Model save
Browse files- README.md +68 -0
- all_results.json +9 -0
- generation_config.json +6 -0
- train_results.json +9 -0
- trainer_state.json +1736 -0
README.md
ADDED
|
@@ -0,0 +1,68 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: alignment-handbook/zephyr-7b-sft-full
|
| 3 |
+
library_name: transformers
|
| 4 |
+
model_name: ultrafeedbackSkyworkAgree_alignmentZephyr7BSftFull_sdpo_score_ebs256_lr1e-07_0
|
| 5 |
+
tags:
|
| 6 |
+
- generated_from_trainer
|
| 7 |
+
- trl
|
| 8 |
+
- dpo
|
| 9 |
+
licence: license
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
# Model Card for ultrafeedbackSkyworkAgree_alignmentZephyr7BSftFull_sdpo_score_ebs256_lr1e-07_0
|
| 13 |
+
|
| 14 |
+
This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full).
|
| 15 |
+
It has been trained using [TRL](https://github.com/huggingface/trl).
|
| 16 |
+
|
| 17 |
+
## Quick start
|
| 18 |
+
|
| 19 |
+
```python
|
| 20 |
+
from transformers import pipeline
|
| 21 |
+
|
| 22 |
+
question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
|
| 23 |
+
generator = pipeline("text-generation", model="YuchenLi01/ultrafeedbackSkyworkAgree_alignmentZephyr7BSftFull_sdpo_score_ebs256_lr1e-07_0", device="cuda")
|
| 24 |
+
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
|
| 25 |
+
print(output["generated_text"])
|
| 26 |
+
```
|
| 27 |
+
|
| 28 |
+
## Training procedure
|
| 29 |
+
|
| 30 |
+
[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/yuchenl4/lmpref/runs/olden_king_smkymkmb3c_0)
|
| 31 |
+
|
| 32 |
+
This model was trained with DPO, a method introduced in [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://huggingface.co/papers/2305.18290).
|
| 33 |
+
|
| 34 |
+
### Framework versions
|
| 35 |
+
|
| 36 |
+
- TRL: 0.12.0
|
| 37 |
+
- Transformers: 4.46.2
|
| 38 |
+
- Pytorch: 2.3.0
|
| 39 |
+
- Datasets: 3.1.0
|
| 40 |
+
- Tokenizers: 0.20.3
|
| 41 |
+
|
| 42 |
+
## Citations
|
| 43 |
+
|
| 44 |
+
Cite DPO as:
|
| 45 |
+
|
| 46 |
+
```bibtex
|
| 47 |
+
@inproceedings{rafailov2023direct,
|
| 48 |
+
title = {{Direct Preference Optimization: Your Language Model is Secretly a Reward Model}},
|
| 49 |
+
author = {Rafael Rafailov and Archit Sharma and Eric Mitchell and Christopher D. Manning and Stefano Ermon and Chelsea Finn},
|
| 50 |
+
year = 2023,
|
| 51 |
+
booktitle = {Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023, NeurIPS 2023, New Orleans, LA, USA, December 10 - 16, 2023},
|
| 52 |
+
url = {http://papers.nips.cc/paper_files/paper/2023/hash/a85b405ed65c6477a4fe8302b5e06ce7-Abstract-Conference.html},
|
| 53 |
+
editor = {Alice Oh and Tristan Naumann and Amir Globerson and Kate Saenko and Moritz Hardt and Sergey Levine},
|
| 54 |
+
}
|
| 55 |
+
```
|
| 56 |
+
|
| 57 |
+
Cite TRL as:
|
| 58 |
+
|
| 59 |
+
```bibtex
|
| 60 |
+
@misc{vonwerra2022trl,
|
| 61 |
+
title = {{TRL: Transformer Reinforcement Learning}},
|
| 62 |
+
author = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallouédec},
|
| 63 |
+
year = 2020,
|
| 64 |
+
journal = {GitHub repository},
|
| 65 |
+
publisher = {GitHub},
|
| 66 |
+
howpublished = {\url{https://github.com/huggingface/trl}}
|
| 67 |
+
}
|
| 68 |
+
```
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9971988795518207,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.6403918319873596,
|
| 5 |
+
"train_runtime": 36690.4382,
|
| 6 |
+
"train_samples": 45608,
|
| 7 |
+
"train_samples_per_second": 1.243,
|
| 8 |
+
"train_steps_per_second": 0.005
|
| 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": "4.46.2"
|
| 6 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9971988795518207,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.6403918319873596,
|
| 5 |
+
"train_runtime": 36690.4382,
|
| 6 |
+
"train_samples": 45608,
|
| 7 |
+
"train_samples_per_second": 1.243,
|
| 8 |
+
"train_steps_per_second": 0.005
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,1736 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.9971988795518207,
|
| 5 |
+
"eval_steps": 2,
|
| 6 |
+
"global_step": 178,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.0056022408963585435,
|
| 13 |
+
"grad_norm": 7.67260987802389,
|
| 14 |
+
"learning_rate": 5.555555555555555e-09,
|
| 15 |
+
"logits/chosen": -3.109375,
|
| 16 |
+
"logits/rejected": -3.1875,
|
| 17 |
+
"logps/chosen": -302.0,
|
| 18 |
+
"logps/rejected": -274.0,
|
| 19 |
+
"loss": 0.6914,
|
| 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": 0.011204481792717087,
|
| 28 |
+
"eval_logits/chosen": -3.171875,
|
| 29 |
+
"eval_logits/rejected": -3.21875,
|
| 30 |
+
"eval_logps/chosen": -342.0,
|
| 31 |
+
"eval_logps/rejected": -284.0,
|
| 32 |
+
"eval_loss": 0.6922929883003235,
|
| 33 |
+
"eval_rewards/accuracies": 0.1927083283662796,
|
| 34 |
+
"eval_rewards/chosen": 5.5789947509765625e-05,
|
| 35 |
+
"eval_rewards/margins": -0.0003662109375,
|
| 36 |
+
"eval_rewards/rejected": 0.00041961669921875,
|
| 37 |
+
"eval_runtime": 133.6177,
|
| 38 |
+
"eval_samples_per_second": 11.129,
|
| 39 |
+
"eval_steps_per_second": 0.18,
|
| 40 |
+
"step": 2
|
| 41 |
+
},
|
| 42 |
+
{
|
| 43 |
+
"epoch": 0.022408963585434174,
|
| 44 |
+
"eval_logits/chosen": -3.171875,
|
| 45 |
+
"eval_logits/rejected": -3.21875,
|
| 46 |
+
"eval_logps/chosen": -342.0,
|
| 47 |
+
"eval_logps/rejected": -284.0,
|
| 48 |
+
"eval_loss": 0.6920216083526611,
|
| 49 |
+
"eval_rewards/accuracies": 0.2083333283662796,
|
| 50 |
+
"eval_rewards/chosen": 0.0003509521484375,
|
| 51 |
+
"eval_rewards/margins": -0.000396728515625,
|
| 52 |
+
"eval_rewards/rejected": 0.0007476806640625,
|
| 53 |
+
"eval_runtime": 134.3568,
|
| 54 |
+
"eval_samples_per_second": 11.068,
|
| 55 |
+
"eval_steps_per_second": 0.179,
|
| 56 |
+
"step": 4
|
| 57 |
+
},
|
| 58 |
+
{
|
| 59 |
+
"epoch": 0.03361344537815126,
|
| 60 |
+
"eval_logits/chosen": -3.171875,
|
| 61 |
+
"eval_logits/rejected": -3.21875,
|
| 62 |
+
"eval_logps/chosen": -342.0,
|
| 63 |
+
"eval_logps/rejected": -284.0,
|
| 64 |
+
"eval_loss": 0.6921105980873108,
|
| 65 |
+
"eval_rewards/accuracies": 0.2057291716337204,
|
| 66 |
+
"eval_rewards/chosen": 0.0001201629638671875,
|
| 67 |
+
"eval_rewards/margins": 0.000484466552734375,
|
| 68 |
+
"eval_rewards/rejected": -0.0003643035888671875,
|
| 69 |
+
"eval_runtime": 134.2477,
|
| 70 |
+
"eval_samples_per_second": 11.077,
|
| 71 |
+
"eval_steps_per_second": 0.179,
|
| 72 |
+
"step": 6
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 0.04481792717086835,
|
| 76 |
+
"eval_logits/chosen": -3.171875,
|
| 77 |
+
"eval_logits/rejected": -3.21875,
|
| 78 |
+
"eval_logps/chosen": -342.0,
|
| 79 |
+
"eval_logps/rejected": -284.0,
|
| 80 |
+
"eval_loss": 0.6920554041862488,
|
| 81 |
+
"eval_rewards/accuracies": 0.21875,
|
| 82 |
+
"eval_rewards/chosen": -9.679794311523438e-05,
|
| 83 |
+
"eval_rewards/margins": -0.00018310546875,
|
| 84 |
+
"eval_rewards/rejected": 8.487701416015625e-05,
|
| 85 |
+
"eval_runtime": 134.4853,
|
| 86 |
+
"eval_samples_per_second": 11.057,
|
| 87 |
+
"eval_steps_per_second": 0.178,
|
| 88 |
+
"step": 8
|
| 89 |
+
},
|
| 90 |
+
{
|
| 91 |
+
"epoch": 0.056022408963585436,
|
| 92 |
+
"grad_norm": 7.0192772162926635,
|
| 93 |
+
"learning_rate": 5.5555555555555555e-08,
|
| 94 |
+
"logits/chosen": -3.1875,
|
| 95 |
+
"logits/rejected": -3.203125,
|
| 96 |
+
"logps/chosen": -304.0,
|
| 97 |
+
"logps/rejected": -278.0,
|
| 98 |
+
"loss": 0.6917,
|
| 99 |
+
"rewards/accuracies": 0.1996527761220932,
|
| 100 |
+
"rewards/chosen": 0.000156402587890625,
|
| 101 |
+
"rewards/margins": 4.744529724121094e-05,
|
| 102 |
+
"rewards/rejected": 0.00010824203491210938,
|
| 103 |
+
"step": 10
|
| 104 |
+
},
|
| 105 |
+
{
|
| 106 |
+
"epoch": 0.056022408963585436,
|
| 107 |
+
"eval_logits/chosen": -3.171875,
|
| 108 |
+
"eval_logits/rejected": -3.21875,
|
| 109 |
+
"eval_logps/chosen": -342.0,
|
| 110 |
+
"eval_logps/rejected": -284.0,
|
| 111 |
+
"eval_loss": 0.692084014415741,
|
| 112 |
+
"eval_rewards/accuracies": 0.2395833283662796,
|
| 113 |
+
"eval_rewards/chosen": -0.0001068115234375,
|
| 114 |
+
"eval_rewards/margins": 0.000270843505859375,
|
| 115 |
+
"eval_rewards/rejected": -0.000377655029296875,
|
| 116 |
+
"eval_runtime": 134.2041,
|
| 117 |
+
"eval_samples_per_second": 11.08,
|
| 118 |
+
"eval_steps_per_second": 0.179,
|
| 119 |
+
"step": 10
|
| 120 |
+
},
|
| 121 |
+
{
|
| 122 |
+
"epoch": 0.06722689075630252,
|
| 123 |
+
"eval_logits/chosen": -3.171875,
|
| 124 |
+
"eval_logits/rejected": -3.21875,
|
| 125 |
+
"eval_logps/chosen": -342.0,
|
| 126 |
+
"eval_logps/rejected": -284.0,
|
| 127 |
+
"eval_loss": 0.6917845010757446,
|
| 128 |
+
"eval_rewards/accuracies": 0.203125,
|
| 129 |
+
"eval_rewards/chosen": 0.00049591064453125,
|
| 130 |
+
"eval_rewards/margins": 0.000331878662109375,
|
| 131 |
+
"eval_rewards/rejected": 0.00016117095947265625,
|
| 132 |
+
"eval_runtime": 134.0059,
|
| 133 |
+
"eval_samples_per_second": 11.097,
|
| 134 |
+
"eval_steps_per_second": 0.179,
|
| 135 |
+
"step": 12
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"epoch": 0.0784313725490196,
|
| 139 |
+
"eval_logits/chosen": -3.171875,
|
| 140 |
+
"eval_logits/rejected": -3.21875,
|
| 141 |
+
"eval_logps/chosen": -342.0,
|
| 142 |
+
"eval_logps/rejected": -284.0,
|
| 143 |
+
"eval_loss": 0.6914895176887512,
|
| 144 |
+
"eval_rewards/accuracies": 0.2317708283662796,
|
| 145 |
+
"eval_rewards/chosen": 0.000972747802734375,
|
| 146 |
+
"eval_rewards/margins": 0.000957489013671875,
|
| 147 |
+
"eval_rewards/rejected": 1.621246337890625e-05,
|
| 148 |
+
"eval_runtime": 134.116,
|
| 149 |
+
"eval_samples_per_second": 11.087,
|
| 150 |
+
"eval_steps_per_second": 0.179,
|
| 151 |
+
"step": 14
|
| 152 |
+
},
|
| 153 |
+
{
|
| 154 |
+
"epoch": 0.0896358543417367,
|
| 155 |
+
"eval_logits/chosen": -3.171875,
|
| 156 |
+
"eval_logits/rejected": -3.21875,
|
| 157 |
+
"eval_logps/chosen": -342.0,
|
| 158 |
+
"eval_logps/rejected": -284.0,
|
| 159 |
+
"eval_loss": 0.6913825869560242,
|
| 160 |
+
"eval_rewards/accuracies": 0.2890625,
|
| 161 |
+
"eval_rewards/chosen": 0.000621795654296875,
|
| 162 |
+
"eval_rewards/margins": 0.00136566162109375,
|
| 163 |
+
"eval_rewards/rejected": -0.000743865966796875,
|
| 164 |
+
"eval_runtime": 133.0747,
|
| 165 |
+
"eval_samples_per_second": 11.174,
|
| 166 |
+
"eval_steps_per_second": 0.18,
|
| 167 |
+
"step": 16
|
| 168 |
+
},
|
| 169 |
+
{
|
| 170 |
+
"epoch": 0.10084033613445378,
|
| 171 |
+
"eval_logits/chosen": -3.15625,
|
| 172 |
+
"eval_logits/rejected": -3.21875,
|
| 173 |
+
"eval_logps/chosen": -342.0,
|
| 174 |
+
"eval_logps/rejected": -284.0,
|
| 175 |
+
"eval_loss": 0.6909087896347046,
|
| 176 |
+
"eval_rewards/accuracies": 0.328125,
|
| 177 |
+
"eval_rewards/chosen": 0.00168609619140625,
|
| 178 |
+
"eval_rewards/margins": 0.002593994140625,
|
| 179 |
+
"eval_rewards/rejected": -0.00091552734375,
|
| 180 |
+
"eval_runtime": 133.8874,
|
| 181 |
+
"eval_samples_per_second": 11.106,
|
| 182 |
+
"eval_steps_per_second": 0.179,
|
| 183 |
+
"step": 18
|
| 184 |
+
},
|
| 185 |
+
{
|
| 186 |
+
"epoch": 0.11204481792717087,
|
| 187 |
+
"grad_norm": 7.546692092094965,
|
| 188 |
+
"learning_rate": 9.996145181203615e-08,
|
| 189 |
+
"logits/chosen": -3.15625,
|
| 190 |
+
"logits/rejected": -3.1875,
|
| 191 |
+
"logps/chosen": -324.0,
|
| 192 |
+
"logps/rejected": -278.0,
|
| 193 |
+
"loss": 0.6915,
|
| 194 |
+
"rewards/accuracies": 0.2593750059604645,
|
| 195 |
+
"rewards/chosen": 0.000823974609375,
|
| 196 |
+
"rewards/margins": 0.0010986328125,
|
| 197 |
+
"rewards/rejected": -0.000274658203125,
|
| 198 |
+
"step": 20
|
| 199 |
+
},
|
| 200 |
+
{
|
| 201 |
+
"epoch": 0.11204481792717087,
|
| 202 |
+
"eval_logits/chosen": -3.15625,
|
| 203 |
+
"eval_logits/rejected": -3.203125,
|
| 204 |
+
"eval_logps/chosen": -342.0,
|
| 205 |
+
"eval_logps/rejected": -284.0,
|
| 206 |
+
"eval_loss": 0.6904590129852295,
|
| 207 |
+
"eval_rewards/accuracies": 0.4270833432674408,
|
| 208 |
+
"eval_rewards/chosen": 0.0024261474609375,
|
| 209 |
+
"eval_rewards/margins": 0.00421142578125,
|
| 210 |
+
"eval_rewards/rejected": -0.0017852783203125,
|
| 211 |
+
"eval_runtime": 134.0613,
|
| 212 |
+
"eval_samples_per_second": 11.092,
|
| 213 |
+
"eval_steps_per_second": 0.179,
|
| 214 |
+
"step": 20
|
| 215 |
+
},
|
| 216 |
+
{
|
| 217 |
+
"epoch": 0.12324929971988796,
|
| 218 |
+
"eval_logits/chosen": -3.15625,
|
| 219 |
+
"eval_logits/rejected": -3.203125,
|
| 220 |
+
"eval_logps/chosen": -342.0,
|
| 221 |
+
"eval_logps/rejected": -284.0,
|
| 222 |
+
"eval_loss": 0.6897143125534058,
|
| 223 |
+
"eval_rewards/accuracies": 0.4947916567325592,
|
| 224 |
+
"eval_rewards/chosen": 0.003570556640625,
|
| 225 |
+
"eval_rewards/margins": 0.00640869140625,
|
| 226 |
+
"eval_rewards/rejected": -0.002838134765625,
|
| 227 |
+
"eval_runtime": 134.4842,
|
| 228 |
+
"eval_samples_per_second": 11.057,
|
| 229 |
+
"eval_steps_per_second": 0.178,
|
| 230 |
+
"step": 22
|
| 231 |
+
},
|
| 232 |
+
{
|
| 233 |
+
"epoch": 0.13445378151260504,
|
| 234 |
+
"eval_logits/chosen": -3.15625,
|
| 235 |
+
"eval_logits/rejected": -3.203125,
|
| 236 |
+
"eval_logps/chosen": -342.0,
|
| 237 |
+
"eval_logps/rejected": -284.0,
|
| 238 |
+
"eval_loss": 0.6890442371368408,
|
| 239 |
+
"eval_rewards/accuracies": 0.5364583134651184,
|
| 240 |
+
"eval_rewards/chosen": 0.0052490234375,
|
| 241 |
+
"eval_rewards/margins": 0.00836181640625,
|
| 242 |
+
"eval_rewards/rejected": -0.0031280517578125,
|
| 243 |
+
"eval_runtime": 133.679,
|
| 244 |
+
"eval_samples_per_second": 11.124,
|
| 245 |
+
"eval_steps_per_second": 0.18,
|
| 246 |
+
"step": 24
|
| 247 |
+
},
|
| 248 |
+
{
|
| 249 |
+
"epoch": 0.14565826330532214,
|
| 250 |
+
"eval_logits/chosen": -3.15625,
|
| 251 |
+
"eval_logits/rejected": -3.203125,
|
| 252 |
+
"eval_logps/chosen": -342.0,
|
| 253 |
+
"eval_logps/rejected": -284.0,
|
| 254 |
+
"eval_loss": 0.6882352232933044,
|
| 255 |
+
"eval_rewards/accuracies": 0.5625,
|
| 256 |
+
"eval_rewards/chosen": 0.006683349609375,
|
| 257 |
+
"eval_rewards/margins": 0.01025390625,
|
| 258 |
+
"eval_rewards/rejected": -0.0035552978515625,
|
| 259 |
+
"eval_runtime": 134.2487,
|
| 260 |
+
"eval_samples_per_second": 11.076,
|
| 261 |
+
"eval_steps_per_second": 0.179,
|
| 262 |
+
"step": 26
|
| 263 |
+
},
|
| 264 |
+
{
|
| 265 |
+
"epoch": 0.1568627450980392,
|
| 266 |
+
"eval_logits/chosen": -3.15625,
|
| 267 |
+
"eval_logits/rejected": -3.203125,
|
| 268 |
+
"eval_logps/chosen": -342.0,
|
| 269 |
+
"eval_logps/rejected": -284.0,
|
| 270 |
+
"eval_loss": 0.6876863241195679,
|
| 271 |
+
"eval_rewards/accuracies": 0.6015625,
|
| 272 |
+
"eval_rewards/chosen": 0.006866455078125,
|
| 273 |
+
"eval_rewards/margins": 0.0118408203125,
|
| 274 |
+
"eval_rewards/rejected": -0.004974365234375,
|
| 275 |
+
"eval_runtime": 134.3926,
|
| 276 |
+
"eval_samples_per_second": 11.065,
|
| 277 |
+
"eval_steps_per_second": 0.179,
|
| 278 |
+
"step": 28
|
| 279 |
+
},
|
| 280 |
+
{
|
| 281 |
+
"epoch": 0.16806722689075632,
|
| 282 |
+
"grad_norm": 6.890557992493745,
|
| 283 |
+
"learning_rate": 9.861849601988382e-08,
|
| 284 |
+
"logits/chosen": -3.171875,
|
| 285 |
+
"logits/rejected": -3.234375,
|
| 286 |
+
"logps/chosen": -340.0,
|
| 287 |
+
"logps/rejected": -266.0,
|
| 288 |
+
"loss": 0.6887,
|
| 289 |
+
"rewards/accuracies": 0.565625011920929,
|
| 290 |
+
"rewards/chosen": 0.005950927734375,
|
| 291 |
+
"rewards/margins": 0.00897216796875,
|
| 292 |
+
"rewards/rejected": -0.0030517578125,
|
| 293 |
+
"step": 30
|
| 294 |
+
},
|
| 295 |
+
{
|
| 296 |
+
"epoch": 0.16806722689075632,
|
| 297 |
+
"eval_logits/chosen": -3.15625,
|
| 298 |
+
"eval_logits/rejected": -3.203125,
|
| 299 |
+
"eval_logps/chosen": -342.0,
|
| 300 |
+
"eval_logps/rejected": -284.0,
|
| 301 |
+
"eval_loss": 0.6867438554763794,
|
| 302 |
+
"eval_rewards/accuracies": 0.6145833134651184,
|
| 303 |
+
"eval_rewards/chosen": 0.007720947265625,
|
| 304 |
+
"eval_rewards/margins": 0.01300048828125,
|
| 305 |
+
"eval_rewards/rejected": -0.005279541015625,
|
| 306 |
+
"eval_runtime": 134.4984,
|
| 307 |
+
"eval_samples_per_second": 11.056,
|
| 308 |
+
"eval_steps_per_second": 0.178,
|
| 309 |
+
"step": 30
|
| 310 |
+
},
|
| 311 |
+
{
|
| 312 |
+
"epoch": 0.1792717086834734,
|
| 313 |
+
"eval_logits/chosen": -3.15625,
|
| 314 |
+
"eval_logits/rejected": -3.203125,
|
| 315 |
+
"eval_logps/chosen": -340.0,
|
| 316 |
+
"eval_logps/rejected": -284.0,
|
| 317 |
+
"eval_loss": 0.6854631304740906,
|
| 318 |
+
"eval_rewards/accuracies": 0.6588541865348816,
|
| 319 |
+
"eval_rewards/chosen": 0.00958251953125,
|
| 320 |
+
"eval_rewards/margins": 0.0169677734375,
|
| 321 |
+
"eval_rewards/rejected": -0.007415771484375,
|
| 322 |
+
"eval_runtime": 133.0213,
|
| 323 |
+
"eval_samples_per_second": 11.179,
|
| 324 |
+
"eval_steps_per_second": 0.18,
|
| 325 |
+
"step": 32
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 0.19047619047619047,
|
| 329 |
+
"eval_logits/chosen": -3.15625,
|
| 330 |
+
"eval_logits/rejected": -3.203125,
|
| 331 |
+
"eval_logps/chosen": -340.0,
|
| 332 |
+
"eval_logps/rejected": -284.0,
|
| 333 |
+
"eval_loss": 0.6838385462760925,
|
| 334 |
+
"eval_rewards/accuracies": 0.6770833134651184,
|
| 335 |
+
"eval_rewards/chosen": 0.01116943359375,
|
| 336 |
+
"eval_rewards/margins": 0.0201416015625,
|
| 337 |
+
"eval_rewards/rejected": -0.009033203125,
|
| 338 |
+
"eval_runtime": 135.0159,
|
| 339 |
+
"eval_samples_per_second": 11.014,
|
| 340 |
+
"eval_steps_per_second": 0.178,
|
| 341 |
+
"step": 34
|
| 342 |
+
},
|
| 343 |
+
{
|
| 344 |
+
"epoch": 0.20168067226890757,
|
| 345 |
+
"eval_logits/chosen": -3.140625,
|
| 346 |
+
"eval_logits/rejected": -3.1875,
|
| 347 |
+
"eval_logps/chosen": -340.0,
|
| 348 |
+
"eval_logps/rejected": -284.0,
|
| 349 |
+
"eval_loss": 0.6825665235519409,
|
| 350 |
+
"eval_rewards/accuracies": 0.6796875,
|
| 351 |
+
"eval_rewards/chosen": 0.01300048828125,
|
| 352 |
+
"eval_rewards/margins": 0.023193359375,
|
| 353 |
+
"eval_rewards/rejected": -0.01019287109375,
|
| 354 |
+
"eval_runtime": 134.4265,
|
| 355 |
+
"eval_samples_per_second": 11.062,
|
| 356 |
+
"eval_steps_per_second": 0.179,
|
| 357 |
+
"step": 36
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 0.21288515406162464,
|
| 361 |
+
"eval_logits/chosen": -3.140625,
|
| 362 |
+
"eval_logits/rejected": -3.1875,
|
| 363 |
+
"eval_logps/chosen": -340.0,
|
| 364 |
+
"eval_logps/rejected": -284.0,
|
| 365 |
+
"eval_loss": 0.6815204620361328,
|
| 366 |
+
"eval_rewards/accuracies": 0.6875,
|
| 367 |
+
"eval_rewards/chosen": 0.0137939453125,
|
| 368 |
+
"eval_rewards/margins": 0.025634765625,
|
| 369 |
+
"eval_rewards/rejected": -0.01177978515625,
|
| 370 |
+
"eval_runtime": 134.3747,
|
| 371 |
+
"eval_samples_per_second": 11.066,
|
| 372 |
+
"eval_steps_per_second": 0.179,
|
| 373 |
+
"step": 38
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 0.22408963585434175,
|
| 377 |
+
"grad_norm": 7.330771538752886,
|
| 378 |
+
"learning_rate": 9.540715869125407e-08,
|
| 379 |
+
"logits/chosen": -3.125,
|
| 380 |
+
"logits/rejected": -3.125,
|
| 381 |
+
"logps/chosen": -308.0,
|
| 382 |
+
"logps/rejected": -276.0,
|
| 383 |
+
"loss": 0.6837,
|
| 384 |
+
"rewards/accuracies": 0.645312488079071,
|
| 385 |
+
"rewards/chosen": 0.01116943359375,
|
| 386 |
+
"rewards/margins": 0.0194091796875,
|
| 387 |
+
"rewards/rejected": -0.00823974609375,
|
| 388 |
+
"step": 40
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 0.22408963585434175,
|
| 392 |
+
"eval_logits/chosen": -3.140625,
|
| 393 |
+
"eval_logits/rejected": -3.1875,
|
| 394 |
+
"eval_logps/chosen": -340.0,
|
| 395 |
+
"eval_logps/rejected": -284.0,
|
| 396 |
+
"eval_loss": 0.6800552010536194,
|
| 397 |
+
"eval_rewards/accuracies": 0.6979166865348816,
|
| 398 |
+
"eval_rewards/chosen": 0.01556396484375,
|
| 399 |
+
"eval_rewards/margins": 0.029296875,
|
| 400 |
+
"eval_rewards/rejected": -0.0137939453125,
|
| 401 |
+
"eval_runtime": 134.815,
|
| 402 |
+
"eval_samples_per_second": 11.03,
|
| 403 |
+
"eval_steps_per_second": 0.178,
|
| 404 |
+
"step": 40
|
| 405 |
+
},
|
| 406 |
+
{
|
| 407 |
+
"epoch": 0.23529411764705882,
|
| 408 |
+
"eval_logits/chosen": -3.140625,
|
| 409 |
+
"eval_logits/rejected": -3.1875,
|
| 410 |
+
"eval_logps/chosen": -340.0,
|
| 411 |
+
"eval_logps/rejected": -284.0,
|
| 412 |
+
"eval_loss": 0.6788628101348877,
|
| 413 |
+
"eval_rewards/accuracies": 0.6979166865348816,
|
| 414 |
+
"eval_rewards/chosen": 0.01507568359375,
|
| 415 |
+
"eval_rewards/margins": 0.0322265625,
|
| 416 |
+
"eval_rewards/rejected": -0.01708984375,
|
| 417 |
+
"eval_runtime": 133.9485,
|
| 418 |
+
"eval_samples_per_second": 11.101,
|
| 419 |
+
"eval_steps_per_second": 0.179,
|
| 420 |
+
"step": 42
|
| 421 |
+
},
|
| 422 |
+
{
|
| 423 |
+
"epoch": 0.24649859943977592,
|
| 424 |
+
"eval_logits/chosen": -3.140625,
|
| 425 |
+
"eval_logits/rejected": -3.1875,
|
| 426 |
+
"eval_logps/chosen": -340.0,
|
| 427 |
+
"eval_logps/rejected": -286.0,
|
| 428 |
+
"eval_loss": 0.677721381187439,
|
| 429 |
+
"eval_rewards/accuracies": 0.7135416865348816,
|
| 430 |
+
"eval_rewards/chosen": 0.0152587890625,
|
| 431 |
+
"eval_rewards/margins": 0.03515625,
|
| 432 |
+
"eval_rewards/rejected": -0.0198974609375,
|
| 433 |
+
"eval_runtime": 134.206,
|
| 434 |
+
"eval_samples_per_second": 11.08,
|
| 435 |
+
"eval_steps_per_second": 0.179,
|
| 436 |
+
"step": 44
|
| 437 |
+
},
|
| 438 |
+
{
|
| 439 |
+
"epoch": 0.25770308123249297,
|
| 440 |
+
"eval_logits/chosen": -3.125,
|
| 441 |
+
"eval_logits/rejected": -3.171875,
|
| 442 |
+
"eval_logps/chosen": -340.0,
|
| 443 |
+
"eval_logps/rejected": -286.0,
|
| 444 |
+
"eval_loss": 0.6766650080680847,
|
| 445 |
+
"eval_rewards/accuracies": 0.7265625,
|
| 446 |
+
"eval_rewards/chosen": 0.01513671875,
|
| 447 |
+
"eval_rewards/margins": 0.03759765625,
|
| 448 |
+
"eval_rewards/rejected": -0.0224609375,
|
| 449 |
+
"eval_runtime": 134.1986,
|
| 450 |
+
"eval_samples_per_second": 11.081,
|
| 451 |
+
"eval_steps_per_second": 0.179,
|
| 452 |
+
"step": 46
|
| 453 |
+
},
|
| 454 |
+
{
|
| 455 |
+
"epoch": 0.2689075630252101,
|
| 456 |
+
"eval_logits/chosen": -3.125,
|
| 457 |
+
"eval_logits/rejected": -3.171875,
|
| 458 |
+
"eval_logps/chosen": -340.0,
|
| 459 |
+
"eval_logps/rejected": -286.0,
|
| 460 |
+
"eval_loss": 0.6755136847496033,
|
| 461 |
+
"eval_rewards/accuracies": 0.7161458134651184,
|
| 462 |
+
"eval_rewards/chosen": 0.0135498046875,
|
| 463 |
+
"eval_rewards/margins": 0.039794921875,
|
| 464 |
+
"eval_rewards/rejected": -0.0262451171875,
|
| 465 |
+
"eval_runtime": 133.4992,
|
| 466 |
+
"eval_samples_per_second": 11.139,
|
| 467 |
+
"eval_steps_per_second": 0.18,
|
| 468 |
+
"step": 48
|
| 469 |
+
},
|
| 470 |
+
{
|
| 471 |
+
"epoch": 0.2801120448179272,
|
| 472 |
+
"grad_norm": 6.881196622326406,
|
| 473 |
+
"learning_rate": 9.045084971874737e-08,
|
| 474 |
+
"logits/chosen": -3.125,
|
| 475 |
+
"logits/rejected": -3.1875,
|
| 476 |
+
"logps/chosen": -330.0,
|
| 477 |
+
"logps/rejected": -286.0,
|
| 478 |
+
"loss": 0.6768,
|
| 479 |
+
"rewards/accuracies": 0.739062488079071,
|
| 480 |
+
"rewards/chosen": 0.01806640625,
|
| 481 |
+
"rewards/margins": 0.0400390625,
|
| 482 |
+
"rewards/rejected": -0.02197265625,
|
| 483 |
+
"step": 50
|
| 484 |
+
},
|
| 485 |
+
{
|
| 486 |
+
"epoch": 0.2801120448179272,
|
| 487 |
+
"eval_logits/chosen": -3.125,
|
| 488 |
+
"eval_logits/rejected": -3.171875,
|
| 489 |
+
"eval_logps/chosen": -340.0,
|
| 490 |
+
"eval_logps/rejected": -286.0,
|
| 491 |
+
"eval_loss": 0.6741231679916382,
|
| 492 |
+
"eval_rewards/accuracies": 0.7317708134651184,
|
| 493 |
+
"eval_rewards/chosen": 0.0123291015625,
|
| 494 |
+
"eval_rewards/margins": 0.043701171875,
|
| 495 |
+
"eval_rewards/rejected": -0.03125,
|
| 496 |
+
"eval_runtime": 134.2614,
|
| 497 |
+
"eval_samples_per_second": 11.075,
|
| 498 |
+
"eval_steps_per_second": 0.179,
|
| 499 |
+
"step": 50
|
| 500 |
+
},
|
| 501 |
+
{
|
| 502 |
+
"epoch": 0.2913165266106443,
|
| 503 |
+
"eval_logits/chosen": -3.125,
|
| 504 |
+
"eval_logits/rejected": -3.171875,
|
| 505 |
+
"eval_logps/chosen": -340.0,
|
| 506 |
+
"eval_logps/rejected": -286.0,
|
| 507 |
+
"eval_loss": 0.672571063041687,
|
| 508 |
+
"eval_rewards/accuracies": 0.7473958134651184,
|
| 509 |
+
"eval_rewards/chosen": 0.0118408203125,
|
| 510 |
+
"eval_rewards/margins": 0.04833984375,
|
| 511 |
+
"eval_rewards/rejected": -0.036376953125,
|
| 512 |
+
"eval_runtime": 134.8474,
|
| 513 |
+
"eval_samples_per_second": 11.027,
|
| 514 |
+
"eval_steps_per_second": 0.178,
|
| 515 |
+
"step": 52
|
| 516 |
+
},
|
| 517 |
+
{
|
| 518 |
+
"epoch": 0.3025210084033613,
|
| 519 |
+
"eval_logits/chosen": -3.125,
|
| 520 |
+
"eval_logits/rejected": -3.171875,
|
| 521 |
+
"eval_logps/chosen": -340.0,
|
| 522 |
+
"eval_logps/rejected": -288.0,
|
| 523 |
+
"eval_loss": 0.6705121397972107,
|
| 524 |
+
"eval_rewards/accuracies": 0.7395833134651184,
|
| 525 |
+
"eval_rewards/chosen": 0.009765625,
|
| 526 |
+
"eval_rewards/margins": 0.052490234375,
|
| 527 |
+
"eval_rewards/rejected": -0.042724609375,
|
| 528 |
+
"eval_runtime": 133.4771,
|
| 529 |
+
"eval_samples_per_second": 11.14,
|
| 530 |
+
"eval_steps_per_second": 0.18,
|
| 531 |
+
"step": 54
|
| 532 |
+
},
|
| 533 |
+
{
|
| 534 |
+
"epoch": 0.3137254901960784,
|
| 535 |
+
"eval_logits/chosen": -3.125,
|
| 536 |
+
"eval_logits/rejected": -3.171875,
|
| 537 |
+
"eval_logps/chosen": -340.0,
|
| 538 |
+
"eval_logps/rejected": -288.0,
|
| 539 |
+
"eval_loss": 0.6688975095748901,
|
| 540 |
+
"eval_rewards/accuracies": 0.7760416865348816,
|
| 541 |
+
"eval_rewards/chosen": 0.00811767578125,
|
| 542 |
+
"eval_rewards/margins": 0.05712890625,
|
| 543 |
+
"eval_rewards/rejected": -0.049072265625,
|
| 544 |
+
"eval_runtime": 133.9848,
|
| 545 |
+
"eval_samples_per_second": 11.098,
|
| 546 |
+
"eval_steps_per_second": 0.179,
|
| 547 |
+
"step": 56
|
| 548 |
+
},
|
| 549 |
+
{
|
| 550 |
+
"epoch": 0.32492997198879553,
|
| 551 |
+
"eval_logits/chosen": -3.125,
|
| 552 |
+
"eval_logits/rejected": -3.171875,
|
| 553 |
+
"eval_logps/chosen": -342.0,
|
| 554 |
+
"eval_logps/rejected": -288.0,
|
| 555 |
+
"eval_loss": 0.6666556000709534,
|
| 556 |
+
"eval_rewards/accuracies": 0.7604166865348816,
|
| 557 |
+
"eval_rewards/chosen": 0.005767822265625,
|
| 558 |
+
"eval_rewards/margins": 0.0615234375,
|
| 559 |
+
"eval_rewards/rejected": -0.0556640625,
|
| 560 |
+
"eval_runtime": 132.6823,
|
| 561 |
+
"eval_samples_per_second": 11.207,
|
| 562 |
+
"eval_steps_per_second": 0.181,
|
| 563 |
+
"step": 58
|
| 564 |
+
},
|
| 565 |
+
{
|
| 566 |
+
"epoch": 0.33613445378151263,
|
| 567 |
+
"grad_norm": 7.102237419157038,
|
| 568 |
+
"learning_rate": 8.394003727664709e-08,
|
| 569 |
+
"logits/chosen": -3.171875,
|
| 570 |
+
"logits/rejected": -3.21875,
|
| 571 |
+
"logps/chosen": -324.0,
|
| 572 |
+
"logps/rejected": -270.0,
|
| 573 |
+
"loss": 0.6683,
|
| 574 |
+
"rewards/accuracies": 0.765625,
|
| 575 |
+
"rewards/chosen": 0.01373291015625,
|
| 576 |
+
"rewards/margins": 0.05859375,
|
| 577 |
+
"rewards/rejected": -0.044921875,
|
| 578 |
+
"step": 60
|
| 579 |
+
},
|
| 580 |
+
{
|
| 581 |
+
"epoch": 0.33613445378151263,
|
| 582 |
+
"eval_logits/chosen": -3.109375,
|
| 583 |
+
"eval_logits/rejected": -3.15625,
|
| 584 |
+
"eval_logps/chosen": -342.0,
|
| 585 |
+
"eval_logps/rejected": -290.0,
|
| 586 |
+
"eval_loss": 0.6645676493644714,
|
| 587 |
+
"eval_rewards/accuracies": 0.7734375,
|
| 588 |
+
"eval_rewards/chosen": 0.004119873046875,
|
| 589 |
+
"eval_rewards/margins": 0.06787109375,
|
| 590 |
+
"eval_rewards/rejected": -0.0634765625,
|
| 591 |
+
"eval_runtime": 134.253,
|
| 592 |
+
"eval_samples_per_second": 11.076,
|
| 593 |
+
"eval_steps_per_second": 0.179,
|
| 594 |
+
"step": 60
|
| 595 |
+
},
|
| 596 |
+
{
|
| 597 |
+
"epoch": 0.3473389355742297,
|
| 598 |
+
"eval_logits/chosen": -3.109375,
|
| 599 |
+
"eval_logits/rejected": -3.15625,
|
| 600 |
+
"eval_logps/chosen": -342.0,
|
| 601 |
+
"eval_logps/rejected": -290.0,
|
| 602 |
+
"eval_loss": 0.6622598767280579,
|
| 603 |
+
"eval_rewards/accuracies": 0.7760416865348816,
|
| 604 |
+
"eval_rewards/chosen": 0.00191497802734375,
|
| 605 |
+
"eval_rewards/margins": 0.07373046875,
|
| 606 |
+
"eval_rewards/rejected": -0.07177734375,
|
| 607 |
+
"eval_runtime": 134.3595,
|
| 608 |
+
"eval_samples_per_second": 11.067,
|
| 609 |
+
"eval_steps_per_second": 0.179,
|
| 610 |
+
"step": 62
|
| 611 |
+
},
|
| 612 |
+
{
|
| 613 |
+
"epoch": 0.3585434173669468,
|
| 614 |
+
"eval_logits/chosen": -3.109375,
|
| 615 |
+
"eval_logits/rejected": -3.15625,
|
| 616 |
+
"eval_logps/chosen": -342.0,
|
| 617 |
+
"eval_logps/rejected": -292.0,
|
| 618 |
+
"eval_loss": 0.6603178977966309,
|
| 619 |
+
"eval_rewards/accuracies": 0.7578125,
|
| 620 |
+
"eval_rewards/chosen": -0.0023193359375,
|
| 621 |
+
"eval_rewards/margins": 0.07763671875,
|
| 622 |
+
"eval_rewards/rejected": -0.080078125,
|
| 623 |
+
"eval_runtime": 135.0017,
|
| 624 |
+
"eval_samples_per_second": 11.015,
|
| 625 |
+
"eval_steps_per_second": 0.178,
|
| 626 |
+
"step": 64
|
| 627 |
+
},
|
| 628 |
+
{
|
| 629 |
+
"epoch": 0.3697478991596639,
|
| 630 |
+
"eval_logits/chosen": -3.109375,
|
| 631 |
+
"eval_logits/rejected": -3.15625,
|
| 632 |
+
"eval_logps/chosen": -342.0,
|
| 633 |
+
"eval_logps/rejected": -292.0,
|
| 634 |
+
"eval_loss": 0.6583611369132996,
|
| 635 |
+
"eval_rewards/accuracies": 0.7630208134651184,
|
| 636 |
+
"eval_rewards/chosen": -0.005096435546875,
|
| 637 |
+
"eval_rewards/margins": 0.083984375,
|
| 638 |
+
"eval_rewards/rejected": -0.0888671875,
|
| 639 |
+
"eval_runtime": 134.9993,
|
| 640 |
+
"eval_samples_per_second": 11.015,
|
| 641 |
+
"eval_steps_per_second": 0.178,
|
| 642 |
+
"step": 66
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 0.38095238095238093,
|
| 646 |
+
"eval_logits/chosen": -3.109375,
|
| 647 |
+
"eval_logits/rejected": -3.15625,
|
| 648 |
+
"eval_logps/chosen": -344.0,
|
| 649 |
+
"eval_logps/rejected": -294.0,
|
| 650 |
+
"eval_loss": 0.6562416553497314,
|
| 651 |
+
"eval_rewards/accuracies": 0.7708333134651184,
|
| 652 |
+
"eval_rewards/chosen": -0.00860595703125,
|
| 653 |
+
"eval_rewards/margins": 0.08984375,
|
| 654 |
+
"eval_rewards/rejected": -0.0986328125,
|
| 655 |
+
"eval_runtime": 134.0273,
|
| 656 |
+
"eval_samples_per_second": 11.095,
|
| 657 |
+
"eval_steps_per_second": 0.179,
|
| 658 |
+
"step": 68
|
| 659 |
+
},
|
| 660 |
+
{
|
| 661 |
+
"epoch": 0.39215686274509803,
|
| 662 |
+
"grad_norm": 6.194643384872316,
|
| 663 |
+
"learning_rate": 7.612492823579743e-08,
|
| 664 |
+
"logits/chosen": -3.09375,
|
| 665 |
+
"logits/rejected": -3.140625,
|
| 666 |
+
"logps/chosen": -308.0,
|
| 667 |
+
"logps/rejected": -280.0,
|
| 668 |
+
"loss": 0.6581,
|
| 669 |
+
"rewards/accuracies": 0.7640625238418579,
|
| 670 |
+
"rewards/chosen": -0.000598907470703125,
|
| 671 |
+
"rewards/margins": 0.08056640625,
|
| 672 |
+
"rewards/rejected": -0.0810546875,
|
| 673 |
+
"step": 70
|
| 674 |
+
},
|
| 675 |
+
{
|
| 676 |
+
"epoch": 0.39215686274509803,
|
| 677 |
+
"eval_logits/chosen": -3.09375,
|
| 678 |
+
"eval_logits/rejected": -3.140625,
|
| 679 |
+
"eval_logps/chosen": -344.0,
|
| 680 |
+
"eval_logps/rejected": -294.0,
|
| 681 |
+
"eval_loss": 0.6542046666145325,
|
| 682 |
+
"eval_rewards/accuracies": 0.7630208134651184,
|
| 683 |
+
"eval_rewards/chosen": -0.01361083984375,
|
| 684 |
+
"eval_rewards/margins": 0.09521484375,
|
| 685 |
+
"eval_rewards/rejected": -0.10888671875,
|
| 686 |
+
"eval_runtime": 133.8942,
|
| 687 |
+
"eval_samples_per_second": 11.106,
|
| 688 |
+
"eval_steps_per_second": 0.179,
|
| 689 |
+
"step": 70
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"epoch": 0.40336134453781514,
|
| 693 |
+
"eval_logits/chosen": -3.09375,
|
| 694 |
+
"eval_logits/rejected": -3.140625,
|
| 695 |
+
"eval_logps/chosen": -344.0,
|
| 696 |
+
"eval_logps/rejected": -296.0,
|
| 697 |
+
"eval_loss": 0.6521748304367065,
|
| 698 |
+
"eval_rewards/accuracies": 0.7708333134651184,
|
| 699 |
+
"eval_rewards/chosen": -0.01904296875,
|
| 700 |
+
"eval_rewards/margins": 0.1015625,
|
| 701 |
+
"eval_rewards/rejected": -0.12060546875,
|
| 702 |
+
"eval_runtime": 135.2867,
|
| 703 |
+
"eval_samples_per_second": 10.991,
|
| 704 |
+
"eval_steps_per_second": 0.177,
|
| 705 |
+
"step": 72
|
| 706 |
+
},
|
| 707 |
+
{
|
| 708 |
+
"epoch": 0.41456582633053224,
|
| 709 |
+
"eval_logits/chosen": -3.09375,
|
| 710 |
+
"eval_logits/rejected": -3.140625,
|
| 711 |
+
"eval_logps/chosen": -344.0,
|
| 712 |
+
"eval_logps/rejected": -296.0,
|
| 713 |
+
"eval_loss": 0.6502817869186401,
|
| 714 |
+
"eval_rewards/accuracies": 0.7786458134651184,
|
| 715 |
+
"eval_rewards/chosen": -0.0244140625,
|
| 716 |
+
"eval_rewards/margins": 0.10693359375,
|
| 717 |
+
"eval_rewards/rejected": -0.130859375,
|
| 718 |
+
"eval_runtime": 133.7003,
|
| 719 |
+
"eval_samples_per_second": 11.122,
|
| 720 |
+
"eval_steps_per_second": 0.18,
|
| 721 |
+
"step": 74
|
| 722 |
+
},
|
| 723 |
+
{
|
| 724 |
+
"epoch": 0.4257703081232493,
|
| 725 |
+
"eval_logits/chosen": -3.09375,
|
| 726 |
+
"eval_logits/rejected": -3.140625,
|
| 727 |
+
"eval_logps/chosen": -344.0,
|
| 728 |
+
"eval_logps/rejected": -298.0,
|
| 729 |
+
"eval_loss": 0.6483878493309021,
|
| 730 |
+
"eval_rewards/accuracies": 0.7864583134651184,
|
| 731 |
+
"eval_rewards/chosen": -0.0308837890625,
|
| 732 |
+
"eval_rewards/margins": 0.111328125,
|
| 733 |
+
"eval_rewards/rejected": -0.142578125,
|
| 734 |
+
"eval_runtime": 134.1515,
|
| 735 |
+
"eval_samples_per_second": 11.084,
|
| 736 |
+
"eval_steps_per_second": 0.179,
|
| 737 |
+
"step": 76
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"epoch": 0.4369747899159664,
|
| 741 |
+
"eval_logits/chosen": -3.09375,
|
| 742 |
+
"eval_logits/rejected": -3.140625,
|
| 743 |
+
"eval_logps/chosen": -346.0,
|
| 744 |
+
"eval_logps/rejected": -298.0,
|
| 745 |
+
"eval_loss": 0.6462110877037048,
|
| 746 |
+
"eval_rewards/accuracies": 0.7708333134651184,
|
| 747 |
+
"eval_rewards/chosen": -0.037109375,
|
| 748 |
+
"eval_rewards/margins": 0.11767578125,
|
| 749 |
+
"eval_rewards/rejected": -0.154296875,
|
| 750 |
+
"eval_runtime": 133.8558,
|
| 751 |
+
"eval_samples_per_second": 11.109,
|
| 752 |
+
"eval_steps_per_second": 0.179,
|
| 753 |
+
"step": 78
|
| 754 |
+
},
|
| 755 |
+
{
|
| 756 |
+
"epoch": 0.4481792717086835,
|
| 757 |
+
"grad_norm": 6.1230379557304975,
|
| 758 |
+
"learning_rate": 6.730585285387464e-08,
|
| 759 |
+
"logits/chosen": -3.125,
|
| 760 |
+
"logits/rejected": -3.15625,
|
| 761 |
+
"logps/chosen": -340.0,
|
| 762 |
+
"logps/rejected": -290.0,
|
| 763 |
+
"loss": 0.6458,
|
| 764 |
+
"rewards/accuracies": 0.8109375238418579,
|
| 765 |
+
"rewards/chosen": -0.01348876953125,
|
| 766 |
+
"rewards/margins": 0.119140625,
|
| 767 |
+
"rewards/rejected": -0.1328125,
|
| 768 |
+
"step": 80
|
| 769 |
+
},
|
| 770 |
+
{
|
| 771 |
+
"epoch": 0.4481792717086835,
|
| 772 |
+
"eval_logits/chosen": -3.09375,
|
| 773 |
+
"eval_logits/rejected": -3.140625,
|
| 774 |
+
"eval_logps/chosen": -346.0,
|
| 775 |
+
"eval_logps/rejected": -300.0,
|
| 776 |
+
"eval_loss": 0.6442796587944031,
|
| 777 |
+
"eval_rewards/accuracies": 0.7708333134651184,
|
| 778 |
+
"eval_rewards/chosen": -0.044677734375,
|
| 779 |
+
"eval_rewards/margins": 0.12255859375,
|
| 780 |
+
"eval_rewards/rejected": -0.1669921875,
|
| 781 |
+
"eval_runtime": 133.5665,
|
| 782 |
+
"eval_samples_per_second": 11.133,
|
| 783 |
+
"eval_steps_per_second": 0.18,
|
| 784 |
+
"step": 80
|
| 785 |
+
},
|
| 786 |
+
{
|
| 787 |
+
"epoch": 0.45938375350140054,
|
| 788 |
+
"eval_logits/chosen": -3.09375,
|
| 789 |
+
"eval_logits/rejected": -3.140625,
|
| 790 |
+
"eval_logps/chosen": -348.0,
|
| 791 |
+
"eval_logps/rejected": -300.0,
|
| 792 |
+
"eval_loss": 0.6423355937004089,
|
| 793 |
+
"eval_rewards/accuracies": 0.7838541865348816,
|
| 794 |
+
"eval_rewards/chosen": -0.049560546875,
|
| 795 |
+
"eval_rewards/margins": 0.1279296875,
|
| 796 |
+
"eval_rewards/rejected": -0.1767578125,
|
| 797 |
+
"eval_runtime": 133.4892,
|
| 798 |
+
"eval_samples_per_second": 11.139,
|
| 799 |
+
"eval_steps_per_second": 0.18,
|
| 800 |
+
"step": 82
|
| 801 |
+
},
|
| 802 |
+
{
|
| 803 |
+
"epoch": 0.47058823529411764,
|
| 804 |
+
"eval_logits/chosen": -3.09375,
|
| 805 |
+
"eval_logits/rejected": -3.125,
|
| 806 |
+
"eval_logps/chosen": -348.0,
|
| 807 |
+
"eval_logps/rejected": -302.0,
|
| 808 |
+
"eval_loss": 0.6405628323554993,
|
| 809 |
+
"eval_rewards/accuracies": 0.7916666865348816,
|
| 810 |
+
"eval_rewards/chosen": -0.052978515625,
|
| 811 |
+
"eval_rewards/margins": 0.1337890625,
|
| 812 |
+
"eval_rewards/rejected": -0.1865234375,
|
| 813 |
+
"eval_runtime": 133.4399,
|
| 814 |
+
"eval_samples_per_second": 11.144,
|
| 815 |
+
"eval_steps_per_second": 0.18,
|
| 816 |
+
"step": 84
|
| 817 |
+
},
|
| 818 |
+
{
|
| 819 |
+
"epoch": 0.48179271708683474,
|
| 820 |
+
"eval_logits/chosen": -3.09375,
|
| 821 |
+
"eval_logits/rejected": -3.125,
|
| 822 |
+
"eval_logps/chosen": -348.0,
|
| 823 |
+
"eval_logps/rejected": -302.0,
|
| 824 |
+
"eval_loss": 0.6387618184089661,
|
| 825 |
+
"eval_rewards/accuracies": 0.7994791865348816,
|
| 826 |
+
"eval_rewards/chosen": -0.056884765625,
|
| 827 |
+
"eval_rewards/margins": 0.138671875,
|
| 828 |
+
"eval_rewards/rejected": -0.1953125,
|
| 829 |
+
"eval_runtime": 133.8344,
|
| 830 |
+
"eval_samples_per_second": 11.111,
|
| 831 |
+
"eval_steps_per_second": 0.179,
|
| 832 |
+
"step": 86
|
| 833 |
+
},
|
| 834 |
+
{
|
| 835 |
+
"epoch": 0.49299719887955185,
|
| 836 |
+
"eval_logits/chosen": -3.078125,
|
| 837 |
+
"eval_logits/rejected": -3.125,
|
| 838 |
+
"eval_logps/chosen": -348.0,
|
| 839 |
+
"eval_logps/rejected": -304.0,
|
| 840 |
+
"eval_loss": 0.6369534730911255,
|
| 841 |
+
"eval_rewards/accuracies": 0.7994791865348816,
|
| 842 |
+
"eval_rewards/chosen": -0.060302734375,
|
| 843 |
+
"eval_rewards/margins": 0.14453125,
|
| 844 |
+
"eval_rewards/rejected": -0.2041015625,
|
| 845 |
+
"eval_runtime": 134.4799,
|
| 846 |
+
"eval_samples_per_second": 11.057,
|
| 847 |
+
"eval_steps_per_second": 0.178,
|
| 848 |
+
"step": 88
|
| 849 |
+
},
|
| 850 |
+
{
|
| 851 |
+
"epoch": 0.5042016806722689,
|
| 852 |
+
"grad_norm": 7.254285729207196,
|
| 853 |
+
"learning_rate": 5.7821723252011546e-08,
|
| 854 |
+
"logits/chosen": -3.0625,
|
| 855 |
+
"logits/rejected": -3.09375,
|
| 856 |
+
"logps/chosen": -330.0,
|
| 857 |
+
"logps/rejected": -288.0,
|
| 858 |
+
"loss": 0.6397,
|
| 859 |
+
"rewards/accuracies": 0.746874988079071,
|
| 860 |
+
"rewards/chosen": -0.05517578125,
|
| 861 |
+
"rewards/margins": 0.1279296875,
|
| 862 |
+
"rewards/rejected": -0.1826171875,
|
| 863 |
+
"step": 90
|
| 864 |
+
},
|
| 865 |
+
{
|
| 866 |
+
"epoch": 0.5042016806722689,
|
| 867 |
+
"eval_logits/chosen": -3.078125,
|
| 868 |
+
"eval_logits/rejected": -3.125,
|
| 869 |
+
"eval_logps/chosen": -348.0,
|
| 870 |
+
"eval_logps/rejected": -304.0,
|
| 871 |
+
"eval_loss": 0.635140061378479,
|
| 872 |
+
"eval_rewards/accuracies": 0.7916666865348816,
|
| 873 |
+
"eval_rewards/chosen": -0.062255859375,
|
| 874 |
+
"eval_rewards/margins": 0.1484375,
|
| 875 |
+
"eval_rewards/rejected": -0.2109375,
|
| 876 |
+
"eval_runtime": 134.0783,
|
| 877 |
+
"eval_samples_per_second": 11.091,
|
| 878 |
+
"eval_steps_per_second": 0.179,
|
| 879 |
+
"step": 90
|
| 880 |
+
},
|
| 881 |
+
{
|
| 882 |
+
"epoch": 0.5154061624649859,
|
| 883 |
+
"eval_logits/chosen": -3.078125,
|
| 884 |
+
"eval_logits/rejected": -3.125,
|
| 885 |
+
"eval_logps/chosen": -348.0,
|
| 886 |
+
"eval_logps/rejected": -306.0,
|
| 887 |
+
"eval_loss": 0.6333230137825012,
|
| 888 |
+
"eval_rewards/accuracies": 0.7942708134651184,
|
| 889 |
+
"eval_rewards/chosen": -0.0654296875,
|
| 890 |
+
"eval_rewards/margins": 0.154296875,
|
| 891 |
+
"eval_rewards/rejected": -0.2197265625,
|
| 892 |
+
"eval_runtime": 133.8735,
|
| 893 |
+
"eval_samples_per_second": 11.108,
|
| 894 |
+
"eval_steps_per_second": 0.179,
|
| 895 |
+
"step": 92
|
| 896 |
+
},
|
| 897 |
+
{
|
| 898 |
+
"epoch": 0.5266106442577031,
|
| 899 |
+
"eval_logits/chosen": -3.078125,
|
| 900 |
+
"eval_logits/rejected": -3.125,
|
| 901 |
+
"eval_logps/chosen": -348.0,
|
| 902 |
+
"eval_logps/rejected": -306.0,
|
| 903 |
+
"eval_loss": 0.6314958930015564,
|
| 904 |
+
"eval_rewards/accuracies": 0.8046875,
|
| 905 |
+
"eval_rewards/chosen": -0.06494140625,
|
| 906 |
+
"eval_rewards/margins": 0.1611328125,
|
| 907 |
+
"eval_rewards/rejected": -0.2255859375,
|
| 908 |
+
"eval_runtime": 133.4379,
|
| 909 |
+
"eval_samples_per_second": 11.144,
|
| 910 |
+
"eval_steps_per_second": 0.18,
|
| 911 |
+
"step": 94
|
| 912 |
+
},
|
| 913 |
+
{
|
| 914 |
+
"epoch": 0.5378151260504201,
|
| 915 |
+
"eval_logits/chosen": -3.078125,
|
| 916 |
+
"eval_logits/rejected": -3.125,
|
| 917 |
+
"eval_logps/chosen": -348.0,
|
| 918 |
+
"eval_logps/rejected": -306.0,
|
| 919 |
+
"eval_loss": 0.6297190189361572,
|
| 920 |
+
"eval_rewards/accuracies": 0.7994791865348816,
|
| 921 |
+
"eval_rewards/chosen": -0.06494140625,
|
| 922 |
+
"eval_rewards/margins": 0.1650390625,
|
| 923 |
+
"eval_rewards/rejected": -0.2294921875,
|
| 924 |
+
"eval_runtime": 134.0434,
|
| 925 |
+
"eval_samples_per_second": 11.093,
|
| 926 |
+
"eval_steps_per_second": 0.179,
|
| 927 |
+
"step": 96
|
| 928 |
+
},
|
| 929 |
+
{
|
| 930 |
+
"epoch": 0.5490196078431373,
|
| 931 |
+
"eval_logits/chosen": -3.078125,
|
| 932 |
+
"eval_logits/rejected": -3.125,
|
| 933 |
+
"eval_logps/chosen": -348.0,
|
| 934 |
+
"eval_logps/rejected": -306.0,
|
| 935 |
+
"eval_loss": 0.6282486915588379,
|
| 936 |
+
"eval_rewards/accuracies": 0.7942708134651184,
|
| 937 |
+
"eval_rewards/chosen": -0.0654296875,
|
| 938 |
+
"eval_rewards/margins": 0.1689453125,
|
| 939 |
+
"eval_rewards/rejected": -0.234375,
|
| 940 |
+
"eval_runtime": 133.4702,
|
| 941 |
+
"eval_samples_per_second": 11.141,
|
| 942 |
+
"eval_steps_per_second": 0.18,
|
| 943 |
+
"step": 98
|
| 944 |
+
},
|
| 945 |
+
{
|
| 946 |
+
"epoch": 0.5602240896358543,
|
| 947 |
+
"grad_norm": 10.597297662849414,
|
| 948 |
+
"learning_rate": 4.803700921204658e-08,
|
| 949 |
+
"logits/chosen": -3.09375,
|
| 950 |
+
"logits/rejected": -3.109375,
|
| 951 |
+
"logps/chosen": -332.0,
|
| 952 |
+
"logps/rejected": -298.0,
|
| 953 |
+
"loss": 0.6281,
|
| 954 |
+
"rewards/accuracies": 0.7828124761581421,
|
| 955 |
+
"rewards/chosen": -0.06591796875,
|
| 956 |
+
"rewards/margins": 0.158203125,
|
| 957 |
+
"rewards/rejected": -0.2236328125,
|
| 958 |
+
"step": 100
|
| 959 |
+
},
|
| 960 |
+
{
|
| 961 |
+
"epoch": 0.5602240896358543,
|
| 962 |
+
"eval_logits/chosen": -3.078125,
|
| 963 |
+
"eval_logits/rejected": -3.125,
|
| 964 |
+
"eval_logps/chosen": -348.0,
|
| 965 |
+
"eval_logps/rejected": -308.0,
|
| 966 |
+
"eval_loss": 0.6264181137084961,
|
| 967 |
+
"eval_rewards/accuracies": 0.7916666865348816,
|
| 968 |
+
"eval_rewards/chosen": -0.0673828125,
|
| 969 |
+
"eval_rewards/margins": 0.173828125,
|
| 970 |
+
"eval_rewards/rejected": -0.2412109375,
|
| 971 |
+
"eval_runtime": 133.3945,
|
| 972 |
+
"eval_samples_per_second": 11.147,
|
| 973 |
+
"eval_steps_per_second": 0.18,
|
| 974 |
+
"step": 100
|
| 975 |
+
},
|
| 976 |
+
{
|
| 977 |
+
"epoch": 0.5714285714285714,
|
| 978 |
+
"eval_logits/chosen": -3.078125,
|
| 979 |
+
"eval_logits/rejected": -3.125,
|
| 980 |
+
"eval_logps/chosen": -348.0,
|
| 981 |
+
"eval_logps/rejected": -308.0,
|
| 982 |
+
"eval_loss": 0.6247581839561462,
|
| 983 |
+
"eval_rewards/accuracies": 0.7994791865348816,
|
| 984 |
+
"eval_rewards/chosen": -0.06591796875,
|
| 985 |
+
"eval_rewards/margins": 0.1787109375,
|
| 986 |
+
"eval_rewards/rejected": -0.2451171875,
|
| 987 |
+
"eval_runtime": 134.9509,
|
| 988 |
+
"eval_samples_per_second": 11.019,
|
| 989 |
+
"eval_steps_per_second": 0.178,
|
| 990 |
+
"step": 102
|
| 991 |
+
},
|
| 992 |
+
{
|
| 993 |
+
"epoch": 0.5826330532212886,
|
| 994 |
+
"eval_logits/chosen": -3.078125,
|
| 995 |
+
"eval_logits/rejected": -3.125,
|
| 996 |
+
"eval_logps/chosen": -348.0,
|
| 997 |
+
"eval_logps/rejected": -308.0,
|
| 998 |
+
"eval_loss": 0.6230900883674622,
|
| 999 |
+
"eval_rewards/accuracies": 0.8020833134651184,
|
| 1000 |
+
"eval_rewards/chosen": -0.0654296875,
|
| 1001 |
+
"eval_rewards/margins": 0.18359375,
|
| 1002 |
+
"eval_rewards/rejected": -0.25,
|
| 1003 |
+
"eval_runtime": 133.2495,
|
| 1004 |
+
"eval_samples_per_second": 11.16,
|
| 1005 |
+
"eval_steps_per_second": 0.18,
|
| 1006 |
+
"step": 104
|
| 1007 |
+
},
|
| 1008 |
+
{
|
| 1009 |
+
"epoch": 0.5938375350140056,
|
| 1010 |
+
"eval_logits/chosen": -3.078125,
|
| 1011 |
+
"eval_logits/rejected": -3.125,
|
| 1012 |
+
"eval_logps/chosen": -348.0,
|
| 1013 |
+
"eval_logps/rejected": -308.0,
|
| 1014 |
+
"eval_loss": 0.621644914150238,
|
| 1015 |
+
"eval_rewards/accuracies": 0.8020833134651184,
|
| 1016 |
+
"eval_rewards/chosen": -0.0654296875,
|
| 1017 |
+
"eval_rewards/margins": 0.1884765625,
|
| 1018 |
+
"eval_rewards/rejected": -0.25390625,
|
| 1019 |
+
"eval_runtime": 134.3837,
|
| 1020 |
+
"eval_samples_per_second": 11.065,
|
| 1021 |
+
"eval_steps_per_second": 0.179,
|
| 1022 |
+
"step": 106
|
| 1023 |
+
},
|
| 1024 |
+
{
|
| 1025 |
+
"epoch": 0.6050420168067226,
|
| 1026 |
+
"eval_logits/chosen": -3.078125,
|
| 1027 |
+
"eval_logits/rejected": -3.125,
|
| 1028 |
+
"eval_logps/chosen": -348.0,
|
| 1029 |
+
"eval_logps/rejected": -308.0,
|
| 1030 |
+
"eval_loss": 0.6203284859657288,
|
| 1031 |
+
"eval_rewards/accuracies": 0.8072916865348816,
|
| 1032 |
+
"eval_rewards/chosen": -0.064453125,
|
| 1033 |
+
"eval_rewards/margins": 0.19140625,
|
| 1034 |
+
"eval_rewards/rejected": -0.255859375,
|
| 1035 |
+
"eval_runtime": 133.785,
|
| 1036 |
+
"eval_samples_per_second": 11.115,
|
| 1037 |
+
"eval_steps_per_second": 0.179,
|
| 1038 |
+
"step": 108
|
| 1039 |
+
},
|
| 1040 |
+
{
|
| 1041 |
+
"epoch": 0.6162464985994398,
|
| 1042 |
+
"grad_norm": 7.480434655836269,
|
| 1043 |
+
"learning_rate": 3.8327731807204744e-08,
|
| 1044 |
+
"logits/chosen": -3.109375,
|
| 1045 |
+
"logits/rejected": -3.09375,
|
| 1046 |
+
"logps/chosen": -326.0,
|
| 1047 |
+
"logps/rejected": -306.0,
|
| 1048 |
+
"loss": 0.6191,
|
| 1049 |
+
"rewards/accuracies": 0.7734375,
|
| 1050 |
+
"rewards/chosen": -0.0673828125,
|
| 1051 |
+
"rewards/margins": 0.1845703125,
|
| 1052 |
+
"rewards/rejected": -0.251953125,
|
| 1053 |
+
"step": 110
|
| 1054 |
+
},
|
| 1055 |
+
{
|
| 1056 |
+
"epoch": 0.6162464985994398,
|
| 1057 |
+
"eval_logits/chosen": -3.078125,
|
| 1058 |
+
"eval_logits/rejected": -3.125,
|
| 1059 |
+
"eval_logps/chosen": -348.0,
|
| 1060 |
+
"eval_logps/rejected": -310.0,
|
| 1061 |
+
"eval_loss": 0.6189648509025574,
|
| 1062 |
+
"eval_rewards/accuracies": 0.8072916865348816,
|
| 1063 |
+
"eval_rewards/chosen": -0.06494140625,
|
| 1064 |
+
"eval_rewards/margins": 0.1953125,
|
| 1065 |
+
"eval_rewards/rejected": -0.259765625,
|
| 1066 |
+
"eval_runtime": 134.1423,
|
| 1067 |
+
"eval_samples_per_second": 11.085,
|
| 1068 |
+
"eval_steps_per_second": 0.179,
|
| 1069 |
+
"step": 110
|
| 1070 |
+
},
|
| 1071 |
+
{
|
| 1072 |
+
"epoch": 0.6274509803921569,
|
| 1073 |
+
"eval_logits/chosen": -3.078125,
|
| 1074 |
+
"eval_logits/rejected": -3.125,
|
| 1075 |
+
"eval_logps/chosen": -348.0,
|
| 1076 |
+
"eval_logps/rejected": -310.0,
|
| 1077 |
+
"eval_loss": 0.6177744269371033,
|
| 1078 |
+
"eval_rewards/accuracies": 0.8046875,
|
| 1079 |
+
"eval_rewards/chosen": -0.06982421875,
|
| 1080 |
+
"eval_rewards/margins": 0.19921875,
|
| 1081 |
+
"eval_rewards/rejected": -0.26953125,
|
| 1082 |
+
"eval_runtime": 133.5083,
|
| 1083 |
+
"eval_samples_per_second": 11.138,
|
| 1084 |
+
"eval_steps_per_second": 0.18,
|
| 1085 |
+
"step": 112
|
| 1086 |
+
},
|
| 1087 |
+
{
|
| 1088 |
+
"epoch": 0.6386554621848739,
|
| 1089 |
+
"eval_logits/chosen": -3.078125,
|
| 1090 |
+
"eval_logits/rejected": -3.125,
|
| 1091 |
+
"eval_logps/chosen": -350.0,
|
| 1092 |
+
"eval_logps/rejected": -312.0,
|
| 1093 |
+
"eval_loss": 0.616416335105896,
|
| 1094 |
+
"eval_rewards/accuracies": 0.7994791865348816,
|
| 1095 |
+
"eval_rewards/chosen": -0.07470703125,
|
| 1096 |
+
"eval_rewards/margins": 0.205078125,
|
| 1097 |
+
"eval_rewards/rejected": -0.279296875,
|
| 1098 |
+
"eval_runtime": 133.8287,
|
| 1099 |
+
"eval_samples_per_second": 11.111,
|
| 1100 |
+
"eval_steps_per_second": 0.179,
|
| 1101 |
+
"step": 114
|
| 1102 |
+
},
|
| 1103 |
+
{
|
| 1104 |
+
"epoch": 0.6498599439775911,
|
| 1105 |
+
"eval_logits/chosen": -3.078125,
|
| 1106 |
+
"eval_logits/rejected": -3.125,
|
| 1107 |
+
"eval_logps/chosen": -350.0,
|
| 1108 |
+
"eval_logps/rejected": -312.0,
|
| 1109 |
+
"eval_loss": 0.6151247620582581,
|
| 1110 |
+
"eval_rewards/accuracies": 0.796875,
|
| 1111 |
+
"eval_rewards/chosen": -0.0791015625,
|
| 1112 |
+
"eval_rewards/margins": 0.208984375,
|
| 1113 |
+
"eval_rewards/rejected": -0.287109375,
|
| 1114 |
+
"eval_runtime": 133.8728,
|
| 1115 |
+
"eval_samples_per_second": 11.108,
|
| 1116 |
+
"eval_steps_per_second": 0.179,
|
| 1117 |
+
"step": 116
|
| 1118 |
+
},
|
| 1119 |
+
{
|
| 1120 |
+
"epoch": 0.6610644257703081,
|
| 1121 |
+
"eval_logits/chosen": -3.078125,
|
| 1122 |
+
"eval_logits/rejected": -3.125,
|
| 1123 |
+
"eval_logps/chosen": -350.0,
|
| 1124 |
+
"eval_logps/rejected": -312.0,
|
| 1125 |
+
"eval_loss": 0.6137887239456177,
|
| 1126 |
+
"eval_rewards/accuracies": 0.7994791865348816,
|
| 1127 |
+
"eval_rewards/chosen": -0.08349609375,
|
| 1128 |
+
"eval_rewards/margins": 0.212890625,
|
| 1129 |
+
"eval_rewards/rejected": -0.294921875,
|
| 1130 |
+
"eval_runtime": 134.3374,
|
| 1131 |
+
"eval_samples_per_second": 11.069,
|
| 1132 |
+
"eval_steps_per_second": 0.179,
|
| 1133 |
+
"step": 118
|
| 1134 |
+
},
|
| 1135 |
+
{
|
| 1136 |
+
"epoch": 0.6722689075630253,
|
| 1137 |
+
"grad_norm": 6.779083992073795,
|
| 1138 |
+
"learning_rate": 2.906701312312861e-08,
|
| 1139 |
+
"logits/chosen": -3.09375,
|
| 1140 |
+
"logits/rejected": -3.109375,
|
| 1141 |
+
"logps/chosen": -346.0,
|
| 1142 |
+
"logps/rejected": -320.0,
|
| 1143 |
+
"loss": 0.6183,
|
| 1144 |
+
"rewards/accuracies": 0.778124988079071,
|
| 1145 |
+
"rewards/chosen": -0.07373046875,
|
| 1146 |
+
"rewards/margins": 0.2001953125,
|
| 1147 |
+
"rewards/rejected": -0.2734375,
|
| 1148 |
+
"step": 120
|
| 1149 |
+
},
|
| 1150 |
+
{
|
| 1151 |
+
"epoch": 0.6722689075630253,
|
| 1152 |
+
"eval_logits/chosen": -3.078125,
|
| 1153 |
+
"eval_logits/rejected": -3.125,
|
| 1154 |
+
"eval_logps/chosen": -350.0,
|
| 1155 |
+
"eval_logps/rejected": -314.0,
|
| 1156 |
+
"eval_loss": 0.6126104593276978,
|
| 1157 |
+
"eval_rewards/accuracies": 0.7994791865348816,
|
| 1158 |
+
"eval_rewards/chosen": -0.0869140625,
|
| 1159 |
+
"eval_rewards/margins": 0.216796875,
|
| 1160 |
+
"eval_rewards/rejected": -0.3046875,
|
| 1161 |
+
"eval_runtime": 134.5655,
|
| 1162 |
+
"eval_samples_per_second": 11.05,
|
| 1163 |
+
"eval_steps_per_second": 0.178,
|
| 1164 |
+
"step": 120
|
| 1165 |
+
},
|
| 1166 |
+
{
|
| 1167 |
+
"epoch": 0.6834733893557423,
|
| 1168 |
+
"eval_logits/chosen": -3.078125,
|
| 1169 |
+
"eval_logits/rejected": -3.125,
|
| 1170 |
+
"eval_logps/chosen": -350.0,
|
| 1171 |
+
"eval_logps/rejected": -314.0,
|
| 1172 |
+
"eval_loss": 0.6114761233329773,
|
| 1173 |
+
"eval_rewards/accuracies": 0.796875,
|
| 1174 |
+
"eval_rewards/chosen": -0.087890625,
|
| 1175 |
+
"eval_rewards/margins": 0.220703125,
|
| 1176 |
+
"eval_rewards/rejected": -0.30859375,
|
| 1177 |
+
"eval_runtime": 134.1574,
|
| 1178 |
+
"eval_samples_per_second": 11.084,
|
| 1179 |
+
"eval_steps_per_second": 0.179,
|
| 1180 |
+
"step": 122
|
| 1181 |
+
},
|
| 1182 |
+
{
|
| 1183 |
+
"epoch": 0.6946778711484594,
|
| 1184 |
+
"eval_logits/chosen": -3.078125,
|
| 1185 |
+
"eval_logits/rejected": -3.125,
|
| 1186 |
+
"eval_logps/chosen": -350.0,
|
| 1187 |
+
"eval_logps/rejected": -314.0,
|
| 1188 |
+
"eval_loss": 0.6107271313667297,
|
| 1189 |
+
"eval_rewards/accuracies": 0.7994791865348816,
|
| 1190 |
+
"eval_rewards/chosen": -0.08837890625,
|
| 1191 |
+
"eval_rewards/margins": 0.2236328125,
|
| 1192 |
+
"eval_rewards/rejected": -0.3125,
|
| 1193 |
+
"eval_runtime": 134.2971,
|
| 1194 |
+
"eval_samples_per_second": 11.072,
|
| 1195 |
+
"eval_steps_per_second": 0.179,
|
| 1196 |
+
"step": 124
|
| 1197 |
+
},
|
| 1198 |
+
{
|
| 1199 |
+
"epoch": 0.7058823529411765,
|
| 1200 |
+
"eval_logits/chosen": -3.078125,
|
| 1201 |
+
"eval_logits/rejected": -3.125,
|
| 1202 |
+
"eval_logps/chosen": -350.0,
|
| 1203 |
+
"eval_logps/rejected": -314.0,
|
| 1204 |
+
"eval_loss": 0.6096285581588745,
|
| 1205 |
+
"eval_rewards/accuracies": 0.7942708134651184,
|
| 1206 |
+
"eval_rewards/chosen": -0.0869140625,
|
| 1207 |
+
"eval_rewards/margins": 0.2265625,
|
| 1208 |
+
"eval_rewards/rejected": -0.314453125,
|
| 1209 |
+
"eval_runtime": 135.0298,
|
| 1210 |
+
"eval_samples_per_second": 11.012,
|
| 1211 |
+
"eval_steps_per_second": 0.178,
|
| 1212 |
+
"step": 126
|
| 1213 |
+
},
|
| 1214 |
+
{
|
| 1215 |
+
"epoch": 0.7170868347338936,
|
| 1216 |
+
"eval_logits/chosen": -3.078125,
|
| 1217 |
+
"eval_logits/rejected": -3.125,
|
| 1218 |
+
"eval_logps/chosen": -350.0,
|
| 1219 |
+
"eval_logps/rejected": -314.0,
|
| 1220 |
+
"eval_loss": 0.6089996099472046,
|
| 1221 |
+
"eval_rewards/accuracies": 0.7942708134651184,
|
| 1222 |
+
"eval_rewards/chosen": -0.0859375,
|
| 1223 |
+
"eval_rewards/margins": 0.2275390625,
|
| 1224 |
+
"eval_rewards/rejected": -0.314453125,
|
| 1225 |
+
"eval_runtime": 134.1686,
|
| 1226 |
+
"eval_samples_per_second": 11.083,
|
| 1227 |
+
"eval_steps_per_second": 0.179,
|
| 1228 |
+
"step": 128
|
| 1229 |
+
},
|
| 1230 |
+
{
|
| 1231 |
+
"epoch": 0.7282913165266106,
|
| 1232 |
+
"grad_norm": 7.520918890430443,
|
| 1233 |
+
"learning_rate": 2.0610737385376347e-08,
|
| 1234 |
+
"logits/chosen": -3.109375,
|
| 1235 |
+
"logits/rejected": -3.125,
|
| 1236 |
+
"logps/chosen": -328.0,
|
| 1237 |
+
"logps/rejected": -308.0,
|
| 1238 |
+
"loss": 0.606,
|
| 1239 |
+
"rewards/accuracies": 0.7906249761581421,
|
| 1240 |
+
"rewards/chosen": -0.0810546875,
|
| 1241 |
+
"rewards/margins": 0.2265625,
|
| 1242 |
+
"rewards/rejected": -0.30859375,
|
| 1243 |
+
"step": 130
|
| 1244 |
+
},
|
| 1245 |
+
{
|
| 1246 |
+
"epoch": 0.7282913165266106,
|
| 1247 |
+
"eval_logits/chosen": -3.078125,
|
| 1248 |
+
"eval_logits/rejected": -3.125,
|
| 1249 |
+
"eval_logps/chosen": -350.0,
|
| 1250 |
+
"eval_logps/rejected": -314.0,
|
| 1251 |
+
"eval_loss": 0.6081994771957397,
|
| 1252 |
+
"eval_rewards/accuracies": 0.796875,
|
| 1253 |
+
"eval_rewards/chosen": -0.08642578125,
|
| 1254 |
+
"eval_rewards/margins": 0.23046875,
|
| 1255 |
+
"eval_rewards/rejected": -0.31640625,
|
| 1256 |
+
"eval_runtime": 134.4089,
|
| 1257 |
+
"eval_samples_per_second": 11.063,
|
| 1258 |
+
"eval_steps_per_second": 0.179,
|
| 1259 |
+
"step": 130
|
| 1260 |
+
},
|
| 1261 |
+
{
|
| 1262 |
+
"epoch": 0.7394957983193278,
|
| 1263 |
+
"eval_logits/chosen": -3.078125,
|
| 1264 |
+
"eval_logits/rejected": -3.125,
|
| 1265 |
+
"eval_logps/chosen": -350.0,
|
| 1266 |
+
"eval_logps/rejected": -316.0,
|
| 1267 |
+
"eval_loss": 0.6074265241622925,
|
| 1268 |
+
"eval_rewards/accuracies": 0.7942708134651184,
|
| 1269 |
+
"eval_rewards/chosen": -0.08740234375,
|
| 1270 |
+
"eval_rewards/margins": 0.2333984375,
|
| 1271 |
+
"eval_rewards/rejected": -0.3203125,
|
| 1272 |
+
"eval_runtime": 133.9872,
|
| 1273 |
+
"eval_samples_per_second": 11.098,
|
| 1274 |
+
"eval_steps_per_second": 0.179,
|
| 1275 |
+
"step": 132
|
| 1276 |
+
},
|
| 1277 |
+
{
|
| 1278 |
+
"epoch": 0.7507002801120448,
|
| 1279 |
+
"eval_logits/chosen": -3.078125,
|
| 1280 |
+
"eval_logits/rejected": -3.125,
|
| 1281 |
+
"eval_logps/chosen": -350.0,
|
| 1282 |
+
"eval_logps/rejected": -316.0,
|
| 1283 |
+
"eval_loss": 0.6068228483200073,
|
| 1284 |
+
"eval_rewards/accuracies": 0.7994791865348816,
|
| 1285 |
+
"eval_rewards/chosen": -0.08837890625,
|
| 1286 |
+
"eval_rewards/margins": 0.2353515625,
|
| 1287 |
+
"eval_rewards/rejected": -0.322265625,
|
| 1288 |
+
"eval_runtime": 133.6241,
|
| 1289 |
+
"eval_samples_per_second": 11.128,
|
| 1290 |
+
"eval_steps_per_second": 0.18,
|
| 1291 |
+
"step": 134
|
| 1292 |
+
},
|
| 1293 |
+
{
|
| 1294 |
+
"epoch": 0.7619047619047619,
|
| 1295 |
+
"eval_logits/chosen": -3.078125,
|
| 1296 |
+
"eval_logits/rejected": -3.125,
|
| 1297 |
+
"eval_logps/chosen": -350.0,
|
| 1298 |
+
"eval_logps/rejected": -316.0,
|
| 1299 |
+
"eval_loss": 0.6062489748001099,
|
| 1300 |
+
"eval_rewards/accuracies": 0.7890625,
|
| 1301 |
+
"eval_rewards/chosen": -0.087890625,
|
| 1302 |
+
"eval_rewards/margins": 0.2353515625,
|
| 1303 |
+
"eval_rewards/rejected": -0.32421875,
|
| 1304 |
+
"eval_runtime": 133.7733,
|
| 1305 |
+
"eval_samples_per_second": 11.116,
|
| 1306 |
+
"eval_steps_per_second": 0.179,
|
| 1307 |
+
"step": 136
|
| 1308 |
+
},
|
| 1309 |
+
{
|
| 1310 |
+
"epoch": 0.773109243697479,
|
| 1311 |
+
"eval_logits/chosen": -3.078125,
|
| 1312 |
+
"eval_logits/rejected": -3.125,
|
| 1313 |
+
"eval_logps/chosen": -350.0,
|
| 1314 |
+
"eval_logps/rejected": -316.0,
|
| 1315 |
+
"eval_loss": 0.6058236360549927,
|
| 1316 |
+
"eval_rewards/accuracies": 0.7916666865348816,
|
| 1317 |
+
"eval_rewards/chosen": -0.0859375,
|
| 1318 |
+
"eval_rewards/margins": 0.23828125,
|
| 1319 |
+
"eval_rewards/rejected": -0.32421875,
|
| 1320 |
+
"eval_runtime": 132.8269,
|
| 1321 |
+
"eval_samples_per_second": 11.195,
|
| 1322 |
+
"eval_steps_per_second": 0.181,
|
| 1323 |
+
"step": 138
|
| 1324 |
+
},
|
| 1325 |
+
{
|
| 1326 |
+
"epoch": 0.7843137254901961,
|
| 1327 |
+
"grad_norm": 6.700409826395709,
|
| 1328 |
+
"learning_rate": 1.3283874528215733e-08,
|
| 1329 |
+
"logits/chosen": -3.109375,
|
| 1330 |
+
"logits/rejected": -3.09375,
|
| 1331 |
+
"logps/chosen": -346.0,
|
| 1332 |
+
"logps/rejected": -318.0,
|
| 1333 |
+
"loss": 0.6045,
|
| 1334 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 1335 |
+
"rewards/chosen": -0.0751953125,
|
| 1336 |
+
"rewards/margins": 0.251953125,
|
| 1337 |
+
"rewards/rejected": -0.326171875,
|
| 1338 |
+
"step": 140
|
| 1339 |
+
},
|
| 1340 |
+
{
|
| 1341 |
+
"epoch": 0.7843137254901961,
|
| 1342 |
+
"eval_logits/chosen": -3.078125,
|
| 1343 |
+
"eval_logits/rejected": -3.125,
|
| 1344 |
+
"eval_logps/chosen": -350.0,
|
| 1345 |
+
"eval_logps/rejected": -316.0,
|
| 1346 |
+
"eval_loss": 0.6051818132400513,
|
| 1347 |
+
"eval_rewards/accuracies": 0.7890625,
|
| 1348 |
+
"eval_rewards/chosen": -0.0849609375,
|
| 1349 |
+
"eval_rewards/margins": 0.240234375,
|
| 1350 |
+
"eval_rewards/rejected": -0.326171875,
|
| 1351 |
+
"eval_runtime": 134.1619,
|
| 1352 |
+
"eval_samples_per_second": 11.084,
|
| 1353 |
+
"eval_steps_per_second": 0.179,
|
| 1354 |
+
"step": 140
|
| 1355 |
+
},
|
| 1356 |
+
{
|
| 1357 |
+
"epoch": 0.7955182072829131,
|
| 1358 |
+
"eval_logits/chosen": -3.078125,
|
| 1359 |
+
"eval_logits/rejected": -3.125,
|
| 1360 |
+
"eval_logps/chosen": -350.0,
|
| 1361 |
+
"eval_logps/rejected": -316.0,
|
| 1362 |
+
"eval_loss": 0.604813277721405,
|
| 1363 |
+
"eval_rewards/accuracies": 0.7994791865348816,
|
| 1364 |
+
"eval_rewards/chosen": -0.08447265625,
|
| 1365 |
+
"eval_rewards/margins": 0.2412109375,
|
| 1366 |
+
"eval_rewards/rejected": -0.326171875,
|
| 1367 |
+
"eval_runtime": 134.0921,
|
| 1368 |
+
"eval_samples_per_second": 11.089,
|
| 1369 |
+
"eval_steps_per_second": 0.179,
|
| 1370 |
+
"step": 142
|
| 1371 |
+
},
|
| 1372 |
+
{
|
| 1373 |
+
"epoch": 0.8067226890756303,
|
| 1374 |
+
"eval_logits/chosen": -3.078125,
|
| 1375 |
+
"eval_logits/rejected": -3.125,
|
| 1376 |
+
"eval_logps/chosen": -350.0,
|
| 1377 |
+
"eval_logps/rejected": -316.0,
|
| 1378 |
+
"eval_loss": 0.6045451164245605,
|
| 1379 |
+
"eval_rewards/accuracies": 0.796875,
|
| 1380 |
+
"eval_rewards/chosen": -0.08251953125,
|
| 1381 |
+
"eval_rewards/margins": 0.2421875,
|
| 1382 |
+
"eval_rewards/rejected": -0.32421875,
|
| 1383 |
+
"eval_runtime": 133.3791,
|
| 1384 |
+
"eval_samples_per_second": 11.149,
|
| 1385 |
+
"eval_steps_per_second": 0.18,
|
| 1386 |
+
"step": 144
|
| 1387 |
+
},
|
| 1388 |
+
{
|
| 1389 |
+
"epoch": 0.8179271708683473,
|
| 1390 |
+
"eval_logits/chosen": -3.078125,
|
| 1391 |
+
"eval_logits/rejected": -3.125,
|
| 1392 |
+
"eval_logps/chosen": -350.0,
|
| 1393 |
+
"eval_logps/rejected": -316.0,
|
| 1394 |
+
"eval_loss": 0.604152262210846,
|
| 1395 |
+
"eval_rewards/accuracies": 0.796875,
|
| 1396 |
+
"eval_rewards/chosen": -0.0830078125,
|
| 1397 |
+
"eval_rewards/margins": 0.2431640625,
|
| 1398 |
+
"eval_rewards/rejected": -0.326171875,
|
| 1399 |
+
"eval_runtime": 136.215,
|
| 1400 |
+
"eval_samples_per_second": 10.917,
|
| 1401 |
+
"eval_steps_per_second": 0.176,
|
| 1402 |
+
"step": 146
|
| 1403 |
+
},
|
| 1404 |
+
{
|
| 1405 |
+
"epoch": 0.8291316526610645,
|
| 1406 |
+
"eval_logits/chosen": -3.078125,
|
| 1407 |
+
"eval_logits/rejected": -3.125,
|
| 1408 |
+
"eval_logps/chosen": -350.0,
|
| 1409 |
+
"eval_logps/rejected": -316.0,
|
| 1410 |
+
"eval_loss": 0.6041899919509888,
|
| 1411 |
+
"eval_rewards/accuracies": 0.796875,
|
| 1412 |
+
"eval_rewards/chosen": -0.08349609375,
|
| 1413 |
+
"eval_rewards/margins": 0.2421875,
|
| 1414 |
+
"eval_rewards/rejected": -0.326171875,
|
| 1415 |
+
"eval_runtime": 133.2436,
|
| 1416 |
+
"eval_samples_per_second": 11.16,
|
| 1417 |
+
"eval_steps_per_second": 0.18,
|
| 1418 |
+
"step": 148
|
| 1419 |
+
},
|
| 1420 |
+
{
|
| 1421 |
+
"epoch": 0.8403361344537815,
|
| 1422 |
+
"grad_norm": 7.912383511701059,
|
| 1423 |
+
"learning_rate": 7.367991782295391e-09,
|
| 1424 |
+
"logits/chosen": -3.09375,
|
| 1425 |
+
"logits/rejected": -3.0625,
|
| 1426 |
+
"logps/chosen": -342.0,
|
| 1427 |
+
"logps/rejected": -318.0,
|
| 1428 |
+
"loss": 0.602,
|
| 1429 |
+
"rewards/accuracies": 0.7953125238418579,
|
| 1430 |
+
"rewards/chosen": -0.07470703125,
|
| 1431 |
+
"rewards/margins": 0.234375,
|
| 1432 |
+
"rewards/rejected": -0.30859375,
|
| 1433 |
+
"step": 150
|
| 1434 |
+
},
|
| 1435 |
+
{
|
| 1436 |
+
"epoch": 0.8403361344537815,
|
| 1437 |
+
"eval_logits/chosen": -3.078125,
|
| 1438 |
+
"eval_logits/rejected": -3.125,
|
| 1439 |
+
"eval_logps/chosen": -350.0,
|
| 1440 |
+
"eval_logps/rejected": -316.0,
|
| 1441 |
+
"eval_loss": 0.6036604642868042,
|
| 1442 |
+
"eval_rewards/accuracies": 0.796875,
|
| 1443 |
+
"eval_rewards/chosen": -0.08203125,
|
| 1444 |
+
"eval_rewards/margins": 0.2451171875,
|
| 1445 |
+
"eval_rewards/rejected": -0.328125,
|
| 1446 |
+
"eval_runtime": 133.8914,
|
| 1447 |
+
"eval_samples_per_second": 11.106,
|
| 1448 |
+
"eval_steps_per_second": 0.179,
|
| 1449 |
+
"step": 150
|
| 1450 |
+
},
|
| 1451 |
+
{
|
| 1452 |
+
"epoch": 0.8515406162464986,
|
| 1453 |
+
"eval_logits/chosen": -3.078125,
|
| 1454 |
+
"eval_logits/rejected": -3.125,
|
| 1455 |
+
"eval_logps/chosen": -350.0,
|
| 1456 |
+
"eval_logps/rejected": -316.0,
|
| 1457 |
+
"eval_loss": 0.6033771634101868,
|
| 1458 |
+
"eval_rewards/accuracies": 0.7916666865348816,
|
| 1459 |
+
"eval_rewards/chosen": -0.08349609375,
|
| 1460 |
+
"eval_rewards/margins": 0.24609375,
|
| 1461 |
+
"eval_rewards/rejected": -0.328125,
|
| 1462 |
+
"eval_runtime": 133.2707,
|
| 1463 |
+
"eval_samples_per_second": 11.158,
|
| 1464 |
+
"eval_steps_per_second": 0.18,
|
| 1465 |
+
"step": 152
|
| 1466 |
+
},
|
| 1467 |
+
{
|
| 1468 |
+
"epoch": 0.8627450980392157,
|
| 1469 |
+
"eval_logits/chosen": -3.078125,
|
| 1470 |
+
"eval_logits/rejected": -3.125,
|
| 1471 |
+
"eval_logps/chosen": -350.0,
|
| 1472 |
+
"eval_logps/rejected": -316.0,
|
| 1473 |
+
"eval_loss": 0.60316401720047,
|
| 1474 |
+
"eval_rewards/accuracies": 0.7864583134651184,
|
| 1475 |
+
"eval_rewards/chosen": -0.0849609375,
|
| 1476 |
+
"eval_rewards/margins": 0.2470703125,
|
| 1477 |
+
"eval_rewards/rejected": -0.33203125,
|
| 1478 |
+
"eval_runtime": 132.7316,
|
| 1479 |
+
"eval_samples_per_second": 11.203,
|
| 1480 |
+
"eval_steps_per_second": 0.181,
|
| 1481 |
+
"step": 154
|
| 1482 |
+
},
|
| 1483 |
+
{
|
| 1484 |
+
"epoch": 0.8739495798319328,
|
| 1485 |
+
"eval_logits/chosen": -3.078125,
|
| 1486 |
+
"eval_logits/rejected": -3.125,
|
| 1487 |
+
"eval_logps/chosen": -350.0,
|
| 1488 |
+
"eval_logps/rejected": -316.0,
|
| 1489 |
+
"eval_loss": 0.6029848456382751,
|
| 1490 |
+
"eval_rewards/accuracies": 0.7942708134651184,
|
| 1491 |
+
"eval_rewards/chosen": -0.08740234375,
|
| 1492 |
+
"eval_rewards/margins": 0.24609375,
|
| 1493 |
+
"eval_rewards/rejected": -0.333984375,
|
| 1494 |
+
"eval_runtime": 134.4869,
|
| 1495 |
+
"eval_samples_per_second": 11.057,
|
| 1496 |
+
"eval_steps_per_second": 0.178,
|
| 1497 |
+
"step": 156
|
| 1498 |
+
},
|
| 1499 |
+
{
|
| 1500 |
+
"epoch": 0.8851540616246498,
|
| 1501 |
+
"eval_logits/chosen": -3.078125,
|
| 1502 |
+
"eval_logits/rejected": -3.125,
|
| 1503 |
+
"eval_logps/chosen": -350.0,
|
| 1504 |
+
"eval_logps/rejected": -316.0,
|
| 1505 |
+
"eval_loss": 0.6026829481124878,
|
| 1506 |
+
"eval_rewards/accuracies": 0.7942708134651184,
|
| 1507 |
+
"eval_rewards/chosen": -0.08837890625,
|
| 1508 |
+
"eval_rewards/margins": 0.248046875,
|
| 1509 |
+
"eval_rewards/rejected": -0.3359375,
|
| 1510 |
+
"eval_runtime": 134.2812,
|
| 1511 |
+
"eval_samples_per_second": 11.074,
|
| 1512 |
+
"eval_steps_per_second": 0.179,
|
| 1513 |
+
"step": 158
|
| 1514 |
+
},
|
| 1515 |
+
{
|
| 1516 |
+
"epoch": 0.896358543417367,
|
| 1517 |
+
"grad_norm": 7.847752860433035,
|
| 1518 |
+
"learning_rate": 3.0904332038757976e-09,
|
| 1519 |
+
"logits/chosen": -3.109375,
|
| 1520 |
+
"logits/rejected": -3.109375,
|
| 1521 |
+
"logps/chosen": -340.0,
|
| 1522 |
+
"logps/rejected": -308.0,
|
| 1523 |
+
"loss": 0.6021,
|
| 1524 |
+
"rewards/accuracies": 0.760937511920929,
|
| 1525 |
+
"rewards/chosen": -0.08251953125,
|
| 1526 |
+
"rewards/margins": 0.2294921875,
|
| 1527 |
+
"rewards/rejected": -0.3125,
|
| 1528 |
+
"step": 160
|
| 1529 |
+
},
|
| 1530 |
+
{
|
| 1531 |
+
"epoch": 0.896358543417367,
|
| 1532 |
+
"eval_logits/chosen": -3.078125,
|
| 1533 |
+
"eval_logits/rejected": -3.125,
|
| 1534 |
+
"eval_logps/chosen": -350.0,
|
| 1535 |
+
"eval_logps/rejected": -316.0,
|
| 1536 |
+
"eval_loss": 0.6027438640594482,
|
| 1537 |
+
"eval_rewards/accuracies": 0.7942708134651184,
|
| 1538 |
+
"eval_rewards/chosen": -0.087890625,
|
| 1539 |
+
"eval_rewards/margins": 0.2490234375,
|
| 1540 |
+
"eval_rewards/rejected": -0.3359375,
|
| 1541 |
+
"eval_runtime": 133.411,
|
| 1542 |
+
"eval_samples_per_second": 11.146,
|
| 1543 |
+
"eval_steps_per_second": 0.18,
|
| 1544 |
+
"step": 160
|
| 1545 |
+
},
|
| 1546 |
+
{
|
| 1547 |
+
"epoch": 0.907563025210084,
|
| 1548 |
+
"eval_logits/chosen": -3.078125,
|
| 1549 |
+
"eval_logits/rejected": -3.125,
|
| 1550 |
+
"eval_logps/chosen": -350.0,
|
| 1551 |
+
"eval_logps/rejected": -316.0,
|
| 1552 |
+
"eval_loss": 0.6028435826301575,
|
| 1553 |
+
"eval_rewards/accuracies": 0.796875,
|
| 1554 |
+
"eval_rewards/chosen": -0.08837890625,
|
| 1555 |
+
"eval_rewards/margins": 0.248046875,
|
| 1556 |
+
"eval_rewards/rejected": -0.3359375,
|
| 1557 |
+
"eval_runtime": 133.846,
|
| 1558 |
+
"eval_samples_per_second": 11.11,
|
| 1559 |
+
"eval_steps_per_second": 0.179,
|
| 1560 |
+
"step": 162
|
| 1561 |
+
},
|
| 1562 |
+
{
|
| 1563 |
+
"epoch": 0.9187675070028011,
|
| 1564 |
+
"eval_logits/chosen": -3.078125,
|
| 1565 |
+
"eval_logits/rejected": -3.125,
|
| 1566 |
+
"eval_logps/chosen": -350.0,
|
| 1567 |
+
"eval_logps/rejected": -318.0,
|
| 1568 |
+
"eval_loss": 0.6026445627212524,
|
| 1569 |
+
"eval_rewards/accuracies": 0.796875,
|
| 1570 |
+
"eval_rewards/chosen": -0.08837890625,
|
| 1571 |
+
"eval_rewards/margins": 0.2490234375,
|
| 1572 |
+
"eval_rewards/rejected": -0.337890625,
|
| 1573 |
+
"eval_runtime": 133.4677,
|
| 1574 |
+
"eval_samples_per_second": 11.141,
|
| 1575 |
+
"eval_steps_per_second": 0.18,
|
| 1576 |
+
"step": 164
|
| 1577 |
+
},
|
| 1578 |
+
{
|
| 1579 |
+
"epoch": 0.9299719887955182,
|
| 1580 |
+
"eval_logits/chosen": -3.078125,
|
| 1581 |
+
"eval_logits/rejected": -3.125,
|
| 1582 |
+
"eval_logps/chosen": -350.0,
|
| 1583 |
+
"eval_logps/rejected": -316.0,
|
| 1584 |
+
"eval_loss": 0.6026737093925476,
|
| 1585 |
+
"eval_rewards/accuracies": 0.796875,
|
| 1586 |
+
"eval_rewards/chosen": -0.0888671875,
|
| 1587 |
+
"eval_rewards/margins": 0.2490234375,
|
| 1588 |
+
"eval_rewards/rejected": -0.337890625,
|
| 1589 |
+
"eval_runtime": 134.151,
|
| 1590 |
+
"eval_samples_per_second": 11.085,
|
| 1591 |
+
"eval_steps_per_second": 0.179,
|
| 1592 |
+
"step": 166
|
| 1593 |
+
},
|
| 1594 |
+
{
|
| 1595 |
+
"epoch": 0.9411764705882353,
|
| 1596 |
+
"eval_logits/chosen": -3.078125,
|
| 1597 |
+
"eval_logits/rejected": -3.125,
|
| 1598 |
+
"eval_logps/chosen": -350.0,
|
| 1599 |
+
"eval_logps/rejected": -318.0,
|
| 1600 |
+
"eval_loss": 0.6025987267494202,
|
| 1601 |
+
"eval_rewards/accuracies": 0.7942708134651184,
|
| 1602 |
+
"eval_rewards/chosen": -0.0888671875,
|
| 1603 |
+
"eval_rewards/margins": 0.2490234375,
|
| 1604 |
+
"eval_rewards/rejected": -0.337890625,
|
| 1605 |
+
"eval_runtime": 135.6171,
|
| 1606 |
+
"eval_samples_per_second": 10.965,
|
| 1607 |
+
"eval_steps_per_second": 0.177,
|
| 1608 |
+
"step": 168
|
| 1609 |
+
},
|
| 1610 |
+
{
|
| 1611 |
+
"epoch": 0.9523809523809523,
|
| 1612 |
+
"grad_norm": 6.811460693897883,
|
| 1613 |
+
"learning_rate": 6.15582970243117e-10,
|
| 1614 |
+
"logits/chosen": -3.09375,
|
| 1615 |
+
"logits/rejected": -3.125,
|
| 1616 |
+
"logps/chosen": -336.0,
|
| 1617 |
+
"logps/rejected": -312.0,
|
| 1618 |
+
"loss": 0.5974,
|
| 1619 |
+
"rewards/accuracies": 0.807812511920929,
|
| 1620 |
+
"rewards/chosen": -0.083984375,
|
| 1621 |
+
"rewards/margins": 0.26953125,
|
| 1622 |
+
"rewards/rejected": -0.353515625,
|
| 1623 |
+
"step": 170
|
| 1624 |
+
},
|
| 1625 |
+
{
|
| 1626 |
+
"epoch": 0.9523809523809523,
|
| 1627 |
+
"eval_logits/chosen": -3.078125,
|
| 1628 |
+
"eval_logits/rejected": -3.125,
|
| 1629 |
+
"eval_logps/chosen": -350.0,
|
| 1630 |
+
"eval_logps/rejected": -318.0,
|
| 1631 |
+
"eval_loss": 0.6025583148002625,
|
| 1632 |
+
"eval_rewards/accuracies": 0.7916666865348816,
|
| 1633 |
+
"eval_rewards/chosen": -0.08935546875,
|
| 1634 |
+
"eval_rewards/margins": 0.248046875,
|
| 1635 |
+
"eval_rewards/rejected": -0.337890625,
|
| 1636 |
+
"eval_runtime": 133.2423,
|
| 1637 |
+
"eval_samples_per_second": 11.16,
|
| 1638 |
+
"eval_steps_per_second": 0.18,
|
| 1639 |
+
"step": 170
|
| 1640 |
+
},
|
| 1641 |
+
{
|
| 1642 |
+
"epoch": 0.9635854341736695,
|
| 1643 |
+
"eval_logits/chosen": -3.078125,
|
| 1644 |
+
"eval_logits/rejected": -3.125,
|
| 1645 |
+
"eval_logps/chosen": -350.0,
|
| 1646 |
+
"eval_logps/rejected": -318.0,
|
| 1647 |
+
"eval_loss": 0.6024807691574097,
|
| 1648 |
+
"eval_rewards/accuracies": 0.7916666865348816,
|
| 1649 |
+
"eval_rewards/chosen": -0.08837890625,
|
| 1650 |
+
"eval_rewards/margins": 0.25,
|
| 1651 |
+
"eval_rewards/rejected": -0.337890625,
|
| 1652 |
+
"eval_runtime": 134.1414,
|
| 1653 |
+
"eval_samples_per_second": 11.085,
|
| 1654 |
+
"eval_steps_per_second": 0.179,
|
| 1655 |
+
"step": 172
|
| 1656 |
+
},
|
| 1657 |
+
{
|
| 1658 |
+
"epoch": 0.9747899159663865,
|
| 1659 |
+
"eval_logits/chosen": -3.078125,
|
| 1660 |
+
"eval_logits/rejected": -3.125,
|
| 1661 |
+
"eval_logps/chosen": -350.0,
|
| 1662 |
+
"eval_logps/rejected": -318.0,
|
| 1663 |
+
"eval_loss": 0.602593183517456,
|
| 1664 |
+
"eval_rewards/accuracies": 0.7942708134651184,
|
| 1665 |
+
"eval_rewards/chosen": -0.08935546875,
|
| 1666 |
+
"eval_rewards/margins": 0.2490234375,
|
| 1667 |
+
"eval_rewards/rejected": -0.337890625,
|
| 1668 |
+
"eval_runtime": 133.5195,
|
| 1669 |
+
"eval_samples_per_second": 11.137,
|
| 1670 |
+
"eval_steps_per_second": 0.18,
|
| 1671 |
+
"step": 174
|
| 1672 |
+
},
|
| 1673 |
+
{
|
| 1674 |
+
"epoch": 0.9859943977591037,
|
| 1675 |
+
"eval_logits/chosen": -3.078125,
|
| 1676 |
+
"eval_logits/rejected": -3.125,
|
| 1677 |
+
"eval_logps/chosen": -350.0,
|
| 1678 |
+
"eval_logps/rejected": -316.0,
|
| 1679 |
+
"eval_loss": 0.6026169061660767,
|
| 1680 |
+
"eval_rewards/accuracies": 0.796875,
|
| 1681 |
+
"eval_rewards/chosen": -0.08984375,
|
| 1682 |
+
"eval_rewards/margins": 0.248046875,
|
| 1683 |
+
"eval_rewards/rejected": -0.337890625,
|
| 1684 |
+
"eval_runtime": 134.5308,
|
| 1685 |
+
"eval_samples_per_second": 11.053,
|
| 1686 |
+
"eval_steps_per_second": 0.178,
|
| 1687 |
+
"step": 176
|
| 1688 |
+
},
|
| 1689 |
+
{
|
| 1690 |
+
"epoch": 0.9971988795518207,
|
| 1691 |
+
"eval_logits/chosen": -3.078125,
|
| 1692 |
+
"eval_logits/rejected": -3.125,
|
| 1693 |
+
"eval_logps/chosen": -350.0,
|
| 1694 |
+
"eval_logps/rejected": -316.0,
|
| 1695 |
+
"eval_loss": 0.6027360558509827,
|
| 1696 |
+
"eval_rewards/accuracies": 0.7916666865348816,
|
| 1697 |
+
"eval_rewards/chosen": -0.0888671875,
|
| 1698 |
+
"eval_rewards/margins": 0.248046875,
|
| 1699 |
+
"eval_rewards/rejected": -0.337890625,
|
| 1700 |
+
"eval_runtime": 134.9547,
|
| 1701 |
+
"eval_samples_per_second": 11.019,
|
| 1702 |
+
"eval_steps_per_second": 0.178,
|
| 1703 |
+
"step": 178
|
| 1704 |
+
},
|
| 1705 |
+
{
|
| 1706 |
+
"epoch": 0.9971988795518207,
|
| 1707 |
+
"step": 178,
|
| 1708 |
+
"total_flos": 0.0,
|
| 1709 |
+
"train_loss": 0.6403918319873596,
|
| 1710 |
+
"train_runtime": 36690.4382,
|
| 1711 |
+
"train_samples_per_second": 1.243,
|
| 1712 |
+
"train_steps_per_second": 0.005
|
| 1713 |
+
}
|
| 1714 |
+
],
|
| 1715 |
+
"logging_steps": 10,
|
| 1716 |
+
"max_steps": 178,
|
| 1717 |
+
"num_input_tokens_seen": 0,
|
| 1718 |
+
"num_train_epochs": 1,
|
| 1719 |
+
"save_steps": 2,
|
| 1720 |
+
"stateful_callbacks": {
|
| 1721 |
+
"TrainerControl": {
|
| 1722 |
+
"args": {
|
| 1723 |
+
"should_epoch_stop": false,
|
| 1724 |
+
"should_evaluate": false,
|
| 1725 |
+
"should_log": false,
|
| 1726 |
+
"should_save": true,
|
| 1727 |
+
"should_training_stop": true
|
| 1728 |
+
},
|
| 1729 |
+
"attributes": {}
|
| 1730 |
+
}
|
| 1731 |
+
},
|
| 1732 |
+
"total_flos": 0.0,
|
| 1733 |
+
"train_batch_size": 32,
|
| 1734 |
+
"trial_name": null,
|
| 1735 |
+
"trial_params": null
|
| 1736 |
+
}
|