Commit
•
aa9c5c8
1
Parent(s):
635e5c2
Model save
Browse files- README.md +80 -0
- all_results.json +9 -0
- train_results.json +9 -0
- trainer_state.json +1098 -0
README.md
ADDED
|
@@ -0,0 +1,80 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: mistralai/Mistral-7B-v0.1
|
| 3 |
+
library_name: peft
|
| 4 |
+
license: apache-2.0
|
| 5 |
+
tags:
|
| 6 |
+
- trl
|
| 7 |
+
- dpo
|
| 8 |
+
- generated_from_trainer
|
| 9 |
+
model-index:
|
| 10 |
+
- name: zephyr-7b-dpo-qlora
|
| 11 |
+
results: []
|
| 12 |
+
---
|
| 13 |
+
|
| 14 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 15 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 16 |
+
|
| 17 |
+
# zephyr-7b-dpo-qlora
|
| 18 |
+
|
| 19 |
+
This model is a fine-tuned version of [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) on an unknown dataset.
|
| 20 |
+
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 0.5021
|
| 22 |
+
- Rewards/chosen: -2.1712
|
| 23 |
+
- Rewards/rejected: -2.9608
|
| 24 |
+
- Rewards/accuracies: 0.7083
|
| 25 |
+
- Rewards/margins: 0.7896
|
| 26 |
+
- Logps/rejected: -535.6057
|
| 27 |
+
- Logps/chosen: -475.6837
|
| 28 |
+
- Logits/rejected: -1.1768
|
| 29 |
+
- Logits/chosen: -1.2825
|
| 30 |
+
|
| 31 |
+
## Model description
|
| 32 |
+
|
| 33 |
+
More information needed
|
| 34 |
+
|
| 35 |
+
## Intended uses & limitations
|
| 36 |
+
|
| 37 |
+
More information needed
|
| 38 |
+
|
| 39 |
+
## Training and evaluation data
|
| 40 |
+
|
| 41 |
+
More information needed
|
| 42 |
+
|
| 43 |
+
## Training procedure
|
| 44 |
+
|
| 45 |
+
### Training hyperparameters
|
| 46 |
+
|
| 47 |
+
The following hyperparameters were used during training:
|
| 48 |
+
- learning_rate: 5e-06
|
| 49 |
+
- train_batch_size: 4
|
| 50 |
+
- eval_batch_size: 4
|
| 51 |
+
- seed: 42
|
| 52 |
+
- distributed_type: multi-GPU
|
| 53 |
+
- num_devices: 6
|
| 54 |
+
- gradient_accumulation_steps: 4
|
| 55 |
+
- total_train_batch_size: 96
|
| 56 |
+
- total_eval_batch_size: 24
|
| 57 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 58 |
+
- lr_scheduler_type: cosine
|
| 59 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 60 |
+
- num_epochs: 1
|
| 61 |
+
|
| 62 |
+
### Training results
|
| 63 |
+
|
| 64 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 65 |
+
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 66 |
+
| 0.6099 | 0.1570 | 100 | 0.5999 | -0.4571 | -0.6755 | 0.6369 | 0.2184 | -307.0734 | -304.2707 | -2.1308 | -2.2191 |
|
| 67 |
+
| 0.5468 | 0.3140 | 200 | 0.5354 | -1.2063 | -1.7025 | 0.6815 | 0.4962 | -409.7740 | -379.1950 | -1.5157 | -1.6078 |
|
| 68 |
+
| 0.5195 | 0.4710 | 300 | 0.5227 | -1.5981 | -2.2831 | 0.7083 | 0.6849 | -467.8293 | -418.3782 | -1.3523 | -1.4527 |
|
| 69 |
+
| 0.4895 | 0.6279 | 400 | 0.5142 | -1.8555 | -2.6654 | 0.6994 | 0.8099 | -506.0622 | -444.1171 | -1.1070 | -1.2180 |
|
| 70 |
+
| 0.4992 | 0.7849 | 500 | 0.5019 | -2.3330 | -3.1029 | 0.7054 | 0.7699 | -549.8137 | -491.8629 | -1.1520 | -1.2589 |
|
| 71 |
+
| 0.5001 | 0.9419 | 600 | 0.5021 | -2.1712 | -2.9608 | 0.7083 | 0.7896 | -535.6057 | -475.6837 | -1.1768 | -1.2825 |
|
| 72 |
+
|
| 73 |
+
|
| 74 |
+
### Framework versions
|
| 75 |
+
|
| 76 |
+
- PEFT 0.13.2
|
| 77 |
+
- Transformers 4.45.2
|
| 78 |
+
- Pytorch 2.1.2
|
| 79 |
+
- Datasets 3.0.1
|
| 80 |
+
- Tokenizers 0.20.1
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5397420258873861,
|
| 5 |
+
"train_runtime": 14909.9591,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 4.1,
|
| 8 |
+
"train_steps_per_second": 0.043
|
| 9 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5397420258873861,
|
| 5 |
+
"train_runtime": 14909.9591,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 4.1,
|
| 8 |
+
"train_steps_per_second": 0.043
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,1098 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.0,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 637,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.0015698587127158557,
|
| 13 |
+
"grad_norm": 0.982062816619873,
|
| 14 |
+
"learning_rate": 7.8125e-08,
|
| 15 |
+
"logits/chosen": -1.9796161651611328,
|
| 16 |
+
"logits/rejected": -1.7990202903747559,
|
| 17 |
+
"logps/chosen": -276.0213317871094,
|
| 18 |
+
"logps/rejected": -220.55433654785156,
|
| 19 |
+
"loss": 0.6931,
|
| 20 |
+
"rewards/accuracies": 0.0,
|
| 21 |
+
"rewards/chosen": 0.0,
|
| 22 |
+
"rewards/margins": 0.0,
|
| 23 |
+
"rewards/rejected": 0.0,
|
| 24 |
+
"step": 1
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.015698587127158554,
|
| 28 |
+
"grad_norm": 0.9399773478507996,
|
| 29 |
+
"learning_rate": 7.8125e-07,
|
| 30 |
+
"logits/chosen": -2.2138850688934326,
|
| 31 |
+
"logits/rejected": -2.1262121200561523,
|
| 32 |
+
"logps/chosen": -290.8321533203125,
|
| 33 |
+
"logps/rejected": -228.361328125,
|
| 34 |
+
"loss": 0.6929,
|
| 35 |
+
"rewards/accuracies": 0.4652777910232544,
|
| 36 |
+
"rewards/chosen": 0.00492230337113142,
|
| 37 |
+
"rewards/margins": 0.0006595260347239673,
|
| 38 |
+
"rewards/rejected": 0.0042627775110304356,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.03139717425431711,
|
| 43 |
+
"grad_norm": 1.019684076309204,
|
| 44 |
+
"learning_rate": 1.5625e-06,
|
| 45 |
+
"logits/chosen": -2.3072190284729004,
|
| 46 |
+
"logits/rejected": -2.201758861541748,
|
| 47 |
+
"logps/chosen": -265.79144287109375,
|
| 48 |
+
"logps/rejected": -252.1865997314453,
|
| 49 |
+
"loss": 0.6895,
|
| 50 |
+
"rewards/accuracies": 0.643750011920929,
|
| 51 |
+
"rewards/chosen": 0.029890870675444603,
|
| 52 |
+
"rewards/margins": 0.009029369801282883,
|
| 53 |
+
"rewards/rejected": 0.02086150087416172,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.04709576138147567,
|
| 58 |
+
"grad_norm": 0.9330567121505737,
|
| 59 |
+
"learning_rate": 2.3437500000000002e-06,
|
| 60 |
+
"logits/chosen": -2.312687635421753,
|
| 61 |
+
"logits/rejected": -2.206019878387451,
|
| 62 |
+
"logps/chosen": -268.0202331542969,
|
| 63 |
+
"logps/rejected": -250.2025146484375,
|
| 64 |
+
"loss": 0.6844,
|
| 65 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 66 |
+
"rewards/chosen": 0.05118671804666519,
|
| 67 |
+
"rewards/margins": 0.019565049558877945,
|
| 68 |
+
"rewards/rejected": 0.03162166848778725,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.06279434850863422,
|
| 73 |
+
"grad_norm": 0.9986972808837891,
|
| 74 |
+
"learning_rate": 3.125e-06,
|
| 75 |
+
"logits/chosen": -2.2849414348602295,
|
| 76 |
+
"logits/rejected": -2.1605541706085205,
|
| 77 |
+
"logps/chosen": -248.8426971435547,
|
| 78 |
+
"logps/rejected": -222.5389404296875,
|
| 79 |
+
"loss": 0.675,
|
| 80 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 81 |
+
"rewards/chosen": 0.06607525795698166,
|
| 82 |
+
"rewards/margins": 0.039594851434230804,
|
| 83 |
+
"rewards/rejected": 0.026480406522750854,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.07849293563579278,
|
| 88 |
+
"grad_norm": 1.18948495388031,
|
| 89 |
+
"learning_rate": 3.90625e-06,
|
| 90 |
+
"logits/chosen": -2.237248659133911,
|
| 91 |
+
"logits/rejected": -2.190307855606079,
|
| 92 |
+
"logps/chosen": -268.77899169921875,
|
| 93 |
+
"logps/rejected": -278.4407958984375,
|
| 94 |
+
"loss": 0.6605,
|
| 95 |
+
"rewards/accuracies": 0.71875,
|
| 96 |
+
"rewards/chosen": 0.00511029502376914,
|
| 97 |
+
"rewards/margins": 0.05497710779309273,
|
| 98 |
+
"rewards/rejected": -0.04986681044101715,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.09419152276295134,
|
| 103 |
+
"grad_norm": 1.285398006439209,
|
| 104 |
+
"learning_rate": 4.6875000000000004e-06,
|
| 105 |
+
"logits/chosen": -2.265840768814087,
|
| 106 |
+
"logits/rejected": -2.1778528690338135,
|
| 107 |
+
"logps/chosen": -272.609130859375,
|
| 108 |
+
"logps/rejected": -277.2625732421875,
|
| 109 |
+
"loss": 0.6429,
|
| 110 |
+
"rewards/accuracies": 0.71875,
|
| 111 |
+
"rewards/chosen": -0.07459741830825806,
|
| 112 |
+
"rewards/margins": 0.12982353568077087,
|
| 113 |
+
"rewards/rejected": -0.20442095398902893,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.10989010989010989,
|
| 118 |
+
"grad_norm": 1.8551888465881348,
|
| 119 |
+
"learning_rate": 4.998647417232375e-06,
|
| 120 |
+
"logits/chosen": -2.1861605644226074,
|
| 121 |
+
"logits/rejected": -2.0909242630004883,
|
| 122 |
+
"logps/chosen": -267.07611083984375,
|
| 123 |
+
"logps/rejected": -259.22210693359375,
|
| 124 |
+
"loss": 0.6386,
|
| 125 |
+
"rewards/accuracies": 0.643750011920929,
|
| 126 |
+
"rewards/chosen": -0.2929137647151947,
|
| 127 |
+
"rewards/margins": 0.1055971160531044,
|
| 128 |
+
"rewards/rejected": -0.3985108435153961,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.12558869701726844,
|
| 133 |
+
"grad_norm": 1.9893733263015747,
|
| 134 |
+
"learning_rate": 4.990386933279973e-06,
|
| 135 |
+
"logits/chosen": -2.184288740158081,
|
| 136 |
+
"logits/rejected": -2.1073215007781982,
|
| 137 |
+
"logps/chosen": -290.60943603515625,
|
| 138 |
+
"logps/rejected": -284.17425537109375,
|
| 139 |
+
"loss": 0.6057,
|
| 140 |
+
"rewards/accuracies": 0.699999988079071,
|
| 141 |
+
"rewards/chosen": -0.30957552790641785,
|
| 142 |
+
"rewards/margins": 0.2696777582168579,
|
| 143 |
+
"rewards/rejected": -0.5792533159255981,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.141287284144427,
|
| 148 |
+
"grad_norm": 3.498678684234619,
|
| 149 |
+
"learning_rate": 4.974642195009681e-06,
|
| 150 |
+
"logits/chosen": -2.1730189323425293,
|
| 151 |
+
"logits/rejected": -2.092939853668213,
|
| 152 |
+
"logps/chosen": -326.27838134765625,
|
| 153 |
+
"logps/rejected": -310.67767333984375,
|
| 154 |
+
"loss": 0.6017,
|
| 155 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 156 |
+
"rewards/chosen": -0.5229882001876831,
|
| 157 |
+
"rewards/margins": 0.20935317873954773,
|
| 158 |
+
"rewards/rejected": -0.7323412895202637,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.15698587127158556,
|
| 163 |
+
"grad_norm": 2.8239424228668213,
|
| 164 |
+
"learning_rate": 4.951460519416228e-06,
|
| 165 |
+
"logits/chosen": -2.254025936126709,
|
| 166 |
+
"logits/rejected": -2.1698710918426514,
|
| 167 |
+
"logps/chosen": -328.6602783203125,
|
| 168 |
+
"logps/rejected": -333.0082092285156,
|
| 169 |
+
"loss": 0.6099,
|
| 170 |
+
"rewards/accuracies": 0.6875,
|
| 171 |
+
"rewards/chosen": -0.4128250479698181,
|
| 172 |
+
"rewards/margins": 0.2569137513637543,
|
| 173 |
+
"rewards/rejected": -0.6697388887405396,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 0.15698587127158556,
|
| 178 |
+
"eval_logits/chosen": -2.2190940380096436,
|
| 179 |
+
"eval_logits/rejected": -2.130751371383667,
|
| 180 |
+
"eval_logps/chosen": -304.2706604003906,
|
| 181 |
+
"eval_logps/rejected": -307.0733642578125,
|
| 182 |
+
"eval_loss": 0.59989333152771,
|
| 183 |
+
"eval_rewards/accuracies": 0.636904776096344,
|
| 184 |
+
"eval_rewards/chosen": -0.45707327127456665,
|
| 185 |
+
"eval_rewards/margins": 0.2184356451034546,
|
| 186 |
+
"eval_rewards/rejected": -0.675508975982666,
|
| 187 |
+
"eval_runtime": 219.8057,
|
| 188 |
+
"eval_samples_per_second": 9.099,
|
| 189 |
+
"eval_steps_per_second": 0.382,
|
| 190 |
+
"step": 100
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 0.1726844583987441,
|
| 194 |
+
"grad_norm": 4.492749214172363,
|
| 195 |
+
"learning_rate": 4.920911573406925e-06,
|
| 196 |
+
"logits/chosen": -2.2060790061950684,
|
| 197 |
+
"logits/rejected": -2.0495288372039795,
|
| 198 |
+
"logps/chosen": -324.7986755371094,
|
| 199 |
+
"logps/rejected": -339.40802001953125,
|
| 200 |
+
"loss": 0.5822,
|
| 201 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 202 |
+
"rewards/chosen": -0.5525876879692078,
|
| 203 |
+
"rewards/margins": 0.44238606095314026,
|
| 204 |
+
"rewards/rejected": -0.9949736595153809,
|
| 205 |
+
"step": 110
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.18838304552590268,
|
| 209 |
+
"grad_norm": 4.563241481781006,
|
| 210 |
+
"learning_rate": 4.883087164434672e-06,
|
| 211 |
+
"logits/chosen": -2.0682525634765625,
|
| 212 |
+
"logits/rejected": -2.0306172370910645,
|
| 213 |
+
"logps/chosen": -367.96966552734375,
|
| 214 |
+
"logps/rejected": -406.1026611328125,
|
| 215 |
+
"loss": 0.5553,
|
| 216 |
+
"rewards/accuracies": 0.71875,
|
| 217 |
+
"rewards/chosen": -0.9666538238525391,
|
| 218 |
+
"rewards/margins": 0.4744674563407898,
|
| 219 |
+
"rewards/rejected": -1.4411213397979736,
|
| 220 |
+
"step": 120
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.20408163265306123,
|
| 224 |
+
"grad_norm": 4.36543607711792,
|
| 225 |
+
"learning_rate": 4.8381009645929044e-06,
|
| 226 |
+
"logits/chosen": -1.958156943321228,
|
| 227 |
+
"logits/rejected": -1.8605877161026,
|
| 228 |
+
"logps/chosen": -372.92083740234375,
|
| 229 |
+
"logps/rejected": -398.33209228515625,
|
| 230 |
+
"loss": 0.5609,
|
| 231 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 232 |
+
"rewards/chosen": -0.9711011052131653,
|
| 233 |
+
"rewards/margins": 0.45527857542037964,
|
| 234 |
+
"rewards/rejected": -1.426379680633545,
|
| 235 |
+
"step": 130
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.21978021978021978,
|
| 239 |
+
"grad_norm": 3.9676990509033203,
|
| 240 |
+
"learning_rate": 4.786088169001671e-06,
|
| 241 |
+
"logits/chosen": -1.9477580785751343,
|
| 242 |
+
"logits/rejected": -1.9076995849609375,
|
| 243 |
+
"logps/chosen": -364.2967224121094,
|
| 244 |
+
"logps/rejected": -398.6386413574219,
|
| 245 |
+
"loss": 0.5805,
|
| 246 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 247 |
+
"rewards/chosen": -1.2606209516525269,
|
| 248 |
+
"rewards/margins": 0.37748971581459045,
|
| 249 |
+
"rewards/rejected": -1.6381107568740845,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.23547880690737832,
|
| 254 |
+
"grad_norm": 4.215748310089111,
|
| 255 |
+
"learning_rate": 4.727205089511466e-06,
|
| 256 |
+
"logits/chosen": -1.8670552968978882,
|
| 257 |
+
"logits/rejected": -1.8230712413787842,
|
| 258 |
+
"logps/chosen": -380.49237060546875,
|
| 259 |
+
"logps/rejected": -442.67596435546875,
|
| 260 |
+
"loss": 0.5594,
|
| 261 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 262 |
+
"rewards/chosen": -1.3134872913360596,
|
| 263 |
+
"rewards/margins": 0.47393256425857544,
|
| 264 |
+
"rewards/rejected": -1.7874199151992798,
|
| 265 |
+
"step": 150
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.25117739403453687,
|
| 269 |
+
"grad_norm": 4.537078380584717,
|
| 270 |
+
"learning_rate": 4.661628684945851e-06,
|
| 271 |
+
"logits/chosen": -1.8435173034667969,
|
| 272 |
+
"logits/rejected": -1.7608001232147217,
|
| 273 |
+
"logps/chosen": -380.8707580566406,
|
| 274 |
+
"logps/rejected": -421.7408142089844,
|
| 275 |
+
"loss": 0.5537,
|
| 276 |
+
"rewards/accuracies": 0.737500011920929,
|
| 277 |
+
"rewards/chosen": -1.0851117372512817,
|
| 278 |
+
"rewards/margins": 0.5639781355857849,
|
| 279 |
+
"rewards/rejected": -1.6490898132324219,
|
| 280 |
+
"step": 160
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.2668759811616955,
|
| 284 |
+
"grad_norm": 4.127157211303711,
|
| 285 |
+
"learning_rate": 4.5895560292946e-06,
|
| 286 |
+
"logits/chosen": -1.8356807231903076,
|
| 287 |
+
"logits/rejected": -1.7069345712661743,
|
| 288 |
+
"logps/chosen": -406.5726623535156,
|
| 289 |
+
"logps/rejected": -423.89862060546875,
|
| 290 |
+
"loss": 0.5395,
|
| 291 |
+
"rewards/accuracies": 0.737500011920929,
|
| 292 |
+
"rewards/chosen": -1.350484848022461,
|
| 293 |
+
"rewards/margins": 0.5463389754295349,
|
| 294 |
+
"rewards/rejected": -1.8968238830566406,
|
| 295 |
+
"step": 170
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 0.282574568288854,
|
| 299 |
+
"grad_norm": 4.295073986053467,
|
| 300 |
+
"learning_rate": 4.511203719455588e-06,
|
| 301 |
+
"logits/chosen": -1.805424690246582,
|
| 302 |
+
"logits/rejected": -1.675389051437378,
|
| 303 |
+
"logps/chosen": -404.70721435546875,
|
| 304 |
+
"logps/rejected": -431.9833984375,
|
| 305 |
+
"loss": 0.539,
|
| 306 |
+
"rewards/accuracies": 0.699999988079071,
|
| 307 |
+
"rewards/chosen": -1.289776086807251,
|
| 308 |
+
"rewards/margins": 0.7749110460281372,
|
| 309 |
+
"rewards/rejected": -2.0646872520446777,
|
| 310 |
+
"step": 180
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.29827315541601257,
|
| 314 |
+
"grad_norm": 4.170295715332031,
|
| 315 |
+
"learning_rate": 4.426807224305315e-06,
|
| 316 |
+
"logits/chosen": -1.7862762212753296,
|
| 317 |
+
"logits/rejected": -1.635716438293457,
|
| 318 |
+
"logps/chosen": -372.5980529785156,
|
| 319 |
+
"logps/rejected": -416.7577209472656,
|
| 320 |
+
"loss": 0.5637,
|
| 321 |
+
"rewards/accuracies": 0.762499988079071,
|
| 322 |
+
"rewards/chosen": -1.134557843208313,
|
| 323 |
+
"rewards/margins": 0.753765344619751,
|
| 324 |
+
"rewards/rejected": -1.888323187828064,
|
| 325 |
+
"step": 190
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 0.3139717425431711,
|
| 329 |
+
"grad_norm": 4.384995937347412,
|
| 330 |
+
"learning_rate": 4.336620177054269e-06,
|
| 331 |
+
"logits/chosen": -1.7506341934204102,
|
| 332 |
+
"logits/rejected": -1.6495962142944336,
|
| 333 |
+
"logps/chosen": -347.5675048828125,
|
| 334 |
+
"logps/rejected": -391.43731689453125,
|
| 335 |
+
"loss": 0.5468,
|
| 336 |
+
"rewards/accuracies": 0.78125,
|
| 337 |
+
"rewards/chosen": -1.009634256362915,
|
| 338 |
+
"rewards/margins": 0.6290403604507446,
|
| 339 |
+
"rewards/rejected": -1.6386747360229492,
|
| 340 |
+
"step": 200
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 0.3139717425431711,
|
| 344 |
+
"eval_logits/chosen": -1.607847809791565,
|
| 345 |
+
"eval_logits/rejected": -1.5156790018081665,
|
| 346 |
+
"eval_logps/chosen": -379.1949768066406,
|
| 347 |
+
"eval_logps/rejected": -409.77398681640625,
|
| 348 |
+
"eval_loss": 0.5353844165802002,
|
| 349 |
+
"eval_rewards/accuracies": 0.6815476417541504,
|
| 350 |
+
"eval_rewards/chosen": -1.2063161134719849,
|
| 351 |
+
"eval_rewards/margins": 0.49619969725608826,
|
| 352 |
+
"eval_rewards/rejected": -1.702515721321106,
|
| 353 |
+
"eval_runtime": 219.5785,
|
| 354 |
+
"eval_samples_per_second": 9.108,
|
| 355 |
+
"eval_steps_per_second": 0.383,
|
| 356 |
+
"step": 200
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 0.32967032967032966,
|
| 360 |
+
"grad_norm": 4.753127098083496,
|
| 361 |
+
"learning_rate": 4.240913613013785e-06,
|
| 362 |
+
"logits/chosen": -1.6041982173919678,
|
| 363 |
+
"logits/rejected": -1.4586137533187866,
|
| 364 |
+
"logps/chosen": -403.050537109375,
|
| 365 |
+
"logps/rejected": -434.6443786621094,
|
| 366 |
+
"loss": 0.5589,
|
| 367 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 368 |
+
"rewards/chosen": -1.2178317308425903,
|
| 369 |
+
"rewards/margins": 0.7227036952972412,
|
| 370 |
+
"rewards/rejected": -1.940535306930542,
|
| 371 |
+
"step": 210
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 0.3453689167974882,
|
| 375 |
+
"grad_norm": 4.243059158325195,
|
| 376 |
+
"learning_rate": 4.139975155065109e-06,
|
| 377 |
+
"logits/chosen": -1.545710802078247,
|
| 378 |
+
"logits/rejected": -1.4544004201889038,
|
| 379 |
+
"logps/chosen": -359.60174560546875,
|
| 380 |
+
"logps/rejected": -431.0966796875,
|
| 381 |
+
"loss": 0.5293,
|
| 382 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 383 |
+
"rewards/chosen": -1.2607206106185913,
|
| 384 |
+
"rewards/margins": 0.6980492472648621,
|
| 385 |
+
"rewards/rejected": -1.9587697982788086,
|
| 386 |
+
"step": 220
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 0.36106750392464676,
|
| 390 |
+
"grad_norm": 4.250452995300293,
|
| 391 |
+
"learning_rate": 4.034108149278544e-06,
|
| 392 |
+
"logits/chosen": -1.5040862560272217,
|
| 393 |
+
"logits/rejected": -1.419395923614502,
|
| 394 |
+
"logps/chosen": -438.34979248046875,
|
| 395 |
+
"logps/rejected": -493.54608154296875,
|
| 396 |
+
"loss": 0.5378,
|
| 397 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 398 |
+
"rewards/chosen": -1.6042639017105103,
|
| 399 |
+
"rewards/margins": 0.6156054735183716,
|
| 400 |
+
"rewards/rejected": -2.2198691368103027,
|
| 401 |
+
"step": 230
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 0.37676609105180536,
|
| 405 |
+
"grad_norm": 5.608371734619141,
|
| 406 |
+
"learning_rate": 3.923630753280358e-06,
|
| 407 |
+
"logits/chosen": -1.5527160167694092,
|
| 408 |
+
"logits/rejected": -1.4736815690994263,
|
| 409 |
+
"logps/chosen": -357.8080749511719,
|
| 410 |
+
"logps/rejected": -439.68218994140625,
|
| 411 |
+
"loss": 0.4945,
|
| 412 |
+
"rewards/accuracies": 0.75,
|
| 413 |
+
"rewards/chosen": -0.9829418063163757,
|
| 414 |
+
"rewards/margins": 0.868217945098877,
|
| 415 |
+
"rewards/rejected": -1.8511598110198975,
|
| 416 |
+
"step": 240
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 0.3924646781789639,
|
| 420 |
+
"grad_norm": 4.367254734039307,
|
| 421 |
+
"learning_rate": 3.80887498010715e-06,
|
| 422 |
+
"logits/chosen": -1.4042975902557373,
|
| 423 |
+
"logits/rejected": -1.283562421798706,
|
| 424 |
+
"logps/chosen": -437.2142639160156,
|
| 425 |
+
"logps/rejected": -488.03826904296875,
|
| 426 |
+
"loss": 0.5232,
|
| 427 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 428 |
+
"rewards/chosen": -1.8854888677597046,
|
| 429 |
+
"rewards/margins": 0.7756179571151733,
|
| 430 |
+
"rewards/rejected": -2.661106824874878,
|
| 431 |
+
"step": 250
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 0.40816326530612246,
|
| 435 |
+
"grad_norm": 3.6756815910339355,
|
| 436 |
+
"learning_rate": 3.690185700421145e-06,
|
| 437 |
+
"logits/chosen": -1.534239411354065,
|
| 438 |
+
"logits/rejected": -1.4822893142700195,
|
| 439 |
+
"logps/chosen": -338.0312805175781,
|
| 440 |
+
"logps/rejected": -372.50274658203125,
|
| 441 |
+
"loss": 0.5643,
|
| 442 |
+
"rewards/accuracies": 0.699999988079071,
|
| 443 |
+
"rewards/chosen": -0.9809848666191101,
|
| 444 |
+
"rewards/margins": 0.43755441904067993,
|
| 445 |
+
"rewards/rejected": -1.41853928565979,
|
| 446 |
+
"step": 260
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 0.423861852433281,
|
| 450 |
+
"grad_norm": 4.209615707397461,
|
| 451 |
+
"learning_rate": 3.567919606085004e-06,
|
| 452 |
+
"logits/chosen": -1.6105585098266602,
|
| 453 |
+
"logits/rejected": -1.5416157245635986,
|
| 454 |
+
"logps/chosen": -353.83245849609375,
|
| 455 |
+
"logps/rejected": -427.657470703125,
|
| 456 |
+
"loss": 0.5363,
|
| 457 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 458 |
+
"rewards/chosen": -0.8973706364631653,
|
| 459 |
+
"rewards/margins": 0.6859413385391235,
|
| 460 |
+
"rewards/rejected": -1.5833120346069336,
|
| 461 |
+
"step": 270
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.43956043956043955,
|
| 465 |
+
"grad_norm": 5.80835485458374,
|
| 466 |
+
"learning_rate": 3.442444138210883e-06,
|
| 467 |
+
"logits/chosen": -1.3900154829025269,
|
| 468 |
+
"logits/rejected": -1.286752700805664,
|
| 469 |
+
"logps/chosen": -383.26959228515625,
|
| 470 |
+
"logps/rejected": -396.59735107421875,
|
| 471 |
+
"loss": 0.5434,
|
| 472 |
+
"rewards/accuracies": 0.675000011920929,
|
| 473 |
+
"rewards/chosen": -1.367221474647522,
|
| 474 |
+
"rewards/margins": 0.605404257774353,
|
| 475 |
+
"rewards/rejected": -1.972625732421875,
|
| 476 |
+
"step": 280
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 0.4552590266875981,
|
| 480 |
+
"grad_norm": 5.807376861572266,
|
| 481 |
+
"learning_rate": 3.314136382905234e-06,
|
| 482 |
+
"logits/chosen": -1.4290958642959595,
|
| 483 |
+
"logits/rejected": -1.3117034435272217,
|
| 484 |
+
"logps/chosen": -405.759765625,
|
| 485 |
+
"logps/rejected": -466.0645446777344,
|
| 486 |
+
"loss": 0.5277,
|
| 487 |
+
"rewards/accuracies": 0.731249988079071,
|
| 488 |
+
"rewards/chosen": -1.4295778274536133,
|
| 489 |
+
"rewards/margins": 0.7734361886978149,
|
| 490 |
+
"rewards/rejected": -2.2030138969421387,
|
| 491 |
+
"step": 290
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.47095761381475665,
|
| 495 |
+
"grad_norm": 4.498719215393066,
|
| 496 |
+
"learning_rate": 3.1833819380279028e-06,
|
| 497 |
+
"logits/chosen": -1.4070186614990234,
|
| 498 |
+
"logits/rejected": -1.3326956033706665,
|
| 499 |
+
"logps/chosen": -398.8482666015625,
|
| 500 |
+
"logps/rejected": -497.5279235839844,
|
| 501 |
+
"loss": 0.5195,
|
| 502 |
+
"rewards/accuracies": 0.737500011920929,
|
| 503 |
+
"rewards/chosen": -1.5320394039154053,
|
| 504 |
+
"rewards/margins": 0.9529226422309875,
|
| 505 |
+
"rewards/rejected": -2.484961986541748,
|
| 506 |
+
"step": 300
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 0.47095761381475665,
|
| 510 |
+
"eval_logits/chosen": -1.452673077583313,
|
| 511 |
+
"eval_logits/rejected": -1.3522714376449585,
|
| 512 |
+
"eval_logps/chosen": -418.378173828125,
|
| 513 |
+
"eval_logps/rejected": -467.829345703125,
|
| 514 |
+
"eval_loss": 0.5227062702178955,
|
| 515 |
+
"eval_rewards/accuracies": 0.7083333134651184,
|
| 516 |
+
"eval_rewards/chosen": -1.5981483459472656,
|
| 517 |
+
"eval_rewards/margins": 0.684920608997345,
|
| 518 |
+
"eval_rewards/rejected": -2.2830684185028076,
|
| 519 |
+
"eval_runtime": 219.6897,
|
| 520 |
+
"eval_samples_per_second": 9.104,
|
| 521 |
+
"eval_steps_per_second": 0.382,
|
| 522 |
+
"step": 300
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 0.48665620094191525,
|
| 526 |
+
"grad_norm": 6.346375465393066,
|
| 527 |
+
"learning_rate": 3.050573754371228e-06,
|
| 528 |
+
"logits/chosen": -1.4891157150268555,
|
| 529 |
+
"logits/rejected": -1.425148844718933,
|
| 530 |
+
"logps/chosen": -399.7137451171875,
|
| 531 |
+
"logps/rejected": -429.02960205078125,
|
| 532 |
+
"loss": 0.5355,
|
| 533 |
+
"rewards/accuracies": 0.71875,
|
| 534 |
+
"rewards/chosen": -1.2991411685943604,
|
| 535 |
+
"rewards/margins": 0.6303215026855469,
|
| 536 |
+
"rewards/rejected": -1.9294626712799072,
|
| 537 |
+
"step": 310
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.5023547880690737,
|
| 541 |
+
"grad_norm": 5.018222808837891,
|
| 542 |
+
"learning_rate": 2.916110954741667e-06,
|
| 543 |
+
"logits/chosen": -1.4688920974731445,
|
| 544 |
+
"logits/rejected": -1.295971155166626,
|
| 545 |
+
"logps/chosen": -429.50250244140625,
|
| 546 |
+
"logps/rejected": -448.509521484375,
|
| 547 |
+
"loss": 0.4945,
|
| 548 |
+
"rewards/accuracies": 0.731249988079071,
|
| 549 |
+
"rewards/chosen": -1.343754529953003,
|
| 550 |
+
"rewards/margins": 0.7055485844612122,
|
| 551 |
+
"rewards/rejected": -2.0493030548095703,
|
| 552 |
+
"step": 320
|
| 553 |
+
},
|
| 554 |
+
{
|
| 555 |
+
"epoch": 0.5180533751962323,
|
| 556 |
+
"grad_norm": 5.533540725708008,
|
| 557 |
+
"learning_rate": 2.7803976344929497e-06,
|
| 558 |
+
"logits/chosen": -1.3964585065841675,
|
| 559 |
+
"logits/rejected": -1.2397369146347046,
|
| 560 |
+
"logps/chosen": -450.6763610839844,
|
| 561 |
+
"logps/rejected": -513.90234375,
|
| 562 |
+
"loss": 0.5211,
|
| 563 |
+
"rewards/accuracies": 0.731249988079071,
|
| 564 |
+
"rewards/chosen": -1.7644832134246826,
|
| 565 |
+
"rewards/margins": 0.8988251686096191,
|
| 566 |
+
"rewards/rejected": -2.663308620452881,
|
| 567 |
+
"step": 330
|
| 568 |
+
},
|
| 569 |
+
{
|
| 570 |
+
"epoch": 0.533751962323391,
|
| 571 |
+
"grad_norm": 4.254801273345947,
|
| 572 |
+
"learning_rate": 2.6438416471154277e-06,
|
| 573 |
+
"logits/chosen": -1.420155644416809,
|
| 574 |
+
"logits/rejected": -1.355759859085083,
|
| 575 |
+
"logps/chosen": -393.2152099609375,
|
| 576 |
+
"logps/rejected": -455.6507263183594,
|
| 577 |
+
"loss": 0.5032,
|
| 578 |
+
"rewards/accuracies": 0.731249988079071,
|
| 579 |
+
"rewards/chosen": -1.4955414533615112,
|
| 580 |
+
"rewards/margins": 0.7609378099441528,
|
| 581 |
+
"rewards/rejected": -2.256479501724243,
|
| 582 |
+
"step": 340
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"epoch": 0.5494505494505495,
|
| 586 |
+
"grad_norm": 5.497036933898926,
|
| 587 |
+
"learning_rate": 2.5068533785312673e-06,
|
| 588 |
+
"logits/chosen": -1.2714860439300537,
|
| 589 |
+
"logits/rejected": -1.206830382347107,
|
| 590 |
+
"logps/chosen": -424.79864501953125,
|
| 591 |
+
"logps/rejected": -484.2491760253906,
|
| 592 |
+
"loss": 0.5265,
|
| 593 |
+
"rewards/accuracies": 0.731249988079071,
|
| 594 |
+
"rewards/chosen": -1.5535635948181152,
|
| 595 |
+
"rewards/margins": 0.7852663993835449,
|
| 596 |
+
"rewards/rejected": -2.33882999420166,
|
| 597 |
+
"step": 350
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 0.565149136577708,
|
| 601 |
+
"grad_norm": 4.5596208572387695,
|
| 602 |
+
"learning_rate": 2.369844513779026e-06,
|
| 603 |
+
"logits/chosen": -1.2018563747406006,
|
| 604 |
+
"logits/rejected": -1.115635871887207,
|
| 605 |
+
"logps/chosen": -440.96942138671875,
|
| 606 |
+
"logps/rejected": -488.6031799316406,
|
| 607 |
+
"loss": 0.5316,
|
| 608 |
+
"rewards/accuracies": 0.762499988079071,
|
| 609 |
+
"rewards/chosen": -1.7989181280136108,
|
| 610 |
+
"rewards/margins": 0.8467502593994141,
|
| 611 |
+
"rewards/rejected": -2.6456682682037354,
|
| 612 |
+
"step": 360
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 0.5808477237048666,
|
| 616 |
+
"grad_norm": 4.705105781555176,
|
| 617 |
+
"learning_rate": 2.2332267997940514e-06,
|
| 618 |
+
"logits/chosen": -1.3089182376861572,
|
| 619 |
+
"logits/rejected": -1.1716320514678955,
|
| 620 |
+
"logps/chosen": -400.7124938964844,
|
| 621 |
+
"logps/rejected": -441.3446350097656,
|
| 622 |
+
"loss": 0.5463,
|
| 623 |
+
"rewards/accuracies": 0.762499988079071,
|
| 624 |
+
"rewards/chosen": -1.4626152515411377,
|
| 625 |
+
"rewards/margins": 0.795612096786499,
|
| 626 |
+
"rewards/rejected": -2.2582273483276367,
|
| 627 |
+
"step": 370
|
| 628 |
+
},
|
| 629 |
+
{
|
| 630 |
+
"epoch": 0.5965463108320251,
|
| 631 |
+
"grad_norm": 4.903963565826416,
|
| 632 |
+
"learning_rate": 2.097410808002869e-06,
|
| 633 |
+
"logits/chosen": -1.3345696926116943,
|
| 634 |
+
"logits/rejected": -1.196610927581787,
|
| 635 |
+
"logps/chosen": -416.953125,
|
| 636 |
+
"logps/rejected": -442.70611572265625,
|
| 637 |
+
"loss": 0.5065,
|
| 638 |
+
"rewards/accuracies": 0.762499988079071,
|
| 639 |
+
"rewards/chosen": -1.265023946762085,
|
| 640 |
+
"rewards/margins": 0.7792559862136841,
|
| 641 |
+
"rewards/rejected": -2.0442800521850586,
|
| 642 |
+
"step": 380
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 0.6122448979591837,
|
| 646 |
+
"grad_norm": 5.188351631164551,
|
| 647 |
+
"learning_rate": 1.962804700450265e-06,
|
| 648 |
+
"logits/chosen": -1.2588812112808228,
|
| 649 |
+
"logits/rejected": -1.1723003387451172,
|
| 650 |
+
"logps/chosen": -375.08428955078125,
|
| 651 |
+
"logps/rejected": -457.165771484375,
|
| 652 |
+
"loss": 0.5228,
|
| 653 |
+
"rewards/accuracies": 0.8187500238418579,
|
| 654 |
+
"rewards/chosen": -1.3058245182037354,
|
| 655 |
+
"rewards/margins": 0.9940802454948425,
|
| 656 |
+
"rewards/rejected": -2.2999048233032227,
|
| 657 |
+
"step": 390
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 0.6279434850863422,
|
| 661 |
+
"grad_norm": 6.60071325302124,
|
| 662 |
+
"learning_rate": 1.8298130031671974e-06,
|
| 663 |
+
"logits/chosen": -1.3610254526138306,
|
| 664 |
+
"logits/rejected": -1.1776959896087646,
|
| 665 |
+
"logps/chosen": -445.6390686035156,
|
| 666 |
+
"logps/rejected": -512.9346923828125,
|
| 667 |
+
"loss": 0.4895,
|
| 668 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 669 |
+
"rewards/chosen": -1.6864149570465088,
|
| 670 |
+
"rewards/margins": 0.9683998823165894,
|
| 671 |
+
"rewards/rejected": -2.6548147201538086,
|
| 672 |
+
"step": 400
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"epoch": 0.6279434850863422,
|
| 676 |
+
"eval_logits/chosen": -1.2180150747299194,
|
| 677 |
+
"eval_logits/rejected": -1.1070232391357422,
|
| 678 |
+
"eval_logps/chosen": -444.1170959472656,
|
| 679 |
+
"eval_logps/rejected": -506.0622253417969,
|
| 680 |
+
"eval_loss": 0.514235258102417,
|
| 681 |
+
"eval_rewards/accuracies": 0.699404776096344,
|
| 682 |
+
"eval_rewards/chosen": -1.8555371761322021,
|
| 683 |
+
"eval_rewards/margins": 0.809860348701477,
|
| 684 |
+
"eval_rewards/rejected": -2.6653974056243896,
|
| 685 |
+
"eval_runtime": 220.0748,
|
| 686 |
+
"eval_samples_per_second": 9.088,
|
| 687 |
+
"eval_steps_per_second": 0.382,
|
| 688 |
+
"step": 400
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"epoch": 0.6436420722135008,
|
| 692 |
+
"grad_norm": 3.98221755027771,
|
| 693 |
+
"learning_rate": 1.6988353904658495e-06,
|
| 694 |
+
"logits/chosen": -1.2132154703140259,
|
| 695 |
+
"logits/rejected": -1.1685426235198975,
|
| 696 |
+
"logps/chosen": -435.8744201660156,
|
| 697 |
+
"logps/rejected": -512.9607543945312,
|
| 698 |
+
"loss": 0.5104,
|
| 699 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 700 |
+
"rewards/chosen": -1.587599515914917,
|
| 701 |
+
"rewards/margins": 1.0233564376831055,
|
| 702 |
+
"rewards/rejected": -2.6109557151794434,
|
| 703 |
+
"step": 410
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.6593406593406593,
|
| 707 |
+
"grad_norm": 6.315725803375244,
|
| 708 |
+
"learning_rate": 1.5702654838153641e-06,
|
| 709 |
+
"logits/chosen": -1.1728157997131348,
|
| 710 |
+
"logits/rejected": -1.1026521921157837,
|
| 711 |
+
"logps/chosen": -421.05963134765625,
|
| 712 |
+
"logps/rejected": -512.4940185546875,
|
| 713 |
+
"loss": 0.4971,
|
| 714 |
+
"rewards/accuracies": 0.731249988079071,
|
| 715 |
+
"rewards/chosen": -1.7201759815216064,
|
| 716 |
+
"rewards/margins": 0.8986526727676392,
|
| 717 |
+
"rewards/rejected": -2.618828535079956,
|
| 718 |
+
"step": 420
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"epoch": 0.6750392464678179,
|
| 722 |
+
"grad_norm": 5.234571933746338,
|
| 723 |
+
"learning_rate": 1.4444896689079142e-06,
|
| 724 |
+
"logits/chosen": -1.1644630432128906,
|
| 725 |
+
"logits/rejected": -1.0723572969436646,
|
| 726 |
+
"logps/chosen": -465.34661865234375,
|
| 727 |
+
"logps/rejected": -545.7144165039062,
|
| 728 |
+
"loss": 0.5029,
|
| 729 |
+
"rewards/accuracies": 0.768750011920929,
|
| 730 |
+
"rewards/chosen": -1.9419937133789062,
|
| 731 |
+
"rewards/margins": 1.0666325092315674,
|
| 732 |
+
"rewards/rejected": -3.0086259841918945,
|
| 733 |
+
"step": 430
|
| 734 |
+
},
|
| 735 |
+
{
|
| 736 |
+
"epoch": 0.6907378335949764,
|
| 737 |
+
"grad_norm": 5.252328395843506,
|
| 738 |
+
"learning_rate": 1.3218859344701634e-06,
|
| 739 |
+
"logits/chosen": -1.1691018342971802,
|
| 740 |
+
"logits/rejected": -1.120827317237854,
|
| 741 |
+
"logps/chosen": -424.86639404296875,
|
| 742 |
+
"logps/rejected": -522.9156494140625,
|
| 743 |
+
"loss": 0.4923,
|
| 744 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 745 |
+
"rewards/chosen": -1.887139081954956,
|
| 746 |
+
"rewards/margins": 1.0102875232696533,
|
| 747 |
+
"rewards/rejected": -2.8974266052246094,
|
| 748 |
+
"step": 440
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"epoch": 0.706436420722135,
|
| 752 |
+
"grad_norm": 7.3930206298828125,
|
| 753 |
+
"learning_rate": 1.2028227363097583e-06,
|
| 754 |
+
"logits/chosen": -1.1967122554779053,
|
| 755 |
+
"logits/rejected": -1.157003402709961,
|
| 756 |
+
"logps/chosen": -483.7662048339844,
|
| 757 |
+
"logps/rejected": -568.5826416015625,
|
| 758 |
+
"loss": 0.4961,
|
| 759 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 760 |
+
"rewards/chosen": -2.155179023742676,
|
| 761 |
+
"rewards/margins": 0.9878844022750854,
|
| 762 |
+
"rewards/rejected": -3.14306378364563,
|
| 763 |
+
"step": 450
|
| 764 |
+
},
|
| 765 |
+
{
|
| 766 |
+
"epoch": 0.7221350078492935,
|
| 767 |
+
"grad_norm": 3.9247488975524902,
|
| 768 |
+
"learning_rate": 1.0876578900107053e-06,
|
| 769 |
+
"logits/chosen": -1.2375166416168213,
|
| 770 |
+
"logits/rejected": -1.1339408159255981,
|
| 771 |
+
"logps/chosen": -465.45458984375,
|
| 772 |
+
"logps/rejected": -551.1287231445312,
|
| 773 |
+
"loss": 0.4842,
|
| 774 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 775 |
+
"rewards/chosen": -1.9650770425796509,
|
| 776 |
+
"rewards/margins": 1.0327082872390747,
|
| 777 |
+
"rewards/rejected": -2.9977853298187256,
|
| 778 |
+
"step": 460
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"epoch": 0.7378335949764521,
|
| 782 |
+
"grad_norm": 7.045323848724365,
|
| 783 |
+
"learning_rate": 9.767374956053584e-07,
|
| 784 |
+
"logits/chosen": -1.1633169651031494,
|
| 785 |
+
"logits/rejected": -1.0927058458328247,
|
| 786 |
+
"logps/chosen": -469.948486328125,
|
| 787 |
+
"logps/rejected": -547.9520874023438,
|
| 788 |
+
"loss": 0.4992,
|
| 789 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 790 |
+
"rewards/chosen": -2.16100811958313,
|
| 791 |
+
"rewards/margins": 0.9476286172866821,
|
| 792 |
+
"rewards/rejected": -3.1086366176605225,
|
| 793 |
+
"step": 470
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"epoch": 0.7535321821036107,
|
| 797 |
+
"grad_norm": 5.355714797973633,
|
| 798 |
+
"learning_rate": 8.703948974546592e-07,
|
| 799 |
+
"logits/chosen": -1.2294596433639526,
|
| 800 |
+
"logits/rejected": -1.1127874851226807,
|
| 801 |
+
"logps/chosen": -500.1397399902344,
|
| 802 |
+
"logps/rejected": -558.7158813476562,
|
| 803 |
+
"loss": 0.508,
|
| 804 |
+
"rewards/accuracies": 0.737500011920929,
|
| 805 |
+
"rewards/chosen": -2.36497163772583,
|
| 806 |
+
"rewards/margins": 0.9023281335830688,
|
| 807 |
+
"rewards/rejected": -3.2672996520996094,
|
| 808 |
+
"step": 480
|
| 809 |
+
},
|
| 810 |
+
{
|
| 811 |
+
"epoch": 0.7692307692307693,
|
| 812 |
+
"grad_norm": 5.423379421234131,
|
| 813 |
+
"learning_rate": 7.689496824624526e-07,
|
| 814 |
+
"logits/chosen": -1.206969976425171,
|
| 815 |
+
"logits/rejected": -1.0124098062515259,
|
| 816 |
+
"logps/chosen": -503.8736267089844,
|
| 817 |
+
"logps/rejected": -560.7037353515625,
|
| 818 |
+
"loss": 0.5118,
|
| 819 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 820 |
+
"rewards/chosen": -2.366530656814575,
|
| 821 |
+
"rewards/margins": 1.0277622938156128,
|
| 822 |
+
"rewards/rejected": -3.3942928314208984,
|
| 823 |
+
"step": 490
|
| 824 |
+
},
|
| 825 |
+
{
|
| 826 |
+
"epoch": 0.7849293563579278,
|
| 827 |
+
"grad_norm": 4.5977630615234375,
|
| 828 |
+
"learning_rate": 6.7270671963451e-07,
|
| 829 |
+
"logits/chosen": -1.24196195602417,
|
| 830 |
+
"logits/rejected": -1.166027307510376,
|
| 831 |
+
"logps/chosen": -503.1126403808594,
|
| 832 |
+
"logps/rejected": -574.1822509765625,
|
| 833 |
+
"loss": 0.4992,
|
| 834 |
+
"rewards/accuracies": 0.75,
|
| 835 |
+
"rewards/chosen": -2.4480373859405518,
|
| 836 |
+
"rewards/margins": 0.874676525592804,
|
| 837 |
+
"rewards/rejected": -3.322713851928711,
|
| 838 |
+
"step": 500
|
| 839 |
+
},
|
| 840 |
+
{
|
| 841 |
+
"epoch": 0.7849293563579278,
|
| 842 |
+
"eval_logits/chosen": -1.2588800191879272,
|
| 843 |
+
"eval_logits/rejected": -1.1520029306411743,
|
| 844 |
+
"eval_logps/chosen": -491.8629455566406,
|
| 845 |
+
"eval_logps/rejected": -549.8136596679688,
|
| 846 |
+
"eval_loss": 0.5018974542617798,
|
| 847 |
+
"eval_rewards/accuracies": 0.7053571343421936,
|
| 848 |
+
"eval_rewards/chosen": -2.332996129989624,
|
| 849 |
+
"eval_rewards/margins": 0.7699158191680908,
|
| 850 |
+
"eval_rewards/rejected": -3.102912187576294,
|
| 851 |
+
"eval_runtime": 219.9359,
|
| 852 |
+
"eval_samples_per_second": 9.094,
|
| 853 |
+
"eval_steps_per_second": 0.382,
|
| 854 |
+
"step": 500
|
| 855 |
+
},
|
| 856 |
+
{
|
| 857 |
+
"epoch": 0.8006279434850864,
|
| 858 |
+
"grad_norm": 4.936287879943848,
|
| 859 |
+
"learning_rate": 5.819552438686238e-07,
|
| 860 |
+
"logits/chosen": -1.2254358530044556,
|
| 861 |
+
"logits/rejected": -1.1503589153289795,
|
| 862 |
+
"logps/chosen": -488.61053466796875,
|
| 863 |
+
"logps/rejected": -558.0599975585938,
|
| 864 |
+
"loss": 0.4946,
|
| 865 |
+
"rewards/accuracies": 0.762499988079071,
|
| 866 |
+
"rewards/chosen": -2.1799943447113037,
|
| 867 |
+
"rewards/margins": 0.8831037282943726,
|
| 868 |
+
"rewards/rejected": -3.063098192214966,
|
| 869 |
+
"step": 510
|
| 870 |
+
},
|
| 871 |
+
{
|
| 872 |
+
"epoch": 0.8163265306122449,
|
| 873 |
+
"grad_norm": 4.393782615661621,
|
| 874 |
+
"learning_rate": 4.969679867292276e-07,
|
| 875 |
+
"logits/chosen": -1.4334200620651245,
|
| 876 |
+
"logits/rejected": -1.2100530862808228,
|
| 877 |
+
"logps/chosen": -525.8604736328125,
|
| 878 |
+
"logps/rejected": -564.2894897460938,
|
| 879 |
+
"loss": 0.5004,
|
| 880 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 881 |
+
"rewards/chosen": -2.246795177459717,
|
| 882 |
+
"rewards/margins": 0.9003856778144836,
|
| 883 |
+
"rewards/rejected": -3.1471807956695557,
|
| 884 |
+
"step": 520
|
| 885 |
+
},
|
| 886 |
+
{
|
| 887 |
+
"epoch": 0.8320251177394035,
|
| 888 |
+
"grad_norm": 4.474452495574951,
|
| 889 |
+
"learning_rate": 4.1800035681877765e-07,
|
| 890 |
+
"logits/chosen": -1.165138840675354,
|
| 891 |
+
"logits/rejected": -1.085632085800171,
|
| 892 |
+
"logps/chosen": -455.230712890625,
|
| 893 |
+
"logps/rejected": -550.22412109375,
|
| 894 |
+
"loss": 0.5097,
|
| 895 |
+
"rewards/accuracies": 0.737500011920929,
|
| 896 |
+
"rewards/chosen": -2.193859338760376,
|
| 897 |
+
"rewards/margins": 0.9264294505119324,
|
| 898 |
+
"rewards/rejected": -3.120288372039795,
|
| 899 |
+
"step": 530
|
| 900 |
+
},
|
| 901 |
+
{
|
| 902 |
+
"epoch": 0.847723704866562,
|
| 903 |
+
"grad_norm": 4.0765767097473145,
|
| 904 |
+
"learning_rate": 3.4528967220911287e-07,
|
| 905 |
+
"logits/chosen": -1.276033639907837,
|
| 906 |
+
"logits/rejected": -1.2467362880706787,
|
| 907 |
+
"logps/chosen": -488.040283203125,
|
| 908 |
+
"logps/rejected": -554.4703979492188,
|
| 909 |
+
"loss": 0.5123,
|
| 910 |
+
"rewards/accuracies": 0.706250011920929,
|
| 911 |
+
"rewards/chosen": -2.10200834274292,
|
| 912 |
+
"rewards/margins": 0.8350952863693237,
|
| 913 |
+
"rewards/rejected": -2.937103748321533,
|
| 914 |
+
"step": 540
|
| 915 |
+
},
|
| 916 |
+
{
|
| 917 |
+
"epoch": 0.8634222919937206,
|
| 918 |
+
"grad_norm": 4.756001949310303,
|
| 919 |
+
"learning_rate": 2.7905444723949765e-07,
|
| 920 |
+
"logits/chosen": -1.2069323062896729,
|
| 921 |
+
"logits/rejected": -1.192297101020813,
|
| 922 |
+
"logps/chosen": -484.1224060058594,
|
| 923 |
+
"logps/rejected": -570.1021118164062,
|
| 924 |
+
"loss": 0.5144,
|
| 925 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 926 |
+
"rewards/chosen": -2.127053737640381,
|
| 927 |
+
"rewards/margins": 0.9687838554382324,
|
| 928 |
+
"rewards/rejected": -3.095837354660034,
|
| 929 |
+
"step": 550
|
| 930 |
+
},
|
| 931 |
+
{
|
| 932 |
+
"epoch": 0.8791208791208791,
|
| 933 |
+
"grad_norm": 4.719667911529541,
|
| 934 |
+
"learning_rate": 2.1949373582475065e-07,
|
| 935 |
+
"logits/chosen": -1.2955992221832275,
|
| 936 |
+
"logits/rejected": -1.2129268646240234,
|
| 937 |
+
"logps/chosen": -453.64794921875,
|
| 938 |
+
"logps/rejected": -541.0159912109375,
|
| 939 |
+
"loss": 0.4979,
|
| 940 |
+
"rewards/accuracies": 0.793749988079071,
|
| 941 |
+
"rewards/chosen": -1.9799339771270752,
|
| 942 |
+
"rewards/margins": 1.0608704090118408,
|
| 943 |
+
"rewards/rejected": -3.040804386138916,
|
| 944 |
+
"step": 560
|
| 945 |
+
},
|
| 946 |
+
{
|
| 947 |
+
"epoch": 0.8948194662480377,
|
| 948 |
+
"grad_norm": 5.24905252456665,
|
| 949 |
+
"learning_rate": 1.667865332469379e-07,
|
| 950 |
+
"logits/chosen": -1.3544114828109741,
|
| 951 |
+
"logits/rejected": -1.2046071290969849,
|
| 952 |
+
"logps/chosen": -456.50482177734375,
|
| 953 |
+
"logps/rejected": -518.5260620117188,
|
| 954 |
+
"loss": 0.4768,
|
| 955 |
+
"rewards/accuracies": 0.84375,
|
| 956 |
+
"rewards/chosen": -1.8672730922698975,
|
| 957 |
+
"rewards/margins": 1.0792367458343506,
|
| 958 |
+
"rewards/rejected": -2.946509838104248,
|
| 959 |
+
"step": 570
|
| 960 |
+
},
|
| 961 |
+
{
|
| 962 |
+
"epoch": 0.9105180533751962,
|
| 963 |
+
"grad_norm": 5.04511833190918,
|
| 964 |
+
"learning_rate": 1.2109123822844653e-07,
|
| 965 |
+
"logits/chosen": -1.314582109451294,
|
| 966 |
+
"logits/rejected": -1.1942718029022217,
|
| 967 |
+
"logps/chosen": -491.97967529296875,
|
| 968 |
+
"logps/rejected": -532.6231689453125,
|
| 969 |
+
"loss": 0.4948,
|
| 970 |
+
"rewards/accuracies": 0.71875,
|
| 971 |
+
"rewards/chosen": -2.1609413623809814,
|
| 972 |
+
"rewards/margins": 0.8695161938667297,
|
| 973 |
+
"rewards/rejected": -3.0304577350616455,
|
| 974 |
+
"step": 580
|
| 975 |
+
},
|
| 976 |
+
{
|
| 977 |
+
"epoch": 0.9262166405023547,
|
| 978 |
+
"grad_norm": 4.632051467895508,
|
| 979 |
+
"learning_rate": 8.254517690300946e-08,
|
| 980 |
+
"logits/chosen": -1.3019325733184814,
|
| 981 |
+
"logits/rejected": -1.1224420070648193,
|
| 982 |
+
"logps/chosen": -463.02874755859375,
|
| 983 |
+
"logps/rejected": -560.708251953125,
|
| 984 |
+
"loss": 0.4826,
|
| 985 |
+
"rewards/accuracies": 0.737500011920929,
|
| 986 |
+
"rewards/chosen": -2.165086507797241,
|
| 987 |
+
"rewards/margins": 1.022458791732788,
|
| 988 |
+
"rewards/rejected": -3.1875455379486084,
|
| 989 |
+
"step": 590
|
| 990 |
+
},
|
| 991 |
+
{
|
| 992 |
+
"epoch": 0.9419152276295133,
|
| 993 |
+
"grad_norm": 5.119118690490723,
|
| 994 |
+
"learning_rate": 5.126419011529993e-08,
|
| 995 |
+
"logits/chosen": -1.251976490020752,
|
| 996 |
+
"logits/rejected": -1.1646479368209839,
|
| 997 |
+
"logps/chosen": -498.5111389160156,
|
| 998 |
+
"logps/rejected": -600.1183471679688,
|
| 999 |
+
"loss": 0.5001,
|
| 1000 |
+
"rewards/accuracies": 0.737500011920929,
|
| 1001 |
+
"rewards/chosen": -2.133627414703369,
|
| 1002 |
+
"rewards/margins": 1.0616191625595093,
|
| 1003 |
+
"rewards/rejected": -3.195246696472168,
|
| 1004 |
+
"step": 600
|
| 1005 |
+
},
|
| 1006 |
+
{
|
| 1007 |
+
"epoch": 0.9419152276295133,
|
| 1008 |
+
"eval_logits/chosen": -1.2825318574905396,
|
| 1009 |
+
"eval_logits/rejected": -1.1767752170562744,
|
| 1010 |
+
"eval_logps/chosen": -475.6836853027344,
|
| 1011 |
+
"eval_logps/rejected": -535.6056518554688,
|
| 1012 |
+
"eval_loss": 0.5020818114280701,
|
| 1013 |
+
"eval_rewards/accuracies": 0.7083333134651184,
|
| 1014 |
+
"eval_rewards/chosen": -2.171203136444092,
|
| 1015 |
+
"eval_rewards/margins": 0.7896292805671692,
|
| 1016 |
+
"eval_rewards/rejected": -2.960832118988037,
|
| 1017 |
+
"eval_runtime": 219.3018,
|
| 1018 |
+
"eval_samples_per_second": 9.12,
|
| 1019 |
+
"eval_steps_per_second": 0.383,
|
| 1020 |
+
"step": 600
|
| 1021 |
+
},
|
| 1022 |
+
{
|
| 1023 |
+
"epoch": 0.957613814756672,
|
| 1024 |
+
"grad_norm": 5.81568717956543,
|
| 1025 |
+
"learning_rate": 2.734228528934679e-08,
|
| 1026 |
+
"logits/chosen": -1.3186975717544556,
|
| 1027 |
+
"logits/rejected": -1.1690412759780884,
|
| 1028 |
+
"logps/chosen": -448.78814697265625,
|
| 1029 |
+
"logps/rejected": -523.2693481445312,
|
| 1030 |
+
"loss": 0.5127,
|
| 1031 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1032 |
+
"rewards/chosen": -2.0488333702087402,
|
| 1033 |
+
"rewards/margins": 1.0742355585098267,
|
| 1034 |
+
"rewards/rejected": -3.1230690479278564,
|
| 1035 |
+
"step": 610
|
| 1036 |
+
},
|
| 1037 |
+
{
|
| 1038 |
+
"epoch": 0.9733124018838305,
|
| 1039 |
+
"grad_norm": 5.066900253295898,
|
| 1040 |
+
"learning_rate": 1.0851353912008644e-08,
|
| 1041 |
+
"logits/chosen": -1.1879260540008545,
|
| 1042 |
+
"logits/rejected": -1.057185411453247,
|
| 1043 |
+
"logps/chosen": -493.63323974609375,
|
| 1044 |
+
"logps/rejected": -551.7728271484375,
|
| 1045 |
+
"loss": 0.4958,
|
| 1046 |
+
"rewards/accuracies": 0.706250011920929,
|
| 1047 |
+
"rewards/chosen": -2.1862809658050537,
|
| 1048 |
+
"rewards/margins": 0.9377900958061218,
|
| 1049 |
+
"rewards/rejected": -3.1240711212158203,
|
| 1050 |
+
"step": 620
|
| 1051 |
+
},
|
| 1052 |
+
{
|
| 1053 |
+
"epoch": 0.989010989010989,
|
| 1054 |
+
"grad_norm": 5.623207092285156,
|
| 1055 |
+
"learning_rate": 1.8409554805329243e-09,
|
| 1056 |
+
"logits/chosen": -1.2257732152938843,
|
| 1057 |
+
"logits/rejected": -1.1151111125946045,
|
| 1058 |
+
"logps/chosen": -482.53857421875,
|
| 1059 |
+
"logps/rejected": -528.3353271484375,
|
| 1060 |
+
"loss": 0.4897,
|
| 1061 |
+
"rewards/accuracies": 0.75,
|
| 1062 |
+
"rewards/chosen": -2.206721544265747,
|
| 1063 |
+
"rewards/margins": 0.8017439842224121,
|
| 1064 |
+
"rewards/rejected": -3.008465528488159,
|
| 1065 |
+
"step": 630
|
| 1066 |
+
},
|
| 1067 |
+
{
|
| 1068 |
+
"epoch": 1.0,
|
| 1069 |
+
"step": 637,
|
| 1070 |
+
"total_flos": 0.0,
|
| 1071 |
+
"train_loss": 0.5397420258873861,
|
| 1072 |
+
"train_runtime": 14909.9591,
|
| 1073 |
+
"train_samples_per_second": 4.1,
|
| 1074 |
+
"train_steps_per_second": 0.043
|
| 1075 |
+
}
|
| 1076 |
+
],
|
| 1077 |
+
"logging_steps": 10,
|
| 1078 |
+
"max_steps": 637,
|
| 1079 |
+
"num_input_tokens_seen": 0,
|
| 1080 |
+
"num_train_epochs": 1,
|
| 1081 |
+
"save_steps": 100,
|
| 1082 |
+
"stateful_callbacks": {
|
| 1083 |
+
"TrainerControl": {
|
| 1084 |
+
"args": {
|
| 1085 |
+
"should_epoch_stop": false,
|
| 1086 |
+
"should_evaluate": false,
|
| 1087 |
+
"should_log": false,
|
| 1088 |
+
"should_save": true,
|
| 1089 |
+
"should_training_stop": true
|
| 1090 |
+
},
|
| 1091 |
+
"attributes": {}
|
| 1092 |
+
}
|
| 1093 |
+
},
|
| 1094 |
+
"total_flos": 0.0,
|
| 1095 |
+
"train_batch_size": 4,
|
| 1096 |
+
"trial_name": null,
|
| 1097 |
+
"trial_params": null
|
| 1098 |
+
}
|