Model save
Browse files- README.md +68 -0
- all_results.json +9 -0
- generation_config.json +6 -0
- model-00001-of-00003.safetensors +1 -1
- model-00002-of-00003.safetensors +1 -1
- model-00003-of-00003.safetensors +1 -1
- train_results.json +9 -0
- trainer_state.json +2546 -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_ebs64_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_ebs64_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_ebs64_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/ultrafeedbackSkyworkAgree_alignmentZephyr7BSftFull_sdpo_score_ebs64_lr1e-07_0try1prJ7sto2sYHlSCIOJJTTEhdKWJiU3dhkNo0dSnqgrczqZp)
|
| 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: 4.46.3
|
| 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": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5638467967928208,
|
| 5 |
+
"train_runtime": 32082.6297,
|
| 6 |
+
"train_samples": 45608,
|
| 7 |
+
"train_samples_per_second": 1.422,
|
| 8 |
+
"train_steps_per_second": 0.022
|
| 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": "4.46.3"
|
| 6 |
+
}
|
model-00001-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4943162336
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:961b26ebfb4b70fee969a499072dbadc0e5b980ab8278407269561cd99946597
|
| 3 |
size 4943162336
|
model-00002-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4999819336
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b468511ff6adc1a7bb835581d7f713f6d5a27f40bf0a97436a833067c658c7ed
|
| 3 |
size 4999819336
|
model-00003-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4540516344
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:796bbafa99f0bd0d7a660d35bdde78d861ab57505eef01785075d37482b2d86c
|
| 3 |
size 4540516344
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5638467967928208,
|
| 5 |
+
"train_runtime": 32082.6297,
|
| 6 |
+
"train_samples": 45608,
|
| 7 |
+
"train_samples_per_second": 1.422,
|
| 8 |
+
"train_steps_per_second": 0.022
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,2546 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.0,
|
| 5 |
+
"eval_steps": 8,
|
| 6 |
+
"global_step": 713,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.001402524544179523,
|
| 13 |
+
"grad_norm": 13.312526408630161,
|
| 14 |
+
"learning_rate": 1.3888888888888888e-09,
|
| 15 |
+
"logits/chosen": -3.09375,
|
| 16 |
+
"logits/rejected": -3.0,
|
| 17 |
+
"logps/chosen": -410.0,
|
| 18 |
+
"logps/rejected": -408.0,
|
| 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": 0.011220196353436185,
|
| 28 |
+
"eval_logits/chosen": -3.234375,
|
| 29 |
+
"eval_logits/rejected": -3.28125,
|
| 30 |
+
"eval_logps/chosen": -322.0,
|
| 31 |
+
"eval_logps/rejected": -276.0,
|
| 32 |
+
"eval_loss": 0.6922186017036438,
|
| 33 |
+
"eval_rewards/accuracies": 0.228723406791687,
|
| 34 |
+
"eval_rewards/chosen": -0.0003604888916015625,
|
| 35 |
+
"eval_rewards/margins": -0.001190185546875,
|
| 36 |
+
"eval_rewards/rejected": 0.000823974609375,
|
| 37 |
+
"eval_runtime": 64.4932,
|
| 38 |
+
"eval_samples_per_second": 23.057,
|
| 39 |
+
"eval_steps_per_second": 0.729,
|
| 40 |
+
"step": 8
|
| 41 |
+
},
|
| 42 |
+
{
|
| 43 |
+
"epoch": 0.014025245441795231,
|
| 44 |
+
"grad_norm": 10.597327421994228,
|
| 45 |
+
"learning_rate": 1.3888888888888889e-08,
|
| 46 |
+
"logits/chosen": -3.1875,
|
| 47 |
+
"logits/rejected": -3.125,
|
| 48 |
+
"logps/chosen": -310.0,
|
| 49 |
+
"logps/rejected": -286.0,
|
| 50 |
+
"loss": 0.6921,
|
| 51 |
+
"rewards/accuracies": 0.2222222238779068,
|
| 52 |
+
"rewards/chosen": 0.00128936767578125,
|
| 53 |
+
"rewards/margins": 0.00201416015625,
|
| 54 |
+
"rewards/rejected": -0.000728607177734375,
|
| 55 |
+
"step": 10
|
| 56 |
+
},
|
| 57 |
+
{
|
| 58 |
+
"epoch": 0.02244039270687237,
|
| 59 |
+
"eval_logits/chosen": -3.234375,
|
| 60 |
+
"eval_logits/rejected": -3.28125,
|
| 61 |
+
"eval_logps/chosen": -322.0,
|
| 62 |
+
"eval_logps/rejected": -276.0,
|
| 63 |
+
"eval_loss": 0.6920655369758606,
|
| 64 |
+
"eval_rewards/accuracies": 0.18617020547389984,
|
| 65 |
+
"eval_rewards/chosen": -0.0014801025390625,
|
| 66 |
+
"eval_rewards/margins": -0.0022125244140625,
|
| 67 |
+
"eval_rewards/rejected": 0.00074005126953125,
|
| 68 |
+
"eval_runtime": 66.7125,
|
| 69 |
+
"eval_samples_per_second": 22.29,
|
| 70 |
+
"eval_steps_per_second": 0.705,
|
| 71 |
+
"step": 16
|
| 72 |
+
},
|
| 73 |
+
{
|
| 74 |
+
"epoch": 0.028050490883590462,
|
| 75 |
+
"grad_norm": 13.836822232155031,
|
| 76 |
+
"learning_rate": 2.7777777777777777e-08,
|
| 77 |
+
"logits/chosen": -3.171875,
|
| 78 |
+
"logits/rejected": -3.25,
|
| 79 |
+
"logps/chosen": -324.0,
|
| 80 |
+
"logps/rejected": -255.0,
|
| 81 |
+
"loss": 0.6922,
|
| 82 |
+
"rewards/accuracies": 0.20000000298023224,
|
| 83 |
+
"rewards/chosen": -0.000904083251953125,
|
| 84 |
+
"rewards/margins": 0.0003910064697265625,
|
| 85 |
+
"rewards/rejected": -0.0012969970703125,
|
| 86 |
+
"step": 20
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 0.033660589060308554,
|
| 90 |
+
"eval_logits/chosen": -3.234375,
|
| 91 |
+
"eval_logits/rejected": -3.28125,
|
| 92 |
+
"eval_logps/chosen": -322.0,
|
| 93 |
+
"eval_logps/rejected": -276.0,
|
| 94 |
+
"eval_loss": 0.6919813752174377,
|
| 95 |
+
"eval_rewards/accuracies": 0.25,
|
| 96 |
+
"eval_rewards/chosen": -0.00095367431640625,
|
| 97 |
+
"eval_rewards/margins": -0.00067138671875,
|
| 98 |
+
"eval_rewards/rejected": -0.0002803802490234375,
|
| 99 |
+
"eval_runtime": 65.6733,
|
| 100 |
+
"eval_samples_per_second": 22.642,
|
| 101 |
+
"eval_steps_per_second": 0.716,
|
| 102 |
+
"step": 24
|
| 103 |
+
},
|
| 104 |
+
{
|
| 105 |
+
"epoch": 0.04207573632538569,
|
| 106 |
+
"grad_norm": 9.808894950380177,
|
| 107 |
+
"learning_rate": 4.166666666666667e-08,
|
| 108 |
+
"logits/chosen": -3.1875,
|
| 109 |
+
"logits/rejected": -3.1875,
|
| 110 |
+
"logps/chosen": -322.0,
|
| 111 |
+
"logps/rejected": -280.0,
|
| 112 |
+
"loss": 0.6917,
|
| 113 |
+
"rewards/accuracies": 0.3375000059604645,
|
| 114 |
+
"rewards/chosen": 0.00023365020751953125,
|
| 115 |
+
"rewards/margins": 0.0014190673828125,
|
| 116 |
+
"rewards/rejected": -0.001190185546875,
|
| 117 |
+
"step": 30
|
| 118 |
+
},
|
| 119 |
+
{
|
| 120 |
+
"epoch": 0.04488078541374474,
|
| 121 |
+
"eval_logits/chosen": -3.234375,
|
| 122 |
+
"eval_logits/rejected": -3.28125,
|
| 123 |
+
"eval_logps/chosen": -322.0,
|
| 124 |
+
"eval_logps/rejected": -276.0,
|
| 125 |
+
"eval_loss": 0.691659152507782,
|
| 126 |
+
"eval_rewards/accuracies": 0.24468085169792175,
|
| 127 |
+
"eval_rewards/chosen": -0.000888824462890625,
|
| 128 |
+
"eval_rewards/margins": 0.0002002716064453125,
|
| 129 |
+
"eval_rewards/rejected": -0.00109100341796875,
|
| 130 |
+
"eval_runtime": 65.5796,
|
| 131 |
+
"eval_samples_per_second": 22.675,
|
| 132 |
+
"eval_steps_per_second": 0.717,
|
| 133 |
+
"step": 32
|
| 134 |
+
},
|
| 135 |
+
{
|
| 136 |
+
"epoch": 0.056100981767180924,
|
| 137 |
+
"grad_norm": 10.03150100621376,
|
| 138 |
+
"learning_rate": 5.5555555555555555e-08,
|
| 139 |
+
"logits/chosen": -3.234375,
|
| 140 |
+
"logits/rejected": -3.234375,
|
| 141 |
+
"logps/chosen": -324.0,
|
| 142 |
+
"logps/rejected": -240.0,
|
| 143 |
+
"loss": 0.6919,
|
| 144 |
+
"rewards/accuracies": 0.1875,
|
| 145 |
+
"rewards/chosen": -0.002655029296875,
|
| 146 |
+
"rewards/margins": -0.001373291015625,
|
| 147 |
+
"rewards/rejected": -0.00128173828125,
|
| 148 |
+
"step": 40
|
| 149 |
+
},
|
| 150 |
+
{
|
| 151 |
+
"epoch": 0.056100981767180924,
|
| 152 |
+
"eval_logits/chosen": -3.234375,
|
| 153 |
+
"eval_logits/rejected": -3.28125,
|
| 154 |
+
"eval_logps/chosen": -322.0,
|
| 155 |
+
"eval_logps/rejected": -276.0,
|
| 156 |
+
"eval_loss": 0.6911664009094238,
|
| 157 |
+
"eval_rewards/accuracies": 0.3085106313228607,
|
| 158 |
+
"eval_rewards/chosen": -0.001922607421875,
|
| 159 |
+
"eval_rewards/margins": 0.00115203857421875,
|
| 160 |
+
"eval_rewards/rejected": -0.0030670166015625,
|
| 161 |
+
"eval_runtime": 65.5097,
|
| 162 |
+
"eval_samples_per_second": 22.699,
|
| 163 |
+
"eval_steps_per_second": 0.717,
|
| 164 |
+
"step": 40
|
| 165 |
+
},
|
| 166 |
+
{
|
| 167 |
+
"epoch": 0.06732117812061711,
|
| 168 |
+
"eval_logits/chosen": -3.234375,
|
| 169 |
+
"eval_logits/rejected": -3.265625,
|
| 170 |
+
"eval_logps/chosen": -322.0,
|
| 171 |
+
"eval_logps/rejected": -276.0,
|
| 172 |
+
"eval_loss": 0.6904101371765137,
|
| 173 |
+
"eval_rewards/accuracies": 0.4095744788646698,
|
| 174 |
+
"eval_rewards/chosen": -0.00360107421875,
|
| 175 |
+
"eval_rewards/margins": 0.0032958984375,
|
| 176 |
+
"eval_rewards/rejected": -0.00689697265625,
|
| 177 |
+
"eval_runtime": 65.8647,
|
| 178 |
+
"eval_samples_per_second": 22.577,
|
| 179 |
+
"eval_steps_per_second": 0.714,
|
| 180 |
+
"step": 48
|
| 181 |
+
},
|
| 182 |
+
{
|
| 183 |
+
"epoch": 0.07012622720897616,
|
| 184 |
+
"grad_norm": 11.024207801738104,
|
| 185 |
+
"learning_rate": 6.944444444444444e-08,
|
| 186 |
+
"logits/chosen": -3.03125,
|
| 187 |
+
"logits/rejected": -3.15625,
|
| 188 |
+
"logps/chosen": -342.0,
|
| 189 |
+
"logps/rejected": -306.0,
|
| 190 |
+
"loss": 0.6917,
|
| 191 |
+
"rewards/accuracies": 0.38749998807907104,
|
| 192 |
+
"rewards/chosen": -0.0007171630859375,
|
| 193 |
+
"rewards/margins": 0.004364013671875,
|
| 194 |
+
"rewards/rejected": -0.005096435546875,
|
| 195 |
+
"step": 50
|
| 196 |
+
},
|
| 197 |
+
{
|
| 198 |
+
"epoch": 0.0785413744740533,
|
| 199 |
+
"eval_logits/chosen": -3.234375,
|
| 200 |
+
"eval_logits/rejected": -3.265625,
|
| 201 |
+
"eval_logps/chosen": -322.0,
|
| 202 |
+
"eval_logps/rejected": -276.0,
|
| 203 |
+
"eval_loss": 0.6893861293792725,
|
| 204 |
+
"eval_rewards/accuracies": 0.5,
|
| 205 |
+
"eval_rewards/chosen": -0.007110595703125,
|
| 206 |
+
"eval_rewards/margins": 0.006195068359375,
|
| 207 |
+
"eval_rewards/rejected": -0.0133056640625,
|
| 208 |
+
"eval_runtime": 66.2121,
|
| 209 |
+
"eval_samples_per_second": 22.458,
|
| 210 |
+
"eval_steps_per_second": 0.71,
|
| 211 |
+
"step": 56
|
| 212 |
+
},
|
| 213 |
+
{
|
| 214 |
+
"epoch": 0.08415147265077139,
|
| 215 |
+
"grad_norm": 11.128580526610264,
|
| 216 |
+
"learning_rate": 8.333333333333334e-08,
|
| 217 |
+
"logits/chosen": -3.1875,
|
| 218 |
+
"logits/rejected": -3.15625,
|
| 219 |
+
"logps/chosen": -286.0,
|
| 220 |
+
"logps/rejected": -266.0,
|
| 221 |
+
"loss": 0.6902,
|
| 222 |
+
"rewards/accuracies": 0.512499988079071,
|
| 223 |
+
"rewards/chosen": -0.00726318359375,
|
| 224 |
+
"rewards/margins": 0.005401611328125,
|
| 225 |
+
"rewards/rejected": -0.0126953125,
|
| 226 |
+
"step": 60
|
| 227 |
+
},
|
| 228 |
+
{
|
| 229 |
+
"epoch": 0.08976157082748948,
|
| 230 |
+
"eval_logits/chosen": -3.234375,
|
| 231 |
+
"eval_logits/rejected": -3.265625,
|
| 232 |
+
"eval_logps/chosen": -322.0,
|
| 233 |
+
"eval_logps/rejected": -278.0,
|
| 234 |
+
"eval_loss": 0.6874608993530273,
|
| 235 |
+
"eval_rewards/accuracies": 0.542553186416626,
|
| 236 |
+
"eval_rewards/chosen": -0.0118408203125,
|
| 237 |
+
"eval_rewards/margins": 0.01043701171875,
|
| 238 |
+
"eval_rewards/rejected": -0.0223388671875,
|
| 239 |
+
"eval_runtime": 65.3339,
|
| 240 |
+
"eval_samples_per_second": 22.76,
|
| 241 |
+
"eval_steps_per_second": 0.719,
|
| 242 |
+
"step": 64
|
| 243 |
+
},
|
| 244 |
+
{
|
| 245 |
+
"epoch": 0.09817671809256662,
|
| 246 |
+
"grad_norm": 11.049642821906803,
|
| 247 |
+
"learning_rate": 9.722222222222221e-08,
|
| 248 |
+
"logits/chosen": -3.21875,
|
| 249 |
+
"logits/rejected": -3.234375,
|
| 250 |
+
"logps/chosen": -290.0,
|
| 251 |
+
"logps/rejected": -280.0,
|
| 252 |
+
"loss": 0.6868,
|
| 253 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 254 |
+
"rewards/chosen": -0.0150146484375,
|
| 255 |
+
"rewards/margins": 0.01202392578125,
|
| 256 |
+
"rewards/rejected": -0.027099609375,
|
| 257 |
+
"step": 70
|
| 258 |
+
},
|
| 259 |
+
{
|
| 260 |
+
"epoch": 0.10098176718092566,
|
| 261 |
+
"eval_logits/chosen": -3.21875,
|
| 262 |
+
"eval_logits/rejected": -3.265625,
|
| 263 |
+
"eval_logps/chosen": -324.0,
|
| 264 |
+
"eval_logps/rejected": -280.0,
|
| 265 |
+
"eval_loss": 0.6850191354751587,
|
| 266 |
+
"eval_rewards/accuracies": 0.6117021441459656,
|
| 267 |
+
"eval_rewards/chosen": -0.0218505859375,
|
| 268 |
+
"eval_rewards/margins": 0.01531982421875,
|
| 269 |
+
"eval_rewards/rejected": -0.037109375,
|
| 270 |
+
"eval_runtime": 65.5283,
|
| 271 |
+
"eval_samples_per_second": 22.692,
|
| 272 |
+
"eval_steps_per_second": 0.717,
|
| 273 |
+
"step": 72
|
| 274 |
+
},
|
| 275 |
+
{
|
| 276 |
+
"epoch": 0.11220196353436185,
|
| 277 |
+
"grad_norm": 12.079775942720834,
|
| 278 |
+
"learning_rate": 9.996157197797842e-08,
|
| 279 |
+
"logits/chosen": -3.1875,
|
| 280 |
+
"logits/rejected": -3.234375,
|
| 281 |
+
"logps/chosen": -332.0,
|
| 282 |
+
"logps/rejected": -286.0,
|
| 283 |
+
"loss": 0.6839,
|
| 284 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 285 |
+
"rewards/chosen": -0.0262451171875,
|
| 286 |
+
"rewards/margins": 0.02587890625,
|
| 287 |
+
"rewards/rejected": -0.052001953125,
|
| 288 |
+
"step": 80
|
| 289 |
+
},
|
| 290 |
+
{
|
| 291 |
+
"epoch": 0.11220196353436185,
|
| 292 |
+
"eval_logits/chosen": -3.21875,
|
| 293 |
+
"eval_logits/rejected": -3.25,
|
| 294 |
+
"eval_logps/chosen": -326.0,
|
| 295 |
+
"eval_logps/rejected": -282.0,
|
| 296 |
+
"eval_loss": 0.6817147135734558,
|
| 297 |
+
"eval_rewards/accuracies": 0.6595744490623474,
|
| 298 |
+
"eval_rewards/chosen": -0.03564453125,
|
| 299 |
+
"eval_rewards/margins": 0.0233154296875,
|
| 300 |
+
"eval_rewards/rejected": -0.058837890625,
|
| 301 |
+
"eval_runtime": 65.1964,
|
| 302 |
+
"eval_samples_per_second": 22.808,
|
| 303 |
+
"eval_steps_per_second": 0.721,
|
| 304 |
+
"step": 80
|
| 305 |
+
},
|
| 306 |
+
{
|
| 307 |
+
"epoch": 0.12342215988779803,
|
| 308 |
+
"eval_logits/chosen": -3.203125,
|
| 309 |
+
"eval_logits/rejected": -3.25,
|
| 310 |
+
"eval_logps/chosen": -326.0,
|
| 311 |
+
"eval_logps/rejected": -284.0,
|
| 312 |
+
"eval_loss": 0.6778491139411926,
|
| 313 |
+
"eval_rewards/accuracies": 0.664893627166748,
|
| 314 |
+
"eval_rewards/chosen": -0.0517578125,
|
| 315 |
+
"eval_rewards/margins": 0.033935546875,
|
| 316 |
+
"eval_rewards/rejected": -0.08544921875,
|
| 317 |
+
"eval_runtime": 65.8317,
|
| 318 |
+
"eval_samples_per_second": 22.588,
|
| 319 |
+
"eval_steps_per_second": 0.714,
|
| 320 |
+
"step": 88
|
| 321 |
+
},
|
| 322 |
+
{
|
| 323 |
+
"epoch": 0.12622720897615708,
|
| 324 |
+
"grad_norm": 13.482170939282454,
|
| 325 |
+
"learning_rate": 9.980555936859367e-08,
|
| 326 |
+
"logits/chosen": -3.171875,
|
| 327 |
+
"logits/rejected": -3.15625,
|
| 328 |
+
"logps/chosen": -338.0,
|
| 329 |
+
"logps/rejected": -304.0,
|
| 330 |
+
"loss": 0.6802,
|
| 331 |
+
"rewards/accuracies": 0.612500011920929,
|
| 332 |
+
"rewards/chosen": -0.05029296875,
|
| 333 |
+
"rewards/margins": 0.0245361328125,
|
| 334 |
+
"rewards/rejected": -0.07470703125,
|
| 335 |
+
"step": 90
|
| 336 |
+
},
|
| 337 |
+
{
|
| 338 |
+
"epoch": 0.13464235624123422,
|
| 339 |
+
"eval_logits/chosen": -3.203125,
|
| 340 |
+
"eval_logits/rejected": -3.234375,
|
| 341 |
+
"eval_logps/chosen": -328.0,
|
| 342 |
+
"eval_logps/rejected": -286.0,
|
| 343 |
+
"eval_loss": 0.6740643978118896,
|
| 344 |
+
"eval_rewards/accuracies": 0.6702127456665039,
|
| 345 |
+
"eval_rewards/chosen": -0.07177734375,
|
| 346 |
+
"eval_rewards/margins": 0.041015625,
|
| 347 |
+
"eval_rewards/rejected": -0.11279296875,
|
| 348 |
+
"eval_runtime": 66.7185,
|
| 349 |
+
"eval_samples_per_second": 22.288,
|
| 350 |
+
"eval_steps_per_second": 0.704,
|
| 351 |
+
"step": 96
|
| 352 |
+
},
|
| 353 |
+
{
|
| 354 |
+
"epoch": 0.1402524544179523,
|
| 355 |
+
"grad_norm": 11.204042194314384,
|
| 356 |
+
"learning_rate": 9.952993480848836e-08,
|
| 357 |
+
"logits/chosen": -3.15625,
|
| 358 |
+
"logits/rejected": -3.171875,
|
| 359 |
+
"logps/chosen": -344.0,
|
| 360 |
+
"logps/rejected": -316.0,
|
| 361 |
+
"loss": 0.6744,
|
| 362 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 363 |
+
"rewards/chosen": -0.0732421875,
|
| 364 |
+
"rewards/margins": 0.03369140625,
|
| 365 |
+
"rewards/rejected": -0.10693359375,
|
| 366 |
+
"step": 100
|
| 367 |
+
},
|
| 368 |
+
{
|
| 369 |
+
"epoch": 0.1458625525946704,
|
| 370 |
+
"eval_logits/chosen": -3.203125,
|
| 371 |
+
"eval_logits/rejected": -3.234375,
|
| 372 |
+
"eval_logps/chosen": -330.0,
|
| 373 |
+
"eval_logps/rejected": -290.0,
|
| 374 |
+
"eval_loss": 0.6696762442588806,
|
| 375 |
+
"eval_rewards/accuracies": 0.6808510422706604,
|
| 376 |
+
"eval_rewards/chosen": -0.08447265625,
|
| 377 |
+
"eval_rewards/margins": 0.052490234375,
|
| 378 |
+
"eval_rewards/rejected": -0.13671875,
|
| 379 |
+
"eval_runtime": 65.3752,
|
| 380 |
+
"eval_samples_per_second": 22.746,
|
| 381 |
+
"eval_steps_per_second": 0.719,
|
| 382 |
+
"step": 104
|
| 383 |
+
},
|
| 384 |
+
{
|
| 385 |
+
"epoch": 0.15427769985974754,
|
| 386 |
+
"grad_norm": 10.547622004632332,
|
| 387 |
+
"learning_rate": 9.913536023162564e-08,
|
| 388 |
+
"logits/chosen": -3.109375,
|
| 389 |
+
"logits/rejected": -3.15625,
|
| 390 |
+
"logps/chosen": -338.0,
|
| 391 |
+
"logps/rejected": -284.0,
|
| 392 |
+
"loss": 0.6694,
|
| 393 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 394 |
+
"rewards/chosen": -0.0966796875,
|
| 395 |
+
"rewards/margins": 0.034423828125,
|
| 396 |
+
"rewards/rejected": -0.130859375,
|
| 397 |
+
"step": 110
|
| 398 |
+
},
|
| 399 |
+
{
|
| 400 |
+
"epoch": 0.1570827489481066,
|
| 401 |
+
"eval_logits/chosen": -3.1875,
|
| 402 |
+
"eval_logits/rejected": -3.21875,
|
| 403 |
+
"eval_logps/chosen": -332.0,
|
| 404 |
+
"eval_logps/rejected": -292.0,
|
| 405 |
+
"eval_loss": 0.6651813387870789,
|
| 406 |
+
"eval_rewards/accuracies": 0.6808510422706604,
|
| 407 |
+
"eval_rewards/chosen": -0.09912109375,
|
| 408 |
+
"eval_rewards/margins": 0.0625,
|
| 409 |
+
"eval_rewards/rejected": -0.1611328125,
|
| 410 |
+
"eval_runtime": 66.8719,
|
| 411 |
+
"eval_samples_per_second": 22.237,
|
| 412 |
+
"eval_steps_per_second": 0.703,
|
| 413 |
+
"step": 112
|
| 414 |
+
},
|
| 415 |
+
{
|
| 416 |
+
"epoch": 0.16830294530154277,
|
| 417 |
+
"grad_norm": 11.442066252813966,
|
| 418 |
+
"learning_rate": 9.862278323974797e-08,
|
| 419 |
+
"logits/chosen": -3.25,
|
| 420 |
+
"logits/rejected": -3.25,
|
| 421 |
+
"logps/chosen": -332.0,
|
| 422 |
+
"logps/rejected": -266.0,
|
| 423 |
+
"loss": 0.6609,
|
| 424 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 425 |
+
"rewards/chosen": -0.11376953125,
|
| 426 |
+
"rewards/margins": 0.0927734375,
|
| 427 |
+
"rewards/rejected": -0.2060546875,
|
| 428 |
+
"step": 120
|
| 429 |
+
},
|
| 430 |
+
{
|
| 431 |
+
"epoch": 0.16830294530154277,
|
| 432 |
+
"eval_logits/chosen": -3.1875,
|
| 433 |
+
"eval_logits/rejected": -3.21875,
|
| 434 |
+
"eval_logps/chosen": -334.0,
|
| 435 |
+
"eval_logps/rejected": -294.0,
|
| 436 |
+
"eval_loss": 0.6606339812278748,
|
| 437 |
+
"eval_rewards/accuracies": 0.707446813583374,
|
| 438 |
+
"eval_rewards/chosen": -0.11572265625,
|
| 439 |
+
"eval_rewards/margins": 0.07568359375,
|
| 440 |
+
"eval_rewards/rejected": -0.19140625,
|
| 441 |
+
"eval_runtime": 65.1545,
|
| 442 |
+
"eval_samples_per_second": 22.823,
|
| 443 |
+
"eval_steps_per_second": 0.721,
|
| 444 |
+
"step": 120
|
| 445 |
+
},
|
| 446 |
+
{
|
| 447 |
+
"epoch": 0.17952314165497896,
|
| 448 |
+
"eval_logits/chosen": -3.171875,
|
| 449 |
+
"eval_logits/rejected": -3.203125,
|
| 450 |
+
"eval_logps/chosen": -334.0,
|
| 451 |
+
"eval_logps/rejected": -298.0,
|
| 452 |
+
"eval_loss": 0.6557604074478149,
|
| 453 |
+
"eval_rewards/accuracies": 0.7234042286872864,
|
| 454 |
+
"eval_rewards/chosen": -0.126953125,
|
| 455 |
+
"eval_rewards/margins": 0.0888671875,
|
| 456 |
+
"eval_rewards/rejected": -0.2158203125,
|
| 457 |
+
"eval_runtime": 65.0762,
|
| 458 |
+
"eval_samples_per_second": 22.85,
|
| 459 |
+
"eval_steps_per_second": 0.722,
|
| 460 |
+
"step": 128
|
| 461 |
+
},
|
| 462 |
+
{
|
| 463 |
+
"epoch": 0.182328190743338,
|
| 464 |
+
"grad_norm": 10.846341171274817,
|
| 465 |
+
"learning_rate": 9.79934348266374e-08,
|
| 466 |
+
"logits/chosen": -2.984375,
|
| 467 |
+
"logits/rejected": -3.1875,
|
| 468 |
+
"logps/chosen": -390.0,
|
| 469 |
+
"logps/rejected": -352.0,
|
| 470 |
+
"loss": 0.6616,
|
| 471 |
+
"rewards/accuracies": 0.699999988079071,
|
| 472 |
+
"rewards/chosen": -0.11474609375,
|
| 473 |
+
"rewards/margins": 0.08251953125,
|
| 474 |
+
"rewards/rejected": -0.197265625,
|
| 475 |
+
"step": 130
|
| 476 |
+
},
|
| 477 |
+
{
|
| 478 |
+
"epoch": 0.19074333800841514,
|
| 479 |
+
"eval_logits/chosen": -3.171875,
|
| 480 |
+
"eval_logits/rejected": -3.203125,
|
| 481 |
+
"eval_logps/chosen": -334.0,
|
| 482 |
+
"eval_logps/rejected": -298.0,
|
| 483 |
+
"eval_loss": 0.6506365537643433,
|
| 484 |
+
"eval_rewards/accuracies": 0.7234042286872864,
|
| 485 |
+
"eval_rewards/chosen": -0.12890625,
|
| 486 |
+
"eval_rewards/margins": 0.10205078125,
|
| 487 |
+
"eval_rewards/rejected": -0.2314453125,
|
| 488 |
+
"eval_runtime": 65.5912,
|
| 489 |
+
"eval_samples_per_second": 22.671,
|
| 490 |
+
"eval_steps_per_second": 0.717,
|
| 491 |
+
"step": 136
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.19635343618513323,
|
| 495 |
+
"grad_norm": 11.36989829334774,
|
| 496 |
+
"learning_rate": 9.724882642178755e-08,
|
| 497 |
+
"logits/chosen": -3.140625,
|
| 498 |
+
"logits/rejected": -3.203125,
|
| 499 |
+
"logps/chosen": -314.0,
|
| 500 |
+
"logps/rejected": -296.0,
|
| 501 |
+
"loss": 0.6564,
|
| 502 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 503 |
+
"rewards/chosen": -0.1533203125,
|
| 504 |
+
"rewards/margins": 0.09765625,
|
| 505 |
+
"rewards/rejected": -0.251953125,
|
| 506 |
+
"step": 140
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 0.20196353436185133,
|
| 510 |
+
"eval_logits/chosen": -3.171875,
|
| 511 |
+
"eval_logits/rejected": -3.203125,
|
| 512 |
+
"eval_logps/chosen": -334.0,
|
| 513 |
+
"eval_logps/rejected": -300.0,
|
| 514 |
+
"eval_loss": 0.6449890732765198,
|
| 515 |
+
"eval_rewards/accuracies": 0.7234042286872864,
|
| 516 |
+
"eval_rewards/chosen": -0.130859375,
|
| 517 |
+
"eval_rewards/margins": 0.11474609375,
|
| 518 |
+
"eval_rewards/rejected": -0.2451171875,
|
| 519 |
+
"eval_runtime": 65.0,
|
| 520 |
+
"eval_samples_per_second": 22.877,
|
| 521 |
+
"eval_steps_per_second": 0.723,
|
| 522 |
+
"step": 144
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 0.21037868162692847,
|
| 526 |
+
"grad_norm": 10.745243280453327,
|
| 527 |
+
"learning_rate": 9.63907462605873e-08,
|
| 528 |
+
"logits/chosen": -3.203125,
|
| 529 |
+
"logits/rejected": -3.140625,
|
| 530 |
+
"logps/chosen": -332.0,
|
| 531 |
+
"logps/rejected": -306.0,
|
| 532 |
+
"loss": 0.6413,
|
| 533 |
+
"rewards/accuracies": 0.737500011920929,
|
| 534 |
+
"rewards/chosen": -0.1513671875,
|
| 535 |
+
"rewards/margins": 0.09423828125,
|
| 536 |
+
"rewards/rejected": -0.24609375,
|
| 537 |
+
"step": 150
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.2131837307152875,
|
| 541 |
+
"eval_logits/chosen": -3.15625,
|
| 542 |
+
"eval_logits/rejected": -3.1875,
|
| 543 |
+
"eval_logps/chosen": -336.0,
|
| 544 |
+
"eval_logps/rejected": -304.0,
|
| 545 |
+
"eval_loss": 0.6383941173553467,
|
| 546 |
+
"eval_rewards/accuracies": 0.7234042286872864,
|
| 547 |
+
"eval_rewards/chosen": -0.1435546875,
|
| 548 |
+
"eval_rewards/margins": 0.1337890625,
|
| 549 |
+
"eval_rewards/rejected": -0.27734375,
|
| 550 |
+
"eval_runtime": 65.4753,
|
| 551 |
+
"eval_samples_per_second": 22.711,
|
| 552 |
+
"eval_steps_per_second": 0.718,
|
| 553 |
+
"step": 152
|
| 554 |
+
},
|
| 555 |
+
{
|
| 556 |
+
"epoch": 0.2244039270687237,
|
| 557 |
+
"grad_norm": 11.27553371495036,
|
| 558 |
+
"learning_rate": 9.542125508973355e-08,
|
| 559 |
+
"logits/chosen": -3.03125,
|
| 560 |
+
"logits/rejected": -3.078125,
|
| 561 |
+
"logps/chosen": -368.0,
|
| 562 |
+
"logps/rejected": -282.0,
|
| 563 |
+
"loss": 0.6307,
|
| 564 |
+
"rewards/accuracies": 0.75,
|
| 565 |
+
"rewards/chosen": -0.12353515625,
|
| 566 |
+
"rewards/margins": 0.189453125,
|
| 567 |
+
"rewards/rejected": -0.314453125,
|
| 568 |
+
"step": 160
|
| 569 |
+
},
|
| 570 |
+
{
|
| 571 |
+
"epoch": 0.2244039270687237,
|
| 572 |
+
"eval_logits/chosen": -3.15625,
|
| 573 |
+
"eval_logits/rejected": -3.1875,
|
| 574 |
+
"eval_logps/chosen": -336.0,
|
| 575 |
+
"eval_logps/rejected": -306.0,
|
| 576 |
+
"eval_loss": 0.6307098269462585,
|
| 577 |
+
"eval_rewards/accuracies": 0.728723406791687,
|
| 578 |
+
"eval_rewards/chosen": -0.1435546875,
|
| 579 |
+
"eval_rewards/margins": 0.154296875,
|
| 580 |
+
"eval_rewards/rejected": -0.296875,
|
| 581 |
+
"eval_runtime": 65.6298,
|
| 582 |
+
"eval_samples_per_second": 22.657,
|
| 583 |
+
"eval_steps_per_second": 0.716,
|
| 584 |
+
"step": 160
|
| 585 |
+
},
|
| 586 |
+
{
|
| 587 |
+
"epoch": 0.23562412342215988,
|
| 588 |
+
"eval_logits/chosen": -3.140625,
|
| 589 |
+
"eval_logits/rejected": -3.171875,
|
| 590 |
+
"eval_logps/chosen": -332.0,
|
| 591 |
+
"eval_logps/rejected": -304.0,
|
| 592 |
+
"eval_loss": 0.6231443881988525,
|
| 593 |
+
"eval_rewards/accuracies": 0.728723406791687,
|
| 594 |
+
"eval_rewards/chosen": -0.1123046875,
|
| 595 |
+
"eval_rewards/margins": 0.1689453125,
|
| 596 |
+
"eval_rewards/rejected": -0.28125,
|
| 597 |
+
"eval_runtime": 65.3606,
|
| 598 |
+
"eval_samples_per_second": 22.751,
|
| 599 |
+
"eval_steps_per_second": 0.719,
|
| 600 |
+
"step": 168
|
| 601 |
+
},
|
| 602 |
+
{
|
| 603 |
+
"epoch": 0.23842917251051893,
|
| 604 |
+
"grad_norm": 10.59379621805741,
|
| 605 |
+
"learning_rate": 9.434268121818663e-08,
|
| 606 |
+
"logits/chosen": -3.234375,
|
| 607 |
+
"logits/rejected": -3.203125,
|
| 608 |
+
"logps/chosen": -292.0,
|
| 609 |
+
"logps/rejected": -268.0,
|
| 610 |
+
"loss": 0.627,
|
| 611 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 612 |
+
"rewards/chosen": -0.18359375,
|
| 613 |
+
"rewards/margins": 0.16015625,
|
| 614 |
+
"rewards/rejected": -0.34375,
|
| 615 |
+
"step": 170
|
| 616 |
+
},
|
| 617 |
+
{
|
| 618 |
+
"epoch": 0.24684431977559607,
|
| 619 |
+
"eval_logits/chosen": -3.140625,
|
| 620 |
+
"eval_logits/rejected": -3.171875,
|
| 621 |
+
"eval_logps/chosen": -334.0,
|
| 622 |
+
"eval_logps/rejected": -308.0,
|
| 623 |
+
"eval_loss": 0.6149851679801941,
|
| 624 |
+
"eval_rewards/accuracies": 0.7393617033958435,
|
| 625 |
+
"eval_rewards/chosen": -0.126953125,
|
| 626 |
+
"eval_rewards/margins": 0.193359375,
|
| 627 |
+
"eval_rewards/rejected": -0.3203125,
|
| 628 |
+
"eval_runtime": 65.2812,
|
| 629 |
+
"eval_samples_per_second": 22.778,
|
| 630 |
+
"eval_steps_per_second": 0.72,
|
| 631 |
+
"step": 176
|
| 632 |
+
},
|
| 633 |
+
{
|
| 634 |
+
"epoch": 0.25245441795231416,
|
| 635 |
+
"grad_norm": 13.319281173183153,
|
| 636 |
+
"learning_rate": 9.315761492555401e-08,
|
| 637 |
+
"logits/chosen": -2.953125,
|
| 638 |
+
"logits/rejected": -3.03125,
|
| 639 |
+
"logps/chosen": -354.0,
|
| 640 |
+
"logps/rejected": -316.0,
|
| 641 |
+
"loss": 0.6139,
|
| 642 |
+
"rewards/accuracies": 0.699999988079071,
|
| 643 |
+
"rewards/chosen": -0.1259765625,
|
| 644 |
+
"rewards/margins": 0.1943359375,
|
| 645 |
+
"rewards/rejected": -0.3203125,
|
| 646 |
+
"step": 180
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"epoch": 0.25806451612903225,
|
| 650 |
+
"eval_logits/chosen": -3.125,
|
| 651 |
+
"eval_logits/rejected": -3.15625,
|
| 652 |
+
"eval_logps/chosen": -336.0,
|
| 653 |
+
"eval_logps/rejected": -310.0,
|
| 654 |
+
"eval_loss": 0.6060777902603149,
|
| 655 |
+
"eval_rewards/accuracies": 0.7340425252914429,
|
| 656 |
+
"eval_rewards/chosen": -0.1337890625,
|
| 657 |
+
"eval_rewards/margins": 0.2197265625,
|
| 658 |
+
"eval_rewards/rejected": -0.353515625,
|
| 659 |
+
"eval_runtime": 65.7676,
|
| 660 |
+
"eval_samples_per_second": 22.61,
|
| 661 |
+
"eval_steps_per_second": 0.715,
|
| 662 |
+
"step": 184
|
| 663 |
+
},
|
| 664 |
+
{
|
| 665 |
+
"epoch": 0.2664796633941094,
|
| 666 |
+
"grad_norm": 13.439011203083904,
|
| 667 |
+
"learning_rate": 9.186890224133106e-08,
|
| 668 |
+
"logits/chosen": -3.03125,
|
| 669 |
+
"logits/rejected": -3.09375,
|
| 670 |
+
"logps/chosen": -356.0,
|
| 671 |
+
"logps/rejected": -314.0,
|
| 672 |
+
"loss": 0.5955,
|
| 673 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 674 |
+
"rewards/chosen": -0.107421875,
|
| 675 |
+
"rewards/margins": 0.2236328125,
|
| 676 |
+
"rewards/rejected": -0.33203125,
|
| 677 |
+
"step": 190
|
| 678 |
+
},
|
| 679 |
+
{
|
| 680 |
+
"epoch": 0.26928471248246844,
|
| 681 |
+
"eval_logits/chosen": -3.125,
|
| 682 |
+
"eval_logits/rejected": -3.15625,
|
| 683 |
+
"eval_logps/chosen": -338.0,
|
| 684 |
+
"eval_logps/rejected": -316.0,
|
| 685 |
+
"eval_loss": 0.5989786386489868,
|
| 686 |
+
"eval_rewards/accuracies": 0.7446808218955994,
|
| 687 |
+
"eval_rewards/chosen": -0.1572265625,
|
| 688 |
+
"eval_rewards/margins": 0.248046875,
|
| 689 |
+
"eval_rewards/rejected": -0.404296875,
|
| 690 |
+
"eval_runtime": 67.4593,
|
| 691 |
+
"eval_samples_per_second": 22.043,
|
| 692 |
+
"eval_steps_per_second": 0.697,
|
| 693 |
+
"step": 192
|
| 694 |
+
},
|
| 695 |
+
{
|
| 696 |
+
"epoch": 0.2805049088359046,
|
| 697 |
+
"grad_norm": 15.24476795667356,
|
| 698 |
+
"learning_rate": 9.047963810993828e-08,
|
| 699 |
+
"logits/chosen": -2.96875,
|
| 700 |
+
"logits/rejected": -3.21875,
|
| 701 |
+
"logps/chosen": -340.0,
|
| 702 |
+
"logps/rejected": -332.0,
|
| 703 |
+
"loss": 0.5975,
|
| 704 |
+
"rewards/accuracies": 0.800000011920929,
|
| 705 |
+
"rewards/chosen": -0.220703125,
|
| 706 |
+
"rewards/margins": 0.318359375,
|
| 707 |
+
"rewards/rejected": -0.5390625,
|
| 708 |
+
"step": 200
|
| 709 |
+
},
|
| 710 |
+
{
|
| 711 |
+
"epoch": 0.2805049088359046,
|
| 712 |
+
"eval_logits/chosen": -3.125,
|
| 713 |
+
"eval_logits/rejected": -3.15625,
|
| 714 |
+
"eval_logps/chosen": -336.0,
|
| 715 |
+
"eval_logps/rejected": -318.0,
|
| 716 |
+
"eval_loss": 0.591622531414032,
|
| 717 |
+
"eval_rewards/accuracies": 0.728723406791687,
|
| 718 |
+
"eval_rewards/chosen": -0.1455078125,
|
| 719 |
+
"eval_rewards/margins": 0.275390625,
|
| 720 |
+
"eval_rewards/rejected": -0.421875,
|
| 721 |
+
"eval_runtime": 65.7395,
|
| 722 |
+
"eval_samples_per_second": 22.62,
|
| 723 |
+
"eval_steps_per_second": 0.715,
|
| 724 |
+
"step": 200
|
| 725 |
+
},
|
| 726 |
+
{
|
| 727 |
+
"epoch": 0.2917251051893408,
|
| 728 |
+
"eval_logits/chosen": -3.125,
|
| 729 |
+
"eval_logits/rejected": -3.15625,
|
| 730 |
+
"eval_logps/chosen": -334.0,
|
| 731 |
+
"eval_logps/rejected": -318.0,
|
| 732 |
+
"eval_loss": 0.5862420797348022,
|
| 733 |
+
"eval_rewards/accuracies": 0.7180851101875305,
|
| 734 |
+
"eval_rewards/chosen": -0.1279296875,
|
| 735 |
+
"eval_rewards/margins": 0.298828125,
|
| 736 |
+
"eval_rewards/rejected": -0.42578125,
|
| 737 |
+
"eval_runtime": 65.2479,
|
| 738 |
+
"eval_samples_per_second": 22.79,
|
| 739 |
+
"eval_steps_per_second": 0.72,
|
| 740 |
+
"step": 208
|
| 741 |
+
},
|
| 742 |
+
{
|
| 743 |
+
"epoch": 0.29453015427769985,
|
| 744 |
+
"grad_norm": 19.945956726222818,
|
| 745 |
+
"learning_rate": 8.899315895796999e-08,
|
| 746 |
+
"logits/chosen": -3.109375,
|
| 747 |
+
"logits/rejected": -3.171875,
|
| 748 |
+
"logps/chosen": -376.0,
|
| 749 |
+
"logps/rejected": -312.0,
|
| 750 |
+
"loss": 0.5887,
|
| 751 |
+
"rewards/accuracies": 0.800000011920929,
|
| 752 |
+
"rewards/chosen": -0.189453125,
|
| 753 |
+
"rewards/margins": 0.369140625,
|
| 754 |
+
"rewards/rejected": -0.55859375,
|
| 755 |
+
"step": 210
|
| 756 |
+
},
|
| 757 |
+
{
|
| 758 |
+
"epoch": 0.302945301542777,
|
| 759 |
+
"eval_logits/chosen": -3.125,
|
| 760 |
+
"eval_logits/rejected": -3.15625,
|
| 761 |
+
"eval_logps/chosen": -334.0,
|
| 762 |
+
"eval_logps/rejected": -318.0,
|
| 763 |
+
"eval_loss": 0.5828134417533875,
|
| 764 |
+
"eval_rewards/accuracies": 0.728723406791687,
|
| 765 |
+
"eval_rewards/chosen": -0.1171875,
|
| 766 |
+
"eval_rewards/margins": 0.310546875,
|
| 767 |
+
"eval_rewards/rejected": -0.427734375,
|
| 768 |
+
"eval_runtime": 65.4231,
|
| 769 |
+
"eval_samples_per_second": 22.729,
|
| 770 |
+
"eval_steps_per_second": 0.718,
|
| 771 |
+
"step": 216
|
| 772 |
+
},
|
| 773 |
+
{
|
| 774 |
+
"epoch": 0.3085553997194951,
|
| 775 |
+
"grad_norm": 13.610682023585936,
|
| 776 |
+
"learning_rate": 8.741303468150459e-08,
|
| 777 |
+
"logits/chosen": -3.140625,
|
| 778 |
+
"logits/rejected": -3.0625,
|
| 779 |
+
"logps/chosen": -262.0,
|
| 780 |
+
"logps/rejected": -272.0,
|
| 781 |
+
"loss": 0.5819,
|
| 782 |
+
"rewards/accuracies": 0.875,
|
| 783 |
+
"rewards/chosen": -0.13671875,
|
| 784 |
+
"rewards/margins": 0.3515625,
|
| 785 |
+
"rewards/rejected": -0.486328125,
|
| 786 |
+
"step": 220
|
| 787 |
+
},
|
| 788 |
+
{
|
| 789 |
+
"epoch": 0.3141654978962132,
|
| 790 |
+
"eval_logits/chosen": -3.125,
|
| 791 |
+
"eval_logits/rejected": -3.15625,
|
| 792 |
+
"eval_logps/chosen": -334.0,
|
| 793 |
+
"eval_logps/rejected": -320.0,
|
| 794 |
+
"eval_loss": 0.5790321826934814,
|
| 795 |
+
"eval_rewards/accuracies": 0.7446808218955994,
|
| 796 |
+
"eval_rewards/chosen": -0.126953125,
|
| 797 |
+
"eval_rewards/margins": 0.322265625,
|
| 798 |
+
"eval_rewards/rejected": -0.44921875,
|
| 799 |
+
"eval_runtime": 65.6686,
|
| 800 |
+
"eval_samples_per_second": 22.644,
|
| 801 |
+
"eval_steps_per_second": 0.716,
|
| 802 |
+
"step": 224
|
| 803 |
+
},
|
| 804 |
+
{
|
| 805 |
+
"epoch": 0.3225806451612903,
|
| 806 |
+
"grad_norm": 16.393253456859437,
|
| 807 |
+
"learning_rate": 8.574306007271956e-08,
|
| 808 |
+
"logits/chosen": -3.09375,
|
| 809 |
+
"logits/rejected": -3.203125,
|
| 810 |
+
"logps/chosen": -372.0,
|
| 811 |
+
"logps/rejected": -344.0,
|
| 812 |
+
"loss": 0.5694,
|
| 813 |
+
"rewards/accuracies": 0.762499988079071,
|
| 814 |
+
"rewards/chosen": -0.14453125,
|
| 815 |
+
"rewards/margins": 0.345703125,
|
| 816 |
+
"rewards/rejected": -0.490234375,
|
| 817 |
+
"step": 230
|
| 818 |
+
},
|
| 819 |
+
{
|
| 820 |
+
"epoch": 0.32538569424964936,
|
| 821 |
+
"eval_logits/chosen": -3.125,
|
| 822 |
+
"eval_logits/rejected": -3.15625,
|
| 823 |
+
"eval_logps/chosen": -340.0,
|
| 824 |
+
"eval_logps/rejected": -328.0,
|
| 825 |
+
"eval_loss": 0.5725518465042114,
|
| 826 |
+
"eval_rewards/accuracies": 0.7180851101875305,
|
| 827 |
+
"eval_rewards/chosen": -0.1748046875,
|
| 828 |
+
"eval_rewards/margins": 0.34765625,
|
| 829 |
+
"eval_rewards/rejected": -0.5234375,
|
| 830 |
+
"eval_runtime": 65.6729,
|
| 831 |
+
"eval_samples_per_second": 22.643,
|
| 832 |
+
"eval_steps_per_second": 0.716,
|
| 833 |
+
"step": 232
|
| 834 |
+
},
|
| 835 |
+
{
|
| 836 |
+
"epoch": 0.33660589060308554,
|
| 837 |
+
"grad_norm": 20.54581221908109,
|
| 838 |
+
"learning_rate": 8.398724570640106e-08,
|
| 839 |
+
"logits/chosen": -3.078125,
|
| 840 |
+
"logits/rejected": -3.0,
|
| 841 |
+
"logps/chosen": -386.0,
|
| 842 |
+
"logps/rejected": -390.0,
|
| 843 |
+
"loss": 0.5678,
|
| 844 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 845 |
+
"rewards/chosen": -0.1826171875,
|
| 846 |
+
"rewards/margins": 0.443359375,
|
| 847 |
+
"rewards/rejected": -0.625,
|
| 848 |
+
"step": 240
|
| 849 |
+
},
|
| 850 |
+
{
|
| 851 |
+
"epoch": 0.33660589060308554,
|
| 852 |
+
"eval_logits/chosen": -3.125,
|
| 853 |
+
"eval_logits/rejected": -3.140625,
|
| 854 |
+
"eval_logps/chosen": -342.0,
|
| 855 |
+
"eval_logps/rejected": -332.0,
|
| 856 |
+
"eval_loss": 0.5683075785636902,
|
| 857 |
+
"eval_rewards/accuracies": 0.7180851101875305,
|
| 858 |
+
"eval_rewards/chosen": -0.1943359375,
|
| 859 |
+
"eval_rewards/margins": 0.3671875,
|
| 860 |
+
"eval_rewards/rejected": -0.5625,
|
| 861 |
+
"eval_runtime": 66.0149,
|
| 862 |
+
"eval_samples_per_second": 22.525,
|
| 863 |
+
"eval_steps_per_second": 0.712,
|
| 864 |
+
"step": 240
|
| 865 |
+
},
|
| 866 |
+
{
|
| 867 |
+
"epoch": 0.34782608695652173,
|
| 868 |
+
"eval_logits/chosen": -3.125,
|
| 869 |
+
"eval_logits/rejected": -3.140625,
|
| 870 |
+
"eval_logps/chosen": -338.0,
|
| 871 |
+
"eval_logps/rejected": -330.0,
|
| 872 |
+
"eval_loss": 0.5656094551086426,
|
| 873 |
+
"eval_rewards/accuracies": 0.7606382966041565,
|
| 874 |
+
"eval_rewards/chosen": -0.16796875,
|
| 875 |
+
"eval_rewards/margins": 0.384765625,
|
| 876 |
+
"eval_rewards/rejected": -0.5546875,
|
| 877 |
+
"eval_runtime": 65.715,
|
| 878 |
+
"eval_samples_per_second": 22.628,
|
| 879 |
+
"eval_steps_per_second": 0.715,
|
| 880 |
+
"step": 248
|
| 881 |
+
},
|
| 882 |
+
{
|
| 883 |
+
"epoch": 0.3506311360448808,
|
| 884 |
+
"grad_norm": 18.794252770652808,
|
| 885 |
+
"learning_rate": 8.214980830823428e-08,
|
| 886 |
+
"logits/chosen": -3.03125,
|
| 887 |
+
"logits/rejected": -3.140625,
|
| 888 |
+
"logps/chosen": -306.0,
|
| 889 |
+
"logps/rejected": -288.0,
|
| 890 |
+
"loss": 0.5712,
|
| 891 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 892 |
+
"rewards/chosen": -0.2109375,
|
| 893 |
+
"rewards/margins": 0.373046875,
|
| 894 |
+
"rewards/rejected": -0.5859375,
|
| 895 |
+
"step": 250
|
| 896 |
+
},
|
| 897 |
+
{
|
| 898 |
+
"epoch": 0.3590462833099579,
|
| 899 |
+
"eval_logits/chosen": -3.125,
|
| 900 |
+
"eval_logits/rejected": -3.140625,
|
| 901 |
+
"eval_logps/chosen": -336.0,
|
| 902 |
+
"eval_logps/rejected": -330.0,
|
| 903 |
+
"eval_loss": 0.5647580027580261,
|
| 904 |
+
"eval_rewards/accuracies": 0.7819148898124695,
|
| 905 |
+
"eval_rewards/chosen": -0.142578125,
|
| 906 |
+
"eval_rewards/margins": 0.392578125,
|
| 907 |
+
"eval_rewards/rejected": -0.53515625,
|
| 908 |
+
"eval_runtime": 66.6553,
|
| 909 |
+
"eval_samples_per_second": 22.309,
|
| 910 |
+
"eval_steps_per_second": 0.705,
|
| 911 |
+
"step": 256
|
| 912 |
+
},
|
| 913 |
+
{
|
| 914 |
+
"epoch": 0.364656381486676,
|
| 915 |
+
"grad_norm": 12.474351601104692,
|
| 916 |
+
"learning_rate": 8.02351606280068e-08,
|
| 917 |
+
"logits/chosen": -3.09375,
|
| 918 |
+
"logits/rejected": -3.078125,
|
| 919 |
+
"logps/chosen": -356.0,
|
| 920 |
+
"logps/rejected": -378.0,
|
| 921 |
+
"loss": 0.5516,
|
| 922 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 923 |
+
"rewards/chosen": -0.212890625,
|
| 924 |
+
"rewards/margins": 0.435546875,
|
| 925 |
+
"rewards/rejected": -0.6484375,
|
| 926 |
+
"step": 260
|
| 927 |
+
},
|
| 928 |
+
{
|
| 929 |
+
"epoch": 0.3702664796633941,
|
| 930 |
+
"eval_logits/chosen": -3.109375,
|
| 931 |
+
"eval_logits/rejected": -3.140625,
|
| 932 |
+
"eval_logps/chosen": -334.0,
|
| 933 |
+
"eval_logps/rejected": -326.0,
|
| 934 |
+
"eval_loss": 0.5650196671485901,
|
| 935 |
+
"eval_rewards/accuracies": 0.771276593208313,
|
| 936 |
+
"eval_rewards/chosen": -0.11376953125,
|
| 937 |
+
"eval_rewards/margins": 0.392578125,
|
| 938 |
+
"eval_rewards/rejected": -0.5078125,
|
| 939 |
+
"eval_runtime": 65.7734,
|
| 940 |
+
"eval_samples_per_second": 22.608,
|
| 941 |
+
"eval_steps_per_second": 0.715,
|
| 942 |
+
"step": 264
|
| 943 |
+
},
|
| 944 |
+
{
|
| 945 |
+
"epoch": 0.37868162692847124,
|
| 946 |
+
"grad_norm": 16.401892485604222,
|
| 947 |
+
"learning_rate": 7.824790084204426e-08,
|
| 948 |
+
"logits/chosen": -3.0625,
|
| 949 |
+
"logits/rejected": -3.140625,
|
| 950 |
+
"logps/chosen": -342.0,
|
| 951 |
+
"logps/rejected": -382.0,
|
| 952 |
+
"loss": 0.5561,
|
| 953 |
+
"rewards/accuracies": 0.75,
|
| 954 |
+
"rewards/chosen": -0.11962890625,
|
| 955 |
+
"rewards/margins": 0.5,
|
| 956 |
+
"rewards/rejected": -0.62109375,
|
| 957 |
+
"step": 270
|
| 958 |
+
},
|
| 959 |
+
{
|
| 960 |
+
"epoch": 0.3814866760168303,
|
| 961 |
+
"eval_logits/chosen": -3.109375,
|
| 962 |
+
"eval_logits/rejected": -3.140625,
|
| 963 |
+
"eval_logps/chosen": -336.0,
|
| 964 |
+
"eval_logps/rejected": -332.0,
|
| 965 |
+
"eval_loss": 0.5578521490097046,
|
| 966 |
+
"eval_rewards/accuracies": 0.7659574747085571,
|
| 967 |
+
"eval_rewards/chosen": -0.1416015625,
|
| 968 |
+
"eval_rewards/margins": 0.421875,
|
| 969 |
+
"eval_rewards/rejected": -0.5625,
|
| 970 |
+
"eval_runtime": 65.6246,
|
| 971 |
+
"eval_samples_per_second": 22.659,
|
| 972 |
+
"eval_steps_per_second": 0.716,
|
| 973 |
+
"step": 272
|
| 974 |
+
},
|
| 975 |
+
{
|
| 976 |
+
"epoch": 0.39270687237026647,
|
| 977 |
+
"grad_norm": 15.668844063971891,
|
| 978 |
+
"learning_rate": 7.619280151032996e-08,
|
| 979 |
+
"logits/chosen": -3.078125,
|
| 980 |
+
"logits/rejected": -3.125,
|
| 981 |
+
"logps/chosen": -308.0,
|
| 982 |
+
"logps/rejected": -310.0,
|
| 983 |
+
"loss": 0.5511,
|
| 984 |
+
"rewards/accuracies": 0.75,
|
| 985 |
+
"rewards/chosen": -0.2041015625,
|
| 986 |
+
"rewards/margins": 0.412109375,
|
| 987 |
+
"rewards/rejected": -0.6171875,
|
| 988 |
+
"step": 280
|
| 989 |
+
},
|
| 990 |
+
{
|
| 991 |
+
"epoch": 0.39270687237026647,
|
| 992 |
+
"eval_logits/chosen": -3.109375,
|
| 993 |
+
"eval_logits/rejected": -3.140625,
|
| 994 |
+
"eval_logps/chosen": -338.0,
|
| 995 |
+
"eval_logps/rejected": -336.0,
|
| 996 |
+
"eval_loss": 0.5538753867149353,
|
| 997 |
+
"eval_rewards/accuracies": 0.7659574747085571,
|
| 998 |
+
"eval_rewards/chosen": -0.15625,
|
| 999 |
+
"eval_rewards/margins": 0.447265625,
|
| 1000 |
+
"eval_rewards/rejected": -0.6015625,
|
| 1001 |
+
"eval_runtime": 66.0541,
|
| 1002 |
+
"eval_samples_per_second": 22.512,
|
| 1003 |
+
"eval_steps_per_second": 0.712,
|
| 1004 |
+
"step": 280
|
| 1005 |
+
},
|
| 1006 |
+
{
|
| 1007 |
+
"epoch": 0.40392706872370265,
|
| 1008 |
+
"eval_logits/chosen": -3.109375,
|
| 1009 |
+
"eval_logits/rejected": -3.125,
|
| 1010 |
+
"eval_logps/chosen": -340.0,
|
| 1011 |
+
"eval_logps/rejected": -340.0,
|
| 1012 |
+
"eval_loss": 0.5501448512077332,
|
| 1013 |
+
"eval_rewards/accuracies": 0.771276593208313,
|
| 1014 |
+
"eval_rewards/chosen": -0.18359375,
|
| 1015 |
+
"eval_rewards/margins": 0.462890625,
|
| 1016 |
+
"eval_rewards/rejected": -0.6484375,
|
| 1017 |
+
"eval_runtime": 65.3015,
|
| 1018 |
+
"eval_samples_per_second": 22.771,
|
| 1019 |
+
"eval_steps_per_second": 0.72,
|
| 1020 |
+
"step": 288
|
| 1021 |
+
},
|
| 1022 |
+
{
|
| 1023 |
+
"epoch": 0.4067321178120617,
|
| 1024 |
+
"grad_norm": 16.413044382075544,
|
| 1025 |
+
"learning_rate": 7.407479811482827e-08,
|
| 1026 |
+
"logits/chosen": -3.09375,
|
| 1027 |
+
"logits/rejected": -3.046875,
|
| 1028 |
+
"logps/chosen": -364.0,
|
| 1029 |
+
"logps/rejected": -338.0,
|
| 1030 |
+
"loss": 0.5385,
|
| 1031 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1032 |
+
"rewards/chosen": -0.1630859375,
|
| 1033 |
+
"rewards/margins": 0.45703125,
|
| 1034 |
+
"rewards/rejected": -0.62109375,
|
| 1035 |
+
"step": 290
|
| 1036 |
+
},
|
| 1037 |
+
{
|
| 1038 |
+
"epoch": 0.41514726507713884,
|
| 1039 |
+
"eval_logits/chosen": -3.109375,
|
| 1040 |
+
"eval_logits/rejected": -3.125,
|
| 1041 |
+
"eval_logps/chosen": -344.0,
|
| 1042 |
+
"eval_logps/rejected": -344.0,
|
| 1043 |
+
"eval_loss": 0.5482958555221558,
|
| 1044 |
+
"eval_rewards/accuracies": 0.75,
|
| 1045 |
+
"eval_rewards/chosen": -0.2197265625,
|
| 1046 |
+
"eval_rewards/margins": 0.466796875,
|
| 1047 |
+
"eval_rewards/rejected": -0.6875,
|
| 1048 |
+
"eval_runtime": 65.3898,
|
| 1049 |
+
"eval_samples_per_second": 22.741,
|
| 1050 |
+
"eval_steps_per_second": 0.719,
|
| 1051 |
+
"step": 296
|
| 1052 |
+
},
|
| 1053 |
+
{
|
| 1054 |
+
"epoch": 0.42075736325385693,
|
| 1055 |
+
"grad_norm": 19.32972616150208,
|
| 1056 |
+
"learning_rate": 7.189897720653835e-08,
|
| 1057 |
+
"logits/chosen": -3.015625,
|
| 1058 |
+
"logits/rejected": -3.109375,
|
| 1059 |
+
"logps/chosen": -338.0,
|
| 1060 |
+
"logps/rejected": -322.0,
|
| 1061 |
+
"loss": 0.5486,
|
| 1062 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 1063 |
+
"rewards/chosen": -0.30859375,
|
| 1064 |
+
"rewards/margins": 0.455078125,
|
| 1065 |
+
"rewards/rejected": -0.765625,
|
| 1066 |
+
"step": 300
|
| 1067 |
+
},
|
| 1068 |
+
{
|
| 1069 |
+
"epoch": 0.426367461430575,
|
| 1070 |
+
"eval_logits/chosen": -3.109375,
|
| 1071 |
+
"eval_logits/rejected": -3.125,
|
| 1072 |
+
"eval_logps/chosen": -340.0,
|
| 1073 |
+
"eval_logps/rejected": -342.0,
|
| 1074 |
+
"eval_loss": 0.5460695028305054,
|
| 1075 |
+
"eval_rewards/accuracies": 0.7659574747085571,
|
| 1076 |
+
"eval_rewards/chosen": -0.1884765625,
|
| 1077 |
+
"eval_rewards/margins": 0.482421875,
|
| 1078 |
+
"eval_rewards/rejected": -0.671875,
|
| 1079 |
+
"eval_runtime": 65.6401,
|
| 1080 |
+
"eval_samples_per_second": 22.654,
|
| 1081 |
+
"eval_steps_per_second": 0.716,
|
| 1082 |
+
"step": 304
|
| 1083 |
+
},
|
| 1084 |
+
{
|
| 1085 |
+
"epoch": 0.43478260869565216,
|
| 1086 |
+
"grad_norm": 20.84038979708752,
|
| 1087 |
+
"learning_rate": 6.967056418974356e-08,
|
| 1088 |
+
"logits/chosen": -3.078125,
|
| 1089 |
+
"logits/rejected": -3.09375,
|
| 1090 |
+
"logps/chosen": -382.0,
|
| 1091 |
+
"logps/rejected": -366.0,
|
| 1092 |
+
"loss": 0.5263,
|
| 1093 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1094 |
+
"rewards/chosen": -0.27734375,
|
| 1095 |
+
"rewards/margins": 0.5234375,
|
| 1096 |
+
"rewards/rejected": -0.80078125,
|
| 1097 |
+
"step": 310
|
| 1098 |
+
},
|
| 1099 |
+
{
|
| 1100 |
+
"epoch": 0.4375876577840112,
|
| 1101 |
+
"eval_logits/chosen": -3.09375,
|
| 1102 |
+
"eval_logits/rejected": -3.125,
|
| 1103 |
+
"eval_logps/chosen": -344.0,
|
| 1104 |
+
"eval_logps/rejected": -346.0,
|
| 1105 |
+
"eval_loss": 0.5433906316757202,
|
| 1106 |
+
"eval_rewards/accuracies": 0.7659574747085571,
|
| 1107 |
+
"eval_rewards/chosen": -0.21484375,
|
| 1108 |
+
"eval_rewards/margins": 0.490234375,
|
| 1109 |
+
"eval_rewards/rejected": -0.703125,
|
| 1110 |
+
"eval_runtime": 65.6624,
|
| 1111 |
+
"eval_samples_per_second": 22.646,
|
| 1112 |
+
"eval_steps_per_second": 0.716,
|
| 1113 |
+
"step": 312
|
| 1114 |
+
},
|
| 1115 |
+
{
|
| 1116 |
+
"epoch": 0.4488078541374474,
|
| 1117 |
+
"grad_norm": 19.39801079549374,
|
| 1118 |
+
"learning_rate": 6.739491077279388e-08,
|
| 1119 |
+
"logits/chosen": -3.09375,
|
| 1120 |
+
"logits/rejected": -3.171875,
|
| 1121 |
+
"logps/chosen": -418.0,
|
| 1122 |
+
"logps/rejected": -354.0,
|
| 1123 |
+
"loss": 0.539,
|
| 1124 |
+
"rewards/accuracies": 0.737500011920929,
|
| 1125 |
+
"rewards/chosen": -0.26171875,
|
| 1126 |
+
"rewards/margins": 0.49609375,
|
| 1127 |
+
"rewards/rejected": -0.7578125,
|
| 1128 |
+
"step": 320
|
| 1129 |
+
},
|
| 1130 |
+
{
|
| 1131 |
+
"epoch": 0.4488078541374474,
|
| 1132 |
+
"eval_logits/chosen": -3.09375,
|
| 1133 |
+
"eval_logits/rejected": -3.125,
|
| 1134 |
+
"eval_logps/chosen": -346.0,
|
| 1135 |
+
"eval_logps/rejected": -350.0,
|
| 1136 |
+
"eval_loss": 0.5414044260978699,
|
| 1137 |
+
"eval_rewards/accuracies": 0.7553191781044006,
|
| 1138 |
+
"eval_rewards/chosen": -0.2490234375,
|
| 1139 |
+
"eval_rewards/margins": 0.48828125,
|
| 1140 |
+
"eval_rewards/rejected": -0.73828125,
|
| 1141 |
+
"eval_runtime": 65.8794,
|
| 1142 |
+
"eval_samples_per_second": 22.572,
|
| 1143 |
+
"eval_steps_per_second": 0.713,
|
| 1144 |
+
"step": 320
|
| 1145 |
+
},
|
| 1146 |
+
{
|
| 1147 |
+
"epoch": 0.4600280504908836,
|
| 1148 |
+
"eval_logits/chosen": -3.09375,
|
| 1149 |
+
"eval_logits/rejected": -3.125,
|
| 1150 |
+
"eval_logps/chosen": -336.0,
|
| 1151 |
+
"eval_logps/rejected": -342.0,
|
| 1152 |
+
"eval_loss": 0.5408417582511902,
|
| 1153 |
+
"eval_rewards/accuracies": 0.7872340679168701,
|
| 1154 |
+
"eval_rewards/chosen": -0.1474609375,
|
| 1155 |
+
"eval_rewards/margins": 0.51171875,
|
| 1156 |
+
"eval_rewards/rejected": -0.65625,
|
| 1157 |
+
"eval_runtime": 65.2872,
|
| 1158 |
+
"eval_samples_per_second": 22.776,
|
| 1159 |
+
"eval_steps_per_second": 0.72,
|
| 1160 |
+
"step": 328
|
| 1161 |
+
},
|
| 1162 |
+
{
|
| 1163 |
+
"epoch": 0.4628330995792426,
|
| 1164 |
+
"grad_norm": 25.683811989115487,
|
| 1165 |
+
"learning_rate": 6.507748211555935e-08,
|
| 1166 |
+
"logits/chosen": -3.0,
|
| 1167 |
+
"logits/rejected": -3.046875,
|
| 1168 |
+
"logps/chosen": -364.0,
|
| 1169 |
+
"logps/rejected": -356.0,
|
| 1170 |
+
"loss": 0.5491,
|
| 1171 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1172 |
+
"rewards/chosen": -0.220703125,
|
| 1173 |
+
"rewards/margins": 0.578125,
|
| 1174 |
+
"rewards/rejected": -0.796875,
|
| 1175 |
+
"step": 330
|
| 1176 |
+
},
|
| 1177 |
+
{
|
| 1178 |
+
"epoch": 0.47124824684431976,
|
| 1179 |
+
"eval_logits/chosen": -3.09375,
|
| 1180 |
+
"eval_logits/rejected": -3.125,
|
| 1181 |
+
"eval_logps/chosen": -344.0,
|
| 1182 |
+
"eval_logps/rejected": -348.0,
|
| 1183 |
+
"eval_loss": 0.5370554327964783,
|
| 1184 |
+
"eval_rewards/accuracies": 0.771276593208313,
|
| 1185 |
+
"eval_rewards/chosen": -0.21875,
|
| 1186 |
+
"eval_rewards/margins": 0.51953125,
|
| 1187 |
+
"eval_rewards/rejected": -0.73828125,
|
| 1188 |
+
"eval_runtime": 65.8828,
|
| 1189 |
+
"eval_samples_per_second": 22.57,
|
| 1190 |
+
"eval_steps_per_second": 0.713,
|
| 1191 |
+
"step": 336
|
| 1192 |
+
},
|
| 1193 |
+
{
|
| 1194 |
+
"epoch": 0.47685834502103785,
|
| 1195 |
+
"grad_norm": 25.38527516497017,
|
| 1196 |
+
"learning_rate": 6.272384370442064e-08,
|
| 1197 |
+
"logits/chosen": -2.953125,
|
| 1198 |
+
"logits/rejected": -2.96875,
|
| 1199 |
+
"logps/chosen": -332.0,
|
| 1200 |
+
"logps/rejected": -362.0,
|
| 1201 |
+
"loss": 0.525,
|
| 1202 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1203 |
+
"rewards/chosen": -0.31640625,
|
| 1204 |
+
"rewards/margins": 0.439453125,
|
| 1205 |
+
"rewards/rejected": -0.75390625,
|
| 1206 |
+
"step": 340
|
| 1207 |
+
},
|
| 1208 |
+
{
|
| 1209 |
+
"epoch": 0.48246844319775595,
|
| 1210 |
+
"eval_logits/chosen": -3.09375,
|
| 1211 |
+
"eval_logits/rejected": -3.109375,
|
| 1212 |
+
"eval_logps/chosen": -350.0,
|
| 1213 |
+
"eval_logps/rejected": -354.0,
|
| 1214 |
+
"eval_loss": 0.5372266173362732,
|
| 1215 |
+
"eval_rewards/accuracies": 0.7606382966041565,
|
| 1216 |
+
"eval_rewards/chosen": -0.28125,
|
| 1217 |
+
"eval_rewards/margins": 0.51171875,
|
| 1218 |
+
"eval_rewards/rejected": -0.79296875,
|
| 1219 |
+
"eval_runtime": 65.5252,
|
| 1220 |
+
"eval_samples_per_second": 22.694,
|
| 1221 |
+
"eval_steps_per_second": 0.717,
|
| 1222 |
+
"step": 344
|
| 1223 |
+
},
|
| 1224 |
+
{
|
| 1225 |
+
"epoch": 0.4908835904628331,
|
| 1226 |
+
"grad_norm": 20.41249212586205,
|
| 1227 |
+
"learning_rate": 6.033964798631775e-08,
|
| 1228 |
+
"logits/chosen": -3.0,
|
| 1229 |
+
"logits/rejected": -3.046875,
|
| 1230 |
+
"logps/chosen": -348.0,
|
| 1231 |
+
"logps/rejected": -356.0,
|
| 1232 |
+
"loss": 0.5509,
|
| 1233 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1234 |
+
"rewards/chosen": -0.4375,
|
| 1235 |
+
"rewards/margins": 0.3828125,
|
| 1236 |
+
"rewards/rejected": -0.8203125,
|
| 1237 |
+
"step": 350
|
| 1238 |
+
},
|
| 1239 |
+
{
|
| 1240 |
+
"epoch": 0.49368863955119213,
|
| 1241 |
+
"eval_logits/chosen": -3.09375,
|
| 1242 |
+
"eval_logits/rejected": -3.125,
|
| 1243 |
+
"eval_logps/chosen": -338.0,
|
| 1244 |
+
"eval_logps/rejected": -344.0,
|
| 1245 |
+
"eval_loss": 0.53655606508255,
|
| 1246 |
+
"eval_rewards/accuracies": 0.7978723645210266,
|
| 1247 |
+
"eval_rewards/chosen": -0.1572265625,
|
| 1248 |
+
"eval_rewards/margins": 0.53515625,
|
| 1249 |
+
"eval_rewards/rejected": -0.6953125,
|
| 1250 |
+
"eval_runtime": 65.8155,
|
| 1251 |
+
"eval_samples_per_second": 22.593,
|
| 1252 |
+
"eval_steps_per_second": 0.714,
|
| 1253 |
+
"step": 352
|
| 1254 |
+
},
|
| 1255 |
+
{
|
| 1256 |
+
"epoch": 0.5049088359046283,
|
| 1257 |
+
"grad_norm": 26.68036848508083,
|
| 1258 |
+
"learning_rate": 5.793062079395602e-08,
|
| 1259 |
+
"logits/chosen": -3.078125,
|
| 1260 |
+
"logits/rejected": -3.140625,
|
| 1261 |
+
"logps/chosen": -384.0,
|
| 1262 |
+
"logps/rejected": -362.0,
|
| 1263 |
+
"loss": 0.5342,
|
| 1264 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1265 |
+
"rewards/chosen": -0.169921875,
|
| 1266 |
+
"rewards/margins": 0.5078125,
|
| 1267 |
+
"rewards/rejected": -0.67578125,
|
| 1268 |
+
"step": 360
|
| 1269 |
+
},
|
| 1270 |
+
{
|
| 1271 |
+
"epoch": 0.5049088359046283,
|
| 1272 |
+
"eval_logits/chosen": -3.09375,
|
| 1273 |
+
"eval_logits/rejected": -3.125,
|
| 1274 |
+
"eval_logps/chosen": -338.0,
|
| 1275 |
+
"eval_logps/rejected": -346.0,
|
| 1276 |
+
"eval_loss": 0.5345030426979065,
|
| 1277 |
+
"eval_rewards/accuracies": 0.7978723645210266,
|
| 1278 |
+
"eval_rewards/chosen": -0.162109375,
|
| 1279 |
+
"eval_rewards/margins": 0.546875,
|
| 1280 |
+
"eval_rewards/rejected": -0.70703125,
|
| 1281 |
+
"eval_runtime": 65.1382,
|
| 1282 |
+
"eval_samples_per_second": 22.828,
|
| 1283 |
+
"eval_steps_per_second": 0.722,
|
| 1284 |
+
"step": 360
|
| 1285 |
+
},
|
| 1286 |
+
{
|
| 1287 |
+
"epoch": 0.5161290322580645,
|
| 1288 |
+
"eval_logits/chosen": -3.09375,
|
| 1289 |
+
"eval_logits/rejected": -3.109375,
|
| 1290 |
+
"eval_logps/chosen": -350.0,
|
| 1291 |
+
"eval_logps/rejected": -358.0,
|
| 1292 |
+
"eval_loss": 0.5322470664978027,
|
| 1293 |
+
"eval_rewards/accuracies": 0.7765957713127136,
|
| 1294 |
+
"eval_rewards/chosen": -0.2890625,
|
| 1295 |
+
"eval_rewards/margins": 0.53515625,
|
| 1296 |
+
"eval_rewards/rejected": -0.82421875,
|
| 1297 |
+
"eval_runtime": 65.4517,
|
| 1298 |
+
"eval_samples_per_second": 22.719,
|
| 1299 |
+
"eval_steps_per_second": 0.718,
|
| 1300 |
+
"step": 368
|
| 1301 |
+
},
|
| 1302 |
+
{
|
| 1303 |
+
"epoch": 0.5189340813464236,
|
| 1304 |
+
"grad_norm": 35.318750139832915,
|
| 1305 |
+
"learning_rate": 5.550254759477064e-08,
|
| 1306 |
+
"logits/chosen": -3.0625,
|
| 1307 |
+
"logits/rejected": -3.0625,
|
| 1308 |
+
"logps/chosen": -342.0,
|
| 1309 |
+
"logps/rejected": -386.0,
|
| 1310 |
+
"loss": 0.5392,
|
| 1311 |
+
"rewards/accuracies": 0.737500011920929,
|
| 1312 |
+
"rewards/chosen": -0.328125,
|
| 1313 |
+
"rewards/margins": 0.4609375,
|
| 1314 |
+
"rewards/rejected": -0.7890625,
|
| 1315 |
+
"step": 370
|
| 1316 |
+
},
|
| 1317 |
+
{
|
| 1318 |
+
"epoch": 0.5273492286115007,
|
| 1319 |
+
"eval_logits/chosen": -3.09375,
|
| 1320 |
+
"eval_logits/rejected": -3.109375,
|
| 1321 |
+
"eval_logps/chosen": -342.0,
|
| 1322 |
+
"eval_logps/rejected": -352.0,
|
| 1323 |
+
"eval_loss": 0.5290676951408386,
|
| 1324 |
+
"eval_rewards/accuracies": 0.7819148898124695,
|
| 1325 |
+
"eval_rewards/chosen": -0.2109375,
|
| 1326 |
+
"eval_rewards/margins": 0.5625,
|
| 1327 |
+
"eval_rewards/rejected": -0.7734375,
|
| 1328 |
+
"eval_runtime": 66.0225,
|
| 1329 |
+
"eval_samples_per_second": 22.523,
|
| 1330 |
+
"eval_steps_per_second": 0.712,
|
| 1331 |
+
"step": 376
|
| 1332 |
+
},
|
| 1333 |
+
{
|
| 1334 |
+
"epoch": 0.5329593267882188,
|
| 1335 |
+
"grad_norm": 25.630480455788526,
|
| 1336 |
+
"learning_rate": 5.3061259596673514e-08,
|
| 1337 |
+
"logits/chosen": -2.875,
|
| 1338 |
+
"logits/rejected": -2.984375,
|
| 1339 |
+
"logps/chosen": -334.0,
|
| 1340 |
+
"logps/rejected": -384.0,
|
| 1341 |
+
"loss": 0.5275,
|
| 1342 |
+
"rewards/accuracies": 0.824999988079071,
|
| 1343 |
+
"rewards/chosen": -0.32421875,
|
| 1344 |
+
"rewards/margins": 0.52734375,
|
| 1345 |
+
"rewards/rejected": -0.8515625,
|
| 1346 |
+
"step": 380
|
| 1347 |
+
},
|
| 1348 |
+
{
|
| 1349 |
+
"epoch": 0.5385694249649369,
|
| 1350 |
+
"eval_logits/chosen": -3.09375,
|
| 1351 |
+
"eval_logits/rejected": -3.125,
|
| 1352 |
+
"eval_logps/chosen": -342.0,
|
| 1353 |
+
"eval_logps/rejected": -350.0,
|
| 1354 |
+
"eval_loss": 0.5304205417633057,
|
| 1355 |
+
"eval_rewards/accuracies": 0.7872340679168701,
|
| 1356 |
+
"eval_rewards/chosen": -0.197265625,
|
| 1357 |
+
"eval_rewards/margins": 0.5546875,
|
| 1358 |
+
"eval_rewards/rejected": -0.75390625,
|
| 1359 |
+
"eval_runtime": 66.8954,
|
| 1360 |
+
"eval_samples_per_second": 22.229,
|
| 1361 |
+
"eval_steps_per_second": 0.703,
|
| 1362 |
+
"step": 384
|
| 1363 |
+
},
|
| 1364 |
+
{
|
| 1365 |
+
"epoch": 0.5469845722300141,
|
| 1366 |
+
"grad_norm": 19.177955351319696,
|
| 1367 |
+
"learning_rate": 5.061261974395087e-08,
|
| 1368 |
+
"logits/chosen": -3.0625,
|
| 1369 |
+
"logits/rejected": -3.078125,
|
| 1370 |
+
"logps/chosen": -340.0,
|
| 1371 |
+
"logps/rejected": -370.0,
|
| 1372 |
+
"loss": 0.5228,
|
| 1373 |
+
"rewards/accuracies": 0.824999988079071,
|
| 1374 |
+
"rewards/chosen": -0.21875,
|
| 1375 |
+
"rewards/margins": 0.578125,
|
| 1376 |
+
"rewards/rejected": -0.80078125,
|
| 1377 |
+
"step": 390
|
| 1378 |
+
},
|
| 1379 |
+
{
|
| 1380 |
+
"epoch": 0.5497896213183731,
|
| 1381 |
+
"eval_logits/chosen": -3.09375,
|
| 1382 |
+
"eval_logits/rejected": -3.125,
|
| 1383 |
+
"eval_logps/chosen": -346.0,
|
| 1384 |
+
"eval_logps/rejected": -356.0,
|
| 1385 |
+
"eval_loss": 0.5286569595336914,
|
| 1386 |
+
"eval_rewards/accuracies": 0.771276593208313,
|
| 1387 |
+
"eval_rewards/chosen": -0.2490234375,
|
| 1388 |
+
"eval_rewards/margins": 0.55859375,
|
| 1389 |
+
"eval_rewards/rejected": -0.80859375,
|
| 1390 |
+
"eval_runtime": 66.2711,
|
| 1391 |
+
"eval_samples_per_second": 22.438,
|
| 1392 |
+
"eval_steps_per_second": 0.709,
|
| 1393 |
+
"step": 392
|
| 1394 |
+
},
|
| 1395 |
+
{
|
| 1396 |
+
"epoch": 0.5610098176718092,
|
| 1397 |
+
"grad_norm": 22.280250814675583,
|
| 1398 |
+
"learning_rate": 4.81625086369435e-08,
|
| 1399 |
+
"logits/chosen": -3.0625,
|
| 1400 |
+
"logits/rejected": -3.125,
|
| 1401 |
+
"logps/chosen": -318.0,
|
| 1402 |
+
"logps/rejected": -350.0,
|
| 1403 |
+
"loss": 0.5174,
|
| 1404 |
+
"rewards/accuracies": 0.762499988079071,
|
| 1405 |
+
"rewards/chosen": -0.3125,
|
| 1406 |
+
"rewards/margins": 0.53125,
|
| 1407 |
+
"rewards/rejected": -0.84375,
|
| 1408 |
+
"step": 400
|
| 1409 |
+
},
|
| 1410 |
+
{
|
| 1411 |
+
"epoch": 0.5610098176718092,
|
| 1412 |
+
"eval_logits/chosen": -3.09375,
|
| 1413 |
+
"eval_logits/rejected": -3.125,
|
| 1414 |
+
"eval_logps/chosen": -346.0,
|
| 1415 |
+
"eval_logps/rejected": -358.0,
|
| 1416 |
+
"eval_loss": 0.5268398523330688,
|
| 1417 |
+
"eval_rewards/accuracies": 0.771276593208313,
|
| 1418 |
+
"eval_rewards/chosen": -0.251953125,
|
| 1419 |
+
"eval_rewards/margins": 0.56640625,
|
| 1420 |
+
"eval_rewards/rejected": -0.8203125,
|
| 1421 |
+
"eval_runtime": 65.9859,
|
| 1422 |
+
"eval_samples_per_second": 22.535,
|
| 1423 |
+
"eval_steps_per_second": 0.712,
|
| 1424 |
+
"step": 400
|
| 1425 |
+
},
|
| 1426 |
+
{
|
| 1427 |
+
"epoch": 0.5722300140252454,
|
| 1428 |
+
"eval_logits/chosen": -3.09375,
|
| 1429 |
+
"eval_logits/rejected": -3.125,
|
| 1430 |
+
"eval_logps/chosen": -344.0,
|
| 1431 |
+
"eval_logps/rejected": -356.0,
|
| 1432 |
+
"eval_loss": 0.5255098938941956,
|
| 1433 |
+
"eval_rewards/accuracies": 0.771276593208313,
|
| 1434 |
+
"eval_rewards/chosen": -0.2265625,
|
| 1435 |
+
"eval_rewards/margins": 0.578125,
|
| 1436 |
+
"eval_rewards/rejected": -0.8046875,
|
| 1437 |
+
"eval_runtime": 65.6155,
|
| 1438 |
+
"eval_samples_per_second": 22.662,
|
| 1439 |
+
"eval_steps_per_second": 0.716,
|
| 1440 |
+
"step": 408
|
| 1441 |
+
},
|
| 1442 |
+
{
|
| 1443 |
+
"epoch": 0.5750350631136045,
|
| 1444 |
+
"grad_norm": 18.983834453019576,
|
| 1445 |
+
"learning_rate": 4.571681040932457e-08,
|
| 1446 |
+
"logits/chosen": -3.046875,
|
| 1447 |
+
"logits/rejected": -3.09375,
|
| 1448 |
+
"logps/chosen": -350.0,
|
| 1449 |
+
"logps/rejected": -346.0,
|
| 1450 |
+
"loss": 0.5184,
|
| 1451 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1452 |
+
"rewards/chosen": -0.27734375,
|
| 1453 |
+
"rewards/margins": 0.59375,
|
| 1454 |
+
"rewards/rejected": -0.8671875,
|
| 1455 |
+
"step": 410
|
| 1456 |
+
},
|
| 1457 |
+
{
|
| 1458 |
+
"epoch": 0.5834502103786816,
|
| 1459 |
+
"eval_logits/chosen": -3.09375,
|
| 1460 |
+
"eval_logits/rejected": -3.109375,
|
| 1461 |
+
"eval_logps/chosen": -350.0,
|
| 1462 |
+
"eval_logps/rejected": -362.0,
|
| 1463 |
+
"eval_loss": 0.524753212928772,
|
| 1464 |
+
"eval_rewards/accuracies": 0.7659574747085571,
|
| 1465 |
+
"eval_rewards/chosen": -0.27734375,
|
| 1466 |
+
"eval_rewards/margins": 0.5859375,
|
| 1467 |
+
"eval_rewards/rejected": -0.859375,
|
| 1468 |
+
"eval_runtime": 65.6143,
|
| 1469 |
+
"eval_samples_per_second": 22.663,
|
| 1470 |
+
"eval_steps_per_second": 0.716,
|
| 1471 |
+
"step": 416
|
| 1472 |
+
},
|
| 1473 |
+
{
|
| 1474 |
+
"epoch": 0.5890603085553997,
|
| 1475 |
+
"grad_norm": 25.592693204204966,
|
| 1476 |
+
"learning_rate": 4.3281398596891846e-08,
|
| 1477 |
+
"logits/chosen": -2.96875,
|
| 1478 |
+
"logits/rejected": -3.046875,
|
| 1479 |
+
"logps/chosen": -354.0,
|
| 1480 |
+
"logps/rejected": -374.0,
|
| 1481 |
+
"loss": 0.5095,
|
| 1482 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 1483 |
+
"rewards/chosen": -0.236328125,
|
| 1484 |
+
"rewards/margins": 0.7421875,
|
| 1485 |
+
"rewards/rejected": -0.9765625,
|
| 1486 |
+
"step": 420
|
| 1487 |
+
},
|
| 1488 |
+
{
|
| 1489 |
+
"epoch": 0.5946704067321178,
|
| 1490 |
+
"eval_logits/chosen": -3.09375,
|
| 1491 |
+
"eval_logits/rejected": -3.109375,
|
| 1492 |
+
"eval_logps/chosen": -348.0,
|
| 1493 |
+
"eval_logps/rejected": -362.0,
|
| 1494 |
+
"eval_loss": 0.523405909538269,
|
| 1495 |
+
"eval_rewards/accuracies": 0.7819148898124695,
|
| 1496 |
+
"eval_rewards/chosen": -0.271484375,
|
| 1497 |
+
"eval_rewards/margins": 0.59765625,
|
| 1498 |
+
"eval_rewards/rejected": -0.8671875,
|
| 1499 |
+
"eval_runtime": 65.4312,
|
| 1500 |
+
"eval_samples_per_second": 22.726,
|
| 1501 |
+
"eval_steps_per_second": 0.718,
|
| 1502 |
+
"step": 424
|
| 1503 |
+
},
|
| 1504 |
+
{
|
| 1505 |
+
"epoch": 0.603085553997195,
|
| 1506 |
+
"grad_norm": 20.945961565616308,
|
| 1507 |
+
"learning_rate": 4.0862122031811584e-08,
|
| 1508 |
+
"logits/chosen": -3.09375,
|
| 1509 |
+
"logits/rejected": -2.96875,
|
| 1510 |
+
"logps/chosen": -398.0,
|
| 1511 |
+
"logps/rejected": -406.0,
|
| 1512 |
+
"loss": 0.5259,
|
| 1513 |
+
"rewards/accuracies": 0.875,
|
| 1514 |
+
"rewards/chosen": -0.2314453125,
|
| 1515 |
+
"rewards/margins": 0.66015625,
|
| 1516 |
+
"rewards/rejected": -0.890625,
|
| 1517 |
+
"step": 430
|
| 1518 |
+
},
|
| 1519 |
+
{
|
| 1520 |
+
"epoch": 0.605890603085554,
|
| 1521 |
+
"eval_logits/chosen": -3.09375,
|
| 1522 |
+
"eval_logits/rejected": -3.109375,
|
| 1523 |
+
"eval_logps/chosen": -346.0,
|
| 1524 |
+
"eval_logps/rejected": -360.0,
|
| 1525 |
+
"eval_loss": 0.5217041373252869,
|
| 1526 |
+
"eval_rewards/accuracies": 0.7819148898124695,
|
| 1527 |
+
"eval_rewards/chosen": -0.2412109375,
|
| 1528 |
+
"eval_rewards/margins": 0.60546875,
|
| 1529 |
+
"eval_rewards/rejected": -0.84765625,
|
| 1530 |
+
"eval_runtime": 66.0803,
|
| 1531 |
+
"eval_samples_per_second": 22.503,
|
| 1532 |
+
"eval_steps_per_second": 0.711,
|
| 1533 |
+
"step": 432
|
| 1534 |
+
},
|
| 1535 |
+
{
|
| 1536 |
+
"epoch": 0.6171107994389902,
|
| 1537 |
+
"grad_norm": 32.43796306320174,
|
| 1538 |
+
"learning_rate": 3.84647907961901e-08,
|
| 1539 |
+
"logits/chosen": -3.0625,
|
| 1540 |
+
"logits/rejected": -2.96875,
|
| 1541 |
+
"logps/chosen": -348.0,
|
| 1542 |
+
"logps/rejected": -366.0,
|
| 1543 |
+
"loss": 0.5131,
|
| 1544 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1545 |
+
"rewards/chosen": -0.279296875,
|
| 1546 |
+
"rewards/margins": 0.67578125,
|
| 1547 |
+
"rewards/rejected": -0.953125,
|
| 1548 |
+
"step": 440
|
| 1549 |
+
},
|
| 1550 |
+
{
|
| 1551 |
+
"epoch": 0.6171107994389902,
|
| 1552 |
+
"eval_logits/chosen": -3.09375,
|
| 1553 |
+
"eval_logits/rejected": -3.109375,
|
| 1554 |
+
"eval_logps/chosen": -346.0,
|
| 1555 |
+
"eval_logps/rejected": -360.0,
|
| 1556 |
+
"eval_loss": 0.520456075668335,
|
| 1557 |
+
"eval_rewards/accuracies": 0.792553186416626,
|
| 1558 |
+
"eval_rewards/chosen": -0.23828125,
|
| 1559 |
+
"eval_rewards/margins": 0.61328125,
|
| 1560 |
+
"eval_rewards/rejected": -0.85546875,
|
| 1561 |
+
"eval_runtime": 65.5788,
|
| 1562 |
+
"eval_samples_per_second": 22.675,
|
| 1563 |
+
"eval_steps_per_second": 0.717,
|
| 1564 |
+
"step": 440
|
| 1565 |
+
},
|
| 1566 |
+
{
|
| 1567 |
+
"epoch": 0.6283309957924264,
|
| 1568 |
+
"eval_logits/chosen": -3.078125,
|
| 1569 |
+
"eval_logits/rejected": -3.109375,
|
| 1570 |
+
"eval_logps/chosen": -344.0,
|
| 1571 |
+
"eval_logps/rejected": -360.0,
|
| 1572 |
+
"eval_loss": 0.5200462937355042,
|
| 1573 |
+
"eval_rewards/accuracies": 0.7978723645210266,
|
| 1574 |
+
"eval_rewards/chosen": -0.2236328125,
|
| 1575 |
+
"eval_rewards/margins": 0.62109375,
|
| 1576 |
+
"eval_rewards/rejected": -0.84765625,
|
| 1577 |
+
"eval_runtime": 65.4678,
|
| 1578 |
+
"eval_samples_per_second": 22.713,
|
| 1579 |
+
"eval_steps_per_second": 0.718,
|
| 1580 |
+
"step": 448
|
| 1581 |
+
},
|
| 1582 |
+
{
|
| 1583 |
+
"epoch": 0.6311360448807855,
|
| 1584 |
+
"grad_norm": 25.78934426006984,
|
| 1585 |
+
"learning_rate": 3.609516226870659e-08,
|
| 1586 |
+
"logits/chosen": -3.015625,
|
| 1587 |
+
"logits/rejected": -3.09375,
|
| 1588 |
+
"logps/chosen": -378.0,
|
| 1589 |
+
"logps/rejected": -394.0,
|
| 1590 |
+
"loss": 0.536,
|
| 1591 |
+
"rewards/accuracies": 0.8125,
|
| 1592 |
+
"rewards/chosen": -0.349609375,
|
| 1593 |
+
"rewards/margins": 0.60546875,
|
| 1594 |
+
"rewards/rejected": -0.953125,
|
| 1595 |
+
"step": 450
|
| 1596 |
+
},
|
| 1597 |
+
{
|
| 1598 |
+
"epoch": 0.6395511921458625,
|
| 1599 |
+
"eval_logits/chosen": -3.078125,
|
| 1600 |
+
"eval_logits/rejected": -3.109375,
|
| 1601 |
+
"eval_logps/chosen": -344.0,
|
| 1602 |
+
"eval_logps/rejected": -360.0,
|
| 1603 |
+
"eval_loss": 0.5196602940559387,
|
| 1604 |
+
"eval_rewards/accuracies": 0.7978723645210266,
|
| 1605 |
+
"eval_rewards/chosen": -0.22265625,
|
| 1606 |
+
"eval_rewards/margins": 0.625,
|
| 1607 |
+
"eval_rewards/rejected": -0.84765625,
|
| 1608 |
+
"eval_runtime": 66.1158,
|
| 1609 |
+
"eval_samples_per_second": 22.491,
|
| 1610 |
+
"eval_steps_per_second": 0.711,
|
| 1611 |
+
"step": 456
|
| 1612 |
+
},
|
| 1613 |
+
{
|
| 1614 |
+
"epoch": 0.6451612903225806,
|
| 1615 |
+
"grad_norm": 27.561675324577404,
|
| 1616 |
+
"learning_rate": 3.375892729781754e-08,
|
| 1617 |
+
"logits/chosen": -2.953125,
|
| 1618 |
+
"logits/rejected": -3.046875,
|
| 1619 |
+
"logps/chosen": -380.0,
|
| 1620 |
+
"logps/rejected": -392.0,
|
| 1621 |
+
"loss": 0.52,
|
| 1622 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1623 |
+
"rewards/chosen": -0.2890625,
|
| 1624 |
+
"rewards/margins": 0.62890625,
|
| 1625 |
+
"rewards/rejected": -0.91796875,
|
| 1626 |
+
"step": 460
|
| 1627 |
+
},
|
| 1628 |
+
{
|
| 1629 |
+
"epoch": 0.6507713884992987,
|
| 1630 |
+
"eval_logits/chosen": -3.078125,
|
| 1631 |
+
"eval_logits/rejected": -3.109375,
|
| 1632 |
+
"eval_logps/chosen": -350.0,
|
| 1633 |
+
"eval_logps/rejected": -366.0,
|
| 1634 |
+
"eval_loss": 0.5189826488494873,
|
| 1635 |
+
"eval_rewards/accuracies": 0.7978723645210266,
|
| 1636 |
+
"eval_rewards/chosen": -0.287109375,
|
| 1637 |
+
"eval_rewards/margins": 0.62109375,
|
| 1638 |
+
"eval_rewards/rejected": -0.90625,
|
| 1639 |
+
"eval_runtime": 66.7501,
|
| 1640 |
+
"eval_samples_per_second": 22.277,
|
| 1641 |
+
"eval_steps_per_second": 0.704,
|
| 1642 |
+
"step": 464
|
| 1643 |
+
},
|
| 1644 |
+
{
|
| 1645 |
+
"epoch": 0.6591865357643759,
|
| 1646 |
+
"grad_norm": 28.5091931919782,
|
| 1647 |
+
"learning_rate": 3.146169653473842e-08,
|
| 1648 |
+
"logits/chosen": -2.953125,
|
| 1649 |
+
"logits/rejected": -3.0,
|
| 1650 |
+
"logps/chosen": -392.0,
|
| 1651 |
+
"logps/rejected": -370.0,
|
| 1652 |
+
"loss": 0.5136,
|
| 1653 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 1654 |
+
"rewards/chosen": -0.2265625,
|
| 1655 |
+
"rewards/margins": 0.69921875,
|
| 1656 |
+
"rewards/rejected": -0.92578125,
|
| 1657 |
+
"step": 470
|
| 1658 |
+
},
|
| 1659 |
+
{
|
| 1660 |
+
"epoch": 0.6619915848527349,
|
| 1661 |
+
"eval_logits/chosen": -3.078125,
|
| 1662 |
+
"eval_logits/rejected": -3.109375,
|
| 1663 |
+
"eval_logps/chosen": -356.0,
|
| 1664 |
+
"eval_logps/rejected": -372.0,
|
| 1665 |
+
"eval_loss": 0.5202246904373169,
|
| 1666 |
+
"eval_rewards/accuracies": 0.7819148898124695,
|
| 1667 |
+
"eval_rewards/chosen": -0.349609375,
|
| 1668 |
+
"eval_rewards/margins": 0.61328125,
|
| 1669 |
+
"eval_rewards/rejected": -0.96484375,
|
| 1670 |
+
"eval_runtime": 65.2435,
|
| 1671 |
+
"eval_samples_per_second": 22.792,
|
| 1672 |
+
"eval_steps_per_second": 0.72,
|
| 1673 |
+
"step": 472
|
| 1674 |
+
},
|
| 1675 |
+
{
|
| 1676 |
+
"epoch": 0.6732117812061711,
|
| 1677 |
+
"grad_norm": 28.307915984199855,
|
| 1678 |
+
"learning_rate": 2.920898695902556e-08,
|
| 1679 |
+
"logits/chosen": -2.96875,
|
| 1680 |
+
"logits/rejected": -3.015625,
|
| 1681 |
+
"logps/chosen": -388.0,
|
| 1682 |
+
"logps/rejected": -386.0,
|
| 1683 |
+
"loss": 0.5402,
|
| 1684 |
+
"rewards/accuracies": 0.762499988079071,
|
| 1685 |
+
"rewards/chosen": -0.46875,
|
| 1686 |
+
"rewards/margins": 0.5703125,
|
| 1687 |
+
"rewards/rejected": -1.0390625,
|
| 1688 |
+
"step": 480
|
| 1689 |
+
},
|
| 1690 |
+
{
|
| 1691 |
+
"epoch": 0.6732117812061711,
|
| 1692 |
+
"eval_logits/chosen": -3.078125,
|
| 1693 |
+
"eval_logits/rejected": -3.109375,
|
| 1694 |
+
"eval_logps/chosen": -358.0,
|
| 1695 |
+
"eval_logps/rejected": -372.0,
|
| 1696 |
+
"eval_loss": 0.5193986296653748,
|
| 1697 |
+
"eval_rewards/accuracies": 0.7765957713127136,
|
| 1698 |
+
"eval_rewards/chosen": -0.357421875,
|
| 1699 |
+
"eval_rewards/margins": 0.6171875,
|
| 1700 |
+
"eval_rewards/rejected": -0.9765625,
|
| 1701 |
+
"eval_runtime": 65.2533,
|
| 1702 |
+
"eval_samples_per_second": 22.788,
|
| 1703 |
+
"eval_steps_per_second": 0.72,
|
| 1704 |
+
"step": 480
|
| 1705 |
+
},
|
| 1706 |
+
{
|
| 1707 |
+
"epoch": 0.6844319775596073,
|
| 1708 |
+
"eval_logits/chosen": -3.078125,
|
| 1709 |
+
"eval_logits/rejected": -3.109375,
|
| 1710 |
+
"eval_logps/chosen": -350.0,
|
| 1711 |
+
"eval_logps/rejected": -366.0,
|
| 1712 |
+
"eval_loss": 0.5179064869880676,
|
| 1713 |
+
"eval_rewards/accuracies": 0.7978723645210266,
|
| 1714 |
+
"eval_rewards/chosen": -0.287109375,
|
| 1715 |
+
"eval_rewards/margins": 0.625,
|
| 1716 |
+
"eval_rewards/rejected": -0.9140625,
|
| 1717 |
+
"eval_runtime": 65.4006,
|
| 1718 |
+
"eval_samples_per_second": 22.737,
|
| 1719 |
+
"eval_steps_per_second": 0.719,
|
| 1720 |
+
"step": 488
|
| 1721 |
+
},
|
| 1722 |
+
{
|
| 1723 |
+
"epoch": 0.6872370266479664,
|
| 1724 |
+
"grad_norm": 34.377948997484125,
|
| 1725 |
+
"learning_rate": 2.7006208629117678e-08,
|
| 1726 |
+
"logits/chosen": -3.046875,
|
| 1727 |
+
"logits/rejected": -3.125,
|
| 1728 |
+
"logps/chosen": -418.0,
|
| 1729 |
+
"logps/rejected": -362.0,
|
| 1730 |
+
"loss": 0.5,
|
| 1731 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1732 |
+
"rewards/chosen": -0.341796875,
|
| 1733 |
+
"rewards/margins": 0.66796875,
|
| 1734 |
+
"rewards/rejected": -1.0078125,
|
| 1735 |
+
"step": 490
|
| 1736 |
+
},
|
| 1737 |
+
{
|
| 1738 |
+
"epoch": 0.6956521739130435,
|
| 1739 |
+
"eval_logits/chosen": -3.078125,
|
| 1740 |
+
"eval_logits/rejected": -3.109375,
|
| 1741 |
+
"eval_logps/chosen": -348.0,
|
| 1742 |
+
"eval_logps/rejected": -366.0,
|
| 1743 |
+
"eval_loss": 0.5165696144104004,
|
| 1744 |
+
"eval_rewards/accuracies": 0.792553186416626,
|
| 1745 |
+
"eval_rewards/chosen": -0.265625,
|
| 1746 |
+
"eval_rewards/margins": 0.6328125,
|
| 1747 |
+
"eval_rewards/rejected": -0.89453125,
|
| 1748 |
+
"eval_runtime": 66.0312,
|
| 1749 |
+
"eval_samples_per_second": 22.52,
|
| 1750 |
+
"eval_steps_per_second": 0.712,
|
| 1751 |
+
"step": 496
|
| 1752 |
+
},
|
| 1753 |
+
{
|
| 1754 |
+
"epoch": 0.7012622720897616,
|
| 1755 |
+
"grad_norm": 24.33249271202181,
|
| 1756 |
+
"learning_rate": 2.485865168965695e-08,
|
| 1757 |
+
"logits/chosen": -2.96875,
|
| 1758 |
+
"logits/rejected": -3.09375,
|
| 1759 |
+
"logps/chosen": -386.0,
|
| 1760 |
+
"logps/rejected": -404.0,
|
| 1761 |
+
"loss": 0.5087,
|
| 1762 |
+
"rewards/accuracies": 0.824999988079071,
|
| 1763 |
+
"rewards/chosen": -0.25390625,
|
| 1764 |
+
"rewards/margins": 0.66796875,
|
| 1765 |
+
"rewards/rejected": -0.921875,
|
| 1766 |
+
"step": 500
|
| 1767 |
+
},
|
| 1768 |
+
{
|
| 1769 |
+
"epoch": 0.7068723702664796,
|
| 1770 |
+
"eval_logits/chosen": -3.078125,
|
| 1771 |
+
"eval_logits/rejected": -3.09375,
|
| 1772 |
+
"eval_logps/chosen": -348.0,
|
| 1773 |
+
"eval_logps/rejected": -366.0,
|
| 1774 |
+
"eval_loss": 0.5151299834251404,
|
| 1775 |
+
"eval_rewards/accuracies": 0.8031914830207825,
|
| 1776 |
+
"eval_rewards/chosen": -0.26171875,
|
| 1777 |
+
"eval_rewards/margins": 0.640625,
|
| 1778 |
+
"eval_rewards/rejected": -0.90234375,
|
| 1779 |
+
"eval_runtime": 65.7259,
|
| 1780 |
+
"eval_samples_per_second": 22.624,
|
| 1781 |
+
"eval_steps_per_second": 0.715,
|
| 1782 |
+
"step": 504
|
| 1783 |
+
},
|
| 1784 |
+
{
|
| 1785 |
+
"epoch": 0.7152875175315568,
|
| 1786 |
+
"grad_norm": 17.54957543271777,
|
| 1787 |
+
"learning_rate": 2.2771473666792496e-08,
|
| 1788 |
+
"logits/chosen": -2.90625,
|
| 1789 |
+
"logits/rejected": -3.03125,
|
| 1790 |
+
"logps/chosen": -380.0,
|
| 1791 |
+
"logps/rejected": -384.0,
|
| 1792 |
+
"loss": 0.5192,
|
| 1793 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1794 |
+
"rewards/chosen": -0.23046875,
|
| 1795 |
+
"rewards/margins": 0.7109375,
|
| 1796 |
+
"rewards/rejected": -0.94140625,
|
| 1797 |
+
"step": 510
|
| 1798 |
+
},
|
| 1799 |
+
{
|
| 1800 |
+
"epoch": 0.7180925666199158,
|
| 1801 |
+
"eval_logits/chosen": -3.078125,
|
| 1802 |
+
"eval_logits/rejected": -3.09375,
|
| 1803 |
+
"eval_logps/chosen": -352.0,
|
| 1804 |
+
"eval_logps/rejected": -370.0,
|
| 1805 |
+
"eval_loss": 0.5147544145584106,
|
| 1806 |
+
"eval_rewards/accuracies": 0.8031914830207825,
|
| 1807 |
+
"eval_rewards/chosen": -0.298828125,
|
| 1808 |
+
"eval_rewards/margins": 0.63671875,
|
| 1809 |
+
"eval_rewards/rejected": -0.9375,
|
| 1810 |
+
"eval_runtime": 65.4306,
|
| 1811 |
+
"eval_samples_per_second": 22.726,
|
| 1812 |
+
"eval_steps_per_second": 0.718,
|
| 1813 |
+
"step": 512
|
| 1814 |
+
},
|
| 1815 |
+
{
|
| 1816 |
+
"epoch": 0.729312762973352,
|
| 1817 |
+
"grad_norm": 35.64036533963828,
|
| 1818 |
+
"learning_rate": 2.0749687081977334e-08,
|
| 1819 |
+
"logits/chosen": -3.03125,
|
| 1820 |
+
"logits/rejected": -3.0,
|
| 1821 |
+
"logps/chosen": -382.0,
|
| 1822 |
+
"logps/rejected": -378.0,
|
| 1823 |
+
"loss": 0.5057,
|
| 1824 |
+
"rewards/accuracies": 0.762499988079071,
|
| 1825 |
+
"rewards/chosen": -0.275390625,
|
| 1826 |
+
"rewards/margins": 0.71484375,
|
| 1827 |
+
"rewards/rejected": -0.98828125,
|
| 1828 |
+
"step": 520
|
| 1829 |
+
},
|
| 1830 |
+
{
|
| 1831 |
+
"epoch": 0.729312762973352,
|
| 1832 |
+
"eval_logits/chosen": -3.0625,
|
| 1833 |
+
"eval_logits/rejected": -3.09375,
|
| 1834 |
+
"eval_logps/chosen": -354.0,
|
| 1835 |
+
"eval_logps/rejected": -372.0,
|
| 1836 |
+
"eval_loss": 0.5146514177322388,
|
| 1837 |
+
"eval_rewards/accuracies": 0.7872340679168701,
|
| 1838 |
+
"eval_rewards/chosen": -0.328125,
|
| 1839 |
+
"eval_rewards/margins": 0.6328125,
|
| 1840 |
+
"eval_rewards/rejected": -0.9609375,
|
| 1841 |
+
"eval_runtime": 65.754,
|
| 1842 |
+
"eval_samples_per_second": 22.615,
|
| 1843 |
+
"eval_steps_per_second": 0.715,
|
| 1844 |
+
"step": 520
|
| 1845 |
+
},
|
| 1846 |
+
{
|
| 1847 |
+
"epoch": 0.7405329593267882,
|
| 1848 |
+
"eval_logits/chosen": -3.0625,
|
| 1849 |
+
"eval_logits/rejected": -3.09375,
|
| 1850 |
+
"eval_logps/chosen": -354.0,
|
| 1851 |
+
"eval_logps/rejected": -372.0,
|
| 1852 |
+
"eval_loss": 0.5139682292938232,
|
| 1853 |
+
"eval_rewards/accuracies": 0.792553186416626,
|
| 1854 |
+
"eval_rewards/chosen": -0.326171875,
|
| 1855 |
+
"eval_rewards/margins": 0.63671875,
|
| 1856 |
+
"eval_rewards/rejected": -0.9609375,
|
| 1857 |
+
"eval_runtime": 65.2389,
|
| 1858 |
+
"eval_samples_per_second": 22.793,
|
| 1859 |
+
"eval_steps_per_second": 0.72,
|
| 1860 |
+
"step": 528
|
| 1861 |
+
},
|
| 1862 |
+
{
|
| 1863 |
+
"epoch": 0.7433380084151473,
|
| 1864 |
+
"grad_norm": 26.831689294819764,
|
| 1865 |
+
"learning_rate": 1.8798147414005737e-08,
|
| 1866 |
+
"logits/chosen": -3.0,
|
| 1867 |
+
"logits/rejected": -2.875,
|
| 1868 |
+
"logps/chosen": -330.0,
|
| 1869 |
+
"logps/rejected": -372.0,
|
| 1870 |
+
"loss": 0.5183,
|
| 1871 |
+
"rewards/accuracies": 0.762499988079071,
|
| 1872 |
+
"rewards/chosen": -0.34765625,
|
| 1873 |
+
"rewards/margins": 0.5859375,
|
| 1874 |
+
"rewards/rejected": -0.93359375,
|
| 1875 |
+
"step": 530
|
| 1876 |
+
},
|
| 1877 |
+
{
|
| 1878 |
+
"epoch": 0.7517531556802244,
|
| 1879 |
+
"eval_logits/chosen": -3.0625,
|
| 1880 |
+
"eval_logits/rejected": -3.09375,
|
| 1881 |
+
"eval_logps/chosen": -352.0,
|
| 1882 |
+
"eval_logps/rejected": -370.0,
|
| 1883 |
+
"eval_loss": 0.5127372741699219,
|
| 1884 |
+
"eval_rewards/accuracies": 0.7978723645210266,
|
| 1885 |
+
"eval_rewards/chosen": -0.298828125,
|
| 1886 |
+
"eval_rewards/margins": 0.64453125,
|
| 1887 |
+
"eval_rewards/rejected": -0.9453125,
|
| 1888 |
+
"eval_runtime": 66.3192,
|
| 1889 |
+
"eval_samples_per_second": 22.422,
|
| 1890 |
+
"eval_steps_per_second": 0.709,
|
| 1891 |
+
"step": 536
|
| 1892 |
+
},
|
| 1893 |
+
{
|
| 1894 |
+
"epoch": 0.7573632538569425,
|
| 1895 |
+
"grad_norm": 46.59871665028949,
|
| 1896 |
+
"learning_rate": 1.692154143820063e-08,
|
| 1897 |
+
"logits/chosen": -3.03125,
|
| 1898 |
+
"logits/rejected": -2.96875,
|
| 1899 |
+
"logps/chosen": -348.0,
|
| 1900 |
+
"logps/rejected": -328.0,
|
| 1901 |
+
"loss": 0.5034,
|
| 1902 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 1903 |
+
"rewards/chosen": -0.388671875,
|
| 1904 |
+
"rewards/margins": 0.515625,
|
| 1905 |
+
"rewards/rejected": -0.90234375,
|
| 1906 |
+
"step": 540
|
| 1907 |
+
},
|
| 1908 |
+
{
|
| 1909 |
+
"epoch": 0.7629733520336606,
|
| 1910 |
+
"eval_logits/chosen": -3.0625,
|
| 1911 |
+
"eval_logits/rejected": -3.09375,
|
| 1912 |
+
"eval_logps/chosen": -350.0,
|
| 1913 |
+
"eval_logps/rejected": -368.0,
|
| 1914 |
+
"eval_loss": 0.5124016404151917,
|
| 1915 |
+
"eval_rewards/accuracies": 0.7978723645210266,
|
| 1916 |
+
"eval_rewards/chosen": -0.271484375,
|
| 1917 |
+
"eval_rewards/margins": 0.65234375,
|
| 1918 |
+
"eval_rewards/rejected": -0.92578125,
|
| 1919 |
+
"eval_runtime": 68.1586,
|
| 1920 |
+
"eval_samples_per_second": 21.817,
|
| 1921 |
+
"eval_steps_per_second": 0.69,
|
| 1922 |
+
"step": 544
|
| 1923 |
+
},
|
| 1924 |
+
{
|
| 1925 |
+
"epoch": 0.7713884992987378,
|
| 1926 |
+
"grad_norm": 17.28019155088413,
|
| 1927 |
+
"learning_rate": 1.5124375970755755e-08,
|
| 1928 |
+
"logits/chosen": -3.078125,
|
| 1929 |
+
"logits/rejected": -3.078125,
|
| 1930 |
+
"logps/chosen": -356.0,
|
| 1931 |
+
"logps/rejected": -392.0,
|
| 1932 |
+
"loss": 0.5075,
|
| 1933 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1934 |
+
"rewards/chosen": -0.390625,
|
| 1935 |
+
"rewards/margins": 0.62890625,
|
| 1936 |
+
"rewards/rejected": -1.015625,
|
| 1937 |
+
"step": 550
|
| 1938 |
+
},
|
| 1939 |
+
{
|
| 1940 |
+
"epoch": 0.7741935483870968,
|
| 1941 |
+
"eval_logits/chosen": -3.0625,
|
| 1942 |
+
"eval_logits/rejected": -3.09375,
|
| 1943 |
+
"eval_logps/chosen": -350.0,
|
| 1944 |
+
"eval_logps/rejected": -368.0,
|
| 1945 |
+
"eval_loss": 0.51242595911026,
|
| 1946 |
+
"eval_rewards/accuracies": 0.792553186416626,
|
| 1947 |
+
"eval_rewards/chosen": -0.28515625,
|
| 1948 |
+
"eval_rewards/margins": 0.65234375,
|
| 1949 |
+
"eval_rewards/rejected": -0.9375,
|
| 1950 |
+
"eval_runtime": 66.0347,
|
| 1951 |
+
"eval_samples_per_second": 22.518,
|
| 1952 |
+
"eval_steps_per_second": 0.712,
|
| 1953 |
+
"step": 552
|
| 1954 |
+
},
|
| 1955 |
+
{
|
| 1956 |
+
"epoch": 0.7854137447405329,
|
| 1957 |
+
"grad_norm": 18.119529412971673,
|
| 1958 |
+
"learning_rate": 1.3410967045263622e-08,
|
| 1959 |
+
"logits/chosen": -3.0,
|
| 1960 |
+
"logits/rejected": -2.90625,
|
| 1961 |
+
"logps/chosen": -348.0,
|
| 1962 |
+
"logps/rejected": -384.0,
|
| 1963 |
+
"loss": 0.5274,
|
| 1964 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1965 |
+
"rewards/chosen": -0.29296875,
|
| 1966 |
+
"rewards/margins": 0.703125,
|
| 1967 |
+
"rewards/rejected": -0.99609375,
|
| 1968 |
+
"step": 560
|
| 1969 |
+
},
|
| 1970 |
+
{
|
| 1971 |
+
"epoch": 0.7854137447405329,
|
| 1972 |
+
"eval_logits/chosen": -3.0625,
|
| 1973 |
+
"eval_logits/rejected": -3.09375,
|
| 1974 |
+
"eval_logps/chosen": -354.0,
|
| 1975 |
+
"eval_logps/rejected": -372.0,
|
| 1976 |
+
"eval_loss": 0.5127307176589966,
|
| 1977 |
+
"eval_rewards/accuracies": 0.792553186416626,
|
| 1978 |
+
"eval_rewards/chosen": -0.31640625,
|
| 1979 |
+
"eval_rewards/margins": 0.64453125,
|
| 1980 |
+
"eval_rewards/rejected": -0.9609375,
|
| 1981 |
+
"eval_runtime": 65.7053,
|
| 1982 |
+
"eval_samples_per_second": 22.631,
|
| 1983 |
+
"eval_steps_per_second": 0.715,
|
| 1984 |
+
"step": 560
|
| 1985 |
+
},
|
| 1986 |
+
{
|
| 1987 |
+
"epoch": 0.7966339410939691,
|
| 1988 |
+
"eval_logits/chosen": -3.0625,
|
| 1989 |
+
"eval_logits/rejected": -3.09375,
|
| 1990 |
+
"eval_logps/chosen": -354.0,
|
| 1991 |
+
"eval_logps/rejected": -372.0,
|
| 1992 |
+
"eval_loss": 0.5127339959144592,
|
| 1993 |
+
"eval_rewards/accuracies": 0.7819148898124695,
|
| 1994 |
+
"eval_rewards/chosen": -0.3203125,
|
| 1995 |
+
"eval_rewards/margins": 0.64453125,
|
| 1996 |
+
"eval_rewards/rejected": -0.96875,
|
| 1997 |
+
"eval_runtime": 65.5474,
|
| 1998 |
+
"eval_samples_per_second": 22.686,
|
| 1999 |
+
"eval_steps_per_second": 0.717,
|
| 2000 |
+
"step": 568
|
| 2001 |
+
},
|
| 2002 |
+
{
|
| 2003 |
+
"epoch": 0.7994389901823282,
|
| 2004 |
+
"grad_norm": 28.16590781200549,
|
| 2005 |
+
"learning_rate": 1.1785429547422909e-08,
|
| 2006 |
+
"logits/chosen": -3.0,
|
| 2007 |
+
"logits/rejected": -3.03125,
|
| 2008 |
+
"logps/chosen": -318.0,
|
| 2009 |
+
"logps/rejected": -362.0,
|
| 2010 |
+
"loss": 0.5138,
|
| 2011 |
+
"rewards/accuracies": 0.699999988079071,
|
| 2012 |
+
"rewards/chosen": -0.462890625,
|
| 2013 |
+
"rewards/margins": 0.48828125,
|
| 2014 |
+
"rewards/rejected": -0.953125,
|
| 2015 |
+
"step": 570
|
| 2016 |
+
},
|
| 2017 |
+
{
|
| 2018 |
+
"epoch": 0.8078541374474053,
|
| 2019 |
+
"eval_logits/chosen": -3.0625,
|
| 2020 |
+
"eval_logits/rejected": -3.09375,
|
| 2021 |
+
"eval_logps/chosen": -354.0,
|
| 2022 |
+
"eval_logps/rejected": -372.0,
|
| 2023 |
+
"eval_loss": 0.5122042298316956,
|
| 2024 |
+
"eval_rewards/accuracies": 0.7872340679168701,
|
| 2025 |
+
"eval_rewards/chosen": -0.3125,
|
| 2026 |
+
"eval_rewards/margins": 0.6484375,
|
| 2027 |
+
"eval_rewards/rejected": -0.9609375,
|
| 2028 |
+
"eval_runtime": 65.369,
|
| 2029 |
+
"eval_samples_per_second": 22.748,
|
| 2030 |
+
"eval_steps_per_second": 0.719,
|
| 2031 |
+
"step": 576
|
| 2032 |
+
},
|
| 2033 |
+
{
|
| 2034 |
+
"epoch": 0.8134642356241234,
|
| 2035 |
+
"grad_norm": 20.331557503849197,
|
| 2036 |
+
"learning_rate": 1.0251667332818215e-08,
|
| 2037 |
+
"logits/chosen": -2.875,
|
| 2038 |
+
"logits/rejected": -2.96875,
|
| 2039 |
+
"logps/chosen": -364.0,
|
| 2040 |
+
"logps/rejected": -390.0,
|
| 2041 |
+
"loss": 0.5087,
|
| 2042 |
+
"rewards/accuracies": 0.824999988079071,
|
| 2043 |
+
"rewards/chosen": -0.4140625,
|
| 2044 |
+
"rewards/margins": 0.75390625,
|
| 2045 |
+
"rewards/rejected": -1.171875,
|
| 2046 |
+
"step": 580
|
| 2047 |
+
},
|
| 2048 |
+
{
|
| 2049 |
+
"epoch": 0.8190743338008415,
|
| 2050 |
+
"eval_logits/chosen": -3.0625,
|
| 2051 |
+
"eval_logits/rejected": -3.09375,
|
| 2052 |
+
"eval_logps/chosen": -352.0,
|
| 2053 |
+
"eval_logps/rejected": -370.0,
|
| 2054 |
+
"eval_loss": 0.5114827156066895,
|
| 2055 |
+
"eval_rewards/accuracies": 0.7978723645210266,
|
| 2056 |
+
"eval_rewards/chosen": -0.30078125,
|
| 2057 |
+
"eval_rewards/margins": 0.65234375,
|
| 2058 |
+
"eval_rewards/rejected": -0.953125,
|
| 2059 |
+
"eval_runtime": 65.3364,
|
| 2060 |
+
"eval_samples_per_second": 22.759,
|
| 2061 |
+
"eval_steps_per_second": 0.719,
|
| 2062 |
+
"step": 584
|
| 2063 |
+
},
|
| 2064 |
+
{
|
| 2065 |
+
"epoch": 0.8274894810659187,
|
| 2066 |
+
"grad_norm": 26.321821934426545,
|
| 2067 |
+
"learning_rate": 8.813363851505284e-09,
|
| 2068 |
+
"logits/chosen": -2.96875,
|
| 2069 |
+
"logits/rejected": -2.96875,
|
| 2070 |
+
"logps/chosen": -412.0,
|
| 2071 |
+
"logps/rejected": -410.0,
|
| 2072 |
+
"loss": 0.5026,
|
| 2073 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 2074 |
+
"rewards/chosen": -0.46875,
|
| 2075 |
+
"rewards/margins": 0.55078125,
|
| 2076 |
+
"rewards/rejected": -1.0234375,
|
| 2077 |
+
"step": 590
|
| 2078 |
+
},
|
| 2079 |
+
{
|
| 2080 |
+
"epoch": 0.8302945301542777,
|
| 2081 |
+
"eval_logits/chosen": -3.0625,
|
| 2082 |
+
"eval_logits/rejected": -3.078125,
|
| 2083 |
+
"eval_logps/chosen": -352.0,
|
| 2084 |
+
"eval_logps/rejected": -370.0,
|
| 2085 |
+
"eval_loss": 0.5113134384155273,
|
| 2086 |
+
"eval_rewards/accuracies": 0.7978723645210266,
|
| 2087 |
+
"eval_rewards/chosen": -0.30078125,
|
| 2088 |
+
"eval_rewards/margins": 0.65234375,
|
| 2089 |
+
"eval_rewards/rejected": -0.953125,
|
| 2090 |
+
"eval_runtime": 65.7466,
|
| 2091 |
+
"eval_samples_per_second": 22.617,
|
| 2092 |
+
"eval_steps_per_second": 0.715,
|
| 2093 |
+
"step": 592
|
| 2094 |
+
},
|
| 2095 |
+
{
|
| 2096 |
+
"epoch": 0.8415147265077139,
|
| 2097 |
+
"grad_norm": 17.927505615509187,
|
| 2098 |
+
"learning_rate": 7.473973301917124e-09,
|
| 2099 |
+
"logits/chosen": -2.90625,
|
| 2100 |
+
"logits/rejected": -2.875,
|
| 2101 |
+
"logps/chosen": -322.0,
|
| 2102 |
+
"logps/rejected": -352.0,
|
| 2103 |
+
"loss": 0.4993,
|
| 2104 |
+
"rewards/accuracies": 0.887499988079071,
|
| 2105 |
+
"rewards/chosen": -0.26171875,
|
| 2106 |
+
"rewards/margins": 0.63671875,
|
| 2107 |
+
"rewards/rejected": -0.8984375,
|
| 2108 |
+
"step": 600
|
| 2109 |
+
},
|
| 2110 |
+
{
|
| 2111 |
+
"epoch": 0.8415147265077139,
|
| 2112 |
+
"eval_logits/chosen": -3.0625,
|
| 2113 |
+
"eval_logits/rejected": -3.078125,
|
| 2114 |
+
"eval_logps/chosen": -352.0,
|
| 2115 |
+
"eval_logps/rejected": -372.0,
|
| 2116 |
+
"eval_loss": 0.5112673044204712,
|
| 2117 |
+
"eval_rewards/accuracies": 0.7978723645210266,
|
| 2118 |
+
"eval_rewards/chosen": -0.3046875,
|
| 2119 |
+
"eval_rewards/margins": 0.65234375,
|
| 2120 |
+
"eval_rewards/rejected": -0.95703125,
|
| 2121 |
+
"eval_runtime": 65.3172,
|
| 2122 |
+
"eval_samples_per_second": 22.766,
|
| 2123 |
+
"eval_steps_per_second": 0.72,
|
| 2124 |
+
"step": 600
|
| 2125 |
+
},
|
| 2126 |
+
{
|
| 2127 |
+
"epoch": 0.85273492286115,
|
| 2128 |
+
"eval_logits/chosen": -3.0625,
|
| 2129 |
+
"eval_logits/rejected": -3.078125,
|
| 2130 |
+
"eval_logps/chosen": -354.0,
|
| 2131 |
+
"eval_logps/rejected": -372.0,
|
| 2132 |
+
"eval_loss": 0.5110276341438293,
|
| 2133 |
+
"eval_rewards/accuracies": 0.7872340679168701,
|
| 2134 |
+
"eval_rewards/chosen": -0.314453125,
|
| 2135 |
+
"eval_rewards/margins": 0.6484375,
|
| 2136 |
+
"eval_rewards/rejected": -0.9609375,
|
| 2137 |
+
"eval_runtime": 65.4306,
|
| 2138 |
+
"eval_samples_per_second": 22.726,
|
| 2139 |
+
"eval_steps_per_second": 0.718,
|
| 2140 |
+
"step": 608
|
| 2141 |
+
},
|
| 2142 |
+
{
|
| 2143 |
+
"epoch": 0.8555399719495091,
|
| 2144 |
+
"grad_norm": 23.917252124640086,
|
| 2145 |
+
"learning_rate": 6.236712335336131e-09,
|
| 2146 |
+
"logits/chosen": -3.0625,
|
| 2147 |
+
"logits/rejected": -2.90625,
|
| 2148 |
+
"logps/chosen": -292.0,
|
| 2149 |
+
"logps/rejected": -342.0,
|
| 2150 |
+
"loss": 0.5145,
|
| 2151 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 2152 |
+
"rewards/chosen": -0.37109375,
|
| 2153 |
+
"rewards/margins": 0.578125,
|
| 2154 |
+
"rewards/rejected": -0.94921875,
|
| 2155 |
+
"step": 610
|
| 2156 |
+
},
|
| 2157 |
+
{
|
| 2158 |
+
"epoch": 0.8639551192145862,
|
| 2159 |
+
"eval_logits/chosen": -3.0625,
|
| 2160 |
+
"eval_logits/rejected": -3.078125,
|
| 2161 |
+
"eval_logps/chosen": -354.0,
|
| 2162 |
+
"eval_logps/rejected": -372.0,
|
| 2163 |
+
"eval_loss": 0.5110043883323669,
|
| 2164 |
+
"eval_rewards/accuracies": 0.7872340679168701,
|
| 2165 |
+
"eval_rewards/chosen": -0.31640625,
|
| 2166 |
+
"eval_rewards/margins": 0.6484375,
|
| 2167 |
+
"eval_rewards/rejected": -0.96484375,
|
| 2168 |
+
"eval_runtime": 65.4869,
|
| 2169 |
+
"eval_samples_per_second": 22.707,
|
| 2170 |
+
"eval_steps_per_second": 0.718,
|
| 2171 |
+
"step": 616
|
| 2172 |
+
},
|
| 2173 |
+
{
|
| 2174 |
+
"epoch": 0.8695652173913043,
|
| 2175 |
+
"grad_norm": 35.23685131659172,
|
| 2176 |
+
"learning_rate": 5.104552330854112e-09,
|
| 2177 |
+
"logits/chosen": -2.921875,
|
| 2178 |
+
"logits/rejected": -2.90625,
|
| 2179 |
+
"logps/chosen": -406.0,
|
| 2180 |
+
"logps/rejected": -434.0,
|
| 2181 |
+
"loss": 0.4857,
|
| 2182 |
+
"rewards/accuracies": 0.862500011920929,
|
| 2183 |
+
"rewards/chosen": -0.353515625,
|
| 2184 |
+
"rewards/margins": 0.76953125,
|
| 2185 |
+
"rewards/rejected": -1.125,
|
| 2186 |
+
"step": 620
|
| 2187 |
+
},
|
| 2188 |
+
{
|
| 2189 |
+
"epoch": 0.8751753155680224,
|
| 2190 |
+
"eval_logits/chosen": -3.0625,
|
| 2191 |
+
"eval_logits/rejected": -3.078125,
|
| 2192 |
+
"eval_logps/chosen": -354.0,
|
| 2193 |
+
"eval_logps/rejected": -372.0,
|
| 2194 |
+
"eval_loss": 0.5109130144119263,
|
| 2195 |
+
"eval_rewards/accuracies": 0.7872340679168701,
|
| 2196 |
+
"eval_rewards/chosen": -0.31640625,
|
| 2197 |
+
"eval_rewards/margins": 0.65234375,
|
| 2198 |
+
"eval_rewards/rejected": -0.96875,
|
| 2199 |
+
"eval_runtime": 65.4952,
|
| 2200 |
+
"eval_samples_per_second": 22.704,
|
| 2201 |
+
"eval_steps_per_second": 0.718,
|
| 2202 |
+
"step": 624
|
| 2203 |
+
},
|
| 2204 |
+
{
|
| 2205 |
+
"epoch": 0.8835904628330996,
|
| 2206 |
+
"grad_norm": 28.24616740578334,
|
| 2207 |
+
"learning_rate": 4.080212259372711e-09,
|
| 2208 |
+
"logits/chosen": -3.0625,
|
| 2209 |
+
"logits/rejected": -3.0625,
|
| 2210 |
+
"logps/chosen": -334.0,
|
| 2211 |
+
"logps/rejected": -356.0,
|
| 2212 |
+
"loss": 0.5163,
|
| 2213 |
+
"rewards/accuracies": 0.887499988079071,
|
| 2214 |
+
"rewards/chosen": -0.33984375,
|
| 2215 |
+
"rewards/margins": 0.73046875,
|
| 2216 |
+
"rewards/rejected": -1.0703125,
|
| 2217 |
+
"step": 630
|
| 2218 |
+
},
|
| 2219 |
+
{
|
| 2220 |
+
"epoch": 0.8863955119214586,
|
| 2221 |
+
"eval_logits/chosen": -3.0625,
|
| 2222 |
+
"eval_logits/rejected": -3.078125,
|
| 2223 |
+
"eval_logps/chosen": -354.0,
|
| 2224 |
+
"eval_logps/rejected": -372.0,
|
| 2225 |
+
"eval_loss": 0.5109508037567139,
|
| 2226 |
+
"eval_rewards/accuracies": 0.7872340679168701,
|
| 2227 |
+
"eval_rewards/chosen": -0.3203125,
|
| 2228 |
+
"eval_rewards/margins": 0.6484375,
|
| 2229 |
+
"eval_rewards/rejected": -0.96875,
|
| 2230 |
+
"eval_runtime": 65.7135,
|
| 2231 |
+
"eval_samples_per_second": 22.629,
|
| 2232 |
+
"eval_steps_per_second": 0.715,
|
| 2233 |
+
"step": 632
|
| 2234 |
+
},
|
| 2235 |
+
{
|
| 2236 |
+
"epoch": 0.8976157082748948,
|
| 2237 |
+
"grad_norm": 23.74887530831686,
|
| 2238 |
+
"learning_rate": 3.1661521537819257e-09,
|
| 2239 |
+
"logits/chosen": -3.03125,
|
| 2240 |
+
"logits/rejected": -2.96875,
|
| 2241 |
+
"logps/chosen": -384.0,
|
| 2242 |
+
"logps/rejected": -402.0,
|
| 2243 |
+
"loss": 0.5258,
|
| 2244 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 2245 |
+
"rewards/chosen": -0.34375,
|
| 2246 |
+
"rewards/margins": 0.73828125,
|
| 2247 |
+
"rewards/rejected": -1.0859375,
|
| 2248 |
+
"step": 640
|
| 2249 |
+
},
|
| 2250 |
+
{
|
| 2251 |
+
"epoch": 0.8976157082748948,
|
| 2252 |
+
"eval_logits/chosen": -3.0625,
|
| 2253 |
+
"eval_logits/rejected": -3.078125,
|
| 2254 |
+
"eval_logps/chosen": -354.0,
|
| 2255 |
+
"eval_logps/rejected": -372.0,
|
| 2256 |
+
"eval_loss": 0.5108342170715332,
|
| 2257 |
+
"eval_rewards/accuracies": 0.7819148898124695,
|
| 2258 |
+
"eval_rewards/chosen": -0.32421875,
|
| 2259 |
+
"eval_rewards/margins": 0.65234375,
|
| 2260 |
+
"eval_rewards/rejected": -0.9765625,
|
| 2261 |
+
"eval_runtime": 65.4352,
|
| 2262 |
+
"eval_samples_per_second": 22.725,
|
| 2263 |
+
"eval_steps_per_second": 0.718,
|
| 2264 |
+
"step": 640
|
| 2265 |
+
},
|
| 2266 |
+
{
|
| 2267 |
+
"epoch": 0.908835904628331,
|
| 2268 |
+
"eval_logits/chosen": -3.0625,
|
| 2269 |
+
"eval_logits/rejected": -3.078125,
|
| 2270 |
+
"eval_logps/chosen": -354.0,
|
| 2271 |
+
"eval_logps/rejected": -374.0,
|
| 2272 |
+
"eval_loss": 0.510999321937561,
|
| 2273 |
+
"eval_rewards/accuracies": 0.7765957713127136,
|
| 2274 |
+
"eval_rewards/chosen": -0.328125,
|
| 2275 |
+
"eval_rewards/margins": 0.65234375,
|
| 2276 |
+
"eval_rewards/rejected": -0.98046875,
|
| 2277 |
+
"eval_runtime": 66.8594,
|
| 2278 |
+
"eval_samples_per_second": 22.241,
|
| 2279 |
+
"eval_steps_per_second": 0.703,
|
| 2280 |
+
"step": 648
|
| 2281 |
+
},
|
| 2282 |
+
{
|
| 2283 |
+
"epoch": 0.9116409537166901,
|
| 2284 |
+
"grad_norm": 23.848849763801983,
|
| 2285 |
+
"learning_rate": 2.3645672009984684e-09,
|
| 2286 |
+
"logits/chosen": -2.90625,
|
| 2287 |
+
"logits/rejected": -2.859375,
|
| 2288 |
+
"logps/chosen": -334.0,
|
| 2289 |
+
"logps/rejected": -372.0,
|
| 2290 |
+
"loss": 0.5206,
|
| 2291 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 2292 |
+
"rewards/chosen": -0.4765625,
|
| 2293 |
+
"rewards/margins": 0.56640625,
|
| 2294 |
+
"rewards/rejected": -1.046875,
|
| 2295 |
+
"step": 650
|
| 2296 |
+
},
|
| 2297 |
+
{
|
| 2298 |
+
"epoch": 0.9200561009817672,
|
| 2299 |
+
"eval_logits/chosen": -3.0625,
|
| 2300 |
+
"eval_logits/rejected": -3.078125,
|
| 2301 |
+
"eval_logps/chosen": -354.0,
|
| 2302 |
+
"eval_logps/rejected": -374.0,
|
| 2303 |
+
"eval_loss": 0.5108804702758789,
|
| 2304 |
+
"eval_rewards/accuracies": 0.7765957713127136,
|
| 2305 |
+
"eval_rewards/chosen": -0.326171875,
|
| 2306 |
+
"eval_rewards/margins": 0.65234375,
|
| 2307 |
+
"eval_rewards/rejected": -0.98046875,
|
| 2308 |
+
"eval_runtime": 66.985,
|
| 2309 |
+
"eval_samples_per_second": 22.199,
|
| 2310 |
+
"eval_steps_per_second": 0.702,
|
| 2311 |
+
"step": 656
|
| 2312 |
+
},
|
| 2313 |
+
{
|
| 2314 |
+
"epoch": 0.9256661991584852,
|
| 2315 |
+
"grad_norm": 20.841266069644966,
|
| 2316 |
+
"learning_rate": 1.677382470052513e-09,
|
| 2317 |
+
"logits/chosen": -3.0,
|
| 2318 |
+
"logits/rejected": -2.984375,
|
| 2319 |
+
"logps/chosen": -412.0,
|
| 2320 |
+
"logps/rejected": -372.0,
|
| 2321 |
+
"loss": 0.4926,
|
| 2322 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 2323 |
+
"rewards/chosen": -0.43359375,
|
| 2324 |
+
"rewards/margins": 0.6875,
|
| 2325 |
+
"rewards/rejected": -1.125,
|
| 2326 |
+
"step": 660
|
| 2327 |
+
},
|
| 2328 |
+
{
|
| 2329 |
+
"epoch": 0.9312762973352033,
|
| 2330 |
+
"eval_logits/chosen": -3.0625,
|
| 2331 |
+
"eval_logits/rejected": -3.078125,
|
| 2332 |
+
"eval_logps/chosen": -354.0,
|
| 2333 |
+
"eval_logps/rejected": -374.0,
|
| 2334 |
+
"eval_loss": 0.5108736753463745,
|
| 2335 |
+
"eval_rewards/accuracies": 0.7819148898124695,
|
| 2336 |
+
"eval_rewards/chosen": -0.328125,
|
| 2337 |
+
"eval_rewards/margins": 0.65234375,
|
| 2338 |
+
"eval_rewards/rejected": -0.98046875,
|
| 2339 |
+
"eval_runtime": 66.0163,
|
| 2340 |
+
"eval_samples_per_second": 22.525,
|
| 2341 |
+
"eval_steps_per_second": 0.712,
|
| 2342 |
+
"step": 664
|
| 2343 |
+
},
|
| 2344 |
+
{
|
| 2345 |
+
"epoch": 0.9396914446002805,
|
| 2346 |
+
"grad_norm": 42.65600681696292,
|
| 2347 |
+
"learning_rate": 1.1062482888836656e-09,
|
| 2348 |
+
"logits/chosen": -3.0625,
|
| 2349 |
+
"logits/rejected": -2.953125,
|
| 2350 |
+
"logps/chosen": -372.0,
|
| 2351 |
+
"logps/rejected": -374.0,
|
| 2352 |
+
"loss": 0.5,
|
| 2353 |
+
"rewards/accuracies": 0.875,
|
| 2354 |
+
"rewards/chosen": -0.349609375,
|
| 2355 |
+
"rewards/margins": 0.8828125,
|
| 2356 |
+
"rewards/rejected": -1.234375,
|
| 2357 |
+
"step": 670
|
| 2358 |
+
},
|
| 2359 |
+
{
|
| 2360 |
+
"epoch": 0.9424964936886395,
|
| 2361 |
+
"eval_logits/chosen": -3.0625,
|
| 2362 |
+
"eval_logits/rejected": -3.078125,
|
| 2363 |
+
"eval_logps/chosen": -354.0,
|
| 2364 |
+
"eval_logps/rejected": -374.0,
|
| 2365 |
+
"eval_loss": 0.5107899904251099,
|
| 2366 |
+
"eval_rewards/accuracies": 0.7819148898124695,
|
| 2367 |
+
"eval_rewards/chosen": -0.326171875,
|
| 2368 |
+
"eval_rewards/margins": 0.65625,
|
| 2369 |
+
"eval_rewards/rejected": -0.984375,
|
| 2370 |
+
"eval_runtime": 65.1766,
|
| 2371 |
+
"eval_samples_per_second": 22.815,
|
| 2372 |
+
"eval_steps_per_second": 0.721,
|
| 2373 |
+
"step": 672
|
| 2374 |
+
},
|
| 2375 |
+
{
|
| 2376 |
+
"epoch": 0.9537166900420757,
|
| 2377 |
+
"grad_norm": 22.14348589619,
|
| 2378 |
+
"learning_rate": 6.525362809492008e-10,
|
| 2379 |
+
"logits/chosen": -2.90625,
|
| 2380 |
+
"logits/rejected": -3.078125,
|
| 2381 |
+
"logps/chosen": -350.0,
|
| 2382 |
+
"logps/rejected": -390.0,
|
| 2383 |
+
"loss": 0.5101,
|
| 2384 |
+
"rewards/accuracies": 0.8125,
|
| 2385 |
+
"rewards/chosen": -0.26953125,
|
| 2386 |
+
"rewards/margins": 0.76171875,
|
| 2387 |
+
"rewards/rejected": -1.03125,
|
| 2388 |
+
"step": 680
|
| 2389 |
+
},
|
| 2390 |
+
{
|
| 2391 |
+
"epoch": 0.9537166900420757,
|
| 2392 |
+
"eval_logits/chosen": -3.0625,
|
| 2393 |
+
"eval_logits/rejected": -3.078125,
|
| 2394 |
+
"eval_logps/chosen": -354.0,
|
| 2395 |
+
"eval_logps/rejected": -372.0,
|
| 2396 |
+
"eval_loss": 0.5109755992889404,
|
| 2397 |
+
"eval_rewards/accuracies": 0.7872340679168701,
|
| 2398 |
+
"eval_rewards/chosen": -0.32421875,
|
| 2399 |
+
"eval_rewards/margins": 0.65234375,
|
| 2400 |
+
"eval_rewards/rejected": -0.9765625,
|
| 2401 |
+
"eval_runtime": 65.4357,
|
| 2402 |
+
"eval_samples_per_second": 22.725,
|
| 2403 |
+
"eval_steps_per_second": 0.718,
|
| 2404 |
+
"step": 680
|
| 2405 |
+
},
|
| 2406 |
+
{
|
| 2407 |
+
"epoch": 0.9649368863955119,
|
| 2408 |
+
"eval_logits/chosen": -3.0625,
|
| 2409 |
+
"eval_logits/rejected": -3.078125,
|
| 2410 |
+
"eval_logps/chosen": -354.0,
|
| 2411 |
+
"eval_logps/rejected": -374.0,
|
| 2412 |
+
"eval_loss": 0.510669469833374,
|
| 2413 |
+
"eval_rewards/accuracies": 0.7819148898124695,
|
| 2414 |
+
"eval_rewards/chosen": -0.32421875,
|
| 2415 |
+
"eval_rewards/margins": 0.65234375,
|
| 2416 |
+
"eval_rewards/rejected": -0.9765625,
|
| 2417 |
+
"eval_runtime": 65.5907,
|
| 2418 |
+
"eval_samples_per_second": 22.671,
|
| 2419 |
+
"eval_steps_per_second": 0.717,
|
| 2420 |
+
"step": 688
|
| 2421 |
+
},
|
| 2422 |
+
{
|
| 2423 |
+
"epoch": 0.967741935483871,
|
| 2424 |
+
"grad_norm": 18.79687876463209,
|
| 2425 |
+
"learning_rate": 3.173360711629325e-10,
|
| 2426 |
+
"logits/chosen": -2.921875,
|
| 2427 |
+
"logits/rejected": -3.0,
|
| 2428 |
+
"logps/chosen": -284.0,
|
| 2429 |
+
"logps/rejected": -380.0,
|
| 2430 |
+
"loss": 0.4995,
|
| 2431 |
+
"rewards/accuracies": 0.824999988079071,
|
| 2432 |
+
"rewards/chosen": -0.357421875,
|
| 2433 |
+
"rewards/margins": 0.64453125,
|
| 2434 |
+
"rewards/rejected": -1.0,
|
| 2435 |
+
"step": 690
|
| 2436 |
+
},
|
| 2437 |
+
{
|
| 2438 |
+
"epoch": 0.9761570827489481,
|
| 2439 |
+
"eval_logits/chosen": -3.0625,
|
| 2440 |
+
"eval_logits/rejected": -3.078125,
|
| 2441 |
+
"eval_logps/chosen": -354.0,
|
| 2442 |
+
"eval_logps/rejected": -372.0,
|
| 2443 |
+
"eval_loss": 0.5105814337730408,
|
| 2444 |
+
"eval_rewards/accuracies": 0.7872340679168701,
|
| 2445 |
+
"eval_rewards/chosen": -0.32421875,
|
| 2446 |
+
"eval_rewards/margins": 0.65234375,
|
| 2447 |
+
"eval_rewards/rejected": -0.9765625,
|
| 2448 |
+
"eval_runtime": 65.6777,
|
| 2449 |
+
"eval_samples_per_second": 22.641,
|
| 2450 |
+
"eval_steps_per_second": 0.716,
|
| 2451 |
+
"step": 696
|
| 2452 |
+
},
|
| 2453 |
+
{
|
| 2454 |
+
"epoch": 0.9817671809256662,
|
| 2455 |
+
"grad_norm": 27.90140864474871,
|
| 2456 |
+
"learning_rate": 1.014526690756634e-10,
|
| 2457 |
+
"logits/chosen": -3.046875,
|
| 2458 |
+
"logits/rejected": -3.09375,
|
| 2459 |
+
"logps/chosen": -368.0,
|
| 2460 |
+
"logps/rejected": -396.0,
|
| 2461 |
+
"loss": 0.5048,
|
| 2462 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 2463 |
+
"rewards/chosen": -0.341796875,
|
| 2464 |
+
"rewards/margins": 0.859375,
|
| 2465 |
+
"rewards/rejected": -1.203125,
|
| 2466 |
+
"step": 700
|
| 2467 |
+
},
|
| 2468 |
+
{
|
| 2469 |
+
"epoch": 0.9873772791023843,
|
| 2470 |
+
"eval_logits/chosen": -3.0625,
|
| 2471 |
+
"eval_logits/rejected": -3.078125,
|
| 2472 |
+
"eval_logps/chosen": -354.0,
|
| 2473 |
+
"eval_logps/rejected": -372.0,
|
| 2474 |
+
"eval_loss": 0.5105742812156677,
|
| 2475 |
+
"eval_rewards/accuracies": 0.7872340679168701,
|
| 2476 |
+
"eval_rewards/chosen": -0.322265625,
|
| 2477 |
+
"eval_rewards/margins": 0.65625,
|
| 2478 |
+
"eval_rewards/rejected": -0.9765625,
|
| 2479 |
+
"eval_runtime": 64.9937,
|
| 2480 |
+
"eval_samples_per_second": 22.879,
|
| 2481 |
+
"eval_steps_per_second": 0.723,
|
| 2482 |
+
"step": 704
|
| 2483 |
+
},
|
| 2484 |
+
{
|
| 2485 |
+
"epoch": 0.9957924263674615,
|
| 2486 |
+
"grad_norm": 25.796305051934237,
|
| 2487 |
+
"learning_rate": 5.404535581693403e-12,
|
| 2488 |
+
"logits/chosen": -3.015625,
|
| 2489 |
+
"logits/rejected": -2.953125,
|
| 2490 |
+
"logps/chosen": -370.0,
|
| 2491 |
+
"logps/rejected": -408.0,
|
| 2492 |
+
"loss": 0.5149,
|
| 2493 |
+
"rewards/accuracies": 0.800000011920929,
|
| 2494 |
+
"rewards/chosen": -0.392578125,
|
| 2495 |
+
"rewards/margins": 0.63671875,
|
| 2496 |
+
"rewards/rejected": -1.03125,
|
| 2497 |
+
"step": 710
|
| 2498 |
+
},
|
| 2499 |
+
{
|
| 2500 |
+
"epoch": 0.9985974754558204,
|
| 2501 |
+
"eval_logits/chosen": -3.0625,
|
| 2502 |
+
"eval_logits/rejected": -3.078125,
|
| 2503 |
+
"eval_logps/chosen": -354.0,
|
| 2504 |
+
"eval_logps/rejected": -374.0,
|
| 2505 |
+
"eval_loss": 0.5107670426368713,
|
| 2506 |
+
"eval_rewards/accuracies": 0.7872340679168701,
|
| 2507 |
+
"eval_rewards/chosen": -0.32421875,
|
| 2508 |
+
"eval_rewards/margins": 0.65234375,
|
| 2509 |
+
"eval_rewards/rejected": -0.9765625,
|
| 2510 |
+
"eval_runtime": 65.9108,
|
| 2511 |
+
"eval_samples_per_second": 22.561,
|
| 2512 |
+
"eval_steps_per_second": 0.713,
|
| 2513 |
+
"step": 712
|
| 2514 |
+
},
|
| 2515 |
+
{
|
| 2516 |
+
"epoch": 1.0,
|
| 2517 |
+
"step": 713,
|
| 2518 |
+
"total_flos": 0.0,
|
| 2519 |
+
"train_loss": 0.5638467967928208,
|
| 2520 |
+
"train_runtime": 32082.6297,
|
| 2521 |
+
"train_samples_per_second": 1.422,
|
| 2522 |
+
"train_steps_per_second": 0.022
|
| 2523 |
+
}
|
| 2524 |
+
],
|
| 2525 |
+
"logging_steps": 10,
|
| 2526 |
+
"max_steps": 713,
|
| 2527 |
+
"num_input_tokens_seen": 0,
|
| 2528 |
+
"num_train_epochs": 1,
|
| 2529 |
+
"save_steps": 8,
|
| 2530 |
+
"stateful_callbacks": {
|
| 2531 |
+
"TrainerControl": {
|
| 2532 |
+
"args": {
|
| 2533 |
+
"should_epoch_stop": false,
|
| 2534 |
+
"should_evaluate": false,
|
| 2535 |
+
"should_log": false,
|
| 2536 |
+
"should_save": true,
|
| 2537 |
+
"should_training_stop": true
|
| 2538 |
+
},
|
| 2539 |
+
"attributes": {}
|
| 2540 |
+
}
|
| 2541 |
+
},
|
| 2542 |
+
"total_flos": 0.0,
|
| 2543 |
+
"train_batch_size": 4,
|
| 2544 |
+
"trial_name": null,
|
| 2545 |
+
"trial_params": null
|
| 2546 |
+
}
|