/accuracies": 1.0,
|
1668 |
+
"rewards/chosen": 3.9618313312530518,
|
1669 |
+
"rewards/margins": 9.469245910644531,
|
1670 |
+
"rewards/rejected": -5.507414817810059,
|
1671 |
+
"step": 820
|
1672 |
+
},
|
1673 |
+
{
|
1674 |
+
"epoch": 3.39,
|
1675 |
+
"grad_norm": 0.00032607687171548605,
|
1676 |
+
"learning_rate": 2.266666666666667e-06,
|
1677 |
+
"logits/chosen": -1.7678247690200806,
|
1678 |
+
"logits/rejected": -1.8045036792755127,
|
1679 |
+
"logps/chosen": -97.9904556274414,
|
1680 |
+
"logps/rejected": -158.86538696289062,
|
1681 |
+
"loss": 0.0114,
|
1682 |
+
"rewards/accuracies": 1.0,
|
1683 |
+
"rewards/chosen": 3.4605960845947266,
|
1684 |
+
"rewards/margins": 9.087323188781738,
|
1685 |
+
"rewards/rejected": -5.626727104187012,
|
1686 |
+
"step": 830
|
1687 |
+
},
|
1688 |
+
{
|
1689 |
+
"epoch": 3.43,
|
1690 |
+
"grad_norm": 9.237614631652832,
|
1691 |
+
"learning_rate": 2.231578947368421e-06,
|
1692 |
+
"logits/chosen": -1.8179492950439453,
|
1693 |
+
"logits/rejected": -1.8643207550048828,
|
1694 |
+
"logps/chosen": -158.73011779785156,
|
1695 |
+
"logps/rejected": -175.10726928710938,
|
1696 |
+
"loss": 0.0529,
|
1697 |
+
"rewards/accuracies": 1.0,
|
1698 |
+
"rewards/chosen": 1.9895429611206055,
|
1699 |
+
"rewards/margins": 7.000767707824707,
|
1700 |
+
"rewards/rejected": -5.011223793029785,
|
1701 |
+
"step": 840
|
1702 |
+
},
|
1703 |
+
{
|
1704 |
+
"epoch": 3.43,
|
1705 |
+
"eval_logits/chosen": -1.842317819595337,
|
1706 |
+
"eval_logits/rejected": -1.8808995485305786,
|
1707 |
+
"eval_logps/chosen": -187.3555145263672,
|
1708 |
+
"eval_logps/rejected": -258.76568603515625,
|
1709 |
+
"eval_loss": 0.2120712250471115,
|
1710 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1711 |
+
"eval_rewards/chosen": 2.1562187671661377,
|
1712 |
+
"eval_rewards/margins": 6.830246925354004,
|
1713 |
+
"eval_rewards/rejected": -4.674027919769287,
|
1714 |
+
"eval_runtime": 26.238,
|
1715 |
+
"eval_samples_per_second": 1.067,
|
1716 |
+
"eval_steps_per_second": 1.067,
|
1717 |
+
"step": 840
|
1718 |
+
},
|
1719 |
+
{
|
1720 |
+
"epoch": 3.47,
|
1721 |
+
"grad_norm": 5.9564328694250435e-05,
|
1722 |
+
"learning_rate": 2.1964912280701754e-06,
|
1723 |
+
"logits/chosen": -1.782867431640625,
|
1724 |
+
"logits/rejected": -1.8579784631729126,
|
1725 |
+
"logps/chosen": -105.4168472290039,
|
1726 |
+
"logps/rejected": -259.2267761230469,
|
1727 |
+
"loss": 0.258,
|
1728 |
+
"rewards/accuracies": 0.8999999761581421,
|
1729 |
+
"rewards/chosen": 4.022895336151123,
|
1730 |
+
"rewards/margins": 12.175846099853516,
|
1731 |
+
"rewards/rejected": -8.152950286865234,
|
1732 |
+
"step": 850
|
1733 |
+
},
|
1734 |
+
{
|
1735 |
+
"epoch": 3.51,
|
1736 |
+
"grad_norm": 1.3973094610264525e-05,
|
1737 |
+
"learning_rate": 2.16140350877193e-06,
|
1738 |
+
"logits/chosen": -1.7417545318603516,
|
1739 |
+
"logits/rejected": -1.8234918117523193,
|
1740 |
+
"logps/chosen": -115.1645736694336,
|
1741 |
+
"logps/rejected": -248.0410614013672,
|
1742 |
+
"loss": 0.039,
|
1743 |
+
"rewards/accuracies": 1.0,
|
1744 |
+
"rewards/chosen": 3.356029987335205,
|
1745 |
+
"rewards/margins": 12.46558952331543,
|
1746 |
+
"rewards/rejected": -9.109560012817383,
|
1747 |
+
"step": 860
|
1748 |
+
},
|
1749 |
+
{
|
1750 |
+
"epoch": 3.55,
|
1751 |
+
"grad_norm": 0.005180311389267445,
|
1752 |
+
"learning_rate": 2.1263157894736844e-06,
|
1753 |
+
"logits/chosen": -1.8123457431793213,
|
1754 |
+
"logits/rejected": -1.9022102355957031,
|
1755 |
+
"logps/chosen": -128.5855255126953,
|
1756 |
+
"logps/rejected": -246.685791015625,
|
1757 |
+
"loss": 0.001,
|
1758 |
+
"rewards/accuracies": 1.0,
|
1759 |
+
"rewards/chosen": 2.4856154918670654,
|
1760 |
+
"rewards/margins": 12.966238021850586,
|
1761 |
+
"rewards/rejected": -10.480623245239258,
|
1762 |
+
"step": 870
|
1763 |
+
},
|
1764 |
+
{
|
1765 |
+
"epoch": 3.55,
|
1766 |
+
"eval_logits/chosen": -1.8409830331802368,
|
1767 |
+
"eval_logits/rejected": -1.87991464138031,
|
1768 |
+
"eval_logps/chosen": -187.5316925048828,
|
1769 |
+
"eval_logps/rejected": -259.3367614746094,
|
1770 |
+
"eval_loss": 0.2091565877199173,
|
1771 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1772 |
+
"eval_rewards/chosen": 2.1033718585968018,
|
1773 |
+
"eval_rewards/margins": 6.94872522354126,
|
1774 |
+
"eval_rewards/rejected": -4.845353126525879,
|
1775 |
+
"eval_runtime": 26.2322,
|
1776 |
+
"eval_samples_per_second": 1.067,
|
1777 |
+
"eval_steps_per_second": 1.067,
|
1778 |
+
"step": 870
|
1779 |
+
},
|
1780 |
+
{
|
1781 |
+
"epoch": 3.59,
|
1782 |
+
"grad_norm": 0.3039734661579132,
|
1783 |
+
"learning_rate": 2.0912280701754387e-06,
|
1784 |
+
"logits/chosen": -1.7311649322509766,
|
1785 |
+
"logits/rejected": -1.8156330585479736,
|
1786 |
+
"logps/chosen": -146.54473876953125,
|
1787 |
+
"logps/rejected": -300.264892578125,
|
1788 |
+
"loss": 0.0179,
|
1789 |
+
"rewards/accuracies": 1.0,
|
1790 |
+
"rewards/chosen": 3.358633041381836,
|
1791 |
+
"rewards/margins": 11.333723068237305,
|
1792 |
+
"rewards/rejected": -7.975089073181152,
|
1793 |
+
"step": 880
|
1794 |
+
},
|
1795 |
+
{
|
1796 |
+
"epoch": 3.63,
|
1797 |
+
"grad_norm": 0.8097418546676636,
|
1798 |
+
"learning_rate": 2.0561403508771934e-06,
|
1799 |
+
"logits/chosen": -1.8806817531585693,
|
1800 |
+
"logits/rejected": -1.922952651977539,
|
1801 |
+
"logps/chosen": -126.9049301147461,
|
1802 |
+
"logps/rejected": -184.9642333984375,
|
1803 |
+
"loss": 0.0541,
|
1804 |
+
"rewards/accuracies": 1.0,
|
1805 |
+
"rewards/chosen": 1.0604941844940186,
|
1806 |
+
"rewards/margins": 6.54372501373291,
|
1807 |
+
"rewards/rejected": -5.4832305908203125,
|
1808 |
+
"step": 890
|
1809 |
+
},
|
1810 |
+
{
|
1811 |
+
"epoch": 3.67,
|
1812 |
+
"grad_norm": 0.025914475321769714,
|
1813 |
+
"learning_rate": 2.0210526315789477e-06,
|
1814 |
+
"logits/chosen": -1.8227866888046265,
|
1815 |
+
"logits/rejected": -1.9303566217422485,
|
1816 |
+
"logps/chosen": -167.85279846191406,
|
1817 |
+
"logps/rejected": -311.27655029296875,
|
1818 |
+
"loss": 0.0284,
|
1819 |
+
"rewards/accuracies": 1.0,
|
1820 |
+
"rewards/chosen": 2.1594152450561523,
|
1821 |
+
"rewards/margins": 9.647980690002441,
|
1822 |
+
"rewards/rejected": -7.488564968109131,
|
1823 |
+
"step": 900
|
1824 |
+
},
|
1825 |
+
{
|
1826 |
+
"epoch": 3.67,
|
1827 |
+
"eval_logits/chosen": -1.8366106748580933,
|
1828 |
+
"eval_logits/rejected": -1.8760128021240234,
|
1829 |
+
"eval_logps/chosen": -187.9383544921875,
|
1830 |
+
"eval_logps/rejected": -260.31500244140625,
|
1831 |
+
"eval_loss": 0.20062340795993805,
|
1832 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1833 |
+
"eval_rewards/chosen": 1.9813728332519531,
|
1834 |
+
"eval_rewards/margins": 7.1201934814453125,
|
1835 |
+
"eval_rewards/rejected": -5.138820171356201,
|
1836 |
+
"eval_runtime": 26.2437,
|
1837 |
+
"eval_samples_per_second": 1.067,
|
1838 |
+
"eval_steps_per_second": 1.067,
|
1839 |
+
"step": 900
|
1840 |
+
},
|
1841 |
+
{
|
1842 |
+
"epoch": 3.71,
|
1843 |
+
"grad_norm": 0.02055520936846733,
|
1844 |
+
"learning_rate": 1.985964912280702e-06,
|
1845 |
+
"logits/chosen": -1.8666080236434937,
|
1846 |
+
"logits/rejected": -1.9068281650543213,
|
1847 |
+
"logps/chosen": -171.9093475341797,
|
1848 |
+
"logps/rejected": -251.93643188476562,
|
1849 |
+
"loss": 0.2622,
|
1850 |
+
"rewards/accuracies": 0.800000011920929,
|
1851 |
+
"rewards/chosen": 0.18651972711086273,
|
1852 |
+
"rewards/margins": 8.190787315368652,
|
1853 |
+
"rewards/rejected": -8.004267692565918,
|
1854 |
+
"step": 910
|
1855 |
+
},
|
1856 |
+
{
|
1857 |
+
"epoch": 3.76,
|
1858 |
+
"grad_norm": 197.38717651367188,
|
1859 |
+
"learning_rate": 1.9508771929824562e-06,
|
1860 |
+
"logits/chosen": -1.8426361083984375,
|
1861 |
+
"logits/rejected": -1.9008222818374634,
|
1862 |
+
"logps/chosen": -150.75900268554688,
|
1863 |
+
"logps/rejected": -251.9907989501953,
|
1864 |
+
"loss": 0.0599,
|
1865 |
+
"rewards/accuracies": 1.0,
|
1866 |
+
"rewards/chosen": 0.7380223870277405,
|
1867 |
+
"rewards/margins": 9.459989547729492,
|
1868 |
+
"rewards/rejected": -8.721967697143555,
|
1869 |
+
"step": 920
|
1870 |
+
},
|
1871 |
+
{
|
1872 |
+
"epoch": 3.8,
|
1873 |
+
"grad_norm": 0.0,
|
1874 |
+
"learning_rate": 1.9157894736842105e-06,
|
1875 |
+
"logits/chosen": -1.8063271045684814,
|
1876 |
+
"logits/rejected": -1.8660167455673218,
|
1877 |
+
"logps/chosen": -198.50723266601562,
|
1878 |
+
"logps/rejected": -222.7967529296875,
|
1879 |
+
"loss": 0.0744,
|
1880 |
+
"rewards/accuracies": 0.8999999761581421,
|
1881 |
+
"rewards/chosen": 1.713536262512207,
|
1882 |
+
"rewards/margins": 9.678624153137207,
|
1883 |
+
"rewards/rejected": -7.965087890625,
|
1884 |
+
"step": 930
|
1885 |
+
},
|
1886 |
+
{
|
1887 |
+
"epoch": 3.8,
|
1888 |
+
"eval_logits/chosen": -1.833928108215332,
|
1889 |
+
"eval_logits/rejected": -1.873318076133728,
|
1890 |
+
"eval_logps/chosen": -188.06385803222656,
|
1891 |
+
"eval_logps/rejected": -260.6358337402344,
|
1892 |
+
"eval_loss": 0.1813022941350937,
|
1893 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1894 |
+
"eval_rewards/chosen": 1.943717122077942,
|
1895 |
+
"eval_rewards/margins": 7.178780555725098,
|
1896 |
+
"eval_rewards/rejected": -5.2350640296936035,
|
1897 |
+
"eval_runtime": 26.2336,
|
1898 |
+
"eval_samples_per_second": 1.067,
|
1899 |
+
"eval_steps_per_second": 1.067,
|
1900 |
+
"step": 930
|
1901 |
+
},
|
1902 |
+
{
|
1903 |
+
"epoch": 3.84,
|
1904 |
+
"grad_norm": 4.8926853679631677e-08,
|
1905 |
+
"learning_rate": 1.8807017543859652e-06,
|
1906 |
+
"logits/chosen": -1.8202307224273682,
|
1907 |
+
"logits/rejected": -1.9166450500488281,
|
1908 |
+
"logps/chosen": -112.94303131103516,
|
1909 |
+
"logps/rejected": -244.4494171142578,
|
1910 |
+
"loss": 0.3721,
|
1911 |
+
"rewards/accuracies": 0.8999999761581421,
|
1912 |
+
"rewards/chosen": 2.255749225616455,
|
1913 |
+
"rewards/margins": 10.898755073547363,
|
1914 |
+
"rewards/rejected": -8.643006324768066,
|
1915 |
+
"step": 940
|
1916 |
+
},
|
1917 |
+
{
|
1918 |
+
"epoch": 3.88,
|
1919 |
+
"grad_norm": 5.456500053405762,
|
1920 |
+
"learning_rate": 1.8456140350877193e-06,
|
1921 |
+
"logits/chosen": -1.9108766317367554,
|
1922 |
+
"logits/rejected": -1.9368479251861572,
|
1923 |
+
"logps/chosen": -132.4052734375,
|
1924 |
+
"logps/rejected": -179.675537109375,
|
1925 |
+
"loss": 0.0257,
|
1926 |
+
"rewards/accuracies": 1.0,
|
1927 |
+
"rewards/chosen": 1.554168701171875,
|
1928 |
+
"rewards/margins": 6.65416955947876,
|
1929 |
+
"rewards/rejected": -5.100000858306885,
|
1930 |
+
"step": 950
|
1931 |
+
},
|
1932 |
+
{
|
1933 |
+
"epoch": 3.92,
|
1934 |
+
"grad_norm": 9.124670028686523,
|
1935 |
+
"learning_rate": 1.810526315789474e-06,
|
1936 |
+
"logits/chosen": -1.7729440927505493,
|
1937 |
+
"logits/rejected": -1.8252880573272705,
|
1938 |
+
"logps/chosen": -139.13653564453125,
|
1939 |
+
"logps/rejected": -193.011962890625,
|
1940 |
+
"loss": 0.091,
|
1941 |
+
"rewards/accuracies": 1.0,
|
1942 |
+
"rewards/chosen": 3.6216557025909424,
|
1943 |
+
"rewards/margins": 8.979948043823242,
|
1944 |
+
"rewards/rejected": -5.3582916259765625,
|
1945 |
+
"step": 960
|
1946 |
+
},
|
1947 |
+
{
|
1948 |
+
"epoch": 3.92,
|
1949 |
+
"eval_logits/chosen": -1.8312675952911377,
|
1950 |
+
"eval_logits/rejected": -1.8707492351531982,
|
1951 |
+
"eval_logps/chosen": -188.43190002441406,
|
1952 |
+
"eval_logps/rejected": -261.29730224609375,
|
1953 |
+
"eval_loss": 0.17215047776699066,
|
1954 |
+
"eval_rewards/accuracies": 0.8928571343421936,
|
1955 |
+
"eval_rewards/chosen": 1.8333097696304321,
|
1956 |
+
"eval_rewards/margins": 7.266810894012451,
|
1957 |
+
"eval_rewards/rejected": -5.433501243591309,
|
1958 |
+
"eval_runtime": 26.2333,
|
1959 |
+
"eval_samples_per_second": 1.067,
|
1960 |
+
"eval_steps_per_second": 1.067,
|
1961 |
+
"step": 960
|
1962 |
+
},
|
1963 |
+
{
|
1964 |
+
"epoch": 3.96,
|
1965 |
+
"grad_norm": 1.8921099353974569e-06,
|
1966 |
+
"learning_rate": 1.7754385964912283e-06,
|
1967 |
+
"logits/chosen": -1.8464349508285522,
|
1968 |
+
"logits/rejected": -1.8927011489868164,
|
1969 |
+
"logps/chosen": -172.9676971435547,
|
1970 |
+
"logps/rejected": -224.15780639648438,
|
1971 |
+
"loss": 0.0795,
|
1972 |
+
"rewards/accuracies": 1.0,
|
1973 |
+
"rewards/chosen": 2.649458169937134,
|
1974 |
+
"rewards/margins": 9.076958656311035,
|
1975 |
+
"rewards/rejected": -6.4275007247924805,
|
1976 |
+
"step": 970
|
1977 |
+
},
|
1978 |
+
{
|
1979 |
+
"epoch": 4.0,
|
1980 |
+
"grad_norm": 318.0891418457031,
|
1981 |
+
"learning_rate": 1.7403508771929826e-06,
|
1982 |
+
"logits/chosen": -1.8083856105804443,
|
1983 |
+
"logits/rejected": -1.8901259899139404,
|
1984 |
+
"logps/chosen": -209.17422485351562,
|
1985 |
+
"logps/rejected": -284.4140625,
|
1986 |
+
"loss": 0.0583,
|
1987 |
+
"rewards/accuracies": 1.0,
|
1988 |
+
"rewards/chosen": 2.1195216178894043,
|
1989 |
+
"rewards/margins": 12.128747940063477,
|
1990 |
+
"rewards/rejected": -10.00922679901123,
|
1991 |
+
"step": 980
|
1992 |
+
},
|
1993 |
+
{
|
1994 |
+
"epoch": 4.04,
|
1995 |
+
"grad_norm": 13.0223388671875,
|
1996 |
+
"learning_rate": 1.705263157894737e-06,
|
1997 |
+
"logits/chosen": -1.847835898399353,
|
1998 |
+
"logits/rejected": -1.868665337562561,
|
1999 |
+
"logps/chosen": -149.8837127685547,
|
2000 |
+
"logps/rejected": -161.1732940673828,
|
2001 |
+
"loss": 0.3504,
|
2002 |
+
"rewards/accuracies": 0.8999999761581421,
|
2003 |
+
"rewards/chosen": 1.0685189962387085,
|
2004 |
+
"rewards/margins": 6.018086910247803,
|
2005 |
+
"rewards/rejected": -4.9495673179626465,
|
2006 |
+
"step": 990
|
2007 |
+
},
|
2008 |
+
{
|
2009 |
+
"epoch": 4.04,
|
2010 |
+
"eval_logits/chosen": -1.8279262781143188,
|
2011 |
+
"eval_logits/rejected": -1.8672196865081787,
|
2012 |
+
"eval_logps/chosen": -188.3168487548828,
|
2013 |
+
"eval_logps/rejected": -261.04876708984375,
|
2014 |
+
"eval_loss": 0.14871637523174286,
|
2015 |
+
"eval_rewards/accuracies": 0.9285714030265808,
|
2016 |
+
"eval_rewards/chosen": 1.8678233623504639,
|
2017 |
+
"eval_rewards/margins": 7.22676944732666,
|
2018 |
+
"eval_rewards/rejected": -5.358946800231934,
|
2019 |
+
"eval_runtime": 26.2331,
|
2020 |
+
"eval_samples_per_second": 1.067,
|
2021 |
+
"eval_steps_per_second": 1.067,
|
2022 |
+
"step": 990
|
2023 |
+
},
|
2024 |
+
{
|
2025 |
+
"epoch": 4.08,
|
2026 |
+
"grad_norm": 0.6881242394447327,
|
2027 |
+
"learning_rate": 1.6701754385964914e-06,
|
2028 |
+
"logits/chosen": -1.9393212795257568,
|
2029 |
+
"logits/rejected": -1.9376901388168335,
|
2030 |
+
"logps/chosen": -155.8722381591797,
|
2031 |
+
"logps/rejected": -191.2816925048828,
|
2032 |
+
"loss": 0.0102,
|
2033 |
+
"rewards/accuracies": 1.0,
|
2034 |
+
"rewards/chosen": 2.3404548168182373,
|
2035 |
+
"rewards/margins": 7.390922546386719,
|
2036 |
+
"rewards/rejected": -5.0504679679870605,
|
2037 |
+
"step": 1000
|
2038 |
+
},
|
2039 |
+
{
|
2040 |
+
"epoch": 4.12,
|
2041 |
+
"grad_norm": 0.0015760493697598577,
|
2042 |
+
"learning_rate": 1.6350877192982456e-06,
|
2043 |
+
"logits/chosen": -1.7732092142105103,
|
2044 |
+
"logits/rejected": -1.8348764181137085,
|
2045 |
+
"logps/chosen": -204.78118896484375,
|
2046 |
+
"logps/rejected": -285.19171142578125,
|
2047 |
+
"loss": 0.0728,
|
2048 |
+
"rewards/accuracies": 0.8999999761581421,
|
2049 |
+
"rewards/chosen": 1.1964905261993408,
|
2050 |
+
"rewards/margins": 11.535087585449219,
|
2051 |
+
"rewards/rejected": -10.338597297668457,
|
2052 |
+
"step": 1010
|
2053 |
+
},
|
2054 |
+
{
|
2055 |
+
"epoch": 4.16,
|
2056 |
+
"grad_norm": 5.320696345734177e-07,
|
2057 |
+
"learning_rate": 1.6000000000000001e-06,
|
2058 |
+
"logits/chosen": -1.7163060903549194,
|
2059 |
+
"logits/rejected": -1.8586156368255615,
|
2060 |
+
"logps/chosen": -123.50263977050781,
|
2061 |
+
"logps/rejected": -320.79693603515625,
|
2062 |
+
"loss": 0.0071,
|
2063 |
+
"rewards/accuracies": 1.0,
|
2064 |
+
"rewards/chosen": 0.7505568265914917,
|
2065 |
+
"rewards/margins": 10.981633186340332,
|
2066 |
+
"rewards/rejected": -10.231074333190918,
|
2067 |
+
"step": 1020
|
2068 |
+
},
|
2069 |
+
{
|
2070 |
+
"epoch": 4.16,
|
2071 |
+
"eval_logits/chosen": -1.824275016784668,
|
2072 |
+
"eval_logits/rejected": -1.8636810779571533,
|
2073 |
+
"eval_logps/chosen": -188.54676818847656,
|
2074 |
+
"eval_logps/rejected": -261.5804748535156,
|
2075 |
+
"eval_loss": 0.14029009640216827,
|
2076 |
+
"eval_rewards/accuracies": 0.9285714030265808,
|
2077 |
+
"eval_rewards/chosen": 1.7988510131835938,
|
2078 |
+
"eval_rewards/margins": 7.317309856414795,
|
2079 |
+
"eval_rewards/rejected": -5.518458843231201,
|
2080 |
+
"eval_runtime": 26.2364,
|
2081 |
+
"eval_samples_per_second": 1.067,
|
2082 |
+
"eval_steps_per_second": 1.067,
|
2083 |
+
"step": 1020
|
2084 |
+
},
|
2085 |
+
{
|
2086 |
+
"epoch": 4.2,
|
2087 |
+
"grad_norm": 0.002488874364644289,
|
2088 |
+
"learning_rate": 1.5649122807017544e-06,
|
2089 |
+
"logits/chosen": -1.7597904205322266,
|
2090 |
+
"logits/rejected": -1.8473920822143555,
|
2091 |
+
"logps/chosen": -165.49359130859375,
|
2092 |
+
"logps/rejected": -273.9131774902344,
|
2093 |
+
"loss": 0.0071,
|
2094 |
+
"rewards/accuracies": 1.0,
|
2095 |
+
"rewards/chosen": 1.4215660095214844,
|
2096 |
+
"rewards/margins": 10.477560043334961,
|
2097 |
+
"rewards/rejected": -9.055994033813477,
|
2098 |
+
"step": 1030
|
2099 |
+
},
|
2100 |
+
{
|
2101 |
+
"epoch": 4.24,
|
2102 |
+
"grad_norm": 0.0017074132338166237,
|
2103 |
+
"learning_rate": 1.529824561403509e-06,
|
2104 |
+
"logits/chosen": -1.7763557434082031,
|
2105 |
+
"logits/rejected": -1.8936008214950562,
|
2106 |
+
"logps/chosen": -133.30006408691406,
|
2107 |
+
"logps/rejected": -336.57427978515625,
|
2108 |
+
"loss": 0.018,
|
2109 |
+
"rewards/accuracies": 1.0,
|
2110 |
+
"rewards/chosen": 4.363164901733398,
|
2111 |
+
"rewards/margins": 14.957908630371094,
|
2112 |
+
"rewards/rejected": -10.594743728637695,
|
2113 |
+
"step": 1040
|
2114 |
+
},
|
2115 |
+
{
|
2116 |
+
"epoch": 4.29,
|
2117 |
+
"grad_norm": 66.57980346679688,
|
2118 |
+
"learning_rate": 1.4947368421052632e-06,
|
2119 |
+
"logits/chosen": -1.835628867149353,
|
2120 |
+
"logits/rejected": -1.8976163864135742,
|
2121 |
+
"logps/chosen": -177.20309448242188,
|
2122 |
+
"logps/rejected": -287.8599548339844,
|
2123 |
+
"loss": 0.0131,
|
2124 |
+
"rewards/accuracies": 1.0,
|
2125 |
+
"rewards/chosen": 2.568208694458008,
|
2126 |
+
"rewards/margins": 11.586195945739746,
|
2127 |
+
"rewards/rejected": -9.017987251281738,
|
2128 |
+
"step": 1050
|
2129 |
+
},
|
2130 |
+
{
|
2131 |
+
"epoch": 4.29,
|
2132 |
+
"eval_logits/chosen": -1.8221967220306396,
|
2133 |
+
"eval_logits/rejected": -1.8615671396255493,
|
2134 |
+
"eval_logps/chosen": -188.5262451171875,
|
2135 |
+
"eval_logps/rejected": -261.6840515136719,
|
2136 |
+
"eval_loss": 0.13124893605709076,
|
2137 |
+
"eval_rewards/accuracies": 0.9285714030265808,
|
2138 |
+
"eval_rewards/chosen": 1.805002212524414,
|
2139 |
+
"eval_rewards/margins": 7.354526042938232,
|
2140 |
+
"eval_rewards/rejected": -5.549524784088135,
|
2141 |
+
"eval_runtime": 26.2328,
|
2142 |
+
"eval_samples_per_second": 1.067,
|
2143 |
+
"eval_steps_per_second": 1.067,
|
2144 |
+
"step": 1050
|
2145 |
+
},
|
2146 |
+
{
|
2147 |
+
"epoch": 4.33,
|
2148 |
+
"grad_norm": 0.027719292789697647,
|
2149 |
+
"learning_rate": 1.4596491228070175e-06,
|
2150 |
+
"logits/chosen": -1.7560068368911743,
|
2151 |
+
"logits/rejected": -1.8672195672988892,
|
2152 |
+
"logps/chosen": -189.530029296875,
|
2153 |
+
"logps/rejected": -330.6167907714844,
|
2154 |
+
"loss": 0.003,
|
2155 |
+
"rewards/accuracies": 1.0,
|
2156 |
+
"rewards/chosen": 1.0810093879699707,
|
2157 |
+
"rewards/margins": 9.271745681762695,
|
2158 |
+
"rewards/rejected": -8.19073486328125,
|
2159 |
+
"step": 1060
|
2160 |
+
},
|
2161 |
+
{
|
2162 |
+
"epoch": 4.37,
|
2163 |
+
"grad_norm": 0.0,
|
2164 |
+
"learning_rate": 1.4245614035087722e-06,
|
2165 |
+
"logits/chosen": -1.8792110681533813,
|
2166 |
+
"logits/rejected": -1.9164355993270874,
|
2167 |
+
"logps/chosen": -212.42788696289062,
|
2168 |
+
"logps/rejected": -238.2472686767578,
|
2169 |
+
"loss": 0.1334,
|
2170 |
+
"rewards/accuracies": 0.8999999761581421,
|
2171 |
+
"rewards/chosen": 1.0076087713241577,
|
2172 |
+
"rewards/margins": 9.041505813598633,
|
2173 |
+
"rewards/rejected": -8.033895492553711,
|
2174 |
+
"step": 1070
|
2175 |
+
},
|
2176 |
+
{
|
2177 |
+
"epoch": 4.41,
|
2178 |
+
"grad_norm": 3.609489340306027e-06,
|
2179 |
+
"learning_rate": 1.3894736842105263e-06,
|
2180 |
+
"logits/chosen": -1.8213157653808594,
|
2181 |
+
"logits/rejected": -1.870356798171997,
|
2182 |
+
"logps/chosen": -80.74689483642578,
|
2183 |
+
"logps/rejected": -158.96109008789062,
|
2184 |
+
"loss": 0.0868,
|
2185 |
+
"rewards/accuracies": 1.0,
|
2186 |
+
"rewards/chosen": 2.509101629257202,
|
2187 |
+
"rewards/margins": 7.621730804443359,
|
2188 |
+
"rewards/rejected": -5.11262845993042,
|
2189 |
+
"step": 1080
|
2190 |
+
},
|
2191 |
+
{
|
2192 |
+
"epoch": 4.41,
|
2193 |
+
"eval_logits/chosen": -1.81948721408844,
|
2194 |
+
"eval_logits/rejected": -1.858702540397644,
|
2195 |
+
"eval_logps/chosen": -188.66748046875,
|
2196 |
+
"eval_logps/rejected": -261.94708251953125,
|
2197 |
+
"eval_loss": 0.1210150271654129,
|
2198 |
+
"eval_rewards/accuracies": 0.9285714030265808,
|
2199 |
+
"eval_rewards/chosen": 1.76263427734375,
|
2200 |
+
"eval_rewards/margins": 7.39107608795166,
|
2201 |
+
"eval_rewards/rejected": -5.62844181060791,
|
2202 |
+
"eval_runtime": 26.2393,
|
2203 |
+
"eval_samples_per_second": 1.067,
|
2204 |
+
"eval_steps_per_second": 1.067,
|
2205 |
+
"step": 1080
|
2206 |
+
},
|
2207 |
+
{
|
2208 |
+
"epoch": 4.45,
|
2209 |
+
"grad_norm": 0.7020970582962036,
|
2210 |
+
"learning_rate": 1.354385964912281e-06,
|
2211 |
+
"logits/chosen": -1.8757814168930054,
|
2212 |
+
"logits/rejected": -1.9657201766967773,
|
2213 |
+
"logps/chosen": -117.080810546875,
|
2214 |
+
"logps/rejected": -241.55252075195312,
|
2215 |
+
"loss": 0.0046,
|
2216 |
+
"rewards/accuracies": 1.0,
|
2217 |
+
"rewards/chosen": 1.6267999410629272,
|
2218 |
+
"rewards/margins": 10.32630729675293,
|
2219 |
+
"rewards/rejected": -8.699506759643555,
|
2220 |
+
"step": 1090
|
2221 |
+
},
|
2222 |
+
{
|
2223 |
+
"epoch": 4.49,
|
2224 |
+
"grad_norm": 0.00016405929636675864,
|
2225 |
+
"learning_rate": 1.3192982456140353e-06,
|
2226 |
+
"logits/chosen": -1.84481680393219,
|
2227 |
+
"logits/rejected": -1.8962256908416748,
|
2228 |
+
"logps/chosen": -176.97457885742188,
|
2229 |
+
"logps/rejected": -248.0132598876953,
|
2230 |
+
"loss": 0.0261,
|
2231 |
+
"rewards/accuracies": 1.0,
|
2232 |
+
"rewards/chosen": 1.7715377807617188,
|
2233 |
+
"rewards/margins": 11.10531234741211,
|
2234 |
+
"rewards/rejected": -9.333776473999023,
|
2235 |
+
"step": 1100
|
2236 |
+
},
|
2237 |
+
{
|
2238 |
+
"epoch": 4.53,
|
2239 |
+
"grad_norm": 0.0024285567924380302,
|
2240 |
+
"learning_rate": 1.2842105263157895e-06,
|
2241 |
+
"logits/chosen": -1.9240976572036743,
|
2242 |
+
"logits/rejected": -1.9567447900772095,
|
2243 |
+
"logps/chosen": -148.41506958007812,
|
2244 |
+
"logps/rejected": -262.0478820800781,
|
2245 |
+
"loss": 0.0041,
|
2246 |
+
"rewards/accuracies": 1.0,
|
2247 |
+
"rewards/chosen": 2.377047538757324,
|
2248 |
+
"rewards/margins": 12.98823070526123,
|
2249 |
+
"rewards/rejected": -10.611183166503906,
|
2250 |
+
"step": 1110
|
2251 |
+
},
|
2252 |
+
{
|
2253 |
+
"epoch": 4.53,
|
2254 |
+
"eval_logits/chosen": -1.8172996044158936,
|
2255 |
+
"eval_logits/rejected": -1.8566371202468872,
|
2256 |
+
"eval_logps/chosen": -188.92127990722656,
|
2257 |
+
"eval_logps/rejected": -262.44561767578125,
|
2258 |
+
"eval_loss": 0.1205989271402359,
|
2259 |
+
"eval_rewards/accuracies": 0.9285714030265808,
|
2260 |
+
"eval_rewards/chosen": 1.686492681503296,
|
2261 |
+
"eval_rewards/margins": 7.464495658874512,
|
2262 |
+
"eval_rewards/rejected": -5.778003215789795,
|
2263 |
+
"eval_runtime": 26.2482,
|
2264 |
+
"eval_samples_per_second": 1.067,
|
2265 |
+
"eval_steps_per_second": 1.067,
|
2266 |
+
"step": 1110
|
2267 |
+
},
|
2268 |
+
{
|
2269 |
+
"epoch": 4.57,
|
2270 |
+
"grad_norm": 26.255502700805664,
|
2271 |
+
"learning_rate": 1.249122807017544e-06,
|
2272 |
+
"logits/chosen": -1.7752002477645874,
|
2273 |
+
"logits/rejected": -1.8067470788955688,
|
2274 |
+
"logps/chosen": -199.05264282226562,
|
2275 |
+
"logps/rejected": -198.15122985839844,
|
2276 |
+
"loss": 0.0289,
|
2277 |
+
"rewards/accuracies": 1.0,
|
2278 |
+
"rewards/chosen": 2.024845600128174,
|
2279 |
+
"rewards/margins": 10.319567680358887,
|
2280 |
+
"rewards/rejected": -8.294721603393555,
|
2281 |
+
"step": 1120
|
2282 |
+
},
|
2283 |
+
{
|
2284 |
+
"epoch": 4.61,
|
2285 |
+
"grad_norm": 170.23622131347656,
|
2286 |
+
"learning_rate": 1.2140350877192983e-06,
|
2287 |
+
"logits/chosen": -1.7735440731048584,
|
2288 |
+
"logits/rejected": -1.8578115701675415,
|
2289 |
+
"logps/chosen": -140.52511596679688,
|
2290 |
+
"logps/rejected": -257.21124267578125,
|
2291 |
+
"loss": 0.0478,
|
2292 |
+
"rewards/accuracies": 1.0,
|
2293 |
+
"rewards/chosen": 3.585864305496216,
|
2294 |
+
"rewards/margins": 12.626684188842773,
|
2295 |
+
"rewards/rejected": -9.04081916809082,
|
2296 |
+
"step": 1130
|
2297 |
+
},
|
2298 |
+
{
|
2299 |
+
"epoch": 4.65,
|
2300 |
+
"grad_norm": 0.09534689784049988,
|
2301 |
+
"learning_rate": 1.1789473684210526e-06,
|
2302 |
+
"logits/chosen": -1.8365780115127563,
|
2303 |
+
"logits/rejected": -1.846933364868164,
|
2304 |
+
"logps/chosen": -90.59309387207031,
|
2305 |
+
"logps/rejected": -117.09541320800781,
|
2306 |
+
"loss": 0.0107,
|
2307 |
+
"rewards/accuracies": 1.0,
|
2308 |
+
"rewards/chosen": 4.559220314025879,
|
2309 |
+
"rewards/margins": 10.236574172973633,
|
2310 |
+
"rewards/rejected": -5.677353382110596,
|
2311 |
+
"step": 1140
|
2312 |
+
},
|
2313 |
+
{
|
2314 |
+
"epoch": 4.65,
|
2315 |
+
"eval_logits/chosen": -1.8170740604400635,
|
2316 |
+
"eval_logits/rejected": -1.8563201427459717,
|
2317 |
+
"eval_logps/chosen": -189.08616638183594,
|
2318 |
+
"eval_logps/rejected": -262.8173828125,
|
2319 |
+
"eval_loss": 0.1178145483136177,
|
2320 |
+
"eval_rewards/accuracies": 0.9642857313156128,
|
2321 |
+
"eval_rewards/chosen": 1.6370302438735962,
|
2322 |
+
"eval_rewards/margins": 7.526562213897705,
|
2323 |
+
"eval_rewards/rejected": -5.88953161239624,
|
2324 |
+
"eval_runtime": 26.2301,
|
2325 |
+
"eval_samples_per_second": 1.067,
|
2326 |
+
"eval_steps_per_second": 1.067,
|
2327 |
+
"step": 1140
|
2328 |
+
},
|
2329 |
+
{
|
2330 |
+
"epoch": 4.69,
|
2331 |
+
"grad_norm": 0.0,
|
2332 |
+
"learning_rate": 1.143859649122807e-06,
|
2333 |
+
"logits/chosen": -1.7817316055297852,
|
2334 |
+
"logits/rejected": -1.8392887115478516,
|
2335 |
+
"logps/chosen": -162.9131317138672,
|
2336 |
+
"logps/rejected": -241.8101043701172,
|
2337 |
+
"loss": 0.0639,
|
2338 |
+
"rewards/accuracies": 1.0,
|
2339 |
+
"rewards/chosen": 1.5432535409927368,
|
2340 |
+
"rewards/margins": 9.659387588500977,
|
2341 |
+
"rewards/rejected": -8.116134643554688,
|
2342 |
+
"step": 1150
|
2343 |
+
},
|
2344 |
+
{
|
2345 |
+
"epoch": 4.73,
|
2346 |
+
"grad_norm": 1.92129647731781,
|
2347 |
+
"learning_rate": 1.1087719298245616e-06,
|
2348 |
+
"logits/chosen": -1.856563925743103,
|
2349 |
+
"logits/rejected": -1.8815357685089111,
|
2350 |
+
"logps/chosen": -140.67437744140625,
|
2351 |
+
"logps/rejected": -180.8864288330078,
|
2352 |
+
"loss": 0.2,
|
2353 |
+
"rewards/accuracies": 0.8999999761581421,
|
2354 |
+
"rewards/chosen": 1.501416563987732,
|
2355 |
+
"rewards/margins": 8.230588912963867,
|
2356 |
+
"rewards/rejected": -6.7291717529296875,
|
2357 |
+
"step": 1160
|
2358 |
+
},
|
2359 |
+
{
|
2360 |
+
"epoch": 4.78,
|
2361 |
+
"grad_norm": 4.147223472595215,
|
2362 |
+
"learning_rate": 1.0736842105263159e-06,
|
2363 |
+
"logits/chosen": -1.7213938236236572,
|
2364 |
+
"logits/rejected": -1.7904603481292725,
|
2365 |
+
"logps/chosen": -125.3969497680664,
|
2366 |
+
"logps/rejected": -262.72662353515625,
|
2367 |
+
"loss": 0.0084,
|
2368 |
+
"rewards/accuracies": 1.0,
|
2369 |
+
"rewards/chosen": 3.621525287628174,
|
2370 |
+
"rewards/margins": 12.910409927368164,
|
2371 |
+
"rewards/rejected": -9.288885116577148,
|
2372 |
+
"step": 1170
|
2373 |
+
},
|
2374 |
+
{
|
2375 |
+
"epoch": 4.78,
|
2376 |
+
"eval_logits/chosen": -1.815858244895935,
|
2377 |
+
"eval_logits/rejected": -1.8551596403121948,
|
2378 |
+
"eval_logps/chosen": -189.174072265625,
|
2379 |
+
"eval_logps/rejected": -262.9738464355469,
|
2380 |
+
"eval_loss": 0.11231832951307297,
|
2381 |
+
"eval_rewards/accuracies": 0.9642857313156128,
|
2382 |
+
"eval_rewards/chosen": 1.610654592514038,
|
2383 |
+
"eval_rewards/margins": 7.547122478485107,
|
2384 |
+
"eval_rewards/rejected": -5.936468601226807,
|
2385 |
+
"eval_runtime": 26.2436,
|
2386 |
+
"eval_samples_per_second": 1.067,
|
2387 |
+
"eval_steps_per_second": 1.067,
|
2388 |
+
"step": 1170
|
2389 |
+
},
|
2390 |
+
{
|
2391 |
+
"epoch": 4.82,
|
2392 |
+
"grad_norm": 8.42303293779878e-09,
|
2393 |
+
"learning_rate": 1.0385964912280702e-06,
|
2394 |
+
"logits/chosen": -1.810766577720642,
|
2395 |
+
"logits/rejected": -1.8864467144012451,
|
2396 |
+
"logps/chosen": -138.02499389648438,
|
2397 |
+
"logps/rejected": -274.3287048339844,
|
2398 |
+
"loss": 0.0587,
|
2399 |
+
"rewards/accuracies": 1.0,
|
2400 |
+
"rewards/chosen": 2.259162187576294,
|
2401 |
+
"rewards/margins": 12.501955032348633,
|
2402 |
+
"rewards/rejected": -10.242792129516602,
|
2403 |
+
"step": 1180
|
2404 |
+
},
|
2405 |
+
{
|
2406 |
+
"epoch": 4.86,
|
2407 |
+
"grad_norm": 1.6972613821053528e-06,
|
2408 |
+
"learning_rate": 1.0035087719298247e-06,
|
2409 |
+
"logits/chosen": -1.8087619543075562,
|
2410 |
+
"logits/rejected": -1.864861249923706,
|
2411 |
+
"logps/chosen": -106.66157531738281,
|
2412 |
+
"logps/rejected": -172.39846801757812,
|
2413 |
+
"loss": 0.0167,
|
2414 |
+
"rewards/accuracies": 1.0,
|
2415 |
+
"rewards/chosen": 1.4979627132415771,
|
2416 |
+
"rewards/margins": 9.330595016479492,
|
2417 |
+
"rewards/rejected": -7.832631587982178,
|
2418 |
+
"step": 1190
|
2419 |
+
},
|
2420 |
+
{
|
2421 |
+
"epoch": 4.9,
|
2422 |
+
"grad_norm": 0.07996608316898346,
|
2423 |
+
"learning_rate": 9.68421052631579e-07,
|
2424 |
+
"logits/chosen": -1.6860878467559814,
|
2425 |
+
"logits/rejected": -1.7491111755371094,
|
2426 |
+
"logps/chosen": -86.7617416381836,
|
2427 |
+
"logps/rejected": -172.6136016845703,
|
2428 |
+
"loss": 0.0049,
|
2429 |
+
"rewards/accuracies": 1.0,
|
2430 |
+
"rewards/chosen": 3.904423952102661,
|
2431 |
+
"rewards/margins": 10.647790908813477,
|
2432 |
+
"rewards/rejected": -6.743367671966553,
|
2433 |
+
"step": 1200
|
2434 |
+
},
|
2435 |
+
{
|
2436 |
+
"epoch": 4.9,
|
2437 |
+
"eval_logits/chosen": -1.8151248693466187,
|
2438 |
+
"eval_logits/rejected": -1.8545377254486084,
|
2439 |
+
"eval_logps/chosen": -189.30612182617188,
|
2440 |
+
"eval_logps/rejected": -263.3506774902344,
|
2441 |
+
"eval_loss": 0.10828591883182526,
|
2442 |
+
"eval_rewards/accuracies": 0.9642857313156128,
|
2443 |
+
"eval_rewards/chosen": 1.5710465908050537,
|
2444 |
+
"eval_rewards/margins": 7.62056303024292,
|
2445 |
+
"eval_rewards/rejected": -6.049516201019287,
|
2446 |
+
"eval_runtime": 26.2353,
|
2447 |
+
"eval_samples_per_second": 1.067,
|
2448 |
+
"eval_steps_per_second": 1.067,
|
2449 |
+
"step": 1200
|
2450 |
+
},
|
2451 |
+
{
|
2452 |
+
"epoch": 4.94,
|
2453 |
+
"grad_norm": 0.0006144978106021881,
|
2454 |
+
"learning_rate": 9.333333333333334e-07,
|
2455 |
+
"logits/chosen": -1.7948280572891235,
|
2456 |
+
"logits/rejected": -1.8062083721160889,
|
2457 |
+
"logps/chosen": -201.0863037109375,
|
2458 |
+
"logps/rejected": -189.9558563232422,
|
2459 |
+
"loss": 0.0226,
|
2460 |
+
"rewards/accuracies": 1.0,
|
2461 |
+
"rewards/chosen": 2.141186237335205,
|
2462 |
+
"rewards/margins": 9.003381729125977,
|
2463 |
+
"rewards/rejected": -6.862195014953613,
|
2464 |
+
"step": 1210
|
2465 |
+
},
|
2466 |
+
{
|
2467 |
+
"epoch": 4.98,
|
2468 |
+
"grad_norm": 0.004125670529901981,
|
2469 |
+
"learning_rate": 8.982456140350877e-07,
|
2470 |
+
"logits/chosen": -1.7143564224243164,
|
2471 |
+
"logits/rejected": -1.755270004272461,
|
2472 |
+
"logps/chosen": -129.1698760986328,
|
2473 |
+
"logps/rejected": -167.7386016845703,
|
2474 |
+
"loss": 0.0439,
|
2475 |
+
"rewards/accuracies": 1.0,
|
2476 |
+
"rewards/chosen": 0.8997966647148132,
|
2477 |
+
"rewards/margins": 8.175169944763184,
|
2478 |
+
"rewards/rejected": -7.275373935699463,
|
2479 |
+
"step": 1220
|
2480 |
+
},
|
2481 |
+
{
|
2482 |
+
"epoch": 5.02,
|
2483 |
+
"grad_norm": 0.000474505708552897,
|
2484 |
+
"learning_rate": 8.631578947368421e-07,
|
2485 |
+
"logits/chosen": -1.8216488361358643,
|
2486 |
+
"logits/rejected": -1.880822777748108,
|
2487 |
+
"logps/chosen": -124.86578369140625,
|
2488 |
+
"logps/rejected": -239.4408721923828,
|
2489 |
+
"loss": 0.0746,
|
2490 |
+
"rewards/accuracies": 0.8999999761581421,
|
2491 |
+
"rewards/chosen": 1.0542094707489014,
|
2492 |
+
"rewards/margins": 9.776211738586426,
|
2493 |
+
"rewards/rejected": -8.722002029418945,
|
2494 |
+
"step": 1230
|
2495 |
+
},
|
2496 |
+
{
|
2497 |
+
"epoch": 5.02,
|
2498 |
+
"eval_logits/chosen": -1.8140125274658203,
|
2499 |
+
"eval_logits/rejected": -1.8534783124923706,
|
2500 |
+
"eval_logps/chosen": -189.43362426757812,
|
2501 |
+
"eval_logps/rejected": -263.6144104003906,
|
2502 |
+
"eval_loss": 0.10336750000715256,
|
2503 |
+
"eval_rewards/accuracies": 0.9642857313156128,
|
2504 |
+
"eval_rewards/chosen": 1.5327917337417603,
|
2505 |
+
"eval_rewards/margins": 7.6614274978637695,
|
2506 |
+
"eval_rewards/rejected": -6.128636837005615,
|
2507 |
+
"eval_runtime": 26.2322,
|
2508 |
+
"eval_samples_per_second": 1.067,
|
2509 |
+
"eval_steps_per_second": 1.067,
|
2510 |
+
"step": 1230
|
2511 |
+
},
|
2512 |
+
{
|
2513 |
+
"epoch": 5.06,
|
2514 |
+
"grad_norm": 0.029490573331713676,
|
2515 |
+
"learning_rate": 8.280701754385966e-07,
|
2516 |
+
"logits/chosen": -1.778794288635254,
|
2517 |
+
"logits/rejected": -1.876511573791504,
|
2518 |
+
"logps/chosen": -113.80903625488281,
|
2519 |
+
"logps/rejected": -301.50201416015625,
|
2520 |
+
"loss": 0.0002,
|
2521 |
+
"rewards/accuracies": 1.0,
|
2522 |
+
"rewards/chosen": 2.4956843852996826,
|
2523 |
+
"rewards/margins": 14.169008255004883,
|
2524 |
+
"rewards/rejected": -11.673322677612305,
|
2525 |
+
"step": 1240
|
2526 |
+
},
|
2527 |
+
{
|
2528 |
+
"epoch": 5.1,
|
2529 |
+
"grad_norm": 0.8206026554107666,
|
2530 |
+
"learning_rate": 7.92982456140351e-07,
|
2531 |
+
"logits/chosen": -1.7988386154174805,
|
2532 |
+
"logits/rejected": -1.8411651849746704,
|
2533 |
+
"logps/chosen": -80.52401733398438,
|
2534 |
+
"logps/rejected": -122.70835876464844,
|
2535 |
+
"loss": 0.0049,
|
2536 |
+
"rewards/accuracies": 1.0,
|
2537 |
+
"rewards/chosen": 3.3624215126037598,
|
2538 |
+
"rewards/margins": 10.361730575561523,
|
2539 |
+
"rewards/rejected": -6.999309539794922,
|
2540 |
+
"step": 1250
|
2541 |
+
},
|
2542 |
+
{
|
2543 |
+
"epoch": 5.14,
|
2544 |
+
"grad_norm": 0.004326160531491041,
|
2545 |
+
"learning_rate": 7.578947368421054e-07,
|
2546 |
+
"logits/chosen": -1.735650658607483,
|
2547 |
+
"logits/rejected": -1.7782208919525146,
|
2548 |
+
"logps/chosen": -143.6359100341797,
|
2549 |
+
"logps/rejected": -203.25271606445312,
|
2550 |
+
"loss": 0.0091,
|
2551 |
+
"rewards/accuracies": 1.0,
|
2552 |
+
"rewards/chosen": 2.7012696266174316,
|
2553 |
+
"rewards/margins": 9.915846824645996,
|
2554 |
+
"rewards/rejected": -7.214578151702881,
|
2555 |
+
"step": 1260
|
2556 |
+
},
|
2557 |
+
{
|
2558 |
+
"epoch": 5.14,
|
2559 |
+
"eval_logits/chosen": -1.8135929107666016,
|
2560 |
+
"eval_logits/rejected": -1.8531488180160522,
|
2561 |
+
"eval_logps/chosen": -189.6214599609375,
|
2562 |
+
"eval_logps/rejected": -264.03973388671875,
|
2563 |
+
"eval_loss": 0.10307549685239792,
|
2564 |
+
"eval_rewards/accuracies": 0.9642857313156128,
|
2565 |
+
"eval_rewards/chosen": 1.4764469861984253,
|
2566 |
+
"eval_rewards/margins": 7.73268461227417,
|
2567 |
+
"eval_rewards/rejected": -6.256237506866455,
|
2568 |
+
"eval_runtime": 26.2397,
|
2569 |
+
"eval_samples_per_second": 1.067,
|
2570 |
+
"eval_steps_per_second": 1.067,
|
2571 |
+
"step": 1260
|
2572 |
+
},
|
2573 |
+
{
|
2574 |
+
"epoch": 5.18,
|
2575 |
+
"grad_norm": 0.05577283725142479,
|
2576 |
+
"learning_rate": 7.228070175438597e-07,
|
2577 |
+
"logits/chosen": -1.9368791580200195,
|
2578 |
+
"logits/rejected": -1.9599593877792358,
|
2579 |
+
"logps/chosen": -194.0889434814453,
|
2580 |
+
"logps/rejected": -283.97613525390625,
|
2581 |
+
"loss": 0.0003,
|
2582 |
+
"rewards/accuracies": 1.0,
|
2583 |
+
"rewards/chosen": -0.7386444211006165,
|
2584 |
+
"rewards/margins": 11.283439636230469,
|
2585 |
+
"rewards/rejected": -12.02208423614502,
|
2586 |
+
"step": 1270
|
2587 |
+
},
|
2588 |
+
{
|
2589 |
+
"epoch": 5.22,
|
2590 |
+
"grad_norm": 6.575052975676954e-05,
|
2591 |
+
"learning_rate": 6.87719298245614e-07,
|
2592 |
+
"logits/chosen": -1.7652225494384766,
|
2593 |
+
"logits/rejected": -1.8098253011703491,
|
2594 |
+
"logps/chosen": -157.82101440429688,
|
2595 |
+
"logps/rejected": -254.159912109375,
|
2596 |
+
"loss": 0.0475,
|
2597 |
+
"rewards/accuracies": 1.0,
|
2598 |
+
"rewards/chosen": 3.464066982269287,
|
2599 |
+
"rewards/margins": 12.162757873535156,
|
2600 |
+
"rewards/rejected": -8.698690414428711,
|
2601 |
+
"step": 1280
|
2602 |
+
},
|
2603 |
+
{
|
2604 |
+
"epoch": 5.27,
|
2605 |
+
"grad_norm": 0.0,
|
2606 |
+
"learning_rate": 6.526315789473684e-07,
|
2607 |
+
"logits/chosen": -1.7878305912017822,
|
2608 |
+
"logits/rejected": -1.8565266132354736,
|
2609 |
+
"logps/chosen": -204.24099731445312,
|
2610 |
+
"logps/rejected": -325.3123779296875,
|
2611 |
+
"loss": 0.0526,
|
2612 |
+
"rewards/accuracies": 1.0,
|
2613 |
+
"rewards/chosen": 2.4723305702209473,
|
2614 |
+
"rewards/margins": 12.587270736694336,
|
2615 |
+
"rewards/rejected": -10.114938735961914,
|
2616 |
+
"step": 1290
|
2617 |
+
},
|
2618 |
+
{
|
2619 |
+
"epoch": 5.27,
|
2620 |
+
"eval_logits/chosen": -1.8132985830307007,
|
2621 |
+
"eval_logits/rejected": -1.8527573347091675,
|
2622 |
+
"eval_logps/chosen": -189.7008819580078,
|
2623 |
+
"eval_logps/rejected": -264.1980895996094,
|
2624 |
+
"eval_loss": 0.09966809302568436,
|
2625 |
+
"eval_rewards/accuracies": 0.9642857313156128,
|
2626 |
+
"eval_rewards/chosen": 1.4526145458221436,
|
2627 |
+
"eval_rewards/margins": 7.756356239318848,
|
2628 |
+
"eval_rewards/rejected": -6.303742408752441,
|
2629 |
+
"eval_runtime": 26.2377,
|
2630 |
+
"eval_samples_per_second": 1.067,
|
2631 |
+
"eval_steps_per_second": 1.067,
|
2632 |
+
"step": 1290
|
2633 |
+
},
|
2634 |
+
{
|
2635 |
+
"epoch": 5.31,
|
2636 |
+
"grad_norm": 6.653774864417983e-09,
|
2637 |
+
"learning_rate": 6.175438596491229e-07,
|
2638 |
+
"logits/chosen": -1.7762054204940796,
|
2639 |
+
"logits/rejected": -1.883049726486206,
|
2640 |
+
"logps/chosen": -146.65390014648438,
|
2641 |
+
"logps/rejected": -295.46466064453125,
|
2642 |
+
"loss": 0.0104,
|
2643 |
+
"rewards/accuracies": 1.0,
|
2644 |
+
"rewards/chosen": 0.7521862983703613,
|
2645 |
+
"rewards/margins": 11.713335037231445,
|
2646 |
+
"rewards/rejected": -10.961150169372559,
|
2647 |
+
"step": 1300
|
2648 |
+
},
|
2649 |
+
{
|
2650 |
+
"epoch": 5.35,
|
2651 |
+
"grad_norm": 2.788616418838501,
|
2652 |
+
"learning_rate": 5.824561403508772e-07,
|
2653 |
+
"logits/chosen": -1.8407360315322876,
|
2654 |
+
"logits/rejected": -1.9248844385147095,
|
2655 |
+
"logps/chosen": -156.55274963378906,
|
2656 |
+
"logps/rejected": -289.77484130859375,
|
2657 |
+
"loss": 0.2168,
|
2658 |
+
"rewards/accuracies": 0.8999999761581421,
|
2659 |
+
"rewards/chosen": 1.0223561525344849,
|
2660 |
+
"rewards/margins": 9.885419845581055,
|
2661 |
+
"rewards/rejected": -8.86306381225586,
|
2662 |
+
"step": 1310
|
2663 |
+
},
|
2664 |
+
{
|
2665 |
+
"epoch": 5.39,
|
2666 |
+
"grad_norm": 0.009542102925479412,
|
2667 |
+
"learning_rate": 5.473684210526316e-07,
|
2668 |
+
"logits/chosen": -1.7713199853897095,
|
2669 |
+
"logits/rejected": -1.8372914791107178,
|
2670 |
+
"logps/chosen": -130.9491729736328,
|
2671 |
+
"logps/rejected": -201.64227294921875,
|
2672 |
+
"loss": 0.0316,
|
2673 |
+
"rewards/accuracies": 1.0,
|
2674 |
+
"rewards/chosen": 1.3896652460098267,
|
2675 |
+
"rewards/margins": 9.91728687286377,
|
2676 |
+
"rewards/rejected": -8.527620315551758,
|
2677 |
+
"step": 1320
|
2678 |
+
},
|
2679 |
+
{
|
2680 |
+
"epoch": 5.39,
|
2681 |
+
"eval_logits/chosen": -1.812363624572754,
|
2682 |
+
"eval_logits/rejected": -1.8517435789108276,
|
2683 |
+
"eval_logps/chosen": -189.71922302246094,
|
2684 |
+
"eval_logps/rejected": -264.22357177734375,
|
2685 |
+
"eval_loss": 0.09649749845266342,
|
2686 |
+
"eval_rewards/accuracies": 0.9642857313156128,
|
2687 |
+
"eval_rewards/chosen": 1.4471155405044556,
|
2688 |
+
"eval_rewards/margins": 7.758502006530762,
|
2689 |
+
"eval_rewards/rejected": -6.3113861083984375,
|
2690 |
+
"eval_runtime": 26.23,
|
2691 |
+
"eval_samples_per_second": 1.067,
|
2692 |
+
"eval_steps_per_second": 1.067,
|
2693 |
+
"step": 1320
|
2694 |
+
},
|
2695 |
+
{
|
2696 |
+
"epoch": 5.43,
|
2697 |
+
"grad_norm": 25.36566162109375,
|
2698 |
+
"learning_rate": 5.12280701754386e-07,
|
2699 |
+
"logits/chosen": -1.800421118736267,
|
2700 |
+
"logits/rejected": -1.853594183921814,
|
2701 |
+
"logps/chosen": -154.73907470703125,
|
2702 |
+
"logps/rejected": -204.53807067871094,
|
2703 |
+
"loss": 0.015,
|
2704 |
+
"rewards/accuracies": 1.0,
|
2705 |
+
"rewards/chosen": 2.8498456478118896,
|
2706 |
+
"rewards/margins": 9.068418502807617,
|
2707 |
+
"rewards/rejected": -6.218573570251465,
|
2708 |
+
"step": 1330
|
2709 |
+
},
|
2710 |
+
{
|
2711 |
+
"epoch": 5.47,
|
2712 |
+
"grad_norm": 0.0038690047804266214,
|
2713 |
+
"learning_rate": 4.771929824561404e-07,
|
2714 |
+
"logits/chosen": -1.7142912149429321,
|
2715 |
+
"logits/rejected": -1.7878398895263672,
|
2716 |
+
"logps/chosen": -123.2061996459961,
|
2717 |
+
"logps/rejected": -255.41015625,
|
2718 |
+
"loss": 0.0066,
|
2719 |
+
"rewards/accuracies": 1.0,
|
2720 |
+
"rewards/chosen": 0.37266087532043457,
|
2721 |
+
"rewards/margins": 11.119722366333008,
|
2722 |
+
"rewards/rejected": -10.747061729431152,
|
2723 |
+
"step": 1340
|
2724 |
+
},
|
2725 |
+
{
|
2726 |
+
"epoch": 5.51,
|
2727 |
+
"grad_norm": 0.021649232134222984,
|
2728 |
+
"learning_rate": 4.421052631578947e-07,
|
2729 |
+
"logits/chosen": -1.895521879196167,
|
2730 |
+
"logits/rejected": -1.9479408264160156,
|
2731 |
+
"logps/chosen": -160.8917999267578,
|
2732 |
+
"logps/rejected": -227.413818359375,
|
2733 |
+
"loss": 0.0249,
|
2734 |
+
"rewards/accuracies": 1.0,
|
2735 |
+
"rewards/chosen": 0.1835147887468338,
|
2736 |
+
"rewards/margins": 7.590100288391113,
|
2737 |
+
"rewards/rejected": -7.406586647033691,
|
2738 |
+
"step": 1350
|
2739 |
+
},
|
2740 |
+
{
|
2741 |
+
"epoch": 5.51,
|
2742 |
+
"eval_logits/chosen": -1.811531662940979,
|
2743 |
+
"eval_logits/rejected": -1.850866436958313,
|
2744 |
+
"eval_logps/chosen": -189.75291442871094,
|
2745 |
+
"eval_logps/rejected": -264.3137512207031,
|
2746 |
+
"eval_loss": 0.09495646506547928,
|
2747 |
+
"eval_rewards/accuracies": 0.9642857313156128,
|
2748 |
+
"eval_rewards/chosen": 1.4370089769363403,
|
2749 |
+
"eval_rewards/margins": 7.775452136993408,
|
2750 |
+
"eval_rewards/rejected": -6.338443279266357,
|
2751 |
+
"eval_runtime": 26.2475,
|
2752 |
+
"eval_samples_per_second": 1.067,
|
2753 |
+
"eval_steps_per_second": 1.067,
|
2754 |
+
"step": 1350
|
2755 |
+
}
|
2756 |
+
],
|
2757 |
+
"logging_steps": 10,
|
2758 |
+
"max_steps": 1470,
|
2759 |
+
"num_input_tokens_seen": 0,
|
2760 |
+
"num_train_epochs": 6,
|
2761 |
+
"save_steps": 90,
|
2762 |
+
"total_flos": 0.0,
|
2763 |
+
"train_batch_size": 1,
|
2764 |
+
"trial_name": null,
|
2765 |
+
"trial_params": null
|
2766 |
+
}
|
checkpoint-1350/training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:41693834eff4b96ebb423cc3037dde0e25267c10f91f2e959260eee7bddf1ddb
|
3 |
+
size 5304
|
checkpoint-1440/README.md
ADDED
@@ -0,0 +1,202 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
library_name: peft
|
3 |
+
base_model: Weni/WeniGPT-Agents-Mistral-1.0.6-SFT-merged
|
4 |
+
---
|
5 |
+
|
6 |
+
# Model Card for Model ID
|
7 |
+
|
8 |
+
<!-- Provide a quick summary of what the model is/does. -->
|
9 |
+
|
10 |
+
|
11 |
+
|
12 |
+
## Model Details
|
13 |
+
|
14 |
+
### Model Description
|
15 |
+
|
16 |
+
<!-- Provide a longer summary of what this model is. -->
|
17 |
+
|
18 |
+
|
19 |
+
|
20 |
+
- **Developed by:** [More Information Needed]
|
21 |
+
- **Funded by [optional]:** [More Information Needed]
|
22 |
+
- **Shared by [optional]:** [More Information Needed]
|
23 |
+
- **Model type:** [More Information Needed]
|
24 |
+
- **Language(s) (NLP):** [More Information Needed]
|
25 |
+
- **License:** [More Information Needed]
|
26 |
+
- **Finetuned from model [optional]:** [More Information Needed]
|
27 |
+
|
28 |
+
### Model Sources [optional]
|
29 |
+
|
30 |
+
<!-- Provide the basic links for the model. -->
|
31 |
+
|
32 |
+
- **Repository:** [More Information Needed]
|
33 |
+
- **Paper [optional]:** [More Information Needed]
|
34 |
+
- **Demo [optional]:** [More Information Needed]
|
35 |
+
|
36 |
+
## Uses
|
37 |
+
|
38 |
+
<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
|
39 |
+
|
40 |
+
### Direct Use
|
41 |
+
|
42 |
+
<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
|
43 |
+
|
44 |
+
[More Information Needed]
|
45 |
+
|
46 |
+
### Downstream Use [optional]
|
47 |
+
|
48 |
+
<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
|
49 |
+
|
50 |
+
[More Information Needed]
|
51 |
+
|
52 |
+
### Out-of-Scope Use
|
53 |
+
|
54 |
+
<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
|
55 |
+
|
56 |
+
[More Information Needed]
|
57 |
+
|
58 |
+
## Bias, Risks, and Limitations
|
59 |
+
|
60 |
+
<!-- This section is meant to convey both technical and sociotechnical limitations. -->
|
61 |
+
|
62 |
+
[More Information Needed]
|
63 |
+
|
64 |
+
### Recommendations
|
65 |
+
|
66 |
+
<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
|
67 |
+
|
68 |
+
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
|
69 |
+
|
70 |
+
## How to Get Started with the Model
|
71 |
+
|
72 |
+
Use the code below to get started with the model.
|
73 |
+
|
74 |
+
[More Information Needed]
|
75 |
+
|
76 |
+
## Training Details
|
77 |
+
|
78 |
+
### Training Data
|
79 |
+
|
80 |
+
<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
|
81 |
+
|
82 |
+
[More Information Needed]
|
83 |
+
|
84 |
+
### Training Procedure
|
85 |
+
|
86 |
+
<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
|
87 |
+
|
88 |
+
#### Preprocessing [optional]
|
89 |
+
|
90 |
+
[More Information Needed]
|
91 |
+
|
92 |
+
|
93 |
+
#### Training Hyperparameters
|
94 |
+
|
95 |
+
- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
|
96 |
+
|
97 |
+
#### Speeds, Sizes, Times [optional]
|
98 |
+
|
99 |
+
<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
|
100 |
+
|
101 |
+
[More Information Needed]
|
102 |
+
|
103 |
+
## Evaluation
|
104 |
+
|
105 |
+
<!-- This section describes the evaluation protocols and provides the results. -->
|
106 |
+
|
107 |
+
### Testing Data, Factors & Metrics
|
108 |
+
|
109 |
+
#### Testing Data
|
110 |
+
|
111 |
+
<!-- This should link to a Dataset Card if possible. -->
|
112 |
+
|
113 |
+
[More Information Needed]
|
114 |
+
|
115 |
+
#### Factors
|
116 |
+
|
117 |
+
<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
|
118 |
+
|
119 |
+
[More Information Needed]
|
120 |
+
|
121 |
+
#### Metrics
|
122 |
+
|
123 |
+
<!-- These are the evaluation metrics being used, ideally with a description of why. -->
|
124 |
+
|
125 |
+
[More Information Needed]
|
126 |
+
|
127 |
+
### Results
|
128 |
+
|
129 |
+
[More Information Needed]
|
130 |
+
|
131 |
+
#### Summary
|
132 |
+
|
133 |
+
|
134 |
+
|
135 |
+
## Model Examination [optional]
|
136 |
+
|
137 |
+
<!-- Relevant interpretability work for the model goes here -->
|
138 |
+
|
139 |
+
[More Information Needed]
|
140 |
+
|
141 |
+
## Environmental Impact
|
142 |
+
|
143 |
+
<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
|
144 |
+
|
145 |
+
Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
|
146 |
+
|
147 |
+
- **Hardware Type:** [More Information Needed]
|
148 |
+
- **Hours used:** [More Information Needed]
|
149 |
+
- **Cloud Provider:** [More Information Needed]
|
150 |
+
- **Compute Region:** [More Information Needed]
|
151 |
+
- **Carbon Emitted:** [More Information Needed]
|
152 |
+
|
153 |
+
## Technical Specifications [optional]
|
154 |
+
|
155 |
+
### Model Architecture and Objective
|
156 |
+
|
157 |
+
[More Information Needed]
|
158 |
+
|
159 |
+
### Compute Infrastructure
|
160 |
+
|
161 |
+
[More Information Needed]
|
162 |
+
|
163 |
+
#### Hardware
|
164 |
+
|
165 |
+
[More Information Needed]
|
166 |
+
|
167 |
+
#### Software
|
168 |
+
|
169 |
+
[More Information Needed]
|
170 |
+
|
171 |
+
## Citation [optional]
|
172 |
+
|
173 |
+
<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
|
174 |
+
|
175 |
+
**BibTeX:**
|
176 |
+
|
177 |
+
[More Information Needed]
|
178 |
+
|
179 |
+
**APA:**
|
180 |
+
|
181 |
+
[More Information Needed]
|
182 |
+
|
183 |
+
## Glossary [optional]
|
184 |
+
|
185 |
+
<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
|
186 |
+
|
187 |
+
[More Information Needed]
|
188 |
+
|
189 |
+
## More Information [optional]
|
190 |
+
|
191 |
+
[More Information Needed]
|
192 |
+
|
193 |
+
## Model Card Authors [optional]
|
194 |
+
|
195 |
+
[More Information Needed]
|
196 |
+
|
197 |
+
## Model Card Contact
|
198 |
+
|
199 |
+
[More Information Needed]
|
200 |
+
### Framework versions
|
201 |
+
|
202 |
+
- PEFT 0.10.0
|