Model save
Browse files- README.md +79 -0
- all_results.json +9 -0
- generation_config.json +6 -0
- train_results.json +9 -0
- trainer_state.json +679 -0
README.md
ADDED
|
@@ -0,0 +1,79 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 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-magpi-reward-scale-01
|
| 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-magpi-reward-scale-01
|
| 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: 0.0001
|
| 21 |
+
- Rewards/chosen: -2.1581
|
| 22 |
+
- Rewards/rejected: -72.1363
|
| 23 |
+
- Rewards/accuracies: 1.0
|
| 24 |
+
- Rewards/margins: 69.9782
|
| 25 |
+
- Logps/rejected: -7854.4189
|
| 26 |
+
- Logps/chosen: -582.7947
|
| 27 |
+
- Logits/rejected: 2.3418
|
| 28 |
+
- Logits/chosen: -1.3156
|
| 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 |
+
| 0.0066 | 0.1420 | 50 | 0.0037 | -1.1750 | -45.6975 | 0.9960 | 44.5225 | -5210.5396 | -484.4839 | -2.4201 | -2.9075 |
|
| 66 |
+
| 0.0019 | 0.2841 | 100 | 0.0003 | -1.8979 | -62.8577 | 1.0 | 60.9597 | -6926.5537 | -556.7739 | -0.2036 | -2.7790 |
|
| 67 |
+
| 0.001 | 0.4261 | 150 | 0.0002 | -1.8922 | -67.8911 | 1.0 | 65.9989 | -7429.9023 | -556.2059 | 1.8870 | -2.6554 |
|
| 68 |
+
| 0.0013 | 0.5682 | 200 | 0.0001 | -2.1555 | -73.7897 | 1.0 | 71.6342 | -8019.7568 | -582.5308 | 2.4298 | -2.2775 |
|
| 69 |
+
| 0.0015 | 0.7102 | 250 | 0.0001 | -2.1984 | -72.3465 | 1.0 | 70.1481 | -7875.4424 | -586.8257 | 2.3430 | -1.6277 |
|
| 70 |
+
| 0.0009 | 0.8523 | 300 | 0.0001 | -2.1293 | -71.2932 | 1.0 | 69.1638 | -7770.1060 | -579.9163 | 2.3229 | -1.4073 |
|
| 71 |
+
| 0.0015 | 0.9943 | 350 | 0.0001 | -2.1581 | -72.1363 | 1.0 | 69.9782 | -7854.4189 | -582.7947 | 2.3418 | -1.3156 |
|
| 72 |
+
|
| 73 |
+
|
| 74 |
+
### Framework versions
|
| 75 |
+
|
| 76 |
+
- Transformers 4.44.0.dev0
|
| 77 |
+
- Pytorch 2.1.2
|
| 78 |
+
- Datasets 2.20.0
|
| 79 |
+
- 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": 0.04425002944373608,
|
| 5 |
+
"train_runtime": 10145.5361,
|
| 6 |
+
"train_samples": 45000,
|
| 7 |
+
"train_samples_per_second": 4.435,
|
| 8 |
+
"train_steps_per_second": 0.035
|
| 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": 0.04425002944373608,
|
| 5 |
+
"train_runtime": 10145.5361,
|
| 6 |
+
"train_samples": 45000,
|
| 7 |
+
"train_samples_per_second": 4.435,
|
| 8 |
+
"train_steps_per_second": 0.035
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,679 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.0,
|
| 5 |
+
"eval_steps": 50,
|
| 6 |
+
"global_step": 352,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.028409090909090908,
|
| 13 |
+
"grad_norm": 38.70626469652959,
|
| 14 |
+
"learning_rate": 1.3888888888888888e-07,
|
| 15 |
+
"logits/chosen": -2.8592591285705566,
|
| 16 |
+
"logits/rejected": -2.64298152923584,
|
| 17 |
+
"logps/chosen": -390.552490234375,
|
| 18 |
+
"logps/rejected": -607.8676147460938,
|
| 19 |
+
"loss": 0.6889,
|
| 20 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 21 |
+
"rewards/chosen": 0.002072929637506604,
|
| 22 |
+
"rewards/margins": 0.013672290369868279,
|
| 23 |
+
"rewards/rejected": -0.011599360033869743,
|
| 24 |
+
"step": 10
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.056818181818181816,
|
| 28 |
+
"grad_norm": 19.199796929875,
|
| 29 |
+
"learning_rate": 2.7777777777777776e-07,
|
| 30 |
+
"logits/chosen": -2.8421695232391357,
|
| 31 |
+
"logits/rejected": -2.6945552825927734,
|
| 32 |
+
"logps/chosen": -328.68048095703125,
|
| 33 |
+
"logps/rejected": -775.0625,
|
| 34 |
+
"loss": 0.5324,
|
| 35 |
+
"rewards/accuracies": 0.987500011920929,
|
| 36 |
+
"rewards/chosen": 0.08464664220809937,
|
| 37 |
+
"rewards/margins": 0.6608749032020569,
|
| 38 |
+
"rewards/rejected": -0.5762282609939575,
|
| 39 |
+
"step": 20
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.08522727272727272,
|
| 43 |
+
"grad_norm": 5.168742449938278,
|
| 44 |
+
"learning_rate": 4.1666666666666667e-07,
|
| 45 |
+
"logits/chosen": -2.85664439201355,
|
| 46 |
+
"logits/rejected": -2.673206329345703,
|
| 47 |
+
"logps/chosen": -295.9004821777344,
|
| 48 |
+
"logps/rejected": -1121.4083251953125,
|
| 49 |
+
"loss": 0.2036,
|
| 50 |
+
"rewards/accuracies": 1.0,
|
| 51 |
+
"rewards/chosen": 0.3638155460357666,
|
| 52 |
+
"rewards/margins": 4.79835319519043,
|
| 53 |
+
"rewards/rejected": -4.434537887573242,
|
| 54 |
+
"step": 30
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.11363636363636363,
|
| 58 |
+
"grad_norm": 1.5631392125120753,
|
| 59 |
+
"learning_rate": 4.998023493068254e-07,
|
| 60 |
+
"logits/chosen": -2.8252196311950684,
|
| 61 |
+
"logits/rejected": -2.627849578857422,
|
| 62 |
+
"logps/chosen": -318.1038818359375,
|
| 63 |
+
"logps/rejected": -2185.43017578125,
|
| 64 |
+
"loss": 0.0507,
|
| 65 |
+
"rewards/accuracies": 1.0,
|
| 66 |
+
"rewards/chosen": 0.28651586174964905,
|
| 67 |
+
"rewards/margins": 15.061859130859375,
|
| 68 |
+
"rewards/rejected": -14.775341987609863,
|
| 69 |
+
"step": 40
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.14204545454545456,
|
| 73 |
+
"grad_norm": 0.27100229342130666,
|
| 74 |
+
"learning_rate": 4.975823666181255e-07,
|
| 75 |
+
"logits/chosen": -2.792081356048584,
|
| 76 |
+
"logits/rejected": -2.4893665313720703,
|
| 77 |
+
"logps/chosen": -460.21923828125,
|
| 78 |
+
"logps/rejected": -4558.29833984375,
|
| 79 |
+
"loss": 0.0066,
|
| 80 |
+
"rewards/accuracies": 1.0,
|
| 81 |
+
"rewards/chosen": -0.9758400917053223,
|
| 82 |
+
"rewards/margins": 38.06774139404297,
|
| 83 |
+
"rewards/rejected": -39.043575286865234,
|
| 84 |
+
"step": 50
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.14204545454545456,
|
| 88 |
+
"eval_logits/chosen": -2.907456398010254,
|
| 89 |
+
"eval_logits/rejected": -2.4200961589813232,
|
| 90 |
+
"eval_logps/chosen": -484.4838562011719,
|
| 91 |
+
"eval_logps/rejected": -5210.53955078125,
|
| 92 |
+
"eval_loss": 0.0037060165777802467,
|
| 93 |
+
"eval_rewards/accuracies": 0.9959677457809448,
|
| 94 |
+
"eval_rewards/chosen": -1.1750215291976929,
|
| 95 |
+
"eval_rewards/margins": 44.522483825683594,
|
| 96 |
+
"eval_rewards/rejected": -45.697505950927734,
|
| 97 |
+
"eval_runtime": 193.8738,
|
| 98 |
+
"eval_samples_per_second": 20.142,
|
| 99 |
+
"eval_steps_per_second": 0.32,
|
| 100 |
+
"step": 50
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.17045454545454544,
|
| 104 |
+
"grad_norm": 0.5320174620153905,
|
| 105 |
+
"learning_rate": 4.929173350101024e-07,
|
| 106 |
+
"logits/chosen": -2.949294090270996,
|
| 107 |
+
"logits/rejected": -2.318241596221924,
|
| 108 |
+
"logps/chosen": -495.1114196777344,
|
| 109 |
+
"logps/rejected": -5547.14013671875,
|
| 110 |
+
"loss": 0.0038,
|
| 111 |
+
"rewards/accuracies": 0.9937499761581421,
|
| 112 |
+
"rewards/chosen": -1.3521640300750732,
|
| 113 |
+
"rewards/margins": 47.690696716308594,
|
| 114 |
+
"rewards/rejected": -49.0428581237793,
|
| 115 |
+
"step": 60
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"epoch": 0.19886363636363635,
|
| 119 |
+
"grad_norm": 0.14876040672715532,
|
| 120 |
+
"learning_rate": 4.858533249305336e-07,
|
| 121 |
+
"logits/chosen": -2.8789420127868652,
|
| 122 |
+
"logits/rejected": -1.8734362125396729,
|
| 123 |
+
"logps/chosen": -535.8917846679688,
|
| 124 |
+
"logps/rejected": -5914.61865234375,
|
| 125 |
+
"loss": 0.0032,
|
| 126 |
+
"rewards/accuracies": 1.0,
|
| 127 |
+
"rewards/chosen": -1.5673115253448486,
|
| 128 |
+
"rewards/margins": 51.4515266418457,
|
| 129 |
+
"rewards/rejected": -53.018836975097656,
|
| 130 |
+
"step": 70
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 0.22727272727272727,
|
| 134 |
+
"grad_norm": 6.953457462206189,
|
| 135 |
+
"learning_rate": 4.764600984163808e-07,
|
| 136 |
+
"logits/chosen": -2.9963574409484863,
|
| 137 |
+
"logits/rejected": -1.690744161605835,
|
| 138 |
+
"logps/chosen": -566.8291015625,
|
| 139 |
+
"logps/rejected": -7217.30712890625,
|
| 140 |
+
"loss": 0.0032,
|
| 141 |
+
"rewards/accuracies": 1.0,
|
| 142 |
+
"rewards/chosen": -1.6050817966461182,
|
| 143 |
+
"rewards/margins": 63.83066940307617,
|
| 144 |
+
"rewards/rejected": -65.43574523925781,
|
| 145 |
+
"step": 80
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 0.2556818181818182,
|
| 149 |
+
"grad_norm": 0.03391351087291168,
|
| 150 |
+
"learning_rate": 4.6483042014491527e-07,
|
| 151 |
+
"logits/chosen": -2.9498190879821777,
|
| 152 |
+
"logits/rejected": -1.7918399572372437,
|
| 153 |
+
"logps/chosen": -580.243408203125,
|
| 154 |
+
"logps/rejected": -6469.92578125,
|
| 155 |
+
"loss": 0.0113,
|
| 156 |
+
"rewards/accuracies": 0.9937499761581421,
|
| 157 |
+
"rewards/chosen": -2.181300640106201,
|
| 158 |
+
"rewards/margins": 56.14410400390625,
|
| 159 |
+
"rewards/rejected": -58.32539749145508,
|
| 160 |
+
"step": 90
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 0.2840909090909091,
|
| 164 |
+
"grad_norm": 0.39266433528373734,
|
| 165 |
+
"learning_rate": 4.510791413176912e-07,
|
| 166 |
+
"logits/chosen": -2.934478521347046,
|
| 167 |
+
"logits/rejected": -0.735981285572052,
|
| 168 |
+
"logps/chosen": -570.57568359375,
|
| 169 |
+
"logps/rejected": -7235.8173828125,
|
| 170 |
+
"loss": 0.0019,
|
| 171 |
+
"rewards/accuracies": 1.0,
|
| 172 |
+
"rewards/chosen": -2.0110573768615723,
|
| 173 |
+
"rewards/margins": 63.289306640625,
|
| 174 |
+
"rewards/rejected": -65.30036926269531,
|
| 175 |
+
"step": 100
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.2840909090909091,
|
| 179 |
+
"eval_logits/chosen": -2.7790369987487793,
|
| 180 |
+
"eval_logits/rejected": -0.2035539597272873,
|
| 181 |
+
"eval_logps/chosen": -556.77392578125,
|
| 182 |
+
"eval_logps/rejected": -6926.5537109375,
|
| 183 |
+
"eval_loss": 0.0003049192891921848,
|
| 184 |
+
"eval_rewards/accuracies": 1.0,
|
| 185 |
+
"eval_rewards/chosen": -1.897922158241272,
|
| 186 |
+
"eval_rewards/margins": 60.95973205566406,
|
| 187 |
+
"eval_rewards/rejected": -62.85765838623047,
|
| 188 |
+
"eval_runtime": 193.4087,
|
| 189 |
+
"eval_samples_per_second": 20.19,
|
| 190 |
+
"eval_steps_per_second": 0.321,
|
| 191 |
+
"step": 100
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.3125,
|
| 195 |
+
"grad_norm": 0.0360149995259237,
|
| 196 |
+
"learning_rate": 4.353420654246546e-07,
|
| 197 |
+
"logits/chosen": -2.6916072368621826,
|
| 198 |
+
"logits/rejected": -0.07679282128810883,
|
| 199 |
+
"logps/chosen": -566.1495361328125,
|
| 200 |
+
"logps/rejected": -6591.94287109375,
|
| 201 |
+
"loss": 0.0025,
|
| 202 |
+
"rewards/accuracies": 1.0,
|
| 203 |
+
"rewards/chosen": -1.907283067703247,
|
| 204 |
+
"rewards/margins": 58.01946258544922,
|
| 205 |
+
"rewards/rejected": -59.92673873901367,
|
| 206 |
+
"step": 110
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.3409090909090909,
|
| 210 |
+
"grad_norm": 0.05338636863642817,
|
| 211 |
+
"learning_rate": 4.177746070897592e-07,
|
| 212 |
+
"logits/chosen": -2.5057804584503174,
|
| 213 |
+
"logits/rejected": 0.8570491671562195,
|
| 214 |
+
"logps/chosen": -578.2913818359375,
|
| 215 |
+
"logps/rejected": -7127.86328125,
|
| 216 |
+
"loss": 0.004,
|
| 217 |
+
"rewards/accuracies": 1.0,
|
| 218 |
+
"rewards/chosen": -1.9677445888519287,
|
| 219 |
+
"rewards/margins": 62.9386100769043,
|
| 220 |
+
"rewards/rejected": -64.90635681152344,
|
| 221 |
+
"step": 120
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.3693181818181818,
|
| 225 |
+
"grad_norm": 0.28121189018180337,
|
| 226 |
+
"learning_rate": 3.9855025724292763e-07,
|
| 227 |
+
"logits/chosen": -2.5544893741607666,
|
| 228 |
+
"logits/rejected": 1.7402244806289673,
|
| 229 |
+
"logps/chosen": -602.7317504882812,
|
| 230 |
+
"logps/rejected": -7376.546875,
|
| 231 |
+
"loss": 0.0015,
|
| 232 |
+
"rewards/accuracies": 1.0,
|
| 233 |
+
"rewards/chosen": -2.3228020668029785,
|
| 234 |
+
"rewards/margins": 65.03727722167969,
|
| 235 |
+
"rewards/rejected": -67.3600845336914,
|
| 236 |
+
"step": 130
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 0.3977272727272727,
|
| 240 |
+
"grad_norm": 0.7726944716269485,
|
| 241 |
+
"learning_rate": 3.7785886977585555e-07,
|
| 242 |
+
"logits/chosen": -2.5471339225769043,
|
| 243 |
+
"logits/rejected": 1.9610217809677124,
|
| 244 |
+
"logps/chosen": -578.9056396484375,
|
| 245 |
+
"logps/rejected": -8475.4873046875,
|
| 246 |
+
"loss": 0.0019,
|
| 247 |
+
"rewards/accuracies": 1.0,
|
| 248 |
+
"rewards/chosen": -2.1286935806274414,
|
| 249 |
+
"rewards/margins": 76.206298828125,
|
| 250 |
+
"rewards/rejected": -78.3349838256836,
|
| 251 |
+
"step": 140
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 0.42613636363636365,
|
| 255 |
+
"grad_norm": 0.5096412009910095,
|
| 256 |
+
"learning_rate": 3.5590478660213206e-07,
|
| 257 |
+
"logits/chosen": -2.5447940826416016,
|
| 258 |
+
"logits/rejected": 1.8056743144989014,
|
| 259 |
+
"logps/chosen": -599.0574340820312,
|
| 260 |
+
"logps/rejected": -7813.42333984375,
|
| 261 |
+
"loss": 0.001,
|
| 262 |
+
"rewards/accuracies": 1.0,
|
| 263 |
+
"rewards/chosen": -2.2099416255950928,
|
| 264 |
+
"rewards/margins": 68.80451965332031,
|
| 265 |
+
"rewards/rejected": -71.01445007324219,
|
| 266 |
+
"step": 150
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 0.42613636363636365,
|
| 270 |
+
"eval_logits/chosen": -2.655405044555664,
|
| 271 |
+
"eval_logits/rejected": 1.886979103088379,
|
| 272 |
+
"eval_logps/chosen": -556.2059326171875,
|
| 273 |
+
"eval_logps/rejected": -7429.90234375,
|
| 274 |
+
"eval_loss": 0.0001985544804483652,
|
| 275 |
+
"eval_rewards/accuracies": 1.0,
|
| 276 |
+
"eval_rewards/chosen": -1.892242193222046,
|
| 277 |
+
"eval_rewards/margins": 65.99889373779297,
|
| 278 |
+
"eval_rewards/rejected": -67.8911361694336,
|
| 279 |
+
"eval_runtime": 192.529,
|
| 280 |
+
"eval_samples_per_second": 20.283,
|
| 281 |
+
"eval_steps_per_second": 0.322,
|
| 282 |
+
"step": 150
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.45454545454545453,
|
| 286 |
+
"grad_norm": 2.608953110629928,
|
| 287 |
+
"learning_rate": 3.3290481963801696e-07,
|
| 288 |
+
"logits/chosen": -2.655552387237549,
|
| 289 |
+
"logits/rejected": 2.1644370555877686,
|
| 290 |
+
"logps/chosen": -535.6851806640625,
|
| 291 |
+
"logps/rejected": -7245.27734375,
|
| 292 |
+
"loss": 0.0005,
|
| 293 |
+
"rewards/accuracies": 1.0,
|
| 294 |
+
"rewards/chosen": -1.6554721593856812,
|
| 295 |
+
"rewards/margins": 64.63240051269531,
|
| 296 |
+
"rewards/rejected": -66.28787231445312,
|
| 297 |
+
"step": 160
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 0.48295454545454547,
|
| 301 |
+
"grad_norm": 0.0342477427475356,
|
| 302 |
+
"learning_rate": 3.0908610963322626e-07,
|
| 303 |
+
"logits/chosen": -2.3891050815582275,
|
| 304 |
+
"logits/rejected": 2.3125836849212646,
|
| 305 |
+
"logps/chosen": -601.9050903320312,
|
| 306 |
+
"logps/rejected": -8139.94921875,
|
| 307 |
+
"loss": 0.0053,
|
| 308 |
+
"rewards/accuracies": 1.0,
|
| 309 |
+
"rewards/chosen": -2.1428682804107666,
|
| 310 |
+
"rewards/margins": 71.97351837158203,
|
| 311 |
+
"rewards/rejected": -74.11639404296875,
|
| 312 |
+
"step": 170
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 0.5113636363636364,
|
| 316 |
+
"grad_norm": 0.011995212283351854,
|
| 317 |
+
"learning_rate": 2.846838829972671e-07,
|
| 318 |
+
"logits/chosen": -2.39284610748291,
|
| 319 |
+
"logits/rejected": 2.5177011489868164,
|
| 320 |
+
"logps/chosen": -605.0557861328125,
|
| 321 |
+
"logps/rejected": -7604.11328125,
|
| 322 |
+
"loss": 0.0022,
|
| 323 |
+
"rewards/accuracies": 1.0,
|
| 324 |
+
"rewards/chosen": -2.441514492034912,
|
| 325 |
+
"rewards/margins": 67.46978759765625,
|
| 326 |
+
"rewards/rejected": -69.91130065917969,
|
| 327 |
+
"step": 180
|
| 328 |
+
},
|
| 329 |
+
{
|
| 330 |
+
"epoch": 0.5397727272727273,
|
| 331 |
+
"grad_norm": 0.17953481614212255,
|
| 332 |
+
"learning_rate": 2.5993912877423147e-07,
|
| 333 |
+
"logits/chosen": -2.4631476402282715,
|
| 334 |
+
"logits/rejected": 1.9089019298553467,
|
| 335 |
+
"logps/chosen": -553.119140625,
|
| 336 |
+
"logps/rejected": -8167.6572265625,
|
| 337 |
+
"loss": 0.0007,
|
| 338 |
+
"rewards/accuracies": 1.0,
|
| 339 |
+
"rewards/chosen": -2.075033664703369,
|
| 340 |
+
"rewards/margins": 72.6854248046875,
|
| 341 |
+
"rewards/rejected": -74.76045989990234,
|
| 342 |
+
"step": 190
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 0.5681818181818182,
|
| 346 |
+
"grad_norm": 0.7077913782197723,
|
| 347 |
+
"learning_rate": 2.3509621870754504e-07,
|
| 348 |
+
"logits/chosen": -2.350708484649658,
|
| 349 |
+
"logits/rejected": 2.6935336589813232,
|
| 350 |
+
"logps/chosen": -562.2257080078125,
|
| 351 |
+
"logps/rejected": -7312.64453125,
|
| 352 |
+
"loss": 0.0013,
|
| 353 |
+
"rewards/accuracies": 1.0,
|
| 354 |
+
"rewards/chosen": -1.970958948135376,
|
| 355 |
+
"rewards/margins": 64.8470687866211,
|
| 356 |
+
"rewards/rejected": -66.81802368164062,
|
| 357 |
+
"step": 200
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 0.5681818181818182,
|
| 361 |
+
"eval_logits/chosen": -2.2775018215179443,
|
| 362 |
+
"eval_logits/rejected": 2.429776430130005,
|
| 363 |
+
"eval_logps/chosen": -582.5308227539062,
|
| 364 |
+
"eval_logps/rejected": -8019.7568359375,
|
| 365 |
+
"eval_loss": 0.00011486691073514521,
|
| 366 |
+
"eval_rewards/accuracies": 1.0,
|
| 367 |
+
"eval_rewards/chosen": -2.155491352081299,
|
| 368 |
+
"eval_rewards/margins": 71.63419342041016,
|
| 369 |
+
"eval_rewards/rejected": -73.78968811035156,
|
| 370 |
+
"eval_runtime": 192.7674,
|
| 371 |
+
"eval_samples_per_second": 20.258,
|
| 372 |
+
"eval_steps_per_second": 0.322,
|
| 373 |
+
"step": 200
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 0.5965909090909091,
|
| 377 |
+
"grad_norm": 0.24855740640565924,
|
| 378 |
+
"learning_rate": 2.1040049389819624e-07,
|
| 379 |
+
"logits/chosen": -2.1356868743896484,
|
| 380 |
+
"logits/rejected": 2.5835635662078857,
|
| 381 |
+
"logps/chosen": -601.9248046875,
|
| 382 |
+
"logps/rejected": -7937.8779296875,
|
| 383 |
+
"loss": 0.0034,
|
| 384 |
+
"rewards/accuracies": 1.0,
|
| 385 |
+
"rewards/chosen": -2.1128151416778564,
|
| 386 |
+
"rewards/margins": 70.94691467285156,
|
| 387 |
+
"rewards/rejected": -73.05973052978516,
|
| 388 |
+
"step": 210
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 0.625,
|
| 392 |
+
"grad_norm": 1.3430282073648296,
|
| 393 |
+
"learning_rate": 1.8609584188988133e-07,
|
| 394 |
+
"logits/chosen": -1.6919233798980713,
|
| 395 |
+
"logits/rejected": 2.1380715370178223,
|
| 396 |
+
"logps/chosen": -634.4395751953125,
|
| 397 |
+
"logps/rejected": -7114.49462890625,
|
| 398 |
+
"loss": 0.0034,
|
| 399 |
+
"rewards/accuracies": 1.0,
|
| 400 |
+
"rewards/chosen": -2.6983890533447266,
|
| 401 |
+
"rewards/margins": 61.95018768310547,
|
| 402 |
+
"rewards/rejected": -64.64857482910156,
|
| 403 |
+
"step": 220
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"epoch": 0.6534090909090909,
|
| 407 |
+
"grad_norm": 35.7863072981701,
|
| 408 |
+
"learning_rate": 1.624222881090439e-07,
|
| 409 |
+
"logits/chosen": -1.9383169412612915,
|
| 410 |
+
"logits/rejected": 1.8241441249847412,
|
| 411 |
+
"logps/chosen": -657.0816040039062,
|
| 412 |
+
"logps/rejected": -8359.109375,
|
| 413 |
+
"loss": 0.0111,
|
| 414 |
+
"rewards/accuracies": 0.9937499761581421,
|
| 415 |
+
"rewards/chosen": -2.6241703033447266,
|
| 416 |
+
"rewards/margins": 74.1111068725586,
|
| 417 |
+
"rewards/rejected": -76.73527526855469,
|
| 418 |
+
"step": 230
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"epoch": 0.6818181818181818,
|
| 422 |
+
"grad_norm": 0.3968538526472747,
|
| 423 |
+
"learning_rate": 1.3961362544602212e-07,
|
| 424 |
+
"logits/chosen": -1.7102636098861694,
|
| 425 |
+
"logits/rejected": 2.2134716510772705,
|
| 426 |
+
"logps/chosen": -606.8711547851562,
|
| 427 |
+
"logps/rejected": -7288.10302734375,
|
| 428 |
+
"loss": 0.001,
|
| 429 |
+
"rewards/accuracies": 1.0,
|
| 430 |
+
"rewards/chosen": -2.3085877895355225,
|
| 431 |
+
"rewards/margins": 64.6317138671875,
|
| 432 |
+
"rewards/rejected": -66.94029998779297,
|
| 433 |
+
"step": 240
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.7102272727272727,
|
| 437 |
+
"grad_norm": 0.345391461678324,
|
| 438 |
+
"learning_rate": 1.1789510538684522e-07,
|
| 439 |
+
"logits/chosen": -1.7613790035247803,
|
| 440 |
+
"logits/rejected": 2.189953088760376,
|
| 441 |
+
"logps/chosen": -568.7935791015625,
|
| 442 |
+
"logps/rejected": -9133.8701171875,
|
| 443 |
+
"loss": 0.0015,
|
| 444 |
+
"rewards/accuracies": 1.0,
|
| 445 |
+
"rewards/chosen": -2.16759991645813,
|
| 446 |
+
"rewards/margins": 81.87528228759766,
|
| 447 |
+
"rewards/rejected": -84.04288482666016,
|
| 448 |
+
"step": 250
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"epoch": 0.7102272727272727,
|
| 452 |
+
"eval_logits/chosen": -1.6277003288269043,
|
| 453 |
+
"eval_logits/rejected": 2.342982769012451,
|
| 454 |
+
"eval_logps/chosen": -586.8257446289062,
|
| 455 |
+
"eval_logps/rejected": -7875.4423828125,
|
| 456 |
+
"eval_loss": 9.437829430680722e-05,
|
| 457 |
+
"eval_rewards/accuracies": 1.0,
|
| 458 |
+
"eval_rewards/chosen": -2.198439836502075,
|
| 459 |
+
"eval_rewards/margins": 70.1480941772461,
|
| 460 |
+
"eval_rewards/rejected": -72.34652709960938,
|
| 461 |
+
"eval_runtime": 192.1809,
|
| 462 |
+
"eval_samples_per_second": 20.319,
|
| 463 |
+
"eval_steps_per_second": 0.323,
|
| 464 |
+
"step": 250
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 0.7386363636363636,
|
| 468 |
+
"grad_norm": 0.30273821428387077,
|
| 469 |
+
"learning_rate": 9.748121349736891e-08,
|
| 470 |
+
"logits/chosen": -1.565211534500122,
|
| 471 |
+
"logits/rejected": 2.2146430015563965,
|
| 472 |
+
"logps/chosen": -623.1561279296875,
|
| 473 |
+
"logps/rejected": -8230.65625,
|
| 474 |
+
"loss": 0.0006,
|
| 475 |
+
"rewards/accuracies": 1.0,
|
| 476 |
+
"rewards/chosen": -2.393279552459717,
|
| 477 |
+
"rewards/margins": 73.13200378417969,
|
| 478 |
+
"rewards/rejected": -75.52528381347656,
|
| 479 |
+
"step": 260
|
| 480 |
+
},
|
| 481 |
+
{
|
| 482 |
+
"epoch": 0.7670454545454546,
|
| 483 |
+
"grad_norm": 0.03514974255734628,
|
| 484 |
+
"learning_rate": 7.857355122839673e-08,
|
| 485 |
+
"logits/chosen": -1.7121340036392212,
|
| 486 |
+
"logits/rejected": 2.091160535812378,
|
| 487 |
+
"logps/chosen": -599.9885864257812,
|
| 488 |
+
"logps/rejected": -7744.3994140625,
|
| 489 |
+
"loss": 0.0007,
|
| 490 |
+
"rewards/accuracies": 1.0,
|
| 491 |
+
"rewards/chosen": -2.191390037536621,
|
| 492 |
+
"rewards/margins": 68.56300354003906,
|
| 493 |
+
"rewards/rejected": -70.75439453125,
|
| 494 |
+
"step": 270
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 0.7954545454545454,
|
| 498 |
+
"grad_norm": 0.1441611887831923,
|
| 499 |
+
"learning_rate": 6.135884496044244e-08,
|
| 500 |
+
"logits/chosen": -1.670058012008667,
|
| 501 |
+
"logits/rejected": 2.3099844455718994,
|
| 502 |
+
"logps/chosen": -586.2249145507812,
|
| 503 |
+
"logps/rejected": -7608.43212890625,
|
| 504 |
+
"loss": 0.0011,
|
| 505 |
+
"rewards/accuracies": 1.0,
|
| 506 |
+
"rewards/chosen": -2.070519208908081,
|
| 507 |
+
"rewards/margins": 67.56490325927734,
|
| 508 |
+
"rewards/rejected": -69.63542175292969,
|
| 509 |
+
"step": 280
|
| 510 |
+
},
|
| 511 |
+
{
|
| 512 |
+
"epoch": 0.8238636363636364,
|
| 513 |
+
"grad_norm": 0.02043160544641207,
|
| 514 |
+
"learning_rate": 4.600710195020982e-08,
|
| 515 |
+
"logits/chosen": -1.508570909500122,
|
| 516 |
+
"logits/rejected": 2.35508131980896,
|
| 517 |
+
"logps/chosen": -602.7320556640625,
|
| 518 |
+
"logps/rejected": -7490.234375,
|
| 519 |
+
"loss": 0.0026,
|
| 520 |
+
"rewards/accuracies": 1.0,
|
| 521 |
+
"rewards/chosen": -2.135887622833252,
|
| 522 |
+
"rewards/margins": 66.35386657714844,
|
| 523 |
+
"rewards/rejected": -68.48976135253906,
|
| 524 |
+
"step": 290
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 0.8522727272727273,
|
| 528 |
+
"grad_norm": 0.6511993649543483,
|
| 529 |
+
"learning_rate": 3.2669931390104374e-08,
|
| 530 |
+
"logits/chosen": -1.581496000289917,
|
| 531 |
+
"logits/rejected": 2.216219186782837,
|
| 532 |
+
"logps/chosen": -553.2108154296875,
|
| 533 |
+
"logps/rejected": -8122.2275390625,
|
| 534 |
+
"loss": 0.0009,
|
| 535 |
+
"rewards/accuracies": 1.0,
|
| 536 |
+
"rewards/chosen": -1.969562292098999,
|
| 537 |
+
"rewards/margins": 73.0444564819336,
|
| 538 |
+
"rewards/rejected": -75.0140151977539,
|
| 539 |
+
"step": 300
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 0.8522727272727273,
|
| 543 |
+
"eval_logits/chosen": -1.4073354005813599,
|
| 544 |
+
"eval_logits/rejected": 2.322908639907837,
|
| 545 |
+
"eval_logps/chosen": -579.916259765625,
|
| 546 |
+
"eval_logps/rejected": -7770.10595703125,
|
| 547 |
+
"eval_loss": 6.592821591766551e-05,
|
| 548 |
+
"eval_rewards/accuracies": 1.0,
|
| 549 |
+
"eval_rewards/chosen": -2.129345178604126,
|
| 550 |
+
"eval_rewards/margins": 69.1638412475586,
|
| 551 |
+
"eval_rewards/rejected": -71.29317474365234,
|
| 552 |
+
"eval_runtime": 193.1056,
|
| 553 |
+
"eval_samples_per_second": 20.222,
|
| 554 |
+
"eval_steps_per_second": 0.321,
|
| 555 |
+
"step": 300
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 0.8806818181818182,
|
| 559 |
+
"grad_norm": 0.14250781664198195,
|
| 560 |
+
"learning_rate": 2.147904716149135e-08,
|
| 561 |
+
"logits/chosen": -1.3581085205078125,
|
| 562 |
+
"logits/rejected": 2.2543134689331055,
|
| 563 |
+
"logps/chosen": -562.1097412109375,
|
| 564 |
+
"logps/rejected": -7898.74072265625,
|
| 565 |
+
"loss": 0.0016,
|
| 566 |
+
"rewards/accuracies": 1.0,
|
| 567 |
+
"rewards/chosen": -2.0004312992095947,
|
| 568 |
+
"rewards/margins": 69.9819564819336,
|
| 569 |
+
"rewards/rejected": -71.98239135742188,
|
| 570 |
+
"step": 310
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"epoch": 0.9090909090909091,
|
| 574 |
+
"grad_norm": 0.015240907682121196,
|
| 575 |
+
"learning_rate": 1.254496706805433e-08,
|
| 576 |
+
"logits/chosen": -1.5214873552322388,
|
| 577 |
+
"logits/rejected": 2.192413806915283,
|
| 578 |
+
"logps/chosen": -591.5843505859375,
|
| 579 |
+
"logps/rejected": -7921.0029296875,
|
| 580 |
+
"loss": 0.0005,
|
| 581 |
+
"rewards/accuracies": 1.0,
|
| 582 |
+
"rewards/chosen": -2.089113235473633,
|
| 583 |
+
"rewards/margins": 70.60108947753906,
|
| 584 |
+
"rewards/rejected": -72.69020080566406,
|
| 585 |
+
"step": 320
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 0.9375,
|
| 589 |
+
"grad_norm": 0.1617057843470206,
|
| 590 |
+
"learning_rate": 5.955921395237318e-09,
|
| 591 |
+
"logits/chosen": -1.4116395711898804,
|
| 592 |
+
"logits/rejected": 2.1853275299072266,
|
| 593 |
+
"logps/chosen": -551.7928466796875,
|
| 594 |
+
"logps/rejected": -7996.5966796875,
|
| 595 |
+
"loss": 0.0004,
|
| 596 |
+
"rewards/accuracies": 1.0,
|
| 597 |
+
"rewards/chosen": -1.9652475118637085,
|
| 598 |
+
"rewards/margins": 71.52317810058594,
|
| 599 |
+
"rewards/rejected": -73.4884262084961,
|
| 600 |
+
"step": 330
|
| 601 |
+
},
|
| 602 |
+
{
|
| 603 |
+
"epoch": 0.9659090909090909,
|
| 604 |
+
"grad_norm": 0.18410596009005747,
|
| 605 |
+
"learning_rate": 1.7769815745066474e-09,
|
| 606 |
+
"logits/chosen": -1.6253635883331299,
|
| 607 |
+
"logits/rejected": 2.031024217605591,
|
| 608 |
+
"logps/chosen": -567.4468994140625,
|
| 609 |
+
"logps/rejected": -7287.84130859375,
|
| 610 |
+
"loss": 0.0006,
|
| 611 |
+
"rewards/accuracies": 1.0,
|
| 612 |
+
"rewards/chosen": -1.9846607446670532,
|
| 613 |
+
"rewards/margins": 64.80608367919922,
|
| 614 |
+
"rewards/rejected": -66.79075622558594,
|
| 615 |
+
"step": 340
|
| 616 |
+
},
|
| 617 |
+
{
|
| 618 |
+
"epoch": 0.9943181818181818,
|
| 619 |
+
"grad_norm": 1.9094053325335274,
|
| 620 |
+
"learning_rate": 4.9417557483610875e-11,
|
| 621 |
+
"logits/chosen": -1.3966957330703735,
|
| 622 |
+
"logits/rejected": 2.161142587661743,
|
| 623 |
+
"logps/chosen": -596.6432495117188,
|
| 624 |
+
"logps/rejected": -7771.0029296875,
|
| 625 |
+
"loss": 0.0015,
|
| 626 |
+
"rewards/accuracies": 1.0,
|
| 627 |
+
"rewards/chosen": -2.2237548828125,
|
| 628 |
+
"rewards/margins": 69.01069641113281,
|
| 629 |
+
"rewards/rejected": -71.23445129394531,
|
| 630 |
+
"step": 350
|
| 631 |
+
},
|
| 632 |
+
{
|
| 633 |
+
"epoch": 0.9943181818181818,
|
| 634 |
+
"eval_logits/chosen": -1.3155968189239502,
|
| 635 |
+
"eval_logits/rejected": 2.341789960861206,
|
| 636 |
+
"eval_logps/chosen": -582.794677734375,
|
| 637 |
+
"eval_logps/rejected": -7854.4189453125,
|
| 638 |
+
"eval_loss": 6.306579598458484e-05,
|
| 639 |
+
"eval_rewards/accuracies": 1.0,
|
| 640 |
+
"eval_rewards/chosen": -2.1581296920776367,
|
| 641 |
+
"eval_rewards/margins": 69.9781723022461,
|
| 642 |
+
"eval_rewards/rejected": -72.13629150390625,
|
| 643 |
+
"eval_runtime": 193.0578,
|
| 644 |
+
"eval_samples_per_second": 20.227,
|
| 645 |
+
"eval_steps_per_second": 0.321,
|
| 646 |
+
"step": 350
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"epoch": 1.0,
|
| 650 |
+
"step": 352,
|
| 651 |
+
"total_flos": 0.0,
|
| 652 |
+
"train_loss": 0.04425002944373608,
|
| 653 |
+
"train_runtime": 10145.5361,
|
| 654 |
+
"train_samples_per_second": 4.435,
|
| 655 |
+
"train_steps_per_second": 0.035
|
| 656 |
+
}
|
| 657 |
+
],
|
| 658 |
+
"logging_steps": 10,
|
| 659 |
+
"max_steps": 352,
|
| 660 |
+
"num_input_tokens_seen": 0,
|
| 661 |
+
"num_train_epochs": 1,
|
| 662 |
+
"save_steps": 100,
|
| 663 |
+
"stateful_callbacks": {
|
| 664 |
+
"TrainerControl": {
|
| 665 |
+
"args": {
|
| 666 |
+
"should_epoch_stop": false,
|
| 667 |
+
"should_evaluate": false,
|
| 668 |
+
"should_log": false,
|
| 669 |
+
"should_save": true,
|
| 670 |
+
"should_training_stop": true
|
| 671 |
+
},
|
| 672 |
+
"attributes": {}
|
| 673 |
+
}
|
| 674 |
+
},
|
| 675 |
+
"total_flos": 0.0,
|
| 676 |
+
"train_batch_size": 8,
|
| 677 |
+
"trial_name": null,
|
| 678 |
+
"trial_params": null
|
| 679 |
+
}
|