Model save
Browse files- README.md +69 -0
- all_results.json +9 -0
- runs/Jan22_20-31-46_ip-172-31-34-75/events.out.tfevents.1737578036.ip-172-31-34-75.4941.0 +2 -2
- train_results.json +9 -0
- trainer_state.json +2262 -0
README.md
ADDED
@@ -0,0 +1,69 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
base_model: philschmid/llama-3-1-8b-math-orca-spectrum-10k-ep1
|
3 |
+
library_name: transformers
|
4 |
+
model_name: dpo-llama-3-1-8b-math-ep3
|
5 |
+
tags:
|
6 |
+
- generated_from_trainer
|
7 |
+
- trl
|
8 |
+
- dpo
|
9 |
+
licence: license
|
10 |
+
---
|
11 |
+
|
12 |
+
# Model Card for dpo-llama-3-1-8b-math-ep3
|
13 |
+
|
14 |
+
This model is a fine-tuned version of [philschmid/llama-3-1-8b-math-orca-spectrum-10k-ep1](https://huggingface.co/philschmid/llama-3-1-8b-math-orca-spectrum-10k-ep1).
|
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="philschmid/dpo-llama-3-1-8b-math-ep3", 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 |
+
|
31 |
+
|
32 |
+
|
33 |
+
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).
|
34 |
+
|
35 |
+
### Framework versions
|
36 |
+
|
37 |
+
- TRL: 0.14.0.dev0
|
38 |
+
- Transformers: 4.48.0
|
39 |
+
- Pytorch: 2.5.1
|
40 |
+
- Datasets: 3.1.0
|
41 |
+
- Tokenizers: 0.21.0
|
42 |
+
|
43 |
+
## Citations
|
44 |
+
|
45 |
+
Cite DPO as:
|
46 |
+
|
47 |
+
```bibtex
|
48 |
+
@inproceedings{rafailov2023direct,
|
49 |
+
title = {{Direct Preference Optimization: Your Language Model is Secretly a Reward Model}},
|
50 |
+
author = {Rafael Rafailov and Archit Sharma and Eric Mitchell and Christopher D. Manning and Stefano Ermon and Chelsea Finn},
|
51 |
+
year = 2023,
|
52 |
+
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},
|
53 |
+
url = {http://papers.nips.cc/paper_files/paper/2023/hash/a85b405ed65c6477a4fe8302b5e06ce7-Abstract-Conference.html},
|
54 |
+
editor = {Alice Oh and Tristan Naumann and Amir Globerson and Kate Saenko and Moritz Hardt and Sergey Levine},
|
55 |
+
}
|
56 |
+
```
|
57 |
+
|
58 |
+
Cite TRL as:
|
59 |
+
|
60 |
+
```bibtex
|
61 |
+
@misc{vonwerra2022trl,
|
62 |
+
title = {{TRL: Transformer Reinforcement Learning}},
|
63 |
+
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},
|
64 |
+
year = 2020,
|
65 |
+
journal = {GitHub repository},
|
66 |
+
publisher = {GitHub},
|
67 |
+
howpublished = {\url{https://github.com/huggingface/trl}}
|
68 |
+
}
|
69 |
+
```
|
all_results.json
ADDED
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 2.989399293286219,
|
3 |
+
"total_flos": 0.0,
|
4 |
+
"train_loss": 0.586526518971653,
|
5 |
+
"train_runtime": 26068.2965,
|
6 |
+
"train_samples": 1981,
|
7 |
+
"train_samples_per_second": 0.228,
|
8 |
+
"train_steps_per_second": 0.028
|
9 |
+
}
|
runs/Jan22_20-31-46_ip-172-31-34-75/events.out.tfevents.1737578036.ip-172-31-34-75.4941.0
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:5cfed40f4894076b59ba6ee487444284342f8ab3075a061278138549e0ae4de2
|
3 |
+
size 108707
|
train_results.json
ADDED
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 2.989399293286219,
|
3 |
+
"total_flos": 0.0,
|
4 |
+
"train_loss": 0.586526518971653,
|
5 |
+
"train_runtime": 26068.2965,
|
6 |
+
"train_samples": 1981,
|
7 |
+
"train_samples_per_second": 0.228,
|
8 |
+
"train_steps_per_second": 0.028
|
9 |
+
}
|
trainer_state.json
ADDED
@@ -0,0 +1,2262 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": null,
|
3 |
+
"best_model_checkpoint": null,
|
4 |
+
"epoch": 2.989399293286219,
|
5 |
+
"eval_steps": 500,
|
6 |
+
"global_step": 741,
|
7 |
+
"is_hyper_param_search": false,
|
8 |
+
"is_local_process_zero": true,
|
9 |
+
"is_world_process_zero": true,
|
10 |
+
"log_history": [
|
11 |
+
{
|
12 |
+
"epoch": 0.020191822311963654,
|
13 |
+
"grad_norm": 30.7789249420166,
|
14 |
+
"learning_rate": 5e-06,
|
15 |
+
"logits/chosen": -1.0492041110992432,
|
16 |
+
"logits/rejected": -1.0134608745574951,
|
17 |
+
"logps/chosen": -127.44047546386719,
|
18 |
+
"logps/rejected": -147.18359375,
|
19 |
+
"loss": 0.6931,
|
20 |
+
"rewards/accuracies": 0.4000000059604645,
|
21 |
+
"rewards/chosen": 0.009235936217010021,
|
22 |
+
"rewards/margins": 0.0008875288185663521,
|
23 |
+
"rewards/rejected": 0.008348408155143261,
|
24 |
+
"step": 5
|
25 |
+
},
|
26 |
+
{
|
27 |
+
"epoch": 0.04038364462392731,
|
28 |
+
"grad_norm": 37.41344451904297,
|
29 |
+
"learning_rate": 5e-06,
|
30 |
+
"logits/chosen": -1.002793312072754,
|
31 |
+
"logits/rejected": -0.9411100149154663,
|
32 |
+
"logps/chosen": -126.66816711425781,
|
33 |
+
"logps/rejected": -159.05479431152344,
|
34 |
+
"loss": 0.6969,
|
35 |
+
"rewards/accuracies": 0.375,
|
36 |
+
"rewards/chosen": 0.006721611134707928,
|
37 |
+
"rewards/margins": -0.006841087248176336,
|
38 |
+
"rewards/rejected": 0.013562696985900402,
|
39 |
+
"step": 10
|
40 |
+
},
|
41 |
+
{
|
42 |
+
"epoch": 0.06057546693589096,
|
43 |
+
"grad_norm": 42.148170471191406,
|
44 |
+
"learning_rate": 5e-06,
|
45 |
+
"logits/chosen": -1.0353584289550781,
|
46 |
+
"logits/rejected": -1.0312868356704712,
|
47 |
+
"logps/chosen": -145.601806640625,
|
48 |
+
"logps/rejected": -159.65481567382812,
|
49 |
+
"loss": 0.6928,
|
50 |
+
"rewards/accuracies": 0.42500001192092896,
|
51 |
+
"rewards/chosen": 0.02560463175177574,
|
52 |
+
"rewards/margins": 0.0021010604687035084,
|
53 |
+
"rewards/rejected": 0.023503568023443222,
|
54 |
+
"step": 15
|
55 |
+
},
|
56 |
+
{
|
57 |
+
"epoch": 0.08076728924785462,
|
58 |
+
"grad_norm": 58.297794342041016,
|
59 |
+
"learning_rate": 5e-06,
|
60 |
+
"logits/chosen": -1.053727388381958,
|
61 |
+
"logits/rejected": -1.0229905843734741,
|
62 |
+
"logps/chosen": -121.37943267822266,
|
63 |
+
"logps/rejected": -130.5764923095703,
|
64 |
+
"loss": 0.6857,
|
65 |
+
"rewards/accuracies": 0.574999988079071,
|
66 |
+
"rewards/chosen": 0.06899537146091461,
|
67 |
+
"rewards/margins": 0.015979710966348648,
|
68 |
+
"rewards/rejected": 0.053015656769275665,
|
69 |
+
"step": 20
|
70 |
+
},
|
71 |
+
{
|
72 |
+
"epoch": 0.10095911155981828,
|
73 |
+
"grad_norm": 158.1420440673828,
|
74 |
+
"learning_rate": 5e-06,
|
75 |
+
"logits/chosen": -1.097388744354248,
|
76 |
+
"logits/rejected": -1.0624990463256836,
|
77 |
+
"logps/chosen": -126.20062255859375,
|
78 |
+
"logps/rejected": -140.00369262695312,
|
79 |
+
"loss": 0.6968,
|
80 |
+
"rewards/accuracies": 0.44999998807907104,
|
81 |
+
"rewards/chosen": 0.040822066366672516,
|
82 |
+
"rewards/margins": -0.004479425959289074,
|
83 |
+
"rewards/rejected": 0.04530149698257446,
|
84 |
+
"step": 25
|
85 |
+
},
|
86 |
+
{
|
87 |
+
"epoch": 0.12115093387178193,
|
88 |
+
"grad_norm": 111.64271545410156,
|
89 |
+
"learning_rate": 5e-06,
|
90 |
+
"logits/chosen": -0.9921103715896606,
|
91 |
+
"logits/rejected": -0.9481407999992371,
|
92 |
+
"logps/chosen": -133.66769409179688,
|
93 |
+
"logps/rejected": -149.52188110351562,
|
94 |
+
"loss": 0.7066,
|
95 |
+
"rewards/accuracies": 0.44999998807907104,
|
96 |
+
"rewards/chosen": -0.17871007323265076,
|
97 |
+
"rewards/margins": -0.001442834734916687,
|
98 |
+
"rewards/rejected": -0.17726723849773407,
|
99 |
+
"step": 30
|
100 |
+
},
|
101 |
+
{
|
102 |
+
"epoch": 0.1413427561837456,
|
103 |
+
"grad_norm": 35.10946273803711,
|
104 |
+
"learning_rate": 5e-06,
|
105 |
+
"logits/chosen": -0.9996519088745117,
|
106 |
+
"logits/rejected": -1.0022809505462646,
|
107 |
+
"logps/chosen": -142.79452514648438,
|
108 |
+
"logps/rejected": -142.8091583251953,
|
109 |
+
"loss": 0.6933,
|
110 |
+
"rewards/accuracies": 0.5249999761581421,
|
111 |
+
"rewards/chosen": -0.1559412181377411,
|
112 |
+
"rewards/margins": 0.018214434385299683,
|
113 |
+
"rewards/rejected": -0.17415565252304077,
|
114 |
+
"step": 35
|
115 |
+
},
|
116 |
+
{
|
117 |
+
"epoch": 0.16153457849570924,
|
118 |
+
"grad_norm": 154.33160400390625,
|
119 |
+
"learning_rate": 5e-06,
|
120 |
+
"logits/chosen": -1.0090668201446533,
|
121 |
+
"logits/rejected": -0.9780625104904175,
|
122 |
+
"logps/chosen": -124.8656234741211,
|
123 |
+
"logps/rejected": -151.53759765625,
|
124 |
+
"loss": 0.6991,
|
125 |
+
"rewards/accuracies": 0.4000000059604645,
|
126 |
+
"rewards/chosen": -0.0610191710293293,
|
127 |
+
"rewards/margins": 0.001720429165288806,
|
128 |
+
"rewards/rejected": -0.06273959577083588,
|
129 |
+
"step": 40
|
130 |
+
},
|
131 |
+
{
|
132 |
+
"epoch": 0.18172640080767288,
|
133 |
+
"grad_norm": 66.54533386230469,
|
134 |
+
"learning_rate": 5e-06,
|
135 |
+
"logits/chosen": -1.0511558055877686,
|
136 |
+
"logits/rejected": -0.9915176630020142,
|
137 |
+
"logps/chosen": -125.61368560791016,
|
138 |
+
"logps/rejected": -140.61505126953125,
|
139 |
+
"loss": 0.6762,
|
140 |
+
"rewards/accuracies": 0.5,
|
141 |
+
"rewards/chosen": -0.11788681894540787,
|
142 |
+
"rewards/margins": 0.049356214702129364,
|
143 |
+
"rewards/rejected": -0.16724303364753723,
|
144 |
+
"step": 45
|
145 |
+
},
|
146 |
+
{
|
147 |
+
"epoch": 0.20191822311963656,
|
148 |
+
"grad_norm": 133.6022491455078,
|
149 |
+
"learning_rate": 5e-06,
|
150 |
+
"logits/chosen": -1.0434271097183228,
|
151 |
+
"logits/rejected": -1.0047496557235718,
|
152 |
+
"logps/chosen": -126.78190612792969,
|
153 |
+
"logps/rejected": -138.11805725097656,
|
154 |
+
"loss": 0.7026,
|
155 |
+
"rewards/accuracies": 0.375,
|
156 |
+
"rewards/chosen": -0.2226967066526413,
|
157 |
+
"rewards/margins": 0.006922038737684488,
|
158 |
+
"rewards/rejected": -0.22961874306201935,
|
159 |
+
"step": 50
|
160 |
+
},
|
161 |
+
{
|
162 |
+
"epoch": 0.2221100454316002,
|
163 |
+
"grad_norm": 61.76374053955078,
|
164 |
+
"learning_rate": 5e-06,
|
165 |
+
"logits/chosen": -1.026590347290039,
|
166 |
+
"logits/rejected": -1.0285896062850952,
|
167 |
+
"logps/chosen": -126.12849426269531,
|
168 |
+
"logps/rejected": -142.56619262695312,
|
169 |
+
"loss": 0.6729,
|
170 |
+
"rewards/accuracies": 0.6000000238418579,
|
171 |
+
"rewards/chosen": -0.26916319131851196,
|
172 |
+
"rewards/margins": 0.06655414402484894,
|
173 |
+
"rewards/rejected": -0.3357173502445221,
|
174 |
+
"step": 55
|
175 |
+
},
|
176 |
+
{
|
177 |
+
"epoch": 0.24230186774356385,
|
178 |
+
"grad_norm": 61.80793380737305,
|
179 |
+
"learning_rate": 5e-06,
|
180 |
+
"logits/chosen": -1.0803353786468506,
|
181 |
+
"logits/rejected": -1.0326142311096191,
|
182 |
+
"logps/chosen": -126.38145446777344,
|
183 |
+
"logps/rejected": -143.6183319091797,
|
184 |
+
"loss": 0.6796,
|
185 |
+
"rewards/accuracies": 0.574999988079071,
|
186 |
+
"rewards/chosen": -0.36116331815719604,
|
187 |
+
"rewards/margins": 0.07047157734632492,
|
188 |
+
"rewards/rejected": -0.43163490295410156,
|
189 |
+
"step": 60
|
190 |
+
},
|
191 |
+
{
|
192 |
+
"epoch": 0.26249369005552753,
|
193 |
+
"grad_norm": 63.28056716918945,
|
194 |
+
"learning_rate": 5e-06,
|
195 |
+
"logits/chosen": -0.9822731018066406,
|
196 |
+
"logits/rejected": -0.9656317830085754,
|
197 |
+
"logps/chosen": -127.26124572753906,
|
198 |
+
"logps/rejected": -142.61508178710938,
|
199 |
+
"loss": 0.6992,
|
200 |
+
"rewards/accuracies": 0.5249999761581421,
|
201 |
+
"rewards/chosen": -0.36886388063430786,
|
202 |
+
"rewards/margins": 0.059492867439985275,
|
203 |
+
"rewards/rejected": -0.42835673689842224,
|
204 |
+
"step": 65
|
205 |
+
},
|
206 |
+
{
|
207 |
+
"epoch": 0.2826855123674912,
|
208 |
+
"grad_norm": 40.66404724121094,
|
209 |
+
"learning_rate": 5e-06,
|
210 |
+
"logits/chosen": -1.0611032247543335,
|
211 |
+
"logits/rejected": -0.9986160397529602,
|
212 |
+
"logps/chosen": -119.45987701416016,
|
213 |
+
"logps/rejected": -137.6642303466797,
|
214 |
+
"loss": 0.635,
|
215 |
+
"rewards/accuracies": 0.6499999761581421,
|
216 |
+
"rewards/chosen": -0.24067869782447815,
|
217 |
+
"rewards/margins": 0.19307085871696472,
|
218 |
+
"rewards/rejected": -0.4337495267391205,
|
219 |
+
"step": 70
|
220 |
+
},
|
221 |
+
{
|
222 |
+
"epoch": 0.3028773346794548,
|
223 |
+
"grad_norm": 141.08375549316406,
|
224 |
+
"learning_rate": 5e-06,
|
225 |
+
"logits/chosen": -1.0468595027923584,
|
226 |
+
"logits/rejected": -1.0398485660552979,
|
227 |
+
"logps/chosen": -123.22066497802734,
|
228 |
+
"logps/rejected": -137.7001190185547,
|
229 |
+
"loss": 0.6822,
|
230 |
+
"rewards/accuracies": 0.5249999761581421,
|
231 |
+
"rewards/chosen": -0.32747992873191833,
|
232 |
+
"rewards/margins": 0.04724354296922684,
|
233 |
+
"rewards/rejected": -0.37472349405288696,
|
234 |
+
"step": 75
|
235 |
+
},
|
236 |
+
{
|
237 |
+
"epoch": 0.32306915699141847,
|
238 |
+
"grad_norm": 76.24655151367188,
|
239 |
+
"learning_rate": 5e-06,
|
240 |
+
"logits/chosen": -1.0108659267425537,
|
241 |
+
"logits/rejected": -0.9696664810180664,
|
242 |
+
"logps/chosen": -132.04534912109375,
|
243 |
+
"logps/rejected": -134.4063262939453,
|
244 |
+
"loss": 0.7582,
|
245 |
+
"rewards/accuracies": 0.44999998807907104,
|
246 |
+
"rewards/chosen": -0.3479716181755066,
|
247 |
+
"rewards/margins": -0.06882990151643753,
|
248 |
+
"rewards/rejected": -0.27914175391197205,
|
249 |
+
"step": 80
|
250 |
+
},
|
251 |
+
{
|
252 |
+
"epoch": 0.3432609793033821,
|
253 |
+
"grad_norm": 294.6365661621094,
|
254 |
+
"learning_rate": 5e-06,
|
255 |
+
"logits/chosen": -1.0133106708526611,
|
256 |
+
"logits/rejected": -1.0042521953582764,
|
257 |
+
"logps/chosen": -134.86880493164062,
|
258 |
+
"logps/rejected": -145.12660217285156,
|
259 |
+
"loss": 0.7659,
|
260 |
+
"rewards/accuracies": 0.5,
|
261 |
+
"rewards/chosen": -0.3877476453781128,
|
262 |
+
"rewards/margins": -0.0979350134730339,
|
263 |
+
"rewards/rejected": -0.2898126542568207,
|
264 |
+
"step": 85
|
265 |
+
},
|
266 |
+
{
|
267 |
+
"epoch": 0.36345280161534577,
|
268 |
+
"grad_norm": 48.36005783081055,
|
269 |
+
"learning_rate": 5e-06,
|
270 |
+
"logits/chosen": -1.0785777568817139,
|
271 |
+
"logits/rejected": -1.0150744915008545,
|
272 |
+
"logps/chosen": -125.0882339477539,
|
273 |
+
"logps/rejected": -134.69866943359375,
|
274 |
+
"loss": 0.7113,
|
275 |
+
"rewards/accuracies": 0.5,
|
276 |
+
"rewards/chosen": -0.36325889825820923,
|
277 |
+
"rewards/margins": 0.047263432294130325,
|
278 |
+
"rewards/rejected": -0.41052237153053284,
|
279 |
+
"step": 90
|
280 |
+
},
|
281 |
+
{
|
282 |
+
"epoch": 0.3836446239273094,
|
283 |
+
"grad_norm": 134.56393432617188,
|
284 |
+
"learning_rate": 5e-06,
|
285 |
+
"logits/chosen": -1.0220086574554443,
|
286 |
+
"logits/rejected": -1.0256017446517944,
|
287 |
+
"logps/chosen": -136.06663513183594,
|
288 |
+
"logps/rejected": -141.03758239746094,
|
289 |
+
"loss": 0.7582,
|
290 |
+
"rewards/accuracies": 0.5,
|
291 |
+
"rewards/chosen": -0.5725785493850708,
|
292 |
+
"rewards/margins": -0.02296329475939274,
|
293 |
+
"rewards/rejected": -0.549615204334259,
|
294 |
+
"step": 95
|
295 |
+
},
|
296 |
+
{
|
297 |
+
"epoch": 0.4038364462392731,
|
298 |
+
"grad_norm": 192.60671997070312,
|
299 |
+
"learning_rate": 5e-06,
|
300 |
+
"logits/chosen": -1.029658555984497,
|
301 |
+
"logits/rejected": -1.015587568283081,
|
302 |
+
"logps/chosen": -135.1102752685547,
|
303 |
+
"logps/rejected": -132.73965454101562,
|
304 |
+
"loss": 0.7576,
|
305 |
+
"rewards/accuracies": 0.574999988079071,
|
306 |
+
"rewards/chosen": -0.6215102076530457,
|
307 |
+
"rewards/margins": -0.022445138543844223,
|
308 |
+
"rewards/rejected": -0.5990650653839111,
|
309 |
+
"step": 100
|
310 |
+
},
|
311 |
+
{
|
312 |
+
"epoch": 0.42402826855123676,
|
313 |
+
"grad_norm": 42.97038269042969,
|
314 |
+
"learning_rate": 5e-06,
|
315 |
+
"logits/chosen": -1.01835036277771,
|
316 |
+
"logits/rejected": -0.9691454768180847,
|
317 |
+
"logps/chosen": -127.7636489868164,
|
318 |
+
"logps/rejected": -131.01290893554688,
|
319 |
+
"loss": 0.8087,
|
320 |
+
"rewards/accuracies": 0.550000011920929,
|
321 |
+
"rewards/chosen": -0.43060892820358276,
|
322 |
+
"rewards/margins": -0.1305285394191742,
|
323 |
+
"rewards/rejected": -0.30008038878440857,
|
324 |
+
"step": 105
|
325 |
+
},
|
326 |
+
{
|
327 |
+
"epoch": 0.4442200908632004,
|
328 |
+
"grad_norm": 73.89049530029297,
|
329 |
+
"learning_rate": 5e-06,
|
330 |
+
"logits/chosen": -1.0535072088241577,
|
331 |
+
"logits/rejected": -1.0358234643936157,
|
332 |
+
"logps/chosen": -131.6114959716797,
|
333 |
+
"logps/rejected": -148.74351501464844,
|
334 |
+
"loss": 0.6739,
|
335 |
+
"rewards/accuracies": 0.6000000238418579,
|
336 |
+
"rewards/chosen": -0.319467693567276,
|
337 |
+
"rewards/margins": 0.11042511463165283,
|
338 |
+
"rewards/rejected": -0.4298928380012512,
|
339 |
+
"step": 110
|
340 |
+
},
|
341 |
+
{
|
342 |
+
"epoch": 0.46441191317516406,
|
343 |
+
"grad_norm": 261.6590881347656,
|
344 |
+
"learning_rate": 5e-06,
|
345 |
+
"logits/chosen": -1.039872407913208,
|
346 |
+
"logits/rejected": -1.061729073524475,
|
347 |
+
"logps/chosen": -133.2412567138672,
|
348 |
+
"logps/rejected": -141.62680053710938,
|
349 |
+
"loss": 0.8194,
|
350 |
+
"rewards/accuracies": 0.44999998807907104,
|
351 |
+
"rewards/chosen": -0.5433087348937988,
|
352 |
+
"rewards/margins": -0.15157613158226013,
|
353 |
+
"rewards/rejected": -0.3917326033115387,
|
354 |
+
"step": 115
|
355 |
+
},
|
356 |
+
{
|
357 |
+
"epoch": 0.4846037354871277,
|
358 |
+
"grad_norm": 153.96530151367188,
|
359 |
+
"learning_rate": 5e-06,
|
360 |
+
"logits/chosen": -1.0134645700454712,
|
361 |
+
"logits/rejected": -1.0050514936447144,
|
362 |
+
"logps/chosen": -140.56578063964844,
|
363 |
+
"logps/rejected": -146.6461944580078,
|
364 |
+
"loss": 0.7222,
|
365 |
+
"rewards/accuracies": 0.4749999940395355,
|
366 |
+
"rewards/chosen": -0.4193994104862213,
|
367 |
+
"rewards/margins": -0.014027280732989311,
|
368 |
+
"rewards/rejected": -0.40537214279174805,
|
369 |
+
"step": 120
|
370 |
+
},
|
371 |
+
{
|
372 |
+
"epoch": 0.5047955577990914,
|
373 |
+
"grad_norm": 92.5348129272461,
|
374 |
+
"learning_rate": 5e-06,
|
375 |
+
"logits/chosen": -1.0472946166992188,
|
376 |
+
"logits/rejected": -1.0149815082550049,
|
377 |
+
"logps/chosen": -126.81514739990234,
|
378 |
+
"logps/rejected": -138.25100708007812,
|
379 |
+
"loss": 0.641,
|
380 |
+
"rewards/accuracies": 0.6499999761581421,
|
381 |
+
"rewards/chosen": -0.2987712025642395,
|
382 |
+
"rewards/margins": 0.14457064867019653,
|
383 |
+
"rewards/rejected": -0.44334182143211365,
|
384 |
+
"step": 125
|
385 |
+
},
|
386 |
+
{
|
387 |
+
"epoch": 0.5249873801110551,
|
388 |
+
"grad_norm": 69.95176696777344,
|
389 |
+
"learning_rate": 5e-06,
|
390 |
+
"logits/chosen": -0.9726712107658386,
|
391 |
+
"logits/rejected": -0.9396511316299438,
|
392 |
+
"logps/chosen": -134.06784057617188,
|
393 |
+
"logps/rejected": -149.26393127441406,
|
394 |
+
"loss": 0.6874,
|
395 |
+
"rewards/accuracies": 0.574999988079071,
|
396 |
+
"rewards/chosen": -0.4227423071861267,
|
397 |
+
"rewards/margins": 0.07707767188549042,
|
398 |
+
"rewards/rejected": -0.4998199939727783,
|
399 |
+
"step": 130
|
400 |
+
},
|
401 |
+
{
|
402 |
+
"epoch": 0.5451792024230186,
|
403 |
+
"grad_norm": 59.02754592895508,
|
404 |
+
"learning_rate": 5e-06,
|
405 |
+
"logits/chosen": -1.0336484909057617,
|
406 |
+
"logits/rejected": -1.0117156505584717,
|
407 |
+
"logps/chosen": -123.12812805175781,
|
408 |
+
"logps/rejected": -138.68785095214844,
|
409 |
+
"loss": 0.6972,
|
410 |
+
"rewards/accuracies": 0.574999988079071,
|
411 |
+
"rewards/chosen": -0.2850368916988373,
|
412 |
+
"rewards/margins": 0.08022015541791916,
|
413 |
+
"rewards/rejected": -0.3652570843696594,
|
414 |
+
"step": 135
|
415 |
+
},
|
416 |
+
{
|
417 |
+
"epoch": 0.5653710247349824,
|
418 |
+
"grad_norm": 164.89817810058594,
|
419 |
+
"learning_rate": 5e-06,
|
420 |
+
"logits/chosen": -0.9582400321960449,
|
421 |
+
"logits/rejected": -0.9342845678329468,
|
422 |
+
"logps/chosen": -136.7896270751953,
|
423 |
+
"logps/rejected": -144.94981384277344,
|
424 |
+
"loss": 0.8658,
|
425 |
+
"rewards/accuracies": 0.5,
|
426 |
+
"rewards/chosen": -0.5358542799949646,
|
427 |
+
"rewards/margins": -0.13312572240829468,
|
428 |
+
"rewards/rejected": -0.40272849798202515,
|
429 |
+
"step": 140
|
430 |
+
},
|
431 |
+
{
|
432 |
+
"epoch": 0.585562847046946,
|
433 |
+
"grad_norm": 273.0749816894531,
|
434 |
+
"learning_rate": 5e-06,
|
435 |
+
"logits/chosen": -1.0799394845962524,
|
436 |
+
"logits/rejected": -1.0522867441177368,
|
437 |
+
"logps/chosen": -133.341796875,
|
438 |
+
"logps/rejected": -137.64601135253906,
|
439 |
+
"loss": 0.7,
|
440 |
+
"rewards/accuracies": 0.550000011920929,
|
441 |
+
"rewards/chosen": -0.5455681681632996,
|
442 |
+
"rewards/margins": 0.1389351338148117,
|
443 |
+
"rewards/rejected": -0.6845033168792725,
|
444 |
+
"step": 145
|
445 |
+
},
|
446 |
+
{
|
447 |
+
"epoch": 0.6057546693589096,
|
448 |
+
"grad_norm": 58.67772674560547,
|
449 |
+
"learning_rate": 5e-06,
|
450 |
+
"logits/chosen": -1.0716097354888916,
|
451 |
+
"logits/rejected": -1.0067239999771118,
|
452 |
+
"logps/chosen": -131.79644775390625,
|
453 |
+
"logps/rejected": -145.97897338867188,
|
454 |
+
"loss": 0.6711,
|
455 |
+
"rewards/accuracies": 0.550000011920929,
|
456 |
+
"rewards/chosen": -0.48813509941101074,
|
457 |
+
"rewards/margins": 0.15445764362812042,
|
458 |
+
"rewards/rejected": -0.64259272813797,
|
459 |
+
"step": 150
|
460 |
+
},
|
461 |
+
{
|
462 |
+
"epoch": 0.6259464916708734,
|
463 |
+
"grad_norm": 153.71763610839844,
|
464 |
+
"learning_rate": 5e-06,
|
465 |
+
"logits/chosen": -0.9819652438163757,
|
466 |
+
"logits/rejected": -0.975143551826477,
|
467 |
+
"logps/chosen": -131.63906860351562,
|
468 |
+
"logps/rejected": -136.08070373535156,
|
469 |
+
"loss": 0.6992,
|
470 |
+
"rewards/accuracies": 0.550000011920929,
|
471 |
+
"rewards/chosen": -0.511472761631012,
|
472 |
+
"rewards/margins": 0.09871827065944672,
|
473 |
+
"rewards/rejected": -0.6101909875869751,
|
474 |
+
"step": 155
|
475 |
+
},
|
476 |
+
{
|
477 |
+
"epoch": 0.6461383139828369,
|
478 |
+
"grad_norm": 140.51803588867188,
|
479 |
+
"learning_rate": 5e-06,
|
480 |
+
"logits/chosen": -0.9384062886238098,
|
481 |
+
"logits/rejected": -0.9456030130386353,
|
482 |
+
"logps/chosen": -137.96107482910156,
|
483 |
+
"logps/rejected": -148.41708374023438,
|
484 |
+
"loss": 0.6837,
|
485 |
+
"rewards/accuracies": 0.574999988079071,
|
486 |
+
"rewards/chosen": -0.5771997570991516,
|
487 |
+
"rewards/margins": 0.1888493001461029,
|
488 |
+
"rewards/rejected": -0.7660490274429321,
|
489 |
+
"step": 160
|
490 |
+
},
|
491 |
+
{
|
492 |
+
"epoch": 0.6663301362948006,
|
493 |
+
"grad_norm": 115.60352325439453,
|
494 |
+
"learning_rate": 5e-06,
|
495 |
+
"logits/chosen": -1.032671332359314,
|
496 |
+
"logits/rejected": -1.032019853591919,
|
497 |
+
"logps/chosen": -136.2175750732422,
|
498 |
+
"logps/rejected": -139.32192993164062,
|
499 |
+
"loss": 0.7872,
|
500 |
+
"rewards/accuracies": 0.4749999940395355,
|
501 |
+
"rewards/chosen": -0.6546095609664917,
|
502 |
+
"rewards/margins": -0.08018063008785248,
|
503 |
+
"rewards/rejected": -0.5744289755821228,
|
504 |
+
"step": 165
|
505 |
+
},
|
506 |
+
{
|
507 |
+
"epoch": 0.6865219586067642,
|
508 |
+
"grad_norm": 79.54869079589844,
|
509 |
+
"learning_rate": 5e-06,
|
510 |
+
"logits/chosen": -1.0461210012435913,
|
511 |
+
"logits/rejected": -1.001206398010254,
|
512 |
+
"logps/chosen": -133.0095672607422,
|
513 |
+
"logps/rejected": -142.75643920898438,
|
514 |
+
"loss": 0.6265,
|
515 |
+
"rewards/accuracies": 0.675000011920929,
|
516 |
+
"rewards/chosen": -0.38921013474464417,
|
517 |
+
"rewards/margins": 0.21428516507148743,
|
518 |
+
"rewards/rejected": -0.6034952998161316,
|
519 |
+
"step": 170
|
520 |
+
},
|
521 |
+
{
|
522 |
+
"epoch": 0.7067137809187279,
|
523 |
+
"grad_norm": 124.25992584228516,
|
524 |
+
"learning_rate": 5e-06,
|
525 |
+
"logits/chosen": -1.0024333000183105,
|
526 |
+
"logits/rejected": -1.0128402709960938,
|
527 |
+
"logps/chosen": -134.9090576171875,
|
528 |
+
"logps/rejected": -141.35768127441406,
|
529 |
+
"loss": 0.6573,
|
530 |
+
"rewards/accuracies": 0.625,
|
531 |
+
"rewards/chosen": -0.43251633644104004,
|
532 |
+
"rewards/margins": 0.22649070620536804,
|
533 |
+
"rewards/rejected": -0.6590070128440857,
|
534 |
+
"step": 175
|
535 |
+
},
|
536 |
+
{
|
537 |
+
"epoch": 0.7269056032306915,
|
538 |
+
"grad_norm": 90.11141204833984,
|
539 |
+
"learning_rate": 5e-06,
|
540 |
+
"logits/chosen": -1.0318266153335571,
|
541 |
+
"logits/rejected": -0.9781469106674194,
|
542 |
+
"logps/chosen": -128.78518676757812,
|
543 |
+
"logps/rejected": -147.6506805419922,
|
544 |
+
"loss": 0.7165,
|
545 |
+
"rewards/accuracies": 0.550000011920929,
|
546 |
+
"rewards/chosen": -0.4871068596839905,
|
547 |
+
"rewards/margins": 0.07786975055932999,
|
548 |
+
"rewards/rejected": -0.5649765729904175,
|
549 |
+
"step": 180
|
550 |
+
},
|
551 |
+
{
|
552 |
+
"epoch": 0.7470974255426552,
|
553 |
+
"grad_norm": 199.0548858642578,
|
554 |
+
"learning_rate": 5e-06,
|
555 |
+
"logits/chosen": -0.984387993812561,
|
556 |
+
"logits/rejected": -0.9830384254455566,
|
557 |
+
"logps/chosen": -132.85769653320312,
|
558 |
+
"logps/rejected": -133.11199951171875,
|
559 |
+
"loss": 0.8698,
|
560 |
+
"rewards/accuracies": 0.4000000059604645,
|
561 |
+
"rewards/chosen": -0.5072764158248901,
|
562 |
+
"rewards/margins": -0.2219184935092926,
|
563 |
+
"rewards/rejected": -0.28535789251327515,
|
564 |
+
"step": 185
|
565 |
+
},
|
566 |
+
{
|
567 |
+
"epoch": 0.7672892478546188,
|
568 |
+
"grad_norm": 260.7405090332031,
|
569 |
+
"learning_rate": 5e-06,
|
570 |
+
"logits/chosen": -1.0293523073196411,
|
571 |
+
"logits/rejected": -1.0172580480575562,
|
572 |
+
"logps/chosen": -126.0751724243164,
|
573 |
+
"logps/rejected": -141.263671875,
|
574 |
+
"loss": 0.6471,
|
575 |
+
"rewards/accuracies": 0.6499999761581421,
|
576 |
+
"rewards/chosen": -0.25228002667427063,
|
577 |
+
"rewards/margins": 0.1569499969482422,
|
578 |
+
"rewards/rejected": -0.4092300534248352,
|
579 |
+
"step": 190
|
580 |
+
},
|
581 |
+
{
|
582 |
+
"epoch": 0.7874810701665825,
|
583 |
+
"grad_norm": 41.43046569824219,
|
584 |
+
"learning_rate": 5e-06,
|
585 |
+
"logits/chosen": -1.071702480316162,
|
586 |
+
"logits/rejected": -1.0639045238494873,
|
587 |
+
"logps/chosen": -124.64006042480469,
|
588 |
+
"logps/rejected": -129.1591796875,
|
589 |
+
"loss": 0.6881,
|
590 |
+
"rewards/accuracies": 0.5,
|
591 |
+
"rewards/chosen": -0.2144135981798172,
|
592 |
+
"rewards/margins": 0.07804916799068451,
|
593 |
+
"rewards/rejected": -0.2924627363681793,
|
594 |
+
"step": 195
|
595 |
+
},
|
596 |
+
{
|
597 |
+
"epoch": 0.8076728924785462,
|
598 |
+
"grad_norm": 35.1358528137207,
|
599 |
+
"learning_rate": 5e-06,
|
600 |
+
"logits/chosen": -0.9687834978103638,
|
601 |
+
"logits/rejected": -0.954856276512146,
|
602 |
+
"logps/chosen": -120.9579849243164,
|
603 |
+
"logps/rejected": -130.809814453125,
|
604 |
+
"loss": 0.7126,
|
605 |
+
"rewards/accuracies": 0.550000011920929,
|
606 |
+
"rewards/chosen": -0.27170878648757935,
|
607 |
+
"rewards/margins": 0.005672874394804239,
|
608 |
+
"rewards/rejected": -0.27738165855407715,
|
609 |
+
"step": 200
|
610 |
+
},
|
611 |
+
{
|
612 |
+
"epoch": 0.8278647147905098,
|
613 |
+
"grad_norm": 27.78827667236328,
|
614 |
+
"learning_rate": 5e-06,
|
615 |
+
"logits/chosen": -1.048870325088501,
|
616 |
+
"logits/rejected": -1.0687897205352783,
|
617 |
+
"logps/chosen": -128.82345581054688,
|
618 |
+
"logps/rejected": -132.6963348388672,
|
619 |
+
"loss": 0.7058,
|
620 |
+
"rewards/accuracies": 0.5249999761581421,
|
621 |
+
"rewards/chosen": -0.2189714014530182,
|
622 |
+
"rewards/margins": 0.05474821850657463,
|
623 |
+
"rewards/rejected": -0.2737196087837219,
|
624 |
+
"step": 205
|
625 |
+
},
|
626 |
+
{
|
627 |
+
"epoch": 0.8480565371024735,
|
628 |
+
"grad_norm": 101.58830261230469,
|
629 |
+
"learning_rate": 5e-06,
|
630 |
+
"logits/chosen": -1.0325753688812256,
|
631 |
+
"logits/rejected": -1.0432530641555786,
|
632 |
+
"logps/chosen": -126.06434631347656,
|
633 |
+
"logps/rejected": -136.6858673095703,
|
634 |
+
"loss": 0.7424,
|
635 |
+
"rewards/accuracies": 0.42500001192092896,
|
636 |
+
"rewards/chosen": -0.38726571202278137,
|
637 |
+
"rewards/margins": -0.062184691429138184,
|
638 |
+
"rewards/rejected": -0.3250810503959656,
|
639 |
+
"step": 210
|
640 |
+
},
|
641 |
+
{
|
642 |
+
"epoch": 0.8682483594144371,
|
643 |
+
"grad_norm": 145.8494110107422,
|
644 |
+
"learning_rate": 5e-06,
|
645 |
+
"logits/chosen": -1.0436766147613525,
|
646 |
+
"logits/rejected": -1.001501441001892,
|
647 |
+
"logps/chosen": -128.81761169433594,
|
648 |
+
"logps/rejected": -146.7082977294922,
|
649 |
+
"loss": 0.649,
|
650 |
+
"rewards/accuracies": 0.6499999761581421,
|
651 |
+
"rewards/chosen": -0.27072566747665405,
|
652 |
+
"rewards/margins": 0.1860719621181488,
|
653 |
+
"rewards/rejected": -0.45679759979248047,
|
654 |
+
"step": 215
|
655 |
+
},
|
656 |
+
{
|
657 |
+
"epoch": 0.8884401817264008,
|
658 |
+
"grad_norm": 122.89201354980469,
|
659 |
+
"learning_rate": 5e-06,
|
660 |
+
"logits/chosen": -0.9502097964286804,
|
661 |
+
"logits/rejected": -0.9250322580337524,
|
662 |
+
"logps/chosen": -136.70542907714844,
|
663 |
+
"logps/rejected": -145.50021362304688,
|
664 |
+
"loss": 0.7171,
|
665 |
+
"rewards/accuracies": 0.5249999761581421,
|
666 |
+
"rewards/chosen": -0.4370450973510742,
|
667 |
+
"rewards/margins": 0.014479175209999084,
|
668 |
+
"rewards/rejected": -0.4515243172645569,
|
669 |
+
"step": 220
|
670 |
+
},
|
671 |
+
{
|
672 |
+
"epoch": 0.9086320040383644,
|
673 |
+
"grad_norm": 136.3062286376953,
|
674 |
+
"learning_rate": 5e-06,
|
675 |
+
"logits/chosen": -1.0122795104980469,
|
676 |
+
"logits/rejected": -0.976104736328125,
|
677 |
+
"logps/chosen": -130.60975646972656,
|
678 |
+
"logps/rejected": -140.47096252441406,
|
679 |
+
"loss": 0.7608,
|
680 |
+
"rewards/accuracies": 0.4749999940395355,
|
681 |
+
"rewards/chosen": -0.3417353332042694,
|
682 |
+
"rewards/margins": -0.07229948043823242,
|
683 |
+
"rewards/rejected": -0.269435852766037,
|
684 |
+
"step": 225
|
685 |
+
},
|
686 |
+
{
|
687 |
+
"epoch": 0.9288238263503281,
|
688 |
+
"grad_norm": 274.8150329589844,
|
689 |
+
"learning_rate": 5e-06,
|
690 |
+
"logits/chosen": -0.9934455156326294,
|
691 |
+
"logits/rejected": -0.9960210919380188,
|
692 |
+
"logps/chosen": -138.07879638671875,
|
693 |
+
"logps/rejected": -148.25611877441406,
|
694 |
+
"loss": 0.7518,
|
695 |
+
"rewards/accuracies": 0.44999998807907104,
|
696 |
+
"rewards/chosen": -0.3778955340385437,
|
697 |
+
"rewards/margins": -0.06377332657575607,
|
698 |
+
"rewards/rejected": -0.3141222298145294,
|
699 |
+
"step": 230
|
700 |
+
},
|
701 |
+
{
|
702 |
+
"epoch": 0.9490156486622918,
|
703 |
+
"grad_norm": 362.4608154296875,
|
704 |
+
"learning_rate": 5e-06,
|
705 |
+
"logits/chosen": -1.028414249420166,
|
706 |
+
"logits/rejected": -1.0202635526657104,
|
707 |
+
"logps/chosen": -126.71832275390625,
|
708 |
+
"logps/rejected": -131.26434326171875,
|
709 |
+
"loss": 0.7528,
|
710 |
+
"rewards/accuracies": 0.4000000059604645,
|
711 |
+
"rewards/chosen": -0.10429117828607559,
|
712 |
+
"rewards/margins": -0.09285052120685577,
|
713 |
+
"rewards/rejected": -0.011440658941864967,
|
714 |
+
"step": 235
|
715 |
+
},
|
716 |
+
{
|
717 |
+
"epoch": 0.9692074709742554,
|
718 |
+
"grad_norm": 200.3191375732422,
|
719 |
+
"learning_rate": 5e-06,
|
720 |
+
"logits/chosen": -1.0229647159576416,
|
721 |
+
"logits/rejected": -0.9428025484085083,
|
722 |
+
"logps/chosen": -126.55399322509766,
|
723 |
+
"logps/rejected": -151.26348876953125,
|
724 |
+
"loss": 0.6815,
|
725 |
+
"rewards/accuracies": 0.550000011920929,
|
726 |
+
"rewards/chosen": 0.17346206307411194,
|
727 |
+
"rewards/margins": 0.028347322717308998,
|
728 |
+
"rewards/rejected": 0.1451147496700287,
|
729 |
+
"step": 240
|
730 |
+
},
|
731 |
+
{
|
732 |
+
"epoch": 0.9893992932862191,
|
733 |
+
"grad_norm": 34.3711051940918,
|
734 |
+
"learning_rate": 5e-06,
|
735 |
+
"logits/chosen": -1.0116575956344604,
|
736 |
+
"logits/rejected": -1.0087472200393677,
|
737 |
+
"logps/chosen": -122.85481262207031,
|
738 |
+
"logps/rejected": -133.4599609375,
|
739 |
+
"loss": 0.6786,
|
740 |
+
"rewards/accuracies": 0.699999988079071,
|
741 |
+
"rewards/chosen": 0.08724311739206314,
|
742 |
+
"rewards/margins": 0.0308841522783041,
|
743 |
+
"rewards/rejected": 0.05635897070169449,
|
744 |
+
"step": 245
|
745 |
+
},
|
746 |
+
{
|
747 |
+
"epoch": 1.0080767289247854,
|
748 |
+
"grad_norm": 27.31838607788086,
|
749 |
+
"learning_rate": 5e-06,
|
750 |
+
"logits/chosen": -1.0463143587112427,
|
751 |
+
"logits/rejected": -0.9973466396331787,
|
752 |
+
"logps/chosen": -124.2419204711914,
|
753 |
+
"logps/rejected": -139.740966796875,
|
754 |
+
"loss": 0.6114,
|
755 |
+
"rewards/accuracies": 0.5945945978164673,
|
756 |
+
"rewards/chosen": 0.005640762392431498,
|
757 |
+
"rewards/margins": 0.07190810143947601,
|
758 |
+
"rewards/rejected": -0.06626733392477036,
|
759 |
+
"step": 250
|
760 |
+
},
|
761 |
+
{
|
762 |
+
"epoch": 1.028268551236749,
|
763 |
+
"grad_norm": 36.600242614746094,
|
764 |
+
"learning_rate": 5e-06,
|
765 |
+
"logits/chosen": -0.9850190877914429,
|
766 |
+
"logits/rejected": -0.9895715713500977,
|
767 |
+
"logps/chosen": -117.77237701416016,
|
768 |
+
"logps/rejected": -139.15866088867188,
|
769 |
+
"loss": 0.623,
|
770 |
+
"rewards/accuracies": 0.800000011920929,
|
771 |
+
"rewards/chosen": 0.06336753815412521,
|
772 |
+
"rewards/margins": 0.15559127926826477,
|
773 |
+
"rewards/rejected": -0.09222372621297836,
|
774 |
+
"step": 255
|
775 |
+
},
|
776 |
+
{
|
777 |
+
"epoch": 1.0484603735487128,
|
778 |
+
"grad_norm": 25.334484100341797,
|
779 |
+
"learning_rate": 5e-06,
|
780 |
+
"logits/chosen": -0.9737665057182312,
|
781 |
+
"logits/rejected": -0.9691070318222046,
|
782 |
+
"logps/chosen": -125.85533142089844,
|
783 |
+
"logps/rejected": -142.1818084716797,
|
784 |
+
"loss": 0.6441,
|
785 |
+
"rewards/accuracies": 0.75,
|
786 |
+
"rewards/chosen": 0.0201385710388422,
|
787 |
+
"rewards/margins": 0.10802127420902252,
|
788 |
+
"rewards/rejected": -0.08788268268108368,
|
789 |
+
"step": 260
|
790 |
+
},
|
791 |
+
{
|
792 |
+
"epoch": 1.0686521958606765,
|
793 |
+
"grad_norm": 170.07984924316406,
|
794 |
+
"learning_rate": 5e-06,
|
795 |
+
"logits/chosen": -0.9972125887870789,
|
796 |
+
"logits/rejected": -0.9916419982910156,
|
797 |
+
"logps/chosen": -128.32261657714844,
|
798 |
+
"logps/rejected": -154.72247314453125,
|
799 |
+
"loss": 0.6211,
|
800 |
+
"rewards/accuracies": 0.7749999761581421,
|
801 |
+
"rewards/chosen": 0.003512249793857336,
|
802 |
+
"rewards/margins": 0.1796935498714447,
|
803 |
+
"rewards/rejected": -0.17618130147457123,
|
804 |
+
"step": 265
|
805 |
+
},
|
806 |
+
{
|
807 |
+
"epoch": 1.08884401817264,
|
808 |
+
"grad_norm": 82.89457702636719,
|
809 |
+
"learning_rate": 5e-06,
|
810 |
+
"logits/chosen": -0.9838020205497742,
|
811 |
+
"logits/rejected": -0.9610651135444641,
|
812 |
+
"logps/chosen": -133.44190979003906,
|
813 |
+
"logps/rejected": -135.03680419921875,
|
814 |
+
"loss": 0.6226,
|
815 |
+
"rewards/accuracies": 0.824999988079071,
|
816 |
+
"rewards/chosen": 0.2014409303665161,
|
817 |
+
"rewards/margins": 0.15444937348365784,
|
818 |
+
"rewards/rejected": 0.04699156433343887,
|
819 |
+
"step": 270
|
820 |
+
},
|
821 |
+
{
|
822 |
+
"epoch": 1.1090358404846037,
|
823 |
+
"grad_norm": 256.55047607421875,
|
824 |
+
"learning_rate": 5e-06,
|
825 |
+
"logits/chosen": -1.0722827911376953,
|
826 |
+
"logits/rejected": -1.0403627157211304,
|
827 |
+
"logps/chosen": -129.58506774902344,
|
828 |
+
"logps/rejected": -135.889892578125,
|
829 |
+
"loss": 0.6594,
|
830 |
+
"rewards/accuracies": 0.6499999761581421,
|
831 |
+
"rewards/chosen": 0.1928495317697525,
|
832 |
+
"rewards/margins": 0.09002382308244705,
|
833 |
+
"rewards/rejected": 0.10282570123672485,
|
834 |
+
"step": 275
|
835 |
+
},
|
836 |
+
{
|
837 |
+
"epoch": 1.1292276627965674,
|
838 |
+
"grad_norm": 157.06471252441406,
|
839 |
+
"learning_rate": 5e-06,
|
840 |
+
"logits/chosen": -1.0328292846679688,
|
841 |
+
"logits/rejected": -1.0051934719085693,
|
842 |
+
"logps/chosen": -126.65974426269531,
|
843 |
+
"logps/rejected": -137.65618896484375,
|
844 |
+
"loss": 0.6243,
|
845 |
+
"rewards/accuracies": 0.675000011920929,
|
846 |
+
"rewards/chosen": 0.24083271622657776,
|
847 |
+
"rewards/margins": 0.17111970484256744,
|
848 |
+
"rewards/rejected": 0.06971298158168793,
|
849 |
+
"step": 280
|
850 |
+
},
|
851 |
+
{
|
852 |
+
"epoch": 1.149419485108531,
|
853 |
+
"grad_norm": 318.4048767089844,
|
854 |
+
"learning_rate": 5e-06,
|
855 |
+
"logits/chosen": -1.0038608312606812,
|
856 |
+
"logits/rejected": -0.9795931577682495,
|
857 |
+
"logps/chosen": -119.45623779296875,
|
858 |
+
"logps/rejected": -147.02752685546875,
|
859 |
+
"loss": 0.5933,
|
860 |
+
"rewards/accuracies": 0.75,
|
861 |
+
"rewards/chosen": 0.10432057082653046,
|
862 |
+
"rewards/margins": 0.235098198056221,
|
863 |
+
"rewards/rejected": -0.13077762722969055,
|
864 |
+
"step": 285
|
865 |
+
},
|
866 |
+
{
|
867 |
+
"epoch": 1.1696113074204948,
|
868 |
+
"grad_norm": 111.46747589111328,
|
869 |
+
"learning_rate": 5e-06,
|
870 |
+
"logits/chosen": -1.0481410026550293,
|
871 |
+
"logits/rejected": -1.0197895765304565,
|
872 |
+
"logps/chosen": -132.78433227539062,
|
873 |
+
"logps/rejected": -139.17845153808594,
|
874 |
+
"loss": 0.615,
|
875 |
+
"rewards/accuracies": 0.7250000238418579,
|
876 |
+
"rewards/chosen": -0.008489435538649559,
|
877 |
+
"rewards/margins": 0.21404604613780975,
|
878 |
+
"rewards/rejected": -0.22253546118736267,
|
879 |
+
"step": 290
|
880 |
+
},
|
881 |
+
{
|
882 |
+
"epoch": 1.1898031297324583,
|
883 |
+
"grad_norm": 118.42349243164062,
|
884 |
+
"learning_rate": 5e-06,
|
885 |
+
"logits/chosen": -0.9705491065979004,
|
886 |
+
"logits/rejected": -0.969982922077179,
|
887 |
+
"logps/chosen": -127.90840911865234,
|
888 |
+
"logps/rejected": -132.5115509033203,
|
889 |
+
"loss": 0.6599,
|
890 |
+
"rewards/accuracies": 0.6000000238418579,
|
891 |
+
"rewards/chosen": -0.07771094888448715,
|
892 |
+
"rewards/margins": 0.09617092460393906,
|
893 |
+
"rewards/rejected": -0.1738818734884262,
|
894 |
+
"step": 295
|
895 |
+
},
|
896 |
+
{
|
897 |
+
"epoch": 1.209994952044422,
|
898 |
+
"grad_norm": 28.792146682739258,
|
899 |
+
"learning_rate": 5e-06,
|
900 |
+
"logits/chosen": -1.0062731504440308,
|
901 |
+
"logits/rejected": -1.0046422481536865,
|
902 |
+
"logps/chosen": -122.89021301269531,
|
903 |
+
"logps/rejected": -128.922119140625,
|
904 |
+
"loss": 0.6536,
|
905 |
+
"rewards/accuracies": 0.6499999761581421,
|
906 |
+
"rewards/chosen": -0.21470877528190613,
|
907 |
+
"rewards/margins": 0.10167400538921356,
|
908 |
+
"rewards/rejected": -0.31638282537460327,
|
909 |
+
"step": 300
|
910 |
+
},
|
911 |
+
{
|
912 |
+
"epoch": 1.2301867743563857,
|
913 |
+
"grad_norm": 30.531999588012695,
|
914 |
+
"learning_rate": 5e-06,
|
915 |
+
"logits/chosen": -1.0686832666397095,
|
916 |
+
"logits/rejected": -0.9891031384468079,
|
917 |
+
"logps/chosen": -131.0304718017578,
|
918 |
+
"logps/rejected": -155.66464233398438,
|
919 |
+
"loss": 0.6276,
|
920 |
+
"rewards/accuracies": 0.6499999761581421,
|
921 |
+
"rewards/chosen": -0.22880509495735168,
|
922 |
+
"rewards/margins": 0.17476414144039154,
|
923 |
+
"rewards/rejected": -0.40356922149658203,
|
924 |
+
"step": 305
|
925 |
+
},
|
926 |
+
{
|
927 |
+
"epoch": 1.2503785966683494,
|
928 |
+
"grad_norm": 30.640419006347656,
|
929 |
+
"learning_rate": 5e-06,
|
930 |
+
"logits/chosen": -1.0535386800765991,
|
931 |
+
"logits/rejected": -1.0263631343841553,
|
932 |
+
"logps/chosen": -131.8798828125,
|
933 |
+
"logps/rejected": -134.03146362304688,
|
934 |
+
"loss": 0.6779,
|
935 |
+
"rewards/accuracies": 0.550000011920929,
|
936 |
+
"rewards/chosen": -0.29857125878334045,
|
937 |
+
"rewards/margins": 0.08670302480459213,
|
938 |
+
"rewards/rejected": -0.3852742612361908,
|
939 |
+
"step": 310
|
940 |
+
},
|
941 |
+
{
|
942 |
+
"epoch": 1.2705704189803129,
|
943 |
+
"grad_norm": 66.49352264404297,
|
944 |
+
"learning_rate": 5e-06,
|
945 |
+
"logits/chosen": -1.0025132894515991,
|
946 |
+
"logits/rejected": -1.0053356885910034,
|
947 |
+
"logps/chosen": -141.34634399414062,
|
948 |
+
"logps/rejected": -155.73513793945312,
|
949 |
+
"loss": 0.5846,
|
950 |
+
"rewards/accuracies": 0.6499999761581421,
|
951 |
+
"rewards/chosen": -0.12694263458251953,
|
952 |
+
"rewards/margins": 0.32664644718170166,
|
953 |
+
"rewards/rejected": -0.4535890519618988,
|
954 |
+
"step": 315
|
955 |
+
},
|
956 |
+
{
|
957 |
+
"epoch": 1.2907622412922766,
|
958 |
+
"grad_norm": 426.6810302734375,
|
959 |
+
"learning_rate": 5e-06,
|
960 |
+
"logits/chosen": -0.9757963418960571,
|
961 |
+
"logits/rejected": -0.971258819103241,
|
962 |
+
"logps/chosen": -133.66746520996094,
|
963 |
+
"logps/rejected": -142.61044311523438,
|
964 |
+
"loss": 0.6566,
|
965 |
+
"rewards/accuracies": 0.675000011920929,
|
966 |
+
"rewards/chosen": -0.2488606721162796,
|
967 |
+
"rewards/margins": 0.13673368096351624,
|
968 |
+
"rewards/rejected": -0.38559430837631226,
|
969 |
+
"step": 320
|
970 |
+
},
|
971 |
+
{
|
972 |
+
"epoch": 1.3109540636042403,
|
973 |
+
"grad_norm": 24.37104034423828,
|
974 |
+
"learning_rate": 5e-06,
|
975 |
+
"logits/chosen": -1.0103740692138672,
|
976 |
+
"logits/rejected": -0.9810827374458313,
|
977 |
+
"logps/chosen": -125.60025787353516,
|
978 |
+
"logps/rejected": -135.81178283691406,
|
979 |
+
"loss": 0.6212,
|
980 |
+
"rewards/accuracies": 0.675000011920929,
|
981 |
+
"rewards/chosen": -0.18190403282642365,
|
982 |
+
"rewards/margins": 0.23106291890144348,
|
983 |
+
"rewards/rejected": -0.4129669666290283,
|
984 |
+
"step": 325
|
985 |
+
},
|
986 |
+
{
|
987 |
+
"epoch": 1.331145885916204,
|
988 |
+
"grad_norm": 25.101844787597656,
|
989 |
+
"learning_rate": 5e-06,
|
990 |
+
"logits/chosen": -0.9837746620178223,
|
991 |
+
"logits/rejected": -0.9348167181015015,
|
992 |
+
"logps/chosen": -123.2785415649414,
|
993 |
+
"logps/rejected": -138.2392578125,
|
994 |
+
"loss": 0.608,
|
995 |
+
"rewards/accuracies": 0.675000011920929,
|
996 |
+
"rewards/chosen": -0.1881856620311737,
|
997 |
+
"rewards/margins": 0.23057310283184052,
|
998 |
+
"rewards/rejected": -0.4187587797641754,
|
999 |
+
"step": 330
|
1000 |
+
},
|
1001 |
+
{
|
1002 |
+
"epoch": 1.3513377082281677,
|
1003 |
+
"grad_norm": 33.728919982910156,
|
1004 |
+
"learning_rate": 5e-06,
|
1005 |
+
"logits/chosen": -0.9812191724777222,
|
1006 |
+
"logits/rejected": -0.960713267326355,
|
1007 |
+
"logps/chosen": -135.54893493652344,
|
1008 |
+
"logps/rejected": -147.19290161132812,
|
1009 |
+
"loss": 0.6661,
|
1010 |
+
"rewards/accuracies": 0.675000011920929,
|
1011 |
+
"rewards/chosen": -0.39554575085639954,
|
1012 |
+
"rewards/margins": 0.18837305903434753,
|
1013 |
+
"rewards/rejected": -0.5839187502861023,
|
1014 |
+
"step": 335
|
1015 |
+
},
|
1016 |
+
{
|
1017 |
+
"epoch": 1.3715295305401312,
|
1018 |
+
"grad_norm": 47.88369369506836,
|
1019 |
+
"learning_rate": 5e-06,
|
1020 |
+
"logits/chosen": -0.9874971508979797,
|
1021 |
+
"logits/rejected": -0.9519085884094238,
|
1022 |
+
"logps/chosen": -128.935791015625,
|
1023 |
+
"logps/rejected": -149.05130004882812,
|
1024 |
+
"loss": 0.5798,
|
1025 |
+
"rewards/accuracies": 0.6499999761581421,
|
1026 |
+
"rewards/chosen": -0.3125215470790863,
|
1027 |
+
"rewards/margins": 0.3500240445137024,
|
1028 |
+
"rewards/rejected": -0.6625455617904663,
|
1029 |
+
"step": 340
|
1030 |
+
},
|
1031 |
+
{
|
1032 |
+
"epoch": 1.3917213528520949,
|
1033 |
+
"grad_norm": 51.914127349853516,
|
1034 |
+
"learning_rate": 5e-06,
|
1035 |
+
"logits/chosen": -1.0254395008087158,
|
1036 |
+
"logits/rejected": -0.9828994870185852,
|
1037 |
+
"logps/chosen": -131.8434295654297,
|
1038 |
+
"logps/rejected": -147.67422485351562,
|
1039 |
+
"loss": 0.5951,
|
1040 |
+
"rewards/accuracies": 0.675000011920929,
|
1041 |
+
"rewards/chosen": -0.3478095829486847,
|
1042 |
+
"rewards/margins": 0.28336477279663086,
|
1043 |
+
"rewards/rejected": -0.6311743855476379,
|
1044 |
+
"step": 345
|
1045 |
+
},
|
1046 |
+
{
|
1047 |
+
"epoch": 1.4119131751640586,
|
1048 |
+
"grad_norm": 202.00729370117188,
|
1049 |
+
"learning_rate": 5e-06,
|
1050 |
+
"logits/chosen": -1.0301780700683594,
|
1051 |
+
"logits/rejected": -0.9973276257514954,
|
1052 |
+
"logps/chosen": -134.11044311523438,
|
1053 |
+
"logps/rejected": -146.69715881347656,
|
1054 |
+
"loss": 0.5718,
|
1055 |
+
"rewards/accuracies": 0.7749999761581421,
|
1056 |
+
"rewards/chosen": -0.36468935012817383,
|
1057 |
+
"rewards/margins": 0.3158627152442932,
|
1058 |
+
"rewards/rejected": -0.680552065372467,
|
1059 |
+
"step": 350
|
1060 |
+
},
|
1061 |
+
{
|
1062 |
+
"epoch": 1.4321049974760223,
|
1063 |
+
"grad_norm": 59.660091400146484,
|
1064 |
+
"learning_rate": 5e-06,
|
1065 |
+
"logits/chosen": -0.9530469179153442,
|
1066 |
+
"logits/rejected": -0.9394823312759399,
|
1067 |
+
"logps/chosen": -129.4771270751953,
|
1068 |
+
"logps/rejected": -139.2011260986328,
|
1069 |
+
"loss": 0.6579,
|
1070 |
+
"rewards/accuracies": 0.574999988079071,
|
1071 |
+
"rewards/chosen": -0.3907158672809601,
|
1072 |
+
"rewards/margins": 0.22507838904857635,
|
1073 |
+
"rewards/rejected": -0.61579430103302,
|
1074 |
+
"step": 355
|
1075 |
+
},
|
1076 |
+
{
|
1077 |
+
"epoch": 1.452296819787986,
|
1078 |
+
"grad_norm": 37.25257110595703,
|
1079 |
+
"learning_rate": 5e-06,
|
1080 |
+
"logits/chosen": -0.9922634363174438,
|
1081 |
+
"logits/rejected": -0.960089385509491,
|
1082 |
+
"logps/chosen": -132.1666717529297,
|
1083 |
+
"logps/rejected": -146.88284301757812,
|
1084 |
+
"loss": 0.5824,
|
1085 |
+
"rewards/accuracies": 0.675000011920929,
|
1086 |
+
"rewards/chosen": -0.30794504284858704,
|
1087 |
+
"rewards/margins": 0.31921353936195374,
|
1088 |
+
"rewards/rejected": -0.627158522605896,
|
1089 |
+
"step": 360
|
1090 |
+
},
|
1091 |
+
{
|
1092 |
+
"epoch": 1.4724886420999495,
|
1093 |
+
"grad_norm": 32.722537994384766,
|
1094 |
+
"learning_rate": 5e-06,
|
1095 |
+
"logits/chosen": -0.9966095685958862,
|
1096 |
+
"logits/rejected": -0.990027904510498,
|
1097 |
+
"logps/chosen": -130.92294311523438,
|
1098 |
+
"logps/rejected": -150.19712829589844,
|
1099 |
+
"loss": 0.6163,
|
1100 |
+
"rewards/accuracies": 0.675000011920929,
|
1101 |
+
"rewards/chosen": -0.5703646540641785,
|
1102 |
+
"rewards/margins": 0.3038683533668518,
|
1103 |
+
"rewards/rejected": -0.874233067035675,
|
1104 |
+
"step": 365
|
1105 |
+
},
|
1106 |
+
{
|
1107 |
+
"epoch": 1.4926804644119132,
|
1108 |
+
"grad_norm": 25.522695541381836,
|
1109 |
+
"learning_rate": 5e-06,
|
1110 |
+
"logits/chosen": -0.9670052528381348,
|
1111 |
+
"logits/rejected": -0.9508792757987976,
|
1112 |
+
"logps/chosen": -126.89973449707031,
|
1113 |
+
"logps/rejected": -139.86825561523438,
|
1114 |
+
"loss": 0.6495,
|
1115 |
+
"rewards/accuracies": 0.6000000238418579,
|
1116 |
+
"rewards/chosen": -0.5382200479507446,
|
1117 |
+
"rewards/margins": 0.1835038661956787,
|
1118 |
+
"rewards/rejected": -0.7217239141464233,
|
1119 |
+
"step": 370
|
1120 |
+
},
|
1121 |
+
{
|
1122 |
+
"epoch": 1.5128722867238769,
|
1123 |
+
"grad_norm": 90.71711730957031,
|
1124 |
+
"learning_rate": 5e-06,
|
1125 |
+
"logits/chosen": -0.9002586603164673,
|
1126 |
+
"logits/rejected": -0.8389655947685242,
|
1127 |
+
"logps/chosen": -139.4822235107422,
|
1128 |
+
"logps/rejected": -149.20494079589844,
|
1129 |
+
"loss": 0.6875,
|
1130 |
+
"rewards/accuracies": 0.6499999761581421,
|
1131 |
+
"rewards/chosen": -0.6395633816719055,
|
1132 |
+
"rewards/margins": 0.2545303702354431,
|
1133 |
+
"rewards/rejected": -0.8940938115119934,
|
1134 |
+
"step": 375
|
1135 |
+
},
|
1136 |
+
{
|
1137 |
+
"epoch": 1.5330641090358403,
|
1138 |
+
"grad_norm": 63.06035232543945,
|
1139 |
+
"learning_rate": 5e-06,
|
1140 |
+
"logits/chosen": -0.969630241394043,
|
1141 |
+
"logits/rejected": -0.9403184056282043,
|
1142 |
+
"logps/chosen": -128.59902954101562,
|
1143 |
+
"logps/rejected": -152.47572326660156,
|
1144 |
+
"loss": 0.568,
|
1145 |
+
"rewards/accuracies": 0.6499999761581421,
|
1146 |
+
"rewards/chosen": -0.4838544726371765,
|
1147 |
+
"rewards/margins": 0.41895875334739685,
|
1148 |
+
"rewards/rejected": -0.9028133153915405,
|
1149 |
+
"step": 380
|
1150 |
+
},
|
1151 |
+
{
|
1152 |
+
"epoch": 1.5532559313478043,
|
1153 |
+
"grad_norm": 31.95730972290039,
|
1154 |
+
"learning_rate": 5e-06,
|
1155 |
+
"logits/chosen": -0.9574828147888184,
|
1156 |
+
"logits/rejected": -0.9362820386886597,
|
1157 |
+
"logps/chosen": -133.1338348388672,
|
1158 |
+
"logps/rejected": -137.119873046875,
|
1159 |
+
"loss": 0.6919,
|
1160 |
+
"rewards/accuracies": 0.625,
|
1161 |
+
"rewards/chosen": -0.6652856469154358,
|
1162 |
+
"rewards/margins": 0.1158902496099472,
|
1163 |
+
"rewards/rejected": -0.7811757922172546,
|
1164 |
+
"step": 385
|
1165 |
+
},
|
1166 |
+
{
|
1167 |
+
"epoch": 1.5734477536597677,
|
1168 |
+
"grad_norm": 82.7573471069336,
|
1169 |
+
"learning_rate": 5e-06,
|
1170 |
+
"logits/chosen": -0.9827107191085815,
|
1171 |
+
"logits/rejected": -0.9468755722045898,
|
1172 |
+
"logps/chosen": -133.9182891845703,
|
1173 |
+
"logps/rejected": -147.95758056640625,
|
1174 |
+
"loss": 0.6258,
|
1175 |
+
"rewards/accuracies": 0.6499999761581421,
|
1176 |
+
"rewards/chosen": -0.508975088596344,
|
1177 |
+
"rewards/margins": 0.27125072479248047,
|
1178 |
+
"rewards/rejected": -0.7802258133888245,
|
1179 |
+
"step": 390
|
1180 |
+
},
|
1181 |
+
{
|
1182 |
+
"epoch": 1.5936395759717314,
|
1183 |
+
"grad_norm": 44.66698455810547,
|
1184 |
+
"learning_rate": 5e-06,
|
1185 |
+
"logits/chosen": -0.9529491662979126,
|
1186 |
+
"logits/rejected": -0.9102163314819336,
|
1187 |
+
"logps/chosen": -127.85603332519531,
|
1188 |
+
"logps/rejected": -165.6941375732422,
|
1189 |
+
"loss": 0.5216,
|
1190 |
+
"rewards/accuracies": 0.800000011920929,
|
1191 |
+
"rewards/chosen": -0.4891042709350586,
|
1192 |
+
"rewards/margins": 0.5509933233261108,
|
1193 |
+
"rewards/rejected": -1.0400975942611694,
|
1194 |
+
"step": 395
|
1195 |
+
},
|
1196 |
+
{
|
1197 |
+
"epoch": 1.6138313982836952,
|
1198 |
+
"grad_norm": 28.743886947631836,
|
1199 |
+
"learning_rate": 5e-06,
|
1200 |
+
"logits/chosen": -0.9382497072219849,
|
1201 |
+
"logits/rejected": -0.9725173115730286,
|
1202 |
+
"logps/chosen": -125.57157135009766,
|
1203 |
+
"logps/rejected": -139.066650390625,
|
1204 |
+
"loss": 0.6589,
|
1205 |
+
"rewards/accuracies": 0.5249999761581421,
|
1206 |
+
"rewards/chosen": -0.4889567792415619,
|
1207 |
+
"rewards/margins": 0.13421431183815002,
|
1208 |
+
"rewards/rejected": -0.6231711506843567,
|
1209 |
+
"step": 400
|
1210 |
+
},
|
1211 |
+
{
|
1212 |
+
"epoch": 1.6340232205956586,
|
1213 |
+
"grad_norm": 26.614904403686523,
|
1214 |
+
"learning_rate": 5e-06,
|
1215 |
+
"logits/chosen": -0.9904226064682007,
|
1216 |
+
"logits/rejected": -0.9899235963821411,
|
1217 |
+
"logps/chosen": -130.1407470703125,
|
1218 |
+
"logps/rejected": -148.65634155273438,
|
1219 |
+
"loss": 0.568,
|
1220 |
+
"rewards/accuracies": 0.7749999761581421,
|
1221 |
+
"rewards/chosen": -0.3976411819458008,
|
1222 |
+
"rewards/margins": 0.3590780794620514,
|
1223 |
+
"rewards/rejected": -0.7567192912101746,
|
1224 |
+
"step": 405
|
1225 |
+
},
|
1226 |
+
{
|
1227 |
+
"epoch": 1.6542150429076226,
|
1228 |
+
"grad_norm": 63.328731536865234,
|
1229 |
+
"learning_rate": 5e-06,
|
1230 |
+
"logits/chosen": -0.9512359499931335,
|
1231 |
+
"logits/rejected": -0.9244702458381653,
|
1232 |
+
"logps/chosen": -142.51605224609375,
|
1233 |
+
"logps/rejected": -146.30654907226562,
|
1234 |
+
"loss": 0.6478,
|
1235 |
+
"rewards/accuracies": 0.699999988079071,
|
1236 |
+
"rewards/chosen": -0.611981213092804,
|
1237 |
+
"rewards/margins": 0.2888861298561096,
|
1238 |
+
"rewards/rejected": -0.9008673429489136,
|
1239 |
+
"step": 410
|
1240 |
+
},
|
1241 |
+
{
|
1242 |
+
"epoch": 1.674406865219586,
|
1243 |
+
"grad_norm": 77.01255798339844,
|
1244 |
+
"learning_rate": 5e-06,
|
1245 |
+
"logits/chosen": -0.957229495048523,
|
1246 |
+
"logits/rejected": -0.9442984461784363,
|
1247 |
+
"logps/chosen": -129.83509826660156,
|
1248 |
+
"logps/rejected": -140.24246215820312,
|
1249 |
+
"loss": 0.6499,
|
1250 |
+
"rewards/accuracies": 0.6499999761581421,
|
1251 |
+
"rewards/chosen": -0.3882976770401001,
|
1252 |
+
"rewards/margins": 0.2001279890537262,
|
1253 |
+
"rewards/rejected": -0.5884256958961487,
|
1254 |
+
"step": 415
|
1255 |
+
},
|
1256 |
+
{
|
1257 |
+
"epoch": 1.6945986875315497,
|
1258 |
+
"grad_norm": 64.73177337646484,
|
1259 |
+
"learning_rate": 5e-06,
|
1260 |
+
"logits/chosen": -0.9330608248710632,
|
1261 |
+
"logits/rejected": -0.9302207827568054,
|
1262 |
+
"logps/chosen": -136.36305236816406,
|
1263 |
+
"logps/rejected": -156.6847381591797,
|
1264 |
+
"loss": 0.5401,
|
1265 |
+
"rewards/accuracies": 0.699999988079071,
|
1266 |
+
"rewards/chosen": -0.5075548887252808,
|
1267 |
+
"rewards/margins": 0.4187861382961273,
|
1268 |
+
"rewards/rejected": -0.9263409376144409,
|
1269 |
+
"step": 420
|
1270 |
+
},
|
1271 |
+
{
|
1272 |
+
"epoch": 1.7147905098435134,
|
1273 |
+
"grad_norm": 33.53404235839844,
|
1274 |
+
"learning_rate": 5e-06,
|
1275 |
+
"logits/chosen": -0.9966014623641968,
|
1276 |
+
"logits/rejected": -0.9909483790397644,
|
1277 |
+
"logps/chosen": -131.6799774169922,
|
1278 |
+
"logps/rejected": -132.3755340576172,
|
1279 |
+
"loss": 0.613,
|
1280 |
+
"rewards/accuracies": 0.75,
|
1281 |
+
"rewards/chosen": -0.41457399725914,
|
1282 |
+
"rewards/margins": 0.2104615867137909,
|
1283 |
+
"rewards/rejected": -0.6250355839729309,
|
1284 |
+
"step": 425
|
1285 |
+
},
|
1286 |
+
{
|
1287 |
+
"epoch": 1.734982332155477,
|
1288 |
+
"grad_norm": 22.384197235107422,
|
1289 |
+
"learning_rate": 5e-06,
|
1290 |
+
"logits/chosen": -0.9789875149726868,
|
1291 |
+
"logits/rejected": -0.9263578653335571,
|
1292 |
+
"logps/chosen": -135.25543212890625,
|
1293 |
+
"logps/rejected": -143.0823211669922,
|
1294 |
+
"loss": 0.6018,
|
1295 |
+
"rewards/accuracies": 0.6499999761581421,
|
1296 |
+
"rewards/chosen": -0.4446907639503479,
|
1297 |
+
"rewards/margins": 0.33062392473220825,
|
1298 |
+
"rewards/rejected": -0.7753147482872009,
|
1299 |
+
"step": 430
|
1300 |
+
},
|
1301 |
+
{
|
1302 |
+
"epoch": 1.7551741544674409,
|
1303 |
+
"grad_norm": 37.48057174682617,
|
1304 |
+
"learning_rate": 5e-06,
|
1305 |
+
"logits/chosen": -1.0088233947753906,
|
1306 |
+
"logits/rejected": -1.007187843322754,
|
1307 |
+
"logps/chosen": -129.21136474609375,
|
1308 |
+
"logps/rejected": -147.24827575683594,
|
1309 |
+
"loss": 0.6201,
|
1310 |
+
"rewards/accuracies": 0.675000011920929,
|
1311 |
+
"rewards/chosen": -0.4158124029636383,
|
1312 |
+
"rewards/margins": 0.2677377462387085,
|
1313 |
+
"rewards/rejected": -0.683550238609314,
|
1314 |
+
"step": 435
|
1315 |
+
},
|
1316 |
+
{
|
1317 |
+
"epoch": 1.7753659767794043,
|
1318 |
+
"grad_norm": 82.85381317138672,
|
1319 |
+
"learning_rate": 5e-06,
|
1320 |
+
"logits/chosen": -0.9587488174438477,
|
1321 |
+
"logits/rejected": -0.9475493431091309,
|
1322 |
+
"logps/chosen": -129.45645141601562,
|
1323 |
+
"logps/rejected": -136.0018310546875,
|
1324 |
+
"loss": 0.6111,
|
1325 |
+
"rewards/accuracies": 0.625,
|
1326 |
+
"rewards/chosen": -0.42559027671813965,
|
1327 |
+
"rewards/margins": 0.2766748368740082,
|
1328 |
+
"rewards/rejected": -0.7022650837898254,
|
1329 |
+
"step": 440
|
1330 |
+
},
|
1331 |
+
{
|
1332 |
+
"epoch": 1.795557799091368,
|
1333 |
+
"grad_norm": 386.5701904296875,
|
1334 |
+
"learning_rate": 5e-06,
|
1335 |
+
"logits/chosen": -0.9646188616752625,
|
1336 |
+
"logits/rejected": -0.922524631023407,
|
1337 |
+
"logps/chosen": -126.2032699584961,
|
1338 |
+
"logps/rejected": -161.66268920898438,
|
1339 |
+
"loss": 0.5059,
|
1340 |
+
"rewards/accuracies": 0.800000011920929,
|
1341 |
+
"rewards/chosen": -0.5310975313186646,
|
1342 |
+
"rewards/margins": 0.5927624702453613,
|
1343 |
+
"rewards/rejected": -1.1238601207733154,
|
1344 |
+
"step": 445
|
1345 |
+
},
|
1346 |
+
{
|
1347 |
+
"epoch": 1.8157496214033317,
|
1348 |
+
"grad_norm": 33.02116012573242,
|
1349 |
+
"learning_rate": 5e-06,
|
1350 |
+
"logits/chosen": -0.9980605840682983,
|
1351 |
+
"logits/rejected": -0.9731265902519226,
|
1352 |
+
"logps/chosen": -124.5853500366211,
|
1353 |
+
"logps/rejected": -143.31353759765625,
|
1354 |
+
"loss": 0.5554,
|
1355 |
+
"rewards/accuracies": 0.75,
|
1356 |
+
"rewards/chosen": -0.6029568314552307,
|
1357 |
+
"rewards/margins": 0.38439229130744934,
|
1358 |
+
"rewards/rejected": -0.9873491525650024,
|
1359 |
+
"step": 450
|
1360 |
+
},
|
1361 |
+
{
|
1362 |
+
"epoch": 1.8359414437152952,
|
1363 |
+
"grad_norm": 72.07835388183594,
|
1364 |
+
"learning_rate": 5e-06,
|
1365 |
+
"logits/chosen": -0.9654960632324219,
|
1366 |
+
"logits/rejected": -0.9319403767585754,
|
1367 |
+
"logps/chosen": -139.59352111816406,
|
1368 |
+
"logps/rejected": -148.42514038085938,
|
1369 |
+
"loss": 0.598,
|
1370 |
+
"rewards/accuracies": 0.75,
|
1371 |
+
"rewards/chosen": -0.7650180459022522,
|
1372 |
+
"rewards/margins": 0.3671188950538635,
|
1373 |
+
"rewards/rejected": -1.1321367025375366,
|
1374 |
+
"step": 455
|
1375 |
+
},
|
1376 |
+
{
|
1377 |
+
"epoch": 1.856133266027259,
|
1378 |
+
"grad_norm": 37.11824417114258,
|
1379 |
+
"learning_rate": 5e-06,
|
1380 |
+
"logits/chosen": -0.9613730311393738,
|
1381 |
+
"logits/rejected": -0.9452563524246216,
|
1382 |
+
"logps/chosen": -131.92416381835938,
|
1383 |
+
"logps/rejected": -146.32144165039062,
|
1384 |
+
"loss": 0.684,
|
1385 |
+
"rewards/accuracies": 0.5249999761581421,
|
1386 |
+
"rewards/chosen": -0.8234866857528687,
|
1387 |
+
"rewards/margins": 0.22403354942798615,
|
1388 |
+
"rewards/rejected": -1.0475202798843384,
|
1389 |
+
"step": 460
|
1390 |
+
},
|
1391 |
+
{
|
1392 |
+
"epoch": 1.8763250883392226,
|
1393 |
+
"grad_norm": 102.86495208740234,
|
1394 |
+
"learning_rate": 5e-06,
|
1395 |
+
"logits/chosen": -0.9701520800590515,
|
1396 |
+
"logits/rejected": -0.9547005891799927,
|
1397 |
+
"logps/chosen": -140.27639770507812,
|
1398 |
+
"logps/rejected": -140.81581115722656,
|
1399 |
+
"loss": 0.6644,
|
1400 |
+
"rewards/accuracies": 0.574999988079071,
|
1401 |
+
"rewards/chosen": -0.8090511560440063,
|
1402 |
+
"rewards/margins": 0.21369917690753937,
|
1403 |
+
"rewards/rejected": -1.0227503776550293,
|
1404 |
+
"step": 465
|
1405 |
+
},
|
1406 |
+
{
|
1407 |
+
"epoch": 1.896516910651186,
|
1408 |
+
"grad_norm": 38.73020553588867,
|
1409 |
+
"learning_rate": 5e-06,
|
1410 |
+
"logits/chosen": -0.9636425971984863,
|
1411 |
+
"logits/rejected": -0.8940737843513489,
|
1412 |
+
"logps/chosen": -128.97019958496094,
|
1413 |
+
"logps/rejected": -144.57125854492188,
|
1414 |
+
"loss": 0.5636,
|
1415 |
+
"rewards/accuracies": 0.75,
|
1416 |
+
"rewards/chosen": -0.7400966882705688,
|
1417 |
+
"rewards/margins": 0.3483065962791443,
|
1418 |
+
"rewards/rejected": -1.0884032249450684,
|
1419 |
+
"step": 470
|
1420 |
+
},
|
1421 |
+
{
|
1422 |
+
"epoch": 1.91670873296315,
|
1423 |
+
"grad_norm": 50.96113204956055,
|
1424 |
+
"learning_rate": 5e-06,
|
1425 |
+
"logits/chosen": -1.0122615098953247,
|
1426 |
+
"logits/rejected": -0.9771004915237427,
|
1427 |
+
"logps/chosen": -130.7502899169922,
|
1428 |
+
"logps/rejected": -147.36337280273438,
|
1429 |
+
"loss": 0.55,
|
1430 |
+
"rewards/accuracies": 0.6499999761581421,
|
1431 |
+
"rewards/chosen": -0.5859622359275818,
|
1432 |
+
"rewards/margins": 0.4816177785396576,
|
1433 |
+
"rewards/rejected": -1.0675798654556274,
|
1434 |
+
"step": 475
|
1435 |
+
},
|
1436 |
+
{
|
1437 |
+
"epoch": 1.9369005552751135,
|
1438 |
+
"grad_norm": 50.336578369140625,
|
1439 |
+
"learning_rate": 5e-06,
|
1440 |
+
"logits/chosen": -1.009649395942688,
|
1441 |
+
"logits/rejected": -0.9977648854255676,
|
1442 |
+
"logps/chosen": -138.2186279296875,
|
1443 |
+
"logps/rejected": -146.3845977783203,
|
1444 |
+
"loss": 0.6772,
|
1445 |
+
"rewards/accuracies": 0.5249999761581421,
|
1446 |
+
"rewards/chosen": -0.772314190864563,
|
1447 |
+
"rewards/margins": 0.2373802661895752,
|
1448 |
+
"rewards/rejected": -1.0096943378448486,
|
1449 |
+
"step": 480
|
1450 |
+
},
|
1451 |
+
{
|
1452 |
+
"epoch": 1.9570923775870772,
|
1453 |
+
"grad_norm": 68.63349151611328,
|
1454 |
+
"learning_rate": 5e-06,
|
1455 |
+
"logits/chosen": -0.9773464202880859,
|
1456 |
+
"logits/rejected": -0.9308035969734192,
|
1457 |
+
"logps/chosen": -139.3035888671875,
|
1458 |
+
"logps/rejected": -161.8266143798828,
|
1459 |
+
"loss": 0.5776,
|
1460 |
+
"rewards/accuracies": 0.75,
|
1461 |
+
"rewards/chosen": -0.8935576677322388,
|
1462 |
+
"rewards/margins": 0.4999828338623047,
|
1463 |
+
"rewards/rejected": -1.393540382385254,
|
1464 |
+
"step": 485
|
1465 |
+
},
|
1466 |
+
{
|
1467 |
+
"epoch": 1.977284199899041,
|
1468 |
+
"grad_norm": 88.56826782226562,
|
1469 |
+
"learning_rate": 5e-06,
|
1470 |
+
"logits/chosen": -1.0616611242294312,
|
1471 |
+
"logits/rejected": -1.0203070640563965,
|
1472 |
+
"logps/chosen": -130.81297302246094,
|
1473 |
+
"logps/rejected": -141.33740234375,
|
1474 |
+
"loss": 0.6037,
|
1475 |
+
"rewards/accuracies": 0.75,
|
1476 |
+
"rewards/chosen": -0.8136008977890015,
|
1477 |
+
"rewards/margins": 0.2671222984790802,
|
1478 |
+
"rewards/rejected": -1.0807230472564697,
|
1479 |
+
"step": 490
|
1480 |
+
},
|
1481 |
+
{
|
1482 |
+
"epoch": 1.9974760222110044,
|
1483 |
+
"grad_norm": 126.10334777832031,
|
1484 |
+
"learning_rate": 5e-06,
|
1485 |
+
"logits/chosen": -1.0034191608428955,
|
1486 |
+
"logits/rejected": -1.010565161705017,
|
1487 |
+
"logps/chosen": -135.72381591796875,
|
1488 |
+
"logps/rejected": -133.8328399658203,
|
1489 |
+
"loss": 0.6832,
|
1490 |
+
"rewards/accuracies": 0.6000000238418579,
|
1491 |
+
"rewards/chosen": -0.7619636654853821,
|
1492 |
+
"rewards/margins": 0.17222894728183746,
|
1493 |
+
"rewards/rejected": -0.9341924786567688,
|
1494 |
+
"step": 495
|
1495 |
+
},
|
1496 |
+
{
|
1497 |
+
"epoch": 2.016153457849571,
|
1498 |
+
"grad_norm": 48.9210205078125,
|
1499 |
+
"learning_rate": 5e-06,
|
1500 |
+
"logits/chosen": -1.023393988609314,
|
1501 |
+
"logits/rejected": -0.9911313652992249,
|
1502 |
+
"logps/chosen": -122.88961029052734,
|
1503 |
+
"logps/rejected": -149.01161193847656,
|
1504 |
+
"loss": 0.3935,
|
1505 |
+
"rewards/accuracies": 0.8648648858070374,
|
1506 |
+
"rewards/chosen": -0.35930752754211426,
|
1507 |
+
"rewards/margins": 0.8153496980667114,
|
1508 |
+
"rewards/rejected": -1.1746571063995361,
|
1509 |
+
"step": 500
|
1510 |
+
},
|
1511 |
+
{
|
1512 |
+
"epoch": 2.0363452801615347,
|
1513 |
+
"grad_norm": 39.74098587036133,
|
1514 |
+
"learning_rate": 5e-06,
|
1515 |
+
"logits/chosen": -1.0313756465911865,
|
1516 |
+
"logits/rejected": -0.9942135810852051,
|
1517 |
+
"logps/chosen": -135.91714477539062,
|
1518 |
+
"logps/rejected": -147.81710815429688,
|
1519 |
+
"loss": 0.4991,
|
1520 |
+
"rewards/accuracies": 0.7749999761581421,
|
1521 |
+
"rewards/chosen": -0.5435918569564819,
|
1522 |
+
"rewards/margins": 0.6506127715110779,
|
1523 |
+
"rewards/rejected": -1.194204568862915,
|
1524 |
+
"step": 505
|
1525 |
+
},
|
1526 |
+
{
|
1527 |
+
"epoch": 2.056537102473498,
|
1528 |
+
"grad_norm": 49.587120056152344,
|
1529 |
+
"learning_rate": 5e-06,
|
1530 |
+
"logits/chosen": -1.0160324573516846,
|
1531 |
+
"logits/rejected": -1.0125465393066406,
|
1532 |
+
"logps/chosen": -131.9548797607422,
|
1533 |
+
"logps/rejected": -152.05862426757812,
|
1534 |
+
"loss": 0.5047,
|
1535 |
+
"rewards/accuracies": 0.800000011920929,
|
1536 |
+
"rewards/chosen": -0.6925584673881531,
|
1537 |
+
"rewards/margins": 0.5985544919967651,
|
1538 |
+
"rewards/rejected": -1.2911128997802734,
|
1539 |
+
"step": 510
|
1540 |
+
},
|
1541 |
+
{
|
1542 |
+
"epoch": 2.0767289247854617,
|
1543 |
+
"grad_norm": 17.234088897705078,
|
1544 |
+
"learning_rate": 5e-06,
|
1545 |
+
"logits/chosen": -0.9918926358222961,
|
1546 |
+
"logits/rejected": -0.9893625378608704,
|
1547 |
+
"logps/chosen": -132.64453125,
|
1548 |
+
"logps/rejected": -158.79873657226562,
|
1549 |
+
"loss": 0.42,
|
1550 |
+
"rewards/accuracies": 0.8500000238418579,
|
1551 |
+
"rewards/chosen": -0.5537530779838562,
|
1552 |
+
"rewards/margins": 0.938607394695282,
|
1553 |
+
"rewards/rejected": -1.4923605918884277,
|
1554 |
+
"step": 515
|
1555 |
+
},
|
1556 |
+
{
|
1557 |
+
"epoch": 2.0969207470974256,
|
1558 |
+
"grad_norm": 162.84471130371094,
|
1559 |
+
"learning_rate": 5e-06,
|
1560 |
+
"logits/chosen": -1.008296012878418,
|
1561 |
+
"logits/rejected": -1.0098919868469238,
|
1562 |
+
"logps/chosen": -139.8673095703125,
|
1563 |
+
"logps/rejected": -160.15660095214844,
|
1564 |
+
"loss": 0.4181,
|
1565 |
+
"rewards/accuracies": 0.824999988079071,
|
1566 |
+
"rewards/chosen": -0.6410013437271118,
|
1567 |
+
"rewards/margins": 0.9079731106758118,
|
1568 |
+
"rewards/rejected": -1.5489743947982788,
|
1569 |
+
"step": 520
|
1570 |
+
},
|
1571 |
+
{
|
1572 |
+
"epoch": 2.117112569409389,
|
1573 |
+
"grad_norm": 29.040531158447266,
|
1574 |
+
"learning_rate": 5e-06,
|
1575 |
+
"logits/chosen": -1.0176864862442017,
|
1576 |
+
"logits/rejected": -1.007263422012329,
|
1577 |
+
"logps/chosen": -131.76019287109375,
|
1578 |
+
"logps/rejected": -149.74111938476562,
|
1579 |
+
"loss": 0.4928,
|
1580 |
+
"rewards/accuracies": 0.800000011920929,
|
1581 |
+
"rewards/chosen": -0.6737687587738037,
|
1582 |
+
"rewards/margins": 0.6350966691970825,
|
1583 |
+
"rewards/rejected": -1.3088653087615967,
|
1584 |
+
"step": 525
|
1585 |
+
},
|
1586 |
+
{
|
1587 |
+
"epoch": 2.137304391721353,
|
1588 |
+
"grad_norm": 97.38697052001953,
|
1589 |
+
"learning_rate": 5e-06,
|
1590 |
+
"logits/chosen": -1.0380053520202637,
|
1591 |
+
"logits/rejected": -0.9718503952026367,
|
1592 |
+
"logps/chosen": -132.33323669433594,
|
1593 |
+
"logps/rejected": -152.97665405273438,
|
1594 |
+
"loss": 0.4573,
|
1595 |
+
"rewards/accuracies": 0.8500000238418579,
|
1596 |
+
"rewards/chosen": -0.6338863372802734,
|
1597 |
+
"rewards/margins": 0.8563046455383301,
|
1598 |
+
"rewards/rejected": -1.4901909828186035,
|
1599 |
+
"step": 530
|
1600 |
+
},
|
1601 |
+
{
|
1602 |
+
"epoch": 2.1574962140333165,
|
1603 |
+
"grad_norm": 77.62568664550781,
|
1604 |
+
"learning_rate": 5e-06,
|
1605 |
+
"logits/chosen": -1.0812318325042725,
|
1606 |
+
"logits/rejected": -1.0583606958389282,
|
1607 |
+
"logps/chosen": -135.34068298339844,
|
1608 |
+
"logps/rejected": -151.8748016357422,
|
1609 |
+
"loss": 0.4322,
|
1610 |
+
"rewards/accuracies": 0.824999988079071,
|
1611 |
+
"rewards/chosen": -0.6975581049919128,
|
1612 |
+
"rewards/margins": 0.8094134330749512,
|
1613 |
+
"rewards/rejected": -1.5069715976715088,
|
1614 |
+
"step": 535
|
1615 |
+
},
|
1616 |
+
{
|
1617 |
+
"epoch": 2.17768803634528,
|
1618 |
+
"grad_norm": 67.6436767578125,
|
1619 |
+
"learning_rate": 5e-06,
|
1620 |
+
"logits/chosen": -1.0246284008026123,
|
1621 |
+
"logits/rejected": -1.014534831047058,
|
1622 |
+
"logps/chosen": -137.48171997070312,
|
1623 |
+
"logps/rejected": -149.2544708251953,
|
1624 |
+
"loss": 0.4063,
|
1625 |
+
"rewards/accuracies": 0.949999988079071,
|
1626 |
+
"rewards/chosen": -0.7033039331436157,
|
1627 |
+
"rewards/margins": 0.945264458656311,
|
1628 |
+
"rewards/rejected": -1.6485683917999268,
|
1629 |
+
"step": 540
|
1630 |
+
},
|
1631 |
+
{
|
1632 |
+
"epoch": 2.197879858657244,
|
1633 |
+
"grad_norm": 63.45933532714844,
|
1634 |
+
"learning_rate": 5e-06,
|
1635 |
+
"logits/chosen": -1.0668962001800537,
|
1636 |
+
"logits/rejected": -1.0191361904144287,
|
1637 |
+
"logps/chosen": -136.38905334472656,
|
1638 |
+
"logps/rejected": -156.759521484375,
|
1639 |
+
"loss": 0.4042,
|
1640 |
+
"rewards/accuracies": 0.8500000238418579,
|
1641 |
+
"rewards/chosen": -0.7140289545059204,
|
1642 |
+
"rewards/margins": 0.8608022928237915,
|
1643 |
+
"rewards/rejected": -1.574831247329712,
|
1644 |
+
"step": 545
|
1645 |
+
},
|
1646 |
+
{
|
1647 |
+
"epoch": 2.2180716809692074,
|
1648 |
+
"grad_norm": 24.154354095458984,
|
1649 |
+
"learning_rate": 5e-06,
|
1650 |
+
"logits/chosen": -1.034576654434204,
|
1651 |
+
"logits/rejected": -1.0220065116882324,
|
1652 |
+
"logps/chosen": -126.9083251953125,
|
1653 |
+
"logps/rejected": -147.03082275390625,
|
1654 |
+
"loss": 0.4812,
|
1655 |
+
"rewards/accuracies": 0.7250000238418579,
|
1656 |
+
"rewards/chosen": -0.6011126041412354,
|
1657 |
+
"rewards/margins": 0.7494941353797913,
|
1658 |
+
"rewards/rejected": -1.3506066799163818,
|
1659 |
+
"step": 550
|
1660 |
+
},
|
1661 |
+
{
|
1662 |
+
"epoch": 2.2382635032811713,
|
1663 |
+
"grad_norm": 51.32468032836914,
|
1664 |
+
"learning_rate": 5e-06,
|
1665 |
+
"logits/chosen": -1.0417560338974,
|
1666 |
+
"logits/rejected": -0.995162844657898,
|
1667 |
+
"logps/chosen": -134.18875122070312,
|
1668 |
+
"logps/rejected": -160.58987426757812,
|
1669 |
+
"loss": 0.4263,
|
1670 |
+
"rewards/accuracies": 0.800000011920929,
|
1671 |
+
"rewards/chosen": -0.7680782079696655,
|
1672 |
+
"rewards/margins": 0.9993975758552551,
|
1673 |
+
"rewards/rejected": -1.7674758434295654,
|
1674 |
+
"step": 555
|
1675 |
+
},
|
1676 |
+
{
|
1677 |
+
"epoch": 2.2584553255931348,
|
1678 |
+
"grad_norm": 31.85598373413086,
|
1679 |
+
"learning_rate": 5e-06,
|
1680 |
+
"logits/chosen": -1.091623306274414,
|
1681 |
+
"logits/rejected": -1.065324068069458,
|
1682 |
+
"logps/chosen": -147.77536010742188,
|
1683 |
+
"logps/rejected": -161.46456909179688,
|
1684 |
+
"loss": 0.4047,
|
1685 |
+
"rewards/accuracies": 0.925000011920929,
|
1686 |
+
"rewards/chosen": -0.9492090940475464,
|
1687 |
+
"rewards/margins": 0.9160175323486328,
|
1688 |
+
"rewards/rejected": -1.8652265071868896,
|
1689 |
+
"step": 560
|
1690 |
+
},
|
1691 |
+
{
|
1692 |
+
"epoch": 2.2786471479050983,
|
1693 |
+
"grad_norm": 40.57142639160156,
|
1694 |
+
"learning_rate": 5e-06,
|
1695 |
+
"logits/chosen": -1.0318306684494019,
|
1696 |
+
"logits/rejected": -1.0202347040176392,
|
1697 |
+
"logps/chosen": -133.6524658203125,
|
1698 |
+
"logps/rejected": -154.2473602294922,
|
1699 |
+
"loss": 0.4398,
|
1700 |
+
"rewards/accuracies": 0.875,
|
1701 |
+
"rewards/chosen": -0.774714469909668,
|
1702 |
+
"rewards/margins": 0.8600656390190125,
|
1703 |
+
"rewards/rejected": -1.6347801685333252,
|
1704 |
+
"step": 565
|
1705 |
+
},
|
1706 |
+
{
|
1707 |
+
"epoch": 2.298838970217062,
|
1708 |
+
"grad_norm": 310.98028564453125,
|
1709 |
+
"learning_rate": 5e-06,
|
1710 |
+
"logits/chosen": -1.0679352283477783,
|
1711 |
+
"logits/rejected": -1.0488876104354858,
|
1712 |
+
"logps/chosen": -143.3402557373047,
|
1713 |
+
"logps/rejected": -157.1344757080078,
|
1714 |
+
"loss": 0.4709,
|
1715 |
+
"rewards/accuracies": 0.824999988079071,
|
1716 |
+
"rewards/chosen": -0.8914315104484558,
|
1717 |
+
"rewards/margins": 0.8738172650337219,
|
1718 |
+
"rewards/rejected": -1.7652488946914673,
|
1719 |
+
"step": 570
|
1720 |
+
},
|
1721 |
+
{
|
1722 |
+
"epoch": 2.3190307925290257,
|
1723 |
+
"grad_norm": 32.55216979980469,
|
1724 |
+
"learning_rate": 5e-06,
|
1725 |
+
"logits/chosen": -1.0195152759552002,
|
1726 |
+
"logits/rejected": -1.0584851503372192,
|
1727 |
+
"logps/chosen": -140.6759796142578,
|
1728 |
+
"logps/rejected": -161.9089813232422,
|
1729 |
+
"loss": 0.5658,
|
1730 |
+
"rewards/accuracies": 0.800000011920929,
|
1731 |
+
"rewards/chosen": -1.0393126010894775,
|
1732 |
+
"rewards/margins": 0.8234487771987915,
|
1733 |
+
"rewards/rejected": -1.8627614974975586,
|
1734 |
+
"step": 575
|
1735 |
+
},
|
1736 |
+
{
|
1737 |
+
"epoch": 2.3392226148409896,
|
1738 |
+
"grad_norm": 68.31248474121094,
|
1739 |
+
"learning_rate": 5e-06,
|
1740 |
+
"logits/chosen": -1.0342488288879395,
|
1741 |
+
"logits/rejected": -1.0230834484100342,
|
1742 |
+
"logps/chosen": -129.74427795410156,
|
1743 |
+
"logps/rejected": -151.37411499023438,
|
1744 |
+
"loss": 0.434,
|
1745 |
+
"rewards/accuracies": 0.8500000238418579,
|
1746 |
+
"rewards/chosen": -0.707876443862915,
|
1747 |
+
"rewards/margins": 1.0140485763549805,
|
1748 |
+
"rewards/rejected": -1.721925139427185,
|
1749 |
+
"step": 580
|
1750 |
+
},
|
1751 |
+
{
|
1752 |
+
"epoch": 2.359414437152953,
|
1753 |
+
"grad_norm": 89.6629638671875,
|
1754 |
+
"learning_rate": 5e-06,
|
1755 |
+
"logits/chosen": -1.1206090450286865,
|
1756 |
+
"logits/rejected": -1.0528461933135986,
|
1757 |
+
"logps/chosen": -128.90150451660156,
|
1758 |
+
"logps/rejected": -145.35079956054688,
|
1759 |
+
"loss": 0.4469,
|
1760 |
+
"rewards/accuracies": 0.8500000238418579,
|
1761 |
+
"rewards/chosen": -0.7581853270530701,
|
1762 |
+
"rewards/margins": 0.7499567270278931,
|
1763 |
+
"rewards/rejected": -1.5081422328948975,
|
1764 |
+
"step": 585
|
1765 |
+
},
|
1766 |
+
{
|
1767 |
+
"epoch": 2.3796062594649166,
|
1768 |
+
"grad_norm": 43.05556869506836,
|
1769 |
+
"learning_rate": 5e-06,
|
1770 |
+
"logits/chosen": -1.065771460533142,
|
1771 |
+
"logits/rejected": -1.0271015167236328,
|
1772 |
+
"logps/chosen": -130.0537872314453,
|
1773 |
+
"logps/rejected": -149.29409790039062,
|
1774 |
+
"loss": 0.4797,
|
1775 |
+
"rewards/accuracies": 0.875,
|
1776 |
+
"rewards/chosen": -0.8601150512695312,
|
1777 |
+
"rewards/margins": 0.7315414547920227,
|
1778 |
+
"rewards/rejected": -1.5916566848754883,
|
1779 |
+
"step": 590
|
1780 |
+
},
|
1781 |
+
{
|
1782 |
+
"epoch": 2.3997980817768805,
|
1783 |
+
"grad_norm": 66.487060546875,
|
1784 |
+
"learning_rate": 5e-06,
|
1785 |
+
"logits/chosen": -1.1071510314941406,
|
1786 |
+
"logits/rejected": -1.0782781839370728,
|
1787 |
+
"logps/chosen": -138.64755249023438,
|
1788 |
+
"logps/rejected": -162.64129638671875,
|
1789 |
+
"loss": 0.4263,
|
1790 |
+
"rewards/accuracies": 0.824999988079071,
|
1791 |
+
"rewards/chosen": -0.9284160733222961,
|
1792 |
+
"rewards/margins": 1.0049647092819214,
|
1793 |
+
"rewards/rejected": -1.9333808422088623,
|
1794 |
+
"step": 595
|
1795 |
+
},
|
1796 |
+
{
|
1797 |
+
"epoch": 2.419989904088844,
|
1798 |
+
"grad_norm": 70.14442443847656,
|
1799 |
+
"learning_rate": 5e-06,
|
1800 |
+
"logits/chosen": -1.0711498260498047,
|
1801 |
+
"logits/rejected": -1.0283135175704956,
|
1802 |
+
"logps/chosen": -139.4008026123047,
|
1803 |
+
"logps/rejected": -173.4601287841797,
|
1804 |
+
"loss": 0.4815,
|
1805 |
+
"rewards/accuracies": 0.699999988079071,
|
1806 |
+
"rewards/chosen": -1.1457339525222778,
|
1807 |
+
"rewards/margins": 0.9765189290046692,
|
1808 |
+
"rewards/rejected": -2.122252941131592,
|
1809 |
+
"step": 600
|
1810 |
+
},
|
1811 |
+
{
|
1812 |
+
"epoch": 2.440181726400808,
|
1813 |
+
"grad_norm": 39.94354248046875,
|
1814 |
+
"learning_rate": 5e-06,
|
1815 |
+
"logits/chosen": -1.0897289514541626,
|
1816 |
+
"logits/rejected": -1.0542991161346436,
|
1817 |
+
"logps/chosen": -140.98178100585938,
|
1818 |
+
"logps/rejected": -154.21890258789062,
|
1819 |
+
"loss": 0.3809,
|
1820 |
+
"rewards/accuracies": 0.8999999761581421,
|
1821 |
+
"rewards/chosen": -0.9919537305831909,
|
1822 |
+
"rewards/margins": 1.0165085792541504,
|
1823 |
+
"rewards/rejected": -2.008462429046631,
|
1824 |
+
"step": 605
|
1825 |
+
},
|
1826 |
+
{
|
1827 |
+
"epoch": 2.4603735487127714,
|
1828 |
+
"grad_norm": 45.401206970214844,
|
1829 |
+
"learning_rate": 5e-06,
|
1830 |
+
"logits/chosen": -1.088404893875122,
|
1831 |
+
"logits/rejected": -1.0834258794784546,
|
1832 |
+
"logps/chosen": -132.83590698242188,
|
1833 |
+
"logps/rejected": -148.06394958496094,
|
1834 |
+
"loss": 0.456,
|
1835 |
+
"rewards/accuracies": 0.8500000238418579,
|
1836 |
+
"rewards/chosen": -0.9679088592529297,
|
1837 |
+
"rewards/margins": 0.8238266110420227,
|
1838 |
+
"rewards/rejected": -1.7917354106903076,
|
1839 |
+
"step": 610
|
1840 |
+
},
|
1841 |
+
{
|
1842 |
+
"epoch": 2.480565371024735,
|
1843 |
+
"grad_norm": 70.00696563720703,
|
1844 |
+
"learning_rate": 5e-06,
|
1845 |
+
"logits/chosen": -1.0881659984588623,
|
1846 |
+
"logits/rejected": -1.0491617918014526,
|
1847 |
+
"logps/chosen": -126.14036560058594,
|
1848 |
+
"logps/rejected": -139.42196655273438,
|
1849 |
+
"loss": 0.4363,
|
1850 |
+
"rewards/accuracies": 0.800000011920929,
|
1851 |
+
"rewards/chosen": -0.7842862010002136,
|
1852 |
+
"rewards/margins": 0.8065754771232605,
|
1853 |
+
"rewards/rejected": -1.5908616781234741,
|
1854 |
+
"step": 615
|
1855 |
+
},
|
1856 |
+
{
|
1857 |
+
"epoch": 2.5007571933366988,
|
1858 |
+
"grad_norm": 93.69404602050781,
|
1859 |
+
"learning_rate": 5e-06,
|
1860 |
+
"logits/chosen": -1.088850975036621,
|
1861 |
+
"logits/rejected": -1.061447262763977,
|
1862 |
+
"logps/chosen": -144.85092163085938,
|
1863 |
+
"logps/rejected": -187.40280151367188,
|
1864 |
+
"loss": 0.334,
|
1865 |
+
"rewards/accuracies": 0.8999999761581421,
|
1866 |
+
"rewards/chosen": -1.2291361093521118,
|
1867 |
+
"rewards/margins": 1.6782220602035522,
|
1868 |
+
"rewards/rejected": -2.907358169555664,
|
1869 |
+
"step": 620
|
1870 |
+
},
|
1871 |
+
{
|
1872 |
+
"epoch": 2.5209490156486623,
|
1873 |
+
"grad_norm": 24.65225601196289,
|
1874 |
+
"learning_rate": 5e-06,
|
1875 |
+
"logits/chosen": -1.1317861080169678,
|
1876 |
+
"logits/rejected": -1.0619679689407349,
|
1877 |
+
"logps/chosen": -150.44094848632812,
|
1878 |
+
"logps/rejected": -178.79635620117188,
|
1879 |
+
"loss": 0.4855,
|
1880 |
+
"rewards/accuracies": 0.7749999761581421,
|
1881 |
+
"rewards/chosen": -1.2844429016113281,
|
1882 |
+
"rewards/margins": 1.1455484628677368,
|
1883 |
+
"rewards/rejected": -2.4299912452697754,
|
1884 |
+
"step": 625
|
1885 |
+
},
|
1886 |
+
{
|
1887 |
+
"epoch": 2.5411408379606257,
|
1888 |
+
"grad_norm": 81.89047241210938,
|
1889 |
+
"learning_rate": 5e-06,
|
1890 |
+
"logits/chosen": -1.108019471168518,
|
1891 |
+
"logits/rejected": -1.1051324605941772,
|
1892 |
+
"logps/chosen": -144.0532989501953,
|
1893 |
+
"logps/rejected": -162.6993865966797,
|
1894 |
+
"loss": 0.4032,
|
1895 |
+
"rewards/accuracies": 0.800000011920929,
|
1896 |
+
"rewards/chosen": -1.3340096473693848,
|
1897 |
+
"rewards/margins": 1.0980908870697021,
|
1898 |
+
"rewards/rejected": -2.432100772857666,
|
1899 |
+
"step": 630
|
1900 |
+
},
|
1901 |
+
{
|
1902 |
+
"epoch": 2.5613326602725897,
|
1903 |
+
"grad_norm": 143.68875122070312,
|
1904 |
+
"learning_rate": 5e-06,
|
1905 |
+
"logits/chosen": -1.0815242528915405,
|
1906 |
+
"logits/rejected": -1.0449540615081787,
|
1907 |
+
"logps/chosen": -138.04147338867188,
|
1908 |
+
"logps/rejected": -156.7818603515625,
|
1909 |
+
"loss": 0.4695,
|
1910 |
+
"rewards/accuracies": 0.75,
|
1911 |
+
"rewards/chosen": -1.1757011413574219,
|
1912 |
+
"rewards/margins": 1.038955807685852,
|
1913 |
+
"rewards/rejected": -2.2146568298339844,
|
1914 |
+
"step": 635
|
1915 |
+
},
|
1916 |
+
{
|
1917 |
+
"epoch": 2.581524482584553,
|
1918 |
+
"grad_norm": 118.63373565673828,
|
1919 |
+
"learning_rate": 5e-06,
|
1920 |
+
"logits/chosen": -1.1306525468826294,
|
1921 |
+
"logits/rejected": -1.1150437593460083,
|
1922 |
+
"logps/chosen": -152.9892578125,
|
1923 |
+
"logps/rejected": -168.2422332763672,
|
1924 |
+
"loss": 0.5398,
|
1925 |
+
"rewards/accuracies": 0.7250000238418579,
|
1926 |
+
"rewards/chosen": -1.6195358037948608,
|
1927 |
+
"rewards/margins": 1.2459442615509033,
|
1928 |
+
"rewards/rejected": -2.8654799461364746,
|
1929 |
+
"step": 640
|
1930 |
+
},
|
1931 |
+
{
|
1932 |
+
"epoch": 2.601716304896517,
|
1933 |
+
"grad_norm": 85.33989715576172,
|
1934 |
+
"learning_rate": 5e-06,
|
1935 |
+
"logits/chosen": -1.1470824480056763,
|
1936 |
+
"logits/rejected": -1.1052191257476807,
|
1937 |
+
"logps/chosen": -144.22109985351562,
|
1938 |
+
"logps/rejected": -171.24172973632812,
|
1939 |
+
"loss": 0.3907,
|
1940 |
+
"rewards/accuracies": 0.8500000238418579,
|
1941 |
+
"rewards/chosen": -1.3878965377807617,
|
1942 |
+
"rewards/margins": 1.330134630203247,
|
1943 |
+
"rewards/rejected": -2.718031167984009,
|
1944 |
+
"step": 645
|
1945 |
+
},
|
1946 |
+
{
|
1947 |
+
"epoch": 2.6219081272084805,
|
1948 |
+
"grad_norm": 28.526046752929688,
|
1949 |
+
"learning_rate": 5e-06,
|
1950 |
+
"logits/chosen": -1.1499074697494507,
|
1951 |
+
"logits/rejected": -1.1181514263153076,
|
1952 |
+
"logps/chosen": -141.45933532714844,
|
1953 |
+
"logps/rejected": -165.86814880371094,
|
1954 |
+
"loss": 0.3265,
|
1955 |
+
"rewards/accuracies": 0.925000011920929,
|
1956 |
+
"rewards/chosen": -1.2811788320541382,
|
1957 |
+
"rewards/margins": 1.2787883281707764,
|
1958 |
+
"rewards/rejected": -2.559967279434204,
|
1959 |
+
"step": 650
|
1960 |
+
},
|
1961 |
+
{
|
1962 |
+
"epoch": 2.6420999495204445,
|
1963 |
+
"grad_norm": 50.67562484741211,
|
1964 |
+
"learning_rate": 5e-06,
|
1965 |
+
"logits/chosen": -1.0879682302474976,
|
1966 |
+
"logits/rejected": -1.0264991521835327,
|
1967 |
+
"logps/chosen": -140.26417541503906,
|
1968 |
+
"logps/rejected": -162.7800750732422,
|
1969 |
+
"loss": 0.3963,
|
1970 |
+
"rewards/accuracies": 0.8999999761581421,
|
1971 |
+
"rewards/chosen": -1.3028833866119385,
|
1972 |
+
"rewards/margins": 1.1044894456863403,
|
1973 |
+
"rewards/rejected": -2.4073727130889893,
|
1974 |
+
"step": 655
|
1975 |
+
},
|
1976 |
+
{
|
1977 |
+
"epoch": 2.662291771832408,
|
1978 |
+
"grad_norm": 71.36746215820312,
|
1979 |
+
"learning_rate": 5e-06,
|
1980 |
+
"logits/chosen": -1.12251877784729,
|
1981 |
+
"logits/rejected": -1.0848548412322998,
|
1982 |
+
"logps/chosen": -137.73040771484375,
|
1983 |
+
"logps/rejected": -154.519287109375,
|
1984 |
+
"loss": 0.45,
|
1985 |
+
"rewards/accuracies": 0.800000011920929,
|
1986 |
+
"rewards/chosen": -1.2700433731079102,
|
1987 |
+
"rewards/margins": 0.8259071111679077,
|
1988 |
+
"rewards/rejected": -2.095950126647949,
|
1989 |
+
"step": 660
|
1990 |
+
},
|
1991 |
+
{
|
1992 |
+
"epoch": 2.6824835941443714,
|
1993 |
+
"grad_norm": 145.13929748535156,
|
1994 |
+
"learning_rate": 5e-06,
|
1995 |
+
"logits/chosen": -1.1168526411056519,
|
1996 |
+
"logits/rejected": -1.0883421897888184,
|
1997 |
+
"logps/chosen": -138.43850708007812,
|
1998 |
+
"logps/rejected": -161.28207397460938,
|
1999 |
+
"loss": 0.4334,
|
2000 |
+
"rewards/accuracies": 0.8999999761581421,
|
2001 |
+
"rewards/chosen": -1.4023481607437134,
|
2002 |
+
"rewards/margins": 1.0314326286315918,
|
2003 |
+
"rewards/rejected": -2.4337806701660156,
|
2004 |
+
"step": 665
|
2005 |
+
},
|
2006 |
+
{
|
2007 |
+
"epoch": 2.7026754164563354,
|
2008 |
+
"grad_norm": 101.96822357177734,
|
2009 |
+
"learning_rate": 5e-06,
|
2010 |
+
"logits/chosen": -1.1315155029296875,
|
2011 |
+
"logits/rejected": -1.1191498041152954,
|
2012 |
+
"logps/chosen": -142.5680694580078,
|
2013 |
+
"logps/rejected": -147.30868530273438,
|
2014 |
+
"loss": 0.4604,
|
2015 |
+
"rewards/accuracies": 0.8500000238418579,
|
2016 |
+
"rewards/chosen": -1.2466862201690674,
|
2017 |
+
"rewards/margins": 0.8339215517044067,
|
2018 |
+
"rewards/rejected": -2.0806076526641846,
|
2019 |
+
"step": 670
|
2020 |
+
},
|
2021 |
+
{
|
2022 |
+
"epoch": 2.722867238768299,
|
2023 |
+
"grad_norm": 35.541744232177734,
|
2024 |
+
"learning_rate": 5e-06,
|
2025 |
+
"logits/chosen": -1.1304982900619507,
|
2026 |
+
"logits/rejected": -1.096704363822937,
|
2027 |
+
"logps/chosen": -140.98458862304688,
|
2028 |
+
"logps/rejected": -153.99526977539062,
|
2029 |
+
"loss": 0.3437,
|
2030 |
+
"rewards/accuracies": 0.875,
|
2031 |
+
"rewards/chosen": -1.1470431089401245,
|
2032 |
+
"rewards/margins": 1.193063735961914,
|
2033 |
+
"rewards/rejected": -2.340106964111328,
|
2034 |
+
"step": 675
|
2035 |
+
},
|
2036 |
+
{
|
2037 |
+
"epoch": 2.7430590610802623,
|
2038 |
+
"grad_norm": 119.16114807128906,
|
2039 |
+
"learning_rate": 5e-06,
|
2040 |
+
"logits/chosen": -1.1810123920440674,
|
2041 |
+
"logits/rejected": -1.1393203735351562,
|
2042 |
+
"logps/chosen": -143.31198120117188,
|
2043 |
+
"logps/rejected": -155.2026824951172,
|
2044 |
+
"loss": 0.4346,
|
2045 |
+
"rewards/accuracies": 0.7749999761581421,
|
2046 |
+
"rewards/chosen": -1.3686244487762451,
|
2047 |
+
"rewards/margins": 0.9831773638725281,
|
2048 |
+
"rewards/rejected": -2.351801633834839,
|
2049 |
+
"step": 680
|
2050 |
+
},
|
2051 |
+
{
|
2052 |
+
"epoch": 2.7632508833922262,
|
2053 |
+
"grad_norm": 112.19788360595703,
|
2054 |
+
"learning_rate": 5e-06,
|
2055 |
+
"logits/chosen": -1.1026071310043335,
|
2056 |
+
"logits/rejected": -1.115681767463684,
|
2057 |
+
"logps/chosen": -145.30560302734375,
|
2058 |
+
"logps/rejected": -155.7900848388672,
|
2059 |
+
"loss": 0.5291,
|
2060 |
+
"rewards/accuracies": 0.7749999761581421,
|
2061 |
+
"rewards/chosen": -1.4804359674453735,
|
2062 |
+
"rewards/margins": 0.9093244671821594,
|
2063 |
+
"rewards/rejected": -2.3897604942321777,
|
2064 |
+
"step": 685
|
2065 |
+
},
|
2066 |
+
{
|
2067 |
+
"epoch": 2.7834427057041897,
|
2068 |
+
"grad_norm": 33.10359191894531,
|
2069 |
+
"learning_rate": 5e-06,
|
2070 |
+
"logits/chosen": -1.1413536071777344,
|
2071 |
+
"logits/rejected": -1.1235370635986328,
|
2072 |
+
"logps/chosen": -137.90969848632812,
|
2073 |
+
"logps/rejected": -170.1824188232422,
|
2074 |
+
"loss": 0.3372,
|
2075 |
+
"rewards/accuracies": 0.875,
|
2076 |
+
"rewards/chosen": -1.2158892154693604,
|
2077 |
+
"rewards/margins": 1.584923505783081,
|
2078 |
+
"rewards/rejected": -2.8008127212524414,
|
2079 |
+
"step": 690
|
2080 |
+
},
|
2081 |
+
{
|
2082 |
+
"epoch": 2.803634528016153,
|
2083 |
+
"grad_norm": 150.4509735107422,
|
2084 |
+
"learning_rate": 5e-06,
|
2085 |
+
"logits/chosen": -1.1553657054901123,
|
2086 |
+
"logits/rejected": -1.1448795795440674,
|
2087 |
+
"logps/chosen": -137.4596405029297,
|
2088 |
+
"logps/rejected": -151.9920196533203,
|
2089 |
+
"loss": 0.4594,
|
2090 |
+
"rewards/accuracies": 0.800000011920929,
|
2091 |
+
"rewards/chosen": -1.3479053974151611,
|
2092 |
+
"rewards/margins": 0.890475869178772,
|
2093 |
+
"rewards/rejected": -2.2383813858032227,
|
2094 |
+
"step": 695
|
2095 |
+
},
|
2096 |
+
{
|
2097 |
+
"epoch": 2.823826350328117,
|
2098 |
+
"grad_norm": 287.857666015625,
|
2099 |
+
"learning_rate": 5e-06,
|
2100 |
+
"logits/chosen": -1.1810104846954346,
|
2101 |
+
"logits/rejected": -1.1226354837417603,
|
2102 |
+
"logps/chosen": -143.69639587402344,
|
2103 |
+
"logps/rejected": -163.6051483154297,
|
2104 |
+
"loss": 0.4457,
|
2105 |
+
"rewards/accuracies": 0.800000011920929,
|
2106 |
+
"rewards/chosen": -1.513098955154419,
|
2107 |
+
"rewards/margins": 1.0407955646514893,
|
2108 |
+
"rewards/rejected": -2.5538947582244873,
|
2109 |
+
"step": 700
|
2110 |
+
},
|
2111 |
+
{
|
2112 |
+
"epoch": 2.8440181726400806,
|
2113 |
+
"grad_norm": 50.71148681640625,
|
2114 |
+
"learning_rate": 5e-06,
|
2115 |
+
"logits/chosen": -1.1496613025665283,
|
2116 |
+
"logits/rejected": -1.12812340259552,
|
2117 |
+
"logps/chosen": -140.67123413085938,
|
2118 |
+
"logps/rejected": -169.8626708984375,
|
2119 |
+
"loss": 0.3212,
|
2120 |
+
"rewards/accuracies": 0.949999988079071,
|
2121 |
+
"rewards/chosen": -1.3559614419937134,
|
2122 |
+
"rewards/margins": 1.4378584623336792,
|
2123 |
+
"rewards/rejected": -2.7938199043273926,
|
2124 |
+
"step": 705
|
2125 |
+
},
|
2126 |
+
{
|
2127 |
+
"epoch": 2.8642099949520445,
|
2128 |
+
"grad_norm": 26.175968170166016,
|
2129 |
+
"learning_rate": 5e-06,
|
2130 |
+
"logits/chosen": -1.2208508253097534,
|
2131 |
+
"logits/rejected": -1.1588385105133057,
|
2132 |
+
"logps/chosen": -129.68093872070312,
|
2133 |
+
"logps/rejected": -161.8412322998047,
|
2134 |
+
"loss": 0.3767,
|
2135 |
+
"rewards/accuracies": 0.8500000238418579,
|
2136 |
+
"rewards/chosen": -1.222773790359497,
|
2137 |
+
"rewards/margins": 1.361515998840332,
|
2138 |
+
"rewards/rejected": -2.58428955078125,
|
2139 |
+
"step": 710
|
2140 |
+
},
|
2141 |
+
{
|
2142 |
+
"epoch": 2.884401817264008,
|
2143 |
+
"grad_norm": 51.06369400024414,
|
2144 |
+
"learning_rate": 5e-06,
|
2145 |
+
"logits/chosen": -1.1848516464233398,
|
2146 |
+
"logits/rejected": -1.1376698017120361,
|
2147 |
+
"logps/chosen": -131.61441040039062,
|
2148 |
+
"logps/rejected": -154.9497528076172,
|
2149 |
+
"loss": 0.3265,
|
2150 |
+
"rewards/accuracies": 0.925000011920929,
|
2151 |
+
"rewards/chosen": -1.088381290435791,
|
2152 |
+
"rewards/margins": 1.4161834716796875,
|
2153 |
+
"rewards/rejected": -2.5045647621154785,
|
2154 |
+
"step": 715
|
2155 |
+
},
|
2156 |
+
{
|
2157 |
+
"epoch": 2.904593639575972,
|
2158 |
+
"grad_norm": 35.54250717163086,
|
2159 |
+
"learning_rate": 5e-06,
|
2160 |
+
"logits/chosen": -1.1926292181015015,
|
2161 |
+
"logits/rejected": -1.157402753829956,
|
2162 |
+
"logps/chosen": -136.55596923828125,
|
2163 |
+
"logps/rejected": -163.8975372314453,
|
2164 |
+
"loss": 0.2961,
|
2165 |
+
"rewards/accuracies": 0.8999999761581421,
|
2166 |
+
"rewards/chosen": -1.1022846698760986,
|
2167 |
+
"rewards/margins": 1.461648941040039,
|
2168 |
+
"rewards/rejected": -2.563933849334717,
|
2169 |
+
"step": 720
|
2170 |
+
},
|
2171 |
+
{
|
2172 |
+
"epoch": 2.9247854618879354,
|
2173 |
+
"grad_norm": 155.65060424804688,
|
2174 |
+
"learning_rate": 5e-06,
|
2175 |
+
"logits/chosen": -1.1305577754974365,
|
2176 |
+
"logits/rejected": -1.085231065750122,
|
2177 |
+
"logps/chosen": -148.7545928955078,
|
2178 |
+
"logps/rejected": -176.36215209960938,
|
2179 |
+
"loss": 0.426,
|
2180 |
+
"rewards/accuracies": 0.824999988079071,
|
2181 |
+
"rewards/chosen": -1.75044846534729,
|
2182 |
+
"rewards/margins": 1.306915521621704,
|
2183 |
+
"rewards/rejected": -3.057363986968994,
|
2184 |
+
"step": 725
|
2185 |
+
},
|
2186 |
+
{
|
2187 |
+
"epoch": 2.944977284199899,
|
2188 |
+
"grad_norm": 83.46701049804688,
|
2189 |
+
"learning_rate": 5e-06,
|
2190 |
+
"logits/chosen": -1.1913352012634277,
|
2191 |
+
"logits/rejected": -1.1349766254425049,
|
2192 |
+
"logps/chosen": -132.3568115234375,
|
2193 |
+
"logps/rejected": -161.4365234375,
|
2194 |
+
"loss": 0.4037,
|
2195 |
+
"rewards/accuracies": 0.800000011920929,
|
2196 |
+
"rewards/chosen": -1.3909555673599243,
|
2197 |
+
"rewards/margins": 1.264525055885315,
|
2198 |
+
"rewards/rejected": -2.6554808616638184,
|
2199 |
+
"step": 730
|
2200 |
+
},
|
2201 |
+
{
|
2202 |
+
"epoch": 2.965169106511863,
|
2203 |
+
"grad_norm": 29.703113555908203,
|
2204 |
+
"learning_rate": 5e-06,
|
2205 |
+
"logits/chosen": -1.1975009441375732,
|
2206 |
+
"logits/rejected": -1.1907942295074463,
|
2207 |
+
"logps/chosen": -126.33477783203125,
|
2208 |
+
"logps/rejected": -158.81204223632812,
|
2209 |
+
"loss": 0.4499,
|
2210 |
+
"rewards/accuracies": 0.7250000238418579,
|
2211 |
+
"rewards/chosen": -1.3245980739593506,
|
2212 |
+
"rewards/margins": 1.180716872215271,
|
2213 |
+
"rewards/rejected": -2.5053153038024902,
|
2214 |
+
"step": 735
|
2215 |
+
},
|
2216 |
+
{
|
2217 |
+
"epoch": 2.9853609288238263,
|
2218 |
+
"grad_norm": 41.88002014160156,
|
2219 |
+
"learning_rate": 5e-06,
|
2220 |
+
"logits/chosen": -1.211860179901123,
|
2221 |
+
"logits/rejected": -1.1881868839263916,
|
2222 |
+
"logps/chosen": -141.8226776123047,
|
2223 |
+
"logps/rejected": -175.9602508544922,
|
2224 |
+
"loss": 0.3985,
|
2225 |
+
"rewards/accuracies": 0.8500000238418579,
|
2226 |
+
"rewards/chosen": -1.6403604745864868,
|
2227 |
+
"rewards/margins": 1.4641687870025635,
|
2228 |
+
"rewards/rejected": -3.1045291423797607,
|
2229 |
+
"step": 740
|
2230 |
+
},
|
2231 |
+
{
|
2232 |
+
"epoch": 2.989399293286219,
|
2233 |
+
"step": 741,
|
2234 |
+
"total_flos": 0.0,
|
2235 |
+
"train_loss": 0.586526518971653,
|
2236 |
+
"train_runtime": 26068.2965,
|
2237 |
+
"train_samples_per_second": 0.228,
|
2238 |
+
"train_steps_per_second": 0.028
|
2239 |
+
}
|
2240 |
+
],
|
2241 |
+
"logging_steps": 5,
|
2242 |
+
"max_steps": 741,
|
2243 |
+
"num_input_tokens_seen": 0,
|
2244 |
+
"num_train_epochs": 3,
|
2245 |
+
"save_steps": 500,
|
2246 |
+
"stateful_callbacks": {
|
2247 |
+
"TrainerControl": {
|
2248 |
+
"args": {
|
2249 |
+
"should_epoch_stop": false,
|
2250 |
+
"should_evaluate": false,
|
2251 |
+
"should_log": false,
|
2252 |
+
"should_save": true,
|
2253 |
+
"should_training_stop": true
|
2254 |
+
},
|
2255 |
+
"attributes": {}
|
2256 |
+
}
|
2257 |
+
},
|
2258 |
+
"total_flos": 0.0,
|
2259 |
+
"train_batch_size": 1,
|
2260 |
+
"trial_name": null,
|
2261 |
+
"trial_params": null
|
2262 |
+
}
|