Model save
Browse files- README.md +76 -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 +1,76 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
base_model: alignment-handbook/zephyr-7b-sft-full
|
| 4 |
+
tags:
|
| 5 |
+
- trl
|
| 6 |
+
- dpo
|
| 7 |
+
- generated_from_trainer
|
| 8 |
+
model-index:
|
| 9 |
+
- name: zephyr-7b-dpo-full
|
| 10 |
+
results: []
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 14 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 15 |
+
|
| 16 |
+
# zephyr-7b-dpo-full
|
| 17 |
+
|
| 18 |
+
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.
|
| 19 |
+
It achieves the following results on the evaluation set:
|
| 20 |
+
- Loss: 0.5032
|
| 21 |
+
- Rewards/chosen: -1.0014
|
| 22 |
+
- Rewards/rejected: -1.9606
|
| 23 |
+
- Rewards/accuracies: 0.7734
|
| 24 |
+
- Rewards/margins: 0.9592
|
| 25 |
+
- Logps/rejected: -458.7184
|
| 26 |
+
- Logps/chosen: -362.7651
|
| 27 |
+
- Logits/rejected: 1.3908
|
| 28 |
+
- Logits/chosen: 0.5145
|
| 29 |
+
|
| 30 |
+
## Model description
|
| 31 |
+
|
| 32 |
+
More information needed
|
| 33 |
+
|
| 34 |
+
## Intended uses & limitations
|
| 35 |
+
|
| 36 |
+
More information needed
|
| 37 |
+
|
| 38 |
+
## Training and evaluation data
|
| 39 |
+
|
| 40 |
+
More information needed
|
| 41 |
+
|
| 42 |
+
## Training procedure
|
| 43 |
+
|
| 44 |
+
### Training hyperparameters
|
| 45 |
+
|
| 46 |
+
The following hyperparameters were used during training:
|
| 47 |
+
- learning_rate: 5e-07
|
| 48 |
+
- train_batch_size: 8
|
| 49 |
+
- eval_batch_size: 8
|
| 50 |
+
- seed: 42
|
| 51 |
+
- distributed_type: multi-GPU
|
| 52 |
+
- num_devices: 8
|
| 53 |
+
- gradient_accumulation_steps: 2
|
| 54 |
+
- total_train_batch_size: 128
|
| 55 |
+
- total_eval_batch_size: 64
|
| 56 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 57 |
+
- lr_scheduler_type: cosine
|
| 58 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 59 |
+
- num_epochs: 1
|
| 60 |
+
|
| 61 |
+
### Training results
|
| 62 |
+
|
| 63 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 64 |
+
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 65 |
+
| 0.5655 | 0.2092 | 100 | 0.5746 | -0.7161 | -1.2966 | 0.7070 | 0.5805 | -392.3193 | -334.2372 | -0.4365 | -0.6501 |
|
| 66 |
+
| 0.5457 | 0.4184 | 200 | 0.5249 | -0.9842 | -1.7929 | 0.7773 | 0.8086 | -441.9472 | -361.0536 | 0.9213 | 0.3160 |
|
| 67 |
+
| 0.4955 | 0.6276 | 300 | 0.5092 | -0.9649 | -1.8656 | 0.7656 | 0.9007 | -449.2220 | -359.1224 | 1.2341 | 0.4705 |
|
| 68 |
+
| 0.5033 | 0.8368 | 400 | 0.5032 | -1.0014 | -1.9606 | 0.7734 | 0.9592 | -458.7184 | -362.7651 | 1.3908 | 0.5145 |
|
| 69 |
+
|
| 70 |
+
|
| 71 |
+
### Framework versions
|
| 72 |
+
|
| 73 |
+
- Transformers 4.44.0
|
| 74 |
+
- Pytorch 2.4.0+cu121
|
| 75 |
+
- Datasets 2.21.0
|
| 76 |
+
- Tokenizers 0.19.1
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5406385645208,
|
| 5 |
+
"train_runtime": 10904.9554,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 5.606,
|
| 8 |
+
"train_steps_per_second": 0.044
|
| 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": "4.44.0"
|
| 6 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5406385645208,
|
| 5 |
+
"train_runtime": 10904.9554,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 5.606,
|
| 8 |
+
"train_steps_per_second": 0.044
|
| 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": 9.170599089942366,
|
| 14 |
+
"learning_rate": 1.0416666666666666e-08,
|
| 15 |
+
"logits/chosen": -2.7655434608459473,
|
| 16 |
+
"logits/rejected": -2.7158141136169434,
|
| 17 |
+
"logps/chosen": -269.8756103515625,
|
| 18 |
+
"logps/rejected": -360.62164306640625,
|
| 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": 9.150565723604885,
|
| 29 |
+
"learning_rate": 1.0416666666666667e-07,
|
| 30 |
+
"logits/chosen": -2.5923984050750732,
|
| 31 |
+
"logits/rejected": -2.562947988510132,
|
| 32 |
+
"logps/chosen": -264.8230895996094,
|
| 33 |
+
"logps/rejected": -251.5508575439453,
|
| 34 |
+
"loss": 0.6931,
|
| 35 |
+
"rewards/accuracies": 0.3888888955116272,
|
| 36 |
+
"rewards/chosen": -0.0003005376784130931,
|
| 37 |
+
"rewards/margins": -0.00048651432734914124,
|
| 38 |
+
"rewards/rejected": 0.0001859766198322177,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.04184100418410042,
|
| 43 |
+
"grad_norm": 8.401794731870115,
|
| 44 |
+
"learning_rate": 2.0833333333333333e-07,
|
| 45 |
+
"logits/chosen": -2.6533799171447754,
|
| 46 |
+
"logits/rejected": -2.6054489612579346,
|
| 47 |
+
"logps/chosen": -281.46368408203125,
|
| 48 |
+
"logps/rejected": -296.8436584472656,
|
| 49 |
+
"loss": 0.6924,
|
| 50 |
+
"rewards/accuracies": 0.512499988079071,
|
| 51 |
+
"rewards/chosen": 0.0004009083495475352,
|
| 52 |
+
"rewards/margins": 0.000830893695820123,
|
| 53 |
+
"rewards/rejected": -0.00042998549179174006,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.06276150627615062,
|
| 58 |
+
"grad_norm": 8.275952755978503,
|
| 59 |
+
"learning_rate": 3.1249999999999997e-07,
|
| 60 |
+
"logits/chosen": -2.6644034385681152,
|
| 61 |
+
"logits/rejected": -2.591947078704834,
|
| 62 |
+
"logps/chosen": -299.4920959472656,
|
| 63 |
+
"logps/rejected": -263.7144470214844,
|
| 64 |
+
"loss": 0.6885,
|
| 65 |
+
"rewards/accuracies": 0.637499988079071,
|
| 66 |
+
"rewards/chosen": 0.011641914024949074,
|
| 67 |
+
"rewards/margins": 0.009539209306240082,
|
| 68 |
+
"rewards/rejected": 0.0021027035545557737,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.08368200836820083,
|
| 73 |
+
"grad_norm": 8.165628930324349,
|
| 74 |
+
"learning_rate": 4.1666666666666667e-07,
|
| 75 |
+
"logits/chosen": -2.557978391647339,
|
| 76 |
+
"logits/rejected": -2.5226192474365234,
|
| 77 |
+
"logps/chosen": -263.15728759765625,
|
| 78 |
+
"logps/rejected": -245.1326141357422,
|
| 79 |
+
"loss": 0.6776,
|
| 80 |
+
"rewards/accuracies": 0.706250011920929,
|
| 81 |
+
"rewards/chosen": 0.03938329964876175,
|
| 82 |
+
"rewards/margins": 0.04691743105649948,
|
| 83 |
+
"rewards/rejected": -0.00753413001075387,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.10460251046025104,
|
| 88 |
+
"grad_norm": 8.408127463371278,
|
| 89 |
+
"learning_rate": 4.999733114418725e-07,
|
| 90 |
+
"logits/chosen": -2.499746084213257,
|
| 91 |
+
"logits/rejected": -2.4598052501678467,
|
| 92 |
+
"logps/chosen": -279.6321716308594,
|
| 93 |
+
"logps/rejected": -297.48968505859375,
|
| 94 |
+
"loss": 0.6615,
|
| 95 |
+
"rewards/accuracies": 0.612500011920929,
|
| 96 |
+
"rewards/chosen": 0.0010753454407677054,
|
| 97 |
+
"rewards/margins": 0.04832205921411514,
|
| 98 |
+
"rewards/rejected": -0.04724670574069023,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.12552301255230125,
|
| 103 |
+
"grad_norm": 11.367364527138742,
|
| 104 |
+
"learning_rate": 4.990398100856366e-07,
|
| 105 |
+
"logits/chosen": -2.5166015625,
|
| 106 |
+
"logits/rejected": -2.4452641010284424,
|
| 107 |
+
"logps/chosen": -273.73309326171875,
|
| 108 |
+
"logps/rejected": -312.8514709472656,
|
| 109 |
+
"loss": 0.639,
|
| 110 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 111 |
+
"rewards/chosen": -0.02069356106221676,
|
| 112 |
+
"rewards/margins": 0.1263645738363266,
|
| 113 |
+
"rewards/rejected": -0.1470581293106079,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.14644351464435146,
|
| 118 |
+
"grad_norm": 10.797907272646912,
|
| 119 |
+
"learning_rate": 4.967775735898179e-07,
|
| 120 |
+
"logits/chosen": -2.369143009185791,
|
| 121 |
+
"logits/rejected": -2.366969585418701,
|
| 122 |
+
"logps/chosen": -270.88665771484375,
|
| 123 |
+
"logps/rejected": -283.3267517089844,
|
| 124 |
+
"loss": 0.6162,
|
| 125 |
+
"rewards/accuracies": 0.762499988079071,
|
| 126 |
+
"rewards/chosen": -0.05217716097831726,
|
| 127 |
+
"rewards/margins": 0.21364183723926544,
|
| 128 |
+
"rewards/rejected": -0.2658190131187439,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.16736401673640167,
|
| 133 |
+
"grad_norm": 20.445807364960142,
|
| 134 |
+
"learning_rate": 4.931986719649298e-07,
|
| 135 |
+
"logits/chosen": -2.560523509979248,
|
| 136 |
+
"logits/rejected": -2.4790120124816895,
|
| 137 |
+
"logps/chosen": -336.60650634765625,
|
| 138 |
+
"logps/rejected": -304.67822265625,
|
| 139 |
+
"loss": 0.6025,
|
| 140 |
+
"rewards/accuracies": 0.675000011920929,
|
| 141 |
+
"rewards/chosen": -0.13829118013381958,
|
| 142 |
+
"rewards/margins": 0.26742225885391235,
|
| 143 |
+
"rewards/rejected": -0.4057134687900543,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.18828451882845187,
|
| 148 |
+
"grad_norm": 16.762310140078398,
|
| 149 |
+
"learning_rate": 4.883222001996351e-07,
|
| 150 |
+
"logits/chosen": -2.073870897293091,
|
| 151 |
+
"logits/rejected": -2.0239508152008057,
|
| 152 |
+
"logps/chosen": -294.5444641113281,
|
| 153 |
+
"logps/rejected": -334.00042724609375,
|
| 154 |
+
"loss": 0.5699,
|
| 155 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 156 |
+
"rewards/chosen": -0.3539494276046753,
|
| 157 |
+
"rewards/margins": 0.46201545000076294,
|
| 158 |
+
"rewards/rejected": -0.8159648180007935,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.20920502092050208,
|
| 163 |
+
"grad_norm": 18.455144867656706,
|
| 164 |
+
"learning_rate": 4.821741763807186e-07,
|
| 165 |
+
"logits/chosen": -1.0050065517425537,
|
| 166 |
+
"logits/rejected": -0.8190995454788208,
|
| 167 |
+
"logps/chosen": -317.0809631347656,
|
| 168 |
+
"logps/rejected": -373.81805419921875,
|
| 169 |
+
"loss": 0.5655,
|
| 170 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 171 |
+
"rewards/chosen": -0.5671249628067017,
|
| 172 |
+
"rewards/margins": 0.6001720428466797,
|
| 173 |
+
"rewards/rejected": -1.1672970056533813,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 0.20920502092050208,
|
| 178 |
+
"eval_logits/chosen": -0.6500550508499146,
|
| 179 |
+
"eval_logits/rejected": -0.4365238547325134,
|
| 180 |
+
"eval_logps/chosen": -334.2371520996094,
|
| 181 |
+
"eval_logps/rejected": -392.31927490234375,
|
| 182 |
+
"eval_loss": 0.5745880603790283,
|
| 183 |
+
"eval_rewards/accuracies": 0.70703125,
|
| 184 |
+
"eval_rewards/chosen": -0.7160719633102417,
|
| 185 |
+
"eval_rewards/margins": 0.5804991126060486,
|
| 186 |
+
"eval_rewards/rejected": -1.2965710163116455,
|
| 187 |
+
"eval_runtime": 110.283,
|
| 188 |
+
"eval_samples_per_second": 18.135,
|
| 189 |
+
"eval_steps_per_second": 0.29,
|
| 190 |
+
"step": 100
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 0.2301255230125523,
|
| 194 |
+
"grad_norm": 17.43281841430265,
|
| 195 |
+
"learning_rate": 4.747874028753375e-07,
|
| 196 |
+
"logits/chosen": -0.7553631067276001,
|
| 197 |
+
"logits/rejected": -0.2396622896194458,
|
| 198 |
+
"logps/chosen": -366.02532958984375,
|
| 199 |
+
"logps/rejected": -376.58050537109375,
|
| 200 |
+
"loss": 0.568,
|
| 201 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 202 |
+
"rewards/chosen": -0.61496901512146,
|
| 203 |
+
"rewards/margins": 0.5723368525505066,
|
| 204 |
+
"rewards/rejected": -1.1873060464859009,
|
| 205 |
+
"step": 110
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.2510460251046025,
|
| 209 |
+
"grad_norm": 19.437266977582095,
|
| 210 |
+
"learning_rate": 4.662012913161997e-07,
|
| 211 |
+
"logits/chosen": -0.5649857521057129,
|
| 212 |
+
"logits/rejected": -0.25259464979171753,
|
| 213 |
+
"logps/chosen": -341.97308349609375,
|
| 214 |
+
"logps/rejected": -376.1507568359375,
|
| 215 |
+
"loss": 0.5592,
|
| 216 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 217 |
+
"rewards/chosen": -0.6079371571540833,
|
| 218 |
+
"rewards/margins": 0.5546227693557739,
|
| 219 |
+
"rewards/rejected": -1.162559986114502,
|
| 220 |
+
"step": 120
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.2719665271966527,
|
| 224 |
+
"grad_norm": 25.375107282483974,
|
| 225 |
+
"learning_rate": 4.5646165232345103e-07,
|
| 226 |
+
"logits/chosen": -0.06077578663825989,
|
| 227 |
+
"logits/rejected": 0.2923804521560669,
|
| 228 |
+
"logps/chosen": -373.219482421875,
|
| 229 |
+
"logps/rejected": -418.9559631347656,
|
| 230 |
+
"loss": 0.5347,
|
| 231 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 232 |
+
"rewards/chosen": -0.9123528599739075,
|
| 233 |
+
"rewards/margins": 0.592673659324646,
|
| 234 |
+
"rewards/rejected": -1.5050265789031982,
|
| 235 |
+
"step": 130
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.2928870292887029,
|
| 239 |
+
"grad_norm": 21.746466725148508,
|
| 240 |
+
"learning_rate": 4.456204510851956e-07,
|
| 241 |
+
"logits/chosen": -0.2530488073825836,
|
| 242 |
+
"logits/rejected": 0.0992981344461441,
|
| 243 |
+
"logps/chosen": -384.7760009765625,
|
| 244 |
+
"logps/rejected": -437.4297790527344,
|
| 245 |
+
"loss": 0.5317,
|
| 246 |
+
"rewards/accuracies": 0.75,
|
| 247 |
+
"rewards/chosen": -0.6880978345870972,
|
| 248 |
+
"rewards/margins": 0.6797319650650024,
|
| 249 |
+
"rewards/rejected": -1.3678295612335205,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.3138075313807531,
|
| 254 |
+
"grad_norm": 26.091710204113372,
|
| 255 |
+
"learning_rate": 4.337355301007335e-07,
|
| 256 |
+
"logits/chosen": 0.01746506430208683,
|
| 257 |
+
"logits/rejected": 0.5026353597640991,
|
| 258 |
+
"logps/chosen": -387.0420227050781,
|
| 259 |
+
"logps/rejected": -434.38580322265625,
|
| 260 |
+
"loss": 0.5459,
|
| 261 |
+
"rewards/accuracies": 0.699999988079071,
|
| 262 |
+
"rewards/chosen": -1.0380537509918213,
|
| 263 |
+
"rewards/margins": 0.5438278913497925,
|
| 264 |
+
"rewards/rejected": -1.5818815231323242,
|
| 265 |
+
"step": 150
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.33472803347280333,
|
| 269 |
+
"grad_norm": 23.15637544568347,
|
| 270 |
+
"learning_rate": 4.2087030056579986e-07,
|
| 271 |
+
"logits/chosen": -0.19735758006572723,
|
| 272 |
+
"logits/rejected": 0.6574625968933105,
|
| 273 |
+
"logps/chosen": -339.94073486328125,
|
| 274 |
+
"logps/rejected": -404.662109375,
|
| 275 |
+
"loss": 0.5375,
|
| 276 |
+
"rewards/accuracies": 0.762499988079071,
|
| 277 |
+
"rewards/chosen": -0.668976902961731,
|
| 278 |
+
"rewards/margins": 0.8034380078315735,
|
| 279 |
+
"rewards/rejected": -1.4724147319793701,
|
| 280 |
+
"step": 160
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.35564853556485354,
|
| 284 |
+
"grad_norm": 26.27445620588168,
|
| 285 |
+
"learning_rate": 4.070934040463998e-07,
|
| 286 |
+
"logits/chosen": 0.8700563311576843,
|
| 287 |
+
"logits/rejected": 1.3576207160949707,
|
| 288 |
+
"logps/chosen": -358.09051513671875,
|
| 289 |
+
"logps/rejected": -400.4471740722656,
|
| 290 |
+
"loss": 0.5423,
|
| 291 |
+
"rewards/accuracies": 0.706250011920929,
|
| 292 |
+
"rewards/chosen": -1.0877528190612793,
|
| 293 |
+
"rewards/margins": 0.6053714156150818,
|
| 294 |
+
"rewards/rejected": -1.6931244134902954,
|
| 295 |
+
"step": 170
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 0.37656903765690375,
|
| 299 |
+
"grad_norm": 36.79313621128292,
|
| 300 |
+
"learning_rate": 3.9247834624635404e-07,
|
| 301 |
+
"logits/chosen": 1.2770946025848389,
|
| 302 |
+
"logits/rejected": 1.9397356510162354,
|
| 303 |
+
"logps/chosen": -342.9526672363281,
|
| 304 |
+
"logps/rejected": -384.0509338378906,
|
| 305 |
+
"loss": 0.5076,
|
| 306 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 307 |
+
"rewards/chosen": -0.8662456274032593,
|
| 308 |
+
"rewards/margins": 0.6985085010528564,
|
| 309 |
+
"rewards/rejected": -1.5647542476654053,
|
| 310 |
+
"step": 180
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.39748953974895396,
|
| 314 |
+
"grad_norm": 23.134859194828028,
|
| 315 |
+
"learning_rate": 3.7710310482256523e-07,
|
| 316 |
+
"logits/chosen": 0.7933472990989685,
|
| 317 |
+
"logits/rejected": 1.352903127670288,
|
| 318 |
+
"logps/chosen": -354.90484619140625,
|
| 319 |
+
"logps/rejected": -407.89801025390625,
|
| 320 |
+
"loss": 0.5173,
|
| 321 |
+
"rewards/accuracies": 0.71875,
|
| 322 |
+
"rewards/chosen": -0.8360937237739563,
|
| 323 |
+
"rewards/margins": 0.6451088190078735,
|
| 324 |
+
"rewards/rejected": -1.481202483177185,
|
| 325 |
+
"step": 190
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 0.41841004184100417,
|
| 329 |
+
"grad_norm": 51.000005370435,
|
| 330 |
+
"learning_rate": 3.610497133404795e-07,
|
| 331 |
+
"logits/chosen": 0.7528682947158813,
|
| 332 |
+
"logits/rejected": 1.2566933631896973,
|
| 333 |
+
"logps/chosen": -360.9665832519531,
|
| 334 |
+
"logps/rejected": -426.8341369628906,
|
| 335 |
+
"loss": 0.5457,
|
| 336 |
+
"rewards/accuracies": 0.737500011920929,
|
| 337 |
+
"rewards/chosen": -1.0107109546661377,
|
| 338 |
+
"rewards/margins": 0.7371687293052673,
|
| 339 |
+
"rewards/rejected": -1.7478797435760498,
|
| 340 |
+
"step": 200
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 0.41841004184100417,
|
| 344 |
+
"eval_logits/chosen": 0.3159633278846741,
|
| 345 |
+
"eval_logits/rejected": 0.9213072061538696,
|
| 346 |
+
"eval_logps/chosen": -361.0535888671875,
|
| 347 |
+
"eval_logps/rejected": -441.9471740722656,
|
| 348 |
+
"eval_loss": 0.5249255299568176,
|
| 349 |
+
"eval_rewards/accuracies": 0.77734375,
|
| 350 |
+
"eval_rewards/chosen": -0.9842361211776733,
|
| 351 |
+
"eval_rewards/margins": 0.808614194393158,
|
| 352 |
+
"eval_rewards/rejected": -1.7928502559661865,
|
| 353 |
+
"eval_runtime": 111.6649,
|
| 354 |
+
"eval_samples_per_second": 17.911,
|
| 355 |
+
"eval_steps_per_second": 0.287,
|
| 356 |
+
"step": 200
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 0.4393305439330544,
|
| 360 |
+
"grad_norm": 25.554968491047823,
|
| 361 |
+
"learning_rate": 3.4440382358952115e-07,
|
| 362 |
+
"logits/chosen": 0.3864581882953644,
|
| 363 |
+
"logits/rejected": 1.292891263961792,
|
| 364 |
+
"logps/chosen": -393.22100830078125,
|
| 365 |
+
"logps/rejected": -420.07012939453125,
|
| 366 |
+
"loss": 0.5389,
|
| 367 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 368 |
+
"rewards/chosen": -1.0355665683746338,
|
| 369 |
+
"rewards/margins": 0.652549684047699,
|
| 370 |
+
"rewards/rejected": -1.6881163120269775,
|
| 371 |
+
"step": 210
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 0.4602510460251046,
|
| 375 |
+
"grad_norm": 29.830177463941375,
|
| 376 |
+
"learning_rate": 3.272542485937368e-07,
|
| 377 |
+
"logits/chosen": 0.6986919641494751,
|
| 378 |
+
"logits/rejected": 1.5640161037445068,
|
| 379 |
+
"logps/chosen": -370.61199951171875,
|
| 380 |
+
"logps/rejected": -428.12554931640625,
|
| 381 |
+
"loss": 0.5263,
|
| 382 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 383 |
+
"rewards/chosen": -1.0523474216461182,
|
| 384 |
+
"rewards/margins": 0.7719280123710632,
|
| 385 |
+
"rewards/rejected": -1.8242753744125366,
|
| 386 |
+
"step": 220
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 0.4811715481171548,
|
| 390 |
+
"grad_norm": 21.180298956536415,
|
| 391 |
+
"learning_rate": 3.096924887558854e-07,
|
| 392 |
+
"logits/chosen": 0.7276783585548401,
|
| 393 |
+
"logits/rejected": 1.484283447265625,
|
| 394 |
+
"logps/chosen": -353.75164794921875,
|
| 395 |
+
"logps/rejected": -443.76275634765625,
|
| 396 |
+
"loss": 0.5424,
|
| 397 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 398 |
+
"rewards/chosen": -1.0479848384857178,
|
| 399 |
+
"rewards/margins": 0.9119135737419128,
|
| 400 |
+
"rewards/rejected": -1.959898591041565,
|
| 401 |
+
"step": 230
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 0.502092050209205,
|
| 405 |
+
"grad_norm": 23.974967302833218,
|
| 406 |
+
"learning_rate": 2.9181224366319943e-07,
|
| 407 |
+
"logits/chosen": 0.2961747646331787,
|
| 408 |
+
"logits/rejected": 1.008892297744751,
|
| 409 |
+
"logps/chosen": -349.42010498046875,
|
| 410 |
+
"logps/rejected": -405.09637451171875,
|
| 411 |
+
"loss": 0.5026,
|
| 412 |
+
"rewards/accuracies": 0.793749988079071,
|
| 413 |
+
"rewards/chosen": -0.8568391799926758,
|
| 414 |
+
"rewards/margins": 0.7035945057868958,
|
| 415 |
+
"rewards/rejected": -1.5604336261749268,
|
| 416 |
+
"step": 240
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 0.5230125523012552,
|
| 420 |
+
"grad_norm": 28.98490083610951,
|
| 421 |
+
"learning_rate": 2.7370891215954565e-07,
|
| 422 |
+
"logits/chosen": 1.0544288158416748,
|
| 423 |
+
"logits/rejected": 1.983277678489685,
|
| 424 |
+
"logps/chosen": -397.53851318359375,
|
| 425 |
+
"logps/rejected": -447.08734130859375,
|
| 426 |
+
"loss": 0.5082,
|
| 427 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 428 |
+
"rewards/chosen": -0.9933551549911499,
|
| 429 |
+
"rewards/margins": 0.8820589780807495,
|
| 430 |
+
"rewards/rejected": -1.8754141330718994,
|
| 431 |
+
"step": 250
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 0.5439330543933054,
|
| 435 |
+
"grad_norm": 34.570387827045344,
|
| 436 |
+
"learning_rate": 2.55479083351317e-07,
|
| 437 |
+
"logits/chosen": 1.1669882535934448,
|
| 438 |
+
"logits/rejected": 2.0455715656280518,
|
| 439 |
+
"logps/chosen": -404.6676940917969,
|
| 440 |
+
"logps/rejected": -453.14019775390625,
|
| 441 |
+
"loss": 0.5115,
|
| 442 |
+
"rewards/accuracies": 0.731249988079071,
|
| 443 |
+
"rewards/chosen": -1.1036981344223022,
|
| 444 |
+
"rewards/margins": 0.8884363174438477,
|
| 445 |
+
"rewards/rejected": -1.992134690284729,
|
| 446 |
+
"step": 260
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 0.5648535564853556,
|
| 450 |
+
"grad_norm": 23.541591723414815,
|
| 451 |
+
"learning_rate": 2.3722002126275822e-07,
|
| 452 |
+
"logits/chosen": 0.38921916484832764,
|
| 453 |
+
"logits/rejected": 1.070547342300415,
|
| 454 |
+
"logps/chosen": -363.3959655761719,
|
| 455 |
+
"logps/rejected": -411.51025390625,
|
| 456 |
+
"loss": 0.5271,
|
| 457 |
+
"rewards/accuracies": 0.675000011920929,
|
| 458 |
+
"rewards/chosen": -0.8641554117202759,
|
| 459 |
+
"rewards/margins": 0.6744648814201355,
|
| 460 |
+
"rewards/rejected": -1.5386203527450562,
|
| 461 |
+
"step": 270
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.5857740585774058,
|
| 465 |
+
"grad_norm": 25.076270854562903,
|
| 466 |
+
"learning_rate": 2.19029145890313e-07,
|
| 467 |
+
"logits/chosen": 0.5906327962875366,
|
| 468 |
+
"logits/rejected": 1.5835926532745361,
|
| 469 |
+
"logps/chosen": -364.9591369628906,
|
| 470 |
+
"logps/rejected": -425.6160583496094,
|
| 471 |
+
"loss": 0.53,
|
| 472 |
+
"rewards/accuracies": 0.699999988079071,
|
| 473 |
+
"rewards/chosen": -1.0323399305343628,
|
| 474 |
+
"rewards/margins": 0.8016790151596069,
|
| 475 |
+
"rewards/rejected": -1.8340189456939697,
|
| 476 |
+
"step": 280
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 0.606694560669456,
|
| 480 |
+
"grad_norm": 26.744088332600196,
|
| 481 |
+
"learning_rate": 2.0100351342479216e-07,
|
| 482 |
+
"logits/chosen": 0.8191035985946655,
|
| 483 |
+
"logits/rejected": 1.6216096878051758,
|
| 484 |
+
"logps/chosen": -373.52532958984375,
|
| 485 |
+
"logps/rejected": -437.02728271484375,
|
| 486 |
+
"loss": 0.5156,
|
| 487 |
+
"rewards/accuracies": 0.737500011920929,
|
| 488 |
+
"rewards/chosen": -1.2243343591690063,
|
| 489 |
+
"rewards/margins": 0.7263683080673218,
|
| 490 |
+
"rewards/rejected": -1.9507026672363281,
|
| 491 |
+
"step": 290
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.6276150627615062,
|
| 495 |
+
"grad_norm": 26.465017968091658,
|
| 496 |
+
"learning_rate": 1.8323929841460178e-07,
|
| 497 |
+
"logits/chosen": 0.5103687047958374,
|
| 498 |
+
"logits/rejected": 1.6966192722320557,
|
| 499 |
+
"logps/chosen": -411.9766540527344,
|
| 500 |
+
"logps/rejected": -446.80499267578125,
|
| 501 |
+
"loss": 0.4955,
|
| 502 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 503 |
+
"rewards/chosen": -1.1094756126403809,
|
| 504 |
+
"rewards/margins": 0.7578223943710327,
|
| 505 |
+
"rewards/rejected": -1.867297887802124,
|
| 506 |
+
"step": 300
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 0.6276150627615062,
|
| 510 |
+
"eval_logits/chosen": 0.4705168604850769,
|
| 511 |
+
"eval_logits/rejected": 1.2340543270111084,
|
| 512 |
+
"eval_logps/chosen": -359.1224060058594,
|
| 513 |
+
"eval_logps/rejected": -449.2220153808594,
|
| 514 |
+
"eval_loss": 0.5091525912284851,
|
| 515 |
+
"eval_rewards/accuracies": 0.765625,
|
| 516 |
+
"eval_rewards/chosen": -0.9649244546890259,
|
| 517 |
+
"eval_rewards/margins": 0.9006747603416443,
|
| 518 |
+
"eval_rewards/rejected": -1.865599274635315,
|
| 519 |
+
"eval_runtime": 113.4568,
|
| 520 |
+
"eval_samples_per_second": 17.628,
|
| 521 |
+
"eval_steps_per_second": 0.282,
|
| 522 |
+
"step": 300
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 0.6485355648535565,
|
| 526 |
+
"grad_norm": 29.87262228525191,
|
| 527 |
+
"learning_rate": 1.6583128063291573e-07,
|
| 528 |
+
"logits/chosen": 0.7746948003768921,
|
| 529 |
+
"logits/rejected": 1.4845497608184814,
|
| 530 |
+
"logps/chosen": -408.56512451171875,
|
| 531 |
+
"logps/rejected": -445.75543212890625,
|
| 532 |
+
"loss": 0.4961,
|
| 533 |
+
"rewards/accuracies": 0.75,
|
| 534 |
+
"rewards/chosen": -1.1326206922531128,
|
| 535 |
+
"rewards/margins": 0.7399401664733887,
|
| 536 |
+
"rewards/rejected": -1.8725608587265015,
|
| 537 |
+
"step": 310
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.6694560669456067,
|
| 541 |
+
"grad_norm": 31.16422293780737,
|
| 542 |
+
"learning_rate": 1.488723393865766e-07,
|
| 543 |
+
"logits/chosen": 1.0418922901153564,
|
| 544 |
+
"logits/rejected": 2.0347073078155518,
|
| 545 |
+
"logps/chosen": -416.1197204589844,
|
| 546 |
+
"logps/rejected": -444.3402404785156,
|
| 547 |
+
"loss": 0.49,
|
| 548 |
+
"rewards/accuracies": 0.75,
|
| 549 |
+
"rewards/chosen": -1.2240684032440186,
|
| 550 |
+
"rewards/margins": 0.7707505226135254,
|
| 551 |
+
"rewards/rejected": -1.9948190450668335,
|
| 552 |
+
"step": 320
|
| 553 |
+
},
|
| 554 |
+
{
|
| 555 |
+
"epoch": 0.6903765690376569,
|
| 556 |
+
"grad_norm": 32.83677496957721,
|
| 557 |
+
"learning_rate": 1.3245295796480788e-07,
|
| 558 |
+
"logits/chosen": 0.1572292298078537,
|
| 559 |
+
"logits/rejected": 1.1684186458587646,
|
| 560 |
+
"logps/chosen": -365.3023986816406,
|
| 561 |
+
"logps/rejected": -434.7137145996094,
|
| 562 |
+
"loss": 0.4983,
|
| 563 |
+
"rewards/accuracies": 0.6875,
|
| 564 |
+
"rewards/chosen": -1.0306357145309448,
|
| 565 |
+
"rewards/margins": 0.7212126851081848,
|
| 566 |
+
"rewards/rejected": -1.7518484592437744,
|
| 567 |
+
"step": 330
|
| 568 |
+
},
|
| 569 |
+
{
|
| 570 |
+
"epoch": 0.7112970711297071,
|
| 571 |
+
"grad_norm": 34.84216926965088,
|
| 572 |
+
"learning_rate": 1.1666074087171627e-07,
|
| 573 |
+
"logits/chosen": 0.42324066162109375,
|
| 574 |
+
"logits/rejected": 1.437130093574524,
|
| 575 |
+
"logps/chosen": -386.4037170410156,
|
| 576 |
+
"logps/rejected": -480.5757751464844,
|
| 577 |
+
"loss": 0.5021,
|
| 578 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 579 |
+
"rewards/chosen": -0.963259220123291,
|
| 580 |
+
"rewards/margins": 1.04145348072052,
|
| 581 |
+
"rewards/rejected": -2.0047125816345215,
|
| 582 |
+
"step": 340
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"epoch": 0.7322175732217573,
|
| 586 |
+
"grad_norm": 26.071760002983055,
|
| 587 |
+
"learning_rate": 1.0157994641835734e-07,
|
| 588 |
+
"logits/chosen": 0.6516727209091187,
|
| 589 |
+
"logits/rejected": 1.5883102416992188,
|
| 590 |
+
"logps/chosen": -368.0363464355469,
|
| 591 |
+
"logps/rejected": -438.5006408691406,
|
| 592 |
+
"loss": 0.4783,
|
| 593 |
+
"rewards/accuracies": 0.75,
|
| 594 |
+
"rewards/chosen": -1.098336935043335,
|
| 595 |
+
"rewards/margins": 0.8796666264533997,
|
| 596 |
+
"rewards/rejected": -1.9780035018920898,
|
| 597 |
+
"step": 350
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 0.7531380753138075,
|
| 601 |
+
"grad_norm": 26.502609026351305,
|
| 602 |
+
"learning_rate": 8.729103716819111e-08,
|
| 603 |
+
"logits/chosen": 0.5831906199455261,
|
| 604 |
+
"logits/rejected": 1.8786232471466064,
|
| 605 |
+
"logps/chosen": -415.86541748046875,
|
| 606 |
+
"logps/rejected": -466.9419860839844,
|
| 607 |
+
"loss": 0.5141,
|
| 608 |
+
"rewards/accuracies": 0.762499988079071,
|
| 609 |
+
"rewards/chosen": -1.1316957473754883,
|
| 610 |
+
"rewards/margins": 0.9251440763473511,
|
| 611 |
+
"rewards/rejected": -2.056839942932129,
|
| 612 |
+
"step": 360
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 0.7740585774058577,
|
| 616 |
+
"grad_norm": 28.539635727706617,
|
| 617 |
+
"learning_rate": 7.387025063449081e-08,
|
| 618 |
+
"logits/chosen": 1.116403341293335,
|
| 619 |
+
"logits/rejected": 1.9285640716552734,
|
| 620 |
+
"logps/chosen": -403.40655517578125,
|
| 621 |
+
"logps/rejected": -438.885498046875,
|
| 622 |
+
"loss": 0.5091,
|
| 623 |
+
"rewards/accuracies": 0.706250011920929,
|
| 624 |
+
"rewards/chosen": -1.280707597732544,
|
| 625 |
+
"rewards/margins": 0.7252805829048157,
|
| 626 |
+
"rewards/rejected": -2.005988121032715,
|
| 627 |
+
"step": 370
|
| 628 |
+
},
|
| 629 |
+
{
|
| 630 |
+
"epoch": 0.7949790794979079,
|
| 631 |
+
"grad_norm": 26.46363151511675,
|
| 632 |
+
"learning_rate": 6.138919252022435e-08,
|
| 633 |
+
"logits/chosen": 1.3486121892929077,
|
| 634 |
+
"logits/rejected": 2.0518882274627686,
|
| 635 |
+
"logps/chosen": -381.86395263671875,
|
| 636 |
+
"logps/rejected": -493.0929260253906,
|
| 637 |
+
"loss": 0.5042,
|
| 638 |
+
"rewards/accuracies": 0.762499988079071,
|
| 639 |
+
"rewards/chosen": -1.423814058303833,
|
| 640 |
+
"rewards/margins": 0.9205785989761353,
|
| 641 |
+
"rewards/rejected": -2.344392776489258,
|
| 642 |
+
"step": 380
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 0.8158995815899581,
|
| 646 |
+
"grad_norm": 23.582088705423008,
|
| 647 |
+
"learning_rate": 4.991445467064689e-08,
|
| 648 |
+
"logits/chosen": 0.8774948120117188,
|
| 649 |
+
"logits/rejected": 1.5052356719970703,
|
| 650 |
+
"logps/chosen": -426.742919921875,
|
| 651 |
+
"logps/rejected": -485.6712341308594,
|
| 652 |
+
"loss": 0.4942,
|
| 653 |
+
"rewards/accuracies": 0.706250011920929,
|
| 654 |
+
"rewards/chosen": -1.211219310760498,
|
| 655 |
+
"rewards/margins": 0.8056495785713196,
|
| 656 |
+
"rewards/rejected": -2.016869068145752,
|
| 657 |
+
"step": 390
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 0.8368200836820083,
|
| 661 |
+
"grad_norm": 25.52262025467119,
|
| 662 |
+
"learning_rate": 3.9507259776993954e-08,
|
| 663 |
+
"logits/chosen": 0.8794261813163757,
|
| 664 |
+
"logits/rejected": 1.9908300638198853,
|
| 665 |
+
"logps/chosen": -388.5022277832031,
|
| 666 |
+
"logps/rejected": -473.42755126953125,
|
| 667 |
+
"loss": 0.5033,
|
| 668 |
+
"rewards/accuracies": 0.75,
|
| 669 |
+
"rewards/chosen": -1.17954683303833,
|
| 670 |
+
"rewards/margins": 0.8852158784866333,
|
| 671 |
+
"rewards/rejected": -2.064762592315674,
|
| 672 |
+
"step": 400
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"epoch": 0.8368200836820083,
|
| 676 |
+
"eval_logits/chosen": 0.5144522190093994,
|
| 677 |
+
"eval_logits/rejected": 1.390769600868225,
|
| 678 |
+
"eval_logps/chosen": -362.76513671875,
|
| 679 |
+
"eval_logps/rejected": -458.7183837890625,
|
| 680 |
+
"eval_loss": 0.5032494068145752,
|
| 681 |
+
"eval_rewards/accuracies": 0.7734375,
|
| 682 |
+
"eval_rewards/chosen": -1.0013517141342163,
|
| 683 |
+
"eval_rewards/margins": 0.9592112302780151,
|
| 684 |
+
"eval_rewards/rejected": -1.9605629444122314,
|
| 685 |
+
"eval_runtime": 104.6004,
|
| 686 |
+
"eval_samples_per_second": 19.12,
|
| 687 |
+
"eval_steps_per_second": 0.306,
|
| 688 |
+
"step": 400
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"epoch": 0.8577405857740585,
|
| 692 |
+
"grad_norm": 24.81847546133683,
|
| 693 |
+
"learning_rate": 3.022313472693447e-08,
|
| 694 |
+
"logits/chosen": 0.6560511589050293,
|
| 695 |
+
"logits/rejected": 1.8641235828399658,
|
| 696 |
+
"logps/chosen": -406.5023193359375,
|
| 697 |
+
"logps/rejected": -463.3063049316406,
|
| 698 |
+
"loss": 0.5077,
|
| 699 |
+
"rewards/accuracies": 0.768750011920929,
|
| 700 |
+
"rewards/chosen": -1.0756852626800537,
|
| 701 |
+
"rewards/margins": 0.8691245913505554,
|
| 702 |
+
"rewards/rejected": -1.9448096752166748,
|
| 703 |
+
"step": 410
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.8786610878661087,
|
| 707 |
+
"grad_norm": 25.68538032885617,
|
| 708 |
+
"learning_rate": 2.2111614344599684e-08,
|
| 709 |
+
"logits/chosen": 0.39021772146224976,
|
| 710 |
+
"logits/rejected": 1.3714463710784912,
|
| 711 |
+
"logps/chosen": -403.7474670410156,
|
| 712 |
+
"logps/rejected": -460.71478271484375,
|
| 713 |
+
"loss": 0.4954,
|
| 714 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 715 |
+
"rewards/chosen": -1.066702961921692,
|
| 716 |
+
"rewards/margins": 0.7960401773452759,
|
| 717 |
+
"rewards/rejected": -1.8627431392669678,
|
| 718 |
+
"step": 420
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"epoch": 0.899581589958159,
|
| 722 |
+
"grad_norm": 28.26563566454169,
|
| 723 |
+
"learning_rate": 1.521597710086439e-08,
|
| 724 |
+
"logits/chosen": 0.8775579333305359,
|
| 725 |
+
"logits/rejected": 1.8475208282470703,
|
| 726 |
+
"logps/chosen": -390.8741149902344,
|
| 727 |
+
"logps/rejected": -458.987060546875,
|
| 728 |
+
"loss": 0.4745,
|
| 729 |
+
"rewards/accuracies": 0.768750011920929,
|
| 730 |
+
"rewards/chosen": -1.1183277368545532,
|
| 731 |
+
"rewards/margins": 0.8938654661178589,
|
| 732 |
+
"rewards/rejected": -2.012192964553833,
|
| 733 |
+
"step": 430
|
| 734 |
+
},
|
| 735 |
+
{
|
| 736 |
+
"epoch": 0.9205020920502092,
|
| 737 |
+
"grad_norm": 28.547233961480817,
|
| 738 |
+
"learning_rate": 9.57301420397924e-09,
|
| 739 |
+
"logits/chosen": 0.518905520439148,
|
| 740 |
+
"logits/rejected": 1.5218467712402344,
|
| 741 |
+
"logps/chosen": -385.42559814453125,
|
| 742 |
+
"logps/rejected": -450.37603759765625,
|
| 743 |
+
"loss": 0.5019,
|
| 744 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 745 |
+
"rewards/chosen": -1.0109506845474243,
|
| 746 |
+
"rewards/margins": 0.8017943501472473,
|
| 747 |
+
"rewards/rejected": -1.8127450942993164,
|
| 748 |
+
"step": 440
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"epoch": 0.9414225941422594,
|
| 752 |
+
"grad_norm": 23.78987065435005,
|
| 753 |
+
"learning_rate": 5.212833302556258e-09,
|
| 754 |
+
"logits/chosen": 0.5571370124816895,
|
| 755 |
+
"logits/rejected": 1.4383113384246826,
|
| 756 |
+
"logps/chosen": -402.0091552734375,
|
| 757 |
+
"logps/rejected": -497.11376953125,
|
| 758 |
+
"loss": 0.4972,
|
| 759 |
+
"rewards/accuracies": 0.71875,
|
| 760 |
+
"rewards/chosen": -1.0828384160995483,
|
| 761 |
+
"rewards/margins": 0.8171746134757996,
|
| 762 |
+
"rewards/rejected": -1.9000129699707031,
|
| 763 |
+
"step": 450
|
| 764 |
+
},
|
| 765 |
+
{
|
| 766 |
+
"epoch": 0.9623430962343096,
|
| 767 |
+
"grad_norm": 25.601990523329984,
|
| 768 |
+
"learning_rate": 2.158697848236607e-09,
|
| 769 |
+
"logits/chosen": 0.730278491973877,
|
| 770 |
+
"logits/rejected": 1.7400062084197998,
|
| 771 |
+
"logps/chosen": -374.495361328125,
|
| 772 |
+
"logps/rejected": -426.7510681152344,
|
| 773 |
+
"loss": 0.5018,
|
| 774 |
+
"rewards/accuracies": 0.731249988079071,
|
| 775 |
+
"rewards/chosen": -1.025061011314392,
|
| 776 |
+
"rewards/margins": 0.8159972429275513,
|
| 777 |
+
"rewards/rejected": -1.8410581350326538,
|
| 778 |
+
"step": 460
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"epoch": 0.9832635983263598,
|
| 782 |
+
"grad_norm": 25.12747739082461,
|
| 783 |
+
"learning_rate": 4.269029751107489e-10,
|
| 784 |
+
"logits/chosen": 0.5695275068283081,
|
| 785 |
+
"logits/rejected": 1.7241246700286865,
|
| 786 |
+
"logps/chosen": -380.4364013671875,
|
| 787 |
+
"logps/rejected": -468.3048400878906,
|
| 788 |
+
"loss": 0.4924,
|
| 789 |
+
"rewards/accuracies": 0.737500011920929,
|
| 790 |
+
"rewards/chosen": -1.0173852443695068,
|
| 791 |
+
"rewards/margins": 0.8419061899185181,
|
| 792 |
+
"rewards/rejected": -1.859291434288025,
|
| 793 |
+
"step": 470
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"epoch": 1.0,
|
| 797 |
+
"step": 478,
|
| 798 |
+
"total_flos": 0.0,
|
| 799 |
+
"train_loss": 0.5406385645208,
|
| 800 |
+
"train_runtime": 10904.9554,
|
| 801 |
+
"train_samples_per_second": 5.606,
|
| 802 |
+
"train_steps_per_second": 0.044
|
| 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 |
+
}
|