Model save
Browse files- README.md +77 -0
- all_results.json +9 -0
- generation_config.json +6 -0
- train_results.json +9 -0
- trainer_state.json +826 -0
README.md
ADDED
|
@@ -0,0 +1,77 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
license: apache-2.0
|
| 4 |
+
base_model: alignment-handbook/zephyr-7b-sft-full
|
| 5 |
+
tags:
|
| 6 |
+
- trl
|
| 7 |
+
- dpo
|
| 8 |
+
- generated_from_trainer
|
| 9 |
+
model-index:
|
| 10 |
+
- name: zephyr-7b-dpo-full
|
| 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-full
|
| 18 |
+
|
| 19 |
+
This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on an unknown dataset.
|
| 20 |
+
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 0.5006
|
| 22 |
+
- Rewards/chosen: -1.0395
|
| 23 |
+
- Rewards/rejected: -2.0108
|
| 24 |
+
- Rewards/accuracies: 0.7852
|
| 25 |
+
- Rewards/margins: 0.9713
|
| 26 |
+
- Logps/rejected: -463.7658
|
| 27 |
+
- Logps/chosen: -366.5239
|
| 28 |
+
- Logits/rejected: 0.6358
|
| 29 |
+
- Logits/chosen: -0.3905
|
| 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-07
|
| 49 |
+
- train_batch_size: 8
|
| 50 |
+
- eval_batch_size: 8
|
| 51 |
+
- seed: 42
|
| 52 |
+
- distributed_type: multi-GPU
|
| 53 |
+
- num_devices: 8
|
| 54 |
+
- gradient_accumulation_steps: 2
|
| 55 |
+
- total_train_batch_size: 128
|
| 56 |
+
- total_eval_batch_size: 64
|
| 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 |
+
| 0.5644 | 0.2092 | 100 | 0.5693 | -0.4760 | -0.9833 | 0.75 | 0.5073 | -361.0131 | -310.1760 | -1.6463 | -1.8042 |
|
| 67 |
+
| 0.5482 | 0.4184 | 200 | 0.5285 | -0.5789 | -1.3324 | 0.7812 | 0.7535 | -395.9196 | -320.4612 | -1.1108 | -1.6512 |
|
| 68 |
+
| 0.4952 | 0.6276 | 300 | 0.5067 | -1.0198 | -1.9482 | 0.7734 | 0.9284 | -457.5016 | -364.5515 | 0.5574 | -0.3940 |
|
| 69 |
+
| 0.5037 | 0.8368 | 400 | 0.5006 | -1.0395 | -2.0108 | 0.7852 | 0.9713 | -463.7658 | -366.5239 | 0.6358 | -0.3905 |
|
| 70 |
+
|
| 71 |
+
|
| 72 |
+
### Framework versions
|
| 73 |
+
|
| 74 |
+
- Transformers 4.45.1
|
| 75 |
+
- Pytorch 2.4.0a0+f70bd71a48.nv24.06
|
| 76 |
+
- Datasets 2.18.0
|
| 77 |
+
- Tokenizers 0.20.0
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5400827195355085,
|
| 5 |
+
"train_runtime": 3932.0785,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 15.548,
|
| 8 |
+
"train_steps_per_second": 0.122
|
| 9 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 1,
|
| 4 |
+
"eos_token_id": 2,
|
| 5 |
+
"transformers_version": "4.45.1"
|
| 6 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5400827195355085,
|
| 5 |
+
"train_runtime": 3932.0785,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 15.548,
|
| 8 |
+
"train_steps_per_second": 0.122
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,826 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.0,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 478,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.0020920502092050207,
|
| 13 |
+
"grad_norm": 9.087554245569063,
|
| 14 |
+
"learning_rate": 1.0416666666666666e-08,
|
| 15 |
+
"logits/chosen": -2.7662220001220703,
|
| 16 |
+
"logits/rejected": -2.7178127765655518,
|
| 17 |
+
"logps/chosen": -269.6776123046875,
|
| 18 |
+
"logps/rejected": -360.6510314941406,
|
| 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": 0.02092050209205021,
|
| 28 |
+
"grad_norm": 9.138148617727442,
|
| 29 |
+
"learning_rate": 1.0416666666666667e-07,
|
| 30 |
+
"logits/chosen": -2.5923047065734863,
|
| 31 |
+
"logits/rejected": -2.562880277633667,
|
| 32 |
+
"logps/chosen": -264.8013916015625,
|
| 33 |
+
"logps/rejected": -251.54592895507812,
|
| 34 |
+
"loss": 0.6931,
|
| 35 |
+
"rewards/accuracies": 0.4375,
|
| 36 |
+
"rewards/chosen": -0.0002887251030188054,
|
| 37 |
+
"rewards/margins": -8.625661575933918e-05,
|
| 38 |
+
"rewards/rejected": -0.00020246852363925427,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.04184100418410042,
|
| 43 |
+
"grad_norm": 8.411516829739382,
|
| 44 |
+
"learning_rate": 2.0833333333333333e-07,
|
| 45 |
+
"logits/chosen": -2.653226613998413,
|
| 46 |
+
"logits/rejected": -2.605504035949707,
|
| 47 |
+
"logps/chosen": -281.4090576171875,
|
| 48 |
+
"logps/rejected": -296.7936096191406,
|
| 49 |
+
"loss": 0.6923,
|
| 50 |
+
"rewards/accuracies": 0.543749988079071,
|
| 51 |
+
"rewards/chosen": 0.0007672592182643712,
|
| 52 |
+
"rewards/margins": 0.0005850494490005076,
|
| 53 |
+
"rewards/rejected": 0.0001822096382966265,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.06276150627615062,
|
| 58 |
+
"grad_norm": 8.341151841493089,
|
| 59 |
+
"learning_rate": 3.1249999999999997e-07,
|
| 60 |
+
"logits/chosen": -2.664233684539795,
|
| 61 |
+
"logits/rejected": -2.5916261672973633,
|
| 62 |
+
"logps/chosen": -299.38519287109375,
|
| 63 |
+
"logps/rejected": -263.70245361328125,
|
| 64 |
+
"loss": 0.6885,
|
| 65 |
+
"rewards/accuracies": 0.668749988079071,
|
| 66 |
+
"rewards/chosen": 0.012760737910866737,
|
| 67 |
+
"rewards/margins": 0.010479705408215523,
|
| 68 |
+
"rewards/rejected": 0.0022810332011431456,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.08368200836820083,
|
| 73 |
+
"grad_norm": 8.161334179415334,
|
| 74 |
+
"learning_rate": 4.1666666666666667e-07,
|
| 75 |
+
"logits/chosen": -2.557464838027954,
|
| 76 |
+
"logits/rejected": -2.522446393966675,
|
| 77 |
+
"logps/chosen": -263.1476745605469,
|
| 78 |
+
"logps/rejected": -245.10128784179688,
|
| 79 |
+
"loss": 0.6776,
|
| 80 |
+
"rewards/accuracies": 0.706250011920929,
|
| 81 |
+
"rewards/chosen": 0.03926760330796242,
|
| 82 |
+
"rewards/margins": 0.04648735374212265,
|
| 83 |
+
"rewards/rejected": -0.007219746708869934,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.10460251046025104,
|
| 88 |
+
"grad_norm": 8.412121997546578,
|
| 89 |
+
"learning_rate": 4.999733114418725e-07,
|
| 90 |
+
"logits/chosen": -2.4982900619506836,
|
| 91 |
+
"logits/rejected": -2.458754301071167,
|
| 92 |
+
"logps/chosen": -279.72418212890625,
|
| 93 |
+
"logps/rejected": -297.51483154296875,
|
| 94 |
+
"loss": 0.6615,
|
| 95 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 96 |
+
"rewards/chosen": 0.0003564223588909954,
|
| 97 |
+
"rewards/margins": 0.0477915033698082,
|
| 98 |
+
"rewards/rejected": -0.04743507504463196,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.12552301255230125,
|
| 103 |
+
"grad_norm": 11.28651377706942,
|
| 104 |
+
"learning_rate": 4.990398100856366e-07,
|
| 105 |
+
"logits/chosen": -2.512681245803833,
|
| 106 |
+
"logits/rejected": -2.4412343502044678,
|
| 107 |
+
"logps/chosen": -273.84716796875,
|
| 108 |
+
"logps/rejected": -313.0481872558594,
|
| 109 |
+
"loss": 0.639,
|
| 110 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 111 |
+
"rewards/chosen": -0.02180738002061844,
|
| 112 |
+
"rewards/margins": 0.1268891841173172,
|
| 113 |
+
"rewards/rejected": -0.14869657158851624,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.14644351464435146,
|
| 118 |
+
"grad_norm": 10.758366048663921,
|
| 119 |
+
"learning_rate": 4.967775735898179e-07,
|
| 120 |
+
"logits/chosen": -2.3633859157562256,
|
| 121 |
+
"logits/rejected": -2.360542058944702,
|
| 122 |
+
"logps/chosen": -271.2507019042969,
|
| 123 |
+
"logps/rejected": -283.77508544921875,
|
| 124 |
+
"loss": 0.6157,
|
| 125 |
+
"rewards/accuracies": 0.762499988079071,
|
| 126 |
+
"rewards/chosen": -0.05616923049092293,
|
| 127 |
+
"rewards/margins": 0.21398130059242249,
|
| 128 |
+
"rewards/rejected": -0.2701505124568939,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.16736401673640167,
|
| 133 |
+
"grad_norm": 16.20901309109112,
|
| 134 |
+
"learning_rate": 4.931986719649298e-07,
|
| 135 |
+
"logits/chosen": -2.5523719787597656,
|
| 136 |
+
"logits/rejected": -2.4694900512695312,
|
| 137 |
+
"logps/chosen": -335.7901306152344,
|
| 138 |
+
"logps/rejected": -304.34521484375,
|
| 139 |
+
"loss": 0.6023,
|
| 140 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 141 |
+
"rewards/chosen": -0.12984994053840637,
|
| 142 |
+
"rewards/margins": 0.2733689248561859,
|
| 143 |
+
"rewards/rejected": -0.4032188355922699,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.18828451882845187,
|
| 148 |
+
"grad_norm": 20.628675413942332,
|
| 149 |
+
"learning_rate": 4.883222001996351e-07,
|
| 150 |
+
"logits/chosen": -2.1346848011016846,
|
| 151 |
+
"logits/rejected": -2.0720489025115967,
|
| 152 |
+
"logps/chosen": -313.75103759765625,
|
| 153 |
+
"logps/rejected": -355.2326965332031,
|
| 154 |
+
"loss": 0.5741,
|
| 155 |
+
"rewards/accuracies": 0.731249988079071,
|
| 156 |
+
"rewards/chosen": -0.546220064163208,
|
| 157 |
+
"rewards/margins": 0.4818921983242035,
|
| 158 |
+
"rewards/rejected": -1.0281121730804443,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.20920502092050208,
|
| 163 |
+
"grad_norm": 15.512372945392126,
|
| 164 |
+
"learning_rate": 4.821741763807186e-07,
|
| 165 |
+
"logits/chosen": -1.6733529567718506,
|
| 166 |
+
"logits/rejected": -1.5271198749542236,
|
| 167 |
+
"logps/chosen": -313.7639465332031,
|
| 168 |
+
"logps/rejected": -366.28643798828125,
|
| 169 |
+
"loss": 0.5644,
|
| 170 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 171 |
+
"rewards/chosen": -0.5339412689208984,
|
| 172 |
+
"rewards/margins": 0.5581195950508118,
|
| 173 |
+
"rewards/rejected": -1.0920608043670654,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 0.20920502092050208,
|
| 178 |
+
"eval_logits/chosen": -1.8042092323303223,
|
| 179 |
+
"eval_logits/rejected": -1.6462740898132324,
|
| 180 |
+
"eval_logps/chosen": -310.176025390625,
|
| 181 |
+
"eval_logps/rejected": -361.0130615234375,
|
| 182 |
+
"eval_loss": 0.5693150758743286,
|
| 183 |
+
"eval_rewards/accuracies": 0.75,
|
| 184 |
+
"eval_rewards/chosen": -0.4760096073150635,
|
| 185 |
+
"eval_rewards/margins": 0.5072795152664185,
|
| 186 |
+
"eval_rewards/rejected": -0.9832891225814819,
|
| 187 |
+
"eval_runtime": 39.9983,
|
| 188 |
+
"eval_samples_per_second": 50.002,
|
| 189 |
+
"eval_steps_per_second": 0.8,
|
| 190 |
+
"step": 100
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 0.2301255230125523,
|
| 194 |
+
"grad_norm": 20.060254091732027,
|
| 195 |
+
"learning_rate": 4.747874028753375e-07,
|
| 196 |
+
"logits/chosen": -1.6645755767822266,
|
| 197 |
+
"logits/rejected": -1.2451729774475098,
|
| 198 |
+
"logps/chosen": -368.3230285644531,
|
| 199 |
+
"logps/rejected": -373.16448974609375,
|
| 200 |
+
"loss": 0.5666,
|
| 201 |
+
"rewards/accuracies": 0.706250011920929,
|
| 202 |
+
"rewards/chosen": -0.6374339461326599,
|
| 203 |
+
"rewards/margins": 0.5157445669174194,
|
| 204 |
+
"rewards/rejected": -1.1531786918640137,
|
| 205 |
+
"step": 110
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.2510460251046025,
|
| 209 |
+
"grad_norm": 20.825225854686142,
|
| 210 |
+
"learning_rate": 4.662012913161997e-07,
|
| 211 |
+
"logits/chosen": -1.753260612487793,
|
| 212 |
+
"logits/rejected": -1.4734609127044678,
|
| 213 |
+
"logps/chosen": -327.0220947265625,
|
| 214 |
+
"logps/rejected": -357.7860107421875,
|
| 215 |
+
"loss": 0.5607,
|
| 216 |
+
"rewards/accuracies": 0.731249988079071,
|
| 217 |
+
"rewards/chosen": -0.4582589268684387,
|
| 218 |
+
"rewards/margins": 0.5211780071258545,
|
| 219 |
+
"rewards/rejected": -0.9794368743896484,
|
| 220 |
+
"step": 120
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.2719665271966527,
|
| 224 |
+
"grad_norm": 23.496638764764725,
|
| 225 |
+
"learning_rate": 4.5646165232345103e-07,
|
| 226 |
+
"logits/chosen": -0.5949552655220032,
|
| 227 |
+
"logits/rejected": -0.22445210814476013,
|
| 228 |
+
"logps/chosen": -385.47186279296875,
|
| 229 |
+
"logps/rejected": -434.91864013671875,
|
| 230 |
+
"loss": 0.54,
|
| 231 |
+
"rewards/accuracies": 0.75,
|
| 232 |
+
"rewards/chosen": -1.0354242324829102,
|
| 233 |
+
"rewards/margins": 0.6292093992233276,
|
| 234 |
+
"rewards/rejected": -1.6646335124969482,
|
| 235 |
+
"step": 130
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.2928870292887029,
|
| 239 |
+
"grad_norm": 19.8432073115127,
|
| 240 |
+
"learning_rate": 4.456204510851956e-07,
|
| 241 |
+
"logits/chosen": -0.9896577596664429,
|
| 242 |
+
"logits/rejected": -0.6126431226730347,
|
| 243 |
+
"logps/chosen": -403.9255676269531,
|
| 244 |
+
"logps/rejected": -456.3795471191406,
|
| 245 |
+
"loss": 0.5379,
|
| 246 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 247 |
+
"rewards/chosen": -0.8796173334121704,
|
| 248 |
+
"rewards/margins": 0.6777242422103882,
|
| 249 |
+
"rewards/rejected": -1.557341456413269,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.3138075313807531,
|
| 254 |
+
"grad_norm": 24.252740229867097,
|
| 255 |
+
"learning_rate": 4.337355301007335e-07,
|
| 256 |
+
"logits/chosen": -1.2636734247207642,
|
| 257 |
+
"logits/rejected": -0.7799659967422485,
|
| 258 |
+
"logps/chosen": -352.56219482421875,
|
| 259 |
+
"logps/rejected": -398.70361328125,
|
| 260 |
+
"loss": 0.5431,
|
| 261 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 262 |
+
"rewards/chosen": -0.6931993365287781,
|
| 263 |
+
"rewards/margins": 0.5318464040756226,
|
| 264 |
+
"rewards/rejected": -1.2250458002090454,
|
| 265 |
+
"step": 150
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.33472803347280333,
|
| 269 |
+
"grad_norm": 23.380475895734964,
|
| 270 |
+
"learning_rate": 4.2087030056579986e-07,
|
| 271 |
+
"logits/chosen": -0.7647526860237122,
|
| 272 |
+
"logits/rejected": -0.008624720387160778,
|
| 273 |
+
"logps/chosen": -355.8943786621094,
|
| 274 |
+
"logps/rejected": -413.7810974121094,
|
| 275 |
+
"loss": 0.5422,
|
| 276 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 277 |
+
"rewards/chosen": -0.8285518884658813,
|
| 278 |
+
"rewards/margins": 0.7353727221488953,
|
| 279 |
+
"rewards/rejected": -1.5639246702194214,
|
| 280 |
+
"step": 160
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.35564853556485354,
|
| 284 |
+
"grad_norm": 25.431140167493865,
|
| 285 |
+
"learning_rate": 4.070934040463998e-07,
|
| 286 |
+
"logits/chosen": -0.6836012601852417,
|
| 287 |
+
"logits/rejected": -0.216752290725708,
|
| 288 |
+
"logps/chosen": -330.0359191894531,
|
| 289 |
+
"logps/rejected": -370.3531799316406,
|
| 290 |
+
"loss": 0.5398,
|
| 291 |
+
"rewards/accuracies": 0.71875,
|
| 292 |
+
"rewards/chosen": -0.8072524070739746,
|
| 293 |
+
"rewards/margins": 0.5846077799797058,
|
| 294 |
+
"rewards/rejected": -1.3918602466583252,
|
| 295 |
+
"step": 170
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 0.37656903765690375,
|
| 299 |
+
"grad_norm": 20.150510606814603,
|
| 300 |
+
"learning_rate": 3.9247834624635404e-07,
|
| 301 |
+
"logits/chosen": -0.19145147502422333,
|
| 302 |
+
"logits/rejected": 0.5919758081436157,
|
| 303 |
+
"logps/chosen": -332.99493408203125,
|
| 304 |
+
"logps/rejected": -376.44049072265625,
|
| 305 |
+
"loss": 0.5038,
|
| 306 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 307 |
+
"rewards/chosen": -0.7668272256851196,
|
| 308 |
+
"rewards/margins": 0.7220631837844849,
|
| 309 |
+
"rewards/rejected": -1.4888904094696045,
|
| 310 |
+
"step": 180
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.39748953974895396,
|
| 314 |
+
"grad_norm": 26.333247990716465,
|
| 315 |
+
"learning_rate": 3.7710310482256523e-07,
|
| 316 |
+
"logits/chosen": -0.44246259331703186,
|
| 317 |
+
"logits/rejected": 0.3344479501247406,
|
| 318 |
+
"logps/chosen": -364.98065185546875,
|
| 319 |
+
"logps/rejected": -421.58978271484375,
|
| 320 |
+
"loss": 0.5202,
|
| 321 |
+
"rewards/accuracies": 0.71875,
|
| 322 |
+
"rewards/chosen": -0.9371756315231323,
|
| 323 |
+
"rewards/margins": 0.6810097098350525,
|
| 324 |
+
"rewards/rejected": -1.6181854009628296,
|
| 325 |
+
"step": 190
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 0.41841004184100417,
|
| 329 |
+
"grad_norm": 61.74429639749924,
|
| 330 |
+
"learning_rate": 3.610497133404795e-07,
|
| 331 |
+
"logits/chosen": -0.9733116030693054,
|
| 332 |
+
"logits/rejected": -0.424935907125473,
|
| 333 |
+
"logps/chosen": -343.17181396484375,
|
| 334 |
+
"logps/rejected": -409.66015625,
|
| 335 |
+
"loss": 0.5482,
|
| 336 |
+
"rewards/accuracies": 0.731249988079071,
|
| 337 |
+
"rewards/chosen": -0.8325435519218445,
|
| 338 |
+
"rewards/margins": 0.7432821393013,
|
| 339 |
+
"rewards/rejected": -1.5758254528045654,
|
| 340 |
+
"step": 200
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 0.41841004184100417,
|
| 344 |
+
"eval_logits/chosen": -1.6512479782104492,
|
| 345 |
+
"eval_logits/rejected": -1.1108059883117676,
|
| 346 |
+
"eval_logps/chosen": -320.461181640625,
|
| 347 |
+
"eval_logps/rejected": -395.9195556640625,
|
| 348 |
+
"eval_loss": 0.5285090208053589,
|
| 349 |
+
"eval_rewards/accuracies": 0.78125,
|
| 350 |
+
"eval_rewards/chosen": -0.5788613557815552,
|
| 351 |
+
"eval_rewards/margins": 0.7534924745559692,
|
| 352 |
+
"eval_rewards/rejected": -1.3323538303375244,
|
| 353 |
+
"eval_runtime": 39.8686,
|
| 354 |
+
"eval_samples_per_second": 50.165,
|
| 355 |
+
"eval_steps_per_second": 0.803,
|
| 356 |
+
"step": 200
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 0.4393305439330544,
|
| 360 |
+
"grad_norm": 23.039933494228325,
|
| 361 |
+
"learning_rate": 3.4440382358952115e-07,
|
| 362 |
+
"logits/chosen": -1.410441279411316,
|
| 363 |
+
"logits/rejected": -0.5446120500564575,
|
| 364 |
+
"logps/chosen": -362.73443603515625,
|
| 365 |
+
"logps/rejected": -392.86572265625,
|
| 366 |
+
"loss": 0.5374,
|
| 367 |
+
"rewards/accuracies": 0.768750011920929,
|
| 368 |
+
"rewards/chosen": -0.7306278944015503,
|
| 369 |
+
"rewards/margins": 0.6853674650192261,
|
| 370 |
+
"rewards/rejected": -1.4159953594207764,
|
| 371 |
+
"step": 210
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 0.4602510460251046,
|
| 375 |
+
"grad_norm": 28.84599390547246,
|
| 376 |
+
"learning_rate": 3.272542485937368e-07,
|
| 377 |
+
"logits/chosen": 0.19500534236431122,
|
| 378 |
+
"logits/rejected": 1.4215036630630493,
|
| 379 |
+
"logps/chosen": -366.9354248046875,
|
| 380 |
+
"logps/rejected": -427.8094787597656,
|
| 381 |
+
"loss": 0.521,
|
| 382 |
+
"rewards/accuracies": 0.731249988079071,
|
| 383 |
+
"rewards/chosen": -1.015799641609192,
|
| 384 |
+
"rewards/margins": 0.8056750297546387,
|
| 385 |
+
"rewards/rejected": -1.8214746713638306,
|
| 386 |
+
"step": 220
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 0.4811715481171548,
|
| 390 |
+
"grad_norm": 23.1895401529341,
|
| 391 |
+
"learning_rate": 3.096924887558854e-07,
|
| 392 |
+
"logits/chosen": 0.23373627662658691,
|
| 393 |
+
"logits/rejected": 1.2636606693267822,
|
| 394 |
+
"logps/chosen": -344.4990234375,
|
| 395 |
+
"logps/rejected": -433.19549560546875,
|
| 396 |
+
"loss": 0.5423,
|
| 397 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 398 |
+
"rewards/chosen": -0.9555310010910034,
|
| 399 |
+
"rewards/margins": 0.8986861109733582,
|
| 400 |
+
"rewards/rejected": -1.8542171716690063,
|
| 401 |
+
"step": 230
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 0.502092050209205,
|
| 405 |
+
"grad_norm": 25.383792279845476,
|
| 406 |
+
"learning_rate": 2.9181224366319943e-07,
|
| 407 |
+
"logits/chosen": -0.8247690200805664,
|
| 408 |
+
"logits/rejected": -0.12202179431915283,
|
| 409 |
+
"logps/chosen": -338.62860107421875,
|
| 410 |
+
"logps/rejected": -393.34130859375,
|
| 411 |
+
"loss": 0.5049,
|
| 412 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 413 |
+
"rewards/chosen": -0.7489091753959656,
|
| 414 |
+
"rewards/margins": 0.6940961480140686,
|
| 415 |
+
"rewards/rejected": -1.4430053234100342,
|
| 416 |
+
"step": 240
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 0.5230125523012552,
|
| 420 |
+
"grad_norm": 23.764621178377002,
|
| 421 |
+
"learning_rate": 2.7370891215954565e-07,
|
| 422 |
+
"logits/chosen": -0.5904697179794312,
|
| 423 |
+
"logits/rejected": 0.471386581659317,
|
| 424 |
+
"logps/chosen": -372.9373474121094,
|
| 425 |
+
"logps/rejected": -412.6297302246094,
|
| 426 |
+
"loss": 0.5051,
|
| 427 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 428 |
+
"rewards/chosen": -0.7474657297134399,
|
| 429 |
+
"rewards/margins": 0.7835052013397217,
|
| 430 |
+
"rewards/rejected": -1.5309709310531616,
|
| 431 |
+
"step": 250
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 0.5439330543933054,
|
| 435 |
+
"grad_norm": 31.749178242054935,
|
| 436 |
+
"learning_rate": 2.55479083351317e-07,
|
| 437 |
+
"logits/chosen": -0.029442792758345604,
|
| 438 |
+
"logits/rejected": 1.2092030048370361,
|
| 439 |
+
"logps/chosen": -399.5458068847656,
|
| 440 |
+
"logps/rejected": -442.3656311035156,
|
| 441 |
+
"loss": 0.5089,
|
| 442 |
+
"rewards/accuracies": 0.75,
|
| 443 |
+
"rewards/chosen": -1.0523649454116821,
|
| 444 |
+
"rewards/margins": 0.8320234417915344,
|
| 445 |
+
"rewards/rejected": -1.8843883275985718,
|
| 446 |
+
"step": 260
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 0.5648535564853556,
|
| 450 |
+
"grad_norm": 32.580466587949495,
|
| 451 |
+
"learning_rate": 2.3722002126275822e-07,
|
| 452 |
+
"logits/chosen": -0.23208408057689667,
|
| 453 |
+
"logits/rejected": 0.542866051197052,
|
| 454 |
+
"logps/chosen": -380.35986328125,
|
| 455 |
+
"logps/rejected": -433.56298828125,
|
| 456 |
+
"loss": 0.5176,
|
| 457 |
+
"rewards/accuracies": 0.706250011920929,
|
| 458 |
+
"rewards/chosen": -1.0335900783538818,
|
| 459 |
+
"rewards/margins": 0.7255193591117859,
|
| 460 |
+
"rewards/rejected": -1.7591092586517334,
|
| 461 |
+
"step": 270
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.5857740585774058,
|
| 465 |
+
"grad_norm": 26.02223175106448,
|
| 466 |
+
"learning_rate": 2.19029145890313e-07,
|
| 467 |
+
"logits/chosen": -0.5966639518737793,
|
| 468 |
+
"logits/rejected": 0.5926642417907715,
|
| 469 |
+
"logps/chosen": -369.0196228027344,
|
| 470 |
+
"logps/rejected": -428.75811767578125,
|
| 471 |
+
"loss": 0.5287,
|
| 472 |
+
"rewards/accuracies": 0.706250011920929,
|
| 473 |
+
"rewards/chosen": -1.0736175775527954,
|
| 474 |
+
"rewards/margins": 0.7921027541160583,
|
| 475 |
+
"rewards/rejected": -1.8657201528549194,
|
| 476 |
+
"step": 280
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 0.606694560669456,
|
| 480 |
+
"grad_norm": 25.62251347650328,
|
| 481 |
+
"learning_rate": 2.0100351342479216e-07,
|
| 482 |
+
"logits/chosen": -0.8950376510620117,
|
| 483 |
+
"logits/rejected": 0.23548343777656555,
|
| 484 |
+
"logps/chosen": -361.40338134765625,
|
| 485 |
+
"logps/rejected": -425.9239196777344,
|
| 486 |
+
"loss": 0.515,
|
| 487 |
+
"rewards/accuracies": 0.737500011920929,
|
| 488 |
+
"rewards/chosen": -1.103497862815857,
|
| 489 |
+
"rewards/margins": 0.7359486222267151,
|
| 490 |
+
"rewards/rejected": -1.8394466638565063,
|
| 491 |
+
"step": 290
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.6276150627615062,
|
| 495 |
+
"grad_norm": 27.37599869158437,
|
| 496 |
+
"learning_rate": 1.8323929841460178e-07,
|
| 497 |
+
"logits/chosen": -0.7678520083427429,
|
| 498 |
+
"logits/rejected": 0.4499804973602295,
|
| 499 |
+
"logps/chosen": -407.6468811035156,
|
| 500 |
+
"logps/rejected": -444.9945373535156,
|
| 501 |
+
"loss": 0.4952,
|
| 502 |
+
"rewards/accuracies": 0.706250011920929,
|
| 503 |
+
"rewards/chosen": -1.0661542415618896,
|
| 504 |
+
"rewards/margins": 0.7832947969436646,
|
| 505 |
+
"rewards/rejected": -1.8494491577148438,
|
| 506 |
+
"step": 300
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 0.6276150627615062,
|
| 510 |
+
"eval_logits/chosen": -0.3939850628376007,
|
| 511 |
+
"eval_logits/rejected": 0.5574444532394409,
|
| 512 |
+
"eval_logps/chosen": -364.551513671875,
|
| 513 |
+
"eval_logps/rejected": -457.5015869140625,
|
| 514 |
+
"eval_loss": 0.506661593914032,
|
| 515 |
+
"eval_rewards/accuracies": 0.7734375,
|
| 516 |
+
"eval_rewards/chosen": -1.0197644233703613,
|
| 517 |
+
"eval_rewards/margins": 0.9284095168113708,
|
| 518 |
+
"eval_rewards/rejected": -1.948173999786377,
|
| 519 |
+
"eval_runtime": 39.937,
|
| 520 |
+
"eval_samples_per_second": 50.079,
|
| 521 |
+
"eval_steps_per_second": 0.801,
|
| 522 |
+
"step": 300
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 0.6485355648535565,
|
| 526 |
+
"grad_norm": 28.6039895111547,
|
| 527 |
+
"learning_rate": 1.6583128063291573e-07,
|
| 528 |
+
"logits/chosen": -0.16188159584999084,
|
| 529 |
+
"logits/rejected": 0.7324084043502808,
|
| 530 |
+
"logps/chosen": -413.48211669921875,
|
| 531 |
+
"logps/rejected": -454.82049560546875,
|
| 532 |
+
"loss": 0.4907,
|
| 533 |
+
"rewards/accuracies": 0.768750011920929,
|
| 534 |
+
"rewards/chosen": -1.1815006732940674,
|
| 535 |
+
"rewards/margins": 0.7815917134284973,
|
| 536 |
+
"rewards/rejected": -1.963092565536499,
|
| 537 |
+
"step": 310
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.6694560669456067,
|
| 541 |
+
"grad_norm": 31.106644428753697,
|
| 542 |
+
"learning_rate": 1.488723393865766e-07,
|
| 543 |
+
"logits/chosen": 0.08699577301740646,
|
| 544 |
+
"logits/rejected": 1.0748140811920166,
|
| 545 |
+
"logps/chosen": -411.9208984375,
|
| 546 |
+
"logps/rejected": -441.51885986328125,
|
| 547 |
+
"loss": 0.4821,
|
| 548 |
+
"rewards/accuracies": 0.762499988079071,
|
| 549 |
+
"rewards/chosen": -1.1822444200515747,
|
| 550 |
+
"rewards/margins": 0.7845796346664429,
|
| 551 |
+
"rewards/rejected": -1.966823935508728,
|
| 552 |
+
"step": 320
|
| 553 |
+
},
|
| 554 |
+
{
|
| 555 |
+
"epoch": 0.6903765690376569,
|
| 556 |
+
"grad_norm": 30.058877077850866,
|
| 557 |
+
"learning_rate": 1.3245295796480788e-07,
|
| 558 |
+
"logits/chosen": -0.6640967130661011,
|
| 559 |
+
"logits/rejected": 0.42964258790016174,
|
| 560 |
+
"logps/chosen": -359.71240234375,
|
| 561 |
+
"logps/rejected": -434.0989685058594,
|
| 562 |
+
"loss": 0.4928,
|
| 563 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 564 |
+
"rewards/chosen": -0.9748628735542297,
|
| 565 |
+
"rewards/margins": 0.7710379362106323,
|
| 566 |
+
"rewards/rejected": -1.7459007501602173,
|
| 567 |
+
"step": 330
|
| 568 |
+
},
|
| 569 |
+
{
|
| 570 |
+
"epoch": 0.7112970711297071,
|
| 571 |
+
"grad_norm": 41.27703909057966,
|
| 572 |
+
"learning_rate": 1.1666074087171627e-07,
|
| 573 |
+
"logits/chosen": -0.4426918923854828,
|
| 574 |
+
"logits/rejected": 0.6907607316970825,
|
| 575 |
+
"logps/chosen": -376.0263671875,
|
| 576 |
+
"logps/rejected": -474.44952392578125,
|
| 577 |
+
"loss": 0.5043,
|
| 578 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 579 |
+
"rewards/chosen": -0.8598421216011047,
|
| 580 |
+
"rewards/margins": 1.083594560623169,
|
| 581 |
+
"rewards/rejected": -1.943436622619629,
|
| 582 |
+
"step": 340
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"epoch": 0.7322175732217573,
|
| 586 |
+
"grad_norm": 25.85661223238129,
|
| 587 |
+
"learning_rate": 1.0157994641835734e-07,
|
| 588 |
+
"logits/chosen": -0.19413986802101135,
|
| 589 |
+
"logits/rejected": 0.754654586315155,
|
| 590 |
+
"logps/chosen": -361.0034484863281,
|
| 591 |
+
"logps/rejected": -430.91253662109375,
|
| 592 |
+
"loss": 0.4784,
|
| 593 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 594 |
+
"rewards/chosen": -1.0284228324890137,
|
| 595 |
+
"rewards/margins": 0.8736444711685181,
|
| 596 |
+
"rewards/rejected": -1.9020671844482422,
|
| 597 |
+
"step": 350
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 0.7531380753138075,
|
| 601 |
+
"grad_norm": 24.132100088486087,
|
| 602 |
+
"learning_rate": 8.729103716819111e-08,
|
| 603 |
+
"logits/chosen": -0.49086451530456543,
|
| 604 |
+
"logits/rejected": 0.8730924725532532,
|
| 605 |
+
"logps/chosen": -414.8470764160156,
|
| 606 |
+
"logps/rejected": -465.10357666015625,
|
| 607 |
+
"loss": 0.5147,
|
| 608 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 609 |
+
"rewards/chosen": -1.1219046115875244,
|
| 610 |
+
"rewards/margins": 0.9166750907897949,
|
| 611 |
+
"rewards/rejected": -2.0385801792144775,
|
| 612 |
+
"step": 360
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 0.7740585774058577,
|
| 616 |
+
"grad_norm": 26.79975843848195,
|
| 617 |
+
"learning_rate": 7.387025063449081e-08,
|
| 618 |
+
"logits/chosen": -0.14118070900440216,
|
| 619 |
+
"logits/rejected": 0.7372316718101501,
|
| 620 |
+
"logps/chosen": -402.12664794921875,
|
| 621 |
+
"logps/rejected": -436.84375,
|
| 622 |
+
"loss": 0.509,
|
| 623 |
+
"rewards/accuracies": 0.71875,
|
| 624 |
+
"rewards/chosen": -1.2679729461669922,
|
| 625 |
+
"rewards/margins": 0.7179969549179077,
|
| 626 |
+
"rewards/rejected": -1.9859699010849,
|
| 627 |
+
"step": 370
|
| 628 |
+
},
|
| 629 |
+
{
|
| 630 |
+
"epoch": 0.7949790794979079,
|
| 631 |
+
"grad_norm": 25.72376549261972,
|
| 632 |
+
"learning_rate": 6.138919252022435e-08,
|
| 633 |
+
"logits/chosen": 0.09815589338541031,
|
| 634 |
+
"logits/rejected": 1.0596524477005005,
|
| 635 |
+
"logps/chosen": -381.30810546875,
|
| 636 |
+
"logps/rejected": -490.9652404785156,
|
| 637 |
+
"loss": 0.5048,
|
| 638 |
+
"rewards/accuracies": 0.75,
|
| 639 |
+
"rewards/chosen": -1.4183294773101807,
|
| 640 |
+
"rewards/margins": 0.9044706225395203,
|
| 641 |
+
"rewards/rejected": -2.3228001594543457,
|
| 642 |
+
"step": 380
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 0.8158995815899581,
|
| 646 |
+
"grad_norm": 24.193050103935505,
|
| 647 |
+
"learning_rate": 4.991445467064689e-08,
|
| 648 |
+
"logits/chosen": -0.3292013108730316,
|
| 649 |
+
"logits/rejected": 0.42220011353492737,
|
| 650 |
+
"logps/chosen": -420.472900390625,
|
| 651 |
+
"logps/rejected": -483.74005126953125,
|
| 652 |
+
"loss": 0.4897,
|
| 653 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 654 |
+
"rewards/chosen": -1.1483708620071411,
|
| 655 |
+
"rewards/margins": 0.8494104146957397,
|
| 656 |
+
"rewards/rejected": -1.99778151512146,
|
| 657 |
+
"step": 390
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 0.8368200836820083,
|
| 661 |
+
"grad_norm": 25.887805811543046,
|
| 662 |
+
"learning_rate": 3.9507259776993954e-08,
|
| 663 |
+
"logits/chosen": -0.24365536868572235,
|
| 664 |
+
"logits/rejected": 1.0439943075180054,
|
| 665 |
+
"logps/chosen": -386.0181579589844,
|
| 666 |
+
"logps/rejected": -473.2196350097656,
|
| 667 |
+
"loss": 0.5037,
|
| 668 |
+
"rewards/accuracies": 0.768750011920929,
|
| 669 |
+
"rewards/chosen": -1.1546484231948853,
|
| 670 |
+
"rewards/margins": 0.9076915979385376,
|
| 671 |
+
"rewards/rejected": -2.062340259552002,
|
| 672 |
+
"step": 400
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"epoch": 0.8368200836820083,
|
| 676 |
+
"eval_logits/chosen": -0.3904561996459961,
|
| 677 |
+
"eval_logits/rejected": 0.6357870101928711,
|
| 678 |
+
"eval_logps/chosen": -366.52386474609375,
|
| 679 |
+
"eval_logps/rejected": -463.7657775878906,
|
| 680 |
+
"eval_loss": 0.5006148815155029,
|
| 681 |
+
"eval_rewards/accuracies": 0.78515625,
|
| 682 |
+
"eval_rewards/chosen": -1.0394879579544067,
|
| 683 |
+
"eval_rewards/margins": 0.9713287949562073,
|
| 684 |
+
"eval_rewards/rejected": -2.010816812515259,
|
| 685 |
+
"eval_runtime": 39.7974,
|
| 686 |
+
"eval_samples_per_second": 50.254,
|
| 687 |
+
"eval_steps_per_second": 0.804,
|
| 688 |
+
"step": 400
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"epoch": 0.8577405857740585,
|
| 692 |
+
"grad_norm": 28.490429904013755,
|
| 693 |
+
"learning_rate": 3.022313472693447e-08,
|
| 694 |
+
"logits/chosen": -0.38060712814331055,
|
| 695 |
+
"logits/rejected": 0.9389992952346802,
|
| 696 |
+
"logps/chosen": -409.4950256347656,
|
| 697 |
+
"logps/rejected": -468.64715576171875,
|
| 698 |
+
"loss": 0.5073,
|
| 699 |
+
"rewards/accuracies": 0.768750011920929,
|
| 700 |
+
"rewards/chosen": -1.1054986715316772,
|
| 701 |
+
"rewards/margins": 0.8928801417350769,
|
| 702 |
+
"rewards/rejected": -1.9983787536621094,
|
| 703 |
+
"step": 410
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.8786610878661087,
|
| 707 |
+
"grad_norm": 24.562220908212364,
|
| 708 |
+
"learning_rate": 2.2111614344599684e-08,
|
| 709 |
+
"logits/chosen": -0.4677561819553375,
|
| 710 |
+
"logits/rejected": 0.5116729736328125,
|
| 711 |
+
"logps/chosen": -406.060546875,
|
| 712 |
+
"logps/rejected": -465.40130615234375,
|
| 713 |
+
"loss": 0.4974,
|
| 714 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 715 |
+
"rewards/chosen": -1.0901445150375366,
|
| 716 |
+
"rewards/margins": 0.8191754221916199,
|
| 717 |
+
"rewards/rejected": -1.9093201160430908,
|
| 718 |
+
"step": 420
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"epoch": 0.899581589958159,
|
| 722 |
+
"grad_norm": 25.456098375916802,
|
| 723 |
+
"learning_rate": 1.521597710086439e-08,
|
| 724 |
+
"logits/chosen": -0.10303229093551636,
|
| 725 |
+
"logits/rejected": 1.026650309562683,
|
| 726 |
+
"logps/chosen": -397.377197265625,
|
| 727 |
+
"logps/rejected": -464.6932067871094,
|
| 728 |
+
"loss": 0.4766,
|
| 729 |
+
"rewards/accuracies": 0.78125,
|
| 730 |
+
"rewards/chosen": -1.182983636856079,
|
| 731 |
+
"rewards/margins": 0.8865542411804199,
|
| 732 |
+
"rewards/rejected": -2.06953763961792,
|
| 733 |
+
"step": 430
|
| 734 |
+
},
|
| 735 |
+
{
|
| 736 |
+
"epoch": 0.9205020920502092,
|
| 737 |
+
"grad_norm": 28.11934736057114,
|
| 738 |
+
"learning_rate": 9.57301420397924e-09,
|
| 739 |
+
"logits/chosen": -0.32768553495407104,
|
| 740 |
+
"logits/rejected": 0.8762621879577637,
|
| 741 |
+
"logps/chosen": -392.1067810058594,
|
| 742 |
+
"logps/rejected": -460.0609436035156,
|
| 743 |
+
"loss": 0.4971,
|
| 744 |
+
"rewards/accuracies": 0.75,
|
| 745 |
+
"rewards/chosen": -1.0779250860214233,
|
| 746 |
+
"rewards/margins": 0.8317632675170898,
|
| 747 |
+
"rewards/rejected": -1.9096883535385132,
|
| 748 |
+
"step": 440
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"epoch": 0.9414225941422594,
|
| 752 |
+
"grad_norm": 24.86592939122457,
|
| 753 |
+
"learning_rate": 5.212833302556258e-09,
|
| 754 |
+
"logits/chosen": -0.24481463432312012,
|
| 755 |
+
"logits/rejected": 0.7605234384536743,
|
| 756 |
+
"logps/chosen": -409.94793701171875,
|
| 757 |
+
"logps/rejected": -506.735595703125,
|
| 758 |
+
"loss": 0.4959,
|
| 759 |
+
"rewards/accuracies": 0.706250011920929,
|
| 760 |
+
"rewards/chosen": -1.1622434854507446,
|
| 761 |
+
"rewards/margins": 0.8344427347183228,
|
| 762 |
+
"rewards/rejected": -1.9966862201690674,
|
| 763 |
+
"step": 450
|
| 764 |
+
},
|
| 765 |
+
{
|
| 766 |
+
"epoch": 0.9623430962343096,
|
| 767 |
+
"grad_norm": 27.408370428716005,
|
| 768 |
+
"learning_rate": 2.158697848236607e-09,
|
| 769 |
+
"logits/chosen": -0.22167810797691345,
|
| 770 |
+
"logits/rejected": 0.9164765477180481,
|
| 771 |
+
"logps/chosen": -382.83819580078125,
|
| 772 |
+
"logps/rejected": -436.36114501953125,
|
| 773 |
+
"loss": 0.498,
|
| 774 |
+
"rewards/accuracies": 0.71875,
|
| 775 |
+
"rewards/chosen": -1.1093404293060303,
|
| 776 |
+
"rewards/margins": 0.827775776386261,
|
| 777 |
+
"rewards/rejected": -1.937116265296936,
|
| 778 |
+
"step": 460
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"epoch": 0.9832635983263598,
|
| 782 |
+
"grad_norm": 25.655126921818987,
|
| 783 |
+
"learning_rate": 4.269029751107489e-10,
|
| 784 |
+
"logits/chosen": -0.3841325640678406,
|
| 785 |
+
"logits/rejected": 0.87626713514328,
|
| 786 |
+
"logps/chosen": -385.66424560546875,
|
| 787 |
+
"logps/rejected": -474.7689514160156,
|
| 788 |
+
"loss": 0.492,
|
| 789 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 790 |
+
"rewards/chosen": -1.0700037479400635,
|
| 791 |
+
"rewards/margins": 0.8540438413619995,
|
| 792 |
+
"rewards/rejected": -1.9240477085113525,
|
| 793 |
+
"step": 470
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"epoch": 1.0,
|
| 797 |
+
"step": 478,
|
| 798 |
+
"total_flos": 0.0,
|
| 799 |
+
"train_loss": 0.5400827195355085,
|
| 800 |
+
"train_runtime": 3932.0785,
|
| 801 |
+
"train_samples_per_second": 15.548,
|
| 802 |
+
"train_steps_per_second": 0.122
|
| 803 |
+
}
|
| 804 |
+
],
|
| 805 |
+
"logging_steps": 10,
|
| 806 |
+
"max_steps": 478,
|
| 807 |
+
"num_input_tokens_seen": 0,
|
| 808 |
+
"num_train_epochs": 1,
|
| 809 |
+
"save_steps": 100,
|
| 810 |
+
"stateful_callbacks": {
|
| 811 |
+
"TrainerControl": {
|
| 812 |
+
"args": {
|
| 813 |
+
"should_epoch_stop": false,
|
| 814 |
+
"should_evaluate": false,
|
| 815 |
+
"should_log": false,
|
| 816 |
+
"should_save": true,
|
| 817 |
+
"should_training_stop": true
|
| 818 |
+
},
|
| 819 |
+
"attributes": {}
|
| 820 |
+
}
|
| 821 |
+
},
|
| 822 |
+
"total_flos": 0.0,
|
| 823 |
+
"train_batch_size": 8,
|
| 824 |
+
"trial_name": null,
|
| 825 |
+
"trial_params": null
|
| 826 |
+
}
|