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_lr5e-07_1
|
| 5 |
+
tags:
|
| 6 |
+
- generated_from_trainer
|
| 7 |
+
- trl
|
| 8 |
+
- dpo
|
| 9 |
+
licence: license
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
# Model Card for ultrafeedbackSkyworkAgree_alignmentZephyr7BSftFull_sdpo_score_ebs64_lr5e-07_1
|
| 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_lr5e-07_1", 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_lr5e-07_1try1S4eJmGXVUhdSrsIAqeWu49ao5UELq4jSJBSvoOVdLYLRFy)
|
| 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": 0.4091535410406212,
|
| 5 |
+
"train_runtime": 30868.5922,
|
| 6 |
+
"train_samples": 45608,
|
| 7 |
+
"train_samples_per_second": 1.477,
|
| 8 |
+
"train_steps_per_second": 0.023
|
| 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 sha256:8cc3c1b33def2af680389b5ce6915b6a7adf0179e5a76ab7bc3936de6b99d9dd
|
| 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 sha256:3bfa92f6339d97133e57d049aa167be3745a90b323fb7f5ef9cef720c12a8f10
|
| 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 sha256:947c2594c61ea8af6f42287d3775e3006a40a61d533e8347c22a20ec6863992a
|
| 3 |
size 4540516344
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.4091535410406212,
|
| 5 |
+
"train_runtime": 30868.5922,
|
| 6 |
+
"train_samples": 45608,
|
| 7 |
+
"train_samples_per_second": 1.477,
|
| 8 |
+
"train_steps_per_second": 0.023
|
| 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": 51.46674233848396,
|
| 14 |
+
"learning_rate": 6.9444444444444435e-09,
|
| 15 |
+
"logits/chosen": -3.328125,
|
| 16 |
+
"logits/rejected": -3.46875,
|
| 17 |
+
"logps/chosen": -468.0,
|
| 18 |
+
"logps/rejected": -300.0,
|
| 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": 0.6924511194229126,
|
| 33 |
+
"eval_rewards/accuracies": 0.21808511018753052,
|
| 34 |
+
"eval_rewards/chosen": 0.001068115234375,
|
| 35 |
+
"eval_rewards/margins": -0.0003376007080078125,
|
| 36 |
+
"eval_rewards/rejected": 0.00139617919921875,
|
| 37 |
+
"eval_runtime": 64.5741,
|
| 38 |
+
"eval_samples_per_second": 23.028,
|
| 39 |
+
"eval_steps_per_second": 0.728,
|
| 40 |
+
"step": 8
|
| 41 |
+
},
|
| 42 |
+
{
|
| 43 |
+
"epoch": 0.014025245441795231,
|
| 44 |
+
"grad_norm": 53.200557685453425,
|
| 45 |
+
"learning_rate": 6.944444444444444e-08,
|
| 46 |
+
"logits/chosen": -3.140625,
|
| 47 |
+
"logits/rejected": -3.21875,
|
| 48 |
+
"logps/chosen": -306.0,
|
| 49 |
+
"logps/rejected": -264.0,
|
| 50 |
+
"loss": 0.694,
|
| 51 |
+
"rewards/accuracies": 0.1666666716337204,
|
| 52 |
+
"rewards/chosen": -0.004241943359375,
|
| 53 |
+
"rewards/margins": -0.00970458984375,
|
| 54 |
+
"rewards/rejected": 0.0054931640625,
|
| 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": 0.6895634531974792,
|
| 64 |
+
"eval_rewards/accuracies": 0.2606382966041565,
|
| 65 |
+
"eval_rewards/chosen": -0.01318359375,
|
| 66 |
+
"eval_rewards/margins": -0.004058837890625,
|
| 67 |
+
"eval_rewards/rejected": -0.0091552734375,
|
| 68 |
+
"eval_runtime": 65.8066,
|
| 69 |
+
"eval_samples_per_second": 22.597,
|
| 70 |
+
"eval_steps_per_second": 0.714,
|
| 71 |
+
"step": 16
|
| 72 |
+
},
|
| 73 |
+
{
|
| 74 |
+
"epoch": 0.028050490883590462,
|
| 75 |
+
"grad_norm": 61.01600920733608,
|
| 76 |
+
"learning_rate": 1.3888888888888888e-07,
|
| 77 |
+
"logits/chosen": -3.171875,
|
| 78 |
+
"logits/rejected": -3.234375,
|
| 79 |
+
"logps/chosen": -314.0,
|
| 80 |
+
"logps/rejected": -294.0,
|
| 81 |
+
"loss": 0.6909,
|
| 82 |
+
"rewards/accuracies": 0.3125,
|
| 83 |
+
"rewards/chosen": -0.006256103515625,
|
| 84 |
+
"rewards/margins": 0.0089111328125,
|
| 85 |
+
"rewards/rejected": -0.01513671875,
|
| 86 |
+
"step": 20
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 0.033660589060308554,
|
| 90 |
+
"eval_logits/chosen": -3.234375,
|
| 91 |
+
"eval_logits/rejected": -3.265625,
|
| 92 |
+
"eval_logps/chosen": -322.0,
|
| 93 |
+
"eval_logps/rejected": -276.0,
|
| 94 |
+
"eval_loss": 0.6792454719543457,
|
| 95 |
+
"eval_rewards/accuracies": 0.48404255509376526,
|
| 96 |
+
"eval_rewards/chosen": -0.04248046875,
|
| 97 |
+
"eval_rewards/margins": 0.027587890625,
|
| 98 |
+
"eval_rewards/rejected": -0.0703125,
|
| 99 |
+
"eval_runtime": 65.5822,
|
| 100 |
+
"eval_samples_per_second": 22.674,
|
| 101 |
+
"eval_steps_per_second": 0.717,
|
| 102 |
+
"step": 24
|
| 103 |
+
},
|
| 104 |
+
{
|
| 105 |
+
"epoch": 0.04207573632538569,
|
| 106 |
+
"grad_norm": 48.23881467629644,
|
| 107 |
+
"learning_rate": 2.0833333333333333e-07,
|
| 108 |
+
"logits/chosen": -3.09375,
|
| 109 |
+
"logits/rejected": -3.25,
|
| 110 |
+
"logps/chosen": -352.0,
|
| 111 |
+
"logps/rejected": -280.0,
|
| 112 |
+
"loss": 0.6754,
|
| 113 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 114 |
+
"rewards/chosen": -0.038330078125,
|
| 115 |
+
"rewards/margins": 0.051513671875,
|
| 116 |
+
"rewards/rejected": -0.08984375,
|
| 117 |
+
"step": 30
|
| 118 |
+
},
|
| 119 |
+
{
|
| 120 |
+
"epoch": 0.04488078541374474,
|
| 121 |
+
"eval_logits/chosen": -3.21875,
|
| 122 |
+
"eval_logits/rejected": -3.25,
|
| 123 |
+
"eval_logps/chosen": -324.0,
|
| 124 |
+
"eval_logps/rejected": -280.0,
|
| 125 |
+
"eval_loss": 0.6562117338180542,
|
| 126 |
+
"eval_rewards/accuracies": 0.6117021441459656,
|
| 127 |
+
"eval_rewards/chosen": -0.111328125,
|
| 128 |
+
"eval_rewards/margins": 0.08544921875,
|
| 129 |
+
"eval_rewards/rejected": -0.197265625,
|
| 130 |
+
"eval_runtime": 65.2452,
|
| 131 |
+
"eval_samples_per_second": 22.791,
|
| 132 |
+
"eval_steps_per_second": 0.72,
|
| 133 |
+
"step": 32
|
| 134 |
+
},
|
| 135 |
+
{
|
| 136 |
+
"epoch": 0.056100981767180924,
|
| 137 |
+
"grad_norm": 47.782407457061694,
|
| 138 |
+
"learning_rate": 2.7777777777777776e-07,
|
| 139 |
+
"logits/chosen": -3.109375,
|
| 140 |
+
"logits/rejected": -3.25,
|
| 141 |
+
"logps/chosen": -346.0,
|
| 142 |
+
"logps/rejected": -276.0,
|
| 143 |
+
"loss": 0.6441,
|
| 144 |
+
"rewards/accuracies": 0.6875,
|
| 145 |
+
"rewards/chosen": -0.1259765625,
|
| 146 |
+
"rewards/margins": 0.1298828125,
|
| 147 |
+
"rewards/rejected": -0.255859375,
|
| 148 |
+
"step": 40
|
| 149 |
+
},
|
| 150 |
+
{
|
| 151 |
+
"epoch": 0.056100981767180924,
|
| 152 |
+
"eval_logits/chosen": -3.203125,
|
| 153 |
+
"eval_logits/rejected": -3.234375,
|
| 154 |
+
"eval_logps/chosen": -326.0,
|
| 155 |
+
"eval_logps/rejected": -284.0,
|
| 156 |
+
"eval_loss": 0.6230681538581848,
|
| 157 |
+
"eval_rewards/accuracies": 0.664893627166748,
|
| 158 |
+
"eval_rewards/chosen": -0.212890625,
|
| 159 |
+
"eval_rewards/margins": 0.1767578125,
|
| 160 |
+
"eval_rewards/rejected": -0.390625,
|
| 161 |
+
"eval_runtime": 65.4966,
|
| 162 |
+
"eval_samples_per_second": 22.703,
|
| 163 |
+
"eval_steps_per_second": 0.718,
|
| 164 |
+
"step": 40
|
| 165 |
+
},
|
| 166 |
+
{
|
| 167 |
+
"epoch": 0.06732117812061711,
|
| 168 |
+
"eval_logits/chosen": -3.1875,
|
| 169 |
+
"eval_logits/rejected": -3.21875,
|
| 170 |
+
"eval_logps/chosen": -328.0,
|
| 171 |
+
"eval_logps/rejected": -288.0,
|
| 172 |
+
"eval_loss": 0.5846173167228699,
|
| 173 |
+
"eval_rewards/accuracies": 0.6914893388748169,
|
| 174 |
+
"eval_rewards/chosen": -0.328125,
|
| 175 |
+
"eval_rewards/margins": 0.328125,
|
| 176 |
+
"eval_rewards/rejected": -0.65625,
|
| 177 |
+
"eval_runtime": 65.9802,
|
| 178 |
+
"eval_samples_per_second": 22.537,
|
| 179 |
+
"eval_steps_per_second": 0.712,
|
| 180 |
+
"step": 48
|
| 181 |
+
},
|
| 182 |
+
{
|
| 183 |
+
"epoch": 0.07012622720897616,
|
| 184 |
+
"grad_norm": 39.07068999911395,
|
| 185 |
+
"learning_rate": 3.472222222222222e-07,
|
| 186 |
+
"logits/chosen": -3.09375,
|
| 187 |
+
"logits/rejected": -3.078125,
|
| 188 |
+
"logps/chosen": -344.0,
|
| 189 |
+
"logps/rejected": -316.0,
|
| 190 |
+
"loss": 0.5952,
|
| 191 |
+
"rewards/accuracies": 0.762499988079071,
|
| 192 |
+
"rewards/chosen": -0.33203125,
|
| 193 |
+
"rewards/margins": 0.2578125,
|
| 194 |
+
"rewards/rejected": -0.58984375,
|
| 195 |
+
"step": 50
|
| 196 |
+
},
|
| 197 |
+
{
|
| 198 |
+
"epoch": 0.0785413744740533,
|
| 199 |
+
"eval_logits/chosen": -3.15625,
|
| 200 |
+
"eval_logits/rejected": -3.1875,
|
| 201 |
+
"eval_logps/chosen": -328.0,
|
| 202 |
+
"eval_logps/rejected": -292.0,
|
| 203 |
+
"eval_loss": 0.5460087060928345,
|
| 204 |
+
"eval_rewards/accuracies": 0.728723406791687,
|
| 205 |
+
"eval_rewards/chosen": -0.326171875,
|
| 206 |
+
"eval_rewards/margins": 0.498046875,
|
| 207 |
+
"eval_rewards/rejected": -0.82421875,
|
| 208 |
+
"eval_runtime": 65.6815,
|
| 209 |
+
"eval_samples_per_second": 22.64,
|
| 210 |
+
"eval_steps_per_second": 0.716,
|
| 211 |
+
"step": 56
|
| 212 |
+
},
|
| 213 |
+
{
|
| 214 |
+
"epoch": 0.08415147265077139,
|
| 215 |
+
"grad_norm": 35.960821270018556,
|
| 216 |
+
"learning_rate": 4.1666666666666667e-07,
|
| 217 |
+
"logits/chosen": -3.046875,
|
| 218 |
+
"logits/rejected": -3.109375,
|
| 219 |
+
"logps/chosen": -346.0,
|
| 220 |
+
"logps/rejected": -292.0,
|
| 221 |
+
"loss": 0.5601,
|
| 222 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 223 |
+
"rewards/chosen": -0.337890625,
|
| 224 |
+
"rewards/margins": 0.48828125,
|
| 225 |
+
"rewards/rejected": -0.828125,
|
| 226 |
+
"step": 60
|
| 227 |
+
},
|
| 228 |
+
{
|
| 229 |
+
"epoch": 0.08976157082748948,
|
| 230 |
+
"eval_logits/chosen": -3.140625,
|
| 231 |
+
"eval_logits/rejected": -3.171875,
|
| 232 |
+
"eval_logps/chosen": -324.0,
|
| 233 |
+
"eval_logps/rejected": -292.0,
|
| 234 |
+
"eval_loss": 0.5130031704902649,
|
| 235 |
+
"eval_rewards/accuracies": 0.7553191781044006,
|
| 236 |
+
"eval_rewards/chosen": -0.10791015625,
|
| 237 |
+
"eval_rewards/margins": 0.68359375,
|
| 238 |
+
"eval_rewards/rejected": -0.7890625,
|
| 239 |
+
"eval_runtime": 65.377,
|
| 240 |
+
"eval_samples_per_second": 22.745,
|
| 241 |
+
"eval_steps_per_second": 0.719,
|
| 242 |
+
"step": 64
|
| 243 |
+
},
|
| 244 |
+
{
|
| 245 |
+
"epoch": 0.09817671809256662,
|
| 246 |
+
"grad_norm": 45.987879679409986,
|
| 247 |
+
"learning_rate": 4.861111111111111e-07,
|
| 248 |
+
"logits/chosen": -2.921875,
|
| 249 |
+
"logits/rejected": -2.96875,
|
| 250 |
+
"logps/chosen": -340.0,
|
| 251 |
+
"logps/rejected": -318.0,
|
| 252 |
+
"loss": 0.5168,
|
| 253 |
+
"rewards/accuracies": 0.699999988079071,
|
| 254 |
+
"rewards/chosen": -0.384765625,
|
| 255 |
+
"rewards/margins": 0.515625,
|
| 256 |
+
"rewards/rejected": -0.8984375,
|
| 257 |
+
"step": 70
|
| 258 |
+
},
|
| 259 |
+
{
|
| 260 |
+
"epoch": 0.10098176718092566,
|
| 261 |
+
"eval_logits/chosen": -3.125,
|
| 262 |
+
"eval_logits/rejected": -3.15625,
|
| 263 |
+
"eval_logps/chosen": -330.0,
|
| 264 |
+
"eval_logps/rejected": -302.0,
|
| 265 |
+
"eval_loss": 0.48529163002967834,
|
| 266 |
+
"eval_rewards/accuracies": 0.7553191781044006,
|
| 267 |
+
"eval_rewards/chosen": -0.3671875,
|
| 268 |
+
"eval_rewards/margins": 0.9375,
|
| 269 |
+
"eval_rewards/rejected": -1.3046875,
|
| 270 |
+
"eval_runtime": 65.3763,
|
| 271 |
+
"eval_samples_per_second": 22.745,
|
| 272 |
+
"eval_steps_per_second": 0.719,
|
| 273 |
+
"step": 72
|
| 274 |
+
},
|
| 275 |
+
{
|
| 276 |
+
"epoch": 0.11220196353436185,
|
| 277 |
+
"grad_norm": 50.85101203918117,
|
| 278 |
+
"learning_rate": 4.998078598898921e-07,
|
| 279 |
+
"logits/chosen": -3.046875,
|
| 280 |
+
"logits/rejected": -3.09375,
|
| 281 |
+
"logps/chosen": -322.0,
|
| 282 |
+
"logps/rejected": -316.0,
|
| 283 |
+
"loss": 0.4973,
|
| 284 |
+
"rewards/accuracies": 0.824999988079071,
|
| 285 |
+
"rewards/chosen": -0.5703125,
|
| 286 |
+
"rewards/margins": 1.1015625,
|
| 287 |
+
"rewards/rejected": -1.671875,
|
| 288 |
+
"step": 80
|
| 289 |
+
},
|
| 290 |
+
{
|
| 291 |
+
"epoch": 0.11220196353436185,
|
| 292 |
+
"eval_logits/chosen": -3.125,
|
| 293 |
+
"eval_logits/rejected": -3.15625,
|
| 294 |
+
"eval_logps/chosen": -324.0,
|
| 295 |
+
"eval_logps/rejected": -298.0,
|
| 296 |
+
"eval_loss": 0.4733535349369049,
|
| 297 |
+
"eval_rewards/accuracies": 0.792553186416626,
|
| 298 |
+
"eval_rewards/chosen": -0.11181640625,
|
| 299 |
+
"eval_rewards/margins": 1.0703125,
|
| 300 |
+
"eval_rewards/rejected": -1.1796875,
|
| 301 |
+
"eval_runtime": 65.986,
|
| 302 |
+
"eval_samples_per_second": 22.535,
|
| 303 |
+
"eval_steps_per_second": 0.712,
|
| 304 |
+
"step": 80
|
| 305 |
+
},
|
| 306 |
+
{
|
| 307 |
+
"epoch": 0.12342215988779803,
|
| 308 |
+
"eval_logits/chosen": -3.125,
|
| 309 |
+
"eval_logits/rejected": -3.15625,
|
| 310 |
+
"eval_logps/chosen": -324.0,
|
| 311 |
+
"eval_logps/rejected": -300.0,
|
| 312 |
+
"eval_loss": 0.46471452713012695,
|
| 313 |
+
"eval_rewards/accuracies": 0.8031914830207825,
|
| 314 |
+
"eval_rewards/chosen": -0.1474609375,
|
| 315 |
+
"eval_rewards/margins": 1.1015625,
|
| 316 |
+
"eval_rewards/rejected": -1.25,
|
| 317 |
+
"eval_runtime": 65.8761,
|
| 318 |
+
"eval_samples_per_second": 22.573,
|
| 319 |
+
"eval_steps_per_second": 0.713,
|
| 320 |
+
"step": 88
|
| 321 |
+
},
|
| 322 |
+
{
|
| 323 |
+
"epoch": 0.12622720897615708,
|
| 324 |
+
"grad_norm": 43.69982337922237,
|
| 325 |
+
"learning_rate": 4.990277968429684e-07,
|
| 326 |
+
"logits/chosen": -3.046875,
|
| 327 |
+
"logits/rejected": -2.953125,
|
| 328 |
+
"logps/chosen": -354.0,
|
| 329 |
+
"logps/rejected": -312.0,
|
| 330 |
+
"loss": 0.4756,
|
| 331 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 332 |
+
"rewards/chosen": -0.11376953125,
|
| 333 |
+
"rewards/margins": 1.1484375,
|
| 334 |
+
"rewards/rejected": -1.2578125,
|
| 335 |
+
"step": 90
|
| 336 |
+
},
|
| 337 |
+
{
|
| 338 |
+
"epoch": 0.13464235624123422,
|
| 339 |
+
"eval_logits/chosen": -3.109375,
|
| 340 |
+
"eval_logits/rejected": -3.140625,
|
| 341 |
+
"eval_logps/chosen": -346.0,
|
| 342 |
+
"eval_logps/rejected": -326.0,
|
| 343 |
+
"eval_loss": 0.4726095497608185,
|
| 344 |
+
"eval_rewards/accuracies": 0.7659574747085571,
|
| 345 |
+
"eval_rewards/chosen": -1.2109375,
|
| 346 |
+
"eval_rewards/margins": 1.28125,
|
| 347 |
+
"eval_rewards/rejected": -2.5,
|
| 348 |
+
"eval_runtime": 66.1434,
|
| 349 |
+
"eval_samples_per_second": 22.481,
|
| 350 |
+
"eval_steps_per_second": 0.711,
|
| 351 |
+
"step": 96
|
| 352 |
+
},
|
| 353 |
+
{
|
| 354 |
+
"epoch": 0.1402524544179523,
|
| 355 |
+
"grad_norm": 65.26391444387855,
|
| 356 |
+
"learning_rate": 4.976496740424417e-07,
|
| 357 |
+
"logits/chosen": -3.078125,
|
| 358 |
+
"logits/rejected": -3.109375,
|
| 359 |
+
"logps/chosen": -330.0,
|
| 360 |
+
"logps/rejected": -302.0,
|
| 361 |
+
"loss": 0.4548,
|
| 362 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 363 |
+
"rewards/chosen": -0.98828125,
|
| 364 |
+
"rewards/margins": 1.3515625,
|
| 365 |
+
"rewards/rejected": -2.34375,
|
| 366 |
+
"step": 100
|
| 367 |
+
},
|
| 368 |
+
{
|
| 369 |
+
"epoch": 0.1458625525946704,
|
| 370 |
+
"eval_logits/chosen": -3.078125,
|
| 371 |
+
"eval_logits/rejected": -3.109375,
|
| 372 |
+
"eval_logps/chosen": -334.0,
|
| 373 |
+
"eval_logps/rejected": -316.0,
|
| 374 |
+
"eval_loss": 0.44555220007896423,
|
| 375 |
+
"eval_rewards/accuracies": 0.7872340679168701,
|
| 376 |
+
"eval_rewards/chosen": -0.6328125,
|
| 377 |
+
"eval_rewards/margins": 1.3828125,
|
| 378 |
+
"eval_rewards/rejected": -2.015625,
|
| 379 |
+
"eval_runtime": 65.9643,
|
| 380 |
+
"eval_samples_per_second": 22.542,
|
| 381 |
+
"eval_steps_per_second": 0.713,
|
| 382 |
+
"step": 104
|
| 383 |
+
},
|
| 384 |
+
{
|
| 385 |
+
"epoch": 0.15427769985974754,
|
| 386 |
+
"grad_norm": 43.09963797811936,
|
| 387 |
+
"learning_rate": 4.956768011581281e-07,
|
| 388 |
+
"logits/chosen": -2.984375,
|
| 389 |
+
"logits/rejected": -3.140625,
|
| 390 |
+
"logps/chosen": -356.0,
|
| 391 |
+
"logps/rejected": -326.0,
|
| 392 |
+
"loss": 0.451,
|
| 393 |
+
"rewards/accuracies": 0.862500011920929,
|
| 394 |
+
"rewards/chosen": -0.53515625,
|
| 395 |
+
"rewards/margins": 1.71875,
|
| 396 |
+
"rewards/rejected": -2.25,
|
| 397 |
+
"step": 110
|
| 398 |
+
},
|
| 399 |
+
{
|
| 400 |
+
"epoch": 0.1570827489481066,
|
| 401 |
+
"eval_logits/chosen": -3.046875,
|
| 402 |
+
"eval_logits/rejected": -3.078125,
|
| 403 |
+
"eval_logps/chosen": -326.0,
|
| 404 |
+
"eval_logps/rejected": -306.0,
|
| 405 |
+
"eval_loss": 0.4476098418235779,
|
| 406 |
+
"eval_rewards/accuracies": 0.813829779624939,
|
| 407 |
+
"eval_rewards/chosen": -0.25390625,
|
| 408 |
+
"eval_rewards/margins": 1.2734375,
|
| 409 |
+
"eval_rewards/rejected": -1.53125,
|
| 410 |
+
"eval_runtime": 65.5408,
|
| 411 |
+
"eval_samples_per_second": 22.688,
|
| 412 |
+
"eval_steps_per_second": 0.717,
|
| 413 |
+
"step": 112
|
| 414 |
+
},
|
| 415 |
+
{
|
| 416 |
+
"epoch": 0.16830294530154277,
|
| 417 |
+
"grad_norm": 35.62825840397104,
|
| 418 |
+
"learning_rate": 4.931139161987398e-07,
|
| 419 |
+
"logits/chosen": -2.953125,
|
| 420 |
+
"logits/rejected": -2.9375,
|
| 421 |
+
"logps/chosen": -366.0,
|
| 422 |
+
"logps/rejected": -360.0,
|
| 423 |
+
"loss": 0.4475,
|
| 424 |
+
"rewards/accuracies": 0.8125,
|
| 425 |
+
"rewards/chosen": -0.5625,
|
| 426 |
+
"rewards/margins": 1.4453125,
|
| 427 |
+
"rewards/rejected": -2.015625,
|
| 428 |
+
"step": 120
|
| 429 |
+
},
|
| 430 |
+
{
|
| 431 |
+
"epoch": 0.16830294530154277,
|
| 432 |
+
"eval_logits/chosen": -3.03125,
|
| 433 |
+
"eval_logits/rejected": -3.0625,
|
| 434 |
+
"eval_logps/chosen": -344.0,
|
| 435 |
+
"eval_logps/rejected": -330.0,
|
| 436 |
+
"eval_loss": 0.4360468089580536,
|
| 437 |
+
"eval_rewards/accuracies": 0.8191489577293396,
|
| 438 |
+
"eval_rewards/chosen": -1.15625,
|
| 439 |
+
"eval_rewards/margins": 1.546875,
|
| 440 |
+
"eval_rewards/rejected": -2.703125,
|
| 441 |
+
"eval_runtime": 67.0228,
|
| 442 |
+
"eval_samples_per_second": 22.186,
|
| 443 |
+
"eval_steps_per_second": 0.701,
|
| 444 |
+
"step": 120
|
| 445 |
+
},
|
| 446 |
+
{
|
| 447 |
+
"epoch": 0.17952314165497896,
|
| 448 |
+
"eval_logits/chosen": -3.03125,
|
| 449 |
+
"eval_logits/rejected": -3.078125,
|
| 450 |
+
"eval_logps/chosen": -338.0,
|
| 451 |
+
"eval_logps/rejected": -322.0,
|
| 452 |
+
"eval_loss": 0.4318736791610718,
|
| 453 |
+
"eval_rewards/accuracies": 0.8404255509376526,
|
| 454 |
+
"eval_rewards/chosen": -0.77734375,
|
| 455 |
+
"eval_rewards/margins": 1.578125,
|
| 456 |
+
"eval_rewards/rejected": -2.359375,
|
| 457 |
+
"eval_runtime": 65.8922,
|
| 458 |
+
"eval_samples_per_second": 22.567,
|
| 459 |
+
"eval_steps_per_second": 0.713,
|
| 460 |
+
"step": 128
|
| 461 |
+
},
|
| 462 |
+
{
|
| 463 |
+
"epoch": 0.182328190743338,
|
| 464 |
+
"grad_norm": 53.386726765592485,
|
| 465 |
+
"learning_rate": 4.89967174133187e-07,
|
| 466 |
+
"logits/chosen": -2.859375,
|
| 467 |
+
"logits/rejected": -2.984375,
|
| 468 |
+
"logps/chosen": -392.0,
|
| 469 |
+
"logps/rejected": -338.0,
|
| 470 |
+
"loss": 0.46,
|
| 471 |
+
"rewards/accuracies": 0.762499988079071,
|
| 472 |
+
"rewards/chosen": -1.2421875,
|
| 473 |
+
"rewards/margins": 1.609375,
|
| 474 |
+
"rewards/rejected": -2.859375,
|
| 475 |
+
"step": 130
|
| 476 |
+
},
|
| 477 |
+
{
|
| 478 |
+
"epoch": 0.19074333800841514,
|
| 479 |
+
"eval_logits/chosen": -3.046875,
|
| 480 |
+
"eval_logits/rejected": -3.078125,
|
| 481 |
+
"eval_logps/chosen": -334.0,
|
| 482 |
+
"eval_logps/rejected": -318.0,
|
| 483 |
+
"eval_loss": 0.427120566368103,
|
| 484 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 485 |
+
"eval_rewards/chosen": -0.58203125,
|
| 486 |
+
"eval_rewards/margins": 1.53125,
|
| 487 |
+
"eval_rewards/rejected": -2.109375,
|
| 488 |
+
"eval_runtime": 65.8368,
|
| 489 |
+
"eval_samples_per_second": 22.586,
|
| 490 |
+
"eval_steps_per_second": 0.714,
|
| 491 |
+
"step": 136
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.19635343618513323,
|
| 495 |
+
"grad_norm": 75.83519043403058,
|
| 496 |
+
"learning_rate": 4.862441321089378e-07,
|
| 497 |
+
"logits/chosen": -3.03125,
|
| 498 |
+
"logits/rejected": -3.140625,
|
| 499 |
+
"logps/chosen": -372.0,
|
| 500 |
+
"logps/rejected": -322.0,
|
| 501 |
+
"loss": 0.4198,
|
| 502 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 503 |
+
"rewards/chosen": -0.5859375,
|
| 504 |
+
"rewards/margins": 1.5546875,
|
| 505 |
+
"rewards/rejected": -2.140625,
|
| 506 |
+
"step": 140
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 0.20196353436185133,
|
| 510 |
+
"eval_logits/chosen": -3.046875,
|
| 511 |
+
"eval_logits/rejected": -3.078125,
|
| 512 |
+
"eval_logps/chosen": -336.0,
|
| 513 |
+
"eval_logps/rejected": -322.0,
|
| 514 |
+
"eval_loss": 0.4199545085430145,
|
| 515 |
+
"eval_rewards/accuracies": 0.8244680762290955,
|
| 516 |
+
"eval_rewards/chosen": -0.67578125,
|
| 517 |
+
"eval_rewards/margins": 1.6953125,
|
| 518 |
+
"eval_rewards/rejected": -2.375,
|
| 519 |
+
"eval_runtime": 65.5367,
|
| 520 |
+
"eval_samples_per_second": 22.69,
|
| 521 |
+
"eval_steps_per_second": 0.717,
|
| 522 |
+
"step": 144
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 0.21037868162692847,
|
| 526 |
+
"grad_norm": 53.861403761625574,
|
| 527 |
+
"learning_rate": 4.819537313029364e-07,
|
| 528 |
+
"logits/chosen": -3.03125,
|
| 529 |
+
"logits/rejected": -3.046875,
|
| 530 |
+
"logps/chosen": -314.0,
|
| 531 |
+
"logps/rejected": -322.0,
|
| 532 |
+
"loss": 0.3987,
|
| 533 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 534 |
+
"rewards/chosen": -0.7890625,
|
| 535 |
+
"rewards/margins": 1.78125,
|
| 536 |
+
"rewards/rejected": -2.5625,
|
| 537 |
+
"step": 150
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.2131837307152875,
|
| 541 |
+
"eval_logits/chosen": -3.03125,
|
| 542 |
+
"eval_logits/rejected": -3.0625,
|
| 543 |
+
"eval_logps/chosen": -338.0,
|
| 544 |
+
"eval_logps/rejected": -326.0,
|
| 545 |
+
"eval_loss": 0.417427122592926,
|
| 546 |
+
"eval_rewards/accuracies": 0.8457446694374084,
|
| 547 |
+
"eval_rewards/chosen": -0.79296875,
|
| 548 |
+
"eval_rewards/margins": 1.75,
|
| 549 |
+
"eval_rewards/rejected": -2.546875,
|
| 550 |
+
"eval_runtime": 65.4524,
|
| 551 |
+
"eval_samples_per_second": 22.719,
|
| 552 |
+
"eval_steps_per_second": 0.718,
|
| 553 |
+
"step": 152
|
| 554 |
+
},
|
| 555 |
+
{
|
| 556 |
+
"epoch": 0.2244039270687237,
|
| 557 |
+
"grad_norm": 76.00085050984279,
|
| 558 |
+
"learning_rate": 4.771062754486677e-07,
|
| 559 |
+
"logits/chosen": -3.046875,
|
| 560 |
+
"logits/rejected": -3.0625,
|
| 561 |
+
"logps/chosen": -350.0,
|
| 562 |
+
"logps/rejected": -316.0,
|
| 563 |
+
"loss": 0.4084,
|
| 564 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 565 |
+
"rewards/chosen": -0.69921875,
|
| 566 |
+
"rewards/margins": 1.921875,
|
| 567 |
+
"rewards/rejected": -2.625,
|
| 568 |
+
"step": 160
|
| 569 |
+
},
|
| 570 |
+
{
|
| 571 |
+
"epoch": 0.2244039270687237,
|
| 572 |
+
"eval_logits/chosen": -3.0625,
|
| 573 |
+
"eval_logits/rejected": -3.09375,
|
| 574 |
+
"eval_logps/chosen": -330.0,
|
| 575 |
+
"eval_logps/rejected": -320.0,
|
| 576 |
+
"eval_loss": 0.4198690950870514,
|
| 577 |
+
"eval_rewards/accuracies": 0.8404255509376526,
|
| 578 |
+
"eval_rewards/chosen": -0.451171875,
|
| 579 |
+
"eval_rewards/margins": 1.828125,
|
| 580 |
+
"eval_rewards/rejected": -2.28125,
|
| 581 |
+
"eval_runtime": 65.9805,
|
| 582 |
+
"eval_samples_per_second": 22.537,
|
| 583 |
+
"eval_steps_per_second": 0.712,
|
| 584 |
+
"step": 160
|
| 585 |
+
},
|
| 586 |
+
{
|
| 587 |
+
"epoch": 0.23562412342215988,
|
| 588 |
+
"eval_logits/chosen": -3.0625,
|
| 589 |
+
"eval_logits/rejected": -3.09375,
|
| 590 |
+
"eval_logps/chosen": -340.0,
|
| 591 |
+
"eval_logps/rejected": -332.0,
|
| 592 |
+
"eval_loss": 0.41553688049316406,
|
| 593 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 594 |
+
"eval_rewards/chosen": -0.9296875,
|
| 595 |
+
"eval_rewards/margins": 1.9296875,
|
| 596 |
+
"eval_rewards/rejected": -2.859375,
|
| 597 |
+
"eval_runtime": 66.758,
|
| 598 |
+
"eval_samples_per_second": 22.274,
|
| 599 |
+
"eval_steps_per_second": 0.704,
|
| 600 |
+
"step": 168
|
| 601 |
+
},
|
| 602 |
+
{
|
| 603 |
+
"epoch": 0.23842917251051893,
|
| 604 |
+
"grad_norm": 81.43616859600893,
|
| 605 |
+
"learning_rate": 4.717134060909331e-07,
|
| 606 |
+
"logits/chosen": -2.9375,
|
| 607 |
+
"logits/rejected": -2.75,
|
| 608 |
+
"logps/chosen": -300.0,
|
| 609 |
+
"logps/rejected": -308.0,
|
| 610 |
+
"loss": 0.4194,
|
| 611 |
+
"rewards/accuracies": 0.887499988079071,
|
| 612 |
+
"rewards/chosen": -1.0,
|
| 613 |
+
"rewards/margins": 2.296875,
|
| 614 |
+
"rewards/rejected": -3.296875,
|
| 615 |
+
"step": 170
|
| 616 |
+
},
|
| 617 |
+
{
|
| 618 |
+
"epoch": 0.24684431977559607,
|
| 619 |
+
"eval_logits/chosen": -3.0625,
|
| 620 |
+
"eval_logits/rejected": -3.109375,
|
| 621 |
+
"eval_logps/chosen": -326.0,
|
| 622 |
+
"eval_logps/rejected": -312.0,
|
| 623 |
+
"eval_loss": 0.4193520247936249,
|
| 624 |
+
"eval_rewards/accuracies": 0.813829779624939,
|
| 625 |
+
"eval_rewards/chosen": -0.2197265625,
|
| 626 |
+
"eval_rewards/margins": 1.6484375,
|
| 627 |
+
"eval_rewards/rejected": -1.8671875,
|
| 628 |
+
"eval_runtime": 67.0437,
|
| 629 |
+
"eval_samples_per_second": 22.18,
|
| 630 |
+
"eval_steps_per_second": 0.701,
|
| 631 |
+
"step": 176
|
| 632 |
+
},
|
| 633 |
+
{
|
| 634 |
+
"epoch": 0.25245441795231416,
|
| 635 |
+
"grad_norm": 41.336420518994295,
|
| 636 |
+
"learning_rate": 4.6578807462777004e-07,
|
| 637 |
+
"logits/chosen": -3.078125,
|
| 638 |
+
"logits/rejected": -3.015625,
|
| 639 |
+
"logps/chosen": -298.0,
|
| 640 |
+
"logps/rejected": -352.0,
|
| 641 |
+
"loss": 0.4338,
|
| 642 |
+
"rewards/accuracies": 0.875,
|
| 643 |
+
"rewards/chosen": -0.6015625,
|
| 644 |
+
"rewards/margins": 1.5234375,
|
| 645 |
+
"rewards/rejected": -2.125,
|
| 646 |
+
"step": 180
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"epoch": 0.25806451612903225,
|
| 650 |
+
"eval_logits/chosen": -3.0625,
|
| 651 |
+
"eval_logits/rejected": -3.109375,
|
| 652 |
+
"eval_logps/chosen": -322.0,
|
| 653 |
+
"eval_logps/rejected": -306.0,
|
| 654 |
+
"eval_loss": 0.4271917939186096,
|
| 655 |
+
"eval_rewards/accuracies": 0.8191489577293396,
|
| 656 |
+
"eval_rewards/chosen": -0.0283203125,
|
| 657 |
+
"eval_rewards/margins": 1.5234375,
|
| 658 |
+
"eval_rewards/rejected": -1.546875,
|
| 659 |
+
"eval_runtime": 65.8831,
|
| 660 |
+
"eval_samples_per_second": 22.57,
|
| 661 |
+
"eval_steps_per_second": 0.713,
|
| 662 |
+
"step": 184
|
| 663 |
+
},
|
| 664 |
+
{
|
| 665 |
+
"epoch": 0.2664796633941094,
|
| 666 |
+
"grad_norm": 32.121143393496304,
|
| 667 |
+
"learning_rate": 4.593445112066553e-07,
|
| 668 |
+
"logits/chosen": -3.015625,
|
| 669 |
+
"logits/rejected": -3.09375,
|
| 670 |
+
"logps/chosen": -394.0,
|
| 671 |
+
"logps/rejected": -354.0,
|
| 672 |
+
"loss": 0.415,
|
| 673 |
+
"rewards/accuracies": 0.875,
|
| 674 |
+
"rewards/chosen": 0.0849609375,
|
| 675 |
+
"rewards/margins": 1.84375,
|
| 676 |
+
"rewards/rejected": -1.7578125,
|
| 677 |
+
"step": 190
|
| 678 |
+
},
|
| 679 |
+
{
|
| 680 |
+
"epoch": 0.26928471248246844,
|
| 681 |
+
"eval_logits/chosen": -3.0625,
|
| 682 |
+
"eval_logits/rejected": -3.109375,
|
| 683 |
+
"eval_logps/chosen": -338.0,
|
| 684 |
+
"eval_logps/rejected": -326.0,
|
| 685 |
+
"eval_loss": 0.4112021028995514,
|
| 686 |
+
"eval_rewards/accuracies": 0.8297872543334961,
|
| 687 |
+
"eval_rewards/chosen": -0.84765625,
|
| 688 |
+
"eval_rewards/margins": 1.7265625,
|
| 689 |
+
"eval_rewards/rejected": -2.578125,
|
| 690 |
+
"eval_runtime": 65.4321,
|
| 691 |
+
"eval_samples_per_second": 22.726,
|
| 692 |
+
"eval_steps_per_second": 0.718,
|
| 693 |
+
"step": 192
|
| 694 |
+
},
|
| 695 |
+
{
|
| 696 |
+
"epoch": 0.2805049088359046,
|
| 697 |
+
"grad_norm": 128.88104657311357,
|
| 698 |
+
"learning_rate": 4.523981905496914e-07,
|
| 699 |
+
"logits/chosen": -2.859375,
|
| 700 |
+
"logits/rejected": -2.9375,
|
| 701 |
+
"logps/chosen": -332.0,
|
| 702 |
+
"logps/rejected": -342.0,
|
| 703 |
+
"loss": 0.4038,
|
| 704 |
+
"rewards/accuracies": 0.925000011920929,
|
| 705 |
+
"rewards/chosen": -0.8203125,
|
| 706 |
+
"rewards/margins": 2.15625,
|
| 707 |
+
"rewards/rejected": -2.984375,
|
| 708 |
+
"step": 200
|
| 709 |
+
},
|
| 710 |
+
{
|
| 711 |
+
"epoch": 0.2805049088359046,
|
| 712 |
+
"eval_logits/chosen": -3.078125,
|
| 713 |
+
"eval_logits/rejected": -3.109375,
|
| 714 |
+
"eval_logps/chosen": -344.0,
|
| 715 |
+
"eval_logps/rejected": -334.0,
|
| 716 |
+
"eval_loss": 0.4111826717853546,
|
| 717 |
+
"eval_rewards/accuracies": 0.792553186416626,
|
| 718 |
+
"eval_rewards/chosen": -1.109375,
|
| 719 |
+
"eval_rewards/margins": 1.8125,
|
| 720 |
+
"eval_rewards/rejected": -2.921875,
|
| 721 |
+
"eval_runtime": 65.5119,
|
| 722 |
+
"eval_samples_per_second": 22.698,
|
| 723 |
+
"eval_steps_per_second": 0.717,
|
| 724 |
+
"step": 200
|
| 725 |
+
},
|
| 726 |
+
{
|
| 727 |
+
"epoch": 0.2917251051893408,
|
| 728 |
+
"eval_logits/chosen": -3.09375,
|
| 729 |
+
"eval_logits/rejected": -3.125,
|
| 730 |
+
"eval_logps/chosen": -326.0,
|
| 731 |
+
"eval_logps/rejected": -314.0,
|
| 732 |
+
"eval_loss": 0.41739681363105774,
|
| 733 |
+
"eval_rewards/accuracies": 0.8191489577293396,
|
| 734 |
+
"eval_rewards/chosen": -0.173828125,
|
| 735 |
+
"eval_rewards/margins": 1.7421875,
|
| 736 |
+
"eval_rewards/rejected": -1.9140625,
|
| 737 |
+
"eval_runtime": 65.6569,
|
| 738 |
+
"eval_samples_per_second": 22.648,
|
| 739 |
+
"eval_steps_per_second": 0.716,
|
| 740 |
+
"step": 208
|
| 741 |
+
},
|
| 742 |
+
{
|
| 743 |
+
"epoch": 0.29453015427769985,
|
| 744 |
+
"grad_norm": 33.6284095972861,
|
| 745 |
+
"learning_rate": 4.4496579478984995e-07,
|
| 746 |
+
"logits/chosen": -3.0,
|
| 747 |
+
"logits/rejected": -2.96875,
|
| 748 |
+
"logps/chosen": -308.0,
|
| 749 |
+
"logps/rejected": -296.0,
|
| 750 |
+
"loss": 0.4119,
|
| 751 |
+
"rewards/accuracies": 0.824999988079071,
|
| 752 |
+
"rewards/chosen": -0.91015625,
|
| 753 |
+
"rewards/margins": 1.84375,
|
| 754 |
+
"rewards/rejected": -2.75,
|
| 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": -330.0,
|
| 762 |
+
"eval_logps/rejected": -320.0,
|
| 763 |
+
"eval_loss": 0.4090508818626404,
|
| 764 |
+
"eval_rewards/accuracies": 0.8191489577293396,
|
| 765 |
+
"eval_rewards/chosen": -0.45703125,
|
| 766 |
+
"eval_rewards/margins": 1.796875,
|
| 767 |
+
"eval_rewards/rejected": -2.25,
|
| 768 |
+
"eval_runtime": 65.408,
|
| 769 |
+
"eval_samples_per_second": 22.734,
|
| 770 |
+
"eval_steps_per_second": 0.719,
|
| 771 |
+
"step": 216
|
| 772 |
+
},
|
| 773 |
+
{
|
| 774 |
+
"epoch": 0.3085553997194951,
|
| 775 |
+
"grad_norm": 50.722207221810784,
|
| 776 |
+
"learning_rate": 4.370651734075229e-07,
|
| 777 |
+
"logits/chosen": -2.890625,
|
| 778 |
+
"logits/rejected": -2.859375,
|
| 779 |
+
"logps/chosen": -352.0,
|
| 780 |
+
"logps/rejected": -356.0,
|
| 781 |
+
"loss": 0.3826,
|
| 782 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 783 |
+
"rewards/chosen": -0.62109375,
|
| 784 |
+
"rewards/margins": 1.953125,
|
| 785 |
+
"rewards/rejected": -2.578125,
|
| 786 |
+
"step": 220
|
| 787 |
+
},
|
| 788 |
+
{
|
| 789 |
+
"epoch": 0.3141654978962132,
|
| 790 |
+
"eval_logits/chosen": -3.109375,
|
| 791 |
+
"eval_logits/rejected": -3.140625,
|
| 792 |
+
"eval_logps/chosen": -352.0,
|
| 793 |
+
"eval_logps/rejected": -342.0,
|
| 794 |
+
"eval_loss": 0.41070857644081116,
|
| 795 |
+
"eval_rewards/accuracies": 0.8031914830207825,
|
| 796 |
+
"eval_rewards/chosen": -1.5,
|
| 797 |
+
"eval_rewards/margins": 1.8125,
|
| 798 |
+
"eval_rewards/rejected": -3.3125,
|
| 799 |
+
"eval_runtime": 65.3108,
|
| 800 |
+
"eval_samples_per_second": 22.768,
|
| 801 |
+
"eval_steps_per_second": 0.72,
|
| 802 |
+
"step": 224
|
| 803 |
+
},
|
| 804 |
+
{
|
| 805 |
+
"epoch": 0.3225806451612903,
|
| 806 |
+
"grad_norm": 30.84942010137543,
|
| 807 |
+
"learning_rate": 4.2871530036359783e-07,
|
| 808 |
+
"logits/chosen": -3.15625,
|
| 809 |
+
"logits/rejected": -3.15625,
|
| 810 |
+
"logps/chosen": -348.0,
|
| 811 |
+
"logps/rejected": -336.0,
|
| 812 |
+
"loss": 0.3878,
|
| 813 |
+
"rewards/accuracies": 0.887499988079071,
|
| 814 |
+
"rewards/chosen": -1.671875,
|
| 815 |
+
"rewards/margins": 2.125,
|
| 816 |
+
"rewards/rejected": -3.796875,
|
| 817 |
+
"step": 230
|
| 818 |
+
},
|
| 819 |
+
{
|
| 820 |
+
"epoch": 0.32538569424964936,
|
| 821 |
+
"eval_logits/chosen": -3.109375,
|
| 822 |
+
"eval_logits/rejected": -3.140625,
|
| 823 |
+
"eval_logps/chosen": -346.0,
|
| 824 |
+
"eval_logps/rejected": -338.0,
|
| 825 |
+
"eval_loss": 0.409851610660553,
|
| 826 |
+
"eval_rewards/accuracies": 0.8244680762290955,
|
| 827 |
+
"eval_rewards/chosen": -1.2265625,
|
| 828 |
+
"eval_rewards/margins": 1.8984375,
|
| 829 |
+
"eval_rewards/rejected": -3.125,
|
| 830 |
+
"eval_runtime": 65.7782,
|
| 831 |
+
"eval_samples_per_second": 22.606,
|
| 832 |
+
"eval_steps_per_second": 0.715,
|
| 833 |
+
"step": 232
|
| 834 |
+
},
|
| 835 |
+
{
|
| 836 |
+
"epoch": 0.33660589060308554,
|
| 837 |
+
"grad_norm": 69.24941711385972,
|
| 838 |
+
"learning_rate": 4.1993622853200526e-07,
|
| 839 |
+
"logits/chosen": -3.078125,
|
| 840 |
+
"logits/rejected": -3.078125,
|
| 841 |
+
"logps/chosen": -316.0,
|
| 842 |
+
"logps/rejected": -336.0,
|
| 843 |
+
"loss": 0.3852,
|
| 844 |
+
"rewards/accuracies": 0.824999988079071,
|
| 845 |
+
"rewards/chosen": -1.3203125,
|
| 846 |
+
"rewards/margins": 1.953125,
|
| 847 |
+
"rewards/rejected": -3.265625,
|
| 848 |
+
"step": 240
|
| 849 |
+
},
|
| 850 |
+
{
|
| 851 |
+
"epoch": 0.33660589060308554,
|
| 852 |
+
"eval_logits/chosen": -3.125,
|
| 853 |
+
"eval_logits/rejected": -3.15625,
|
| 854 |
+
"eval_logps/chosen": -336.0,
|
| 855 |
+
"eval_logps/rejected": -328.0,
|
| 856 |
+
"eval_loss": 0.40991342067718506,
|
| 857 |
+
"eval_rewards/accuracies": 0.8244680762290955,
|
| 858 |
+
"eval_rewards/chosen": -0.69140625,
|
| 859 |
+
"eval_rewards/margins": 1.9453125,
|
| 860 |
+
"eval_rewards/rejected": -2.640625,
|
| 861 |
+
"eval_runtime": 65.801,
|
| 862 |
+
"eval_samples_per_second": 22.598,
|
| 863 |
+
"eval_steps_per_second": 0.714,
|
| 864 |
+
"step": 240
|
| 865 |
+
},
|
| 866 |
+
{
|
| 867 |
+
"epoch": 0.34782608695652173,
|
| 868 |
+
"eval_logits/chosen": -3.109375,
|
| 869 |
+
"eval_logits/rejected": -3.140625,
|
| 870 |
+
"eval_logps/chosen": -340.0,
|
| 871 |
+
"eval_logps/rejected": -330.0,
|
| 872 |
+
"eval_loss": 0.4066612720489502,
|
| 873 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 874 |
+
"eval_rewards/chosen": -0.87890625,
|
| 875 |
+
"eval_rewards/margins": 1.8828125,
|
| 876 |
+
"eval_rewards/rejected": -2.765625,
|
| 877 |
+
"eval_runtime": 65.3251,
|
| 878 |
+
"eval_samples_per_second": 22.763,
|
| 879 |
+
"eval_steps_per_second": 0.719,
|
| 880 |
+
"step": 248
|
| 881 |
+
},
|
| 882 |
+
{
|
| 883 |
+
"epoch": 0.3506311360448808,
|
| 884 |
+
"grad_norm": 47.10180307111659,
|
| 885 |
+
"learning_rate": 4.107490415411714e-07,
|
| 886 |
+
"logits/chosen": -3.03125,
|
| 887 |
+
"logits/rejected": -3.078125,
|
| 888 |
+
"logps/chosen": -422.0,
|
| 889 |
+
"logps/rejected": -310.0,
|
| 890 |
+
"loss": 0.4338,
|
| 891 |
+
"rewards/accuracies": 0.800000011920929,
|
| 892 |
+
"rewards/chosen": -1.09375,
|
| 893 |
+
"rewards/margins": 1.515625,
|
| 894 |
+
"rewards/rejected": -2.609375,
|
| 895 |
+
"step": 250
|
| 896 |
+
},
|
| 897 |
+
{
|
| 898 |
+
"epoch": 0.3590462833099579,
|
| 899 |
+
"eval_logits/chosen": -3.09375,
|
| 900 |
+
"eval_logits/rejected": -3.140625,
|
| 901 |
+
"eval_logps/chosen": -342.0,
|
| 902 |
+
"eval_logps/rejected": -334.0,
|
| 903 |
+
"eval_loss": 0.4039236903190613,
|
| 904 |
+
"eval_rewards/accuracies": 0.8244680762290955,
|
| 905 |
+
"eval_rewards/chosen": -1.0390625,
|
| 906 |
+
"eval_rewards/margins": 1.9296875,
|
| 907 |
+
"eval_rewards/rejected": -2.96875,
|
| 908 |
+
"eval_runtime": 65.7057,
|
| 909 |
+
"eval_samples_per_second": 22.631,
|
| 910 |
+
"eval_steps_per_second": 0.715,
|
| 911 |
+
"step": 256
|
| 912 |
+
},
|
| 913 |
+
{
|
| 914 |
+
"epoch": 0.364656381486676,
|
| 915 |
+
"grad_norm": 37.27672395240747,
|
| 916 |
+
"learning_rate": 4.01175803140034e-07,
|
| 917 |
+
"logits/chosen": -3.09375,
|
| 918 |
+
"logits/rejected": -3.140625,
|
| 919 |
+
"logps/chosen": -370.0,
|
| 920 |
+
"logps/rejected": -332.0,
|
| 921 |
+
"loss": 0.3974,
|
| 922 |
+
"rewards/accuracies": 0.887499988079071,
|
| 923 |
+
"rewards/chosen": -1.328125,
|
| 924 |
+
"rewards/margins": 2.265625,
|
| 925 |
+
"rewards/rejected": -3.59375,
|
| 926 |
+
"step": 260
|
| 927 |
+
},
|
| 928 |
+
{
|
| 929 |
+
"epoch": 0.3702664796633941,
|
| 930 |
+
"eval_logits/chosen": -3.09375,
|
| 931 |
+
"eval_logits/rejected": -3.125,
|
| 932 |
+
"eval_logps/chosen": -336.0,
|
| 933 |
+
"eval_logps/rejected": -328.0,
|
| 934 |
+
"eval_loss": 0.40224307775497437,
|
| 935 |
+
"eval_rewards/accuracies": 0.8191489577293396,
|
| 936 |
+
"eval_rewards/chosen": -0.72265625,
|
| 937 |
+
"eval_rewards/margins": 1.875,
|
| 938 |
+
"eval_rewards/rejected": -2.59375,
|
| 939 |
+
"eval_runtime": 65.6767,
|
| 940 |
+
"eval_samples_per_second": 22.641,
|
| 941 |
+
"eval_steps_per_second": 0.716,
|
| 942 |
+
"step": 264
|
| 943 |
+
},
|
| 944 |
+
{
|
| 945 |
+
"epoch": 0.37868162692847124,
|
| 946 |
+
"grad_norm": 32.86123128156331,
|
| 947 |
+
"learning_rate": 3.9123950421022135e-07,
|
| 948 |
+
"logits/chosen": -3.09375,
|
| 949 |
+
"logits/rejected": -3.015625,
|
| 950 |
+
"logps/chosen": -326.0,
|
| 951 |
+
"logps/rejected": -328.0,
|
| 952 |
+
"loss": 0.4019,
|
| 953 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 954 |
+
"rewards/chosen": -0.921875,
|
| 955 |
+
"rewards/margins": 1.90625,
|
| 956 |
+
"rewards/rejected": -2.828125,
|
| 957 |
+
"step": 270
|
| 958 |
+
},
|
| 959 |
+
{
|
| 960 |
+
"epoch": 0.3814866760168303,
|
| 961 |
+
"eval_logits/chosen": -3.09375,
|
| 962 |
+
"eval_logits/rejected": -3.125,
|
| 963 |
+
"eval_logps/chosen": -332.0,
|
| 964 |
+
"eval_logps/rejected": -320.0,
|
| 965 |
+
"eval_loss": 0.40509188175201416,
|
| 966 |
+
"eval_rewards/accuracies": 0.8457446694374084,
|
| 967 |
+
"eval_rewards/chosen": -0.5546875,
|
| 968 |
+
"eval_rewards/margins": 1.7109375,
|
| 969 |
+
"eval_rewards/rejected": -2.265625,
|
| 970 |
+
"eval_runtime": 65.6219,
|
| 971 |
+
"eval_samples_per_second": 22.66,
|
| 972 |
+
"eval_steps_per_second": 0.716,
|
| 973 |
+
"step": 272
|
| 974 |
+
},
|
| 975 |
+
{
|
| 976 |
+
"epoch": 0.39270687237026647,
|
| 977 |
+
"grad_norm": 34.91842459517312,
|
| 978 |
+
"learning_rate": 3.8096400755164976e-07,
|
| 979 |
+
"logits/chosen": -3.0625,
|
| 980 |
+
"logits/rejected": -3.03125,
|
| 981 |
+
"logps/chosen": -340.0,
|
| 982 |
+
"logps/rejected": -328.0,
|
| 983 |
+
"loss": 0.3903,
|
| 984 |
+
"rewards/accuracies": 0.887499988079071,
|
| 985 |
+
"rewards/chosen": -0.984375,
|
| 986 |
+
"rewards/margins": 1.7265625,
|
| 987 |
+
"rewards/rejected": -2.703125,
|
| 988 |
+
"step": 280
|
| 989 |
+
},
|
| 990 |
+
{
|
| 991 |
+
"epoch": 0.39270687237026647,
|
| 992 |
+
"eval_logits/chosen": -3.09375,
|
| 993 |
+
"eval_logits/rejected": -3.125,
|
| 994 |
+
"eval_logps/chosen": -338.0,
|
| 995 |
+
"eval_logps/rejected": -330.0,
|
| 996 |
+
"eval_loss": 0.4004907011985779,
|
| 997 |
+
"eval_rewards/accuracies": 0.8617021441459656,
|
| 998 |
+
"eval_rewards/chosen": -0.84765625,
|
| 999 |
+
"eval_rewards/margins": 1.875,
|
| 1000 |
+
"eval_rewards/rejected": -2.71875,
|
| 1001 |
+
"eval_runtime": 66.0203,
|
| 1002 |
+
"eval_samples_per_second": 22.523,
|
| 1003 |
+
"eval_steps_per_second": 0.712,
|
| 1004 |
+
"step": 280
|
| 1005 |
+
},
|
| 1006 |
+
{
|
| 1007 |
+
"epoch": 0.40392706872370265,
|
| 1008 |
+
"eval_logits/chosen": -3.078125,
|
| 1009 |
+
"eval_logits/rejected": -3.109375,
|
| 1010 |
+
"eval_logps/chosen": -348.0,
|
| 1011 |
+
"eval_logps/rejected": -340.0,
|
| 1012 |
+
"eval_loss": 0.3978319466114044,
|
| 1013 |
+
"eval_rewards/accuracies": 0.8563829660415649,
|
| 1014 |
+
"eval_rewards/chosen": -1.265625,
|
| 1015 |
+
"eval_rewards/margins": 2.0,
|
| 1016 |
+
"eval_rewards/rejected": -3.265625,
|
| 1017 |
+
"eval_runtime": 65.7904,
|
| 1018 |
+
"eval_samples_per_second": 22.602,
|
| 1019 |
+
"eval_steps_per_second": 0.714,
|
| 1020 |
+
"step": 288
|
| 1021 |
+
},
|
| 1022 |
+
{
|
| 1023 |
+
"epoch": 0.4067321178120617,
|
| 1024 |
+
"grad_norm": 48.315535907518516,
|
| 1025 |
+
"learning_rate": 3.7037399057414135e-07,
|
| 1026 |
+
"logits/chosen": -2.890625,
|
| 1027 |
+
"logits/rejected": -2.96875,
|
| 1028 |
+
"logps/chosen": -314.0,
|
| 1029 |
+
"logps/rejected": -332.0,
|
| 1030 |
+
"loss": 0.3948,
|
| 1031 |
+
"rewards/accuracies": 0.9375,
|
| 1032 |
+
"rewards/chosen": -1.2421875,
|
| 1033 |
+
"rewards/margins": 2.15625,
|
| 1034 |
+
"rewards/rejected": -3.40625,
|
| 1035 |
+
"step": 290
|
| 1036 |
+
},
|
| 1037 |
+
{
|
| 1038 |
+
"epoch": 0.41514726507713884,
|
| 1039 |
+
"eval_logits/chosen": -3.09375,
|
| 1040 |
+
"eval_logits/rejected": -3.125,
|
| 1041 |
+
"eval_logps/chosen": -334.0,
|
| 1042 |
+
"eval_logps/rejected": -326.0,
|
| 1043 |
+
"eval_loss": 0.39696547389030457,
|
| 1044 |
+
"eval_rewards/accuracies": 0.8297872543334961,
|
| 1045 |
+
"eval_rewards/chosen": -0.65234375,
|
| 1046 |
+
"eval_rewards/margins": 1.90625,
|
| 1047 |
+
"eval_rewards/rejected": -2.5625,
|
| 1048 |
+
"eval_runtime": 65.6246,
|
| 1049 |
+
"eval_samples_per_second": 22.659,
|
| 1050 |
+
"eval_steps_per_second": 0.716,
|
| 1051 |
+
"step": 296
|
| 1052 |
+
},
|
| 1053 |
+
{
|
| 1054 |
+
"epoch": 0.42075736325385693,
|
| 1055 |
+
"grad_norm": 31.56207798451172,
|
| 1056 |
+
"learning_rate": 3.594948860326918e-07,
|
| 1057 |
+
"logits/chosen": -3.078125,
|
| 1058 |
+
"logits/rejected": -3.09375,
|
| 1059 |
+
"logps/chosen": -304.0,
|
| 1060 |
+
"logps/rejected": -314.0,
|
| 1061 |
+
"loss": 0.3894,
|
| 1062 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1063 |
+
"rewards/chosen": -0.8203125,
|
| 1064 |
+
"rewards/margins": 1.9453125,
|
| 1065 |
+
"rewards/rejected": -2.765625,
|
| 1066 |
+
"step": 300
|
| 1067 |
+
},
|
| 1068 |
+
{
|
| 1069 |
+
"epoch": 0.426367461430575,
|
| 1070 |
+
"eval_logits/chosen": -3.078125,
|
| 1071 |
+
"eval_logits/rejected": -3.109375,
|
| 1072 |
+
"eval_logps/chosen": -344.0,
|
| 1073 |
+
"eval_logps/rejected": -334.0,
|
| 1074 |
+
"eval_loss": 0.39327114820480347,
|
| 1075 |
+
"eval_rewards/accuracies": 0.8404255509376526,
|
| 1076 |
+
"eval_rewards/chosen": -1.109375,
|
| 1077 |
+
"eval_rewards/margins": 1.8671875,
|
| 1078 |
+
"eval_rewards/rejected": -2.984375,
|
| 1079 |
+
"eval_runtime": 66.1549,
|
| 1080 |
+
"eval_samples_per_second": 22.478,
|
| 1081 |
+
"eval_steps_per_second": 0.71,
|
| 1082 |
+
"step": 304
|
| 1083 |
+
},
|
| 1084 |
+
{
|
| 1085 |
+
"epoch": 0.43478260869565216,
|
| 1086 |
+
"grad_norm": 33.392648021583874,
|
| 1087 |
+
"learning_rate": 3.4835282094871775e-07,
|
| 1088 |
+
"logits/chosen": -3.046875,
|
| 1089 |
+
"logits/rejected": -3.078125,
|
| 1090 |
+
"logps/chosen": -380.0,
|
| 1091 |
+
"logps/rejected": -360.0,
|
| 1092 |
+
"loss": 0.3824,
|
| 1093 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1094 |
+
"rewards/chosen": -1.2421875,
|
| 1095 |
+
"rewards/margins": 2.0,
|
| 1096 |
+
"rewards/rejected": -3.25,
|
| 1097 |
+
"step": 310
|
| 1098 |
+
},
|
| 1099 |
+
{
|
| 1100 |
+
"epoch": 0.4375876577840112,
|
| 1101 |
+
"eval_logits/chosen": -3.078125,
|
| 1102 |
+
"eval_logits/rejected": -3.109375,
|
| 1103 |
+
"eval_logps/chosen": -350.0,
|
| 1104 |
+
"eval_logps/rejected": -342.0,
|
| 1105 |
+
"eval_loss": 0.39230719208717346,
|
| 1106 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 1107 |
+
"eval_rewards/chosen": -1.4140625,
|
| 1108 |
+
"eval_rewards/margins": 1.9296875,
|
| 1109 |
+
"eval_rewards/rejected": -3.34375,
|
| 1110 |
+
"eval_runtime": 65.8562,
|
| 1111 |
+
"eval_samples_per_second": 22.58,
|
| 1112 |
+
"eval_steps_per_second": 0.714,
|
| 1113 |
+
"step": 312
|
| 1114 |
+
},
|
| 1115 |
+
{
|
| 1116 |
+
"epoch": 0.4488078541374474,
|
| 1117 |
+
"grad_norm": 92.67447234773125,
|
| 1118 |
+
"learning_rate": 3.369745538639694e-07,
|
| 1119 |
+
"logits/chosen": -3.0,
|
| 1120 |
+
"logits/rejected": -2.96875,
|
| 1121 |
+
"logps/chosen": -378.0,
|
| 1122 |
+
"logps/rejected": -316.0,
|
| 1123 |
+
"loss": 0.422,
|
| 1124 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1125 |
+
"rewards/chosen": -1.3046875,
|
| 1126 |
+
"rewards/margins": 2.1875,
|
| 1127 |
+
"rewards/rejected": -3.484375,
|
| 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": -336.0,
|
| 1135 |
+
"eval_logps/rejected": -332.0,
|
| 1136 |
+
"eval_loss": 0.38914069533348083,
|
| 1137 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 1138 |
+
"eval_rewards/chosen": -0.74609375,
|
| 1139 |
+
"eval_rewards/margins": 2.046875,
|
| 1140 |
+
"eval_rewards/rejected": -2.796875,
|
| 1141 |
+
"eval_runtime": 65.8877,
|
| 1142 |
+
"eval_samples_per_second": 22.569,
|
| 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": -340.0,
|
| 1151 |
+
"eval_logps/rejected": -336.0,
|
| 1152 |
+
"eval_loss": 0.3886280059814453,
|
| 1153 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 1154 |
+
"eval_rewards/chosen": -0.9296875,
|
| 1155 |
+
"eval_rewards/margins": 2.078125,
|
| 1156 |
+
"eval_rewards/rejected": -3.0,
|
| 1157 |
+
"eval_runtime": 65.7911,
|
| 1158 |
+
"eval_samples_per_second": 22.602,
|
| 1159 |
+
"eval_steps_per_second": 0.714,
|
| 1160 |
+
"step": 328
|
| 1161 |
+
},
|
| 1162 |
+
{
|
| 1163 |
+
"epoch": 0.4628330995792426,
|
| 1164 |
+
"grad_norm": 44.34559171874966,
|
| 1165 |
+
"learning_rate": 3.2538741057779675e-07,
|
| 1166 |
+
"logits/chosen": -2.953125,
|
| 1167 |
+
"logits/rejected": -2.84375,
|
| 1168 |
+
"logps/chosen": -376.0,
|
| 1169 |
+
"logps/rejected": -360.0,
|
| 1170 |
+
"loss": 0.3761,
|
| 1171 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1172 |
+
"rewards/chosen": -0.87890625,
|
| 1173 |
+
"rewards/margins": 2.46875,
|
| 1174 |
+
"rewards/rejected": -3.34375,
|
| 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": -336.0,
|
| 1182 |
+
"eval_logps/rejected": -330.0,
|
| 1183 |
+
"eval_loss": 0.38978397846221924,
|
| 1184 |
+
"eval_rewards/accuracies": 0.8191489577293396,
|
| 1185 |
+
"eval_rewards/chosen": -0.6796875,
|
| 1186 |
+
"eval_rewards/margins": 2.0625,
|
| 1187 |
+
"eval_rewards/rejected": -2.75,
|
| 1188 |
+
"eval_runtime": 66.1797,
|
| 1189 |
+
"eval_samples_per_second": 22.469,
|
| 1190 |
+
"eval_steps_per_second": 0.71,
|
| 1191 |
+
"step": 336
|
| 1192 |
+
},
|
| 1193 |
+
{
|
| 1194 |
+
"epoch": 0.47685834502103785,
|
| 1195 |
+
"grad_norm": 35.10773780257282,
|
| 1196 |
+
"learning_rate": 3.136192185221032e-07,
|
| 1197 |
+
"logits/chosen": -3.03125,
|
| 1198 |
+
"logits/rejected": -2.96875,
|
| 1199 |
+
"logps/chosen": -350.0,
|
| 1200 |
+
"logps/rejected": -366.0,
|
| 1201 |
+
"loss": 0.3534,
|
| 1202 |
+
"rewards/accuracies": 0.875,
|
| 1203 |
+
"rewards/chosen": -0.9609375,
|
| 1204 |
+
"rewards/margins": 2.28125,
|
| 1205 |
+
"rewards/rejected": -3.25,
|
| 1206 |
+
"step": 340
|
| 1207 |
+
},
|
| 1208 |
+
{
|
| 1209 |
+
"epoch": 0.48246844319775595,
|
| 1210 |
+
"eval_logits/chosen": -3.078125,
|
| 1211 |
+
"eval_logits/rejected": -3.109375,
|
| 1212 |
+
"eval_logps/chosen": -340.0,
|
| 1213 |
+
"eval_logps/rejected": -336.0,
|
| 1214 |
+
"eval_loss": 0.3868924677371979,
|
| 1215 |
+
"eval_rewards/accuracies": 0.8563829660415649,
|
| 1216 |
+
"eval_rewards/chosen": -0.953125,
|
| 1217 |
+
"eval_rewards/margins": 2.078125,
|
| 1218 |
+
"eval_rewards/rejected": -3.03125,
|
| 1219 |
+
"eval_runtime": 66.0426,
|
| 1220 |
+
"eval_samples_per_second": 22.516,
|
| 1221 |
+
"eval_steps_per_second": 0.712,
|
| 1222 |
+
"step": 344
|
| 1223 |
+
},
|
| 1224 |
+
{
|
| 1225 |
+
"epoch": 0.4908835904628331,
|
| 1226 |
+
"grad_norm": 31.140190663491566,
|
| 1227 |
+
"learning_rate": 3.016982399315888e-07,
|
| 1228 |
+
"logits/chosen": -3.078125,
|
| 1229 |
+
"logits/rejected": -3.09375,
|
| 1230 |
+
"logps/chosen": -374.0,
|
| 1231 |
+
"logps/rejected": -332.0,
|
| 1232 |
+
"loss": 0.3601,
|
| 1233 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1234 |
+
"rewards/chosen": -1.234375,
|
| 1235 |
+
"rewards/margins": 2.421875,
|
| 1236 |
+
"rewards/rejected": -3.65625,
|
| 1237 |
+
"step": 350
|
| 1238 |
+
},
|
| 1239 |
+
{
|
| 1240 |
+
"epoch": 0.49368863955119213,
|
| 1241 |
+
"eval_logits/chosen": -3.0625,
|
| 1242 |
+
"eval_logits/rejected": -3.09375,
|
| 1243 |
+
"eval_logps/chosen": -342.0,
|
| 1244 |
+
"eval_logps/rejected": -338.0,
|
| 1245 |
+
"eval_loss": 0.38676464557647705,
|
| 1246 |
+
"eval_rewards/accuracies": 0.8510638475418091,
|
| 1247 |
+
"eval_rewards/chosen": -1.0078125,
|
| 1248 |
+
"eval_rewards/margins": 2.140625,
|
| 1249 |
+
"eval_rewards/rejected": -3.15625,
|
| 1250 |
+
"eval_runtime": 65.739,
|
| 1251 |
+
"eval_samples_per_second": 22.62,
|
| 1252 |
+
"eval_steps_per_second": 0.715,
|
| 1253 |
+
"step": 352
|
| 1254 |
+
},
|
| 1255 |
+
{
|
| 1256 |
+
"epoch": 0.5049088359046283,
|
| 1257 |
+
"grad_norm": 47.07813704000649,
|
| 1258 |
+
"learning_rate": 2.896531039697801e-07,
|
| 1259 |
+
"logits/chosen": -3.03125,
|
| 1260 |
+
"logits/rejected": -3.078125,
|
| 1261 |
+
"logps/chosen": -334.0,
|
| 1262 |
+
"logps/rejected": -330.0,
|
| 1263 |
+
"loss": 0.3996,
|
| 1264 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1265 |
+
"rewards/chosen": -1.40625,
|
| 1266 |
+
"rewards/margins": 2.265625,
|
| 1267 |
+
"rewards/rejected": -3.671875,
|
| 1268 |
+
"step": 360
|
| 1269 |
+
},
|
| 1270 |
+
{
|
| 1271 |
+
"epoch": 0.5049088359046283,
|
| 1272 |
+
"eval_logits/chosen": -3.03125,
|
| 1273 |
+
"eval_logits/rejected": -3.0625,
|
| 1274 |
+
"eval_logps/chosen": -348.0,
|
| 1275 |
+
"eval_logps/rejected": -344.0,
|
| 1276 |
+
"eval_loss": 0.3909456133842468,
|
| 1277 |
+
"eval_rewards/accuracies": 0.8404255509376526,
|
| 1278 |
+
"eval_rewards/chosen": -1.2734375,
|
| 1279 |
+
"eval_rewards/margins": 2.171875,
|
| 1280 |
+
"eval_rewards/rejected": -3.4375,
|
| 1281 |
+
"eval_runtime": 66.2809,
|
| 1282 |
+
"eval_samples_per_second": 22.435,
|
| 1283 |
+
"eval_steps_per_second": 0.709,
|
| 1284 |
+
"step": 360
|
| 1285 |
+
},
|
| 1286 |
+
{
|
| 1287 |
+
"epoch": 0.5161290322580645,
|
| 1288 |
+
"eval_logits/chosen": -3.03125,
|
| 1289 |
+
"eval_logits/rejected": -3.046875,
|
| 1290 |
+
"eval_logps/chosen": -350.0,
|
| 1291 |
+
"eval_logps/rejected": -344.0,
|
| 1292 |
+
"eval_loss": 0.3894420564174652,
|
| 1293 |
+
"eval_rewards/accuracies": 0.8297872543334961,
|
| 1294 |
+
"eval_rewards/chosen": -1.375,
|
| 1295 |
+
"eval_rewards/margins": 2.109375,
|
| 1296 |
+
"eval_rewards/rejected": -3.484375,
|
| 1297 |
+
"eval_runtime": 65.5848,
|
| 1298 |
+
"eval_samples_per_second": 22.673,
|
| 1299 |
+
"eval_steps_per_second": 0.717,
|
| 1300 |
+
"step": 368
|
| 1301 |
+
},
|
| 1302 |
+
{
|
| 1303 |
+
"epoch": 0.5189340813464236,
|
| 1304 |
+
"grad_norm": 33.796854317256766,
|
| 1305 |
+
"learning_rate": 2.7751273797385324e-07,
|
| 1306 |
+
"logits/chosen": -2.921875,
|
| 1307 |
+
"logits/rejected": -2.796875,
|
| 1308 |
+
"logps/chosen": -364.0,
|
| 1309 |
+
"logps/rejected": -338.0,
|
| 1310 |
+
"loss": 0.3591,
|
| 1311 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 1312 |
+
"rewards/chosen": -1.4921875,
|
| 1313 |
+
"rewards/margins": 2.390625,
|
| 1314 |
+
"rewards/rejected": -3.890625,
|
| 1315 |
+
"step": 370
|
| 1316 |
+
},
|
| 1317 |
+
{
|
| 1318 |
+
"epoch": 0.5273492286115007,
|
| 1319 |
+
"eval_logits/chosen": -3.046875,
|
| 1320 |
+
"eval_logits/rejected": -3.078125,
|
| 1321 |
+
"eval_logps/chosen": -338.0,
|
| 1322 |
+
"eval_logps/rejected": -334.0,
|
| 1323 |
+
"eval_loss": 0.38463667035102844,
|
| 1324 |
+
"eval_rewards/accuracies": 0.8404255509376526,
|
| 1325 |
+
"eval_rewards/chosen": -0.8203125,
|
| 1326 |
+
"eval_rewards/margins": 2.15625,
|
| 1327 |
+
"eval_rewards/rejected": -2.984375,
|
| 1328 |
+
"eval_runtime": 65.7596,
|
| 1329 |
+
"eval_samples_per_second": 22.613,
|
| 1330 |
+
"eval_steps_per_second": 0.715,
|
| 1331 |
+
"step": 376
|
| 1332 |
+
},
|
| 1333 |
+
{
|
| 1334 |
+
"epoch": 0.5329593267882188,
|
| 1335 |
+
"grad_norm": 40.02993475158031,
|
| 1336 |
+
"learning_rate": 2.6530629798336756e-07,
|
| 1337 |
+
"logits/chosen": -3.0,
|
| 1338 |
+
"logits/rejected": -2.96875,
|
| 1339 |
+
"logps/chosen": -364.0,
|
| 1340 |
+
"logps/rejected": -372.0,
|
| 1341 |
+
"loss": 0.3726,
|
| 1342 |
+
"rewards/accuracies": 0.875,
|
| 1343 |
+
"rewards/chosen": -0.90625,
|
| 1344 |
+
"rewards/margins": 2.34375,
|
| 1345 |
+
"rewards/rejected": -3.25,
|
| 1346 |
+
"step": 380
|
| 1347 |
+
},
|
| 1348 |
+
{
|
| 1349 |
+
"epoch": 0.5385694249649369,
|
| 1350 |
+
"eval_logits/chosen": -3.0625,
|
| 1351 |
+
"eval_logits/rejected": -3.09375,
|
| 1352 |
+
"eval_logps/chosen": -340.0,
|
| 1353 |
+
"eval_logps/rejected": -334.0,
|
| 1354 |
+
"eval_loss": 0.3832222521305084,
|
| 1355 |
+
"eval_rewards/accuracies": 0.8404255509376526,
|
| 1356 |
+
"eval_rewards/chosen": -0.87109375,
|
| 1357 |
+
"eval_rewards/margins": 2.078125,
|
| 1358 |
+
"eval_rewards/rejected": -2.953125,
|
| 1359 |
+
"eval_runtime": 65.913,
|
| 1360 |
+
"eval_samples_per_second": 22.56,
|
| 1361 |
+
"eval_steps_per_second": 0.713,
|
| 1362 |
+
"step": 384
|
| 1363 |
+
},
|
| 1364 |
+
{
|
| 1365 |
+
"epoch": 0.5469845722300141,
|
| 1366 |
+
"grad_norm": 24.220998890998306,
|
| 1367 |
+
"learning_rate": 2.5306309871975437e-07,
|
| 1368 |
+
"logits/chosen": -2.96875,
|
| 1369 |
+
"logits/rejected": -2.96875,
|
| 1370 |
+
"logps/chosen": -358.0,
|
| 1371 |
+
"logps/rejected": -366.0,
|
| 1372 |
+
"loss": 0.371,
|
| 1373 |
+
"rewards/accuracies": 0.875,
|
| 1374 |
+
"rewards/chosen": -0.953125,
|
| 1375 |
+
"rewards/margins": 2.265625,
|
| 1376 |
+
"rewards/rejected": -3.21875,
|
| 1377 |
+
"step": 390
|
| 1378 |
+
},
|
| 1379 |
+
{
|
| 1380 |
+
"epoch": 0.5497896213183731,
|
| 1381 |
+
"eval_logits/chosen": -3.078125,
|
| 1382 |
+
"eval_logits/rejected": -3.109375,
|
| 1383 |
+
"eval_logps/chosen": -340.0,
|
| 1384 |
+
"eval_logps/rejected": -334.0,
|
| 1385 |
+
"eval_loss": 0.38178983330726624,
|
| 1386 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 1387 |
+
"eval_rewards/chosen": -0.8984375,
|
| 1388 |
+
"eval_rewards/margins": 2.046875,
|
| 1389 |
+
"eval_rewards/rejected": -2.953125,
|
| 1390 |
+
"eval_runtime": 65.9539,
|
| 1391 |
+
"eval_samples_per_second": 22.546,
|
| 1392 |
+
"eval_steps_per_second": 0.713,
|
| 1393 |
+
"step": 392
|
| 1394 |
+
},
|
| 1395 |
+
{
|
| 1396 |
+
"epoch": 0.5610098176718092,
|
| 1397 |
+
"grad_norm": 40.61458230374078,
|
| 1398 |
+
"learning_rate": 2.408125431847175e-07,
|
| 1399 |
+
"logits/chosen": -2.828125,
|
| 1400 |
+
"logits/rejected": -2.984375,
|
| 1401 |
+
"logps/chosen": -326.0,
|
| 1402 |
+
"logps/rejected": -360.0,
|
| 1403 |
+
"loss": 0.3546,
|
| 1404 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1405 |
+
"rewards/chosen": -1.0390625,
|
| 1406 |
+
"rewards/margins": 2.3125,
|
| 1407 |
+
"rewards/rejected": -3.34375,
|
| 1408 |
+
"step": 400
|
| 1409 |
+
},
|
| 1410 |
+
{
|
| 1411 |
+
"epoch": 0.5610098176718092,
|
| 1412 |
+
"eval_logits/chosen": -3.0625,
|
| 1413 |
+
"eval_logits/rejected": -3.09375,
|
| 1414 |
+
"eval_logps/chosen": -344.0,
|
| 1415 |
+
"eval_logps/rejected": -338.0,
|
| 1416 |
+
"eval_loss": 0.3821606934070587,
|
| 1417 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 1418 |
+
"eval_rewards/chosen": -1.1328125,
|
| 1419 |
+
"eval_rewards/margins": 2.03125,
|
| 1420 |
+
"eval_rewards/rejected": -3.171875,
|
| 1421 |
+
"eval_runtime": 65.6255,
|
| 1422 |
+
"eval_samples_per_second": 22.659,
|
| 1423 |
+
"eval_steps_per_second": 0.716,
|
| 1424 |
+
"step": 400
|
| 1425 |
+
},
|
| 1426 |
+
{
|
| 1427 |
+
"epoch": 0.5722300140252454,
|
| 1428 |
+
"eval_logits/chosen": -3.078125,
|
| 1429 |
+
"eval_logits/rejected": -3.109375,
|
| 1430 |
+
"eval_logps/chosen": -336.0,
|
| 1431 |
+
"eval_logps/rejected": -332.0,
|
| 1432 |
+
"eval_loss": 0.3827318847179413,
|
| 1433 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 1434 |
+
"eval_rewards/chosen": -0.72265625,
|
| 1435 |
+
"eval_rewards/margins": 2.09375,
|
| 1436 |
+
"eval_rewards/rejected": -2.8125,
|
| 1437 |
+
"eval_runtime": 66.6876,
|
| 1438 |
+
"eval_samples_per_second": 22.298,
|
| 1439 |
+
"eval_steps_per_second": 0.705,
|
| 1440 |
+
"step": 408
|
| 1441 |
+
},
|
| 1442 |
+
{
|
| 1443 |
+
"epoch": 0.5750350631136045,
|
| 1444 |
+
"grad_norm": 37.20301916256947,
|
| 1445 |
+
"learning_rate": 2.2858405204662287e-07,
|
| 1446 |
+
"logits/chosen": -3.0,
|
| 1447 |
+
"logits/rejected": -3.015625,
|
| 1448 |
+
"logps/chosen": -336.0,
|
| 1449 |
+
"logps/rejected": -318.0,
|
| 1450 |
+
"loss": 0.3815,
|
| 1451 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1452 |
+
"rewards/chosen": -1.2421875,
|
| 1453 |
+
"rewards/margins": 2.515625,
|
| 1454 |
+
"rewards/rejected": -3.765625,
|
| 1455 |
+
"step": 410
|
| 1456 |
+
},
|
| 1457 |
+
{
|
| 1458 |
+
"epoch": 0.5834502103786816,
|
| 1459 |
+
"eval_logits/chosen": -3.09375,
|
| 1460 |
+
"eval_logits/rejected": -3.125,
|
| 1461 |
+
"eval_logps/chosen": -344.0,
|
| 1462 |
+
"eval_logps/rejected": -338.0,
|
| 1463 |
+
"eval_loss": 0.38075175881385803,
|
| 1464 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 1465 |
+
"eval_rewards/chosen": -1.1171875,
|
| 1466 |
+
"eval_rewards/margins": 2.015625,
|
| 1467 |
+
"eval_rewards/rejected": -3.125,
|
| 1468 |
+
"eval_runtime": 66.0902,
|
| 1469 |
+
"eval_samples_per_second": 22.5,
|
| 1470 |
+
"eval_steps_per_second": 0.711,
|
| 1471 |
+
"step": 416
|
| 1472 |
+
},
|
| 1473 |
+
{
|
| 1474 |
+
"epoch": 0.5890603085553997,
|
| 1475 |
+
"grad_norm": 27.9879490973163,
|
| 1476 |
+
"learning_rate": 2.164069929844592e-07,
|
| 1477 |
+
"logits/chosen": -2.9375,
|
| 1478 |
+
"logits/rejected": -3.09375,
|
| 1479 |
+
"logps/chosen": -350.0,
|
| 1480 |
+
"logps/rejected": -324.0,
|
| 1481 |
+
"loss": 0.36,
|
| 1482 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1483 |
+
"rewards/chosen": -0.97265625,
|
| 1484 |
+
"rewards/margins": 2.03125,
|
| 1485 |
+
"rewards/rejected": -3.015625,
|
| 1486 |
+
"step": 420
|
| 1487 |
+
},
|
| 1488 |
+
{
|
| 1489 |
+
"epoch": 0.5946704067321178,
|
| 1490 |
+
"eval_logits/chosen": -3.09375,
|
| 1491 |
+
"eval_logits/rejected": -3.125,
|
| 1492 |
+
"eval_logps/chosen": -338.0,
|
| 1493 |
+
"eval_logps/rejected": -334.0,
|
| 1494 |
+
"eval_loss": 0.37844687700271606,
|
| 1495 |
+
"eval_rewards/accuracies": 0.8457446694374084,
|
| 1496 |
+
"eval_rewards/chosen": -0.84375,
|
| 1497 |
+
"eval_rewards/margins": 2.078125,
|
| 1498 |
+
"eval_rewards/rejected": -2.921875,
|
| 1499 |
+
"eval_runtime": 66.0062,
|
| 1500 |
+
"eval_samples_per_second": 22.528,
|
| 1501 |
+
"eval_steps_per_second": 0.712,
|
| 1502 |
+
"step": 424
|
| 1503 |
+
},
|
| 1504 |
+
{
|
| 1505 |
+
"epoch": 0.603085553997195,
|
| 1506 |
+
"grad_norm": 41.322997758051834,
|
| 1507 |
+
"learning_rate": 2.0431061015905793e-07,
|
| 1508 |
+
"logits/chosen": -3.09375,
|
| 1509 |
+
"logits/rejected": -3.109375,
|
| 1510 |
+
"logps/chosen": -318.0,
|
| 1511 |
+
"logps/rejected": -360.0,
|
| 1512 |
+
"loss": 0.3522,
|
| 1513 |
+
"rewards/accuracies": 0.9375,
|
| 1514 |
+
"rewards/chosen": -1.046875,
|
| 1515 |
+
"rewards/margins": 2.6875,
|
| 1516 |
+
"rewards/rejected": -3.734375,
|
| 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": -342.0,
|
| 1524 |
+
"eval_logps/rejected": -338.0,
|
| 1525 |
+
"eval_loss": 0.37742072343826294,
|
| 1526 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 1527 |
+
"eval_rewards/chosen": -1.03125,
|
| 1528 |
+
"eval_rewards/margins": 2.109375,
|
| 1529 |
+
"eval_rewards/rejected": -3.140625,
|
| 1530 |
+
"eval_runtime": 65.9149,
|
| 1531 |
+
"eval_samples_per_second": 22.559,
|
| 1532 |
+
"eval_steps_per_second": 0.713,
|
| 1533 |
+
"step": 432
|
| 1534 |
+
},
|
| 1535 |
+
{
|
| 1536 |
+
"epoch": 0.6171107994389902,
|
| 1537 |
+
"grad_norm": 31.204018506373746,
|
| 1538 |
+
"learning_rate": 1.923239539809505e-07,
|
| 1539 |
+
"logits/chosen": -3.0,
|
| 1540 |
+
"logits/rejected": -3.046875,
|
| 1541 |
+
"logps/chosen": -338.0,
|
| 1542 |
+
"logps/rejected": -334.0,
|
| 1543 |
+
"loss": 0.3323,
|
| 1544 |
+
"rewards/accuracies": 0.875,
|
| 1545 |
+
"rewards/chosen": -1.1953125,
|
| 1546 |
+
"rewards/margins": 2.15625,
|
| 1547 |
+
"rewards/rejected": -3.359375,
|
| 1548 |
+
"step": 440
|
| 1549 |
+
},
|
| 1550 |
+
{
|
| 1551 |
+
"epoch": 0.6171107994389902,
|
| 1552 |
+
"eval_logits/chosen": -3.078125,
|
| 1553 |
+
"eval_logits/rejected": -3.109375,
|
| 1554 |
+
"eval_logps/chosen": -344.0,
|
| 1555 |
+
"eval_logps/rejected": -342.0,
|
| 1556 |
+
"eval_loss": 0.37543419003486633,
|
| 1557 |
+
"eval_rewards/accuracies": 0.8457446694374084,
|
| 1558 |
+
"eval_rewards/chosen": -1.125,
|
| 1559 |
+
"eval_rewards/margins": 2.171875,
|
| 1560 |
+
"eval_rewards/rejected": -3.296875,
|
| 1561 |
+
"eval_runtime": 66.2693,
|
| 1562 |
+
"eval_samples_per_second": 22.439,
|
| 1563 |
+
"eval_steps_per_second": 0.709,
|
| 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": -346.0,
|
| 1571 |
+
"eval_logps/rejected": -342.0,
|
| 1572 |
+
"eval_loss": 0.3757575452327728,
|
| 1573 |
+
"eval_rewards/accuracies": 0.8297872543334961,
|
| 1574 |
+
"eval_rewards/chosen": -1.1953125,
|
| 1575 |
+
"eval_rewards/margins": 2.171875,
|
| 1576 |
+
"eval_rewards/rejected": -3.375,
|
| 1577 |
+
"eval_runtime": 66.0335,
|
| 1578 |
+
"eval_samples_per_second": 22.519,
|
| 1579 |
+
"eval_steps_per_second": 0.712,
|
| 1580 |
+
"step": 448
|
| 1581 |
+
},
|
| 1582 |
+
{
|
| 1583 |
+
"epoch": 0.6311360448807855,
|
| 1584 |
+
"grad_norm": 29.323680763100512,
|
| 1585 |
+
"learning_rate": 1.8047581134353296e-07,
|
| 1586 |
+
"logits/chosen": -2.96875,
|
| 1587 |
+
"logits/rejected": -2.984375,
|
| 1588 |
+
"logps/chosen": -394.0,
|
| 1589 |
+
"logps/rejected": -380.0,
|
| 1590 |
+
"loss": 0.3768,
|
| 1591 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1592 |
+
"rewards/chosen": -1.4375,
|
| 1593 |
+
"rewards/margins": 2.234375,
|
| 1594 |
+
"rewards/rejected": -3.671875,
|
| 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": -340.0,
|
| 1603 |
+
"eval_loss": 0.3760261535644531,
|
| 1604 |
+
"eval_rewards/accuracies": 0.8457446694374084,
|
| 1605 |
+
"eval_rewards/chosen": -1.078125,
|
| 1606 |
+
"eval_rewards/margins": 2.1875,
|
| 1607 |
+
"eval_rewards/rejected": -3.265625,
|
| 1608 |
+
"eval_runtime": 66.2078,
|
| 1609 |
+
"eval_samples_per_second": 22.46,
|
| 1610 |
+
"eval_steps_per_second": 0.71,
|
| 1611 |
+
"step": 456
|
| 1612 |
+
},
|
| 1613 |
+
{
|
| 1614 |
+
"epoch": 0.6451612903225806,
|
| 1615 |
+
"grad_norm": 36.95205536065495,
|
| 1616 |
+
"learning_rate": 1.687946364890877e-07,
|
| 1617 |
+
"logits/chosen": -2.96875,
|
| 1618 |
+
"logits/rejected": -2.96875,
|
| 1619 |
+
"logps/chosen": -338.0,
|
| 1620 |
+
"logps/rejected": -332.0,
|
| 1621 |
+
"loss": 0.3637,
|
| 1622 |
+
"rewards/accuracies": 0.9375,
|
| 1623 |
+
"rewards/chosen": -1.2734375,
|
| 1624 |
+
"rewards/margins": 2.40625,
|
| 1625 |
+
"rewards/rejected": -3.6875,
|
| 1626 |
+
"step": 460
|
| 1627 |
+
},
|
| 1628 |
+
{
|
| 1629 |
+
"epoch": 0.6507713884992987,
|
| 1630 |
+
"eval_logits/chosen": -3.078125,
|
| 1631 |
+
"eval_logits/rejected": -3.09375,
|
| 1632 |
+
"eval_logps/chosen": -342.0,
|
| 1633 |
+
"eval_logps/rejected": -340.0,
|
| 1634 |
+
"eval_loss": 0.3759002089500427,
|
| 1635 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 1636 |
+
"eval_rewards/chosen": -1.0625,
|
| 1637 |
+
"eval_rewards/margins": 2.21875,
|
| 1638 |
+
"eval_rewards/rejected": -3.28125,
|
| 1639 |
+
"eval_runtime": 65.6388,
|
| 1640 |
+
"eval_samples_per_second": 22.654,
|
| 1641 |
+
"eval_steps_per_second": 0.716,
|
| 1642 |
+
"step": 464
|
| 1643 |
+
},
|
| 1644 |
+
{
|
| 1645 |
+
"epoch": 0.6591865357643759,
|
| 1646 |
+
"grad_norm": 50.750866736233036,
|
| 1647 |
+
"learning_rate": 1.573084826736921e-07,
|
| 1648 |
+
"logits/chosen": -2.984375,
|
| 1649 |
+
"logits/rejected": -3.046875,
|
| 1650 |
+
"logps/chosen": -332.0,
|
| 1651 |
+
"logps/rejected": -340.0,
|
| 1652 |
+
"loss": 0.3901,
|
| 1653 |
+
"rewards/accuracies": 0.875,
|
| 1654 |
+
"rewards/chosen": -1.484375,
|
| 1655 |
+
"rewards/margins": 2.21875,
|
| 1656 |
+
"rewards/rejected": -3.703125,
|
| 1657 |
+
"step": 470
|
| 1658 |
+
},
|
| 1659 |
+
{
|
| 1660 |
+
"epoch": 0.6619915848527349,
|
| 1661 |
+
"eval_logits/chosen": -3.0625,
|
| 1662 |
+
"eval_logits/rejected": -3.09375,
|
| 1663 |
+
"eval_logps/chosen": -350.0,
|
| 1664 |
+
"eval_logps/rejected": -346.0,
|
| 1665 |
+
"eval_loss": 0.3759775161743164,
|
| 1666 |
+
"eval_rewards/accuracies": 0.8457446694374084,
|
| 1667 |
+
"eval_rewards/chosen": -1.4375,
|
| 1668 |
+
"eval_rewards/margins": 2.109375,
|
| 1669 |
+
"eval_rewards/rejected": -3.546875,
|
| 1670 |
+
"eval_runtime": 66.2324,
|
| 1671 |
+
"eval_samples_per_second": 22.451,
|
| 1672 |
+
"eval_steps_per_second": 0.71,
|
| 1673 |
+
"step": 472
|
| 1674 |
+
},
|
| 1675 |
+
{
|
| 1676 |
+
"epoch": 0.6732117812061711,
|
| 1677 |
+
"grad_norm": 46.86048650983821,
|
| 1678 |
+
"learning_rate": 1.460449347951278e-07,
|
| 1679 |
+
"logits/chosen": -2.875,
|
| 1680 |
+
"logits/rejected": -3.0,
|
| 1681 |
+
"logps/chosen": -328.0,
|
| 1682 |
+
"logps/rejected": -324.0,
|
| 1683 |
+
"loss": 0.3555,
|
| 1684 |
+
"rewards/accuracies": 0.875,
|
| 1685 |
+
"rewards/chosen": -1.390625,
|
| 1686 |
+
"rewards/margins": 2.453125,
|
| 1687 |
+
"rewards/rejected": -3.84375,
|
| 1688 |
+
"step": 480
|
| 1689 |
+
},
|
| 1690 |
+
{
|
| 1691 |
+
"epoch": 0.6732117812061711,
|
| 1692 |
+
"eval_logits/chosen": -3.078125,
|
| 1693 |
+
"eval_logits/rejected": -3.09375,
|
| 1694 |
+
"eval_logps/chosen": -350.0,
|
| 1695 |
+
"eval_logps/rejected": -346.0,
|
| 1696 |
+
"eval_loss": 0.3772374987602234,
|
| 1697 |
+
"eval_rewards/accuracies": 0.8404255509376526,
|
| 1698 |
+
"eval_rewards/chosen": -1.3671875,
|
| 1699 |
+
"eval_rewards/margins": 2.140625,
|
| 1700 |
+
"eval_rewards/rejected": -3.5,
|
| 1701 |
+
"eval_runtime": 66.1888,
|
| 1702 |
+
"eval_samples_per_second": 22.466,
|
| 1703 |
+
"eval_steps_per_second": 0.71,
|
| 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": -344.0,
|
| 1711 |
+
"eval_logps/rejected": -340.0,
|
| 1712 |
+
"eval_loss": 0.37612324953079224,
|
| 1713 |
+
"eval_rewards/accuracies": 0.8563829660415649,
|
| 1714 |
+
"eval_rewards/chosen": -1.0859375,
|
| 1715 |
+
"eval_rewards/margins": 2.171875,
|
| 1716 |
+
"eval_rewards/rejected": -3.25,
|
| 1717 |
+
"eval_runtime": 66.0573,
|
| 1718 |
+
"eval_samples_per_second": 22.511,
|
| 1719 |
+
"eval_steps_per_second": 0.712,
|
| 1720 |
+
"step": 488
|
| 1721 |
+
},
|
| 1722 |
+
{
|
| 1723 |
+
"epoch": 0.6872370266479664,
|
| 1724 |
+
"grad_norm": 38.26875632825401,
|
| 1725 |
+
"learning_rate": 1.3503104314558838e-07,
|
| 1726 |
+
"logits/chosen": -3.046875,
|
| 1727 |
+
"logits/rejected": -3.0625,
|
| 1728 |
+
"logps/chosen": -382.0,
|
| 1729 |
+
"logps/rejected": -350.0,
|
| 1730 |
+
"loss": 0.373,
|
| 1731 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1732 |
+
"rewards/chosen": -1.40625,
|
| 1733 |
+
"rewards/margins": 2.15625,
|
| 1734 |
+
"rewards/rejected": -3.5625,
|
| 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": -340.0,
|
| 1742 |
+
"eval_logps/rejected": -334.0,
|
| 1743 |
+
"eval_loss": 0.37762168049812317,
|
| 1744 |
+
"eval_rewards/accuracies": 0.8457446694374084,
|
| 1745 |
+
"eval_rewards/chosen": -0.90625,
|
| 1746 |
+
"eval_rewards/margins": 2.046875,
|
| 1747 |
+
"eval_rewards/rejected": -2.953125,
|
| 1748 |
+
"eval_runtime": 65.9608,
|
| 1749 |
+
"eval_samples_per_second": 22.544,
|
| 1750 |
+
"eval_steps_per_second": 0.713,
|
| 1751 |
+
"step": 496
|
| 1752 |
+
},
|
| 1753 |
+
{
|
| 1754 |
+
"epoch": 0.7012622720897616,
|
| 1755 |
+
"grad_norm": 37.30046756155309,
|
| 1756 |
+
"learning_rate": 1.2429325844828475e-07,
|
| 1757 |
+
"logits/chosen": -2.90625,
|
| 1758 |
+
"logits/rejected": -3.015625,
|
| 1759 |
+
"logps/chosen": -336.0,
|
| 1760 |
+
"logps/rejected": -328.0,
|
| 1761 |
+
"loss": 0.3617,
|
| 1762 |
+
"rewards/accuracies": 0.9375,
|
| 1763 |
+
"rewards/chosen": -0.9296875,
|
| 1764 |
+
"rewards/margins": 2.59375,
|
| 1765 |
+
"rewards/rejected": -3.53125,
|
| 1766 |
+
"step": 500
|
| 1767 |
+
},
|
| 1768 |
+
{
|
| 1769 |
+
"epoch": 0.7068723702664796,
|
| 1770 |
+
"eval_logits/chosen": -3.078125,
|
| 1771 |
+
"eval_logits/rejected": -3.109375,
|
| 1772 |
+
"eval_logps/chosen": -340.0,
|
| 1773 |
+
"eval_logps/rejected": -334.0,
|
| 1774 |
+
"eval_loss": 0.3774646520614624,
|
| 1775 |
+
"eval_rewards/accuracies": 0.8404255509376526,
|
| 1776 |
+
"eval_rewards/chosen": -0.94921875,
|
| 1777 |
+
"eval_rewards/margins": 2.0,
|
| 1778 |
+
"eval_rewards/rejected": -2.953125,
|
| 1779 |
+
"eval_runtime": 65.763,
|
| 1780 |
+
"eval_samples_per_second": 22.612,
|
| 1781 |
+
"eval_steps_per_second": 0.715,
|
| 1782 |
+
"step": 504
|
| 1783 |
+
},
|
| 1784 |
+
{
|
| 1785 |
+
"epoch": 0.7152875175315568,
|
| 1786 |
+
"grad_norm": 28.616464108884514,
|
| 1787 |
+
"learning_rate": 1.1385736833396248e-07,
|
| 1788 |
+
"logits/chosen": -2.984375,
|
| 1789 |
+
"logits/rejected": -3.0625,
|
| 1790 |
+
"logps/chosen": -392.0,
|
| 1791 |
+
"logps/rejected": -338.0,
|
| 1792 |
+
"loss": 0.3665,
|
| 1793 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 1794 |
+
"rewards/chosen": -1.2578125,
|
| 1795 |
+
"rewards/margins": 1.90625,
|
| 1796 |
+
"rewards/rejected": -3.15625,
|
| 1797 |
+
"step": 510
|
| 1798 |
+
},
|
| 1799 |
+
{
|
| 1800 |
+
"epoch": 0.7180925666199158,
|
| 1801 |
+
"eval_logits/chosen": -3.078125,
|
| 1802 |
+
"eval_logits/rejected": -3.109375,
|
| 1803 |
+
"eval_logps/chosen": -338.0,
|
| 1804 |
+
"eval_logps/rejected": -334.0,
|
| 1805 |
+
"eval_loss": 0.3764929175376892,
|
| 1806 |
+
"eval_rewards/accuracies": 0.8457446694374084,
|
| 1807 |
+
"eval_rewards/chosen": -0.8359375,
|
| 1808 |
+
"eval_rewards/margins": 2.0625,
|
| 1809 |
+
"eval_rewards/rejected": -2.90625,
|
| 1810 |
+
"eval_runtime": 65.6772,
|
| 1811 |
+
"eval_samples_per_second": 22.641,
|
| 1812 |
+
"eval_steps_per_second": 0.716,
|
| 1813 |
+
"step": 512
|
| 1814 |
+
},
|
| 1815 |
+
{
|
| 1816 |
+
"epoch": 0.729312762973352,
|
| 1817 |
+
"grad_norm": 26.70960973383744,
|
| 1818 |
+
"learning_rate": 1.0374843540988668e-07,
|
| 1819 |
+
"logits/chosen": -2.984375,
|
| 1820 |
+
"logits/rejected": -2.9375,
|
| 1821 |
+
"logps/chosen": -314.0,
|
| 1822 |
+
"logps/rejected": -318.0,
|
| 1823 |
+
"loss": 0.3612,
|
| 1824 |
+
"rewards/accuracies": 0.9375,
|
| 1825 |
+
"rewards/chosen": -1.0546875,
|
| 1826 |
+
"rewards/margins": 2.390625,
|
| 1827 |
+
"rewards/rejected": -3.4375,
|
| 1828 |
+
"step": 520
|
| 1829 |
+
},
|
| 1830 |
+
{
|
| 1831 |
+
"epoch": 0.729312762973352,
|
| 1832 |
+
"eval_logits/chosen": -3.078125,
|
| 1833 |
+
"eval_logits/rejected": -3.109375,
|
| 1834 |
+
"eval_logps/chosen": -340.0,
|
| 1835 |
+
"eval_logps/rejected": -336.0,
|
| 1836 |
+
"eval_loss": 0.3756672441959381,
|
| 1837 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 1838 |
+
"eval_rewards/chosen": -0.9296875,
|
| 1839 |
+
"eval_rewards/margins": 2.125,
|
| 1840 |
+
"eval_rewards/rejected": -3.0625,
|
| 1841 |
+
"eval_runtime": 66.0986,
|
| 1842 |
+
"eval_samples_per_second": 22.497,
|
| 1843 |
+
"eval_steps_per_second": 0.711,
|
| 1844 |
+
"step": 520
|
| 1845 |
+
},
|
| 1846 |
+
{
|
| 1847 |
+
"epoch": 0.7405329593267882,
|
| 1848 |
+
"eval_logits/chosen": -3.078125,
|
| 1849 |
+
"eval_logits/rejected": -3.109375,
|
| 1850 |
+
"eval_logps/chosen": -340.0,
|
| 1851 |
+
"eval_logps/rejected": -338.0,
|
| 1852 |
+
"eval_loss": 0.3768278658390045,
|
| 1853 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 1854 |
+
"eval_rewards/chosen": -0.9375,
|
| 1855 |
+
"eval_rewards/margins": 2.1875,
|
| 1856 |
+
"eval_rewards/rejected": -3.109375,
|
| 1857 |
+
"eval_runtime": 66.3604,
|
| 1858 |
+
"eval_samples_per_second": 22.408,
|
| 1859 |
+
"eval_steps_per_second": 0.708,
|
| 1860 |
+
"step": 528
|
| 1861 |
+
},
|
| 1862 |
+
{
|
| 1863 |
+
"epoch": 0.7433380084151473,
|
| 1864 |
+
"grad_norm": 44.52616987508156,
|
| 1865 |
+
"learning_rate": 9.39907370700287e-08,
|
| 1866 |
+
"logits/chosen": -3.078125,
|
| 1867 |
+
"logits/rejected": -3.0625,
|
| 1868 |
+
"logps/chosen": -328.0,
|
| 1869 |
+
"logps/rejected": -320.0,
|
| 1870 |
+
"loss": 0.3761,
|
| 1871 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1872 |
+
"rewards/chosen": -1.109375,
|
| 1873 |
+
"rewards/margins": 2.1875,
|
| 1874 |
+
"rewards/rejected": -3.28125,
|
| 1875 |
+
"step": 530
|
| 1876 |
+
},
|
| 1877 |
+
{
|
| 1878 |
+
"epoch": 0.7517531556802244,
|
| 1879 |
+
"eval_logits/chosen": -3.09375,
|
| 1880 |
+
"eval_logits/rejected": -3.109375,
|
| 1881 |
+
"eval_logps/chosen": -340.0,
|
| 1882 |
+
"eval_logps/rejected": -336.0,
|
| 1883 |
+
"eval_loss": 0.3774166703224182,
|
| 1884 |
+
"eval_rewards/accuracies": 0.8457446694374084,
|
| 1885 |
+
"eval_rewards/chosen": -0.9453125,
|
| 1886 |
+
"eval_rewards/margins": 2.125,
|
| 1887 |
+
"eval_rewards/rejected": -3.0625,
|
| 1888 |
+
"eval_runtime": 66.0445,
|
| 1889 |
+
"eval_samples_per_second": 22.515,
|
| 1890 |
+
"eval_steps_per_second": 0.712,
|
| 1891 |
+
"step": 536
|
| 1892 |
+
},
|
| 1893 |
+
{
|
| 1894 |
+
"epoch": 0.7573632538569425,
|
| 1895 |
+
"grad_norm": 31.39879813498721,
|
| 1896 |
+
"learning_rate": 8.460770719100316e-08,
|
| 1897 |
+
"logits/chosen": -3.0,
|
| 1898 |
+
"logits/rejected": -3.0,
|
| 1899 |
+
"logps/chosen": -326.0,
|
| 1900 |
+
"logps/rejected": -346.0,
|
| 1901 |
+
"loss": 0.3852,
|
| 1902 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1903 |
+
"rewards/chosen": -1.015625,
|
| 1904 |
+
"rewards/margins": 2.046875,
|
| 1905 |
+
"rewards/rejected": -3.0625,
|
| 1906 |
+
"step": 540
|
| 1907 |
+
},
|
| 1908 |
+
{
|
| 1909 |
+
"epoch": 0.7629733520336606,
|
| 1910 |
+
"eval_logits/chosen": -3.09375,
|
| 1911 |
+
"eval_logits/rejected": -3.125,
|
| 1912 |
+
"eval_logps/chosen": -340.0,
|
| 1913 |
+
"eval_logps/rejected": -336.0,
|
| 1914 |
+
"eval_loss": 0.37610381841659546,
|
| 1915 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 1916 |
+
"eval_rewards/chosen": -0.94140625,
|
| 1917 |
+
"eval_rewards/margins": 2.0625,
|
| 1918 |
+
"eval_rewards/rejected": -3.0,
|
| 1919 |
+
"eval_runtime": 66.8835,
|
| 1920 |
+
"eval_samples_per_second": 22.233,
|
| 1921 |
+
"eval_steps_per_second": 0.703,
|
| 1922 |
+
"step": 544
|
| 1923 |
+
},
|
| 1924 |
+
{
|
| 1925 |
+
"epoch": 0.7713884992987378,
|
| 1926 |
+
"grad_norm": 33.163388614762034,
|
| 1927 |
+
"learning_rate": 7.562187985377877e-08,
|
| 1928 |
+
"logits/chosen": -3.03125,
|
| 1929 |
+
"logits/rejected": -3.0625,
|
| 1930 |
+
"logps/chosen": -302.0,
|
| 1931 |
+
"logps/rejected": -336.0,
|
| 1932 |
+
"loss": 0.3887,
|
| 1933 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1934 |
+
"rewards/chosen": -1.0,
|
| 1935 |
+
"rewards/margins": 2.234375,
|
| 1936 |
+
"rewards/rejected": -3.234375,
|
| 1937 |
+
"step": 550
|
| 1938 |
+
},
|
| 1939 |
+
{
|
| 1940 |
+
"epoch": 0.7741935483870968,
|
| 1941 |
+
"eval_logits/chosen": -3.09375,
|
| 1942 |
+
"eval_logits/rejected": -3.125,
|
| 1943 |
+
"eval_logps/chosen": -340.0,
|
| 1944 |
+
"eval_logps/rejected": -334.0,
|
| 1945 |
+
"eval_loss": 0.3767978847026825,
|
| 1946 |
+
"eval_rewards/accuracies": 0.8191489577293396,
|
| 1947 |
+
"eval_rewards/chosen": -0.8828125,
|
| 1948 |
+
"eval_rewards/margins": 2.03125,
|
| 1949 |
+
"eval_rewards/rejected": -2.921875,
|
| 1950 |
+
"eval_runtime": 66.2121,
|
| 1951 |
+
"eval_samples_per_second": 22.458,
|
| 1952 |
+
"eval_steps_per_second": 0.71,
|
| 1953 |
+
"step": 552
|
| 1954 |
+
},
|
| 1955 |
+
{
|
| 1956 |
+
"epoch": 0.7854137447405329,
|
| 1957 |
+
"grad_norm": 33.17915364920084,
|
| 1958 |
+
"learning_rate": 6.705483522631811e-08,
|
| 1959 |
+
"logits/chosen": -3.203125,
|
| 1960 |
+
"logits/rejected": -3.109375,
|
| 1961 |
+
"logps/chosen": -284.0,
|
| 1962 |
+
"logps/rejected": -296.0,
|
| 1963 |
+
"loss": 0.3653,
|
| 1964 |
+
"rewards/accuracies": 0.875,
|
| 1965 |
+
"rewards/chosen": -1.2578125,
|
| 1966 |
+
"rewards/margins": 1.953125,
|
| 1967 |
+
"rewards/rejected": -3.21875,
|
| 1968 |
+
"step": 560
|
| 1969 |
+
},
|
| 1970 |
+
{
|
| 1971 |
+
"epoch": 0.7854137447405329,
|
| 1972 |
+
"eval_logits/chosen": -3.09375,
|
| 1973 |
+
"eval_logits/rejected": -3.125,
|
| 1974 |
+
"eval_logps/chosen": -338.0,
|
| 1975 |
+
"eval_logps/rejected": -334.0,
|
| 1976 |
+
"eval_loss": 0.37570077180862427,
|
| 1977 |
+
"eval_rewards/accuracies": 0.8244680762290955,
|
| 1978 |
+
"eval_rewards/chosen": -0.84765625,
|
| 1979 |
+
"eval_rewards/margins": 2.078125,
|
| 1980 |
+
"eval_rewards/rejected": -2.921875,
|
| 1981 |
+
"eval_runtime": 66.2859,
|
| 1982 |
+
"eval_samples_per_second": 22.433,
|
| 1983 |
+
"eval_steps_per_second": 0.709,
|
| 1984 |
+
"step": 560
|
| 1985 |
+
},
|
| 1986 |
+
{
|
| 1987 |
+
"epoch": 0.7966339410939691,
|
| 1988 |
+
"eval_logits/chosen": -3.09375,
|
| 1989 |
+
"eval_logits/rejected": -3.125,
|
| 1990 |
+
"eval_logps/chosen": -336.0,
|
| 1991 |
+
"eval_logps/rejected": -332.0,
|
| 1992 |
+
"eval_loss": 0.37514832615852356,
|
| 1993 |
+
"eval_rewards/accuracies": 0.8297872543334961,
|
| 1994 |
+
"eval_rewards/chosen": -0.76171875,
|
| 1995 |
+
"eval_rewards/margins": 2.109375,
|
| 1996 |
+
"eval_rewards/rejected": -2.875,
|
| 1997 |
+
"eval_runtime": 66.2321,
|
| 1998 |
+
"eval_samples_per_second": 22.451,
|
| 1999 |
+
"eval_steps_per_second": 0.71,
|
| 2000 |
+
"step": 568
|
| 2001 |
+
},
|
| 2002 |
+
{
|
| 2003 |
+
"epoch": 0.7994389901823282,
|
| 2004 |
+
"grad_norm": 39.45560582368654,
|
| 2005 |
+
"learning_rate": 5.8927147737114546e-08,
|
| 2006 |
+
"logits/chosen": -3.09375,
|
| 2007 |
+
"logits/rejected": -3.046875,
|
| 2008 |
+
"logps/chosen": -322.0,
|
| 2009 |
+
"logps/rejected": -306.0,
|
| 2010 |
+
"loss": 0.3595,
|
| 2011 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 2012 |
+
"rewards/chosen": -1.171875,
|
| 2013 |
+
"rewards/margins": 2.171875,
|
| 2014 |
+
"rewards/rejected": -3.34375,
|
| 2015 |
+
"step": 570
|
| 2016 |
+
},
|
| 2017 |
+
{
|
| 2018 |
+
"epoch": 0.8078541374474053,
|
| 2019 |
+
"eval_logits/chosen": -3.09375,
|
| 2020 |
+
"eval_logits/rejected": -3.125,
|
| 2021 |
+
"eval_logps/chosen": -338.0,
|
| 2022 |
+
"eval_logps/rejected": -334.0,
|
| 2023 |
+
"eval_loss": 0.3752020299434662,
|
| 2024 |
+
"eval_rewards/accuracies": 0.8297872543334961,
|
| 2025 |
+
"eval_rewards/chosen": -0.78125,
|
| 2026 |
+
"eval_rewards/margins": 2.109375,
|
| 2027 |
+
"eval_rewards/rejected": -2.890625,
|
| 2028 |
+
"eval_runtime": 66.0782,
|
| 2029 |
+
"eval_samples_per_second": 22.504,
|
| 2030 |
+
"eval_steps_per_second": 0.711,
|
| 2031 |
+
"step": 576
|
| 2032 |
+
},
|
| 2033 |
+
{
|
| 2034 |
+
"epoch": 0.8134642356241234,
|
| 2035 |
+
"grad_norm": 56.20015949598879,
|
| 2036 |
+
"learning_rate": 5.125833666409107e-08,
|
| 2037 |
+
"logits/chosen": -2.9375,
|
| 2038 |
+
"logits/rejected": -3.0,
|
| 2039 |
+
"logps/chosen": -316.0,
|
| 2040 |
+
"logps/rejected": -304.0,
|
| 2041 |
+
"loss": 0.3618,
|
| 2042 |
+
"rewards/accuracies": 0.887499988079071,
|
| 2043 |
+
"rewards/chosen": -1.0078125,
|
| 2044 |
+
"rewards/margins": 2.21875,
|
| 2045 |
+
"rewards/rejected": -3.234375,
|
| 2046 |
+
"step": 580
|
| 2047 |
+
},
|
| 2048 |
+
{
|
| 2049 |
+
"epoch": 0.8190743338008415,
|
| 2050 |
+
"eval_logits/chosen": -3.09375,
|
| 2051 |
+
"eval_logits/rejected": -3.125,
|
| 2052 |
+
"eval_logps/chosen": -340.0,
|
| 2053 |
+
"eval_logps/rejected": -336.0,
|
| 2054 |
+
"eval_loss": 0.3744948208332062,
|
| 2055 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 2056 |
+
"eval_rewards/chosen": -0.91015625,
|
| 2057 |
+
"eval_rewards/margins": 2.09375,
|
| 2058 |
+
"eval_rewards/rejected": -3.0,
|
| 2059 |
+
"eval_runtime": 66.6042,
|
| 2060 |
+
"eval_samples_per_second": 22.326,
|
| 2061 |
+
"eval_steps_per_second": 0.706,
|
| 2062 |
+
"step": 584
|
| 2063 |
+
},
|
| 2064 |
+
{
|
| 2065 |
+
"epoch": 0.8274894810659187,
|
| 2066 |
+
"grad_norm": 46.520579632902326,
|
| 2067 |
+
"learning_rate": 4.4066819257526423e-08,
|
| 2068 |
+
"logits/chosen": -3.0625,
|
| 2069 |
+
"logits/rejected": -3.015625,
|
| 2070 |
+
"logps/chosen": -326.0,
|
| 2071 |
+
"logps/rejected": -402.0,
|
| 2072 |
+
"loss": 0.3668,
|
| 2073 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 2074 |
+
"rewards/chosen": -0.953125,
|
| 2075 |
+
"rewards/margins": 2.40625,
|
| 2076 |
+
"rewards/rejected": -3.375,
|
| 2077 |
+
"step": 590
|
| 2078 |
+
},
|
| 2079 |
+
{
|
| 2080 |
+
"epoch": 0.8302945301542777,
|
| 2081 |
+
"eval_logits/chosen": -3.09375,
|
| 2082 |
+
"eval_logits/rejected": -3.125,
|
| 2083 |
+
"eval_logps/chosen": -340.0,
|
| 2084 |
+
"eval_logps/rejected": -336.0,
|
| 2085 |
+
"eval_loss": 0.3740004003047943,
|
| 2086 |
+
"eval_rewards/accuracies": 0.8404255509376526,
|
| 2087 |
+
"eval_rewards/chosen": -0.921875,
|
| 2088 |
+
"eval_rewards/margins": 2.09375,
|
| 2089 |
+
"eval_rewards/rejected": -3.015625,
|
| 2090 |
+
"eval_runtime": 66.3956,
|
| 2091 |
+
"eval_samples_per_second": 22.396,
|
| 2092 |
+
"eval_steps_per_second": 0.708,
|
| 2093 |
+
"step": 592
|
| 2094 |
+
},
|
| 2095 |
+
{
|
| 2096 |
+
"epoch": 0.8415147265077139,
|
| 2097 |
+
"grad_norm": 47.468270807569716,
|
| 2098 |
+
"learning_rate": 3.736986650958562e-08,
|
| 2099 |
+
"logits/chosen": -3.03125,
|
| 2100 |
+
"logits/rejected": -3.15625,
|
| 2101 |
+
"logps/chosen": -408.0,
|
| 2102 |
+
"logps/rejected": -378.0,
|
| 2103 |
+
"loss": 0.3709,
|
| 2104 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2105 |
+
"rewards/chosen": -0.72265625,
|
| 2106 |
+
"rewards/margins": 2.921875,
|
| 2107 |
+
"rewards/rejected": -3.640625,
|
| 2108 |
+
"step": 600
|
| 2109 |
+
},
|
| 2110 |
+
{
|
| 2111 |
+
"epoch": 0.8415147265077139,
|
| 2112 |
+
"eval_logits/chosen": -3.09375,
|
| 2113 |
+
"eval_logits/rejected": -3.125,
|
| 2114 |
+
"eval_logps/chosen": -342.0,
|
| 2115 |
+
"eval_logps/rejected": -336.0,
|
| 2116 |
+
"eval_loss": 0.374103307723999,
|
| 2117 |
+
"eval_rewards/accuracies": 0.8297872543334961,
|
| 2118 |
+
"eval_rewards/chosen": -0.97265625,
|
| 2119 |
+
"eval_rewards/margins": 2.0625,
|
| 2120 |
+
"eval_rewards/rejected": -3.046875,
|
| 2121 |
+
"eval_runtime": 66.1384,
|
| 2122 |
+
"eval_samples_per_second": 22.483,
|
| 2123 |
+
"eval_steps_per_second": 0.711,
|
| 2124 |
+
"step": 600
|
| 2125 |
+
},
|
| 2126 |
+
{
|
| 2127 |
+
"epoch": 0.85273492286115,
|
| 2128 |
+
"eval_logits/chosen": -3.09375,
|
| 2129 |
+
"eval_logits/rejected": -3.125,
|
| 2130 |
+
"eval_logps/chosen": -342.0,
|
| 2131 |
+
"eval_logps/rejected": -338.0,
|
| 2132 |
+
"eval_loss": 0.3734557628631592,
|
| 2133 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 2134 |
+
"eval_rewards/chosen": -1.015625,
|
| 2135 |
+
"eval_rewards/margins": 2.078125,
|
| 2136 |
+
"eval_rewards/rejected": -3.09375,
|
| 2137 |
+
"eval_runtime": 66.2586,
|
| 2138 |
+
"eval_samples_per_second": 22.442,
|
| 2139 |
+
"eval_steps_per_second": 0.709,
|
| 2140 |
+
"step": 608
|
| 2141 |
+
},
|
| 2142 |
+
{
|
| 2143 |
+
"epoch": 0.8555399719495091,
|
| 2144 |
+
"grad_norm": 56.20453228147611,
|
| 2145 |
+
"learning_rate": 3.118356167668065e-08,
|
| 2146 |
+
"logits/chosen": -3.109375,
|
| 2147 |
+
"logits/rejected": -3.078125,
|
| 2148 |
+
"logps/chosen": -320.0,
|
| 2149 |
+
"logps/rejected": -330.0,
|
| 2150 |
+
"loss": 0.3574,
|
| 2151 |
+
"rewards/accuracies": 0.925000011920929,
|
| 2152 |
+
"rewards/chosen": -1.015625,
|
| 2153 |
+
"rewards/margins": 2.671875,
|
| 2154 |
+
"rewards/rejected": -3.6875,
|
| 2155 |
+
"step": 610
|
| 2156 |
+
},
|
| 2157 |
+
{
|
| 2158 |
+
"epoch": 0.8639551192145862,
|
| 2159 |
+
"eval_logits/chosen": -3.09375,
|
| 2160 |
+
"eval_logits/rejected": -3.125,
|
| 2161 |
+
"eval_logps/chosen": -342.0,
|
| 2162 |
+
"eval_logps/rejected": -338.0,
|
| 2163 |
+
"eval_loss": 0.3729269802570343,
|
| 2164 |
+
"eval_rewards/accuracies": 0.8297872543334961,
|
| 2165 |
+
"eval_rewards/chosen": -1.0390625,
|
| 2166 |
+
"eval_rewards/margins": 2.09375,
|
| 2167 |
+
"eval_rewards/rejected": -3.125,
|
| 2168 |
+
"eval_runtime": 66.4655,
|
| 2169 |
+
"eval_samples_per_second": 22.372,
|
| 2170 |
+
"eval_steps_per_second": 0.707,
|
| 2171 |
+
"step": 616
|
| 2172 |
+
},
|
| 2173 |
+
{
|
| 2174 |
+
"epoch": 0.8695652173913043,
|
| 2175 |
+
"grad_norm": 40.32752366240338,
|
| 2176 |
+
"learning_rate": 2.552276165427056e-08,
|
| 2177 |
+
"logits/chosen": -3.046875,
|
| 2178 |
+
"logits/rejected": -3.03125,
|
| 2179 |
+
"logps/chosen": -368.0,
|
| 2180 |
+
"logps/rejected": -394.0,
|
| 2181 |
+
"loss": 0.3611,
|
| 2182 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2183 |
+
"rewards/chosen": -1.2421875,
|
| 2184 |
+
"rewards/margins": 2.5,
|
| 2185 |
+
"rewards/rejected": -3.734375,
|
| 2186 |
+
"step": 620
|
| 2187 |
+
},
|
| 2188 |
+
{
|
| 2189 |
+
"epoch": 0.8751753155680224,
|
| 2190 |
+
"eval_logits/chosen": -3.09375,
|
| 2191 |
+
"eval_logits/rejected": -3.125,
|
| 2192 |
+
"eval_logps/chosen": -342.0,
|
| 2193 |
+
"eval_logps/rejected": -338.0,
|
| 2194 |
+
"eval_loss": 0.3727641999721527,
|
| 2195 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 2196 |
+
"eval_rewards/chosen": -1.0625,
|
| 2197 |
+
"eval_rewards/margins": 2.09375,
|
| 2198 |
+
"eval_rewards/rejected": -3.15625,
|
| 2199 |
+
"eval_runtime": 66.0542,
|
| 2200 |
+
"eval_samples_per_second": 22.512,
|
| 2201 |
+
"eval_steps_per_second": 0.712,
|
| 2202 |
+
"step": 624
|
| 2203 |
+
},
|
| 2204 |
+
{
|
| 2205 |
+
"epoch": 0.8835904628330996,
|
| 2206 |
+
"grad_norm": 39.39615195808181,
|
| 2207 |
+
"learning_rate": 2.040106129686356e-08,
|
| 2208 |
+
"logits/chosen": -3.0,
|
| 2209 |
+
"logits/rejected": -3.078125,
|
| 2210 |
+
"logps/chosen": -374.0,
|
| 2211 |
+
"logps/rejected": -348.0,
|
| 2212 |
+
"loss": 0.3413,
|
| 2213 |
+
"rewards/accuracies": 0.887499988079071,
|
| 2214 |
+
"rewards/chosen": -0.953125,
|
| 2215 |
+
"rewards/margins": 2.53125,
|
| 2216 |
+
"rewards/rejected": -3.484375,
|
| 2217 |
+
"step": 630
|
| 2218 |
+
},
|
| 2219 |
+
{
|
| 2220 |
+
"epoch": 0.8863955119214586,
|
| 2221 |
+
"eval_logits/chosen": -3.09375,
|
| 2222 |
+
"eval_logits/rejected": -3.125,
|
| 2223 |
+
"eval_logps/chosen": -344.0,
|
| 2224 |
+
"eval_logps/rejected": -340.0,
|
| 2225 |
+
"eval_loss": 0.3731161057949066,
|
| 2226 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 2227 |
+
"eval_rewards/chosen": -1.1015625,
|
| 2228 |
+
"eval_rewards/margins": 2.09375,
|
| 2229 |
+
"eval_rewards/rejected": -3.203125,
|
| 2230 |
+
"eval_runtime": 66.117,
|
| 2231 |
+
"eval_samples_per_second": 22.49,
|
| 2232 |
+
"eval_steps_per_second": 0.711,
|
| 2233 |
+
"step": 632
|
| 2234 |
+
},
|
| 2235 |
+
{
|
| 2236 |
+
"epoch": 0.8976157082748948,
|
| 2237 |
+
"grad_norm": 31.05579865660738,
|
| 2238 |
+
"learning_rate": 1.583076076890963e-08,
|
| 2239 |
+
"logits/chosen": -3.109375,
|
| 2240 |
+
"logits/rejected": -3.109375,
|
| 2241 |
+
"logps/chosen": -360.0,
|
| 2242 |
+
"logps/rejected": -328.0,
|
| 2243 |
+
"loss": 0.3556,
|
| 2244 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 2245 |
+
"rewards/chosen": -1.171875,
|
| 2246 |
+
"rewards/margins": 2.21875,
|
| 2247 |
+
"rewards/rejected": -3.390625,
|
| 2248 |
+
"step": 640
|
| 2249 |
+
},
|
| 2250 |
+
{
|
| 2251 |
+
"epoch": 0.8976157082748948,
|
| 2252 |
+
"eval_logits/chosen": -3.09375,
|
| 2253 |
+
"eval_logits/rejected": -3.125,
|
| 2254 |
+
"eval_logps/chosen": -344.0,
|
| 2255 |
+
"eval_logps/rejected": -340.0,
|
| 2256 |
+
"eval_loss": 0.372738242149353,
|
| 2257 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 2258 |
+
"eval_rewards/chosen": -1.1015625,
|
| 2259 |
+
"eval_rewards/margins": 2.109375,
|
| 2260 |
+
"eval_rewards/rejected": -3.21875,
|
| 2261 |
+
"eval_runtime": 66.4596,
|
| 2262 |
+
"eval_samples_per_second": 22.374,
|
| 2263 |
+
"eval_steps_per_second": 0.707,
|
| 2264 |
+
"step": 640
|
| 2265 |
+
},
|
| 2266 |
+
{
|
| 2267 |
+
"epoch": 0.908835904628331,
|
| 2268 |
+
"eval_logits/chosen": -3.09375,
|
| 2269 |
+
"eval_logits/rejected": -3.125,
|
| 2270 |
+
"eval_logps/chosen": -344.0,
|
| 2271 |
+
"eval_logps/rejected": -340.0,
|
| 2272 |
+
"eval_loss": 0.37327107787132263,
|
| 2273 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 2274 |
+
"eval_rewards/chosen": -1.0625,
|
| 2275 |
+
"eval_rewards/margins": 2.125,
|
| 2276 |
+
"eval_rewards/rejected": -3.1875,
|
| 2277 |
+
"eval_runtime": 67.9699,
|
| 2278 |
+
"eval_samples_per_second": 21.877,
|
| 2279 |
+
"eval_steps_per_second": 0.691,
|
| 2280 |
+
"step": 648
|
| 2281 |
+
},
|
| 2282 |
+
{
|
| 2283 |
+
"epoch": 0.9116409537166901,
|
| 2284 |
+
"grad_norm": 42.559333930514235,
|
| 2285 |
+
"learning_rate": 1.1822836004992343e-08,
|
| 2286 |
+
"logits/chosen": -3.046875,
|
| 2287 |
+
"logits/rejected": -2.96875,
|
| 2288 |
+
"logps/chosen": -296.0,
|
| 2289 |
+
"logps/rejected": -324.0,
|
| 2290 |
+
"loss": 0.3569,
|
| 2291 |
+
"rewards/accuracies": 0.9375,
|
| 2292 |
+
"rewards/chosen": -1.4375,
|
| 2293 |
+
"rewards/margins": 2.3125,
|
| 2294 |
+
"rewards/rejected": -3.75,
|
| 2295 |
+
"step": 650
|
| 2296 |
+
},
|
| 2297 |
+
{
|
| 2298 |
+
"epoch": 0.9200561009817672,
|
| 2299 |
+
"eval_logits/chosen": -3.09375,
|
| 2300 |
+
"eval_logits/rejected": -3.125,
|
| 2301 |
+
"eval_logps/chosen": -342.0,
|
| 2302 |
+
"eval_logps/rejected": -338.0,
|
| 2303 |
+
"eval_loss": 0.3728487491607666,
|
| 2304 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 2305 |
+
"eval_rewards/chosen": -1.03125,
|
| 2306 |
+
"eval_rewards/margins": 2.140625,
|
| 2307 |
+
"eval_rewards/rejected": -3.171875,
|
| 2308 |
+
"eval_runtime": 66.1242,
|
| 2309 |
+
"eval_samples_per_second": 22.488,
|
| 2310 |
+
"eval_steps_per_second": 0.711,
|
| 2311 |
+
"step": 656
|
| 2312 |
+
},
|
| 2313 |
+
{
|
| 2314 |
+
"epoch": 0.9256661991584852,
|
| 2315 |
+
"grad_norm": 33.71718914979295,
|
| 2316 |
+
"learning_rate": 8.386912350262566e-09,
|
| 2317 |
+
"logits/chosen": -2.96875,
|
| 2318 |
+
"logits/rejected": -3.0,
|
| 2319 |
+
"logps/chosen": -352.0,
|
| 2320 |
+
"logps/rejected": -362.0,
|
| 2321 |
+
"loss": 0.3563,
|
| 2322 |
+
"rewards/accuracies": 0.862500011920929,
|
| 2323 |
+
"rewards/chosen": -1.2265625,
|
| 2324 |
+
"rewards/margins": 2.03125,
|
| 2325 |
+
"rewards/rejected": -3.265625,
|
| 2326 |
+
"step": 660
|
| 2327 |
+
},
|
| 2328 |
+
{
|
| 2329 |
+
"epoch": 0.9312762973352033,
|
| 2330 |
+
"eval_logits/chosen": -3.09375,
|
| 2331 |
+
"eval_logits/rejected": -3.125,
|
| 2332 |
+
"eval_logps/chosen": -342.0,
|
| 2333 |
+
"eval_logps/rejected": -338.0,
|
| 2334 |
+
"eval_loss": 0.37276628613471985,
|
| 2335 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 2336 |
+
"eval_rewards/chosen": -1.015625,
|
| 2337 |
+
"eval_rewards/margins": 2.140625,
|
| 2338 |
+
"eval_rewards/rejected": -3.15625,
|
| 2339 |
+
"eval_runtime": 66.0656,
|
| 2340 |
+
"eval_samples_per_second": 22.508,
|
| 2341 |
+
"eval_steps_per_second": 0.711,
|
| 2342 |
+
"step": 664
|
| 2343 |
+
},
|
| 2344 |
+
{
|
| 2345 |
+
"epoch": 0.9396914446002805,
|
| 2346 |
+
"grad_norm": 52.307264892507675,
|
| 2347 |
+
"learning_rate": 5.5312414444183276e-09,
|
| 2348 |
+
"logits/chosen": -3.046875,
|
| 2349 |
+
"logits/rejected": -2.859375,
|
| 2350 |
+
"logps/chosen": -354.0,
|
| 2351 |
+
"logps/rejected": -370.0,
|
| 2352 |
+
"loss": 0.3694,
|
| 2353 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 2354 |
+
"rewards/chosen": -1.109375,
|
| 2355 |
+
"rewards/margins": 2.3125,
|
| 2356 |
+
"rewards/rejected": -3.421875,
|
| 2357 |
+
"step": 670
|
| 2358 |
+
},
|
| 2359 |
+
{
|
| 2360 |
+
"epoch": 0.9424964936886395,
|
| 2361 |
+
"eval_logits/chosen": -3.09375,
|
| 2362 |
+
"eval_logits/rejected": -3.125,
|
| 2363 |
+
"eval_logps/chosen": -342.0,
|
| 2364 |
+
"eval_logps/rejected": -338.0,
|
| 2365 |
+
"eval_loss": 0.3723788857460022,
|
| 2366 |
+
"eval_rewards/accuracies": 0.8404255509376526,
|
| 2367 |
+
"eval_rewards/chosen": -1.0078125,
|
| 2368 |
+
"eval_rewards/margins": 2.140625,
|
| 2369 |
+
"eval_rewards/rejected": -3.15625,
|
| 2370 |
+
"eval_runtime": 65.8691,
|
| 2371 |
+
"eval_samples_per_second": 22.575,
|
| 2372 |
+
"eval_steps_per_second": 0.714,
|
| 2373 |
+
"step": 672
|
| 2374 |
+
},
|
| 2375 |
+
{
|
| 2376 |
+
"epoch": 0.9537166900420757,
|
| 2377 |
+
"grad_norm": 39.5471805424214,
|
| 2378 |
+
"learning_rate": 3.262681404746004e-09,
|
| 2379 |
+
"logits/chosen": -3.09375,
|
| 2380 |
+
"logits/rejected": -3.015625,
|
| 2381 |
+
"logps/chosen": -368.0,
|
| 2382 |
+
"logps/rejected": -368.0,
|
| 2383 |
+
"loss": 0.3447,
|
| 2384 |
+
"rewards/accuracies": 0.887499988079071,
|
| 2385 |
+
"rewards/chosen": -1.0546875,
|
| 2386 |
+
"rewards/margins": 2.671875,
|
| 2387 |
+
"rewards/rejected": -3.71875,
|
| 2388 |
+
"step": 680
|
| 2389 |
+
},
|
| 2390 |
+
{
|
| 2391 |
+
"epoch": 0.9537166900420757,
|
| 2392 |
+
"eval_logits/chosen": -3.09375,
|
| 2393 |
+
"eval_logits/rejected": -3.125,
|
| 2394 |
+
"eval_logps/chosen": -342.0,
|
| 2395 |
+
"eval_logps/rejected": -338.0,
|
| 2396 |
+
"eval_loss": 0.37281739711761475,
|
| 2397 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 2398 |
+
"eval_rewards/chosen": -1.015625,
|
| 2399 |
+
"eval_rewards/margins": 2.140625,
|
| 2400 |
+
"eval_rewards/rejected": -3.15625,
|
| 2401 |
+
"eval_runtime": 66.1705,
|
| 2402 |
+
"eval_samples_per_second": 22.472,
|
| 2403 |
+
"eval_steps_per_second": 0.71,
|
| 2404 |
+
"step": 680
|
| 2405 |
+
},
|
| 2406 |
+
{
|
| 2407 |
+
"epoch": 0.9649368863955119,
|
| 2408 |
+
"eval_logits/chosen": -3.09375,
|
| 2409 |
+
"eval_logits/rejected": -3.125,
|
| 2410 |
+
"eval_logps/chosen": -342.0,
|
| 2411 |
+
"eval_logps/rejected": -338.0,
|
| 2412 |
+
"eval_loss": 0.3726153075695038,
|
| 2413 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 2414 |
+
"eval_rewards/chosen": -1.0078125,
|
| 2415 |
+
"eval_rewards/margins": 2.140625,
|
| 2416 |
+
"eval_rewards/rejected": -3.15625,
|
| 2417 |
+
"eval_runtime": 66.3119,
|
| 2418 |
+
"eval_samples_per_second": 22.424,
|
| 2419 |
+
"eval_steps_per_second": 0.709,
|
| 2420 |
+
"step": 688
|
| 2421 |
+
},
|
| 2422 |
+
{
|
| 2423 |
+
"epoch": 0.967741935483871,
|
| 2424 |
+
"grad_norm": 40.36229007754571,
|
| 2425 |
+
"learning_rate": 1.5866803558146624e-09,
|
| 2426 |
+
"logits/chosen": -2.984375,
|
| 2427 |
+
"logits/rejected": -2.96875,
|
| 2428 |
+
"logps/chosen": -322.0,
|
| 2429 |
+
"logps/rejected": -372.0,
|
| 2430 |
+
"loss": 0.3766,
|
| 2431 |
+
"rewards/accuracies": 0.949999988079071,
|
| 2432 |
+
"rewards/chosen": -1.359375,
|
| 2433 |
+
"rewards/margins": 2.046875,
|
| 2434 |
+
"rewards/rejected": -3.40625,
|
| 2435 |
+
"step": 690
|
| 2436 |
+
},
|
| 2437 |
+
{
|
| 2438 |
+
"epoch": 0.9761570827489481,
|
| 2439 |
+
"eval_logits/chosen": -3.09375,
|
| 2440 |
+
"eval_logits/rejected": -3.125,
|
| 2441 |
+
"eval_logps/chosen": -342.0,
|
| 2442 |
+
"eval_logps/rejected": -338.0,
|
| 2443 |
+
"eval_loss": 0.3727646470069885,
|
| 2444 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 2445 |
+
"eval_rewards/chosen": -1.0078125,
|
| 2446 |
+
"eval_rewards/margins": 2.140625,
|
| 2447 |
+
"eval_rewards/rejected": -3.140625,
|
| 2448 |
+
"eval_runtime": 65.8839,
|
| 2449 |
+
"eval_samples_per_second": 22.57,
|
| 2450 |
+
"eval_steps_per_second": 0.713,
|
| 2451 |
+
"step": 696
|
| 2452 |
+
},
|
| 2453 |
+
{
|
| 2454 |
+
"epoch": 0.9817671809256662,
|
| 2455 |
+
"grad_norm": 41.008248288595254,
|
| 2456 |
+
"learning_rate": 5.07263345378317e-10,
|
| 2457 |
+
"logits/chosen": -2.96875,
|
| 2458 |
+
"logits/rejected": -3.015625,
|
| 2459 |
+
"logps/chosen": -346.0,
|
| 2460 |
+
"logps/rejected": -358.0,
|
| 2461 |
+
"loss": 0.3699,
|
| 2462 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 2463 |
+
"rewards/chosen": -0.8984375,
|
| 2464 |
+
"rewards/margins": 2.578125,
|
| 2465 |
+
"rewards/rejected": -3.484375,
|
| 2466 |
+
"step": 700
|
| 2467 |
+
},
|
| 2468 |
+
{
|
| 2469 |
+
"epoch": 0.9873772791023843,
|
| 2470 |
+
"eval_logits/chosen": -3.09375,
|
| 2471 |
+
"eval_logits/rejected": -3.125,
|
| 2472 |
+
"eval_logps/chosen": -342.0,
|
| 2473 |
+
"eval_logps/rejected": -338.0,
|
| 2474 |
+
"eval_loss": 0.372467964887619,
|
| 2475 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 2476 |
+
"eval_rewards/chosen": -1.015625,
|
| 2477 |
+
"eval_rewards/margins": 2.125,
|
| 2478 |
+
"eval_rewards/rejected": -3.140625,
|
| 2479 |
+
"eval_runtime": 65.8027,
|
| 2480 |
+
"eval_samples_per_second": 22.598,
|
| 2481 |
+
"eval_steps_per_second": 0.714,
|
| 2482 |
+
"step": 704
|
| 2483 |
+
},
|
| 2484 |
+
{
|
| 2485 |
+
"epoch": 0.9957924263674615,
|
| 2486 |
+
"grad_norm": 50.076560093112214,
|
| 2487 |
+
"learning_rate": 2.7022677908467016e-11,
|
| 2488 |
+
"logits/chosen": -3.03125,
|
| 2489 |
+
"logits/rejected": -3.015625,
|
| 2490 |
+
"logps/chosen": -394.0,
|
| 2491 |
+
"logps/rejected": -354.0,
|
| 2492 |
+
"loss": 0.3758,
|
| 2493 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 2494 |
+
"rewards/chosen": -1.1015625,
|
| 2495 |
+
"rewards/margins": 1.9296875,
|
| 2496 |
+
"rewards/rejected": -3.03125,
|
| 2497 |
+
"step": 710
|
| 2498 |
+
},
|
| 2499 |
+
{
|
| 2500 |
+
"epoch": 0.9985974754558204,
|
| 2501 |
+
"eval_logits/chosen": -3.09375,
|
| 2502 |
+
"eval_logits/rejected": -3.125,
|
| 2503 |
+
"eval_logps/chosen": -342.0,
|
| 2504 |
+
"eval_logps/rejected": -338.0,
|
| 2505 |
+
"eval_loss": 0.37290096282958984,
|
| 2506 |
+
"eval_rewards/accuracies": 0.835106372833252,
|
| 2507 |
+
"eval_rewards/chosen": -1.0078125,
|
| 2508 |
+
"eval_rewards/margins": 2.140625,
|
| 2509 |
+
"eval_rewards/rejected": -3.140625,
|
| 2510 |
+
"eval_runtime": 66.4035,
|
| 2511 |
+
"eval_samples_per_second": 22.393,
|
| 2512 |
+
"eval_steps_per_second": 0.708,
|
| 2513 |
+
"step": 712
|
| 2514 |
+
},
|
| 2515 |
+
{
|
| 2516 |
+
"epoch": 1.0,
|
| 2517 |
+
"step": 713,
|
| 2518 |
+
"total_flos": 0.0,
|
| 2519 |
+
"train_loss": 0.4091535410406212,
|
| 2520 |
+
"train_runtime": 30868.5922,
|
| 2521 |
+
"train_samples_per_second": 1.477,
|
| 2522 |
+
"train_steps_per_second": 0.023
|
| 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 |
+
}
|