Siddeshwar1625 commited on
Commit
712001d
·
verified ·
1 Parent(s): 6cb9f23

Upload generator checkpoints for round 001

Browse files
Files changed (30) hide show
  1. self_play_hf_a10g_train/round_001/generator_train/README.md +67 -0
  2. self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/chat_template.jinja +54 -0
  3. self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/config.json +57 -0
  4. self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/generation_config.json +13 -0
  5. self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/model.safetensors +3 -0
  6. self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/optimizer.pt +3 -0
  7. self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/rng_state.pth +3 -0
  8. self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/scheduler.pt +3 -0
  9. self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/tokenizer.json +3 -0
  10. self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/tokenizer_config.json +32 -0
  11. self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/trainer_state.json +764 -0
  12. self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/training_args.bin +3 -0
  13. self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/chat_template.jinja +54 -0
  14. self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/config.json +57 -0
  15. self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/generation_config.json +13 -0
  16. self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/model.safetensors +3 -0
  17. self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/optimizer.pt +3 -0
  18. self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/rng_state.pth +3 -0
  19. self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/scheduler.pt +3 -0
  20. self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/tokenizer.json +3 -0
  21. self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/tokenizer_config.json +32 -0
  22. self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/trainer_state.json +953 -0
  23. self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/training_args.bin +3 -0
  24. self_play_hf_a10g_train/round_001/generator_train/final_model/chat_template.jinja +54 -0
  25. self_play_hf_a10g_train/round_001/generator_train/final_model/config.json +57 -0
  26. self_play_hf_a10g_train/round_001/generator_train/final_model/generation_config.json +13 -0
  27. self_play_hf_a10g_train/round_001/generator_train/final_model/model.safetensors +3 -0
  28. self_play_hf_a10g_train/round_001/generator_train/final_model/tokenizer.json +3 -0
  29. self_play_hf_a10g_train/round_001/generator_train/final_model/tokenizer_config.json +32 -0
  30. self_play_hf_a10g_train/round_001/generator_train/final_model/training_args.bin +3 -0
self_play_hf_a10g_train/round_001/generator_train/README.md ADDED
@@ -0,0 +1,67 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: Qwen/Qwen2.5-0.5B-Instruct
3
+ library_name: transformers
4
+ model_name: generator_train
5
+ tags:
6
+ - generated_from_trainer
7
+ - grpo
8
+ - trl
9
+ licence: license
10
+ ---
11
+
12
+ # Model Card for generator_train
13
+
14
+ This model is a fine-tuned version of [Qwen/Qwen2.5-0.5B-Instruct](https://huggingface.co/Qwen/Qwen2.5-0.5B-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="None", 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/siddeshwar2004-international-institute-of-information-te/osint-self-play-train/runs/1tah27vo)
31
+
32
+
33
+
34
+ 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).
35
+
36
+ ### Framework versions
37
+
38
+ - TRL: 1.2.0
39
+ - Transformers: 5.6.2
40
+ - Pytorch: 2.11.0
41
+ - Datasets: 4.8.4
42
+ - Tokenizers: 0.22.2
43
+
44
+ ## Citations
45
+
46
+ Cite GRPO as:
47
+
48
+ ```bibtex
49
+ @article{shao2024deepseekmath,
50
+ title = {{DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models}},
51
+ 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},
52
+ year = 2024,
53
+ eprint = {arXiv:2402.03300},
54
+ }
55
+ ```
56
+
57
+ Cite TRL as:
58
+
59
+ ```bibtex
60
+ @software{vonwerra2020trl,
61
+ title = {{TRL: Transformers Reinforcement Learning}},
62
+ author = {von Werra, Leandro and Belkada, Younes and Tunstall, Lewis and Beeching, Edward and Thrush, Tristan and Lambert, Nathan and Huang, Shengyi and Rasul, Kashif and Gallouédec, Quentin},
63
+ license = {Apache-2.0},
64
+ url = {https://github.com/huggingface/trl},
65
+ year = {2020}
66
+ }
67
+ ```
self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/chat_template.jinja ADDED
@@ -0,0 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {%- if tools %}
2
+ {{- '<|im_start|>system\n' }}
3
+ {%- if messages[0]['role'] == 'system' %}
4
+ {{- messages[0]['content'] }}
5
+ {%- else %}
6
+ {{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }}
7
+ {%- endif %}
8
+ {{- "\n\n# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
9
+ {%- for tool in tools %}
10
+ {{- "\n" }}
11
+ {{- tool | tojson }}
12
+ {%- endfor %}
13
+ {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
14
+ {%- else %}
15
+ {%- if messages[0]['role'] == 'system' %}
16
+ {{- '<|im_start|>system\n' + messages[0]['content'] + '<|im_end|>\n' }}
17
+ {%- else %}
18
+ {{- '<|im_start|>system\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\n' }}
19
+ {%- endif %}
20
+ {%- endif %}
21
+ {%- for message in messages %}
22
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) or (message.role == "assistant" and not message.tool_calls) %}
23
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
24
+ {%- elif message.role == "assistant" %}
25
+ {{- '<|im_start|>' + message.role }}
26
+ {%- if message.content %}
27
+ {{- '\n' + message.content }}
28
+ {%- endif %}
29
+ {%- for tool_call in message.tool_calls %}
30
+ {%- if tool_call.function is defined %}
31
+ {%- set tool_call = tool_call.function %}
32
+ {%- endif %}
33
+ {{- '\n<tool_call>\n{"name": "' }}
34
+ {{- tool_call.name }}
35
+ {{- '", "arguments": ' }}
36
+ {{- tool_call.arguments | tojson }}
37
+ {{- '}\n</tool_call>' }}
38
+ {%- endfor %}
39
+ {{- '<|im_end|>\n' }}
40
+ {%- elif message.role == "tool" %}
41
+ {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != "tool") %}
42
+ {{- '<|im_start|>user' }}
43
+ {%- endif %}
44
+ {{- '\n<tool_response>\n' }}
45
+ {{- message.content }}
46
+ {{- '\n</tool_response>' }}
47
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
48
+ {{- '<|im_end|>\n' }}
49
+ {%- endif %}
50
+ {%- endif %}
51
+ {%- endfor %}
52
+ {%- if add_generation_prompt %}
53
+ {{- '<|im_start|>assistant\n' }}
54
+ {%- endif %}
self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/config.json ADDED
@@ -0,0 +1,57 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Qwen2ForCausalLM"
4
+ ],
5
+ "attention_dropout": 0.0,
6
+ "bos_token_id": null,
7
+ "dtype": "float32",
8
+ "eos_token_id": 151645,
9
+ "hidden_act": "silu",
10
+ "hidden_size": 896,
11
+ "initializer_range": 0.02,
12
+ "intermediate_size": 4864,
13
+ "layer_types": [
14
+ "full_attention",
15
+ "full_attention",
16
+ "full_attention",
17
+ "full_attention",
18
+ "full_attention",
19
+ "full_attention",
20
+ "full_attention",
21
+ "full_attention",
22
+ "full_attention",
23
+ "full_attention",
24
+ "full_attention",
25
+ "full_attention",
26
+ "full_attention",
27
+ "full_attention",
28
+ "full_attention",
29
+ "full_attention",
30
+ "full_attention",
31
+ "full_attention",
32
+ "full_attention",
33
+ "full_attention",
34
+ "full_attention",
35
+ "full_attention",
36
+ "full_attention",
37
+ "full_attention"
38
+ ],
39
+ "max_position_embeddings": 32768,
40
+ "max_window_layers": 21,
41
+ "model_type": "qwen2",
42
+ "num_attention_heads": 14,
43
+ "num_hidden_layers": 24,
44
+ "num_key_value_heads": 2,
45
+ "pad_token_id": 151643,
46
+ "rms_norm_eps": 1e-06,
47
+ "rope_parameters": {
48
+ "rope_theta": 1000000.0,
49
+ "rope_type": "default"
50
+ },
51
+ "sliding_window": null,
52
+ "tie_word_embeddings": true,
53
+ "transformers_version": "5.6.2",
54
+ "use_cache": false,
55
+ "use_sliding_window": false,
56
+ "vocab_size": 151936
57
+ }
self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/generation_config.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_sample": true,
3
+ "eos_token_id": [
4
+ 151645,
5
+ 151643
6
+ ],
7
+ "pad_token_id": 151643,
8
+ "repetition_penalty": 1.1,
9
+ "temperature": 0.7,
10
+ "top_k": 20,
11
+ "top_p": 0.8,
12
+ "transformers_version": "5.6.2"
13
+ }
self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:af6dc3d2420ea21956bb1d12f3b95a123b0b8df6807959734b3a992fb4a284c5
3
+ size 1976163472
self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e18a61cd56a6cbde84b8be5685dc8ccd8c446b5783bb0ecc590e018319d0c085
3
+ size 3952509771
self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b5725c96886e63d92a4804b9e1b509a94ed72b0ac9da7f6955ce8a319b258d37
3
+ size 14645
self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:44b7842b78ac3e944fa674f4961bef93e595fe53f0c4495643408e144ea2a074
3
+ size 1465
self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3fd169731d2cbde95e10bf356d66d5997fd885dd8dbb6fb4684da3f23b2585d8
3
+ size 11421892
self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/tokenizer_config.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "backend": "tokenizers",
4
+ "bos_token": null,
5
+ "clean_up_tokenization_spaces": false,
6
+ "eos_token": "<|im_end|>",
7
+ "errors": "replace",
8
+ "extra_special_tokens": [
9
+ "<|im_start|>",
10
+ "<|im_end|>",
11
+ "<|object_ref_start|>",
12
+ "<|object_ref_end|>",
13
+ "<|box_start|>",
14
+ "<|box_end|>",
15
+ "<|quad_start|>",
16
+ "<|quad_end|>",
17
+ "<|vision_start|>",
18
+ "<|vision_end|>",
19
+ "<|vision_pad|>",
20
+ "<|image_pad|>",
21
+ "<|video_pad|>"
22
+ ],
23
+ "is_local": false,
24
+ "local_files_only": false,
25
+ "model_max_length": 131072,
26
+ "pad_token": "<|endoftext|>",
27
+ "padding_side": "left",
28
+ "split_special_tokens": false,
29
+ "tokenizer_class": "Qwen2Tokenizer",
30
+ "truncation_side": "left",
31
+ "unk_token": null
32
+ }
self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/trainer_state.json ADDED
@@ -0,0 +1,764 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 1.6666666666666665,
6
+ "eval_steps": 500,
7
+ "global_step": 40,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "clip_ratio/high_max": 0.001953125,
14
+ "clip_ratio/high_mean": 0.001953125,
15
+ "clip_ratio/low_mean": 0.0,
16
+ "clip_ratio/low_min": 0.0,
17
+ "clip_ratio/region_mean": 0.001953125,
18
+ "completions/clipped_ratio": 1.0,
19
+ "completions/max_length": 384.0,
20
+ "completions/max_terminated_length": 0.0,
21
+ "completions/mean_length": 384.0,
22
+ "completions/mean_terminated_length": 0.0,
23
+ "completions/min_length": 384.0,
24
+ "completions/min_terminated_length": 0.0,
25
+ "entropy": 1.9362258911132812,
26
+ "epoch": 0.041666666666666664,
27
+ "frac_reward_zero_std": 0.0,
28
+ "grad_norm": 3.342789649963379,
29
+ "kl": 0.0005498419050127268,
30
+ "learning_rate": 5e-06,
31
+ "loss": 0.13995857536792755,
32
+ "num_tokens": 25244.0,
33
+ "reward": -0.4352343678474426,
34
+ "reward_std": 0.306624174118042,
35
+ "rewards/GeneratorRewardFunction/mean": -0.4352343678474426,
36
+ "rewards/GeneratorRewardFunction/std": 0.306624174118042,
37
+ "step": 1,
38
+ "step_time": 12.708289422999997
39
+ },
40
+ {
41
+ "clip_ratio/high_max": 0.0045572915114462376,
42
+ "clip_ratio/high_mean": 0.0045572915114462376,
43
+ "clip_ratio/low_mean": 0.0026041667442768812,
44
+ "clip_ratio/low_min": 0.0026041667442768812,
45
+ "clip_ratio/region_mean": 0.0071614584885537624,
46
+ "entropy": 1.2719587087631226,
47
+ "epoch": 0.08333333333333333,
48
+ "grad_norm": 2.8023905754089355,
49
+ "kl": 0.0014687416842207313,
50
+ "learning_rate": 4.9000000000000005e-06,
51
+ "loss": -0.0669216588139534,
52
+ "step": 2,
53
+ "step_time": 0.221013198999799
54
+ },
55
+ {
56
+ "clip_ratio/high_max": 0.010416666977107525,
57
+ "clip_ratio/high_mean": 0.010416666977107525,
58
+ "clip_ratio/low_mean": 0.014322916977107525,
59
+ "clip_ratio/low_min": 0.014322916977107525,
60
+ "clip_ratio/region_mean": 0.02473958395421505,
61
+ "entropy": 1.8711100816726685,
62
+ "epoch": 0.125,
63
+ "grad_norm": 2.356076717376709,
64
+ "kl": 0.004658207297325134,
65
+ "learning_rate": 4.800000000000001e-06,
66
+ "loss": 0.010263405740261078,
67
+ "step": 3,
68
+ "step_time": 0.21947432299998582
69
+ },
70
+ {
71
+ "clip_ratio/high_max": 0.02018229104578495,
72
+ "clip_ratio/high_mean": 0.02018229104578495,
73
+ "clip_ratio/low_mean": 0.009765625,
74
+ "clip_ratio/low_min": 0.009765625,
75
+ "clip_ratio/region_mean": 0.02994791604578495,
76
+ "entropy": 1.1866708993911743,
77
+ "epoch": 0.16666666666666666,
78
+ "grad_norm": 1.5606235265731812,
79
+ "kl": 0.005778420716524124,
80
+ "learning_rate": 4.7e-06,
81
+ "loss": -0.08204194158315659,
82
+ "step": 4,
83
+ "step_time": 0.21935116999998172
84
+ },
85
+ {
86
+ "clip_ratio/high_max": 0.0013020833721384406,
87
+ "clip_ratio/high_mean": 0.0013020833721384406,
88
+ "clip_ratio/low_mean": 0.0006510416860692203,
89
+ "clip_ratio/low_min": 0.0006510416860692203,
90
+ "clip_ratio/region_mean": 0.001953125,
91
+ "completions/clipped_ratio": 1.0,
92
+ "completions/max_length": 384.0,
93
+ "completions/max_terminated_length": 0.0,
94
+ "completions/mean_length": 384.0,
95
+ "completions/mean_terminated_length": 0.0,
96
+ "completions/min_length": 384.0,
97
+ "completions/min_terminated_length": 0.0,
98
+ "entropy": 1.1421619653701782,
99
+ "epoch": 0.20833333333333334,
100
+ "frac_reward_zero_std": 0.0,
101
+ "grad_norm": 3.0044524669647217,
102
+ "kl": 0.017580831423401833,
103
+ "learning_rate": 4.600000000000001e-06,
104
+ "loss": 0.022549070417881012,
105
+ "num_tokens": 50440.0,
106
+ "reward": -0.3424999713897705,
107
+ "reward_std": 0.3177734613418579,
108
+ "rewards/GeneratorRewardFunction/mean": -0.3424999713897705,
109
+ "rewards/GeneratorRewardFunction/std": 0.3177734911441803,
110
+ "step": 5,
111
+ "step_time": 12.087754471000153
112
+ },
113
+ {
114
+ "clip_ratio/high_max": 0.008463541977107525,
115
+ "clip_ratio/high_mean": 0.008463541977107525,
116
+ "clip_ratio/low_mean": 0.0052083334885537624,
117
+ "clip_ratio/low_min": 0.0052083334885537624,
118
+ "clip_ratio/region_mean": 0.013671875,
119
+ "entropy": 1.110270619392395,
120
+ "epoch": 0.25,
121
+ "grad_norm": 2.2674970626831055,
122
+ "kl": 0.01950421743094921,
123
+ "learning_rate": 4.5e-06,
124
+ "loss": 0.00025658251252025366,
125
+ "step": 6,
126
+ "step_time": 0.2263194110000768
127
+ },
128
+ {
129
+ "clip_ratio/high_max": 0.01171875,
130
+ "clip_ratio/high_mean": 0.01171875,
131
+ "clip_ratio/low_mean": 0.008463541977107525,
132
+ "clip_ratio/low_min": 0.008463541977107525,
133
+ "clip_ratio/region_mean": 0.02018229104578495,
134
+ "entropy": 1.1433427333831787,
135
+ "epoch": 0.2916666666666667,
136
+ "grad_norm": 2.097913980484009,
137
+ "kl": 0.050842806696891785,
138
+ "learning_rate": 4.4e-06,
139
+ "loss": -0.02153957635164261,
140
+ "step": 7,
141
+ "step_time": 0.2258150579998528
142
+ },
143
+ {
144
+ "clip_ratio/high_max": 0.021484375,
145
+ "clip_ratio/high_mean": 0.021484375,
146
+ "clip_ratio/low_mean": 0.013020833022892475,
147
+ "clip_ratio/low_min": 0.013020833022892475,
148
+ "clip_ratio/region_mean": 0.0345052070915699,
149
+ "entropy": 1.5583876371383667,
150
+ "epoch": 0.3333333333333333,
151
+ "grad_norm": 2.776447057723999,
152
+ "kl": 0.06861245632171631,
153
+ "learning_rate": 4.3e-06,
154
+ "loss": 0.0012576957233250141,
155
+ "step": 8,
156
+ "step_time": 0.22490879599990876
157
+ },
158
+ {
159
+ "clip_ratio/high_max": 0.0,
160
+ "clip_ratio/high_mean": 0.0,
161
+ "clip_ratio/low_mean": 0.0,
162
+ "clip_ratio/low_min": 0.0,
163
+ "clip_ratio/region_mean": 0.0,
164
+ "completions/clipped_ratio": 1.0,
165
+ "completions/max_length": 384.0,
166
+ "completions/max_terminated_length": 0.0,
167
+ "completions/mean_length": 384.0,
168
+ "completions/mean_terminated_length": 0.0,
169
+ "completions/min_length": 384.0,
170
+ "completions/min_terminated_length": 0.0,
171
+ "entropy": 1.256606101989746,
172
+ "epoch": 0.375,
173
+ "frac_reward_zero_std": 0.0,
174
+ "grad_norm": 3.1693882942199707,
175
+ "kl": 0.034918785095214844,
176
+ "learning_rate": 4.2000000000000004e-06,
177
+ "loss": 0.06234124302864075,
178
+ "num_tokens": 75884.0,
179
+ "reward": -0.3112500011920929,
180
+ "reward_std": 0.3181116282939911,
181
+ "rewards/GeneratorRewardFunction/mean": -0.3112500011920929,
182
+ "rewards/GeneratorRewardFunction/std": 0.3181115984916687,
183
+ "step": 9,
184
+ "step_time": 12.085582495999915
185
+ },
186
+ {
187
+ "clip_ratio/high_max": 0.00390625,
188
+ "clip_ratio/high_mean": 0.00390625,
189
+ "clip_ratio/low_mean": 0.001953125,
190
+ "clip_ratio/low_min": 0.001953125,
191
+ "clip_ratio/region_mean": 0.005859375,
192
+ "entropy": 1.150619387626648,
193
+ "epoch": 0.4166666666666667,
194
+ "grad_norm": 2.827258348464966,
195
+ "kl": 0.026281872764229774,
196
+ "learning_rate": 4.1e-06,
197
+ "loss": -0.04307486489415169,
198
+ "step": 10,
199
+ "step_time": 0.22665007600016907
200
+ },
201
+ {
202
+ "clip_ratio/high_max": 0.013020833022892475,
203
+ "clip_ratio/high_mean": 0.013020833022892475,
204
+ "clip_ratio/low_mean": 0.012369791977107525,
205
+ "clip_ratio/low_min": 0.012369791977107525,
206
+ "clip_ratio/region_mean": 0.025390625,
207
+ "entropy": 1.353300929069519,
208
+ "epoch": 0.4583333333333333,
209
+ "grad_norm": 1.517728328704834,
210
+ "kl": 0.043695781379938126,
211
+ "learning_rate": 4.000000000000001e-06,
212
+ "loss": 0.0004213004431221634,
213
+ "step": 11,
214
+ "step_time": 0.2250163410001278
215
+ },
216
+ {
217
+ "clip_ratio/high_max": 0.015625,
218
+ "clip_ratio/high_mean": 0.015625,
219
+ "clip_ratio/low_mean": 0.01953125,
220
+ "clip_ratio/low_min": 0.01953125,
221
+ "clip_ratio/region_mean": 0.03515625,
222
+ "entropy": 1.5499415397644043,
223
+ "epoch": 0.5,
224
+ "grad_norm": 2.2769711017608643,
225
+ "kl": 0.04331633448600769,
226
+ "learning_rate": 3.900000000000001e-06,
227
+ "loss": -0.019014613702893257,
228
+ "step": 12,
229
+ "step_time": 0.2254276790001768
230
+ },
231
+ {
232
+ "clip_ratio/high_max": 0.0026041667442768812,
233
+ "clip_ratio/high_mean": 0.0026041667442768812,
234
+ "clip_ratio/low_mean": 0.0,
235
+ "clip_ratio/low_min": 0.0,
236
+ "clip_ratio/region_mean": 0.0026041667442768812,
237
+ "completions/clipped_ratio": 1.0,
238
+ "completions/max_length": 384.0,
239
+ "completions/max_terminated_length": 0.0,
240
+ "completions/mean_length": 384.0,
241
+ "completions/mean_terminated_length": 0.0,
242
+ "completions/min_length": 384.0,
243
+ "completions/min_terminated_length": 0.0,
244
+ "entropy": 0.9478046298027039,
245
+ "epoch": 0.5416666666666666,
246
+ "frac_reward_zero_std": 0.0,
247
+ "grad_norm": 1.8451720476150513,
248
+ "kl": 0.042272429913282394,
249
+ "learning_rate": 3.8000000000000005e-06,
250
+ "loss": -0.06487596780061722,
251
+ "num_tokens": 101112.0,
252
+ "reward": -0.3031249940395355,
253
+ "reward_std": 0.3281609117984772,
254
+ "rewards/GeneratorRewardFunction/mean": -0.3031249940395355,
255
+ "rewards/GeneratorRewardFunction/std": 0.3281609117984772,
256
+ "step": 13,
257
+ "step_time": 12.174987709000106
258
+ },
259
+ {
260
+ "clip_ratio/high_max": 0.0065104165114462376,
261
+ "clip_ratio/high_mean": 0.0065104165114462376,
262
+ "clip_ratio/low_mean": 0.0,
263
+ "clip_ratio/low_min": 0.0,
264
+ "clip_ratio/region_mean": 0.0065104165114462376,
265
+ "entropy": 1.186617374420166,
266
+ "epoch": 0.5833333333333334,
267
+ "grad_norm": 1.483729600906372,
268
+ "kl": 0.05555545911192894,
269
+ "learning_rate": 3.7e-06,
270
+ "loss": -0.12233645468950272,
271
+ "step": 14,
272
+ "step_time": 0.21759630599990487
273
+ },
274
+ {
275
+ "clip_ratio/high_max": 0.0,
276
+ "clip_ratio/high_mean": 0.0,
277
+ "clip_ratio/low_mean": 0.008463541977107525,
278
+ "clip_ratio/low_min": 0.008463541977107525,
279
+ "clip_ratio/region_mean": 0.008463541977107525,
280
+ "entropy": 0.8805699348449707,
281
+ "epoch": 0.625,
282
+ "grad_norm": 3.184286594390869,
283
+ "kl": 0.05239926651120186,
284
+ "learning_rate": 3.6000000000000003e-06,
285
+ "loss": 0.27255433797836304,
286
+ "step": 15,
287
+ "step_time": 0.21799612400013757
288
+ },
289
+ {
290
+ "clip_ratio/high_max": 0.014322916977107525,
291
+ "clip_ratio/high_mean": 0.014322916977107525,
292
+ "clip_ratio/low_mean": 0.0032552082557231188,
293
+ "clip_ratio/low_min": 0.0032552082557231188,
294
+ "clip_ratio/region_mean": 0.017578125,
295
+ "entropy": 1.7632579803466797,
296
+ "epoch": 0.6666666666666666,
297
+ "grad_norm": 2.479675531387329,
298
+ "kl": 0.09853244572877884,
299
+ "learning_rate": 3.5e-06,
300
+ "loss": -0.08373244851827621,
301
+ "step": 16,
302
+ "step_time": 0.21777955800007476
303
+ },
304
+ {
305
+ "clip_ratio/high_max": 0.0,
306
+ "clip_ratio/high_mean": 0.0,
307
+ "clip_ratio/low_mean": 0.0,
308
+ "clip_ratio/low_min": 0.0,
309
+ "clip_ratio/region_mean": 0.0,
310
+ "completions/clipped_ratio": 1.0,
311
+ "completions/max_length": 384.0,
312
+ "completions/max_terminated_length": 0.0,
313
+ "completions/mean_length": 384.0,
314
+ "completions/mean_terminated_length": 0.0,
315
+ "completions/min_length": 384.0,
316
+ "completions/min_terminated_length": 0.0,
317
+ "entropy": 1.2678290605545044,
318
+ "epoch": 0.7083333333333334,
319
+ "frac_reward_zero_std": 0.0,
320
+ "grad_norm": 3.134302854537964,
321
+ "kl": 0.06070971116423607,
322
+ "learning_rate": 3.4000000000000005e-06,
323
+ "loss": -0.052310068160295486,
324
+ "num_tokens": 126436.0,
325
+ "reward": -0.2212499976158142,
326
+ "reward_std": 0.27250993251800537,
327
+ "rewards/GeneratorRewardFunction/mean": -0.2212499976158142,
328
+ "rewards/GeneratorRewardFunction/std": 0.27250996232032776,
329
+ "step": 17,
330
+ "step_time": 12.04144509899993
331
+ },
332
+ {
333
+ "clip_ratio/high_max": 0.0,
334
+ "clip_ratio/high_mean": 0.0,
335
+ "clip_ratio/low_mean": 0.0006510416860692203,
336
+ "clip_ratio/low_min": 0.0006510416860692203,
337
+ "clip_ratio/region_mean": 0.0006510416860692203,
338
+ "entropy": 1.7929935455322266,
339
+ "epoch": 0.75,
340
+ "grad_norm": 3.1320412158966064,
341
+ "kl": 0.062457334250211716,
342
+ "learning_rate": 3.3000000000000006e-06,
343
+ "loss": -0.03248615190386772,
344
+ "step": 18,
345
+ "step_time": 0.2239023360000374
346
+ },
347
+ {
348
+ "clip_ratio/high_max": 0.01627604104578495,
349
+ "clip_ratio/high_mean": 0.01627604104578495,
350
+ "clip_ratio/low_mean": 0.0,
351
+ "clip_ratio/low_min": 0.0,
352
+ "clip_ratio/region_mean": 0.01627604104578495,
353
+ "entropy": 1.145819067955017,
354
+ "epoch": 0.7916666666666666,
355
+ "grad_norm": 1.5332239866256714,
356
+ "kl": 0.07924208045005798,
357
+ "learning_rate": 3.2000000000000003e-06,
358
+ "loss": -0.08211782574653625,
359
+ "step": 19,
360
+ "step_time": 0.22507186799998635
361
+ },
362
+ {
363
+ "clip_ratio/high_max": 0.0,
364
+ "clip_ratio/high_mean": 0.0,
365
+ "clip_ratio/low_mean": 0.012369791977107525,
366
+ "clip_ratio/low_min": 0.012369791977107525,
367
+ "clip_ratio/region_mean": 0.012369791977107525,
368
+ "entropy": 0.9164741635322571,
369
+ "epoch": 0.8333333333333334,
370
+ "grad_norm": 2.2384536266326904,
371
+ "kl": 0.06750863045454025,
372
+ "learning_rate": 3.1000000000000004e-06,
373
+ "loss": 0.16943609714508057,
374
+ "step": 20,
375
+ "step_time": 0.22430561300006957
376
+ },
377
+ {
378
+ "clip_ratio/high_max": 0.0013020833721384406,
379
+ "clip_ratio/high_mean": 0.0013020833721384406,
380
+ "clip_ratio/low_mean": 0.0,
381
+ "clip_ratio/low_min": 0.0,
382
+ "clip_ratio/region_mean": 0.0013020833721384406,
383
+ "completions/clipped_ratio": 1.0,
384
+ "completions/max_length": 384.0,
385
+ "completions/max_terminated_length": 0.0,
386
+ "completions/mean_length": 384.0,
387
+ "completions/mean_terminated_length": 0.0,
388
+ "completions/min_length": 384.0,
389
+ "completions/min_terminated_length": 0.0,
390
+ "entropy": 1.862565040588379,
391
+ "epoch": 0.875,
392
+ "frac_reward_zero_std": 0.0,
393
+ "grad_norm": 2.588430643081665,
394
+ "kl": 0.04901580885052681,
395
+ "learning_rate": 3e-06,
396
+ "loss": -0.12625738978385925,
397
+ "num_tokens": 151596.0,
398
+ "reward": -0.2653124928474426,
399
+ "reward_std": 0.2914073169231415,
400
+ "rewards/GeneratorRewardFunction/mean": -0.2653124928474426,
401
+ "rewards/GeneratorRewardFunction/std": 0.2914073169231415,
402
+ "step": 21,
403
+ "step_time": 12.2114765299998
404
+ },
405
+ {
406
+ "clip_ratio/high_max": 0.0032552082557231188,
407
+ "clip_ratio/high_mean": 0.0032552082557231188,
408
+ "clip_ratio/low_mean": 0.0,
409
+ "clip_ratio/low_min": 0.0,
410
+ "clip_ratio/region_mean": 0.0032552082557231188,
411
+ "entropy": 1.3516654968261719,
412
+ "epoch": 0.9166666666666666,
413
+ "grad_norm": 2.4181604385375977,
414
+ "kl": 0.061538487672805786,
415
+ "learning_rate": 2.9e-06,
416
+ "loss": 0.0242471843957901,
417
+ "step": 22,
418
+ "step_time": 0.21940968300009445
419
+ },
420
+ {
421
+ "clip_ratio/high_max": 0.0065104165114462376,
422
+ "clip_ratio/high_mean": 0.0065104165114462376,
423
+ "clip_ratio/low_mean": 0.0013020833721384406,
424
+ "clip_ratio/low_min": 0.0013020833721384406,
425
+ "clip_ratio/region_mean": 0.0078125,
426
+ "entropy": 1.3538661003112793,
427
+ "epoch": 0.9583333333333334,
428
+ "grad_norm": 2.905212879180908,
429
+ "kl": 0.07095103710889816,
430
+ "learning_rate": 2.8000000000000003e-06,
431
+ "loss": 0.08377016335725784,
432
+ "step": 23,
433
+ "step_time": 0.22010638500000823
434
+ },
435
+ {
436
+ "clip_ratio/high_max": 0.013020833022892475,
437
+ "clip_ratio/high_mean": 0.013020833022892475,
438
+ "clip_ratio/low_mean": 0.001953125,
439
+ "clip_ratio/low_min": 0.001953125,
440
+ "clip_ratio/region_mean": 0.014973958022892475,
441
+ "entropy": 1.703626275062561,
442
+ "epoch": 1.0,
443
+ "grad_norm": 2.835925340652466,
444
+ "kl": 0.0820431113243103,
445
+ "learning_rate": 2.7000000000000004e-06,
446
+ "loss": 0.018686100840568542,
447
+ "step": 24,
448
+ "step_time": 0.21947678399988035
449
+ },
450
+ {
451
+ "clip_ratio/high_max": 0.0,
452
+ "clip_ratio/high_mean": 0.0,
453
+ "clip_ratio/low_mean": 0.0,
454
+ "clip_ratio/low_min": 0.0,
455
+ "clip_ratio/region_mean": 0.0,
456
+ "completions/clipped_ratio": 1.0,
457
+ "completions/max_length": 384.0,
458
+ "completions/max_terminated_length": 0.0,
459
+ "completions/mean_length": 384.0,
460
+ "completions/mean_terminated_length": 0.0,
461
+ "completions/min_length": 384.0,
462
+ "completions/min_terminated_length": 0.0,
463
+ "entropy": 1.6442562341690063,
464
+ "epoch": 1.0416666666666667,
465
+ "frac_reward_zero_std": 0.0,
466
+ "grad_norm": 2.382233142852783,
467
+ "kl": 0.0744522213935852,
468
+ "learning_rate": 2.6e-06,
469
+ "loss": 0.060507725924253464,
470
+ "num_tokens": 177212.0,
471
+ "reward": -0.2524999976158142,
472
+ "reward_std": 0.28701913356781006,
473
+ "rewards/GeneratorRewardFunction/mean": -0.2524999976158142,
474
+ "rewards/GeneratorRewardFunction/std": 0.28701916337013245,
475
+ "step": 25,
476
+ "step_time": 12.102920622000056
477
+ },
478
+ {
479
+ "clip_ratio/high_max": 0.001953125,
480
+ "clip_ratio/high_mean": 0.001953125,
481
+ "clip_ratio/low_mean": 0.0006510416860692203,
482
+ "clip_ratio/low_min": 0.0006510416860692203,
483
+ "clip_ratio/region_mean": 0.0026041667442768812,
484
+ "entropy": 1.1888453960418701,
485
+ "epoch": 1.0833333333333333,
486
+ "grad_norm": 2.2921271324157715,
487
+ "kl": 0.09428787231445312,
488
+ "learning_rate": 2.5e-06,
489
+ "loss": 0.06040751934051514,
490
+ "step": 26,
491
+ "step_time": 0.22665847199982636
492
+ },
493
+ {
494
+ "clip_ratio/high_max": 0.011067708022892475,
495
+ "clip_ratio/high_mean": 0.011067708022892475,
496
+ "clip_ratio/low_mean": 0.0,
497
+ "clip_ratio/low_min": 0.0,
498
+ "clip_ratio/region_mean": 0.011067708022892475,
499
+ "entropy": 1.1056007146835327,
500
+ "epoch": 1.125,
501
+ "grad_norm": 1.509630560874939,
502
+ "kl": 0.0742887407541275,
503
+ "learning_rate": 2.4000000000000003e-06,
504
+ "loss": -0.12057904154062271,
505
+ "step": 27,
506
+ "step_time": 0.22739392499988753
507
+ },
508
+ {
509
+ "clip_ratio/high_max": 0.005859375,
510
+ "clip_ratio/high_mean": 0.005859375,
511
+ "clip_ratio/low_mean": 0.0026041667442768812,
512
+ "clip_ratio/low_min": 0.0026041667442768812,
513
+ "clip_ratio/region_mean": 0.008463541977107525,
514
+ "entropy": 0.8066513538360596,
515
+ "epoch": 1.1666666666666667,
516
+ "grad_norm": 2.760502576828003,
517
+ "kl": 0.10300976783037186,
518
+ "learning_rate": 2.3000000000000004e-06,
519
+ "loss": 0.00034140495699830353,
520
+ "step": 28,
521
+ "step_time": 0.22687044499980402
522
+ },
523
+ {
524
+ "clip_ratio/high_max": 0.0,
525
+ "clip_ratio/high_mean": 0.0,
526
+ "clip_ratio/low_mean": 0.0013020833721384406,
527
+ "clip_ratio/low_min": 0.0013020833721384406,
528
+ "clip_ratio/region_mean": 0.0013020833721384406,
529
+ "completions/clipped_ratio": 1.0,
530
+ "completions/max_length": 384.0,
531
+ "completions/max_terminated_length": 0.0,
532
+ "completions/mean_length": 384.0,
533
+ "completions/mean_terminated_length": 0.0,
534
+ "completions/min_length": 384.0,
535
+ "completions/min_terminated_length": 0.0,
536
+ "entropy": 1.6604108810424805,
537
+ "epoch": 1.2083333333333333,
538
+ "frac_reward_zero_std": 0.0,
539
+ "grad_norm": 1.5455973148345947,
540
+ "kl": 0.09443795680999756,
541
+ "learning_rate": 2.2e-06,
542
+ "loss": 0.06687173247337341,
543
+ "num_tokens": 202200.0,
544
+ "reward": -0.12000000476837158,
545
+ "reward_std": 0.20759736001491547,
546
+ "rewards/GeneratorRewardFunction/mean": -0.12000000476837158,
547
+ "rewards/GeneratorRewardFunction/std": 0.20759736001491547,
548
+ "step": 29,
549
+ "step_time": 12.208307090000062
550
+ },
551
+ {
552
+ "clip_ratio/high_max": 0.0006510416860692203,
553
+ "clip_ratio/high_mean": 0.0006510416860692203,
554
+ "clip_ratio/low_mean": 0.0006510416860692203,
555
+ "clip_ratio/low_min": 0.0006510416860692203,
556
+ "clip_ratio/region_mean": 0.0013020833721384406,
557
+ "entropy": 1.0410054922103882,
558
+ "epoch": 1.25,
559
+ "grad_norm": 3.1164426803588867,
560
+ "kl": 0.07589004188776016,
561
+ "learning_rate": 2.1000000000000002e-06,
562
+ "loss": -0.12521938979625702,
563
+ "step": 30,
564
+ "step_time": 0.22144749400013097
565
+ },
566
+ {
567
+ "clip_ratio/high_max": 0.00390625,
568
+ "clip_ratio/high_mean": 0.00390625,
569
+ "clip_ratio/low_mean": 0.0,
570
+ "clip_ratio/low_min": 0.0,
571
+ "clip_ratio/region_mean": 0.00390625,
572
+ "entropy": 1.3995355367660522,
573
+ "epoch": 1.2916666666666667,
574
+ "grad_norm": 2.3790829181671143,
575
+ "kl": 0.10240024328231812,
576
+ "learning_rate": 2.0000000000000003e-06,
577
+ "loss": -0.006538359913975,
578
+ "step": 31,
579
+ "step_time": 0.22054820099992867
580
+ },
581
+ {
582
+ "clip_ratio/high_max": 0.0032552082557231188,
583
+ "clip_ratio/high_mean": 0.0032552082557231188,
584
+ "clip_ratio/low_mean": 0.0071614584885537624,
585
+ "clip_ratio/low_min": 0.0071614584885537624,
586
+ "clip_ratio/region_mean": 0.010416666977107525,
587
+ "entropy": 1.3301101922988892,
588
+ "epoch": 1.3333333333333333,
589
+ "grad_norm": 1.3749467134475708,
590
+ "kl": 0.09631212800741196,
591
+ "learning_rate": 1.9000000000000002e-06,
592
+ "loss": 0.06641887873411179,
593
+ "step": 32,
594
+ "step_time": 0.21976359399991452
595
+ },
596
+ {
597
+ "clip_ratio/high_max": 0.0,
598
+ "clip_ratio/high_mean": 0.0,
599
+ "clip_ratio/low_mean": 0.0,
600
+ "clip_ratio/low_min": 0.0,
601
+ "clip_ratio/region_mean": 0.0,
602
+ "completions/clipped_ratio": 1.0,
603
+ "completions/max_length": 384.0,
604
+ "completions/max_terminated_length": 0.0,
605
+ "completions/mean_length": 384.0,
606
+ "completions/mean_terminated_length": 0.0,
607
+ "completions/min_length": 384.0,
608
+ "completions/min_terminated_length": 0.0,
609
+ "entropy": 1.4248062372207642,
610
+ "epoch": 1.375,
611
+ "frac_reward_zero_std": 0.0,
612
+ "grad_norm": 2.76702618598938,
613
+ "kl": 0.08123808354139328,
614
+ "learning_rate": 1.8000000000000001e-06,
615
+ "loss": -0.20796377956867218,
616
+ "num_tokens": 227556.0,
617
+ "reward": -0.1759375035762787,
618
+ "reward_std": 0.23178452253341675,
619
+ "rewards/GeneratorRewardFunction/mean": -0.1759375035762787,
620
+ "rewards/GeneratorRewardFunction/std": 0.23178452253341675,
621
+ "step": 33,
622
+ "step_time": 12.118321306000098
623
+ },
624
+ {
625
+ "clip_ratio/high_max": 0.0,
626
+ "clip_ratio/high_mean": 0.0,
627
+ "clip_ratio/low_mean": 0.001953125,
628
+ "clip_ratio/low_min": 0.001953125,
629
+ "clip_ratio/region_mean": 0.001953125,
630
+ "entropy": 1.7949985265731812,
631
+ "epoch": 1.4166666666666667,
632
+ "grad_norm": 4.808130264282227,
633
+ "kl": 0.17208479344844818,
634
+ "learning_rate": 1.7000000000000002e-06,
635
+ "loss": 0.277148574590683,
636
+ "step": 34,
637
+ "step_time": 0.22698771800014583
638
+ },
639
+ {
640
+ "clip_ratio/high_max": 0.00390625,
641
+ "clip_ratio/high_mean": 0.00390625,
642
+ "clip_ratio/low_mean": 0.0026041667442768812,
643
+ "clip_ratio/low_min": 0.0026041667442768812,
644
+ "clip_ratio/region_mean": 0.0065104165114462376,
645
+ "entropy": 1.6398433446884155,
646
+ "epoch": 1.4583333333333333,
647
+ "grad_norm": 3.2052321434020996,
648
+ "kl": 0.12288280576467514,
649
+ "learning_rate": 1.6000000000000001e-06,
650
+ "loss": 0.052221912890672684,
651
+ "step": 35,
652
+ "step_time": 0.22688573699997505
653
+ },
654
+ {
655
+ "clip_ratio/high_max": 0.0045572915114462376,
656
+ "clip_ratio/high_mean": 0.0045572915114462376,
657
+ "clip_ratio/low_mean": 0.0,
658
+ "clip_ratio/low_min": 0.0,
659
+ "clip_ratio/region_mean": 0.0045572915114462376,
660
+ "entropy": 1.7194310426712036,
661
+ "epoch": 1.5,
662
+ "grad_norm": 1.6972845792770386,
663
+ "kl": 0.09935706853866577,
664
+ "learning_rate": 1.5e-06,
665
+ "loss": -0.11925359815359116,
666
+ "step": 36,
667
+ "step_time": 0.2262224169999172
668
+ },
669
+ {
670
+ "clip_ratio/high_max": 0.0,
671
+ "clip_ratio/high_mean": 0.0,
672
+ "clip_ratio/low_mean": 0.0,
673
+ "clip_ratio/low_min": 0.0,
674
+ "clip_ratio/region_mean": 0.0,
675
+ "completions/clipped_ratio": 1.0,
676
+ "completions/max_length": 384.0,
677
+ "completions/max_terminated_length": 0.0,
678
+ "completions/mean_length": 384.0,
679
+ "completions/mean_terminated_length": 0.0,
680
+ "completions/min_length": 384.0,
681
+ "completions/min_terminated_length": 0.0,
682
+ "entropy": 1.8657582998275757,
683
+ "epoch": 1.5416666666666665,
684
+ "frac_reward_zero_std": 0.25,
685
+ "grad_norm": 1.8829715251922607,
686
+ "kl": 0.11106264591217041,
687
+ "learning_rate": 1.4000000000000001e-06,
688
+ "loss": -0.08045297861099243,
689
+ "num_tokens": 252640.0,
690
+ "reward": -0.17375001311302185,
691
+ "reward_std": 0.2488071471452713,
692
+ "rewards/GeneratorRewardFunction/mean": -0.17375001311302185,
693
+ "rewards/GeneratorRewardFunction/std": 0.2488071620464325,
694
+ "step": 37,
695
+ "step_time": 12.129672501999949
696
+ },
697
+ {
698
+ "clip_ratio/high_max": 0.001953125,
699
+ "clip_ratio/high_mean": 0.001953125,
700
+ "clip_ratio/low_mean": 0.0013020833721384406,
701
+ "clip_ratio/low_min": 0.0013020833721384406,
702
+ "clip_ratio/region_mean": 0.0032552082557231188,
703
+ "entropy": 0.8961874842643738,
704
+ "epoch": 1.5833333333333335,
705
+ "grad_norm": 3.2175376415252686,
706
+ "kl": 0.09107788652181625,
707
+ "learning_rate": 1.3e-06,
708
+ "loss": 0.1592753380537033,
709
+ "step": 38,
710
+ "step_time": 0.2210742439999649
711
+ },
712
+ {
713
+ "clip_ratio/high_max": 0.001953125,
714
+ "clip_ratio/high_mean": 0.001953125,
715
+ "clip_ratio/low_mean": 0.0006510416860692203,
716
+ "clip_ratio/low_min": 0.0006510416860692203,
717
+ "clip_ratio/region_mean": 0.0026041667442768812,
718
+ "entropy": 1.1073484420776367,
719
+ "epoch": 1.625,
720
+ "grad_norm": 2.0094873905181885,
721
+ "kl": 0.07970862090587616,
722
+ "learning_rate": 1.2000000000000002e-06,
723
+ "loss": 0.03532649576663971,
724
+ "step": 39,
725
+ "step_time": 0.2208535810000285
726
+ },
727
+ {
728
+ "clip_ratio/high_max": 0.0045572915114462376,
729
+ "clip_ratio/high_mean": 0.0045572915114462376,
730
+ "clip_ratio/low_mean": 0.0,
731
+ "clip_ratio/low_min": 0.0,
732
+ "clip_ratio/region_mean": 0.0045572915114462376,
733
+ "entropy": 1.5075187683105469,
734
+ "epoch": 1.6666666666666665,
735
+ "grad_norm": 1.7352609634399414,
736
+ "kl": 0.09652527421712875,
737
+ "learning_rate": 1.1e-06,
738
+ "loss": -0.114167220890522,
739
+ "step": 40,
740
+ "step_time": 0.2206852459999027
741
+ }
742
+ ],
743
+ "logging_steps": 1,
744
+ "max_steps": 50,
745
+ "num_input_tokens_seen": 252640,
746
+ "num_train_epochs": 3,
747
+ "save_steps": 10,
748
+ "stateful_callbacks": {
749
+ "TrainerControl": {
750
+ "args": {
751
+ "should_epoch_stop": false,
752
+ "should_evaluate": false,
753
+ "should_log": false,
754
+ "should_save": true,
755
+ "should_training_stop": false
756
+ },
757
+ "attributes": {}
758
+ }
759
+ },
760
+ "total_flos": 0.0,
761
+ "train_batch_size": 4,
762
+ "trial_name": null,
763
+ "trial_params": null
764
+ }
self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:31ec66b64f432daf7616434296713e432d134face96e308f2ebc175e2e26f025
3
+ size 7249
self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/chat_template.jinja ADDED
@@ -0,0 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {%- if tools %}
2
+ {{- '<|im_start|>system\n' }}
3
+ {%- if messages[0]['role'] == 'system' %}
4
+ {{- messages[0]['content'] }}
5
+ {%- else %}
6
+ {{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }}
7
+ {%- endif %}
8
+ {{- "\n\n# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
9
+ {%- for tool in tools %}
10
+ {{- "\n" }}
11
+ {{- tool | tojson }}
12
+ {%- endfor %}
13
+ {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
14
+ {%- else %}
15
+ {%- if messages[0]['role'] == 'system' %}
16
+ {{- '<|im_start|>system\n' + messages[0]['content'] + '<|im_end|>\n' }}
17
+ {%- else %}
18
+ {{- '<|im_start|>system\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\n' }}
19
+ {%- endif %}
20
+ {%- endif %}
21
+ {%- for message in messages %}
22
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) or (message.role == "assistant" and not message.tool_calls) %}
23
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
24
+ {%- elif message.role == "assistant" %}
25
+ {{- '<|im_start|>' + message.role }}
26
+ {%- if message.content %}
27
+ {{- '\n' + message.content }}
28
+ {%- endif %}
29
+ {%- for tool_call in message.tool_calls %}
30
+ {%- if tool_call.function is defined %}
31
+ {%- set tool_call = tool_call.function %}
32
+ {%- endif %}
33
+ {{- '\n<tool_call>\n{"name": "' }}
34
+ {{- tool_call.name }}
35
+ {{- '", "arguments": ' }}
36
+ {{- tool_call.arguments | tojson }}
37
+ {{- '}\n</tool_call>' }}
38
+ {%- endfor %}
39
+ {{- '<|im_end|>\n' }}
40
+ {%- elif message.role == "tool" %}
41
+ {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != "tool") %}
42
+ {{- '<|im_start|>user' }}
43
+ {%- endif %}
44
+ {{- '\n<tool_response>\n' }}
45
+ {{- message.content }}
46
+ {{- '\n</tool_response>' }}
47
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
48
+ {{- '<|im_end|>\n' }}
49
+ {%- endif %}
50
+ {%- endif %}
51
+ {%- endfor %}
52
+ {%- if add_generation_prompt %}
53
+ {{- '<|im_start|>assistant\n' }}
54
+ {%- endif %}
self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/config.json ADDED
@@ -0,0 +1,57 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Qwen2ForCausalLM"
4
+ ],
5
+ "attention_dropout": 0.0,
6
+ "bos_token_id": null,
7
+ "dtype": "float32",
8
+ "eos_token_id": 151645,
9
+ "hidden_act": "silu",
10
+ "hidden_size": 896,
11
+ "initializer_range": 0.02,
12
+ "intermediate_size": 4864,
13
+ "layer_types": [
14
+ "full_attention",
15
+ "full_attention",
16
+ "full_attention",
17
+ "full_attention",
18
+ "full_attention",
19
+ "full_attention",
20
+ "full_attention",
21
+ "full_attention",
22
+ "full_attention",
23
+ "full_attention",
24
+ "full_attention",
25
+ "full_attention",
26
+ "full_attention",
27
+ "full_attention",
28
+ "full_attention",
29
+ "full_attention",
30
+ "full_attention",
31
+ "full_attention",
32
+ "full_attention",
33
+ "full_attention",
34
+ "full_attention",
35
+ "full_attention",
36
+ "full_attention",
37
+ "full_attention"
38
+ ],
39
+ "max_position_embeddings": 32768,
40
+ "max_window_layers": 21,
41
+ "model_type": "qwen2",
42
+ "num_attention_heads": 14,
43
+ "num_hidden_layers": 24,
44
+ "num_key_value_heads": 2,
45
+ "pad_token_id": 151643,
46
+ "rms_norm_eps": 1e-06,
47
+ "rope_parameters": {
48
+ "rope_theta": 1000000.0,
49
+ "rope_type": "default"
50
+ },
51
+ "sliding_window": null,
52
+ "tie_word_embeddings": true,
53
+ "transformers_version": "5.6.2",
54
+ "use_cache": false,
55
+ "use_sliding_window": false,
56
+ "vocab_size": 151936
57
+ }
self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/generation_config.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_sample": true,
3
+ "eos_token_id": [
4
+ 151645,
5
+ 151643
6
+ ],
7
+ "pad_token_id": 151643,
8
+ "repetition_penalty": 1.1,
9
+ "temperature": 0.7,
10
+ "top_k": 20,
11
+ "top_p": 0.8,
12
+ "transformers_version": "5.6.2"
13
+ }
self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:419f71328f66d9eaa2ba1d9037f5033a97db1d93862f5eb35b2f04ea57340b82
3
+ size 1976163472
self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:114213b8482ef3e88d23c48f5e68f5306cee49c75c01f38181569ce6c5f53878
3
+ size 3952509771
self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7876773bbbd765b5b540a43519e4809b559e65cdfa3f4e9508024dc7702f2f6e
3
+ size 14645
self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3cf220ca534359fdb729d8e74ed3b0c609c54e6d591e1d2478f5521fc51fba05
3
+ size 1465
self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3fd169731d2cbde95e10bf356d66d5997fd885dd8dbb6fb4684da3f23b2585d8
3
+ size 11421892
self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/tokenizer_config.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "backend": "tokenizers",
4
+ "bos_token": null,
5
+ "clean_up_tokenization_spaces": false,
6
+ "eos_token": "<|im_end|>",
7
+ "errors": "replace",
8
+ "extra_special_tokens": [
9
+ "<|im_start|>",
10
+ "<|im_end|>",
11
+ "<|object_ref_start|>",
12
+ "<|object_ref_end|>",
13
+ "<|box_start|>",
14
+ "<|box_end|>",
15
+ "<|quad_start|>",
16
+ "<|quad_end|>",
17
+ "<|vision_start|>",
18
+ "<|vision_end|>",
19
+ "<|vision_pad|>",
20
+ "<|image_pad|>",
21
+ "<|video_pad|>"
22
+ ],
23
+ "is_local": false,
24
+ "local_files_only": false,
25
+ "model_max_length": 131072,
26
+ "pad_token": "<|endoftext|>",
27
+ "padding_side": "left",
28
+ "split_special_tokens": false,
29
+ "tokenizer_class": "Qwen2Tokenizer",
30
+ "truncation_side": "left",
31
+ "unk_token": null
32
+ }
self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/trainer_state.json ADDED
@@ -0,0 +1,953 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 2.0833333333333335,
6
+ "eval_steps": 500,
7
+ "global_step": 50,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "clip_ratio/high_max": 0.001953125,
14
+ "clip_ratio/high_mean": 0.001953125,
15
+ "clip_ratio/low_mean": 0.0,
16
+ "clip_ratio/low_min": 0.0,
17
+ "clip_ratio/region_mean": 0.001953125,
18
+ "completions/clipped_ratio": 1.0,
19
+ "completions/max_length": 384.0,
20
+ "completions/max_terminated_length": 0.0,
21
+ "completions/mean_length": 384.0,
22
+ "completions/mean_terminated_length": 0.0,
23
+ "completions/min_length": 384.0,
24
+ "completions/min_terminated_length": 0.0,
25
+ "entropy": 1.9362258911132812,
26
+ "epoch": 0.041666666666666664,
27
+ "frac_reward_zero_std": 0.0,
28
+ "grad_norm": 3.342789649963379,
29
+ "kl": 0.0005498419050127268,
30
+ "learning_rate": 5e-06,
31
+ "loss": 0.13995857536792755,
32
+ "num_tokens": 25244.0,
33
+ "reward": -0.4352343678474426,
34
+ "reward_std": 0.306624174118042,
35
+ "rewards/GeneratorRewardFunction/mean": -0.4352343678474426,
36
+ "rewards/GeneratorRewardFunction/std": 0.306624174118042,
37
+ "step": 1,
38
+ "step_time": 12.708289422999997
39
+ },
40
+ {
41
+ "clip_ratio/high_max": 0.0045572915114462376,
42
+ "clip_ratio/high_mean": 0.0045572915114462376,
43
+ "clip_ratio/low_mean": 0.0026041667442768812,
44
+ "clip_ratio/low_min": 0.0026041667442768812,
45
+ "clip_ratio/region_mean": 0.0071614584885537624,
46
+ "entropy": 1.2719587087631226,
47
+ "epoch": 0.08333333333333333,
48
+ "grad_norm": 2.8023905754089355,
49
+ "kl": 0.0014687416842207313,
50
+ "learning_rate": 4.9000000000000005e-06,
51
+ "loss": -0.0669216588139534,
52
+ "step": 2,
53
+ "step_time": 0.221013198999799
54
+ },
55
+ {
56
+ "clip_ratio/high_max": 0.010416666977107525,
57
+ "clip_ratio/high_mean": 0.010416666977107525,
58
+ "clip_ratio/low_mean": 0.014322916977107525,
59
+ "clip_ratio/low_min": 0.014322916977107525,
60
+ "clip_ratio/region_mean": 0.02473958395421505,
61
+ "entropy": 1.8711100816726685,
62
+ "epoch": 0.125,
63
+ "grad_norm": 2.356076717376709,
64
+ "kl": 0.004658207297325134,
65
+ "learning_rate": 4.800000000000001e-06,
66
+ "loss": 0.010263405740261078,
67
+ "step": 3,
68
+ "step_time": 0.21947432299998582
69
+ },
70
+ {
71
+ "clip_ratio/high_max": 0.02018229104578495,
72
+ "clip_ratio/high_mean": 0.02018229104578495,
73
+ "clip_ratio/low_mean": 0.009765625,
74
+ "clip_ratio/low_min": 0.009765625,
75
+ "clip_ratio/region_mean": 0.02994791604578495,
76
+ "entropy": 1.1866708993911743,
77
+ "epoch": 0.16666666666666666,
78
+ "grad_norm": 1.5606235265731812,
79
+ "kl": 0.005778420716524124,
80
+ "learning_rate": 4.7e-06,
81
+ "loss": -0.08204194158315659,
82
+ "step": 4,
83
+ "step_time": 0.21935116999998172
84
+ },
85
+ {
86
+ "clip_ratio/high_max": 0.0013020833721384406,
87
+ "clip_ratio/high_mean": 0.0013020833721384406,
88
+ "clip_ratio/low_mean": 0.0006510416860692203,
89
+ "clip_ratio/low_min": 0.0006510416860692203,
90
+ "clip_ratio/region_mean": 0.001953125,
91
+ "completions/clipped_ratio": 1.0,
92
+ "completions/max_length": 384.0,
93
+ "completions/max_terminated_length": 0.0,
94
+ "completions/mean_length": 384.0,
95
+ "completions/mean_terminated_length": 0.0,
96
+ "completions/min_length": 384.0,
97
+ "completions/min_terminated_length": 0.0,
98
+ "entropy": 1.1421619653701782,
99
+ "epoch": 0.20833333333333334,
100
+ "frac_reward_zero_std": 0.0,
101
+ "grad_norm": 3.0044524669647217,
102
+ "kl": 0.017580831423401833,
103
+ "learning_rate": 4.600000000000001e-06,
104
+ "loss": 0.022549070417881012,
105
+ "num_tokens": 50440.0,
106
+ "reward": -0.3424999713897705,
107
+ "reward_std": 0.3177734613418579,
108
+ "rewards/GeneratorRewardFunction/mean": -0.3424999713897705,
109
+ "rewards/GeneratorRewardFunction/std": 0.3177734911441803,
110
+ "step": 5,
111
+ "step_time": 12.087754471000153
112
+ },
113
+ {
114
+ "clip_ratio/high_max": 0.008463541977107525,
115
+ "clip_ratio/high_mean": 0.008463541977107525,
116
+ "clip_ratio/low_mean": 0.0052083334885537624,
117
+ "clip_ratio/low_min": 0.0052083334885537624,
118
+ "clip_ratio/region_mean": 0.013671875,
119
+ "entropy": 1.110270619392395,
120
+ "epoch": 0.25,
121
+ "grad_norm": 2.2674970626831055,
122
+ "kl": 0.01950421743094921,
123
+ "learning_rate": 4.5e-06,
124
+ "loss": 0.00025658251252025366,
125
+ "step": 6,
126
+ "step_time": 0.2263194110000768
127
+ },
128
+ {
129
+ "clip_ratio/high_max": 0.01171875,
130
+ "clip_ratio/high_mean": 0.01171875,
131
+ "clip_ratio/low_mean": 0.008463541977107525,
132
+ "clip_ratio/low_min": 0.008463541977107525,
133
+ "clip_ratio/region_mean": 0.02018229104578495,
134
+ "entropy": 1.1433427333831787,
135
+ "epoch": 0.2916666666666667,
136
+ "grad_norm": 2.097913980484009,
137
+ "kl": 0.050842806696891785,
138
+ "learning_rate": 4.4e-06,
139
+ "loss": -0.02153957635164261,
140
+ "step": 7,
141
+ "step_time": 0.2258150579998528
142
+ },
143
+ {
144
+ "clip_ratio/high_max": 0.021484375,
145
+ "clip_ratio/high_mean": 0.021484375,
146
+ "clip_ratio/low_mean": 0.013020833022892475,
147
+ "clip_ratio/low_min": 0.013020833022892475,
148
+ "clip_ratio/region_mean": 0.0345052070915699,
149
+ "entropy": 1.5583876371383667,
150
+ "epoch": 0.3333333333333333,
151
+ "grad_norm": 2.776447057723999,
152
+ "kl": 0.06861245632171631,
153
+ "learning_rate": 4.3e-06,
154
+ "loss": 0.0012576957233250141,
155
+ "step": 8,
156
+ "step_time": 0.22490879599990876
157
+ },
158
+ {
159
+ "clip_ratio/high_max": 0.0,
160
+ "clip_ratio/high_mean": 0.0,
161
+ "clip_ratio/low_mean": 0.0,
162
+ "clip_ratio/low_min": 0.0,
163
+ "clip_ratio/region_mean": 0.0,
164
+ "completions/clipped_ratio": 1.0,
165
+ "completions/max_length": 384.0,
166
+ "completions/max_terminated_length": 0.0,
167
+ "completions/mean_length": 384.0,
168
+ "completions/mean_terminated_length": 0.0,
169
+ "completions/min_length": 384.0,
170
+ "completions/min_terminated_length": 0.0,
171
+ "entropy": 1.256606101989746,
172
+ "epoch": 0.375,
173
+ "frac_reward_zero_std": 0.0,
174
+ "grad_norm": 3.1693882942199707,
175
+ "kl": 0.034918785095214844,
176
+ "learning_rate": 4.2000000000000004e-06,
177
+ "loss": 0.06234124302864075,
178
+ "num_tokens": 75884.0,
179
+ "reward": -0.3112500011920929,
180
+ "reward_std": 0.3181116282939911,
181
+ "rewards/GeneratorRewardFunction/mean": -0.3112500011920929,
182
+ "rewards/GeneratorRewardFunction/std": 0.3181115984916687,
183
+ "step": 9,
184
+ "step_time": 12.085582495999915
185
+ },
186
+ {
187
+ "clip_ratio/high_max": 0.00390625,
188
+ "clip_ratio/high_mean": 0.00390625,
189
+ "clip_ratio/low_mean": 0.001953125,
190
+ "clip_ratio/low_min": 0.001953125,
191
+ "clip_ratio/region_mean": 0.005859375,
192
+ "entropy": 1.150619387626648,
193
+ "epoch": 0.4166666666666667,
194
+ "grad_norm": 2.827258348464966,
195
+ "kl": 0.026281872764229774,
196
+ "learning_rate": 4.1e-06,
197
+ "loss": -0.04307486489415169,
198
+ "step": 10,
199
+ "step_time": 0.22665007600016907
200
+ },
201
+ {
202
+ "clip_ratio/high_max": 0.013020833022892475,
203
+ "clip_ratio/high_mean": 0.013020833022892475,
204
+ "clip_ratio/low_mean": 0.012369791977107525,
205
+ "clip_ratio/low_min": 0.012369791977107525,
206
+ "clip_ratio/region_mean": 0.025390625,
207
+ "entropy": 1.353300929069519,
208
+ "epoch": 0.4583333333333333,
209
+ "grad_norm": 1.517728328704834,
210
+ "kl": 0.043695781379938126,
211
+ "learning_rate": 4.000000000000001e-06,
212
+ "loss": 0.0004213004431221634,
213
+ "step": 11,
214
+ "step_time": 0.2250163410001278
215
+ },
216
+ {
217
+ "clip_ratio/high_max": 0.015625,
218
+ "clip_ratio/high_mean": 0.015625,
219
+ "clip_ratio/low_mean": 0.01953125,
220
+ "clip_ratio/low_min": 0.01953125,
221
+ "clip_ratio/region_mean": 0.03515625,
222
+ "entropy": 1.5499415397644043,
223
+ "epoch": 0.5,
224
+ "grad_norm": 2.2769711017608643,
225
+ "kl": 0.04331633448600769,
226
+ "learning_rate": 3.900000000000001e-06,
227
+ "loss": -0.019014613702893257,
228
+ "step": 12,
229
+ "step_time": 0.2254276790001768
230
+ },
231
+ {
232
+ "clip_ratio/high_max": 0.0026041667442768812,
233
+ "clip_ratio/high_mean": 0.0026041667442768812,
234
+ "clip_ratio/low_mean": 0.0,
235
+ "clip_ratio/low_min": 0.0,
236
+ "clip_ratio/region_mean": 0.0026041667442768812,
237
+ "completions/clipped_ratio": 1.0,
238
+ "completions/max_length": 384.0,
239
+ "completions/max_terminated_length": 0.0,
240
+ "completions/mean_length": 384.0,
241
+ "completions/mean_terminated_length": 0.0,
242
+ "completions/min_length": 384.0,
243
+ "completions/min_terminated_length": 0.0,
244
+ "entropy": 0.9478046298027039,
245
+ "epoch": 0.5416666666666666,
246
+ "frac_reward_zero_std": 0.0,
247
+ "grad_norm": 1.8451720476150513,
248
+ "kl": 0.042272429913282394,
249
+ "learning_rate": 3.8000000000000005e-06,
250
+ "loss": -0.06487596780061722,
251
+ "num_tokens": 101112.0,
252
+ "reward": -0.3031249940395355,
253
+ "reward_std": 0.3281609117984772,
254
+ "rewards/GeneratorRewardFunction/mean": -0.3031249940395355,
255
+ "rewards/GeneratorRewardFunction/std": 0.3281609117984772,
256
+ "step": 13,
257
+ "step_time": 12.174987709000106
258
+ },
259
+ {
260
+ "clip_ratio/high_max": 0.0065104165114462376,
261
+ "clip_ratio/high_mean": 0.0065104165114462376,
262
+ "clip_ratio/low_mean": 0.0,
263
+ "clip_ratio/low_min": 0.0,
264
+ "clip_ratio/region_mean": 0.0065104165114462376,
265
+ "entropy": 1.186617374420166,
266
+ "epoch": 0.5833333333333334,
267
+ "grad_norm": 1.483729600906372,
268
+ "kl": 0.05555545911192894,
269
+ "learning_rate": 3.7e-06,
270
+ "loss": -0.12233645468950272,
271
+ "step": 14,
272
+ "step_time": 0.21759630599990487
273
+ },
274
+ {
275
+ "clip_ratio/high_max": 0.0,
276
+ "clip_ratio/high_mean": 0.0,
277
+ "clip_ratio/low_mean": 0.008463541977107525,
278
+ "clip_ratio/low_min": 0.008463541977107525,
279
+ "clip_ratio/region_mean": 0.008463541977107525,
280
+ "entropy": 0.8805699348449707,
281
+ "epoch": 0.625,
282
+ "grad_norm": 3.184286594390869,
283
+ "kl": 0.05239926651120186,
284
+ "learning_rate": 3.6000000000000003e-06,
285
+ "loss": 0.27255433797836304,
286
+ "step": 15,
287
+ "step_time": 0.21799612400013757
288
+ },
289
+ {
290
+ "clip_ratio/high_max": 0.014322916977107525,
291
+ "clip_ratio/high_mean": 0.014322916977107525,
292
+ "clip_ratio/low_mean": 0.0032552082557231188,
293
+ "clip_ratio/low_min": 0.0032552082557231188,
294
+ "clip_ratio/region_mean": 0.017578125,
295
+ "entropy": 1.7632579803466797,
296
+ "epoch": 0.6666666666666666,
297
+ "grad_norm": 2.479675531387329,
298
+ "kl": 0.09853244572877884,
299
+ "learning_rate": 3.5e-06,
300
+ "loss": -0.08373244851827621,
301
+ "step": 16,
302
+ "step_time": 0.21777955800007476
303
+ },
304
+ {
305
+ "clip_ratio/high_max": 0.0,
306
+ "clip_ratio/high_mean": 0.0,
307
+ "clip_ratio/low_mean": 0.0,
308
+ "clip_ratio/low_min": 0.0,
309
+ "clip_ratio/region_mean": 0.0,
310
+ "completions/clipped_ratio": 1.0,
311
+ "completions/max_length": 384.0,
312
+ "completions/max_terminated_length": 0.0,
313
+ "completions/mean_length": 384.0,
314
+ "completions/mean_terminated_length": 0.0,
315
+ "completions/min_length": 384.0,
316
+ "completions/min_terminated_length": 0.0,
317
+ "entropy": 1.2678290605545044,
318
+ "epoch": 0.7083333333333334,
319
+ "frac_reward_zero_std": 0.0,
320
+ "grad_norm": 3.134302854537964,
321
+ "kl": 0.06070971116423607,
322
+ "learning_rate": 3.4000000000000005e-06,
323
+ "loss": -0.052310068160295486,
324
+ "num_tokens": 126436.0,
325
+ "reward": -0.2212499976158142,
326
+ "reward_std": 0.27250993251800537,
327
+ "rewards/GeneratorRewardFunction/mean": -0.2212499976158142,
328
+ "rewards/GeneratorRewardFunction/std": 0.27250996232032776,
329
+ "step": 17,
330
+ "step_time": 12.04144509899993
331
+ },
332
+ {
333
+ "clip_ratio/high_max": 0.0,
334
+ "clip_ratio/high_mean": 0.0,
335
+ "clip_ratio/low_mean": 0.0006510416860692203,
336
+ "clip_ratio/low_min": 0.0006510416860692203,
337
+ "clip_ratio/region_mean": 0.0006510416860692203,
338
+ "entropy": 1.7929935455322266,
339
+ "epoch": 0.75,
340
+ "grad_norm": 3.1320412158966064,
341
+ "kl": 0.062457334250211716,
342
+ "learning_rate": 3.3000000000000006e-06,
343
+ "loss": -0.03248615190386772,
344
+ "step": 18,
345
+ "step_time": 0.2239023360000374
346
+ },
347
+ {
348
+ "clip_ratio/high_max": 0.01627604104578495,
349
+ "clip_ratio/high_mean": 0.01627604104578495,
350
+ "clip_ratio/low_mean": 0.0,
351
+ "clip_ratio/low_min": 0.0,
352
+ "clip_ratio/region_mean": 0.01627604104578495,
353
+ "entropy": 1.145819067955017,
354
+ "epoch": 0.7916666666666666,
355
+ "grad_norm": 1.5332239866256714,
356
+ "kl": 0.07924208045005798,
357
+ "learning_rate": 3.2000000000000003e-06,
358
+ "loss": -0.08211782574653625,
359
+ "step": 19,
360
+ "step_time": 0.22507186799998635
361
+ },
362
+ {
363
+ "clip_ratio/high_max": 0.0,
364
+ "clip_ratio/high_mean": 0.0,
365
+ "clip_ratio/low_mean": 0.012369791977107525,
366
+ "clip_ratio/low_min": 0.012369791977107525,
367
+ "clip_ratio/region_mean": 0.012369791977107525,
368
+ "entropy": 0.9164741635322571,
369
+ "epoch": 0.8333333333333334,
370
+ "grad_norm": 2.2384536266326904,
371
+ "kl": 0.06750863045454025,
372
+ "learning_rate": 3.1000000000000004e-06,
373
+ "loss": 0.16943609714508057,
374
+ "step": 20,
375
+ "step_time": 0.22430561300006957
376
+ },
377
+ {
378
+ "clip_ratio/high_max": 0.0013020833721384406,
379
+ "clip_ratio/high_mean": 0.0013020833721384406,
380
+ "clip_ratio/low_mean": 0.0,
381
+ "clip_ratio/low_min": 0.0,
382
+ "clip_ratio/region_mean": 0.0013020833721384406,
383
+ "completions/clipped_ratio": 1.0,
384
+ "completions/max_length": 384.0,
385
+ "completions/max_terminated_length": 0.0,
386
+ "completions/mean_length": 384.0,
387
+ "completions/mean_terminated_length": 0.0,
388
+ "completions/min_length": 384.0,
389
+ "completions/min_terminated_length": 0.0,
390
+ "entropy": 1.862565040588379,
391
+ "epoch": 0.875,
392
+ "frac_reward_zero_std": 0.0,
393
+ "grad_norm": 2.588430643081665,
394
+ "kl": 0.04901580885052681,
395
+ "learning_rate": 3e-06,
396
+ "loss": -0.12625738978385925,
397
+ "num_tokens": 151596.0,
398
+ "reward": -0.2653124928474426,
399
+ "reward_std": 0.2914073169231415,
400
+ "rewards/GeneratorRewardFunction/mean": -0.2653124928474426,
401
+ "rewards/GeneratorRewardFunction/std": 0.2914073169231415,
402
+ "step": 21,
403
+ "step_time": 12.2114765299998
404
+ },
405
+ {
406
+ "clip_ratio/high_max": 0.0032552082557231188,
407
+ "clip_ratio/high_mean": 0.0032552082557231188,
408
+ "clip_ratio/low_mean": 0.0,
409
+ "clip_ratio/low_min": 0.0,
410
+ "clip_ratio/region_mean": 0.0032552082557231188,
411
+ "entropy": 1.3516654968261719,
412
+ "epoch": 0.9166666666666666,
413
+ "grad_norm": 2.4181604385375977,
414
+ "kl": 0.061538487672805786,
415
+ "learning_rate": 2.9e-06,
416
+ "loss": 0.0242471843957901,
417
+ "step": 22,
418
+ "step_time": 0.21940968300009445
419
+ },
420
+ {
421
+ "clip_ratio/high_max": 0.0065104165114462376,
422
+ "clip_ratio/high_mean": 0.0065104165114462376,
423
+ "clip_ratio/low_mean": 0.0013020833721384406,
424
+ "clip_ratio/low_min": 0.0013020833721384406,
425
+ "clip_ratio/region_mean": 0.0078125,
426
+ "entropy": 1.3538661003112793,
427
+ "epoch": 0.9583333333333334,
428
+ "grad_norm": 2.905212879180908,
429
+ "kl": 0.07095103710889816,
430
+ "learning_rate": 2.8000000000000003e-06,
431
+ "loss": 0.08377016335725784,
432
+ "step": 23,
433
+ "step_time": 0.22010638500000823
434
+ },
435
+ {
436
+ "clip_ratio/high_max": 0.013020833022892475,
437
+ "clip_ratio/high_mean": 0.013020833022892475,
438
+ "clip_ratio/low_mean": 0.001953125,
439
+ "clip_ratio/low_min": 0.001953125,
440
+ "clip_ratio/region_mean": 0.014973958022892475,
441
+ "entropy": 1.703626275062561,
442
+ "epoch": 1.0,
443
+ "grad_norm": 2.835925340652466,
444
+ "kl": 0.0820431113243103,
445
+ "learning_rate": 2.7000000000000004e-06,
446
+ "loss": 0.018686100840568542,
447
+ "step": 24,
448
+ "step_time": 0.21947678399988035
449
+ },
450
+ {
451
+ "clip_ratio/high_max": 0.0,
452
+ "clip_ratio/high_mean": 0.0,
453
+ "clip_ratio/low_mean": 0.0,
454
+ "clip_ratio/low_min": 0.0,
455
+ "clip_ratio/region_mean": 0.0,
456
+ "completions/clipped_ratio": 1.0,
457
+ "completions/max_length": 384.0,
458
+ "completions/max_terminated_length": 0.0,
459
+ "completions/mean_length": 384.0,
460
+ "completions/mean_terminated_length": 0.0,
461
+ "completions/min_length": 384.0,
462
+ "completions/min_terminated_length": 0.0,
463
+ "entropy": 1.6442562341690063,
464
+ "epoch": 1.0416666666666667,
465
+ "frac_reward_zero_std": 0.0,
466
+ "grad_norm": 2.382233142852783,
467
+ "kl": 0.0744522213935852,
468
+ "learning_rate": 2.6e-06,
469
+ "loss": 0.060507725924253464,
470
+ "num_tokens": 177212.0,
471
+ "reward": -0.2524999976158142,
472
+ "reward_std": 0.28701913356781006,
473
+ "rewards/GeneratorRewardFunction/mean": -0.2524999976158142,
474
+ "rewards/GeneratorRewardFunction/std": 0.28701916337013245,
475
+ "step": 25,
476
+ "step_time": 12.102920622000056
477
+ },
478
+ {
479
+ "clip_ratio/high_max": 0.001953125,
480
+ "clip_ratio/high_mean": 0.001953125,
481
+ "clip_ratio/low_mean": 0.0006510416860692203,
482
+ "clip_ratio/low_min": 0.0006510416860692203,
483
+ "clip_ratio/region_mean": 0.0026041667442768812,
484
+ "entropy": 1.1888453960418701,
485
+ "epoch": 1.0833333333333333,
486
+ "grad_norm": 2.2921271324157715,
487
+ "kl": 0.09428787231445312,
488
+ "learning_rate": 2.5e-06,
489
+ "loss": 0.06040751934051514,
490
+ "step": 26,
491
+ "step_time": 0.22665847199982636
492
+ },
493
+ {
494
+ "clip_ratio/high_max": 0.011067708022892475,
495
+ "clip_ratio/high_mean": 0.011067708022892475,
496
+ "clip_ratio/low_mean": 0.0,
497
+ "clip_ratio/low_min": 0.0,
498
+ "clip_ratio/region_mean": 0.011067708022892475,
499
+ "entropy": 1.1056007146835327,
500
+ "epoch": 1.125,
501
+ "grad_norm": 1.509630560874939,
502
+ "kl": 0.0742887407541275,
503
+ "learning_rate": 2.4000000000000003e-06,
504
+ "loss": -0.12057904154062271,
505
+ "step": 27,
506
+ "step_time": 0.22739392499988753
507
+ },
508
+ {
509
+ "clip_ratio/high_max": 0.005859375,
510
+ "clip_ratio/high_mean": 0.005859375,
511
+ "clip_ratio/low_mean": 0.0026041667442768812,
512
+ "clip_ratio/low_min": 0.0026041667442768812,
513
+ "clip_ratio/region_mean": 0.008463541977107525,
514
+ "entropy": 0.8066513538360596,
515
+ "epoch": 1.1666666666666667,
516
+ "grad_norm": 2.760502576828003,
517
+ "kl": 0.10300976783037186,
518
+ "learning_rate": 2.3000000000000004e-06,
519
+ "loss": 0.00034140495699830353,
520
+ "step": 28,
521
+ "step_time": 0.22687044499980402
522
+ },
523
+ {
524
+ "clip_ratio/high_max": 0.0,
525
+ "clip_ratio/high_mean": 0.0,
526
+ "clip_ratio/low_mean": 0.0013020833721384406,
527
+ "clip_ratio/low_min": 0.0013020833721384406,
528
+ "clip_ratio/region_mean": 0.0013020833721384406,
529
+ "completions/clipped_ratio": 1.0,
530
+ "completions/max_length": 384.0,
531
+ "completions/max_terminated_length": 0.0,
532
+ "completions/mean_length": 384.0,
533
+ "completions/mean_terminated_length": 0.0,
534
+ "completions/min_length": 384.0,
535
+ "completions/min_terminated_length": 0.0,
536
+ "entropy": 1.6604108810424805,
537
+ "epoch": 1.2083333333333333,
538
+ "frac_reward_zero_std": 0.0,
539
+ "grad_norm": 1.5455973148345947,
540
+ "kl": 0.09443795680999756,
541
+ "learning_rate": 2.2e-06,
542
+ "loss": 0.06687173247337341,
543
+ "num_tokens": 202200.0,
544
+ "reward": -0.12000000476837158,
545
+ "reward_std": 0.20759736001491547,
546
+ "rewards/GeneratorRewardFunction/mean": -0.12000000476837158,
547
+ "rewards/GeneratorRewardFunction/std": 0.20759736001491547,
548
+ "step": 29,
549
+ "step_time": 12.208307090000062
550
+ },
551
+ {
552
+ "clip_ratio/high_max": 0.0006510416860692203,
553
+ "clip_ratio/high_mean": 0.0006510416860692203,
554
+ "clip_ratio/low_mean": 0.0006510416860692203,
555
+ "clip_ratio/low_min": 0.0006510416860692203,
556
+ "clip_ratio/region_mean": 0.0013020833721384406,
557
+ "entropy": 1.0410054922103882,
558
+ "epoch": 1.25,
559
+ "grad_norm": 3.1164426803588867,
560
+ "kl": 0.07589004188776016,
561
+ "learning_rate": 2.1000000000000002e-06,
562
+ "loss": -0.12521938979625702,
563
+ "step": 30,
564
+ "step_time": 0.22144749400013097
565
+ },
566
+ {
567
+ "clip_ratio/high_max": 0.00390625,
568
+ "clip_ratio/high_mean": 0.00390625,
569
+ "clip_ratio/low_mean": 0.0,
570
+ "clip_ratio/low_min": 0.0,
571
+ "clip_ratio/region_mean": 0.00390625,
572
+ "entropy": 1.3995355367660522,
573
+ "epoch": 1.2916666666666667,
574
+ "grad_norm": 2.3790829181671143,
575
+ "kl": 0.10240024328231812,
576
+ "learning_rate": 2.0000000000000003e-06,
577
+ "loss": -0.006538359913975,
578
+ "step": 31,
579
+ "step_time": 0.22054820099992867
580
+ },
581
+ {
582
+ "clip_ratio/high_max": 0.0032552082557231188,
583
+ "clip_ratio/high_mean": 0.0032552082557231188,
584
+ "clip_ratio/low_mean": 0.0071614584885537624,
585
+ "clip_ratio/low_min": 0.0071614584885537624,
586
+ "clip_ratio/region_mean": 0.010416666977107525,
587
+ "entropy": 1.3301101922988892,
588
+ "epoch": 1.3333333333333333,
589
+ "grad_norm": 1.3749467134475708,
590
+ "kl": 0.09631212800741196,
591
+ "learning_rate": 1.9000000000000002e-06,
592
+ "loss": 0.06641887873411179,
593
+ "step": 32,
594
+ "step_time": 0.21976359399991452
595
+ },
596
+ {
597
+ "clip_ratio/high_max": 0.0,
598
+ "clip_ratio/high_mean": 0.0,
599
+ "clip_ratio/low_mean": 0.0,
600
+ "clip_ratio/low_min": 0.0,
601
+ "clip_ratio/region_mean": 0.0,
602
+ "completions/clipped_ratio": 1.0,
603
+ "completions/max_length": 384.0,
604
+ "completions/max_terminated_length": 0.0,
605
+ "completions/mean_length": 384.0,
606
+ "completions/mean_terminated_length": 0.0,
607
+ "completions/min_length": 384.0,
608
+ "completions/min_terminated_length": 0.0,
609
+ "entropy": 1.4248062372207642,
610
+ "epoch": 1.375,
611
+ "frac_reward_zero_std": 0.0,
612
+ "grad_norm": 2.76702618598938,
613
+ "kl": 0.08123808354139328,
614
+ "learning_rate": 1.8000000000000001e-06,
615
+ "loss": -0.20796377956867218,
616
+ "num_tokens": 227556.0,
617
+ "reward": -0.1759375035762787,
618
+ "reward_std": 0.23178452253341675,
619
+ "rewards/GeneratorRewardFunction/mean": -0.1759375035762787,
620
+ "rewards/GeneratorRewardFunction/std": 0.23178452253341675,
621
+ "step": 33,
622
+ "step_time": 12.118321306000098
623
+ },
624
+ {
625
+ "clip_ratio/high_max": 0.0,
626
+ "clip_ratio/high_mean": 0.0,
627
+ "clip_ratio/low_mean": 0.001953125,
628
+ "clip_ratio/low_min": 0.001953125,
629
+ "clip_ratio/region_mean": 0.001953125,
630
+ "entropy": 1.7949985265731812,
631
+ "epoch": 1.4166666666666667,
632
+ "grad_norm": 4.808130264282227,
633
+ "kl": 0.17208479344844818,
634
+ "learning_rate": 1.7000000000000002e-06,
635
+ "loss": 0.277148574590683,
636
+ "step": 34,
637
+ "step_time": 0.22698771800014583
638
+ },
639
+ {
640
+ "clip_ratio/high_max": 0.00390625,
641
+ "clip_ratio/high_mean": 0.00390625,
642
+ "clip_ratio/low_mean": 0.0026041667442768812,
643
+ "clip_ratio/low_min": 0.0026041667442768812,
644
+ "clip_ratio/region_mean": 0.0065104165114462376,
645
+ "entropy": 1.6398433446884155,
646
+ "epoch": 1.4583333333333333,
647
+ "grad_norm": 3.2052321434020996,
648
+ "kl": 0.12288280576467514,
649
+ "learning_rate": 1.6000000000000001e-06,
650
+ "loss": 0.052221912890672684,
651
+ "step": 35,
652
+ "step_time": 0.22688573699997505
653
+ },
654
+ {
655
+ "clip_ratio/high_max": 0.0045572915114462376,
656
+ "clip_ratio/high_mean": 0.0045572915114462376,
657
+ "clip_ratio/low_mean": 0.0,
658
+ "clip_ratio/low_min": 0.0,
659
+ "clip_ratio/region_mean": 0.0045572915114462376,
660
+ "entropy": 1.7194310426712036,
661
+ "epoch": 1.5,
662
+ "grad_norm": 1.6972845792770386,
663
+ "kl": 0.09935706853866577,
664
+ "learning_rate": 1.5e-06,
665
+ "loss": -0.11925359815359116,
666
+ "step": 36,
667
+ "step_time": 0.2262224169999172
668
+ },
669
+ {
670
+ "clip_ratio/high_max": 0.0,
671
+ "clip_ratio/high_mean": 0.0,
672
+ "clip_ratio/low_mean": 0.0,
673
+ "clip_ratio/low_min": 0.0,
674
+ "clip_ratio/region_mean": 0.0,
675
+ "completions/clipped_ratio": 1.0,
676
+ "completions/max_length": 384.0,
677
+ "completions/max_terminated_length": 0.0,
678
+ "completions/mean_length": 384.0,
679
+ "completions/mean_terminated_length": 0.0,
680
+ "completions/min_length": 384.0,
681
+ "completions/min_terminated_length": 0.0,
682
+ "entropy": 1.8657582998275757,
683
+ "epoch": 1.5416666666666665,
684
+ "frac_reward_zero_std": 0.25,
685
+ "grad_norm": 1.8829715251922607,
686
+ "kl": 0.11106264591217041,
687
+ "learning_rate": 1.4000000000000001e-06,
688
+ "loss": -0.08045297861099243,
689
+ "num_tokens": 252640.0,
690
+ "reward": -0.17375001311302185,
691
+ "reward_std": 0.2488071471452713,
692
+ "rewards/GeneratorRewardFunction/mean": -0.17375001311302185,
693
+ "rewards/GeneratorRewardFunction/std": 0.2488071620464325,
694
+ "step": 37,
695
+ "step_time": 12.129672501999949
696
+ },
697
+ {
698
+ "clip_ratio/high_max": 0.001953125,
699
+ "clip_ratio/high_mean": 0.001953125,
700
+ "clip_ratio/low_mean": 0.0013020833721384406,
701
+ "clip_ratio/low_min": 0.0013020833721384406,
702
+ "clip_ratio/region_mean": 0.0032552082557231188,
703
+ "entropy": 0.8961874842643738,
704
+ "epoch": 1.5833333333333335,
705
+ "grad_norm": 3.2175376415252686,
706
+ "kl": 0.09107788652181625,
707
+ "learning_rate": 1.3e-06,
708
+ "loss": 0.1592753380537033,
709
+ "step": 38,
710
+ "step_time": 0.2210742439999649
711
+ },
712
+ {
713
+ "clip_ratio/high_max": 0.001953125,
714
+ "clip_ratio/high_mean": 0.001953125,
715
+ "clip_ratio/low_mean": 0.0006510416860692203,
716
+ "clip_ratio/low_min": 0.0006510416860692203,
717
+ "clip_ratio/region_mean": 0.0026041667442768812,
718
+ "entropy": 1.1073484420776367,
719
+ "epoch": 1.625,
720
+ "grad_norm": 2.0094873905181885,
721
+ "kl": 0.07970862090587616,
722
+ "learning_rate": 1.2000000000000002e-06,
723
+ "loss": 0.03532649576663971,
724
+ "step": 39,
725
+ "step_time": 0.2208535810000285
726
+ },
727
+ {
728
+ "clip_ratio/high_max": 0.0045572915114462376,
729
+ "clip_ratio/high_mean": 0.0045572915114462376,
730
+ "clip_ratio/low_mean": 0.0,
731
+ "clip_ratio/low_min": 0.0,
732
+ "clip_ratio/region_mean": 0.0045572915114462376,
733
+ "entropy": 1.5075187683105469,
734
+ "epoch": 1.6666666666666665,
735
+ "grad_norm": 1.7352609634399414,
736
+ "kl": 0.09652527421712875,
737
+ "learning_rate": 1.1e-06,
738
+ "loss": -0.114167220890522,
739
+ "step": 40,
740
+ "step_time": 0.2206852459999027
741
+ },
742
+ {
743
+ "clip_ratio/high_max": 0.001953125,
744
+ "clip_ratio/high_mean": 0.001953125,
745
+ "clip_ratio/low_mean": 0.0,
746
+ "clip_ratio/low_min": 0.0,
747
+ "clip_ratio/region_mean": 0.001953125,
748
+ "completions/clipped_ratio": 1.0,
749
+ "completions/max_length": 384.0,
750
+ "completions/max_terminated_length": 0.0,
751
+ "completions/mean_length": 384.0,
752
+ "completions/mean_terminated_length": 0.0,
753
+ "completions/min_length": 384.0,
754
+ "completions/min_terminated_length": 0.0,
755
+ "entropy": 1.53101646900177,
756
+ "epoch": 1.7083333333333335,
757
+ "frac_reward_zero_std": 0.0,
758
+ "grad_norm": 2.1645283699035645,
759
+ "kl": 0.12367743253707886,
760
+ "learning_rate": 1.0000000000000002e-06,
761
+ "loss": -0.07645408064126968,
762
+ "num_tokens": 277896.0,
763
+ "reward": -0.11020445823669434,
764
+ "reward_std": 0.3714827597141266,
765
+ "rewards/GeneratorRewardFunction/mean": -0.11020445823669434,
766
+ "rewards/GeneratorRewardFunction/std": 0.37148281931877136,
767
+ "step": 41,
768
+ "step_time": 13.437096242999814
769
+ },
770
+ {
771
+ "clip_ratio/high_max": 0.0013020833721384406,
772
+ "clip_ratio/high_mean": 0.0013020833721384406,
773
+ "clip_ratio/low_mean": 0.0026041667442768812,
774
+ "clip_ratio/low_min": 0.0026041667442768812,
775
+ "clip_ratio/region_mean": 0.00390625,
776
+ "entropy": 1.8844527006149292,
777
+ "epoch": 1.75,
778
+ "grad_norm": 4.297327041625977,
779
+ "kl": 0.11400499194860458,
780
+ "learning_rate": 9.000000000000001e-07,
781
+ "loss": 0.1845421940088272,
782
+ "step": 42,
783
+ "step_time": 0.22780594799996834
784
+ },
785
+ {
786
+ "clip_ratio/high_max": 0.0006510416860692203,
787
+ "clip_ratio/high_mean": 0.0006510416860692203,
788
+ "clip_ratio/low_mean": 0.00390625,
789
+ "clip_ratio/low_min": 0.00390625,
790
+ "clip_ratio/region_mean": 0.0045572915114462376,
791
+ "entropy": 1.5398229360580444,
792
+ "epoch": 1.7916666666666665,
793
+ "grad_norm": 2.3809876441955566,
794
+ "kl": 0.09886562079191208,
795
+ "learning_rate": 8.000000000000001e-07,
796
+ "loss": 0.0956474170088768,
797
+ "step": 43,
798
+ "step_time": 0.22696597999993173
799
+ },
800
+ {
801
+ "clip_ratio/high_max": 0.001953125,
802
+ "clip_ratio/high_mean": 0.001953125,
803
+ "clip_ratio/low_mean": 0.0,
804
+ "clip_ratio/low_min": 0.0,
805
+ "clip_ratio/region_mean": 0.001953125,
806
+ "entropy": 1.144031047821045,
807
+ "epoch": 1.8333333333333335,
808
+ "grad_norm": 2.3327653408050537,
809
+ "kl": 0.12570063769817352,
810
+ "learning_rate": 7.000000000000001e-07,
811
+ "loss": -0.20216596126556396,
812
+ "step": 44,
813
+ "step_time": 0.22737932700010788
814
+ },
815
+ {
816
+ "clip_ratio/high_max": 0.0013020833721384406,
817
+ "clip_ratio/high_mean": 0.0013020833721384406,
818
+ "clip_ratio/low_mean": 0.0006510416860692203,
819
+ "clip_ratio/low_min": 0.0006510416860692203,
820
+ "clip_ratio/region_mean": 0.001953125,
821
+ "completions/clipped_ratio": 1.0,
822
+ "completions/max_length": 384.0,
823
+ "completions/max_terminated_length": 0.0,
824
+ "completions/mean_length": 384.0,
825
+ "completions/mean_terminated_length": 0.0,
826
+ "completions/min_length": 384.0,
827
+ "completions/min_terminated_length": 0.0,
828
+ "entropy": 1.589156150817871,
829
+ "epoch": 1.875,
830
+ "frac_reward_zero_std": 0.0,
831
+ "grad_norm": 2.0184178352355957,
832
+ "kl": 0.11923892050981522,
833
+ "learning_rate": 6.000000000000001e-07,
834
+ "loss": -0.02861384116113186,
835
+ "num_tokens": 303192.0,
836
+ "reward": -0.18860293924808502,
837
+ "reward_std": 0.3568896949291229,
838
+ "rewards/GeneratorRewardFunction/mean": -0.18860293924808502,
839
+ "rewards/GeneratorRewardFunction/std": 0.3568896949291229,
840
+ "step": 45,
841
+ "step_time": 12.666136853000353
842
+ },
843
+ {
844
+ "clip_ratio/high_max": 0.0006510416860692203,
845
+ "clip_ratio/high_mean": 0.0006510416860692203,
846
+ "clip_ratio/low_mean": 0.0006510416860692203,
847
+ "clip_ratio/low_min": 0.0006510416860692203,
848
+ "clip_ratio/region_mean": 0.0013020833721384406,
849
+ "entropy": 1.8437665700912476,
850
+ "epoch": 1.9166666666666665,
851
+ "grad_norm": 3.0968308448791504,
852
+ "kl": 0.13262896239757538,
853
+ "learning_rate": 5.000000000000001e-07,
854
+ "loss": -0.13969974219799042,
855
+ "step": 46,
856
+ "step_time": 0.2212220750002416
857
+ },
858
+ {
859
+ "clip_ratio/high_max": 0.0,
860
+ "clip_ratio/high_mean": 0.0,
861
+ "clip_ratio/low_mean": 0.0,
862
+ "clip_ratio/low_min": 0.0,
863
+ "clip_ratio/region_mean": 0.0,
864
+ "entropy": 1.501238226890564,
865
+ "epoch": 1.9583333333333335,
866
+ "grad_norm": 1.9655606746673584,
867
+ "kl": 0.1233946904540062,
868
+ "learning_rate": 4.0000000000000003e-07,
869
+ "loss": 0.015763644129037857,
870
+ "step": 47,
871
+ "step_time": 0.2197655360000681
872
+ },
873
+ {
874
+ "clip_ratio/high_max": 0.0006510416860692203,
875
+ "clip_ratio/high_mean": 0.0006510416860692203,
876
+ "clip_ratio/low_mean": 0.0,
877
+ "clip_ratio/low_min": 0.0,
878
+ "clip_ratio/region_mean": 0.0006510416860692203,
879
+ "entropy": 2.031977891921997,
880
+ "epoch": 2.0,
881
+ "grad_norm": 3.66928768157959,
882
+ "kl": 0.102005235850811,
883
+ "learning_rate": 3.0000000000000004e-07,
884
+ "loss": 0.15419435501098633,
885
+ "step": 48,
886
+ "step_time": 0.21971565600006215
887
+ },
888
+ {
889
+ "clip_ratio/high_max": 0.0,
890
+ "clip_ratio/high_mean": 0.0,
891
+ "clip_ratio/low_mean": 0.0,
892
+ "clip_ratio/low_min": 0.0,
893
+ "clip_ratio/region_mean": 0.0,
894
+ "completions/clipped_ratio": 1.0,
895
+ "completions/max_length": 384.0,
896
+ "completions/max_terminated_length": 0.0,
897
+ "completions/mean_length": 384.0,
898
+ "completions/mean_terminated_length": 0.0,
899
+ "completions/min_length": 384.0,
900
+ "completions/min_terminated_length": 0.0,
901
+ "entropy": 1.4279413223266602,
902
+ "epoch": 2.0416666666666665,
903
+ "frac_reward_zero_std": 0.0,
904
+ "grad_norm": 2.270608901977539,
905
+ "kl": 0.0915692076086998,
906
+ "learning_rate": 2.0000000000000002e-07,
907
+ "loss": 0.017728755250573158,
908
+ "num_tokens": 328804.0,
909
+ "reward": 0.06775510311126709,
910
+ "reward_std": 0.3357759118080139,
911
+ "rewards/GeneratorRewardFunction/mean": 0.06775510311126709,
912
+ "rewards/GeneratorRewardFunction/std": 0.3357759118080139,
913
+ "step": 49,
914
+ "step_time": 13.325419247000355
915
+ },
916
+ {
917
+ "clip_ratio/high_max": 0.0026041667442768812,
918
+ "clip_ratio/high_mean": 0.0026041667442768812,
919
+ "clip_ratio/low_mean": 0.0,
920
+ "clip_ratio/low_min": 0.0,
921
+ "clip_ratio/region_mean": 0.0026041667442768812,
922
+ "entropy": 1.4744024276733398,
923
+ "epoch": 2.0833333333333335,
924
+ "grad_norm": 2.7985856533050537,
925
+ "kl": 0.11476268619298935,
926
+ "learning_rate": 1.0000000000000001e-07,
927
+ "loss": 0.06151821091771126,
928
+ "step": 50,
929
+ "step_time": 0.2295348450002166
930
+ }
931
+ ],
932
+ "logging_steps": 1,
933
+ "max_steps": 50,
934
+ "num_input_tokens_seen": 328804,
935
+ "num_train_epochs": 3,
936
+ "save_steps": 10,
937
+ "stateful_callbacks": {
938
+ "TrainerControl": {
939
+ "args": {
940
+ "should_epoch_stop": false,
941
+ "should_evaluate": false,
942
+ "should_log": false,
943
+ "should_save": true,
944
+ "should_training_stop": true
945
+ },
946
+ "attributes": {}
947
+ }
948
+ },
949
+ "total_flos": 0.0,
950
+ "train_batch_size": 4,
951
+ "trial_name": null,
952
+ "trial_params": null
953
+ }
self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:31ec66b64f432daf7616434296713e432d134face96e308f2ebc175e2e26f025
3
+ size 7249
self_play_hf_a10g_train/round_001/generator_train/final_model/chat_template.jinja ADDED
@@ -0,0 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {%- if tools %}
2
+ {{- '<|im_start|>system\n' }}
3
+ {%- if messages[0]['role'] == 'system' %}
4
+ {{- messages[0]['content'] }}
5
+ {%- else %}
6
+ {{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }}
7
+ {%- endif %}
8
+ {{- "\n\n# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
9
+ {%- for tool in tools %}
10
+ {{- "\n" }}
11
+ {{- tool | tojson }}
12
+ {%- endfor %}
13
+ {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
14
+ {%- else %}
15
+ {%- if messages[0]['role'] == 'system' %}
16
+ {{- '<|im_start|>system\n' + messages[0]['content'] + '<|im_end|>\n' }}
17
+ {%- else %}
18
+ {{- '<|im_start|>system\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\n' }}
19
+ {%- endif %}
20
+ {%- endif %}
21
+ {%- for message in messages %}
22
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) or (message.role == "assistant" and not message.tool_calls) %}
23
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
24
+ {%- elif message.role == "assistant" %}
25
+ {{- '<|im_start|>' + message.role }}
26
+ {%- if message.content %}
27
+ {{- '\n' + message.content }}
28
+ {%- endif %}
29
+ {%- for tool_call in message.tool_calls %}
30
+ {%- if tool_call.function is defined %}
31
+ {%- set tool_call = tool_call.function %}
32
+ {%- endif %}
33
+ {{- '\n<tool_call>\n{"name": "' }}
34
+ {{- tool_call.name }}
35
+ {{- '", "arguments": ' }}
36
+ {{- tool_call.arguments | tojson }}
37
+ {{- '}\n</tool_call>' }}
38
+ {%- endfor %}
39
+ {{- '<|im_end|>\n' }}
40
+ {%- elif message.role == "tool" %}
41
+ {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != "tool") %}
42
+ {{- '<|im_start|>user' }}
43
+ {%- endif %}
44
+ {{- '\n<tool_response>\n' }}
45
+ {{- message.content }}
46
+ {{- '\n</tool_response>' }}
47
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
48
+ {{- '<|im_end|>\n' }}
49
+ {%- endif %}
50
+ {%- endif %}
51
+ {%- endfor %}
52
+ {%- if add_generation_prompt %}
53
+ {{- '<|im_start|>assistant\n' }}
54
+ {%- endif %}
self_play_hf_a10g_train/round_001/generator_train/final_model/config.json ADDED
@@ -0,0 +1,57 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Qwen2ForCausalLM"
4
+ ],
5
+ "attention_dropout": 0.0,
6
+ "bos_token_id": null,
7
+ "dtype": "float32",
8
+ "eos_token_id": 151645,
9
+ "hidden_act": "silu",
10
+ "hidden_size": 896,
11
+ "initializer_range": 0.02,
12
+ "intermediate_size": 4864,
13
+ "layer_types": [
14
+ "full_attention",
15
+ "full_attention",
16
+ "full_attention",
17
+ "full_attention",
18
+ "full_attention",
19
+ "full_attention",
20
+ "full_attention",
21
+ "full_attention",
22
+ "full_attention",
23
+ "full_attention",
24
+ "full_attention",
25
+ "full_attention",
26
+ "full_attention",
27
+ "full_attention",
28
+ "full_attention",
29
+ "full_attention",
30
+ "full_attention",
31
+ "full_attention",
32
+ "full_attention",
33
+ "full_attention",
34
+ "full_attention",
35
+ "full_attention",
36
+ "full_attention",
37
+ "full_attention"
38
+ ],
39
+ "max_position_embeddings": 32768,
40
+ "max_window_layers": 21,
41
+ "model_type": "qwen2",
42
+ "num_attention_heads": 14,
43
+ "num_hidden_layers": 24,
44
+ "num_key_value_heads": 2,
45
+ "pad_token_id": 151643,
46
+ "rms_norm_eps": 1e-06,
47
+ "rope_parameters": {
48
+ "rope_theta": 1000000.0,
49
+ "rope_type": "default"
50
+ },
51
+ "sliding_window": null,
52
+ "tie_word_embeddings": true,
53
+ "transformers_version": "5.6.2",
54
+ "use_cache": false,
55
+ "use_sliding_window": false,
56
+ "vocab_size": 151936
57
+ }
self_play_hf_a10g_train/round_001/generator_train/final_model/generation_config.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_sample": true,
3
+ "eos_token_id": [
4
+ 151645,
5
+ 151643
6
+ ],
7
+ "pad_token_id": 151643,
8
+ "repetition_penalty": 1.1,
9
+ "temperature": 0.7,
10
+ "top_k": 20,
11
+ "top_p": 0.8,
12
+ "transformers_version": "5.6.2"
13
+ }
self_play_hf_a10g_train/round_001/generator_train/final_model/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:419f71328f66d9eaa2ba1d9037f5033a97db1d93862f5eb35b2f04ea57340b82
3
+ size 1976163472
self_play_hf_a10g_train/round_001/generator_train/final_model/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3fd169731d2cbde95e10bf356d66d5997fd885dd8dbb6fb4684da3f23b2585d8
3
+ size 11421892
self_play_hf_a10g_train/round_001/generator_train/final_model/tokenizer_config.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "backend": "tokenizers",
4
+ "bos_token": null,
5
+ "clean_up_tokenization_spaces": false,
6
+ "eos_token": "<|im_end|>",
7
+ "errors": "replace",
8
+ "extra_special_tokens": [
9
+ "<|im_start|>",
10
+ "<|im_end|>",
11
+ "<|object_ref_start|>",
12
+ "<|object_ref_end|>",
13
+ "<|box_start|>",
14
+ "<|box_end|>",
15
+ "<|quad_start|>",
16
+ "<|quad_end|>",
17
+ "<|vision_start|>",
18
+ "<|vision_end|>",
19
+ "<|vision_pad|>",
20
+ "<|image_pad|>",
21
+ "<|video_pad|>"
22
+ ],
23
+ "is_local": false,
24
+ "local_files_only": false,
25
+ "model_max_length": 131072,
26
+ "pad_token": "<|endoftext|>",
27
+ "padding_side": "left",
28
+ "split_special_tokens": false,
29
+ "tokenizer_class": "Qwen2Tokenizer",
30
+ "truncation_side": "left",
31
+ "unk_token": null
32
+ }
self_play_hf_a10g_train/round_001/generator_train/final_model/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:31ec66b64f432daf7616434296713e432d134face96e308f2ebc175e2e26f025
3
+ size 7249