Model save
Browse files- README.md +80 -0
- all_results.json +9 -0
- generation_config.json +6 -0
- train_results.json +9 -0
- trainer_state.json +815 -0
README.md
ADDED
|
@@ -0,0 +1,80 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 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-gpt_consistent-reward-scale-1
|
| 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-gpt_consistent-reward-scale-1
|
| 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 the None dataset.
|
| 19 |
+
It achieves the following results on the evaluation set:
|
| 20 |
+
- Loss: 0.4815
|
| 21 |
+
- Rewards/chosen: -1.5849
|
| 22 |
+
- Rewards/rejected: -2.8045
|
| 23 |
+
- Rewards/accuracies: 0.7328
|
| 24 |
+
- Rewards/margins: 1.2196
|
| 25 |
+
- Logps/rejected: -526.9686
|
| 26 |
+
- Logps/chosen: -443.5758
|
| 27 |
+
- Logits/rejected: 3.4838
|
| 28 |
+
- Logits/chosen: 2.3333
|
| 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: 55
|
| 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 |
+
| 0.6607 | 0.1147 | 50 | 0.6447 | -0.0044 | -0.1452 | 0.6897 | 0.1408 | -261.0398 | -285.5275 | -2.4923 | -2.5735 |
|
| 66 |
+
| 0.5616 | 0.2294 | 100 | 0.5464 | -0.8527 | -1.5772 | 0.6853 | 0.7245 | -404.2408 | -370.3625 | 0.2075 | -0.2410 |
|
| 67 |
+
| 0.5333 | 0.3440 | 150 | 0.5195 | -1.0024 | -1.8820 | 0.7112 | 0.8797 | -434.7274 | -385.3255 | 1.4808 | 0.5890 |
|
| 68 |
+
| 0.5219 | 0.4587 | 200 | 0.5010 | -1.0719 | -2.0541 | 0.7328 | 0.9822 | -451.9354 | -392.2838 | 2.4260 | 1.4256 |
|
| 69 |
+
| 0.5007 | 0.5734 | 250 | 0.4917 | -1.2321 | -2.3291 | 0.7241 | 1.0970 | -479.4298 | -408.2994 | 2.6738 | 1.4527 |
|
| 70 |
+
| 0.5109 | 0.6881 | 300 | 0.4878 | -1.3356 | -2.5048 | 0.7284 | 1.1691 | -496.9991 | -418.6534 | 2.8884 | 1.5762 |
|
| 71 |
+
| 0.5063 | 0.8028 | 350 | 0.4814 | -1.4870 | -2.6833 | 0.7371 | 1.1963 | -514.8549 | -433.7904 | 3.3469 | 2.1699 |
|
| 72 |
+
| 0.4936 | 0.9174 | 400 | 0.4815 | -1.5849 | -2.8045 | 0.7328 | 1.2196 | -526.9686 | -443.5758 | 3.4838 | 2.3333 |
|
| 73 |
+
|
| 74 |
+
|
| 75 |
+
### Framework versions
|
| 76 |
+
|
| 77 |
+
- Transformers 4.44.0.dev0
|
| 78 |
+
- Pytorch 2.1.2
|
| 79 |
+
- Datasets 2.20.0
|
| 80 |
+
- 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": 0.5380998532706445,
|
| 5 |
+
"train_runtime": 11735.5354,
|
| 6 |
+
"train_samples": 55758,
|
| 7 |
+
"train_samples_per_second": 4.751,
|
| 8 |
+
"train_steps_per_second": 0.037
|
| 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.dev0"
|
| 6 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5380998532706445,
|
| 5 |
+
"train_runtime": 11735.5354,
|
| 6 |
+
"train_samples": 55758,
|
| 7 |
+
"train_samples_per_second": 4.751,
|
| 8 |
+
"train_steps_per_second": 0.037
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,815 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.0,
|
| 5 |
+
"eval_steps": 50,
|
| 6 |
+
"global_step": 436,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.022935779816513763,
|
| 13 |
+
"grad_norm": 6.666838198680302,
|
| 14 |
+
"learning_rate": 1.1363636363636363e-07,
|
| 15 |
+
"logits/chosen": -2.6193127632141113,
|
| 16 |
+
"logits/rejected": -2.5524516105651855,
|
| 17 |
+
"logps/chosen": -265.4483337402344,
|
| 18 |
+
"logps/rejected": -236.09951782226562,
|
| 19 |
+
"loss": 0.6931,
|
| 20 |
+
"rewards/accuracies": 0.35624998807907104,
|
| 21 |
+
"rewards/chosen": -2.520665293559432e-05,
|
| 22 |
+
"rewards/margins": -0.0003298574883956462,
|
| 23 |
+
"rewards/rejected": 0.00030465080635622144,
|
| 24 |
+
"step": 10
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.045871559633027525,
|
| 28 |
+
"grad_norm": 6.256434066153107,
|
| 29 |
+
"learning_rate": 2.2727272727272726e-07,
|
| 30 |
+
"logits/chosen": -2.6579253673553467,
|
| 31 |
+
"logits/rejected": -2.5760178565979004,
|
| 32 |
+
"logps/chosen": -298.82305908203125,
|
| 33 |
+
"logps/rejected": -274.31829833984375,
|
| 34 |
+
"loss": 0.6922,
|
| 35 |
+
"rewards/accuracies": 0.612500011920929,
|
| 36 |
+
"rewards/chosen": -0.0003779557882808149,
|
| 37 |
+
"rewards/margins": 0.0017056630458682775,
|
| 38 |
+
"rewards/rejected": -0.0020836188923567533,
|
| 39 |
+
"step": 20
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.06880733944954129,
|
| 43 |
+
"grad_norm": 6.014004575787888,
|
| 44 |
+
"learning_rate": 3.4090909090909085e-07,
|
| 45 |
+
"logits/chosen": -2.675771713256836,
|
| 46 |
+
"logits/rejected": -2.6020450592041016,
|
| 47 |
+
"logps/chosen": -290.2829895019531,
|
| 48 |
+
"logps/rejected": -234.431396484375,
|
| 49 |
+
"loss": 0.6883,
|
| 50 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 51 |
+
"rewards/chosen": 0.01049796212464571,
|
| 52 |
+
"rewards/margins": 0.01499337237328291,
|
| 53 |
+
"rewards/rejected": -0.004495412111282349,
|
| 54 |
+
"step": 30
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.09174311926605505,
|
| 58 |
+
"grad_norm": 6.545633202821474,
|
| 59 |
+
"learning_rate": 4.545454545454545e-07,
|
| 60 |
+
"logits/chosen": -2.6612305641174316,
|
| 61 |
+
"logits/rejected": -2.6114158630371094,
|
| 62 |
+
"logps/chosen": -280.93109130859375,
|
| 63 |
+
"logps/rejected": -267.76605224609375,
|
| 64 |
+
"loss": 0.6776,
|
| 65 |
+
"rewards/accuracies": 0.668749988079071,
|
| 66 |
+
"rewards/chosen": 0.04238061234354973,
|
| 67 |
+
"rewards/margins": 0.04211103171110153,
|
| 68 |
+
"rewards/rejected": 0.00026957987574860454,
|
| 69 |
+
"step": 40
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.11467889908256881,
|
| 73 |
+
"grad_norm": 7.0518705334996215,
|
| 74 |
+
"learning_rate": 4.997110275491701e-07,
|
| 75 |
+
"logits/chosen": -2.6194958686828613,
|
| 76 |
+
"logits/rejected": -2.6111502647399902,
|
| 77 |
+
"logps/chosen": -294.24957275390625,
|
| 78 |
+
"logps/rejected": -305.2260437011719,
|
| 79 |
+
"loss": 0.6607,
|
| 80 |
+
"rewards/accuracies": 0.706250011920929,
|
| 81 |
+
"rewards/chosen": 0.020016254857182503,
|
| 82 |
+
"rewards/margins": 0.08043506741523743,
|
| 83 |
+
"rewards/rejected": -0.060418806970119476,
|
| 84 |
+
"step": 50
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.11467889908256881,
|
| 88 |
+
"eval_logits/chosen": -2.5735373497009277,
|
| 89 |
+
"eval_logits/rejected": -2.4922757148742676,
|
| 90 |
+
"eval_logps/chosen": -285.52752685546875,
|
| 91 |
+
"eval_logps/rejected": -261.0398254394531,
|
| 92 |
+
"eval_loss": 0.6447442770004272,
|
| 93 |
+
"eval_rewards/accuracies": 0.6896551847457886,
|
| 94 |
+
"eval_rewards/chosen": -0.004374094773083925,
|
| 95 |
+
"eval_rewards/margins": 0.140799880027771,
|
| 96 |
+
"eval_rewards/rejected": -0.14517399668693542,
|
| 97 |
+
"eval_runtime": 94.492,
|
| 98 |
+
"eval_samples_per_second": 19.24,
|
| 99 |
+
"eval_steps_per_second": 0.307,
|
| 100 |
+
"step": 50
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.13761467889908258,
|
| 104 |
+
"grad_norm": 10.342302388199876,
|
| 105 |
+
"learning_rate": 4.979475034558115e-07,
|
| 106 |
+
"logits/chosen": -2.559204578399658,
|
| 107 |
+
"logits/rejected": -2.5000882148742676,
|
| 108 |
+
"logps/chosen": -296.8924255371094,
|
| 109 |
+
"logps/rejected": -279.6778869628906,
|
| 110 |
+
"loss": 0.6362,
|
| 111 |
+
"rewards/accuracies": 0.668749988079071,
|
| 112 |
+
"rewards/chosen": -0.07407740503549576,
|
| 113 |
+
"rewards/margins": 0.14713598787784576,
|
| 114 |
+
"rewards/rejected": -0.22121338546276093,
|
| 115 |
+
"step": 60
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"epoch": 0.16055045871559634,
|
| 119 |
+
"grad_norm": 11.57375336560156,
|
| 120 |
+
"learning_rate": 4.945923025551788e-07,
|
| 121 |
+
"logits/chosen": -2.439746141433716,
|
| 122 |
+
"logits/rejected": -2.39266037940979,
|
| 123 |
+
"logps/chosen": -341.58990478515625,
|
| 124 |
+
"logps/rejected": -305.513427734375,
|
| 125 |
+
"loss": 0.614,
|
| 126 |
+
"rewards/accuracies": 0.706250011920929,
|
| 127 |
+
"rewards/chosen": -0.14018920063972473,
|
| 128 |
+
"rewards/margins": 0.3008686304092407,
|
| 129 |
+
"rewards/rejected": -0.44105786085128784,
|
| 130 |
+
"step": 70
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 0.1834862385321101,
|
| 134 |
+
"grad_norm": 13.101676053286978,
|
| 135 |
+
"learning_rate": 4.896669632591651e-07,
|
| 136 |
+
"logits/chosen": -2.333707332611084,
|
| 137 |
+
"logits/rejected": -2.2206273078918457,
|
| 138 |
+
"logps/chosen": -311.94189453125,
|
| 139 |
+
"logps/rejected": -312.6253662109375,
|
| 140 |
+
"loss": 0.5993,
|
| 141 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 142 |
+
"rewards/chosen": -0.2609206438064575,
|
| 143 |
+
"rewards/margins": 0.33877214789390564,
|
| 144 |
+
"rewards/rejected": -0.5996928215026855,
|
| 145 |
+
"step": 80
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 0.20642201834862386,
|
| 149 |
+
"grad_norm": 18.464805323764754,
|
| 150 |
+
"learning_rate": 4.832031033425662e-07,
|
| 151 |
+
"logits/chosen": -1.2291042804718018,
|
| 152 |
+
"logits/rejected": -1.0986605882644653,
|
| 153 |
+
"logps/chosen": -349.7447814941406,
|
| 154 |
+
"logps/rejected": -355.8329162597656,
|
| 155 |
+
"loss": 0.5841,
|
| 156 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 157 |
+
"rewards/chosen": -0.6215134859085083,
|
| 158 |
+
"rewards/margins": 0.4688030779361725,
|
| 159 |
+
"rewards/rejected": -1.0903165340423584,
|
| 160 |
+
"step": 90
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 0.22935779816513763,
|
| 164 |
+
"grad_norm": 18.44194365148455,
|
| 165 |
+
"learning_rate": 4.752422169756047e-07,
|
| 166 |
+
"logits/chosen": -0.5803453326225281,
|
| 167 |
+
"logits/rejected": -0.4585650861263275,
|
| 168 |
+
"logps/chosen": -330.8556823730469,
|
| 169 |
+
"logps/rejected": -383.45281982421875,
|
| 170 |
+
"loss": 0.5616,
|
| 171 |
+
"rewards/accuracies": 0.706250011920929,
|
| 172 |
+
"rewards/chosen": -0.5843100547790527,
|
| 173 |
+
"rewards/margins": 0.5309332609176636,
|
| 174 |
+
"rewards/rejected": -1.1152434349060059,
|
| 175 |
+
"step": 100
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.22935779816513763,
|
| 179 |
+
"eval_logits/chosen": -0.241001695394516,
|
| 180 |
+
"eval_logits/rejected": 0.20754042267799377,
|
| 181 |
+
"eval_logps/chosen": -370.36248779296875,
|
| 182 |
+
"eval_logps/rejected": -404.2408142089844,
|
| 183 |
+
"eval_loss": 0.5463992357254028,
|
| 184 |
+
"eval_rewards/accuracies": 0.6853448152542114,
|
| 185 |
+
"eval_rewards/chosen": -0.8527240753173828,
|
| 186 |
+
"eval_rewards/margins": 0.7244595289230347,
|
| 187 |
+
"eval_rewards/rejected": -1.5771836042404175,
|
| 188 |
+
"eval_runtime": 94.4096,
|
| 189 |
+
"eval_samples_per_second": 19.257,
|
| 190 |
+
"eval_steps_per_second": 0.307,
|
| 191 |
+
"step": 100
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.25229357798165136,
|
| 195 |
+
"grad_norm": 20.029871948259476,
|
| 196 |
+
"learning_rate": 4.658354083558188e-07,
|
| 197 |
+
"logits/chosen": -0.10621030628681183,
|
| 198 |
+
"logits/rejected": 0.371669203042984,
|
| 199 |
+
"logps/chosen": -369.4322204589844,
|
| 200 |
+
"logps/rejected": -397.5771179199219,
|
| 201 |
+
"loss": 0.5451,
|
| 202 |
+
"rewards/accuracies": 0.71875,
|
| 203 |
+
"rewards/chosen": -1.0424408912658691,
|
| 204 |
+
"rewards/margins": 0.5959984660148621,
|
| 205 |
+
"rewards/rejected": -1.638439416885376,
|
| 206 |
+
"step": 110
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.27522935779816515,
|
| 210 |
+
"grad_norm": 25.325536766276052,
|
| 211 |
+
"learning_rate": 4.550430636492389e-07,
|
| 212 |
+
"logits/chosen": 0.43674445152282715,
|
| 213 |
+
"logits/rejected": 0.7346484065055847,
|
| 214 |
+
"logps/chosen": -387.1654968261719,
|
| 215 |
+
"logps/rejected": -426.51898193359375,
|
| 216 |
+
"loss": 0.5604,
|
| 217 |
+
"rewards/accuracies": 0.6875,
|
| 218 |
+
"rewards/chosen": -1.1134614944458008,
|
| 219 |
+
"rewards/margins": 0.6370891332626343,
|
| 220 |
+
"rewards/rejected": -1.7505505084991455,
|
| 221 |
+
"step": 120
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.2981651376146789,
|
| 225 |
+
"grad_norm": 25.1538794496433,
|
| 226 |
+
"learning_rate": 4.429344633468004e-07,
|
| 227 |
+
"logits/chosen": 0.7902742624282837,
|
| 228 |
+
"logits/rejected": 1.2213919162750244,
|
| 229 |
+
"logps/chosen": -351.0482482910156,
|
| 230 |
+
"logps/rejected": -406.3229675292969,
|
| 231 |
+
"loss": 0.5509,
|
| 232 |
+
"rewards/accuracies": 0.699999988079071,
|
| 233 |
+
"rewards/chosen": -0.9486812353134155,
|
| 234 |
+
"rewards/margins": 0.6884225606918335,
|
| 235 |
+
"rewards/rejected": -1.6371036767959595,
|
| 236 |
+
"step": 130
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 0.3211009174311927,
|
| 240 |
+
"grad_norm": 21.483029077159443,
|
| 241 |
+
"learning_rate": 4.2958733752443187e-07,
|
| 242 |
+
"logits/chosen": 0.7886204123497009,
|
| 243 |
+
"logits/rejected": 1.4716787338256836,
|
| 244 |
+
"logps/chosen": -352.66986083984375,
|
| 245 |
+
"logps/rejected": -371.2418212890625,
|
| 246 |
+
"loss": 0.5467,
|
| 247 |
+
"rewards/accuracies": 0.731249988079071,
|
| 248 |
+
"rewards/chosen": -0.764846920967102,
|
| 249 |
+
"rewards/margins": 0.704605758190155,
|
| 250 |
+
"rewards/rejected": -1.4694526195526123,
|
| 251 |
+
"step": 140
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 0.3440366972477064,
|
| 255 |
+
"grad_norm": 21.063627942952767,
|
| 256 |
+
"learning_rate": 4.150873668617898e-07,
|
| 257 |
+
"logits/chosen": 1.314632534980774,
|
| 258 |
+
"logits/rejected": 2.00125789642334,
|
| 259 |
+
"logps/chosen": -385.40447998046875,
|
| 260 |
+
"logps/rejected": -437.3175354003906,
|
| 261 |
+
"loss": 0.5333,
|
| 262 |
+
"rewards/accuracies": 0.71875,
|
| 263 |
+
"rewards/chosen": -1.1886088848114014,
|
| 264 |
+
"rewards/margins": 0.8526908755302429,
|
| 265 |
+
"rewards/rejected": -2.04129958152771,
|
| 266 |
+
"step": 150
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 0.3440366972477064,
|
| 270 |
+
"eval_logits/chosen": 0.5890395045280457,
|
| 271 |
+
"eval_logits/rejected": 1.4808454513549805,
|
| 272 |
+
"eval_logps/chosen": -385.3254699707031,
|
| 273 |
+
"eval_logps/rejected": -434.7274169921875,
|
| 274 |
+
"eval_loss": 0.5195103883743286,
|
| 275 |
+
"eval_rewards/accuracies": 0.7112069129943848,
|
| 276 |
+
"eval_rewards/chosen": -1.0023537874221802,
|
| 277 |
+
"eval_rewards/margins": 0.8796959519386292,
|
| 278 |
+
"eval_rewards/rejected": -1.8820499181747437,
|
| 279 |
+
"eval_runtime": 94.3575,
|
| 280 |
+
"eval_samples_per_second": 19.267,
|
| 281 |
+
"eval_steps_per_second": 0.307,
|
| 282 |
+
"step": 150
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.3669724770642202,
|
| 286 |
+
"grad_norm": 27.196294375049977,
|
| 287 |
+
"learning_rate": 3.9952763262280397e-07,
|
| 288 |
+
"logits/chosen": 0.8305681943893433,
|
| 289 |
+
"logits/rejected": 1.4138782024383545,
|
| 290 |
+
"logps/chosen": -414.48675537109375,
|
| 291 |
+
"logps/rejected": -488.97027587890625,
|
| 292 |
+
"loss": 0.5169,
|
| 293 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 294 |
+
"rewards/chosen": -1.179356336593628,
|
| 295 |
+
"rewards/margins": 0.8506600260734558,
|
| 296 |
+
"rewards/rejected": -2.0300164222717285,
|
| 297 |
+
"step": 160
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 0.38990825688073394,
|
| 301 |
+
"grad_norm": 20.448285631169824,
|
| 302 |
+
"learning_rate": 3.8300801912883414e-07,
|
| 303 |
+
"logits/chosen": 1.4580880403518677,
|
| 304 |
+
"logits/rejected": 1.959507703781128,
|
| 305 |
+
"logps/chosen": -411.11505126953125,
|
| 306 |
+
"logps/rejected": -499.7334899902344,
|
| 307 |
+
"loss": 0.5017,
|
| 308 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 309 |
+
"rewards/chosen": -1.395141363143921,
|
| 310 |
+
"rewards/margins": 0.9015046954154968,
|
| 311 |
+
"rewards/rejected": -2.2966461181640625,
|
| 312 |
+
"step": 170
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 0.41284403669724773,
|
| 316 |
+
"grad_norm": 19.388105918914484,
|
| 317 |
+
"learning_rate": 3.6563457256020884e-07,
|
| 318 |
+
"logits/chosen": 1.2716890573501587,
|
| 319 |
+
"logits/rejected": 2.1560723781585693,
|
| 320 |
+
"logps/chosen": -418.70068359375,
|
| 321 |
+
"logps/rejected": -435.9356994628906,
|
| 322 |
+
"loss": 0.5267,
|
| 323 |
+
"rewards/accuracies": 0.731249988079071,
|
| 324 |
+
"rewards/chosen": -1.1534276008605957,
|
| 325 |
+
"rewards/margins": 0.8209800720214844,
|
| 326 |
+
"rewards/rejected": -1.9744075536727905,
|
| 327 |
+
"step": 180
|
| 328 |
+
},
|
| 329 |
+
{
|
| 330 |
+
"epoch": 0.43577981651376146,
|
| 331 |
+
"grad_norm": 24.752350050203567,
|
| 332 |
+
"learning_rate": 3.475188202022617e-07,
|
| 333 |
+
"logits/chosen": 1.6485626697540283,
|
| 334 |
+
"logits/rejected": 2.1269421577453613,
|
| 335 |
+
"logps/chosen": -365.346435546875,
|
| 336 |
+
"logps/rejected": -476.385009765625,
|
| 337 |
+
"loss": 0.5115,
|
| 338 |
+
"rewards/accuracies": 0.762499988079071,
|
| 339 |
+
"rewards/chosen": -1.0723192691802979,
|
| 340 |
+
"rewards/margins": 1.048086404800415,
|
| 341 |
+
"rewards/rejected": -2.120405673980713,
|
| 342 |
+
"step": 190
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 0.45871559633027525,
|
| 346 |
+
"grad_norm": 20.591094290068703,
|
| 347 |
+
"learning_rate": 3.287770545059052e-07,
|
| 348 |
+
"logits/chosen": 1.790414810180664,
|
| 349 |
+
"logits/rejected": 2.4998342990875244,
|
| 350 |
+
"logps/chosen": -400.47991943359375,
|
| 351 |
+
"logps/rejected": -456.0809020996094,
|
| 352 |
+
"loss": 0.5219,
|
| 353 |
+
"rewards/accuracies": 0.71875,
|
| 354 |
+
"rewards/chosen": -1.221982717514038,
|
| 355 |
+
"rewards/margins": 0.900266170501709,
|
| 356 |
+
"rewards/rejected": -2.122248888015747,
|
| 357 |
+
"step": 200
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 0.45871559633027525,
|
| 361 |
+
"eval_logits/chosen": 1.4256442785263062,
|
| 362 |
+
"eval_logits/rejected": 2.4260053634643555,
|
| 363 |
+
"eval_logps/chosen": -392.2837829589844,
|
| 364 |
+
"eval_logps/rejected": -451.93536376953125,
|
| 365 |
+
"eval_loss": 0.5009589195251465,
|
| 366 |
+
"eval_rewards/accuracies": 0.732758641242981,
|
| 367 |
+
"eval_rewards/chosen": -1.0719367265701294,
|
| 368 |
+
"eval_rewards/margins": 0.9821925759315491,
|
| 369 |
+
"eval_rewards/rejected": -2.0541296005249023,
|
| 370 |
+
"eval_runtime": 96.4986,
|
| 371 |
+
"eval_samples_per_second": 18.84,
|
| 372 |
+
"eval_steps_per_second": 0.301,
|
| 373 |
+
"step": 200
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 0.481651376146789,
|
| 377 |
+
"grad_norm": 21.25864225686741,
|
| 378 |
+
"learning_rate": 3.0952958655864954e-07,
|
| 379 |
+
"logits/chosen": 1.83803391456604,
|
| 380 |
+
"logits/rejected": 2.399480104446411,
|
| 381 |
+
"logps/chosen": -394.57037353515625,
|
| 382 |
+
"logps/rejected": -468.97021484375,
|
| 383 |
+
"loss": 0.5006,
|
| 384 |
+
"rewards/accuracies": 0.75,
|
| 385 |
+
"rewards/chosen": -1.212659478187561,
|
| 386 |
+
"rewards/margins": 0.9336206316947937,
|
| 387 |
+
"rewards/rejected": -2.14628005027771,
|
| 388 |
+
"step": 210
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 0.5045871559633027,
|
| 392 |
+
"grad_norm": 24.95309197907959,
|
| 393 |
+
"learning_rate": 2.898999737583448e-07,
|
| 394 |
+
"logits/chosen": 1.8371950387954712,
|
| 395 |
+
"logits/rejected": 2.8440494537353516,
|
| 396 |
+
"logps/chosen": -448.06109619140625,
|
| 397 |
+
"logps/rejected": -526.2056884765625,
|
| 398 |
+
"loss": 0.491,
|
| 399 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 400 |
+
"rewards/chosen": -1.2538080215454102,
|
| 401 |
+
"rewards/margins": 1.158546805381775,
|
| 402 |
+
"rewards/rejected": -2.4123549461364746,
|
| 403 |
+
"step": 220
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"epoch": 0.5275229357798165,
|
| 407 |
+
"grad_norm": 24.874258979226397,
|
| 408 |
+
"learning_rate": 2.7001422664752333e-07,
|
| 409 |
+
"logits/chosen": 1.71869695186615,
|
| 410 |
+
"logits/rejected": 2.2633819580078125,
|
| 411 |
+
"logps/chosen": -376.9873046875,
|
| 412 |
+
"logps/rejected": -471.9263610839844,
|
| 413 |
+
"loss": 0.5124,
|
| 414 |
+
"rewards/accuracies": 0.731249988079071,
|
| 415 |
+
"rewards/chosen": -1.0904380083084106,
|
| 416 |
+
"rewards/margins": 0.9726268649101257,
|
| 417 |
+
"rewards/rejected": -2.0630650520324707,
|
| 418 |
+
"step": 230
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"epoch": 0.5504587155963303,
|
| 422 |
+
"grad_norm": 18.86072207414666,
|
| 423 |
+
"learning_rate": 2.5e-07,
|
| 424 |
+
"logits/chosen": 1.5722936391830444,
|
| 425 |
+
"logits/rejected": 2.684537410736084,
|
| 426 |
+
"logps/chosen": -418.7516174316406,
|
| 427 |
+
"logps/rejected": -480.7166442871094,
|
| 428 |
+
"loss": 0.5197,
|
| 429 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 430 |
+
"rewards/chosen": -1.3121980428695679,
|
| 431 |
+
"rewards/margins": 0.8814684152603149,
|
| 432 |
+
"rewards/rejected": -2.193666696548462,
|
| 433 |
+
"step": 240
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.573394495412844,
|
| 437 |
+
"grad_norm": 19.046011375519186,
|
| 438 |
+
"learning_rate": 2.2998577335247667e-07,
|
| 439 |
+
"logits/chosen": 1.6455481052398682,
|
| 440 |
+
"logits/rejected": 2.8547589778900146,
|
| 441 |
+
"logps/chosen": -438.38238525390625,
|
| 442 |
+
"logps/rejected": -492.9747619628906,
|
| 443 |
+
"loss": 0.5007,
|
| 444 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 445 |
+
"rewards/chosen": -1.3458855152130127,
|
| 446 |
+
"rewards/margins": 0.9794231653213501,
|
| 447 |
+
"rewards/rejected": -2.3253085613250732,
|
| 448 |
+
"step": 250
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"epoch": 0.573394495412844,
|
| 452 |
+
"eval_logits/chosen": 1.4527463912963867,
|
| 453 |
+
"eval_logits/rejected": 2.673773765563965,
|
| 454 |
+
"eval_logps/chosen": -408.2994384765625,
|
| 455 |
+
"eval_logps/rejected": -479.4298095703125,
|
| 456 |
+
"eval_loss": 0.4917016923427582,
|
| 457 |
+
"eval_rewards/accuracies": 0.7241379022598267,
|
| 458 |
+
"eval_rewards/chosen": -1.232093095779419,
|
| 459 |
+
"eval_rewards/margins": 1.096980333328247,
|
| 460 |
+
"eval_rewards/rejected": -2.329073429107666,
|
| 461 |
+
"eval_runtime": 95.5197,
|
| 462 |
+
"eval_samples_per_second": 19.033,
|
| 463 |
+
"eval_steps_per_second": 0.304,
|
| 464 |
+
"step": 250
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 0.5963302752293578,
|
| 468 |
+
"grad_norm": 19.810153184401507,
|
| 469 |
+
"learning_rate": 2.1010002624165524e-07,
|
| 470 |
+
"logits/chosen": 1.6456598043441772,
|
| 471 |
+
"logits/rejected": 2.469041347503662,
|
| 472 |
+
"logps/chosen": -406.33642578125,
|
| 473 |
+
"logps/rejected": -532.65869140625,
|
| 474 |
+
"loss": 0.4916,
|
| 475 |
+
"rewards/accuracies": 0.800000011920929,
|
| 476 |
+
"rewards/chosen": -1.356218934059143,
|
| 477 |
+
"rewards/margins": 1.226430892944336,
|
| 478 |
+
"rewards/rejected": -2.5826497077941895,
|
| 479 |
+
"step": 260
|
| 480 |
+
},
|
| 481 |
+
{
|
| 482 |
+
"epoch": 0.6192660550458715,
|
| 483 |
+
"grad_norm": 24.810334002988427,
|
| 484 |
+
"learning_rate": 1.9047041344135043e-07,
|
| 485 |
+
"logits/chosen": 2.157437801361084,
|
| 486 |
+
"logits/rejected": 2.676581621170044,
|
| 487 |
+
"logps/chosen": -414.63427734375,
|
| 488 |
+
"logps/rejected": -517.6594848632812,
|
| 489 |
+
"loss": 0.4963,
|
| 490 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 491 |
+
"rewards/chosen": -1.4857197999954224,
|
| 492 |
+
"rewards/margins": 1.172404408454895,
|
| 493 |
+
"rewards/rejected": -2.6581242084503174,
|
| 494 |
+
"step": 270
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 0.6422018348623854,
|
| 498 |
+
"grad_norm": 22.618579944183196,
|
| 499 |
+
"learning_rate": 1.7122294549409482e-07,
|
| 500 |
+
"logits/chosen": 1.9851901531219482,
|
| 501 |
+
"logits/rejected": 2.9032249450683594,
|
| 502 |
+
"logps/chosen": -424.8306579589844,
|
| 503 |
+
"logps/rejected": -531.9037475585938,
|
| 504 |
+
"loss": 0.5145,
|
| 505 |
+
"rewards/accuracies": 0.800000011920929,
|
| 506 |
+
"rewards/chosen": -1.5183179378509521,
|
| 507 |
+
"rewards/margins": 1.1627616882324219,
|
| 508 |
+
"rewards/rejected": -2.681079387664795,
|
| 509 |
+
"step": 280
|
| 510 |
+
},
|
| 511 |
+
{
|
| 512 |
+
"epoch": 0.6651376146788991,
|
| 513 |
+
"grad_norm": 29.728057206179926,
|
| 514 |
+
"learning_rate": 1.524811797977383e-07,
|
| 515 |
+
"logits/chosen": 1.9606460332870483,
|
| 516 |
+
"logits/rejected": 3.0971603393554688,
|
| 517 |
+
"logps/chosen": -424.40155029296875,
|
| 518 |
+
"logps/rejected": -510.2555236816406,
|
| 519 |
+
"loss": 0.4931,
|
| 520 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 521 |
+
"rewards/chosen": -1.3667072057724,
|
| 522 |
+
"rewards/margins": 1.1966798305511475,
|
| 523 |
+
"rewards/rejected": -2.563387155532837,
|
| 524 |
+
"step": 290
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 0.6880733944954128,
|
| 528 |
+
"grad_norm": 32.56443619783417,
|
| 529 |
+
"learning_rate": 1.3436542743979125e-07,
|
| 530 |
+
"logits/chosen": 1.548921823501587,
|
| 531 |
+
"logits/rejected": 2.2454886436462402,
|
| 532 |
+
"logps/chosen": -435.3851623535156,
|
| 533 |
+
"logps/rejected": -480.7173767089844,
|
| 534 |
+
"loss": 0.5109,
|
| 535 |
+
"rewards/accuracies": 0.762499988079071,
|
| 536 |
+
"rewards/chosen": -1.2943581342697144,
|
| 537 |
+
"rewards/margins": 0.9242169260978699,
|
| 538 |
+
"rewards/rejected": -2.2185752391815186,
|
| 539 |
+
"step": 300
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 0.6880733944954128,
|
| 543 |
+
"eval_logits/chosen": 1.5762183666229248,
|
| 544 |
+
"eval_logits/rejected": 2.8883702754974365,
|
| 545 |
+
"eval_logps/chosen": -418.6533508300781,
|
| 546 |
+
"eval_logps/rejected": -496.99908447265625,
|
| 547 |
+
"eval_loss": 0.48780226707458496,
|
| 548 |
+
"eval_rewards/accuracies": 0.7284482717514038,
|
| 549 |
+
"eval_rewards/chosen": -1.3356326818466187,
|
| 550 |
+
"eval_rewards/margins": 1.1691336631774902,
|
| 551 |
+
"eval_rewards/rejected": -2.5047662258148193,
|
| 552 |
+
"eval_runtime": 94.5476,
|
| 553 |
+
"eval_samples_per_second": 19.228,
|
| 554 |
+
"eval_steps_per_second": 0.307,
|
| 555 |
+
"step": 300
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 0.7110091743119266,
|
| 559 |
+
"grad_norm": 28.8657602827159,
|
| 560 |
+
"learning_rate": 1.1699198087116588e-07,
|
| 561 |
+
"logits/chosen": 1.6565430164337158,
|
| 562 |
+
"logits/rejected": 2.6408467292785645,
|
| 563 |
+
"logps/chosen": -422.91943359375,
|
| 564 |
+
"logps/rejected": -511.9986267089844,
|
| 565 |
+
"loss": 0.5225,
|
| 566 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 567 |
+
"rewards/chosen": -1.449542760848999,
|
| 568 |
+
"rewards/margins": 1.0059467554092407,
|
| 569 |
+
"rewards/rejected": -2.4554896354675293,
|
| 570 |
+
"step": 310
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"epoch": 0.7339449541284404,
|
| 574 |
+
"grad_norm": 28.01306501480046,
|
| 575 |
+
"learning_rate": 1.00472367377196e-07,
|
| 576 |
+
"logits/chosen": 2.107029914855957,
|
| 577 |
+
"logits/rejected": 3.2613253593444824,
|
| 578 |
+
"logps/chosen": -415.86041259765625,
|
| 579 |
+
"logps/rejected": -515.1480712890625,
|
| 580 |
+
"loss": 0.4977,
|
| 581 |
+
"rewards/accuracies": 0.762499988079071,
|
| 582 |
+
"rewards/chosen": -1.4162843227386475,
|
| 583 |
+
"rewards/margins": 1.390293836593628,
|
| 584 |
+
"rewards/rejected": -2.8065781593322754,
|
| 585 |
+
"step": 320
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 0.7568807339449541,
|
| 589 |
+
"grad_norm": 23.410627824071433,
|
| 590 |
+
"learning_rate": 8.49126331382102e-08,
|
| 591 |
+
"logits/chosen": 2.485326051712036,
|
| 592 |
+
"logits/rejected": 3.1795496940612793,
|
| 593 |
+
"logps/chosen": -442.5235290527344,
|
| 594 |
+
"logps/rejected": -500.9856872558594,
|
| 595 |
+
"loss": 0.4944,
|
| 596 |
+
"rewards/accuracies": 0.6875,
|
| 597 |
+
"rewards/chosen": -1.68001389503479,
|
| 598 |
+
"rewards/margins": 0.8588422536849976,
|
| 599 |
+
"rewards/rejected": -2.538856029510498,
|
| 600 |
+
"step": 330
|
| 601 |
+
},
|
| 602 |
+
{
|
| 603 |
+
"epoch": 0.7798165137614679,
|
| 604 |
+
"grad_norm": 36.6413379461555,
|
| 605 |
+
"learning_rate": 7.041266247556812e-08,
|
| 606 |
+
"logits/chosen": 2.2517406940460205,
|
| 607 |
+
"logits/rejected": 2.871661901473999,
|
| 608 |
+
"logps/chosen": -440.378662109375,
|
| 609 |
+
"logps/rejected": -506.90625,
|
| 610 |
+
"loss": 0.4867,
|
| 611 |
+
"rewards/accuracies": 0.793749988079071,
|
| 612 |
+
"rewards/chosen": -1.5238606929779053,
|
| 613 |
+
"rewards/margins": 0.9453619718551636,
|
| 614 |
+
"rewards/rejected": -2.4692225456237793,
|
| 615 |
+
"step": 340
|
| 616 |
+
},
|
| 617 |
+
{
|
| 618 |
+
"epoch": 0.8027522935779816,
|
| 619 |
+
"grad_norm": 32.4776553295204,
|
| 620 |
+
"learning_rate": 5.706553665319955e-08,
|
| 621 |
+
"logits/chosen": 2.2310845851898193,
|
| 622 |
+
"logits/rejected": 3.158881187438965,
|
| 623 |
+
"logps/chosen": -426.0604553222656,
|
| 624 |
+
"logps/rejected": -490.872314453125,
|
| 625 |
+
"loss": 0.5063,
|
| 626 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 627 |
+
"rewards/chosen": -1.5156935453414917,
|
| 628 |
+
"rewards/margins": 1.016414761543274,
|
| 629 |
+
"rewards/rejected": -2.5321083068847656,
|
| 630 |
+
"step": 350
|
| 631 |
+
},
|
| 632 |
+
{
|
| 633 |
+
"epoch": 0.8027522935779816,
|
| 634 |
+
"eval_logits/chosen": 2.1698544025421143,
|
| 635 |
+
"eval_logits/rejected": 3.3468618392944336,
|
| 636 |
+
"eval_logps/chosen": -433.7904052734375,
|
| 637 |
+
"eval_logps/rejected": -514.8548583984375,
|
| 638 |
+
"eval_loss": 0.4814243018627167,
|
| 639 |
+
"eval_rewards/accuracies": 0.7370689511299133,
|
| 640 |
+
"eval_rewards/chosen": -1.487002968788147,
|
| 641 |
+
"eval_rewards/margins": 1.1963213682174683,
|
| 642 |
+
"eval_rewards/rejected": -2.683324098587036,
|
| 643 |
+
"eval_runtime": 95.0017,
|
| 644 |
+
"eval_samples_per_second": 19.136,
|
| 645 |
+
"eval_steps_per_second": 0.305,
|
| 646 |
+
"step": 350
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"epoch": 0.8256880733944955,
|
| 650 |
+
"grad_norm": 22.48584624269231,
|
| 651 |
+
"learning_rate": 4.4956936350761005e-08,
|
| 652 |
+
"logits/chosen": 2.1727540493011475,
|
| 653 |
+
"logits/rejected": 2.8084352016448975,
|
| 654 |
+
"logps/chosen": -391.9521179199219,
|
| 655 |
+
"logps/rejected": -506.19482421875,
|
| 656 |
+
"loss": 0.4802,
|
| 657 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 658 |
+
"rewards/chosen": -1.3963208198547363,
|
| 659 |
+
"rewards/margins": 1.0802565813064575,
|
| 660 |
+
"rewards/rejected": -2.4765772819519043,
|
| 661 |
+
"step": 360
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 0.8486238532110092,
|
| 665 |
+
"grad_norm": 26.981108661605717,
|
| 666 |
+
"learning_rate": 3.416459164418123e-08,
|
| 667 |
+
"logits/chosen": 1.7840194702148438,
|
| 668 |
+
"logits/rejected": 2.861525535583496,
|
| 669 |
+
"logps/chosen": -447.0101013183594,
|
| 670 |
+
"logps/rejected": -527.9534912109375,
|
| 671 |
+
"loss": 0.4861,
|
| 672 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 673 |
+
"rewards/chosen": -1.4742560386657715,
|
| 674 |
+
"rewards/margins": 1.179532527923584,
|
| 675 |
+
"rewards/rejected": -2.6537883281707764,
|
| 676 |
+
"step": 370
|
| 677 |
+
},
|
| 678 |
+
{
|
| 679 |
+
"epoch": 0.8715596330275229,
|
| 680 |
+
"grad_norm": 32.47929591425802,
|
| 681 |
+
"learning_rate": 2.475778302439524e-08,
|
| 682 |
+
"logits/chosen": 2.0354526042938232,
|
| 683 |
+
"logits/rejected": 3.16379976272583,
|
| 684 |
+
"logps/chosen": -440.89593505859375,
|
| 685 |
+
"logps/rejected": -530.92919921875,
|
| 686 |
+
"loss": 0.4928,
|
| 687 |
+
"rewards/accuracies": 0.762499988079071,
|
| 688 |
+
"rewards/chosen": -1.4451792240142822,
|
| 689 |
+
"rewards/margins": 1.287550687789917,
|
| 690 |
+
"rewards/rejected": -2.732729911804199,
|
| 691 |
+
"step": 380
|
| 692 |
+
},
|
| 693 |
+
{
|
| 694 |
+
"epoch": 0.8944954128440367,
|
| 695 |
+
"grad_norm": 33.85969586257753,
|
| 696 |
+
"learning_rate": 1.6796896657433805e-08,
|
| 697 |
+
"logits/chosen": 2.4787354469299316,
|
| 698 |
+
"logits/rejected": 3.341116428375244,
|
| 699 |
+
"logps/chosen": -407.4842834472656,
|
| 700 |
+
"logps/rejected": -490.75555419921875,
|
| 701 |
+
"loss": 0.503,
|
| 702 |
+
"rewards/accuracies": 0.75,
|
| 703 |
+
"rewards/chosen": -1.5552680492401123,
|
| 704 |
+
"rewards/margins": 1.0592255592346191,
|
| 705 |
+
"rewards/rejected": -2.6144936084747314,
|
| 706 |
+
"step": 390
|
| 707 |
+
},
|
| 708 |
+
{
|
| 709 |
+
"epoch": 0.9174311926605505,
|
| 710 |
+
"grad_norm": 23.099468591696,
|
| 711 |
+
"learning_rate": 1.0333036740834855e-08,
|
| 712 |
+
"logits/chosen": 2.7448534965515137,
|
| 713 |
+
"logits/rejected": 3.38175630569458,
|
| 714 |
+
"logps/chosen": -370.7669372558594,
|
| 715 |
+
"logps/rejected": -485.293701171875,
|
| 716 |
+
"loss": 0.4936,
|
| 717 |
+
"rewards/accuracies": 0.762499988079071,
|
| 718 |
+
"rewards/chosen": -1.4620729684829712,
|
| 719 |
+
"rewards/margins": 1.098707914352417,
|
| 720 |
+
"rewards/rejected": -2.5607810020446777,
|
| 721 |
+
"step": 400
|
| 722 |
+
},
|
| 723 |
+
{
|
| 724 |
+
"epoch": 0.9174311926605505,
|
| 725 |
+
"eval_logits/chosen": 2.333301305770874,
|
| 726 |
+
"eval_logits/rejected": 3.4838266372680664,
|
| 727 |
+
"eval_logps/chosen": -443.5758361816406,
|
| 728 |
+
"eval_logps/rejected": -526.9686279296875,
|
| 729 |
+
"eval_loss": 0.4815457761287689,
|
| 730 |
+
"eval_rewards/accuracies": 0.732758641242981,
|
| 731 |
+
"eval_rewards/chosen": -1.5848578214645386,
|
| 732 |
+
"eval_rewards/margins": 1.2196037769317627,
|
| 733 |
+
"eval_rewards/rejected": -2.8044614791870117,
|
| 734 |
+
"eval_runtime": 95.9268,
|
| 735 |
+
"eval_samples_per_second": 18.952,
|
| 736 |
+
"eval_steps_per_second": 0.302,
|
| 737 |
+
"step": 400
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"epoch": 0.9403669724770642,
|
| 741 |
+
"grad_norm": 24.65582545776771,
|
| 742 |
+
"learning_rate": 5.4076974448211685e-09,
|
| 743 |
+
"logits/chosen": 2.678440809249878,
|
| 744 |
+
"logits/rejected": 3.7608656883239746,
|
| 745 |
+
"logps/chosen": -434.31890869140625,
|
| 746 |
+
"logps/rejected": -525.4384765625,
|
| 747 |
+
"loss": 0.4901,
|
| 748 |
+
"rewards/accuracies": 0.737500011920929,
|
| 749 |
+
"rewards/chosen": -1.633670449256897,
|
| 750 |
+
"rewards/margins": 1.2402476072311401,
|
| 751 |
+
"rewards/rejected": -2.873918056488037,
|
| 752 |
+
"step": 410
|
| 753 |
+
},
|
| 754 |
+
{
|
| 755 |
+
"epoch": 0.963302752293578,
|
| 756 |
+
"grad_norm": 24.35589696145217,
|
| 757 |
+
"learning_rate": 2.052496544188487e-09,
|
| 758 |
+
"logits/chosen": 2.744503974914551,
|
| 759 |
+
"logits/rejected": 3.5964274406433105,
|
| 760 |
+
"logps/chosen": -424.82373046875,
|
| 761 |
+
"logps/rejected": -535.3244018554688,
|
| 762 |
+
"loss": 0.4891,
|
| 763 |
+
"rewards/accuracies": 0.768750011920929,
|
| 764 |
+
"rewards/chosen": -1.6538474559783936,
|
| 765 |
+
"rewards/margins": 1.2406232357025146,
|
| 766 |
+
"rewards/rejected": -2.894470691680908,
|
| 767 |
+
"step": 420
|
| 768 |
+
},
|
| 769 |
+
{
|
| 770 |
+
"epoch": 0.9862385321100917,
|
| 771 |
+
"grad_norm": 24.120958058828865,
|
| 772 |
+
"learning_rate": 2.889724508297886e-10,
|
| 773 |
+
"logits/chosen": 2.3434717655181885,
|
| 774 |
+
"logits/rejected": 3.6963977813720703,
|
| 775 |
+
"logps/chosen": -459.7005310058594,
|
| 776 |
+
"logps/rejected": -510.67413330078125,
|
| 777 |
+
"loss": 0.4816,
|
| 778 |
+
"rewards/accuracies": 0.731249988079071,
|
| 779 |
+
"rewards/chosen": -1.5652509927749634,
|
| 780 |
+
"rewards/margins": 1.1011250019073486,
|
| 781 |
+
"rewards/rejected": -2.6663756370544434,
|
| 782 |
+
"step": 430
|
| 783 |
+
},
|
| 784 |
+
{
|
| 785 |
+
"epoch": 1.0,
|
| 786 |
+
"step": 436,
|
| 787 |
+
"total_flos": 0.0,
|
| 788 |
+
"train_loss": 0.5380998532706445,
|
| 789 |
+
"train_runtime": 11735.5354,
|
| 790 |
+
"train_samples_per_second": 4.751,
|
| 791 |
+
"train_steps_per_second": 0.037
|
| 792 |
+
}
|
| 793 |
+
],
|
| 794 |
+
"logging_steps": 10,
|
| 795 |
+
"max_steps": 436,
|
| 796 |
+
"num_input_tokens_seen": 0,
|
| 797 |
+
"num_train_epochs": 1,
|
| 798 |
+
"save_steps": 100,
|
| 799 |
+
"stateful_callbacks": {
|
| 800 |
+
"TrainerControl": {
|
| 801 |
+
"args": {
|
| 802 |
+
"should_epoch_stop": false,
|
| 803 |
+
"should_evaluate": false,
|
| 804 |
+
"should_log": false,
|
| 805 |
+
"should_save": true,
|
| 806 |
+
"should_training_stop": true
|
| 807 |
+
},
|
| 808 |
+
"attributes": {}
|
| 809 |
+
}
|
| 810 |
+
},
|
| 811 |
+
"total_flos": 0.0,
|
| 812 |
+
"train_batch_size": 8,
|
| 813 |
+
"trial_name": null,
|
| 814 |
+
"trial_params": null
|
| 815 |
+
}
|