iaa01 commited on
Commit
6756c6b
·
verified ·
1 Parent(s): 77a6e6c

Model save

Browse files
Files changed (6) hide show
  1. README.md +68 -0
  2. all_results.json +8 -0
  3. config.json +1 -1
  4. generation_config.json +14 -0
  5. train_results.json +8 -0
  6. trainer_state.json +3192 -0
README.md ADDED
@@ -0,0 +1,68 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: Qwen/Qwen2.5-3B-Instruct
3
+ library_name: transformers
4
+ model_name: qwen-2.5-3b-r1-countdown
5
+ tags:
6
+ - generated_from_trainer
7
+ - trl
8
+ - grpo
9
+ licence: license
10
+ ---
11
+
12
+ # Model Card for qwen-2.5-3b-r1-countdown
13
+
14
+ This model is a fine-tuned version of [Qwen/Qwen2.5-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct).
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="iaa01/qwen-2.5-3b-r1-countdown", device="cuda")
24
+ output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
25
+ print(output["generated_text"])
26
+ ```
27
+
28
+ ## Training procedure
29
+
30
+ [<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/personal-iaa/huggingface/runs/21moehkl)
31
+
32
+
33
+ This model was trained with GRPO, a method introduced in [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://huggingface.co/papers/2402.03300).
34
+
35
+ ### Framework versions
36
+
37
+ - TRL: 0.16.0.dev0
38
+ - Transformers: 4.49.0
39
+ - Pytorch: 2.5.1+cu121
40
+ - Datasets: 3.4.1
41
+ - Tokenizers: 0.21.1
42
+
43
+ ## Citations
44
+
45
+ Cite GRPO as:
46
+
47
+ ```bibtex
48
+ @article{zhihong2024deepseekmath,
49
+ title = {{DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models}},
50
+ author = {Zhihong Shao and Peiyi Wang and Qihao Zhu and Runxin Xu and Junxiao Song and Mingchuan Zhang and Y. K. Li and Y. Wu and Daya Guo},
51
+ year = 2024,
52
+ eprint = {arXiv:2402.03300},
53
+ }
54
+
55
+ ```
56
+
57
+ Cite TRL as:
58
+
59
+ ```bibtex
60
+ @misc{vonwerra2022trl,
61
+ title = {{TRL: Transformer Reinforcement Learning}},
62
+ author = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallouédec},
63
+ year = 2020,
64
+ journal = {GitHub repository},
65
+ publisher = {GitHub},
66
+ howpublished = {\url{https://github.com/huggingface/trl}}
67
+ }
68
+ ```
all_results.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "total_flos": 0.0,
3
+ "train_loss": 0.04409932630071934,
4
+ "train_runtime": 12236.3143,
5
+ "train_samples": 45000,
6
+ "train_samples_per_second": 1.177,
7
+ "train_steps_per_second": 0.037
8
+ }
config.json CHANGED
@@ -23,7 +23,7 @@
23
  "tie_word_embeddings": true,
24
  "torch_dtype": "bfloat16",
25
  "transformers_version": "4.49.0",
26
- "use_cache": false,
27
  "use_sliding_window": false,
28
  "vocab_size": 151936
29
  }
 
23
  "tie_word_embeddings": true,
24
  "torch_dtype": "bfloat16",
25
  "transformers_version": "4.49.0",
26
+ "use_cache": true,
27
  "use_sliding_window": false,
28
  "vocab_size": 151936
29
  }
generation_config.json ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 151643,
3
+ "do_sample": true,
4
+ "eos_token_id": [
5
+ 151645,
6
+ 151643
7
+ ],
8
+ "pad_token_id": 151643,
9
+ "repetition_penalty": 1.05,
10
+ "temperature": 0.7,
11
+ "top_k": 20,
12
+ "top_p": 0.8,
13
+ "transformers_version": "4.49.0"
14
+ }
train_results.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "total_flos": 0.0,
3
+ "train_loss": 0.04409932630071934,
4
+ "train_runtime": 12236.3143,
5
+ "train_samples": 45000,
6
+ "train_samples_per_second": 1.177,
7
+ "train_steps_per_second": 0.037
8
+ }
trainer_state.json ADDED
@@ -0,0 +1,3192 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 0.08,
5
+ "eval_steps": 500,
6
+ "global_step": 450,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "clip_ratio": 0.0,
13
+ "completion_length": 497.109375,
14
+ "epoch": 0.00035555555555555557,
15
+ "grad_norm": 1.5540773635891456,
16
+ "kl": 0.0,
17
+ "learning_rate": 7.142857142857142e-08,
18
+ "loss": 0.1037,
19
+ "reward": 0.328125,
20
+ "reward_std": 0.4159187823534012,
21
+ "rewards/equation_reward_func": 0.0625,
22
+ "rewards/format_reward_func": 0.265625,
23
+ "step": 2
24
+ },
25
+ {
26
+ "clip_ratio": 0.0,
27
+ "completion_length": 532.40625,
28
+ "epoch": 0.0007111111111111111,
29
+ "grad_norm": 1.2357520393415342,
30
+ "kl": 0.0007323026657104492,
31
+ "learning_rate": 1.4285714285714285e-07,
32
+ "loss": 0.1034,
33
+ "reward": 0.25,
34
+ "reward_std": 0.3582531735301018,
35
+ "rewards/equation_reward_func": 0.0625,
36
+ "rewards/format_reward_func": 0.1875,
37
+ "step": 4
38
+ },
39
+ {
40
+ "clip_ratio": 0.0,
41
+ "completion_length": 501.78125,
42
+ "epoch": 0.0010666666666666667,
43
+ "grad_norm": 1.207958912839064,
44
+ "kl": 0.0007942914962768555,
45
+ "learning_rate": 2.1428571428571426e-07,
46
+ "loss": 0.1091,
47
+ "reward": 0.234375,
48
+ "reward_std": 0.3534187823534012,
49
+ "rewards/equation_reward_func": 0.03125,
50
+ "rewards/format_reward_func": 0.203125,
51
+ "step": 6
52
+ },
53
+ {
54
+ "clip_ratio": 0.0,
55
+ "completion_length": 472.109375,
56
+ "epoch": 0.0014222222222222223,
57
+ "grad_norm": 1.6667490438178998,
58
+ "kl": 0.000766754150390625,
59
+ "learning_rate": 2.857142857142857e-07,
60
+ "loss": 0.1158,
61
+ "reward": 0.390625,
62
+ "reward_std": 0.4387340322136879,
63
+ "rewards/equation_reward_func": 0.140625,
64
+ "rewards/format_reward_func": 0.25,
65
+ "step": 8
66
+ },
67
+ {
68
+ "clip_ratio": 0.0,
69
+ "completion_length": 508.5,
70
+ "epoch": 0.0017777777777777779,
71
+ "grad_norm": 1.7217421147986447,
72
+ "kl": 0.00079345703125,
73
+ "learning_rate": 3.5714285714285716e-07,
74
+ "loss": 0.0744,
75
+ "reward": 0.375,
76
+ "reward_std": 0.48542676120996475,
77
+ "rewards/equation_reward_func": 0.03125,
78
+ "rewards/format_reward_func": 0.34375,
79
+ "step": 10
80
+ },
81
+ {
82
+ "clip_ratio": 0.0,
83
+ "completion_length": 563.03125,
84
+ "epoch": 0.0021333333333333334,
85
+ "grad_norm": 1.4870770624622829,
86
+ "kl": 0.0007028579711914062,
87
+ "learning_rate": 4.285714285714285e-07,
88
+ "loss": 0.16,
89
+ "reward": 0.28125,
90
+ "reward_std": 0.3846687823534012,
91
+ "rewards/equation_reward_func": 0.03125,
92
+ "rewards/format_reward_func": 0.25,
93
+ "step": 12
94
+ },
95
+ {
96
+ "clip_ratio": 0.0,
97
+ "completion_length": 463.609375,
98
+ "epoch": 0.002488888888888889,
99
+ "grad_norm": 1.7635576195780378,
100
+ "kl": 0.0008461475372314453,
101
+ "learning_rate": 5e-07,
102
+ "loss": 0.0834,
103
+ "reward": 0.328125,
104
+ "reward_std": 0.46167195588350296,
105
+ "rewards/equation_reward_func": 0.03125,
106
+ "rewards/format_reward_func": 0.296875,
107
+ "step": 14
108
+ },
109
+ {
110
+ "clip_ratio": 0.0,
111
+ "completion_length": 485.484375,
112
+ "epoch": 0.0028444444444444446,
113
+ "grad_norm": 1.5351814757263196,
114
+ "kl": 0.0009121894836425781,
115
+ "learning_rate": 4.999740409224932e-07,
116
+ "loss": 0.1649,
117
+ "reward": 0.4375,
118
+ "reward_std": 0.45151595771312714,
119
+ "rewards/equation_reward_func": 0.109375,
120
+ "rewards/format_reward_func": 0.328125,
121
+ "step": 16
122
+ },
123
+ {
124
+ "clip_ratio": 0.0,
125
+ "completion_length": 473.578125,
126
+ "epoch": 0.0032,
127
+ "grad_norm": 1.790949387851571,
128
+ "kl": 0.0012161731719970703,
129
+ "learning_rate": 4.998961690809627e-07,
130
+ "loss": 0.1101,
131
+ "reward": 0.40625,
132
+ "reward_std": 0.5236847400665283,
133
+ "rewards/equation_reward_func": 0.046875,
134
+ "rewards/format_reward_func": 0.359375,
135
+ "step": 18
136
+ },
137
+ {
138
+ "clip_ratio": 0.0,
139
+ "completion_length": 437.8125,
140
+ "epoch": 0.0035555555555555557,
141
+ "grad_norm": 1.8537703194309358,
142
+ "kl": 0.0023336410522460938,
143
+ "learning_rate": 4.997664006472578e-07,
144
+ "loss": 0.1244,
145
+ "reward": 0.546875,
146
+ "reward_std": 0.42558756470680237,
147
+ "rewards/equation_reward_func": 0.046875,
148
+ "rewards/format_reward_func": 0.5,
149
+ "step": 20
150
+ },
151
+ {
152
+ "clip_ratio": 0.0,
153
+ "completion_length": 406.515625,
154
+ "epoch": 0.003911111111111111,
155
+ "grad_norm": 1.7394706416500154,
156
+ "kl": 0.0038557052612304688,
157
+ "learning_rate": 4.995847625707292e-07,
158
+ "loss": 0.0125,
159
+ "reward": 0.59375,
160
+ "reward_std": 0.5879059880971909,
161
+ "rewards/equation_reward_func": 0.0625,
162
+ "rewards/format_reward_func": 0.53125,
163
+ "step": 22
164
+ },
165
+ {
166
+ "clip_ratio": 0.0,
167
+ "completion_length": 459.703125,
168
+ "epoch": 0.004266666666666667,
169
+ "grad_norm": 1.454676166117581,
170
+ "kl": 0.004199504852294922,
171
+ "learning_rate": 4.993512925726318e-07,
172
+ "loss": 0.0336,
173
+ "reward": 0.625,
174
+ "reward_std": 0.40400634706020355,
175
+ "rewards/equation_reward_func": 0.015625,
176
+ "rewards/format_reward_func": 0.609375,
177
+ "step": 24
178
+ },
179
+ {
180
+ "clip_ratio": 0.0,
181
+ "completion_length": 481.953125,
182
+ "epoch": 0.004622222222222222,
183
+ "grad_norm": 1.7838057007741426,
184
+ "kl": 0.0038619041442871094,
185
+ "learning_rate": 4.990660391382923e-07,
186
+ "loss": 0.0952,
187
+ "reward": 0.765625,
188
+ "reward_std": 0.34194982051849365,
189
+ "rewards/equation_reward_func": 0.03125,
190
+ "rewards/format_reward_func": 0.734375,
191
+ "step": 26
192
+ },
193
+ {
194
+ "clip_ratio": 0.0,
195
+ "completion_length": 443.234375,
196
+ "epoch": 0.004977777777777778,
197
+ "grad_norm": 2.0657041070771247,
198
+ "kl": 0.00693511962890625,
199
+ "learning_rate": 4.987290615070384e-07,
200
+ "loss": 0.0515,
201
+ "reward": 0.796875,
202
+ "reward_std": 0.5276496410369873,
203
+ "rewards/equation_reward_func": 0.078125,
204
+ "rewards/format_reward_func": 0.71875,
205
+ "step": 28
206
+ },
207
+ {
208
+ "clip_ratio": 0.0,
209
+ "completion_length": 524.359375,
210
+ "epoch": 0.005333333333333333,
211
+ "grad_norm": 1.5368658141890732,
212
+ "kl": 0.007770538330078125,
213
+ "learning_rate": 4.983404296598978e-07,
214
+ "loss": 0.1967,
215
+ "reward": 0.75,
216
+ "reward_std": 0.5364577993750572,
217
+ "rewards/equation_reward_func": 0.0625,
218
+ "rewards/format_reward_func": 0.6875,
219
+ "step": 30
220
+ },
221
+ {
222
+ "clip_ratio": 0.0,
223
+ "completion_length": 455.6875,
224
+ "epoch": 0.005688888888888889,
225
+ "grad_norm": 1.7349608259853453,
226
+ "kl": 0.009317398071289062,
227
+ "learning_rate": 4.979002243050646e-07,
228
+ "loss": 0.1176,
229
+ "reward": 0.875,
230
+ "reward_std": 0.47960907220840454,
231
+ "rewards/equation_reward_func": 0.109375,
232
+ "rewards/format_reward_func": 0.765625,
233
+ "step": 32
234
+ },
235
+ {
236
+ "clip_ratio": 0.0,
237
+ "completion_length": 477.53125,
238
+ "epoch": 0.006044444444444444,
239
+ "grad_norm": 1.2174715114943861,
240
+ "kl": 0.011419296264648438,
241
+ "learning_rate": 4.974085368611381e-07,
242
+ "loss": 0.0599,
243
+ "reward": 0.96875,
244
+ "reward_std": 0.3978152498602867,
245
+ "rewards/equation_reward_func": 0.140625,
246
+ "rewards/format_reward_func": 0.828125,
247
+ "step": 34
248
+ },
249
+ {
250
+ "clip_ratio": 0.0,
251
+ "completion_length": 544.34375,
252
+ "epoch": 0.0064,
253
+ "grad_norm": 0.9373070959302399,
254
+ "kl": 0.009111404418945312,
255
+ "learning_rate": 4.968654694381379e-07,
256
+ "loss": 0.0615,
257
+ "reward": 0.8125,
258
+ "reward_std": 0.35645299404859543,
259
+ "rewards/equation_reward_func": 0.03125,
260
+ "rewards/format_reward_func": 0.78125,
261
+ "step": 36
262
+ },
263
+ {
264
+ "clip_ratio": 0.0,
265
+ "completion_length": 522.078125,
266
+ "epoch": 0.0067555555555555554,
267
+ "grad_norm": 1.69272651339043,
268
+ "kl": 0.01029205322265625,
269
+ "learning_rate": 4.962711348162987e-07,
270
+ "loss": 0.1866,
271
+ "reward": 0.84375,
272
+ "reward_std": 0.5025907382369041,
273
+ "rewards/equation_reward_func": 0.125,
274
+ "rewards/format_reward_func": 0.71875,
275
+ "step": 38
276
+ },
277
+ {
278
+ "clip_ratio": 0.0,
279
+ "completion_length": 507.984375,
280
+ "epoch": 0.0071111111111111115,
281
+ "grad_norm": 1.2015156610010047,
282
+ "kl": 0.012607574462890625,
283
+ "learning_rate": 4.956256564226487e-07,
284
+ "loss": 0.0963,
285
+ "reward": 1.015625,
286
+ "reward_std": 0.407927505671978,
287
+ "rewards/equation_reward_func": 0.140625,
288
+ "rewards/format_reward_func": 0.875,
289
+ "step": 40
290
+ },
291
+ {
292
+ "clip_ratio": 0.0,
293
+ "completion_length": 444.296875,
294
+ "epoch": 0.007466666666666667,
295
+ "grad_norm": 1.1020450998312568,
296
+ "kl": 0.015117645263671875,
297
+ "learning_rate": 4.949291683053768e-07,
298
+ "loss": 0.1049,
299
+ "reward": 0.953125,
300
+ "reward_std": 0.38331207633018494,
301
+ "rewards/equation_reward_func": 0.078125,
302
+ "rewards/format_reward_func": 0.875,
303
+ "step": 42
304
+ },
305
+ {
306
+ "clip_ratio": 0.0,
307
+ "completion_length": 467.03125,
308
+ "epoch": 0.007822222222222222,
309
+ "grad_norm": 1.4356691747538686,
310
+ "kl": 0.015163421630859375,
311
+ "learning_rate": 4.941818151059955e-07,
312
+ "loss": 0.0563,
313
+ "reward": 1.0,
314
+ "reward_std": 0.31069982051849365,
315
+ "rewards/equation_reward_func": 0.140625,
316
+ "rewards/format_reward_func": 0.859375,
317
+ "step": 44
318
+ },
319
+ {
320
+ "clip_ratio": 0.0,
321
+ "completion_length": 466.1875,
322
+ "epoch": 0.008177777777777779,
323
+ "grad_norm": 1.1138441770229877,
324
+ "kl": 0.014316558837890625,
325
+ "learning_rate": 4.933837520293017e-07,
326
+ "loss": 0.0513,
327
+ "reward": 0.953125,
328
+ "reward_std": 0.20728103816509247,
329
+ "rewards/equation_reward_func": 0.03125,
330
+ "rewards/format_reward_func": 0.921875,
331
+ "step": 46
332
+ },
333
+ {
334
+ "clip_ratio": 0.0,
335
+ "completion_length": 483.21875,
336
+ "epoch": 0.008533333333333334,
337
+ "grad_norm": 1.5073854647894624,
338
+ "kl": 0.0155029296875,
339
+ "learning_rate": 4.925351448111454e-07,
340
+ "loss": 0.0312,
341
+ "reward": 0.984375,
342
+ "reward_std": 0.37053901702165604,
343
+ "rewards/equation_reward_func": 0.109375,
344
+ "rewards/format_reward_func": 0.875,
345
+ "step": 48
346
+ },
347
+ {
348
+ "clip_ratio": 0.0,
349
+ "completion_length": 441.59375,
350
+ "epoch": 0.008888888888888889,
351
+ "grad_norm": 1.0379492454910968,
352
+ "kl": 0.015254974365234375,
353
+ "learning_rate": 4.91636169684011e-07,
354
+ "loss": 0.0637,
355
+ "reward": 1.015625,
356
+ "reward_std": 0.24336542934179306,
357
+ "rewards/equation_reward_func": 0.09375,
358
+ "rewards/format_reward_func": 0.921875,
359
+ "step": 50
360
+ },
361
+ {
362
+ "clip_ratio": 0.0,
363
+ "completion_length": 491.46875,
364
+ "epoch": 0.009244444444444444,
365
+ "grad_norm": 1.3198501259898483,
366
+ "kl": 0.016269683837890625,
367
+ "learning_rate": 4.906870133404186e-07,
368
+ "loss": 0.062,
369
+ "reward": 1.0,
370
+ "reward_std": 0.31069982051849365,
371
+ "rewards/equation_reward_func": 0.09375,
372
+ "rewards/format_reward_func": 0.90625,
373
+ "step": 52
374
+ },
375
+ {
376
+ "clip_ratio": 0.0,
377
+ "completion_length": 450.234375,
378
+ "epoch": 0.0096,
379
+ "grad_norm": 1.1484527235656556,
380
+ "kl": 0.021270751953125,
381
+ "learning_rate": 4.896878728941531e-07,
382
+ "loss": 0.0533,
383
+ "reward": 1.015625,
384
+ "reward_std": 0.408423587679863,
385
+ "rewards/equation_reward_func": 0.140625,
386
+ "rewards/format_reward_func": 0.875,
387
+ "step": 54
388
+ },
389
+ {
390
+ "clip_ratio": 0.0,
391
+ "completion_length": 408.03125,
392
+ "epoch": 0.009955555555555556,
393
+ "grad_norm": 1.0788691812319753,
394
+ "kl": 0.0235595703125,
395
+ "learning_rate": 4.886389558393284e-07,
396
+ "loss": -0.0358,
397
+ "reward": 1.015625,
398
+ "reward_std": 0.20728103816509247,
399
+ "rewards/equation_reward_func": 0.0625,
400
+ "rewards/format_reward_func": 0.953125,
401
+ "step": 56
402
+ },
403
+ {
404
+ "clip_ratio": 0.0,
405
+ "completion_length": 487.09375,
406
+ "epoch": 0.010311111111111111,
407
+ "grad_norm": 1.4656646934435877,
408
+ "kl": 0.02265167236328125,
409
+ "learning_rate": 4.875404800072976e-07,
410
+ "loss": 0.0995,
411
+ "reward": 0.953125,
412
+ "reward_std": 0.26978103816509247,
413
+ "rewards/equation_reward_func": 0.046875,
414
+ "rewards/format_reward_func": 0.90625,
415
+ "step": 58
416
+ },
417
+ {
418
+ "clip_ratio": 0.0,
419
+ "completion_length": 435.6875,
420
+ "epoch": 0.010666666666666666,
421
+ "grad_norm": 1.1291328060832448,
422
+ "kl": 0.02149200439453125,
423
+ "learning_rate": 4.86392673521415e-07,
424
+ "loss": 0.0992,
425
+ "reward": 1.046875,
426
+ "reward_std": 0.1923343911767006,
427
+ "rewards/equation_reward_func": 0.078125,
428
+ "rewards/format_reward_func": 0.96875,
429
+ "step": 60
430
+ },
431
+ {
432
+ "clip_ratio": 0.0,
433
+ "completion_length": 453.59375,
434
+ "epoch": 0.011022222222222221,
435
+ "grad_norm": 0.7889399494362057,
436
+ "kl": 0.025543212890625,
437
+ "learning_rate": 4.851957747496606e-07,
438
+ "loss": 0.0177,
439
+ "reward": 1.03125,
440
+ "reward_std": 0.09858439117670059,
441
+ "rewards/equation_reward_func": 0.046875,
442
+ "rewards/format_reward_func": 0.984375,
443
+ "step": 62
444
+ },
445
+ {
446
+ "clip_ratio": 0.0,
447
+ "completion_length": 413.859375,
448
+ "epoch": 0.011377777777777778,
449
+ "grad_norm": 0.8220739984789721,
450
+ "kl": 0.02587890625,
451
+ "learning_rate": 4.839500322551386e-07,
452
+ "loss": 0.0939,
453
+ "reward": 1.0625,
454
+ "reward_std": 0.22706207633018494,
455
+ "rewards/equation_reward_func": 0.09375,
456
+ "rewards/format_reward_func": 0.96875,
457
+ "step": 64
458
+ },
459
+ {
460
+ "clip_ratio": 0.0,
461
+ "completion_length": 394.328125,
462
+ "epoch": 0.011733333333333333,
463
+ "grad_norm": 0.3782612004884283,
464
+ "kl": 0.02734375,
465
+ "learning_rate": 4.826557047444563e-07,
466
+ "loss": 0.0026,
467
+ "reward": 1.015625,
468
+ "reward_std": 0.09375,
469
+ "rewards/equation_reward_func": 0.03125,
470
+ "rewards/format_reward_func": 0.984375,
471
+ "step": 66
472
+ },
473
+ {
474
+ "clip_ratio": 0.0,
475
+ "completion_length": 437.765625,
476
+ "epoch": 0.012088888888888889,
477
+ "grad_norm": 1.0169159547009763,
478
+ "kl": 0.0303802490234375,
479
+ "learning_rate": 4.813130610139993e-07,
480
+ "loss": 0.1134,
481
+ "reward": 1.09375,
482
+ "reward_std": 0.34678421169519424,
483
+ "rewards/equation_reward_func": 0.15625,
484
+ "rewards/format_reward_func": 0.9375,
485
+ "step": 68
486
+ },
487
+ {
488
+ "clip_ratio": 0.0,
489
+ "completion_length": 395.046875,
490
+ "epoch": 0.012444444444444444,
491
+ "grad_norm": 1.1708246723062659,
492
+ "kl": 0.0349884033203125,
493
+ "learning_rate": 4.799223798941089e-07,
494
+ "loss": 0.0274,
495
+ "reward": 1.0625,
496
+ "reward_std": 0.1610843911767006,
497
+ "rewards/equation_reward_func": 0.078125,
498
+ "rewards/format_reward_func": 0.984375,
499
+ "step": 70
500
+ },
501
+ {
502
+ "clip_ratio": 0.0,
503
+ "completion_length": 409.109375,
504
+ "epoch": 0.0128,
505
+ "grad_norm": 0.9619605643432636,
506
+ "kl": 0.041015625,
507
+ "learning_rate": 4.78483950191177e-07,
508
+ "loss": 0.0364,
509
+ "reward": 1.03125,
510
+ "reward_std": 0.1610843911767006,
511
+ "rewards/equation_reward_func": 0.0625,
512
+ "rewards/format_reward_func": 0.96875,
513
+ "step": 72
514
+ },
515
+ {
516
+ "clip_ratio": 0.0,
517
+ "completion_length": 406.109375,
518
+ "epoch": 0.013155555555555556,
519
+ "grad_norm": 1.5109017153953268,
520
+ "kl": 0.036956787109375,
521
+ "learning_rate": 4.769980706276687e-07,
522
+ "loss": 0.1356,
523
+ "reward": 1.15625,
524
+ "reward_std": 0.3713996410369873,
525
+ "rewards/equation_reward_func": 0.203125,
526
+ "rewards/format_reward_func": 0.953125,
527
+ "step": 74
528
+ },
529
+ {
530
+ "clip_ratio": 0.0,
531
+ "completion_length": 401.671875,
532
+ "epoch": 0.013511111111111111,
533
+ "grad_norm": 1.4832479712654314,
534
+ "kl": 0.0571441650390625,
535
+ "learning_rate": 4.7546504978008595e-07,
536
+ "loss": 0.0329,
537
+ "reward": 1.078125,
538
+ "reward_std": 0.15625,
539
+ "rewards/equation_reward_func": 0.109375,
540
+ "rewards/format_reward_func": 0.96875,
541
+ "step": 76
542
+ },
543
+ {
544
+ "clip_ratio": 0.0,
545
+ "completion_length": 383.390625,
546
+ "epoch": 0.013866666666666666,
547
+ "grad_norm": 1.6506361354745738,
548
+ "kl": 0.04904937744140625,
549
+ "learning_rate": 4.738852060148848e-07,
550
+ "loss": 0.0359,
551
+ "reward": 1.109375,
552
+ "reward_std": 0.1923343911767006,
553
+ "rewards/equation_reward_func": 0.125,
554
+ "rewards/format_reward_func": 0.984375,
555
+ "step": 78
556
+ },
557
+ {
558
+ "clip_ratio": 0.0,
559
+ "completion_length": 428.203125,
560
+ "epoch": 0.014222222222222223,
561
+ "grad_norm": 1.2805072360716652,
562
+ "kl": 0.0862884521484375,
563
+ "learning_rate": 4.722588674223593e-07,
564
+ "loss": 0.0975,
565
+ "reward": 1.0625,
566
+ "reward_std": 0.2596687823534012,
567
+ "rewards/equation_reward_func": 0.109375,
568
+ "rewards/format_reward_func": 0.953125,
569
+ "step": 80
570
+ },
571
+ {
572
+ "clip_ratio": 0.0,
573
+ "completion_length": 334.765625,
574
+ "epoch": 0.014577777777777778,
575
+ "grad_norm": 1.5003659203947817,
576
+ "kl": 0.083343505859375,
577
+ "learning_rate": 4.70586371748506e-07,
578
+ "loss": 0.0372,
579
+ "reward": 1.140625,
580
+ "reward_std": 0.20200317353010178,
581
+ "rewards/equation_reward_func": 0.140625,
582
+ "rewards/format_reward_func": 1.0,
583
+ "step": 82
584
+ },
585
+ {
586
+ "clip_ratio": 0.0,
587
+ "completion_length": 375.625,
588
+ "epoch": 0.014933333333333333,
589
+ "grad_norm": 1.4069013627610942,
590
+ "kl": 0.0662841796875,
591
+ "learning_rate": 4.6886806632488363e-07,
592
+ "loss": 0.0729,
593
+ "reward": 1.140625,
594
+ "reward_std": 0.21694982051849365,
595
+ "rewards/equation_reward_func": 0.15625,
596
+ "rewards/format_reward_func": 0.984375,
597
+ "step": 84
598
+ },
599
+ {
600
+ "clip_ratio": 0.0,
601
+ "completion_length": 374.171875,
602
+ "epoch": 0.015288888888888888,
603
+ "grad_norm": 1.533899888374318,
604
+ "kl": 0.11968994140625,
605
+ "learning_rate": 4.6710430799648143e-07,
606
+ "loss": 0.0581,
607
+ "reward": 1.15625,
608
+ "reward_std": 0.2235843911767006,
609
+ "rewards/equation_reward_func": 0.15625,
610
+ "rewards/format_reward_func": 1.0,
611
+ "step": 86
612
+ },
613
+ {
614
+ "clip_ratio": 0.0,
615
+ "completion_length": 338.90625,
616
+ "epoch": 0.015644444444444443,
617
+ "grad_norm": 1.213080736189347,
618
+ "kl": 0.24658203125,
619
+ "learning_rate": 4.652954630476127e-07,
620
+ "loss": 0.0379,
621
+ "reward": 1.078125,
622
+ "reward_std": 0.1298343911767006,
623
+ "rewards/equation_reward_func": 0.078125,
624
+ "rewards/format_reward_func": 1.0,
625
+ "step": 88
626
+ },
627
+ {
628
+ "clip_ratio": 0.0,
629
+ "completion_length": 371.796875,
630
+ "epoch": 0.016,
631
+ "grad_norm": 1.709623394747104,
632
+ "kl": 1.02197265625,
633
+ "learning_rate": 4.6344190712584713e-07,
634
+ "loss": -0.0009,
635
+ "reward": 1.15625,
636
+ "reward_std": 0.23325317353010178,
637
+ "rewards/equation_reward_func": 0.15625,
638
+ "rewards/format_reward_func": 1.0,
639
+ "step": 90
640
+ },
641
+ {
642
+ "clip_ratio": 0.0,
643
+ "completion_length": 372.4375,
644
+ "epoch": 0.016355555555555557,
645
+ "grad_norm": 1.1039741000193948,
646
+ "kl": 0.51763916015625,
647
+ "learning_rate": 4.615440251639995e-07,
648
+ "loss": 0.0138,
649
+ "reward": 1.03125,
650
+ "reward_std": 0.09858439117670059,
651
+ "rewards/equation_reward_func": 0.046875,
652
+ "rewards/format_reward_func": 0.984375,
653
+ "step": 92
654
+ },
655
+ {
656
+ "clip_ratio": 0.0,
657
+ "completion_length": 383.03125,
658
+ "epoch": 0.01671111111111111,
659
+ "grad_norm": 1.0615946335422568,
660
+ "kl": 0.32763671875,
661
+ "learning_rate": 4.596022113001894e-07,
662
+ "loss": 0.0168,
663
+ "reward": 1.0625,
664
+ "reward_std": 0.09858439117670059,
665
+ "rewards/equation_reward_func": 0.0625,
666
+ "rewards/format_reward_func": 1.0,
667
+ "step": 94
668
+ },
669
+ {
670
+ "clip_ratio": 0.0,
671
+ "completion_length": 387.046875,
672
+ "epoch": 0.017066666666666667,
673
+ "grad_norm": 1.0567721975237463,
674
+ "kl": 0.186248779296875,
675
+ "learning_rate": 4.576168687959895e-07,
676
+ "loss": 0.014,
677
+ "reward": 1.125,
678
+ "reward_std": 0.17075317353010178,
679
+ "rewards/equation_reward_func": 0.125,
680
+ "rewards/format_reward_func": 1.0,
681
+ "step": 96
682
+ },
683
+ {
684
+ "clip_ratio": 0.0,
685
+ "completion_length": 357.375,
686
+ "epoch": 0.01742222222222222,
687
+ "grad_norm": 117.02412491992783,
688
+ "kl": 8.14306640625,
689
+ "learning_rate": 4.555884099526793e-07,
690
+ "loss": 0.0505,
691
+ "reward": 1.09375,
692
+ "reward_std": 0.13466878235340118,
693
+ "rewards/equation_reward_func": 0.09375,
694
+ "rewards/format_reward_func": 1.0,
695
+ "step": 98
696
+ },
697
+ {
698
+ "clip_ratio": 0.0,
699
+ "completion_length": 361.265625,
700
+ "epoch": 0.017777777777777778,
701
+ "grad_norm": 1.2109959204361576,
702
+ "kl": 0.1463623046875,
703
+ "learning_rate": 4.5351725602562174e-07,
704
+ "loss": -0.0073,
705
+ "reward": 1.125,
706
+ "reward_std": 0.24819982051849365,
707
+ "rewards/equation_reward_func": 0.140625,
708
+ "rewards/format_reward_func": 0.984375,
709
+ "step": 100
710
+ },
711
+ {
712
+ "clip_ratio": 0.0,
713
+ "completion_length": 374.34375,
714
+ "epoch": 0.018133333333333335,
715
+ "grad_norm": 1.3629062892707553,
716
+ "kl": 0.22503662109375,
717
+ "learning_rate": 4.514038371367791e-07,
718
+ "loss": 0.0304,
719
+ "reward": 1.109375,
720
+ "reward_std": 0.16591878235340118,
721
+ "rewards/equation_reward_func": 0.109375,
722
+ "rewards/format_reward_func": 1.0,
723
+ "step": 102
724
+ },
725
+ {
726
+ "clip_ratio": 0.0,
727
+ "completion_length": 355.671875,
728
+ "epoch": 0.018488888888888888,
729
+ "grad_norm": 1.066354910378942,
730
+ "kl": 0.237884521484375,
731
+ "learning_rate": 4.4924859218538936e-07,
732
+ "loss": 0.0266,
733
+ "reward": 1.0625,
734
+ "reward_std": 0.125,
735
+ "rewards/equation_reward_func": 0.0625,
736
+ "rewards/format_reward_func": 1.0,
737
+ "step": 104
738
+ },
739
+ {
740
+ "clip_ratio": 0.0,
741
+ "completion_length": 346.40625,
742
+ "epoch": 0.018844444444444445,
743
+ "grad_norm": 1.244076703781482,
744
+ "kl": 0.2713623046875,
745
+ "learning_rate": 4.470519687568185e-07,
746
+ "loss": 0.0573,
747
+ "reward": 1.15625,
748
+ "reward_std": 0.1610843911767006,
749
+ "rewards/equation_reward_func": 0.15625,
750
+ "rewards/format_reward_func": 1.0,
751
+ "step": 106
752
+ },
753
+ {
754
+ "clip_ratio": 0.0,
755
+ "completion_length": 343.875,
756
+ "epoch": 0.0192,
757
+ "grad_norm": 1.5317514782871564,
758
+ "kl": 0.193603515625,
759
+ "learning_rate": 4.4481442302960923e-07,
760
+ "loss": 0.0681,
761
+ "reward": 1.125,
762
+ "reward_std": 0.22178421169519424,
763
+ "rewards/equation_reward_func": 0.140625,
764
+ "rewards/format_reward_func": 0.984375,
765
+ "step": 108
766
+ },
767
+ {
768
+ "clip_ratio": 0.0,
769
+ "completion_length": 357.453125,
770
+ "epoch": 0.019555555555555555,
771
+ "grad_norm": 1.3615194783712425,
772
+ "kl": 0.257049560546875,
773
+ "learning_rate": 4.4253641968074505e-07,
774
+ "loss": 0.0324,
775
+ "reward": 1.046875,
776
+ "reward_std": 0.1298343911767006,
777
+ "rewards/equation_reward_func": 0.0625,
778
+ "rewards/format_reward_func": 0.984375,
779
+ "step": 110
780
+ },
781
+ {
782
+ "clip_ratio": 0.0,
783
+ "completion_length": 330.875,
784
+ "epoch": 0.019911111111111112,
785
+ "grad_norm": 1.6404324826318923,
786
+ "kl": 0.42608642578125,
787
+ "learning_rate": 4.402184317891501e-07,
788
+ "loss": 0.0444,
789
+ "reward": 1.15625,
790
+ "reward_std": 0.26933756470680237,
791
+ "rewards/equation_reward_func": 0.171875,
792
+ "rewards/format_reward_func": 0.984375,
793
+ "step": 112
794
+ },
795
+ {
796
+ "clip_ratio": 0.0,
797
+ "completion_length": 327.734375,
798
+ "epoch": 0.020266666666666665,
799
+ "grad_norm": 1.1970449285988152,
800
+ "kl": 0.30352783203125,
801
+ "learning_rate": 4.37860940737443e-07,
802
+ "loss": 0.0289,
803
+ "reward": 1.03125,
804
+ "reward_std": 0.125,
805
+ "rewards/equation_reward_func": 0.046875,
806
+ "rewards/format_reward_func": 0.984375,
807
+ "step": 114
808
+ },
809
+ {
810
+ "clip_ratio": 0.0,
811
+ "completion_length": 345.984375,
812
+ "epoch": 0.020622222222222222,
813
+ "grad_norm": 3.6830472868585975,
814
+ "kl": 0.693603515625,
815
+ "learning_rate": 4.354644361119671e-07,
816
+ "loss": 0.0444,
817
+ "reward": 1.03125,
818
+ "reward_std": 0.15358919650316238,
819
+ "rewards/equation_reward_func": 0.0625,
820
+ "rewards/format_reward_func": 0.96875,
821
+ "step": 116
822
+ },
823
+ {
824
+ "clip_ratio": 0.0,
825
+ "completion_length": 317.34375,
826
+ "epoch": 0.02097777777777778,
827
+ "grad_norm": 0.8803960644083081,
828
+ "kl": 0.40740966796875,
829
+ "learning_rate": 4.3302941560111716e-07,
830
+ "loss": 0.0669,
831
+ "reward": 1.09375,
832
+ "reward_std": 0.1875,
833
+ "rewards/equation_reward_func": 0.09375,
834
+ "rewards/format_reward_func": 1.0,
835
+ "step": 118
836
+ },
837
+ {
838
+ "clip_ratio": 0.0,
839
+ "completion_length": 325.359375,
840
+ "epoch": 0.021333333333333333,
841
+ "grad_norm": 1.7222894595317144,
842
+ "kl": 0.9599609375,
843
+ "learning_rate": 4.3055638489198236e-07,
844
+ "loss": 0.021,
845
+ "reward": 1.0625,
846
+ "reward_std": 0.1610843911767006,
847
+ "rewards/equation_reward_func": 0.078125,
848
+ "rewards/format_reward_func": 0.984375,
849
+ "step": 120
850
+ },
851
+ {
852
+ "clip_ratio": 0.0,
853
+ "completion_length": 343.03125,
854
+ "epoch": 0.02168888888888889,
855
+ "grad_norm": 3.1092809656603206,
856
+ "kl": 1.03076171875,
857
+ "learning_rate": 4.280458575653296e-07,
858
+ "loss": 0.0242,
859
+ "reward": 1.09375,
860
+ "reward_std": 0.1875,
861
+ "rewards/equation_reward_func": 0.09375,
862
+ "rewards/format_reward_func": 1.0,
863
+ "step": 122
864
+ },
865
+ {
866
+ "clip_ratio": 0.0,
867
+ "completion_length": 311.359375,
868
+ "epoch": 0.022044444444444443,
869
+ "grad_norm": 2.3189720464471533,
870
+ "kl": 0.631591796875,
871
+ "learning_rate": 4.2549835498894665e-07,
872
+ "loss": 0.0467,
873
+ "reward": 1.09375,
874
+ "reward_std": 0.2235843911767006,
875
+ "rewards/equation_reward_func": 0.125,
876
+ "rewards/format_reward_func": 0.96875,
877
+ "step": 124
878
+ },
879
+ {
880
+ "clip_ratio": 0.0,
881
+ "completion_length": 307.046875,
882
+ "epoch": 0.0224,
883
+ "grad_norm": 0.7441216613827443,
884
+ "kl": 0.5086669921875,
885
+ "learning_rate": 4.229144062093679e-07,
886
+ "loss": -0.0085,
887
+ "reward": 1.140625,
888
+ "reward_std": 0.20200317353010178,
889
+ "rewards/equation_reward_func": 0.140625,
890
+ "rewards/format_reward_func": 1.0,
891
+ "step": 126
892
+ },
893
+ {
894
+ "clip_ratio": 0.0,
895
+ "completion_length": 291.640625,
896
+ "epoch": 0.022755555555555557,
897
+ "grad_norm": 1.7398896558997667,
898
+ "kl": 0.682373046875,
899
+ "learning_rate": 4.2029454784200675e-07,
900
+ "loss": 0.0178,
901
+ "reward": 1.03125,
902
+ "reward_std": 0.0625,
903
+ "rewards/equation_reward_func": 0.046875,
904
+ "rewards/format_reward_func": 0.984375,
905
+ "step": 128
906
+ },
907
+ {
908
+ "clip_ratio": 0.0,
909
+ "completion_length": 286.921875,
910
+ "epoch": 0.02311111111111111,
911
+ "grad_norm": 1.4445137095871459,
912
+ "kl": 0.564208984375,
913
+ "learning_rate": 4.1763932395971433e-07,
914
+ "loss": 0.0149,
915
+ "reward": 1.078125,
916
+ "reward_std": 0.22841878235340118,
917
+ "rewards/equation_reward_func": 0.109375,
918
+ "rewards/format_reward_func": 0.96875,
919
+ "step": 130
920
+ },
921
+ {
922
+ "clip_ratio": 0.0,
923
+ "completion_length": 299.28125,
924
+ "epoch": 0.023466666666666667,
925
+ "grad_norm": 1.082942893318675,
926
+ "kl": 0.3787841796875,
927
+ "learning_rate": 4.1494928597979117e-07,
928
+ "loss": 0.0099,
929
+ "reward": 1.140625,
930
+ "reward_std": 0.22841878235340118,
931
+ "rewards/equation_reward_func": 0.140625,
932
+ "rewards/format_reward_func": 1.0,
933
+ "step": 132
934
+ },
935
+ {
936
+ "clip_ratio": 0.0,
937
+ "completion_length": 298.84375,
938
+ "epoch": 0.023822222222222224,
939
+ "grad_norm": 2.813667402859087,
940
+ "kl": 1.5950927734375,
941
+ "learning_rate": 4.122249925494726e-07,
942
+ "loss": 0.0468,
943
+ "reward": 1.21875,
944
+ "reward_std": 0.2957531735301018,
945
+ "rewards/equation_reward_func": 0.21875,
946
+ "rewards/format_reward_func": 1.0,
947
+ "step": 134
948
+ },
949
+ {
950
+ "clip_ratio": 0.0,
951
+ "completion_length": 304.90625,
952
+ "epoch": 0.024177777777777777,
953
+ "grad_norm": 1.6133284823417406,
954
+ "kl": 0.628173828125,
955
+ "learning_rate": 4.094670094299131e-07,
956
+ "loss": 0.0012,
957
+ "reward": 1.0625,
958
+ "reward_std": 0.09858439117670059,
959
+ "rewards/equation_reward_func": 0.0625,
960
+ "rewards/format_reward_func": 1.0,
961
+ "step": 136
962
+ },
963
+ {
964
+ "clip_ratio": 0.0,
965
+ "completion_length": 330.609375,
966
+ "epoch": 0.024533333333333334,
967
+ "grad_norm": 1.5618980968427136,
968
+ "kl": 0.620361328125,
969
+ "learning_rate": 4.066759093786931e-07,
970
+ "loss": 0.0299,
971
+ "reward": 1.046875,
972
+ "reward_std": 0.1298343911767006,
973
+ "rewards/equation_reward_func": 0.0625,
974
+ "rewards/format_reward_func": 0.984375,
975
+ "step": 138
976
+ },
977
+ {
978
+ "clip_ratio": 0.0,
979
+ "completion_length": 315.84375,
980
+ "epoch": 0.024888888888888887,
981
+ "grad_norm": 0.7496675007517801,
982
+ "kl": 0.5653076171875,
983
+ "learning_rate": 4.038522720308732e-07,
984
+ "loss": 0.0383,
985
+ "reward": 1.046875,
986
+ "reward_std": 0.06733439117670059,
987
+ "rewards/equation_reward_func": 0.046875,
988
+ "rewards/format_reward_func": 1.0,
989
+ "step": 140
990
+ },
991
+ {
992
+ "clip_ratio": 0.0,
993
+ "completion_length": 279.46875,
994
+ "epoch": 0.025244444444444444,
995
+ "grad_norm": 1.623499251970735,
996
+ "kl": 0.6290283203125,
997
+ "learning_rate": 4.009966837786194e-07,
998
+ "loss": 0.0041,
999
+ "reward": 1.046875,
1000
+ "reward_std": 0.09375,
1001
+ "rewards/equation_reward_func": 0.046875,
1002
+ "rewards/format_reward_func": 1.0,
1003
+ "step": 142
1004
+ },
1005
+ {
1006
+ "clip_ratio": 0.0,
1007
+ "completion_length": 301.1875,
1008
+ "epoch": 0.0256,
1009
+ "grad_norm": 1.5816454578936159,
1010
+ "kl": 0.6375732421875,
1011
+ "learning_rate": 3.981097376494259e-07,
1012
+ "loss": -0.0138,
1013
+ "reward": 1.0625,
1014
+ "reward_std": 0.09858439117670059,
1015
+ "rewards/equation_reward_func": 0.0625,
1016
+ "rewards/format_reward_func": 1.0,
1017
+ "step": 144
1018
+ },
1019
+ {
1020
+ "clip_ratio": 0.0,
1021
+ "completion_length": 286.390625,
1022
+ "epoch": 0.025955555555555555,
1023
+ "grad_norm": 0.9976268635346477,
1024
+ "kl": 0.594970703125,
1025
+ "learning_rate": 3.951920331829592e-07,
1026
+ "loss": 0.0283,
1027
+ "reward": 1.15625,
1028
+ "reward_std": 0.125,
1029
+ "rewards/equation_reward_func": 0.15625,
1030
+ "rewards/format_reward_func": 1.0,
1031
+ "step": 146
1032
+ },
1033
+ {
1034
+ "clip_ratio": 0.0,
1035
+ "completion_length": 290.21875,
1036
+ "epoch": 0.02631111111111111,
1037
+ "grad_norm": 1.453799485818376,
1038
+ "kl": 1.276611328125,
1039
+ "learning_rate": 3.922441763065506e-07,
1040
+ "loss": 0.0691,
1041
+ "reward": 1.0625,
1042
+ "reward_std": 0.1610843911767006,
1043
+ "rewards/equation_reward_func": 0.078125,
1044
+ "rewards/format_reward_func": 0.984375,
1045
+ "step": 148
1046
+ },
1047
+ {
1048
+ "clip_ratio": 0.0,
1049
+ "completion_length": 330.203125,
1050
+ "epoch": 0.02666666666666667,
1051
+ "grad_norm": 1.4585609222742653,
1052
+ "kl": 0.7332763671875,
1053
+ "learning_rate": 3.8926677920936093e-07,
1054
+ "loss": 0.0235,
1055
+ "reward": 1.0625,
1056
+ "reward_std": 0.125,
1057
+ "rewards/equation_reward_func": 0.0625,
1058
+ "rewards/format_reward_func": 1.0,
1059
+ "step": 150
1060
+ },
1061
+ {
1062
+ "clip_ratio": 0.0,
1063
+ "completion_length": 314.703125,
1064
+ "epoch": 0.027022222222222222,
1065
+ "grad_norm": 1.2450868511160544,
1066
+ "kl": 0.74951171875,
1067
+ "learning_rate": 3.862604602152464e-07,
1068
+ "loss": 0.0255,
1069
+ "reward": 1.03125,
1070
+ "reward_std": 0.0625,
1071
+ "rewards/equation_reward_func": 0.03125,
1072
+ "rewards/format_reward_func": 1.0,
1073
+ "step": 152
1074
+ },
1075
+ {
1076
+ "clip_ratio": 0.0,
1077
+ "completion_length": 313.203125,
1078
+ "epoch": 0.02737777777777778,
1079
+ "grad_norm": 1.3495788554258188,
1080
+ "kl": 0.805908203125,
1081
+ "learning_rate": 3.8322584365434934e-07,
1082
+ "loss": 0.0098,
1083
+ "reward": 1.03125,
1084
+ "reward_std": 0.09858439117670059,
1085
+ "rewards/equation_reward_func": 0.046875,
1086
+ "rewards/format_reward_func": 0.984375,
1087
+ "step": 154
1088
+ },
1089
+ {
1090
+ "clip_ratio": 0.0,
1091
+ "completion_length": 324.9375,
1092
+ "epoch": 0.027733333333333332,
1093
+ "grad_norm": 2.1466128173407117,
1094
+ "kl": 1.1510009765625,
1095
+ "learning_rate": 3.8016355973344173e-07,
1096
+ "loss": 0.09,
1097
+ "reward": 1.140625,
1098
+ "reward_std": 0.23808756470680237,
1099
+ "rewards/equation_reward_func": 0.15625,
1100
+ "rewards/format_reward_func": 0.984375,
1101
+ "step": 156
1102
+ },
1103
+ {
1104
+ "clip_ratio": 0.0,
1105
+ "completion_length": 324.859375,
1106
+ "epoch": 0.02808888888888889,
1107
+ "grad_norm": 2.204629350726093,
1108
+ "kl": 1.16015625,
1109
+ "learning_rate": 3.7707424440504863e-07,
1110
+ "loss": 0.0803,
1111
+ "reward": 1.140625,
1112
+ "reward_std": 0.2548343911767006,
1113
+ "rewards/equation_reward_func": 0.140625,
1114
+ "rewards/format_reward_func": 1.0,
1115
+ "step": 158
1116
+ },
1117
+ {
1118
+ "clip_ratio": 0.0,
1119
+ "completion_length": 302.859375,
1120
+ "epoch": 0.028444444444444446,
1121
+ "grad_norm": 1.8398150430903868,
1122
+ "kl": 1.251708984375,
1123
+ "learning_rate": 3.739585392353787e-07,
1124
+ "loss": 0.0197,
1125
+ "reward": 1.15625,
1126
+ "reward_std": 0.1610843911767006,
1127
+ "rewards/equation_reward_func": 0.15625,
1128
+ "rewards/format_reward_func": 1.0,
1129
+ "step": 160
1130
+ },
1131
+ {
1132
+ "clip_ratio": 0.0,
1133
+ "completion_length": 292.640625,
1134
+ "epoch": 0.0288,
1135
+ "grad_norm": 2.13847808834117,
1136
+ "kl": 1.3173828125,
1137
+ "learning_rate": 3.7081709127108767e-07,
1138
+ "loss": 0.0155,
1139
+ "reward": 1.1875,
1140
+ "reward_std": 0.2235843911767006,
1141
+ "rewards/equation_reward_func": 0.1875,
1142
+ "rewards/format_reward_func": 1.0,
1143
+ "step": 162
1144
+ },
1145
+ {
1146
+ "clip_ratio": 0.0,
1147
+ "completion_length": 261.234375,
1148
+ "epoch": 0.029155555555555556,
1149
+ "grad_norm": 3.506581199733204,
1150
+ "kl": 1.736328125,
1151
+ "learning_rate": 3.6765055290490513e-07,
1152
+ "loss": -0.0109,
1153
+ "reward": 1.171875,
1154
+ "reward_std": 0.28125,
1155
+ "rewards/equation_reward_func": 0.171875,
1156
+ "rewards/format_reward_func": 1.0,
1157
+ "step": 164
1158
+ },
1159
+ {
1160
+ "clip_ratio": 0.0,
1161
+ "completion_length": 302.953125,
1162
+ "epoch": 0.02951111111111111,
1163
+ "grad_norm": 2.5724923821581025,
1164
+ "kl": 6.1572265625,
1165
+ "learning_rate": 3.644595817401501e-07,
1166
+ "loss": 0.0332,
1167
+ "reward": 1.09375,
1168
+ "reward_std": 0.1610843911767006,
1169
+ "rewards/equation_reward_func": 0.09375,
1170
+ "rewards/format_reward_func": 1.0,
1171
+ "step": 166
1172
+ },
1173
+ {
1174
+ "clip_ratio": 0.0,
1175
+ "completion_length": 280.03125,
1176
+ "epoch": 0.029866666666666666,
1177
+ "grad_norm": 1.5126543710719036,
1178
+ "kl": 1.48974609375,
1179
+ "learning_rate": 3.6124484045416483e-07,
1180
+ "loss": 0.0267,
1181
+ "reward": 1.140625,
1182
+ "reward_std": 0.1923343911767006,
1183
+ "rewards/equation_reward_func": 0.140625,
1184
+ "rewards/format_reward_func": 1.0,
1185
+ "step": 168
1186
+ },
1187
+ {
1188
+ "clip_ratio": 0.0,
1189
+ "completion_length": 302.546875,
1190
+ "epoch": 0.030222222222222223,
1191
+ "grad_norm": 3.4982226449385156,
1192
+ "kl": 3.032470703125,
1193
+ "learning_rate": 3.580069966606949e-07,
1194
+ "loss": 0.0559,
1195
+ "reward": 1.03125,
1196
+ "reward_std": 0.1875,
1197
+ "rewards/equation_reward_func": 0.078125,
1198
+ "rewards/format_reward_func": 0.953125,
1199
+ "step": 170
1200
+ },
1201
+ {
1202
+ "clip_ratio": 0.0,
1203
+ "completion_length": 334.515625,
1204
+ "epoch": 0.030577777777777777,
1205
+ "grad_norm": 3.0985982569829638,
1206
+ "kl": 14.5616455078125,
1207
+ "learning_rate": 3.547467227712444e-07,
1208
+ "loss": 0.0302,
1209
+ "reward": 1.03125,
1210
+ "reward_std": 0.23853103816509247,
1211
+ "rewards/equation_reward_func": 0.078125,
1212
+ "rewards/format_reward_func": 0.953125,
1213
+ "step": 172
1214
+ },
1215
+ {
1216
+ "clip_ratio": 0.0,
1217
+ "completion_length": 300.71875,
1218
+ "epoch": 0.030933333333333334,
1219
+ "grad_norm": 10.688658199362251,
1220
+ "kl": 3.142333984375,
1221
+ "learning_rate": 3.5146469585543386e-07,
1222
+ "loss": -0.0049,
1223
+ "reward": 1.171875,
1224
+ "reward_std": 0.16591878235340118,
1225
+ "rewards/equation_reward_func": 0.171875,
1226
+ "rewards/format_reward_func": 1.0,
1227
+ "step": 174
1228
+ },
1229
+ {
1230
+ "clip_ratio": 0.0,
1231
+ "completion_length": 292.484375,
1232
+ "epoch": 0.03128888888888889,
1233
+ "grad_norm": 6.967303370780864,
1234
+ "kl": 4.5751953125,
1235
+ "learning_rate": 3.481615975003922e-07,
1236
+ "loss": 0.0089,
1237
+ "reward": 1.125,
1238
+ "reward_std": 0.13466878235340118,
1239
+ "rewards/equation_reward_func": 0.15625,
1240
+ "rewards/format_reward_func": 0.96875,
1241
+ "step": 176
1242
+ },
1243
+ {
1244
+ "clip_ratio": 0.0,
1245
+ "completion_length": 292.1875,
1246
+ "epoch": 0.03164444444444445,
1247
+ "grad_norm": 2.379129578155761,
1248
+ "kl": 1.24462890625,
1249
+ "learning_rate": 3.448381136692089e-07,
1250
+ "loss": 0.0412,
1251
+ "reward": 1.15625,
1252
+ "reward_std": 0.1610843911767006,
1253
+ "rewards/equation_reward_func": 0.15625,
1254
+ "rewards/format_reward_func": 1.0,
1255
+ "step": 178
1256
+ },
1257
+ {
1258
+ "clip_ratio": 0.0,
1259
+ "completion_length": 336.0625,
1260
+ "epoch": 0.032,
1261
+ "grad_norm": 1.8115427356791605,
1262
+ "kl": 0.7757568359375,
1263
+ "learning_rate": 3.4149493455847897e-07,
1264
+ "loss": 0.0274,
1265
+ "reward": 1.015625,
1266
+ "reward_std": 0.09375,
1267
+ "rewards/equation_reward_func": 0.03125,
1268
+ "rewards/format_reward_func": 0.984375,
1269
+ "step": 180
1270
+ },
1271
+ {
1272
+ "clip_ratio": 0.0,
1273
+ "completion_length": 297.34375,
1274
+ "epoch": 0.032355555555555554,
1275
+ "grad_norm": 2.279790073120627,
1276
+ "kl": 0.8720703125,
1277
+ "learning_rate": 3.3813275445496766e-07,
1278
+ "loss": 0.0637,
1279
+ "reward": 1.09375,
1280
+ "reward_std": 0.1610843911767006,
1281
+ "rewards/equation_reward_func": 0.09375,
1282
+ "rewards/format_reward_func": 1.0,
1283
+ "step": 182
1284
+ },
1285
+ {
1286
+ "clip_ratio": 0.0,
1287
+ "completion_length": 311.6875,
1288
+ "epoch": 0.032711111111111114,
1289
+ "grad_norm": 1.9111976240917574,
1290
+ "kl": 1.3289794921875,
1291
+ "learning_rate": 3.347522715914262e-07,
1292
+ "loss": -0.0126,
1293
+ "reward": 1.0625,
1294
+ "reward_std": 0.125,
1295
+ "rewards/equation_reward_func": 0.0625,
1296
+ "rewards/format_reward_func": 1.0,
1297
+ "step": 184
1298
+ },
1299
+ {
1300
+ "clip_ratio": 0.0,
1301
+ "completion_length": 284.0625,
1302
+ "epoch": 0.03306666666666667,
1303
+ "grad_norm": 48.68973900874978,
1304
+ "kl": 8.955810546875,
1305
+ "learning_rate": 3.313541880015877e-07,
1306
+ "loss": 0.0674,
1307
+ "reward": 1.171875,
1308
+ "reward_std": 0.1923343911767006,
1309
+ "rewards/equation_reward_func": 0.171875,
1310
+ "rewards/format_reward_func": 1.0,
1311
+ "step": 186
1312
+ },
1313
+ {
1314
+ "clip_ratio": 0.0,
1315
+ "completion_length": 334.1875,
1316
+ "epoch": 0.03342222222222222,
1317
+ "grad_norm": 0.3837227064051435,
1318
+ "kl": 0.7593994140625,
1319
+ "learning_rate": 3.279392093743747e-07,
1320
+ "loss": 0.0074,
1321
+ "reward": 1.015625,
1322
+ "reward_std": 0.03125,
1323
+ "rewards/equation_reward_func": 0.015625,
1324
+ "rewards/format_reward_func": 1.0,
1325
+ "step": 188
1326
+ },
1327
+ {
1328
+ "clip_ratio": 0.0,
1329
+ "completion_length": 276.140625,
1330
+ "epoch": 0.033777777777777775,
1331
+ "grad_norm": 2.2048786067003374,
1332
+ "kl": 0.738037109375,
1333
+ "learning_rate": 3.245080449073459e-07,
1334
+ "loss": 0.0079,
1335
+ "reward": 1.1875,
1336
+ "reward_std": 0.2596687823534012,
1337
+ "rewards/equation_reward_func": 0.1875,
1338
+ "rewards/format_reward_func": 1.0,
1339
+ "step": 190
1340
+ },
1341
+ {
1342
+ "clip_ratio": 0.0,
1343
+ "completion_length": 324.796875,
1344
+ "epoch": 0.034133333333333335,
1345
+ "grad_norm": 1.4988476602299554,
1346
+ "kl": 0.73046875,
1347
+ "learning_rate": 3.210614071594162e-07,
1348
+ "loss": 0.0788,
1349
+ "reward": 1.125,
1350
+ "reward_std": 0.19716878235340118,
1351
+ "rewards/equation_reward_func": 0.125,
1352
+ "rewards/format_reward_func": 1.0,
1353
+ "step": 192
1354
+ },
1355
+ {
1356
+ "clip_ratio": 0.0,
1357
+ "completion_length": 260.8125,
1358
+ "epoch": 0.03448888888888889,
1359
+ "grad_norm": 3.9875059912502744,
1360
+ "kl": 1.3214111328125,
1361
+ "learning_rate": 3.1760001190287695e-07,
1362
+ "loss": 0.0088,
1363
+ "reward": 1.140625,
1364
+ "reward_std": 0.1923343911767006,
1365
+ "rewards/equation_reward_func": 0.140625,
1366
+ "rewards/format_reward_func": 1.0,
1367
+ "step": 194
1368
+ },
1369
+ {
1370
+ "clip_ratio": 0.0,
1371
+ "completion_length": 278.03125,
1372
+ "epoch": 0.03484444444444444,
1373
+ "grad_norm": 2.1591229522344846,
1374
+ "kl": 0.6678466796875,
1375
+ "learning_rate": 3.141245779747502e-07,
1376
+ "loss": 0.0746,
1377
+ "reward": 1.203125,
1378
+ "reward_std": 0.30058756470680237,
1379
+ "rewards/equation_reward_func": 0.203125,
1380
+ "rewards/format_reward_func": 1.0,
1381
+ "step": 196
1382
+ },
1383
+ {
1384
+ "clip_ratio": 0.0,
1385
+ "completion_length": 333.765625,
1386
+ "epoch": 0.0352,
1387
+ "grad_norm": 3.410997217745853,
1388
+ "kl": 0.9820556640625,
1389
+ "learning_rate": 3.106358271275056e-07,
1390
+ "loss": 0.0738,
1391
+ "reward": 1.09375,
1392
+ "reward_std": 0.2596687823534012,
1393
+ "rewards/equation_reward_func": 0.125,
1394
+ "rewards/format_reward_func": 0.96875,
1395
+ "step": 198
1396
+ },
1397
+ {
1398
+ "clip_ratio": 0.0,
1399
+ "completion_length": 327.84375,
1400
+ "epoch": 0.035555555555555556,
1401
+ "grad_norm": 2.2453024821477534,
1402
+ "kl": 1.880126953125,
1403
+ "learning_rate": 3.0713448387917227e-07,
1404
+ "loss": -0.0002,
1405
+ "reward": 1.046875,
1406
+ "reward_std": 0.09375,
1407
+ "rewards/equation_reward_func": 0.046875,
1408
+ "rewards/format_reward_func": 1.0,
1409
+ "step": 200
1410
+ },
1411
+ {
1412
+ "clip_ratio": 0.0,
1413
+ "completion_length": 311.515625,
1414
+ "epoch": 0.03591111111111111,
1415
+ "grad_norm": 2.3752697048728186,
1416
+ "kl": 2.144775390625,
1417
+ "learning_rate": 3.0362127536287636e-07,
1418
+ "loss": 0.0192,
1419
+ "reward": 1.09375,
1420
+ "reward_std": 0.18569982051849365,
1421
+ "rewards/equation_reward_func": 0.109375,
1422
+ "rewards/format_reward_func": 0.984375,
1423
+ "step": 202
1424
+ },
1425
+ {
1426
+ "clip_ratio": 0.0,
1427
+ "completion_length": 284.3125,
1428
+ "epoch": 0.03626666666666667,
1429
+ "grad_norm": 135.00953365889976,
1430
+ "kl": 28.689697265625,
1431
+ "learning_rate": 3.0009693117583523e-07,
1432
+ "loss": 0.0403,
1433
+ "reward": 1.0625,
1434
+ "reward_std": 0.09858439117670059,
1435
+ "rewards/equation_reward_func": 0.0625,
1436
+ "rewards/format_reward_func": 1.0,
1437
+ "step": 204
1438
+ },
1439
+ {
1440
+ "clip_ratio": 0.0,
1441
+ "completion_length": 299.71875,
1442
+ "epoch": 0.03662222222222222,
1443
+ "grad_norm": 1.8674915516374395,
1444
+ "kl": 2.016845703125,
1445
+ "learning_rate": 2.965621832278401e-07,
1446
+ "loss": 0.049,
1447
+ "reward": 1.109375,
1448
+ "reward_std": 0.13950317353010178,
1449
+ "rewards/equation_reward_func": 0.109375,
1450
+ "rewards/format_reward_func": 1.0,
1451
+ "step": 206
1452
+ },
1453
+ {
1454
+ "clip_ratio": 0.0,
1455
+ "completion_length": 334.296875,
1456
+ "epoch": 0.036977777777777776,
1457
+ "grad_norm": 1.6829162960798076,
1458
+ "kl": 2.673828125,
1459
+ "learning_rate": 2.9301776558925875e-07,
1460
+ "loss": 0.0532,
1461
+ "reward": 1.0625,
1462
+ "reward_std": 0.12717358767986298,
1463
+ "rewards/equation_reward_func": 0.078125,
1464
+ "rewards/format_reward_func": 0.984375,
1465
+ "step": 208
1466
+ },
1467
+ {
1468
+ "clip_ratio": 0.0,
1469
+ "completion_length": 342.4375,
1470
+ "epoch": 0.037333333333333336,
1471
+ "grad_norm": 0.8086021793065252,
1472
+ "kl": 1.893310546875,
1473
+ "learning_rate": 2.894644143385885e-07,
1474
+ "loss": 0.0337,
1475
+ "reward": 1.078125,
1476
+ "reward_std": 0.09375,
1477
+ "rewards/equation_reward_func": 0.078125,
1478
+ "rewards/format_reward_func": 1.0,
1479
+ "step": 210
1480
+ },
1481
+ {
1482
+ "clip_ratio": 0.0,
1483
+ "completion_length": 307.765625,
1484
+ "epoch": 0.03768888888888889,
1485
+ "grad_norm": 2.4643376957229237,
1486
+ "kl": 1.91650390625,
1487
+ "learning_rate": 2.859028674095937e-07,
1488
+ "loss": 0.0027,
1489
+ "reward": 1.203125,
1490
+ "reward_std": 0.30058756470680237,
1491
+ "rewards/equation_reward_func": 0.203125,
1492
+ "rewards/format_reward_func": 1.0,
1493
+ "step": 212
1494
+ },
1495
+ {
1496
+ "clip_ratio": 0.0,
1497
+ "completion_length": 298.515625,
1498
+ "epoch": 0.03804444444444444,
1499
+ "grad_norm": 3.029946510557144,
1500
+ "kl": 1.754638671875,
1501
+ "learning_rate": 2.823338644380566e-07,
1502
+ "loss": 0.022,
1503
+ "reward": 1.15625,
1504
+ "reward_std": 0.17075317353010178,
1505
+ "rewards/equation_reward_func": 0.15625,
1506
+ "rewards/format_reward_func": 1.0,
1507
+ "step": 214
1508
+ },
1509
+ {
1510
+ "clip_ratio": 0.0,
1511
+ "completion_length": 315.765625,
1512
+ "epoch": 0.0384,
1513
+ "grad_norm": 1.810123541436162,
1514
+ "kl": 1.518310546875,
1515
+ "learning_rate": 2.7875814660817504e-07,
1516
+ "loss": -0.0159,
1517
+ "reward": 1.078125,
1518
+ "reward_std": 0.1298343911767006,
1519
+ "rewards/equation_reward_func": 0.09375,
1520
+ "rewards/format_reward_func": 0.984375,
1521
+ "step": 216
1522
+ },
1523
+ {
1524
+ "clip_ratio": 0.0,
1525
+ "completion_length": 269.6875,
1526
+ "epoch": 0.03875555555555556,
1527
+ "grad_norm": 2.337741190093351,
1528
+ "kl": 1.03564453125,
1529
+ "learning_rate": 2.751764564986396e-07,
1530
+ "loss": 0.0113,
1531
+ "reward": 1.140625,
1532
+ "reward_std": 0.22841878235340118,
1533
+ "rewards/equation_reward_func": 0.140625,
1534
+ "rewards/format_reward_func": 1.0,
1535
+ "step": 218
1536
+ },
1537
+ {
1538
+ "clip_ratio": 0.0,
1539
+ "completion_length": 329.484375,
1540
+ "epoch": 0.03911111111111111,
1541
+ "grad_norm": 0.7489189831580635,
1542
+ "kl": 1.310302734375,
1543
+ "learning_rate": 2.715895379284194e-07,
1544
+ "loss": 0.0084,
1545
+ "reward": 1.015625,
1546
+ "reward_std": 0.03125,
1547
+ "rewards/equation_reward_func": 0.015625,
1548
+ "rewards/format_reward_func": 1.0,
1549
+ "step": 220
1550
+ },
1551
+ {
1552
+ "clip_ratio": 0.0,
1553
+ "completion_length": 306.265625,
1554
+ "epoch": 0.039466666666666664,
1555
+ "grad_norm": 0.8093400893734075,
1556
+ "kl": 1.0770263671875,
1557
+ "learning_rate": 2.6799813580229174e-07,
1558
+ "loss": 0.0144,
1559
+ "reward": 1.015625,
1560
+ "reward_std": 0.03125,
1561
+ "rewards/equation_reward_func": 0.015625,
1562
+ "rewards/format_reward_func": 1.0,
1563
+ "step": 222
1564
+ },
1565
+ {
1566
+ "clip_ratio": 0.0,
1567
+ "completion_length": 286.640625,
1568
+ "epoch": 0.039822222222222224,
1569
+ "grad_norm": 1.6055403629730776,
1570
+ "kl": 1.1387939453125,
1571
+ "learning_rate": 2.6440299595614606e-07,
1572
+ "loss": 0.0386,
1573
+ "reward": 1.09375,
1574
+ "reward_std": 0.1610843911767006,
1575
+ "rewards/equation_reward_func": 0.09375,
1576
+ "rewards/format_reward_func": 1.0,
1577
+ "step": 224
1578
+ },
1579
+ {
1580
+ "clip_ratio": 0.0,
1581
+ "completion_length": 304.046875,
1582
+ "epoch": 0.04017777777777778,
1583
+ "grad_norm": 2.0302718515892164,
1584
+ "kl": 1.0672607421875,
1585
+ "learning_rate": 2.6080486500209347e-07,
1586
+ "loss": 0.0873,
1587
+ "reward": 1.1875,
1588
+ "reward_std": 0.2596687823534012,
1589
+ "rewards/equation_reward_func": 0.203125,
1590
+ "rewards/format_reward_func": 0.984375,
1591
+ "step": 226
1592
+ },
1593
+ {
1594
+ "clip_ratio": 0.0,
1595
+ "completion_length": 332.578125,
1596
+ "epoch": 0.04053333333333333,
1597
+ "grad_norm": 109.08080062943176,
1598
+ "kl": 12.7587890625,
1599
+ "learning_rate": 2.572044901734166e-07,
1600
+ "loss": 0.0036,
1601
+ "reward": 1.09375,
1602
+ "reward_std": 0.14961542934179306,
1603
+ "rewards/equation_reward_func": 0.109375,
1604
+ "rewards/format_reward_func": 0.984375,
1605
+ "step": 228
1606
+ },
1607
+ {
1608
+ "clip_ratio": 0.0,
1609
+ "completion_length": 361.203125,
1610
+ "epoch": 0.04088888888888889,
1611
+ "grad_norm": 3.7661219080037074,
1612
+ "kl": 2.75927734375,
1613
+ "learning_rate": 2.536026191693893e-07,
1614
+ "loss": 0.1178,
1615
+ "reward": 1.140625,
1616
+ "reward_std": 0.2645031735301018,
1617
+ "rewards/equation_reward_func": 0.171875,
1618
+ "rewards/format_reward_func": 0.96875,
1619
+ "step": 230
1620
+ },
1621
+ {
1622
+ "clip_ratio": 0.0,
1623
+ "completion_length": 325.75,
1624
+ "epoch": 0.041244444444444445,
1625
+ "grad_norm": 2.0246518816727224,
1626
+ "kl": 1.3505859375,
1627
+ "learning_rate": 2.5e-07,
1628
+ "loss": 0.0314,
1629
+ "reward": 1.15625,
1630
+ "reward_std": 0.20683756470680237,
1631
+ "rewards/equation_reward_func": 0.15625,
1632
+ "rewards/format_reward_func": 1.0,
1633
+ "step": 232
1634
+ },
1635
+ {
1636
+ "clip_ratio": 0.0,
1637
+ "completion_length": 301.09375,
1638
+ "epoch": 0.0416,
1639
+ "grad_norm": 3.4777576883520958,
1640
+ "kl": 1.7216796875,
1641
+ "learning_rate": 2.4639738083061073e-07,
1642
+ "loss": 0.0796,
1643
+ "reward": 1.140625,
1644
+ "reward_std": 0.28911860287189484,
1645
+ "rewards/equation_reward_func": 0.171875,
1646
+ "rewards/format_reward_func": 0.96875,
1647
+ "step": 234
1648
+ },
1649
+ {
1650
+ "clip_ratio": 0.0,
1651
+ "completion_length": 323.640625,
1652
+ "epoch": 0.04195555555555556,
1653
+ "grad_norm": 2.269233789387794,
1654
+ "kl": 1.789794921875,
1655
+ "learning_rate": 2.4279550982658345e-07,
1656
+ "loss": 0.0343,
1657
+ "reward": 1.0625,
1658
+ "reward_std": 0.1610843911767006,
1659
+ "rewards/equation_reward_func": 0.09375,
1660
+ "rewards/format_reward_func": 0.96875,
1661
+ "step": 236
1662
+ },
1663
+ {
1664
+ "clip_ratio": 0.0,
1665
+ "completion_length": 285.75,
1666
+ "epoch": 0.04231111111111111,
1667
+ "grad_norm": 3.232143067478558,
1668
+ "kl": 1.150146484375,
1669
+ "learning_rate": 2.3919513499790646e-07,
1670
+ "loss": 0.0351,
1671
+ "reward": 1.109375,
1672
+ "reward_std": 0.24733919650316238,
1673
+ "rewards/equation_reward_func": 0.125,
1674
+ "rewards/format_reward_func": 0.984375,
1675
+ "step": 238
1676
+ },
1677
+ {
1678
+ "clip_ratio": 0.0,
1679
+ "completion_length": 301.1875,
1680
+ "epoch": 0.042666666666666665,
1681
+ "grad_norm": 3.9490792042976843,
1682
+ "kl": 1.3486328125,
1683
+ "learning_rate": 2.3559700404385394e-07,
1684
+ "loss": 0.0803,
1685
+ "reward": 1.140625,
1686
+ "reward_std": 0.2909187823534012,
1687
+ "rewards/equation_reward_func": 0.171875,
1688
+ "rewards/format_reward_func": 0.96875,
1689
+ "step": 240
1690
+ },
1691
+ {
1692
+ "clip_ratio": 0.0,
1693
+ "completion_length": 290.25,
1694
+ "epoch": 0.043022222222222226,
1695
+ "grad_norm": 3.0823221892184796,
1696
+ "kl": 2.60009765625,
1697
+ "learning_rate": 2.3200186419770823e-07,
1698
+ "loss": 0.019,
1699
+ "reward": 1.125,
1700
+ "reward_std": 0.2596687823534012,
1701
+ "rewards/equation_reward_func": 0.140625,
1702
+ "rewards/format_reward_func": 0.984375,
1703
+ "step": 242
1704
+ },
1705
+ {
1706
+ "clip_ratio": 0.0,
1707
+ "completion_length": 330.953125,
1708
+ "epoch": 0.04337777777777778,
1709
+ "grad_norm": 8.855287463518104,
1710
+ "kl": 4.67138671875,
1711
+ "learning_rate": 2.284104620715807e-07,
1712
+ "loss": -0.002,
1713
+ "reward": 1.03125,
1714
+ "reward_std": 0.125,
1715
+ "rewards/equation_reward_func": 0.046875,
1716
+ "rewards/format_reward_func": 0.984375,
1717
+ "step": 244
1718
+ },
1719
+ {
1720
+ "clip_ratio": 0.0,
1721
+ "completion_length": 289.71875,
1722
+ "epoch": 0.04373333333333333,
1723
+ "grad_norm": 9.922124025430394,
1724
+ "kl": 7.11962890625,
1725
+ "learning_rate": 2.2482354350136043e-07,
1726
+ "loss": 0.0296,
1727
+ "reward": 1.046875,
1728
+ "reward_std": 0.09375,
1729
+ "rewards/equation_reward_func": 0.0625,
1730
+ "rewards/format_reward_func": 0.984375,
1731
+ "step": 246
1732
+ },
1733
+ {
1734
+ "clip_ratio": 0.0,
1735
+ "completion_length": 347.328125,
1736
+ "epoch": 0.044088888888888886,
1737
+ "grad_norm": 4.9145352811838805,
1738
+ "kl": 2.637939453125,
1739
+ "learning_rate": 2.2124185339182496e-07,
1740
+ "loss": 0.1017,
1741
+ "reward": 1.140625,
1742
+ "reward_std": 0.28911860287189484,
1743
+ "rewards/equation_reward_func": 0.171875,
1744
+ "rewards/format_reward_func": 0.96875,
1745
+ "step": 248
1746
+ },
1747
+ {
1748
+ "clip_ratio": 0.0,
1749
+ "completion_length": 308.96875,
1750
+ "epoch": 0.044444444444444446,
1751
+ "grad_norm": 2.1751878201806485,
1752
+ "kl": 3.7567138671875,
1753
+ "learning_rate": 2.1766613556194344e-07,
1754
+ "loss": 0.0106,
1755
+ "reward": 1.0625,
1756
+ "reward_std": 0.1875,
1757
+ "rewards/equation_reward_func": 0.078125,
1758
+ "rewards/format_reward_func": 0.984375,
1759
+ "step": 250
1760
+ },
1761
+ {
1762
+ "clip_ratio": 0.0,
1763
+ "completion_length": 304.1875,
1764
+ "epoch": 0.0448,
1765
+ "grad_norm": 0.5379516641958847,
1766
+ "kl": 1.821044921875,
1767
+ "learning_rate": 2.1409713259040628e-07,
1768
+ "loss": 0.0195,
1769
+ "reward": 1.0625,
1770
+ "reward_std": 0.0625,
1771
+ "rewards/equation_reward_func": 0.0625,
1772
+ "rewards/format_reward_func": 1.0,
1773
+ "step": 252
1774
+ },
1775
+ {
1776
+ "clip_ratio": 0.0,
1777
+ "completion_length": 309.921875,
1778
+ "epoch": 0.04515555555555555,
1779
+ "grad_norm": 1.5422659826295908,
1780
+ "kl": 2.447998046875,
1781
+ "learning_rate": 2.105355856614115e-07,
1782
+ "loss": 0.0521,
1783
+ "reward": 1.1875,
1784
+ "reward_std": 0.25,
1785
+ "rewards/equation_reward_func": 0.1875,
1786
+ "rewards/format_reward_func": 1.0,
1787
+ "step": 254
1788
+ },
1789
+ {
1790
+ "clip_ratio": 0.0,
1791
+ "completion_length": 312.9375,
1792
+ "epoch": 0.04551111111111111,
1793
+ "grad_norm": 1.881389397761525,
1794
+ "kl": 4.099609375,
1795
+ "learning_rate": 2.069822344107413e-07,
1796
+ "loss": -0.0096,
1797
+ "reward": 1.09375,
1798
+ "reward_std": 0.1610843911767006,
1799
+ "rewards/equation_reward_func": 0.09375,
1800
+ "rewards/format_reward_func": 1.0,
1801
+ "step": 256
1802
+ },
1803
+ {
1804
+ "clip_ratio": 0.0,
1805
+ "completion_length": 339.296875,
1806
+ "epoch": 0.04586666666666667,
1807
+ "grad_norm": 2.7062053922429947,
1808
+ "kl": 2.461669921875,
1809
+ "learning_rate": 2.034378167721599e-07,
1810
+ "loss": 0.0942,
1811
+ "reward": 1.140625,
1812
+ "reward_std": 0.20200317353010178,
1813
+ "rewards/equation_reward_func": 0.140625,
1814
+ "rewards/format_reward_func": 1.0,
1815
+ "step": 258
1816
+ },
1817
+ {
1818
+ "clip_ratio": 0.0,
1819
+ "completion_length": 309.84375,
1820
+ "epoch": 0.04622222222222222,
1821
+ "grad_norm": 3.420752165914464,
1822
+ "kl": 1.4273681640625,
1823
+ "learning_rate": 1.9990306882416485e-07,
1824
+ "loss": 0.0293,
1825
+ "reward": 1.109375,
1826
+ "reward_std": 0.22841878235340118,
1827
+ "rewards/equation_reward_func": 0.125,
1828
+ "rewards/format_reward_func": 0.984375,
1829
+ "step": 260
1830
+ },
1831
+ {
1832
+ "clip_ratio": 0.0,
1833
+ "completion_length": 330.09375,
1834
+ "epoch": 0.04657777777777778,
1835
+ "grad_norm": 1.4438609155615563,
1836
+ "kl": 3.030517578125,
1837
+ "learning_rate": 1.9637872463712362e-07,
1838
+ "loss": 0.0386,
1839
+ "reward": 1.078125,
1840
+ "reward_std": 0.1298343911767006,
1841
+ "rewards/equation_reward_func": 0.078125,
1842
+ "rewards/format_reward_func": 1.0,
1843
+ "step": 262
1844
+ },
1845
+ {
1846
+ "clip_ratio": 0.0,
1847
+ "completion_length": 293.234375,
1848
+ "epoch": 0.046933333333333334,
1849
+ "grad_norm": 3.756782187930511,
1850
+ "kl": 1.635498046875,
1851
+ "learning_rate": 1.9286551612082773e-07,
1852
+ "loss": 0.0679,
1853
+ "reward": 1.15625,
1854
+ "reward_std": 0.32036860287189484,
1855
+ "rewards/equation_reward_func": 0.1875,
1856
+ "rewards/format_reward_func": 0.96875,
1857
+ "step": 264
1858
+ },
1859
+ {
1860
+ "clip_ratio": 0.0,
1861
+ "completion_length": 355.5625,
1862
+ "epoch": 0.04728888888888889,
1863
+ "grad_norm": 5.013875423554075,
1864
+ "kl": 2.47119140625,
1865
+ "learning_rate": 1.8936417287249446e-07,
1866
+ "loss": 0.0631,
1867
+ "reward": 1.0625,
1868
+ "reward_std": 0.27461542934179306,
1869
+ "rewards/equation_reward_func": 0.109375,
1870
+ "rewards/format_reward_func": 0.953125,
1871
+ "step": 266
1872
+ },
1873
+ {
1874
+ "clip_ratio": 0.0,
1875
+ "completion_length": 310.15625,
1876
+ "epoch": 0.04764444444444445,
1877
+ "grad_norm": 0.5561076088635254,
1878
+ "kl": 1.971435546875,
1879
+ "learning_rate": 1.8587542202524985e-07,
1880
+ "loss": 0.0363,
1881
+ "reward": 1.0625,
1882
+ "reward_std": 0.09858439117670059,
1883
+ "rewards/equation_reward_func": 0.0625,
1884
+ "rewards/format_reward_func": 1.0,
1885
+ "step": 268
1886
+ },
1887
+ {
1888
+ "clip_ratio": 0.0,
1889
+ "completion_length": 337.90625,
1890
+ "epoch": 0.048,
1891
+ "grad_norm": 2.53746279303349,
1892
+ "kl": 4.04541015625,
1893
+ "learning_rate": 1.82399988097123e-07,
1894
+ "loss": 0.0317,
1895
+ "reward": 1.109375,
1896
+ "reward_std": 0.1923343911767006,
1897
+ "rewards/equation_reward_func": 0.125,
1898
+ "rewards/format_reward_func": 0.984375,
1899
+ "step": 270
1900
+ },
1901
+ {
1902
+ "clip_ratio": 0.0,
1903
+ "completion_length": 333.828125,
1904
+ "epoch": 0.048355555555555554,
1905
+ "grad_norm": 2.2242108525324134,
1906
+ "kl": 3.0537109375,
1907
+ "learning_rate": 1.7893859284058378e-07,
1908
+ "loss": 0.0213,
1909
+ "reward": 1.0,
1910
+ "reward_std": 0.17603103816509247,
1911
+ "rewards/equation_reward_func": 0.046875,
1912
+ "rewards/format_reward_func": 0.953125,
1913
+ "step": 272
1914
+ },
1915
+ {
1916
+ "clip_ratio": 0.0,
1917
+ "completion_length": 303.53125,
1918
+ "epoch": 0.04871111111111111,
1919
+ "grad_norm": 2.6756650642649373,
1920
+ "kl": 2.84130859375,
1921
+ "learning_rate": 1.7549195509265407e-07,
1922
+ "loss": 0.0452,
1923
+ "reward": 1.109375,
1924
+ "reward_std": 0.1923343911767006,
1925
+ "rewards/equation_reward_func": 0.125,
1926
+ "rewards/format_reward_func": 0.984375,
1927
+ "step": 274
1928
+ },
1929
+ {
1930
+ "clip_ratio": 0.0,
1931
+ "completion_length": 282.046875,
1932
+ "epoch": 0.04906666666666667,
1933
+ "grad_norm": 73.95606383267277,
1934
+ "kl": 28.233154296875,
1935
+ "learning_rate": 1.7206079062562536e-07,
1936
+ "loss": 0.0546,
1937
+ "reward": 1.109375,
1938
+ "reward_std": 0.15625,
1939
+ "rewards/equation_reward_func": 0.109375,
1940
+ "rewards/format_reward_func": 1.0,
1941
+ "step": 276
1942
+ },
1943
+ {
1944
+ "clip_ratio": 0.0,
1945
+ "completion_length": 325.859375,
1946
+ "epoch": 0.04942222222222222,
1947
+ "grad_norm": 1.7788841527513326,
1948
+ "kl": 29.791015625,
1949
+ "learning_rate": 1.6864581199841226e-07,
1950
+ "loss": 0.0807,
1951
+ "reward": 1.09375,
1952
+ "reward_std": 0.19716878235340118,
1953
+ "rewards/equation_reward_func": 0.109375,
1954
+ "rewards/format_reward_func": 0.984375,
1955
+ "step": 278
1956
+ },
1957
+ {
1958
+ "clip_ratio": 0.0,
1959
+ "completion_length": 298.21875,
1960
+ "epoch": 0.049777777777777775,
1961
+ "grad_norm": 1.5764903296991766,
1962
+ "kl": 3.01513671875,
1963
+ "learning_rate": 1.6524772840857388e-07,
1964
+ "loss": 0.0145,
1965
+ "reward": 1.140625,
1966
+ "reward_std": 0.15625,
1967
+ "rewards/equation_reward_func": 0.140625,
1968
+ "rewards/format_reward_func": 1.0,
1969
+ "step": 280
1970
+ },
1971
+ {
1972
+ "clip_ratio": 0.0,
1973
+ "completion_length": 309.125,
1974
+ "epoch": 0.050133333333333335,
1975
+ "grad_norm": 7.596181720979034,
1976
+ "kl": 4.2314453125,
1977
+ "learning_rate": 1.6186724554503237e-07,
1978
+ "loss": -0.0037,
1979
+ "reward": 1.140625,
1980
+ "reward_std": 0.20200317353010178,
1981
+ "rewards/equation_reward_func": 0.140625,
1982
+ "rewards/format_reward_func": 1.0,
1983
+ "step": 282
1984
+ },
1985
+ {
1986
+ "clip_ratio": 0.0,
1987
+ "completion_length": 290.640625,
1988
+ "epoch": 0.05048888888888889,
1989
+ "grad_norm": 2.848445546750997,
1990
+ "kl": 3.4404296875,
1991
+ "learning_rate": 1.5850506544152103e-07,
1992
+ "loss": 0.0475,
1993
+ "reward": 1.125,
1994
+ "reward_std": 0.19716878235340118,
1995
+ "rewards/equation_reward_func": 0.125,
1996
+ "rewards/format_reward_func": 1.0,
1997
+ "step": 284
1998
+ },
1999
+ {
2000
+ "clip_ratio": 0.0,
2001
+ "completion_length": 294.75,
2002
+ "epoch": 0.05084444444444444,
2003
+ "grad_norm": 1.415636309836905,
2004
+ "kl": 1.5673828125,
2005
+ "learning_rate": 1.5516188633079107e-07,
2006
+ "loss": 0.0296,
2007
+ "reward": 1.046875,
2008
+ "reward_std": 0.06733439117670059,
2009
+ "rewards/equation_reward_func": 0.046875,
2010
+ "rewards/format_reward_func": 1.0,
2011
+ "step": 286
2012
+ },
2013
+ {
2014
+ "clip_ratio": 0.0,
2015
+ "completion_length": 335.703125,
2016
+ "epoch": 0.0512,
2017
+ "grad_norm": 1.949097702749259,
2018
+ "kl": 2.3369140625,
2019
+ "learning_rate": 1.5183840249960784e-07,
2020
+ "loss": 0.0612,
2021
+ "reward": 1.09375,
2022
+ "reward_std": 0.19716878235340118,
2023
+ "rewards/equation_reward_func": 0.109375,
2024
+ "rewards/format_reward_func": 0.984375,
2025
+ "step": 288
2026
+ },
2027
+ {
2028
+ "clip_ratio": 0.0,
2029
+ "completion_length": 280.515625,
2030
+ "epoch": 0.051555555555555556,
2031
+ "grad_norm": 2.8460028252876772,
2032
+ "kl": 1.85888671875,
2033
+ "learning_rate": 1.4853530414456612e-07,
2034
+ "loss": 0.0418,
2035
+ "reward": 1.125,
2036
+ "reward_std": 0.2235843911767006,
2037
+ "rewards/equation_reward_func": 0.125,
2038
+ "rewards/format_reward_func": 1.0,
2039
+ "step": 290
2040
+ },
2041
+ {
2042
+ "clip_ratio": 0.0,
2043
+ "completion_length": 271.75,
2044
+ "epoch": 0.05191111111111111,
2045
+ "grad_norm": 1.6592349286689605,
2046
+ "kl": 2.0166015625,
2047
+ "learning_rate": 1.4525327722875568e-07,
2048
+ "loss": 0.0283,
2049
+ "reward": 1.09375,
2050
+ "reward_std": 0.13466878235340118,
2051
+ "rewards/equation_reward_func": 0.09375,
2052
+ "rewards/format_reward_func": 1.0,
2053
+ "step": 292
2054
+ },
2055
+ {
2056
+ "clip_ratio": 0.0,
2057
+ "completion_length": 254.75,
2058
+ "epoch": 0.05226666666666667,
2059
+ "grad_norm": 2.3645749075795472,
2060
+ "kl": 1.789794921875,
2061
+ "learning_rate": 1.4199300333930515e-07,
2062
+ "loss": 0.0835,
2063
+ "reward": 1.1875,
2064
+ "reward_std": 0.3221687823534012,
2065
+ "rewards/equation_reward_func": 0.203125,
2066
+ "rewards/format_reward_func": 0.984375,
2067
+ "step": 294
2068
+ },
2069
+ {
2070
+ "clip_ratio": 0.0,
2071
+ "completion_length": 319.671875,
2072
+ "epoch": 0.05262222222222222,
2073
+ "grad_norm": 4.343843945416777,
2074
+ "kl": 2.50732421875,
2075
+ "learning_rate": 1.3875515954583523e-07,
2076
+ "loss": 0.0624,
2077
+ "reward": 1.109375,
2078
+ "reward_std": 0.2645031735301018,
2079
+ "rewards/equation_reward_func": 0.140625,
2080
+ "rewards/format_reward_func": 0.96875,
2081
+ "step": 296
2082
+ },
2083
+ {
2084
+ "clip_ratio": 0.0,
2085
+ "completion_length": 306.578125,
2086
+ "epoch": 0.052977777777777776,
2087
+ "grad_norm": 2.377005295350322,
2088
+ "kl": 1.9296875,
2089
+ "learning_rate": 1.3554041825985e-07,
2090
+ "loss": 0.0352,
2091
+ "reward": 1.09375,
2092
+ "reward_std": 0.2992308586835861,
2093
+ "rewards/equation_reward_func": 0.140625,
2094
+ "rewards/format_reward_func": 0.953125,
2095
+ "step": 298
2096
+ },
2097
+ {
2098
+ "clip_ratio": 0.0,
2099
+ "completion_length": 321.609375,
2100
+ "epoch": 0.05333333333333334,
2101
+ "grad_norm": 2.611273886267531,
2102
+ "kl": 2.335693359375,
2103
+ "learning_rate": 1.323494470950949e-07,
2104
+ "loss": 0.0688,
2105
+ "reward": 1.0625,
2106
+ "reward_std": 0.2235843911767006,
2107
+ "rewards/equation_reward_func": 0.09375,
2108
+ "rewards/format_reward_func": 0.96875,
2109
+ "step": 300
2110
+ },
2111
+ {
2112
+ "clip_ratio": 0.0,
2113
+ "completion_length": 298.859375,
2114
+ "epoch": 0.05368888888888889,
2115
+ "grad_norm": 2.213608739672042,
2116
+ "kl": 2.317138671875,
2117
+ "learning_rate": 1.2918290872891236e-07,
2118
+ "loss": 0.045,
2119
+ "reward": 1.046875,
2120
+ "reward_std": 0.15625,
2121
+ "rewards/equation_reward_func": 0.0625,
2122
+ "rewards/format_reward_func": 0.984375,
2123
+ "step": 302
2124
+ },
2125
+ {
2126
+ "clip_ratio": 0.0,
2127
+ "completion_length": 319.390625,
2128
+ "epoch": 0.054044444444444444,
2129
+ "grad_norm": 1.1361079950653166,
2130
+ "kl": 2.716796875,
2131
+ "learning_rate": 1.260414607646213e-07,
2132
+ "loss": 0.0547,
2133
+ "reward": 1.015625,
2134
+ "reward_std": 0.09375,
2135
+ "rewards/equation_reward_func": 0.03125,
2136
+ "rewards/format_reward_func": 0.984375,
2137
+ "step": 304
2138
+ },
2139
+ {
2140
+ "clip_ratio": 0.0,
2141
+ "completion_length": 315.734375,
2142
+ "epoch": 0.0544,
2143
+ "grad_norm": 2.694956280462497,
2144
+ "kl": 3.01123046875,
2145
+ "learning_rate": 1.2292575559495143e-07,
2146
+ "loss": 0.0712,
2147
+ "reward": 1.078125,
2148
+ "reward_std": 0.21694982051849365,
2149
+ "rewards/equation_reward_func": 0.109375,
2150
+ "rewards/format_reward_func": 0.96875,
2151
+ "step": 306
2152
+ },
2153
+ {
2154
+ "clip_ratio": 0.0,
2155
+ "completion_length": 314.15625,
2156
+ "epoch": 0.05475555555555556,
2157
+ "grad_norm": 1.6293408130888805,
2158
+ "kl": 2.3837890625,
2159
+ "learning_rate": 1.1983644026655835e-07,
2160
+ "loss": 0.0147,
2161
+ "reward": 1.125,
2162
+ "reward_std": 0.13466878235340118,
2163
+ "rewards/equation_reward_func": 0.125,
2164
+ "rewards/format_reward_func": 1.0,
2165
+ "step": 308
2166
+ },
2167
+ {
2168
+ "clip_ratio": 0.0,
2169
+ "completion_length": 273.125,
2170
+ "epoch": 0.05511111111111111,
2171
+ "grad_norm": 2.1302515062721126,
2172
+ "kl": 2.05712890625,
2173
+ "learning_rate": 1.1677415634565066e-07,
2174
+ "loss": 0.0406,
2175
+ "reward": 1.15625,
2176
+ "reward_std": 0.19716878235340118,
2177
+ "rewards/equation_reward_func": 0.15625,
2178
+ "rewards/format_reward_func": 1.0,
2179
+ "step": 310
2180
+ },
2181
+ {
2182
+ "clip_ratio": 0.0,
2183
+ "completion_length": 337.59375,
2184
+ "epoch": 0.055466666666666664,
2185
+ "grad_norm": 15.014740699660255,
2186
+ "kl": 7.84521484375,
2187
+ "learning_rate": 1.1373953978475353e-07,
2188
+ "loss": 0.0368,
2189
+ "reward": 1.0625,
2190
+ "reward_std": 0.125,
2191
+ "rewards/equation_reward_func": 0.0625,
2192
+ "rewards/format_reward_func": 1.0,
2193
+ "step": 312
2194
+ },
2195
+ {
2196
+ "clip_ratio": 0.0,
2197
+ "completion_length": 312.1875,
2198
+ "epoch": 0.055822222222222224,
2199
+ "grad_norm": 2.0788629281112727,
2200
+ "kl": 4.335693359375,
2201
+ "learning_rate": 1.1073322079063913e-07,
2202
+ "loss": 0.0683,
2203
+ "reward": 1.15625,
2204
+ "reward_std": 0.2235843911767006,
2205
+ "rewards/equation_reward_func": 0.15625,
2206
+ "rewards/format_reward_func": 1.0,
2207
+ "step": 314
2208
+ },
2209
+ {
2210
+ "clip_ratio": 0.0,
2211
+ "completion_length": 276.34375,
2212
+ "epoch": 0.05617777777777778,
2213
+ "grad_norm": 11.038251821932738,
2214
+ "kl": 4.507568359375,
2215
+ "learning_rate": 1.0775582369344946e-07,
2216
+ "loss": 0.0259,
2217
+ "reward": 1.09375,
2218
+ "reward_std": 0.1610843911767006,
2219
+ "rewards/equation_reward_func": 0.09375,
2220
+ "rewards/format_reward_func": 1.0,
2221
+ "step": 316
2222
+ },
2223
+ {
2224
+ "clip_ratio": 0.0,
2225
+ "completion_length": 285.59375,
2226
+ "epoch": 0.05653333333333333,
2227
+ "grad_norm": 4.009960857728169,
2228
+ "kl": 2.4130859375,
2229
+ "learning_rate": 1.0480796681704077e-07,
2230
+ "loss": -0.0194,
2231
+ "reward": 1.125,
2232
+ "reward_std": 0.125,
2233
+ "rewards/equation_reward_func": 0.125,
2234
+ "rewards/format_reward_func": 1.0,
2235
+ "step": 318
2236
+ },
2237
+ {
2238
+ "clip_ratio": 0.0,
2239
+ "completion_length": 277.296875,
2240
+ "epoch": 0.05688888888888889,
2241
+ "grad_norm": 4.499177733550329,
2242
+ "kl": 5.369873046875,
2243
+ "learning_rate": 1.018902623505741e-07,
2244
+ "loss": 0.0504,
2245
+ "reward": 1.078125,
2246
+ "reward_std": 0.1298343911767006,
2247
+ "rewards/equation_reward_func": 0.078125,
2248
+ "rewards/format_reward_func": 1.0,
2249
+ "step": 320
2250
+ },
2251
+ {
2252
+ "clip_ratio": 0.0,
2253
+ "completion_length": 316.578125,
2254
+ "epoch": 0.057244444444444445,
2255
+ "grad_norm": 3.3047156898766272,
2256
+ "kl": 3.43701171875,
2257
+ "learning_rate": 9.900331622138063e-08,
2258
+ "loss": 0.0595,
2259
+ "reward": 1.078125,
2260
+ "reward_std": 0.16591878235340118,
2261
+ "rewards/equation_reward_func": 0.09375,
2262
+ "rewards/format_reward_func": 0.984375,
2263
+ "step": 322
2264
+ },
2265
+ {
2266
+ "clip_ratio": 0.0,
2267
+ "completion_length": 297.765625,
2268
+ "epoch": 0.0576,
2269
+ "grad_norm": 4.938077502561252,
2270
+ "kl": 2.2734375,
2271
+ "learning_rate": 9.614772796912681e-08,
2272
+ "loss": 0.02,
2273
+ "reward": 1.15625,
2274
+ "reward_std": 0.1610843911767006,
2275
+ "rewards/equation_reward_func": 0.15625,
2276
+ "rewards/format_reward_func": 1.0,
2277
+ "step": 324
2278
+ },
2279
+ {
2280
+ "clip_ratio": 0.0,
2281
+ "completion_length": 337.640625,
2282
+ "epoch": 0.05795555555555556,
2283
+ "grad_norm": 3.2448715644280974,
2284
+ "kl": 2.689453125,
2285
+ "learning_rate": 9.332409062130686e-08,
2286
+ "loss": 0.0284,
2287
+ "reward": 1.0625,
2288
+ "reward_std": 0.1610843911767006,
2289
+ "rewards/equation_reward_func": 0.078125,
2290
+ "rewards/format_reward_func": 0.984375,
2291
+ "step": 326
2292
+ },
2293
+ {
2294
+ "clip_ratio": 0.0,
2295
+ "completion_length": 297.21875,
2296
+ "epoch": 0.05831111111111111,
2297
+ "grad_norm": 3.6089469790675452,
2298
+ "kl": 1.907470703125,
2299
+ "learning_rate": 9.053299057008699e-08,
2300
+ "loss": 0.0125,
2301
+ "reward": 0.96875,
2302
+ "reward_std": 0.125,
2303
+ "rewards/equation_reward_func": 0.015625,
2304
+ "rewards/format_reward_func": 0.953125,
2305
+ "step": 328
2306
+ },
2307
+ {
2308
+ "clip_ratio": 0.0,
2309
+ "completion_length": 298.953125,
2310
+ "epoch": 0.058666666666666666,
2311
+ "grad_norm": 3.9902002814055497,
2312
+ "kl": 2.693603515625,
2313
+ "learning_rate": 8.777500745052743e-08,
2314
+ "loss": 0.054,
2315
+ "reward": 1.109375,
2316
+ "reward_std": 0.1923343911767006,
2317
+ "rewards/equation_reward_func": 0.125,
2318
+ "rewards/format_reward_func": 0.984375,
2319
+ "step": 330
2320
+ },
2321
+ {
2322
+ "clip_ratio": 0.0,
2323
+ "completion_length": 310.515625,
2324
+ "epoch": 0.05902222222222222,
2325
+ "grad_norm": 2.3258942923733033,
2326
+ "kl": 1.953857421875,
2327
+ "learning_rate": 8.505071402020892e-08,
2328
+ "loss": 0.0404,
2329
+ "reward": 1.078125,
2330
+ "reward_std": 0.22841878235340118,
2331
+ "rewards/equation_reward_func": 0.109375,
2332
+ "rewards/format_reward_func": 0.96875,
2333
+ "step": 332
2334
+ },
2335
+ {
2336
+ "clip_ratio": 0.0,
2337
+ "completion_length": 306.328125,
2338
+ "epoch": 0.05937777777777778,
2339
+ "grad_norm": 2.105282555937082,
2340
+ "kl": 2.130615234375,
2341
+ "learning_rate": 8.236067604028562e-08,
2342
+ "loss": 0.0711,
2343
+ "reward": 1.0625,
2344
+ "reward_std": 0.1610843911767006,
2345
+ "rewards/equation_reward_func": 0.078125,
2346
+ "rewards/format_reward_func": 0.984375,
2347
+ "step": 334
2348
+ },
2349
+ {
2350
+ "clip_ratio": 0.0,
2351
+ "completion_length": 333.03125,
2352
+ "epoch": 0.05973333333333333,
2353
+ "grad_norm": 1.6921617524371768,
2354
+ "kl": 2.76611328125,
2355
+ "learning_rate": 7.970545215799327e-08,
2356
+ "loss": -0.0171,
2357
+ "reward": 1.09375,
2358
+ "reward_std": 0.19716878235340118,
2359
+ "rewards/equation_reward_func": 0.109375,
2360
+ "rewards/format_reward_func": 0.984375,
2361
+ "step": 336
2362
+ },
2363
+ {
2364
+ "clip_ratio": 0.0,
2365
+ "completion_length": 295.71875,
2366
+ "epoch": 0.060088888888888886,
2367
+ "grad_norm": 2.2366407518555658,
2368
+ "kl": 2.571533203125,
2369
+ "learning_rate": 7.708559379063204e-08,
2370
+ "loss": 0.0171,
2371
+ "reward": 1.078125,
2372
+ "reward_std": 0.15625,
2373
+ "rewards/equation_reward_func": 0.09375,
2374
+ "rewards/format_reward_func": 0.984375,
2375
+ "step": 338
2376
+ },
2377
+ {
2378
+ "clip_ratio": 0.0,
2379
+ "completion_length": 296.125,
2380
+ "epoch": 0.060444444444444446,
2381
+ "grad_norm": 1.7453930694713657,
2382
+ "kl": 1.99072265625,
2383
+ "learning_rate": 7.45016450110534e-08,
2384
+ "loss": 0.086,
2385
+ "reward": 1.078125,
2386
+ "reward_std": 0.2548343911767006,
2387
+ "rewards/equation_reward_func": 0.125,
2388
+ "rewards/format_reward_func": 0.953125,
2389
+ "step": 340
2390
+ },
2391
+ {
2392
+ "clip_ratio": 0.0,
2393
+ "completion_length": 287.390625,
2394
+ "epoch": 0.0608,
2395
+ "grad_norm": 2.7894086215156477,
2396
+ "kl": 3.01318359375,
2397
+ "learning_rate": 7.195414243467029e-08,
2398
+ "loss": 0.0482,
2399
+ "reward": 1.171875,
2400
+ "reward_std": 0.22841878235340118,
2401
+ "rewards/equation_reward_func": 0.171875,
2402
+ "rewards/format_reward_func": 1.0,
2403
+ "step": 342
2404
+ },
2405
+ {
2406
+ "clip_ratio": 0.0,
2407
+ "completion_length": 317.40625,
2408
+ "epoch": 0.06115555555555555,
2409
+ "grad_norm": 0.931178773836098,
2410
+ "kl": 1.2216796875,
2411
+ "learning_rate": 6.944361510801763e-08,
2412
+ "loss": -0.0173,
2413
+ "reward": 1.015625,
2414
+ "reward_std": 0.09375,
2415
+ "rewards/equation_reward_func": 0.03125,
2416
+ "rewards/format_reward_func": 0.984375,
2417
+ "step": 344
2418
+ },
2419
+ {
2420
+ "clip_ratio": 0.0,
2421
+ "completion_length": 299.359375,
2422
+ "epoch": 0.061511111111111114,
2423
+ "grad_norm": 3.753569436457584,
2424
+ "kl": 2.353759765625,
2425
+ "learning_rate": 6.697058439888283e-08,
2426
+ "loss": 0.0152,
2427
+ "reward": 1.078125,
2428
+ "reward_std": 0.10341878235340118,
2429
+ "rewards/equation_reward_func": 0.078125,
2430
+ "rewards/format_reward_func": 1.0,
2431
+ "step": 346
2432
+ },
2433
+ {
2434
+ "clip_ratio": 0.0,
2435
+ "completion_length": 260.09375,
2436
+ "epoch": 0.06186666666666667,
2437
+ "grad_norm": 2.9761032955845192,
2438
+ "kl": 1.854736328125,
2439
+ "learning_rate": 6.453556388803288e-08,
2440
+ "loss": 0.025,
2441
+ "reward": 1.078125,
2442
+ "reward_std": 0.1923343911767006,
2443
+ "rewards/equation_reward_func": 0.09375,
2444
+ "rewards/format_reward_func": 0.984375,
2445
+ "step": 348
2446
+ },
2447
+ {
2448
+ "clip_ratio": 0.0,
2449
+ "completion_length": 306.0625,
2450
+ "epoch": 0.06222222222222222,
2451
+ "grad_norm": 2.583072279031626,
2452
+ "kl": 1.90185546875,
2453
+ "learning_rate": 6.213905926255697e-08,
2454
+ "loss": -0.0007,
2455
+ "reward": 1.09375,
2456
+ "reward_std": 0.1610843911767006,
2457
+ "rewards/equation_reward_func": 0.09375,
2458
+ "rewards/format_reward_func": 1.0,
2459
+ "step": 350
2460
+ },
2461
+ {
2462
+ "clip_ratio": 0.0,
2463
+ "completion_length": 284.453125,
2464
+ "epoch": 0.06257777777777777,
2465
+ "grad_norm": 2.5644083370227495,
2466
+ "kl": 10.295654296875,
2467
+ "learning_rate": 5.978156821084987e-08,
2468
+ "loss": 0.011,
2469
+ "reward": 1.0625,
2470
+ "reward_std": 0.1610843911767006,
2471
+ "rewards/equation_reward_func": 0.078125,
2472
+ "rewards/format_reward_func": 0.984375,
2473
+ "step": 352
2474
+ },
2475
+ {
2476
+ "clip_ratio": 0.0,
2477
+ "completion_length": 304.84375,
2478
+ "epoch": 0.06293333333333333,
2479
+ "grad_norm": 3.260671051240977,
2480
+ "kl": 7.114013671875,
2481
+ "learning_rate": 5.7463580319254853e-08,
2482
+ "loss": 0.078,
2483
+ "reward": 1.09375,
2484
+ "reward_std": 0.2235843911767006,
2485
+ "rewards/equation_reward_func": 0.109375,
2486
+ "rewards/format_reward_func": 0.984375,
2487
+ "step": 354
2488
+ },
2489
+ {
2490
+ "clip_ratio": 0.0,
2491
+ "completion_length": 281.296875,
2492
+ "epoch": 0.0632888888888889,
2493
+ "grad_norm": 2.0339670428860477,
2494
+ "kl": 2.173095703125,
2495
+ "learning_rate": 5.518557697039081e-08,
2496
+ "loss": 0.0621,
2497
+ "reward": 1.140625,
2498
+ "reward_std": 0.20200317353010178,
2499
+ "rewards/equation_reward_func": 0.171875,
2500
+ "rewards/format_reward_func": 0.96875,
2501
+ "step": 356
2502
+ },
2503
+ {
2504
+ "clip_ratio": 0.0,
2505
+ "completion_length": 294.5625,
2506
+ "epoch": 0.06364444444444445,
2507
+ "grad_norm": 3.3851967041509154,
2508
+ "kl": 2.346923828125,
2509
+ "learning_rate": 5.294803124318145e-08,
2510
+ "loss": -0.008,
2511
+ "reward": 1.046875,
2512
+ "reward_std": 0.06733439117670059,
2513
+ "rewards/equation_reward_func": 0.046875,
2514
+ "rewards/format_reward_func": 1.0,
2515
+ "step": 358
2516
+ },
2517
+ {
2518
+ "clip_ratio": 0.0,
2519
+ "completion_length": 319.390625,
2520
+ "epoch": 0.064,
2521
+ "grad_norm": 1.6809943974752608,
2522
+ "kl": 2.185791015625,
2523
+ "learning_rate": 5.07514078146106e-08,
2524
+ "loss": 0.0332,
2525
+ "reward": 1.109375,
2526
+ "reward_std": 0.19053421169519424,
2527
+ "rewards/equation_reward_func": 0.125,
2528
+ "rewards/format_reward_func": 0.984375,
2529
+ "step": 360
2530
+ },
2531
+ {
2532
+ "clip_ratio": 0.0,
2533
+ "completion_length": 261.21875,
2534
+ "epoch": 0.06435555555555555,
2535
+ "grad_norm": 1.105930554758287,
2536
+ "kl": 1.884521484375,
2537
+ "learning_rate": 4.859616286322094e-08,
2538
+ "loss": 0.0335,
2539
+ "reward": 1.0625,
2540
+ "reward_std": 0.09858439117670059,
2541
+ "rewards/equation_reward_func": 0.0625,
2542
+ "rewards/format_reward_func": 1.0,
2543
+ "step": 362
2544
+ },
2545
+ {
2546
+ "clip_ratio": 0.0,
2547
+ "completion_length": 293.71875,
2548
+ "epoch": 0.06471111111111111,
2549
+ "grad_norm": 2.075449397863208,
2550
+ "kl": 2.149169921875,
2551
+ "learning_rate": 4.648274397437829e-08,
2552
+ "loss": -0.0078,
2553
+ "reward": 1.109375,
2554
+ "reward_std": 0.15625,
2555
+ "rewards/equation_reward_func": 0.109375,
2556
+ "rewards/format_reward_func": 1.0,
2557
+ "step": 364
2558
+ },
2559
+ {
2560
+ "clip_ratio": 0.0,
2561
+ "completion_length": 299.0625,
2562
+ "epoch": 0.06506666666666666,
2563
+ "grad_norm": 2.440666439161297,
2564
+ "kl": 2.136474609375,
2565
+ "learning_rate": 4.4411590047320617e-08,
2566
+ "loss": 0.0491,
2567
+ "reward": 1.140625,
2568
+ "reward_std": 0.20200317353010178,
2569
+ "rewards/equation_reward_func": 0.140625,
2570
+ "rewards/format_reward_func": 1.0,
2571
+ "step": 366
2572
+ },
2573
+ {
2574
+ "clip_ratio": 0.0,
2575
+ "completion_length": 320.890625,
2576
+ "epoch": 0.06542222222222223,
2577
+ "grad_norm": 4.619211137682817,
2578
+ "kl": 1.549072265625,
2579
+ "learning_rate": 4.2383131204010494e-08,
2580
+ "loss": 0.0484,
2581
+ "reward": 1.078125,
2582
+ "reward_std": 0.26978103816509247,
2583
+ "rewards/equation_reward_func": 0.109375,
2584
+ "rewards/format_reward_func": 0.96875,
2585
+ "step": 368
2586
+ },
2587
+ {
2588
+ "clip_ratio": 0.0,
2589
+ "completion_length": 288.234375,
2590
+ "epoch": 0.06577777777777778,
2591
+ "grad_norm": 2.1383834468441396,
2592
+ "kl": 1.26123046875,
2593
+ "learning_rate": 4.039778869981064e-08,
2594
+ "loss": 0.017,
2595
+ "reward": 1.078125,
2596
+ "reward_std": 0.1298343911767006,
2597
+ "rewards/equation_reward_func": 0.078125,
2598
+ "rewards/format_reward_func": 1.0,
2599
+ "step": 370
2600
+ },
2601
+ {
2602
+ "clip_ratio": 0.0,
2603
+ "completion_length": 316.71875,
2604
+ "epoch": 0.06613333333333334,
2605
+ "grad_norm": 2.9751121792977497,
2606
+ "kl": 1.9384765625,
2607
+ "learning_rate": 3.845597483600049e-08,
2608
+ "loss": 0.0462,
2609
+ "reward": 1.0625,
2610
+ "reward_std": 0.125,
2611
+ "rewards/equation_reward_func": 0.0625,
2612
+ "rewards/format_reward_func": 1.0,
2613
+ "step": 372
2614
+ },
2615
+ {
2616
+ "clip_ratio": 0.0,
2617
+ "completion_length": 302.296875,
2618
+ "epoch": 0.06648888888888889,
2619
+ "grad_norm": 1.7788468837628437,
2620
+ "kl": 1.295654296875,
2621
+ "learning_rate": 3.655809287415284e-08,
2622
+ "loss": -0.0181,
2623
+ "reward": 1.09375,
2624
+ "reward_std": 0.2235843911767006,
2625
+ "rewards/equation_reward_func": 0.109375,
2626
+ "rewards/format_reward_func": 0.984375,
2627
+ "step": 374
2628
+ },
2629
+ {
2630
+ "clip_ratio": 0.0,
2631
+ "completion_length": 282.734375,
2632
+ "epoch": 0.06684444444444444,
2633
+ "grad_norm": 1.7936851877118818,
2634
+ "kl": 1.361083984375,
2635
+ "learning_rate": 3.4704536952387285e-08,
2636
+ "loss": 0.0238,
2637
+ "reward": 1.078125,
2638
+ "reward_std": 0.16591878235340118,
2639
+ "rewards/equation_reward_func": 0.09375,
2640
+ "rewards/format_reward_func": 0.984375,
2641
+ "step": 376
2642
+ },
2643
+ {
2644
+ "clip_ratio": 0.0,
2645
+ "completion_length": 310.375,
2646
+ "epoch": 0.0672,
2647
+ "grad_norm": 4.685423790152177,
2648
+ "kl": 3.947998046875,
2649
+ "learning_rate": 3.2895692003518575e-08,
2650
+ "loss": 0.0234,
2651
+ "reward": 1.03125,
2652
+ "reward_std": 0.125,
2653
+ "rewards/equation_reward_func": 0.046875,
2654
+ "rewards/format_reward_func": 0.984375,
2655
+ "step": 378
2656
+ },
2657
+ {
2658
+ "clip_ratio": 0.0,
2659
+ "completion_length": 298.34375,
2660
+ "epoch": 0.06755555555555555,
2661
+ "grad_norm": 3.1205307234975543,
2662
+ "kl": 1.413330078125,
2663
+ "learning_rate": 3.113193367511635e-08,
2664
+ "loss": 0.0445,
2665
+ "reward": 1.046875,
2666
+ "reward_std": 0.21875,
2667
+ "rewards/equation_reward_func": 0.078125,
2668
+ "rewards/format_reward_func": 0.96875,
2669
+ "step": 380
2670
+ },
2671
+ {
2672
+ "clip_ratio": 0.0,
2673
+ "completion_length": 319.703125,
2674
+ "epoch": 0.06791111111111112,
2675
+ "grad_norm": 1.1720140167792736,
2676
+ "kl": 1.87060546875,
2677
+ "learning_rate": 2.9413628251493934e-08,
2678
+ "loss": 0.0189,
2679
+ "reward": 1.03125,
2680
+ "reward_std": 0.125,
2681
+ "rewards/equation_reward_func": 0.046875,
2682
+ "rewards/format_reward_func": 0.984375,
2683
+ "step": 382
2684
+ },
2685
+ {
2686
+ "clip_ratio": 0.0,
2687
+ "completion_length": 310.0,
2688
+ "epoch": 0.06826666666666667,
2689
+ "grad_norm": 2.6084931637781725,
2690
+ "kl": 3.1181640625,
2691
+ "learning_rate": 2.774113257764066e-08,
2692
+ "loss": 0.0243,
2693
+ "reward": 1.046875,
2694
+ "reward_std": 0.06733439117670059,
2695
+ "rewards/equation_reward_func": 0.046875,
2696
+ "rewards/format_reward_func": 1.0,
2697
+ "step": 384
2698
+ },
2699
+ {
2700
+ "clip_ratio": 0.0,
2701
+ "completion_length": 313.3125,
2702
+ "epoch": 0.06862222222222222,
2703
+ "grad_norm": 4.358043629813717,
2704
+ "kl": 2.083251953125,
2705
+ "learning_rate": 2.611479398511518e-08,
2706
+ "loss": 0.0497,
2707
+ "reward": 1.15625,
2708
+ "reward_std": 0.2235843911767006,
2709
+ "rewards/equation_reward_func": 0.171875,
2710
+ "rewards/format_reward_func": 0.984375,
2711
+ "step": 386
2712
+ },
2713
+ {
2714
+ "clip_ratio": 0.0,
2715
+ "completion_length": 311.5625,
2716
+ "epoch": 0.06897777777777778,
2717
+ "grad_norm": 1.7314295380277487,
2718
+ "kl": 2.2861328125,
2719
+ "learning_rate": 2.4534950219914057e-08,
2720
+ "loss": 0.0184,
2721
+ "reward": 1.03125,
2722
+ "reward_std": 0.0625,
2723
+ "rewards/equation_reward_func": 0.03125,
2724
+ "rewards/format_reward_func": 1.0,
2725
+ "step": 388
2726
+ },
2727
+ {
2728
+ "clip_ratio": 0.0,
2729
+ "completion_length": 315.078125,
2730
+ "epoch": 0.06933333333333333,
2731
+ "grad_norm": 3.085044311379691,
2732
+ "kl": 1.55078125,
2733
+ "learning_rate": 2.300192937233128e-08,
2734
+ "loss": 0.0237,
2735
+ "reward": 1.125,
2736
+ "reward_std": 0.1610843911767006,
2737
+ "rewards/equation_reward_func": 0.140625,
2738
+ "rewards/format_reward_func": 0.984375,
2739
+ "step": 390
2740
+ },
2741
+ {
2742
+ "clip_ratio": 0.0,
2743
+ "completion_length": 321.1875,
2744
+ "epoch": 0.06968888888888888,
2745
+ "grad_norm": 2.1144825124318696,
2746
+ "kl": 1.6259765625,
2747
+ "learning_rate": 2.1516049808822935e-08,
2748
+ "loss": 0.0228,
2749
+ "reward": 1.078125,
2750
+ "reward_std": 0.1923343911767006,
2751
+ "rewards/equation_reward_func": 0.09375,
2752
+ "rewards/format_reward_func": 0.984375,
2753
+ "step": 392
2754
+ },
2755
+ {
2756
+ "clip_ratio": 0.0,
2757
+ "completion_length": 296.390625,
2758
+ "epoch": 0.07004444444444445,
2759
+ "grad_norm": 2.2495222116539715,
2760
+ "kl": 1.490234375,
2761
+ "learning_rate": 2.007762010589098e-08,
2762
+ "loss": 0.0232,
2763
+ "reward": 1.078125,
2764
+ "reward_std": 0.1923343911767006,
2765
+ "rewards/equation_reward_func": 0.09375,
2766
+ "rewards/format_reward_func": 0.984375,
2767
+ "step": 394
2768
+ },
2769
+ {
2770
+ "clip_ratio": 0.0,
2771
+ "completion_length": 286.09375,
2772
+ "epoch": 0.0704,
2773
+ "grad_norm": 3.5477202571420716,
2774
+ "kl": 3.229248046875,
2775
+ "learning_rate": 1.8686938986000627e-08,
2776
+ "loss": -0.0136,
2777
+ "reward": 1.109375,
2778
+ "reward_std": 0.20200317353010178,
2779
+ "rewards/equation_reward_func": 0.125,
2780
+ "rewards/format_reward_func": 0.984375,
2781
+ "step": 396
2782
+ },
2783
+ {
2784
+ "clip_ratio": 0.0,
2785
+ "completion_length": 292.890625,
2786
+ "epoch": 0.07075555555555556,
2787
+ "grad_norm": 3.401342482083026,
2788
+ "kl": 1.608642578125,
2789
+ "learning_rate": 1.734429525554365e-08,
2790
+ "loss": -0.0173,
2791
+ "reward": 1.015625,
2792
+ "reward_std": 0.08228103816509247,
2793
+ "rewards/equation_reward_func": 0.03125,
2794
+ "rewards/format_reward_func": 0.984375,
2795
+ "step": 398
2796
+ },
2797
+ {
2798
+ "clip_ratio": 0.0,
2799
+ "completion_length": 290.140625,
2800
+ "epoch": 0.07111111111111111,
2801
+ "grad_norm": 2.4346116264171203,
2802
+ "kl": 2.428955078125,
2803
+ "learning_rate": 1.604996774486145e-08,
2804
+ "loss": 0.0619,
2805
+ "reward": 1.046875,
2806
+ "reward_std": 0.1298343911767006,
2807
+ "rewards/equation_reward_func": 0.0625,
2808
+ "rewards/format_reward_func": 0.984375,
2809
+ "step": 400
2810
+ },
2811
+ {
2812
+ "clip_ratio": 0.0,
2813
+ "completion_length": 327.09375,
2814
+ "epoch": 0.07146666666666666,
2815
+ "grad_norm": 1.4096767097340954,
2816
+ "kl": 2.732421875,
2817
+ "learning_rate": 1.4804225250339281e-08,
2818
+ "loss": 0.0063,
2819
+ "reward": 1.09375,
2820
+ "reward_std": 0.09858439117670059,
2821
+ "rewards/equation_reward_func": 0.09375,
2822
+ "rewards/format_reward_func": 1.0,
2823
+ "step": 402
2824
+ },
2825
+ {
2826
+ "clip_ratio": 0.0,
2827
+ "completion_length": 308.53125,
2828
+ "epoch": 0.07182222222222222,
2829
+ "grad_norm": 11.032142152384289,
2830
+ "kl": 3.81396484375,
2831
+ "learning_rate": 1.360732647858498e-08,
2832
+ "loss": 0.0763,
2833
+ "reward": 1.203125,
2834
+ "reward_std": 0.2645031735301018,
2835
+ "rewards/equation_reward_func": 0.203125,
2836
+ "rewards/format_reward_func": 1.0,
2837
+ "step": 404
2838
+ },
2839
+ {
2840
+ "clip_ratio": 0.0,
2841
+ "completion_length": 308.3125,
2842
+ "epoch": 0.07217777777777777,
2843
+ "grad_norm": 2.3580842283452412,
2844
+ "kl": 2.40478515625,
2845
+ "learning_rate": 1.2459519992702311e-08,
2846
+ "loss": 0.0281,
2847
+ "reward": 1.015625,
2848
+ "reward_std": 0.09375,
2849
+ "rewards/equation_reward_func": 0.03125,
2850
+ "rewards/format_reward_func": 0.984375,
2851
+ "step": 406
2852
+ },
2853
+ {
2854
+ "clip_ratio": 0.0,
2855
+ "completion_length": 276.265625,
2856
+ "epoch": 0.07253333333333334,
2857
+ "grad_norm": 1.807591388336414,
2858
+ "kl": 2.02685546875,
2859
+ "learning_rate": 1.1361044160671629e-08,
2860
+ "loss": 0.0073,
2861
+ "reward": 1.15625,
2862
+ "reward_std": 0.19716878235340118,
2863
+ "rewards/equation_reward_func": 0.15625,
2864
+ "rewards/format_reward_func": 1.0,
2865
+ "step": 408
2866
+ },
2867
+ {
2868
+ "clip_ratio": 0.0,
2869
+ "completion_length": 305.59375,
2870
+ "epoch": 0.07288888888888889,
2871
+ "grad_norm": 397.3498098291439,
2872
+ "kl": 50.008544921875,
2873
+ "learning_rate": 1.0312127105846947e-08,
2874
+ "loss": 0.1459,
2875
+ "reward": 1.03125,
2876
+ "reward_std": 0.23853103816509247,
2877
+ "rewards/equation_reward_func": 0.078125,
2878
+ "rewards/format_reward_func": 0.953125,
2879
+ "step": 410
2880
+ },
2881
+ {
2882
+ "clip_ratio": 0.0,
2883
+ "completion_length": 307.015625,
2884
+ "epoch": 0.07324444444444445,
2885
+ "grad_norm": 2.197104654614589,
2886
+ "kl": 2.2783203125,
2887
+ "learning_rate": 9.312986659581301e-09,
2888
+ "loss": 0.0373,
2889
+ "reward": 1.09375,
2890
+ "reward_std": 0.1610843911767006,
2891
+ "rewards/equation_reward_func": 0.09375,
2892
+ "rewards/format_reward_func": 1.0,
2893
+ "step": 412
2894
+ },
2895
+ {
2896
+ "clip_ratio": 0.0,
2897
+ "completion_length": 307.390625,
2898
+ "epoch": 0.0736,
2899
+ "grad_norm": 2.2032918507669015,
2900
+ "kl": 1.651611328125,
2901
+ "learning_rate": 8.363830315988945e-09,
2902
+ "loss": 0.0537,
2903
+ "reward": 1.078125,
2904
+ "reward_std": 0.1298343911767006,
2905
+ "rewards/equation_reward_func": 0.09375,
2906
+ "rewards/format_reward_func": 0.984375,
2907
+ "step": 414
2908
+ },
2909
+ {
2910
+ "clip_ratio": 0.0,
2911
+ "completion_length": 296.71875,
2912
+ "epoch": 0.07395555555555555,
2913
+ "grad_norm": 2.6002311285352175,
2914
+ "kl": 2.895751953125,
2915
+ "learning_rate": 7.46485518885462e-09,
2916
+ "loss": 0.0162,
2917
+ "reward": 1.09375,
2918
+ "reward_std": 0.125,
2919
+ "rewards/equation_reward_func": 0.09375,
2920
+ "rewards/format_reward_func": 1.0,
2921
+ "step": 416
2922
+ },
2923
+ {
2924
+ "clip_ratio": 0.0,
2925
+ "completion_length": 314.734375,
2926
+ "epoch": 0.0743111111111111,
2927
+ "grad_norm": 1.4366325174606742,
2928
+ "kl": 1.68115234375,
2929
+ "learning_rate": 6.616247970698319e-09,
2930
+ "loss": 0.0527,
2931
+ "reward": 1.0625,
2932
+ "reward_std": 0.22575797885656357,
2933
+ "rewards/equation_reward_func": 0.109375,
2934
+ "rewards/format_reward_func": 0.953125,
2935
+ "step": 418
2936
+ },
2937
+ {
2938
+ "clip_ratio": 0.0,
2939
+ "completion_length": 274.125,
2940
+ "epoch": 0.07466666666666667,
2941
+ "grad_norm": 2.832174519961065,
2942
+ "kl": 1.653076171875,
2943
+ "learning_rate": 5.8181848940044855e-09,
2944
+ "loss": 0.0479,
2945
+ "reward": 1.15625,
2946
+ "reward_std": 0.25,
2947
+ "rewards/equation_reward_func": 0.15625,
2948
+ "rewards/format_reward_func": 1.0,
2949
+ "step": 420
2950
+ },
2951
+ {
2952
+ "clip_ratio": 0.0,
2953
+ "completion_length": 293.578125,
2954
+ "epoch": 0.07502222222222223,
2955
+ "grad_norm": 3.2917629073928247,
2956
+ "kl": 1.63916015625,
2957
+ "learning_rate": 5.070831694623135e-09,
2958
+ "loss": 0.0314,
2959
+ "reward": 1.1875,
2960
+ "reward_std": 0.3221687823534012,
2961
+ "rewards/equation_reward_func": 0.1875,
2962
+ "rewards/format_reward_func": 1.0,
2963
+ "step": 422
2964
+ },
2965
+ {
2966
+ "clip_ratio": 0.0,
2967
+ "completion_length": 268.71875,
2968
+ "epoch": 0.07537777777777778,
2969
+ "grad_norm": 1.3697523023561404,
2970
+ "kl": 2.22021484375,
2971
+ "learning_rate": 4.374343577351336e-09,
2972
+ "loss": 0.0136,
2973
+ "reward": 1.046875,
2974
+ "reward_std": 0.1298343911767006,
2975
+ "rewards/equation_reward_func": 0.0625,
2976
+ "rewards/format_reward_func": 0.984375,
2977
+ "step": 424
2978
+ },
2979
+ {
2980
+ "clip_ratio": 0.0,
2981
+ "completion_length": 312.59375,
2982
+ "epoch": 0.07573333333333333,
2983
+ "grad_norm": 2.896324603562049,
2984
+ "kl": 1.552734375,
2985
+ "learning_rate": 3.7288651837012745e-09,
2986
+ "loss": 0.0834,
2987
+ "reward": 1.140625,
2988
+ "reward_std": 0.2909187823534012,
2989
+ "rewards/equation_reward_func": 0.15625,
2990
+ "rewards/format_reward_func": 0.984375,
2991
+ "step": 426
2992
+ },
2993
+ {
2994
+ "clip_ratio": 0.0,
2995
+ "completion_length": 279.921875,
2996
+ "epoch": 0.07608888888888889,
2997
+ "grad_norm": 2.2563282918149166,
2998
+ "kl": 2.084228515625,
2999
+ "learning_rate": 3.134530561862081e-09,
3000
+ "loss": 0.0184,
3001
+ "reward": 1.109375,
3002
+ "reward_std": 0.10341878235340118,
3003
+ "rewards/equation_reward_func": 0.125,
3004
+ "rewards/format_reward_func": 0.984375,
3005
+ "step": 428
3006
+ },
3007
+ {
3008
+ "clip_ratio": 0.0,
3009
+ "completion_length": 299.8125,
3010
+ "epoch": 0.07644444444444444,
3011
+ "grad_norm": 2.805702161500462,
3012
+ "kl": 1.2802734375,
3013
+ "learning_rate": 2.5914631388619103e-09,
3014
+ "loss": 0.022,
3015
+ "reward": 1.0625,
3016
+ "reward_std": 0.0625,
3017
+ "rewards/equation_reward_func": 0.0625,
3018
+ "rewards/format_reward_func": 1.0,
3019
+ "step": 430
3020
+ },
3021
+ {
3022
+ "clip_ratio": 0.0,
3023
+ "completion_length": 280.1875,
3024
+ "epoch": 0.0768,
3025
+ "grad_norm": 2.3897542479107328,
3026
+ "kl": 1.429931640625,
3027
+ "learning_rate": 2.0997756949353297e-09,
3028
+ "loss": 0.0494,
3029
+ "reward": 1.109375,
3030
+ "reward_std": 0.24336542934179306,
3031
+ "rewards/equation_reward_func": 0.140625,
3032
+ "rewards/format_reward_func": 0.96875,
3033
+ "step": 432
3034
+ },
3035
+ {
3036
+ "clip_ratio": 0.0,
3037
+ "completion_length": 309.828125,
3038
+ "epoch": 0.07715555555555556,
3039
+ "grad_norm": 3.547940603348454,
3040
+ "kl": 1.690185546875,
3041
+ "learning_rate": 1.6595703401020844e-09,
3042
+ "loss": 0.0557,
3043
+ "reward": 1.109375,
3044
+ "reward_std": 0.1923343911767006,
3045
+ "rewards/equation_reward_func": 0.109375,
3046
+ "rewards/format_reward_func": 1.0,
3047
+ "step": 434
3048
+ },
3049
+ {
3050
+ "clip_ratio": 0.0,
3051
+ "completion_length": 295.828125,
3052
+ "epoch": 0.07751111111111111,
3053
+ "grad_norm": 109.41721675323444,
3054
+ "kl": 28.513916015625,
3055
+ "learning_rate": 1.2709384929615596e-09,
3056
+ "loss": 0.0721,
3057
+ "reward": 1.09375,
3058
+ "reward_std": 0.13466878235340118,
3059
+ "rewards/equation_reward_func": 0.09375,
3060
+ "rewards/format_reward_func": 1.0,
3061
+ "step": 436
3062
+ },
3063
+ {
3064
+ "clip_ratio": 0.0,
3065
+ "completion_length": 280.578125,
3066
+ "epoch": 0.07786666666666667,
3067
+ "grad_norm": 1.9228717734559753,
3068
+ "kl": 2.058837890625,
3069
+ "learning_rate": 9.339608617077165e-10,
3070
+ "loss": -0.034,
3071
+ "reward": 1.015625,
3072
+ "reward_std": 0.09375,
3073
+ "rewards/equation_reward_func": 0.03125,
3074
+ "rewards/format_reward_func": 0.984375,
3075
+ "step": 438
3076
+ },
3077
+ {
3078
+ "clip_ratio": 0.0,
3079
+ "completion_length": 292.5,
3080
+ "epoch": 0.07822222222222222,
3081
+ "grad_norm": 3.2983736045582552,
3082
+ "kl": 1.70654296875,
3083
+ "learning_rate": 6.487074273681114e-10,
3084
+ "loss": 0.0022,
3085
+ "reward": 1.140625,
3086
+ "reward_std": 0.22841878235340118,
3087
+ "rewards/equation_reward_func": 0.140625,
3088
+ "rewards/format_reward_func": 1.0,
3089
+ "step": 440
3090
+ },
3091
+ {
3092
+ "clip_ratio": 0.0,
3093
+ "completion_length": 280.296875,
3094
+ "epoch": 0.07857777777777777,
3095
+ "grad_norm": 2.7643778642220154,
3096
+ "kl": 1.887939453125,
3097
+ "learning_rate": 4.152374292708538e-10,
3098
+ "loss": 0.0083,
3099
+ "reward": 1.140625,
3100
+ "reward_std": 0.1923343911767006,
3101
+ "rewards/equation_reward_func": 0.140625,
3102
+ "rewards/format_reward_func": 1.0,
3103
+ "step": 442
3104
+ },
3105
+ {
3106
+ "clip_ratio": 0.0,
3107
+ "completion_length": 287.140625,
3108
+ "epoch": 0.07893333333333333,
3109
+ "grad_norm": 2.995996483175849,
3110
+ "kl": 1.871337890625,
3111
+ "learning_rate": 2.3359935274214204e-10,
3112
+ "loss": 0.0448,
3113
+ "reward": 1.265625,
3114
+ "reward_std": 0.33667195588350296,
3115
+ "rewards/equation_reward_func": 0.28125,
3116
+ "rewards/format_reward_func": 0.984375,
3117
+ "step": 444
3118
+ },
3119
+ {
3120
+ "clip_ratio": 0.0,
3121
+ "completion_length": 307.890625,
3122
+ "epoch": 0.0792888888888889,
3123
+ "grad_norm": 2.600208741694735,
3124
+ "kl": 2.14111328125,
3125
+ "learning_rate": 1.0383091903720665e-10,
3126
+ "loss": 0.0051,
3127
+ "reward": 1.0,
3128
+ "reward_std": 0.0625,
3129
+ "rewards/equation_reward_func": 0.015625,
3130
+ "rewards/format_reward_func": 0.984375,
3131
+ "step": 446
3132
+ },
3133
+ {
3134
+ "clip_ratio": 0.0,
3135
+ "completion_length": 283.359375,
3136
+ "epoch": 0.07964444444444445,
3137
+ "grad_norm": 2.5889730669425286,
3138
+ "kl": 377.768310546875,
3139
+ "learning_rate": 2.595907750671533e-11,
3140
+ "loss": 0.4151,
3141
+ "reward": 1.046875,
3142
+ "reward_std": 0.20728103816509247,
3143
+ "rewards/equation_reward_func": 0.078125,
3144
+ "rewards/format_reward_func": 0.96875,
3145
+ "step": 448
3146
+ },
3147
+ {
3148
+ "clip_ratio": 0.0,
3149
+ "completion_length": 283.171875,
3150
+ "epoch": 0.08,
3151
+ "grad_norm": 2.8580107740928953,
3152
+ "kl": 2.1181640625,
3153
+ "learning_rate": 0.0,
3154
+ "loss": 0.0826,
3155
+ "reward": 1.125,
3156
+ "reward_std": 0.2235843911767006,
3157
+ "rewards/equation_reward_func": 0.125,
3158
+ "rewards/format_reward_func": 1.0,
3159
+ "step": 450
3160
+ },
3161
+ {
3162
+ "epoch": 0.08,
3163
+ "step": 450,
3164
+ "total_flos": 0.0,
3165
+ "train_loss": 0.04409932630071934,
3166
+ "train_runtime": 12236.3143,
3167
+ "train_samples_per_second": 1.177,
3168
+ "train_steps_per_second": 0.037
3169
+ }
3170
+ ],
3171
+ "logging_steps": 2,
3172
+ "max_steps": 450,
3173
+ "num_input_tokens_seen": 0,
3174
+ "num_train_epochs": 1,
3175
+ "save_steps": 25,
3176
+ "stateful_callbacks": {
3177
+ "TrainerControl": {
3178
+ "args": {
3179
+ "should_epoch_stop": false,
3180
+ "should_evaluate": false,
3181
+ "should_log": false,
3182
+ "should_save": true,
3183
+ "should_training_stop": true
3184
+ },
3185
+ "attributes": {}
3186
+ }
3187
+ },
3188
+ "total_flos": 0.0,
3189
+ "train_batch_size": 4,
3190
+ "trial_name": null,
3191
+ "trial_params": null
3192
+ }