Upload generator checkpoints for round 001
Browse files- .gitattributes +3 -0
- self_play_hf_a10g_train/round_001/generator_train/README.md +67 -0
- self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/chat_template.jinja +54 -0
- self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/config.json +57 -0
- self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/generation_config.json +13 -0
- self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/model.safetensors +3 -0
- self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/optimizer.pt +3 -0
- self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/rng_state.pth +3 -0
- self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/scheduler.pt +3 -0
- self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/tokenizer.json +3 -0
- self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/tokenizer_config.json +32 -0
- self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/trainer_state.json +764 -0
- self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/training_args.bin +3 -0
- self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/chat_template.jinja +54 -0
- self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/config.json +57 -0
- self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/generation_config.json +13 -0
- self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/model.safetensors +3 -0
- self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/optimizer.pt +3 -0
- self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/rng_state.pth +3 -0
- self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/scheduler.pt +3 -0
- self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/tokenizer.json +3 -0
- self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/tokenizer_config.json +32 -0
- self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/trainer_state.json +953 -0
- self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/training_args.bin +3 -0
- self_play_hf_a10g_train/round_001/generator_train/final_model/chat_template.jinja +54 -0
- self_play_hf_a10g_train/round_001/generator_train/final_model/config.json +57 -0
- self_play_hf_a10g_train/round_001/generator_train/final_model/generation_config.json +13 -0
- self_play_hf_a10g_train/round_001/generator_train/final_model/model.safetensors +3 -0
- self_play_hf_a10g_train/round_001/generator_train/final_model/tokenizer.json +3 -0
- self_play_hf_a10g_train/round_001/generator_train/final_model/tokenizer_config.json +32 -0
- self_play_hf_a10g_train/round_001/generator_train/final_model/training_args.bin +3 -0
.gitattributes
CHANGED
|
@@ -33,3 +33,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
|
|
|
|
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
| 36 |
+
self_play_hf_a10g_train/round_001/generator_train/checkpoint-40/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 37 |
+
self_play_hf_a10g_train/round_001/generator_train/checkpoint-50/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 38 |
+
self_play_hf_a10g_train/round_001/generator_train/final_model/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
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/w4yxkqbv)
|
| 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:f104793ef80b632081adc349f9a54bede0112ea26d13ce2c2a8312cf61dbbfae
|
| 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 sha256:2d04b529c8641e602da72a33b67693f5b64694dac6252d75dad9f985ff685e6b
|
| 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": 3.343350648880005,
|
| 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": 12.520110770000002
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"clip_ratio/high_max": 0.0045572915114462376,
|
| 42 |
+
"clip_ratio/high_mean": 0.0045572915114462376,
|
| 43 |
+
"clip_ratio/low_mean": 0.0013020833721384406,
|
| 44 |
+
"clip_ratio/low_min": 0.0013020833721384406,
|
| 45 |
+
"clip_ratio/region_mean": 0.005859375,
|
| 46 |
+
"entropy": 1.2710224390029907,
|
| 47 |
+
"epoch": 0.08333333333333333,
|
| 48 |
+
"grad_norm": 2.8392181396484375,
|
| 49 |
+
"kl": 0.001467077643610537,
|
| 50 |
+
"learning_rate": 4.9000000000000005e-06,
|
| 51 |
+
"loss": -0.06676606088876724,
|
| 52 |
+
"step": 2,
|
| 53 |
+
"step_time": 0.22065511100001345
|
| 54 |
+
},
|
| 55 |
+
{
|
| 56 |
+
"clip_ratio/high_max": 0.013671875,
|
| 57 |
+
"clip_ratio/high_mean": 0.013671875,
|
| 58 |
+
"clip_ratio/low_mean": 0.014322916977107525,
|
| 59 |
+
"clip_ratio/low_min": 0.014322916977107525,
|
| 60 |
+
"clip_ratio/region_mean": 0.02799479104578495,
|
| 61 |
+
"entropy": 1.871756911277771,
|
| 62 |
+
"epoch": 0.125,
|
| 63 |
+
"grad_norm": 2.4508721828460693,
|
| 64 |
+
"kl": 0.004840313456952572,
|
| 65 |
+
"learning_rate": 4.800000000000001e-06,
|
| 66 |
+
"loss": 0.010330882854759693,
|
| 67 |
+
"step": 3,
|
| 68 |
+
"step_time": 0.21905603899998027
|
| 69 |
+
},
|
| 70 |
+
{
|
| 71 |
+
"clip_ratio/high_max": 0.01822916604578495,
|
| 72 |
+
"clip_ratio/high_mean": 0.01822916604578495,
|
| 73 |
+
"clip_ratio/low_mean": 0.010416666977107525,
|
| 74 |
+
"clip_ratio/low_min": 0.010416666977107525,
|
| 75 |
+
"clip_ratio/region_mean": 0.02864583395421505,
|
| 76 |
+
"entropy": 1.1871482133865356,
|
| 77 |
+
"epoch": 0.16666666666666666,
|
| 78 |
+
"grad_norm": 1.5818687677383423,
|
| 79 |
+
"kl": 0.005701068323105574,
|
| 80 |
+
"learning_rate": 4.7e-06,
|
| 81 |
+
"loss": -0.08211664110422134,
|
| 82 |
+
"step": 4,
|
| 83 |
+
"step_time": 0.21982950500000698
|
| 84 |
+
},
|
| 85 |
+
{
|
| 86 |
+
"clip_ratio/high_max": 0.0006510416860692203,
|
| 87 |
+
"clip_ratio/high_mean": 0.0006510416860692203,
|
| 88 |
+
"clip_ratio/low_mean": 0.0006510416860692203,
|
| 89 |
+
"clip_ratio/low_min": 0.0006510416860692203,
|
| 90 |
+
"clip_ratio/region_mean": 0.0013020833721384406,
|
| 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": 1.0733331441879272,
|
| 99 |
+
"epoch": 0.20833333333333334,
|
| 100 |
+
"frac_reward_zero_std": 0.0,
|
| 101 |
+
"grad_norm": 3.4188177585601807,
|
| 102 |
+
"kl": 0.018397731706500053,
|
| 103 |
+
"learning_rate": 4.600000000000001e-06,
|
| 104 |
+
"loss": 0.14518485963344574,
|
| 105 |
+
"num_tokens": 50440.0,
|
| 106 |
+
"reward": -0.2228125035762787,
|
| 107 |
+
"reward_std": 0.274566113948822,
|
| 108 |
+
"rewards/GeneratorRewardFunction/mean": -0.2228125035762787,
|
| 109 |
+
"rewards/GeneratorRewardFunction/std": 0.2745661437511444,
|
| 110 |
+
"step": 5,
|
| 111 |
+
"step_time": 12.121955963999994
|
| 112 |
+
},
|
| 113 |
+
{
|
| 114 |
+
"clip_ratio/high_max": 0.0052083334885537624,
|
| 115 |
+
"clip_ratio/high_mean": 0.0052083334885537624,
|
| 116 |
+
"clip_ratio/low_mean": 0.001953125,
|
| 117 |
+
"clip_ratio/low_min": 0.001953125,
|
| 118 |
+
"clip_ratio/region_mean": 0.0071614584885537624,
|
| 119 |
+
"entropy": 1.327884316444397,
|
| 120 |
+
"epoch": 0.25,
|
| 121 |
+
"grad_norm": 2.6934618949890137,
|
| 122 |
+
"kl": 0.01829482428729534,
|
| 123 |
+
"learning_rate": 4.5e-06,
|
| 124 |
+
"loss": -0.037107061594724655,
|
| 125 |
+
"step": 6,
|
| 126 |
+
"step_time": 0.22655425900001092
|
| 127 |
+
},
|
| 128 |
+
{
|
| 129 |
+
"clip_ratio/high_max": 0.0065104165114462376,
|
| 130 |
+
"clip_ratio/high_mean": 0.0065104165114462376,
|
| 131 |
+
"clip_ratio/low_mean": 0.0052083334885537624,
|
| 132 |
+
"clip_ratio/low_min": 0.0052083334885537624,
|
| 133 |
+
"clip_ratio/region_mean": 0.01171875,
|
| 134 |
+
"entropy": 1.2937031984329224,
|
| 135 |
+
"epoch": 0.2916666666666667,
|
| 136 |
+
"grad_norm": 2.8983969688415527,
|
| 137 |
+
"kl": 0.021993428468704224,
|
| 138 |
+
"learning_rate": 4.4e-06,
|
| 139 |
+
"loss": 0.013194209896028042,
|
| 140 |
+
"step": 7,
|
| 141 |
+
"step_time": 0.22578505299998142
|
| 142 |
+
},
|
| 143 |
+
{
|
| 144 |
+
"clip_ratio/high_max": 0.029296875,
|
| 145 |
+
"clip_ratio/high_mean": 0.029296875,
|
| 146 |
+
"clip_ratio/low_mean": 0.0,
|
| 147 |
+
"clip_ratio/low_min": 0.0,
|
| 148 |
+
"clip_ratio/region_mean": 0.029296875,
|
| 149 |
+
"entropy": 0.7918106913566589,
|
| 150 |
+
"epoch": 0.3333333333333333,
|
| 151 |
+
"grad_norm": 1.268328309059143,
|
| 152 |
+
"kl": 0.03330208733677864,
|
| 153 |
+
"learning_rate": 4.3e-06,
|
| 154 |
+
"loss": -0.12033451348543167,
|
| 155 |
+
"step": 8,
|
| 156 |
+
"step_time": 0.2253496229999996
|
| 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": 1.0043877363204956,
|
| 172 |
+
"epoch": 0.375,
|
| 173 |
+
"frac_reward_zero_std": 0.25,
|
| 174 |
+
"grad_norm": 1.3302299976348877,
|
| 175 |
+
"kl": 0.02948068641126156,
|
| 176 |
+
"learning_rate": 4.2000000000000004e-06,
|
| 177 |
+
"loss": -0.07878098636865616,
|
| 178 |
+
"num_tokens": 75884.0,
|
| 179 |
+
"reward": -0.12250000238418579,
|
| 180 |
+
"reward_std": 0.21038061380386353,
|
| 181 |
+
"rewards/GeneratorRewardFunction/mean": -0.12250000238418579,
|
| 182 |
+
"rewards/GeneratorRewardFunction/std": 0.21038061380386353,
|
| 183 |
+
"step": 9,
|
| 184 |
+
"step_time": 11.949294435000013
|
| 185 |
+
},
|
| 186 |
+
{
|
| 187 |
+
"clip_ratio/high_max": 0.010416666977107525,
|
| 188 |
+
"clip_ratio/high_mean": 0.010416666977107525,
|
| 189 |
+
"clip_ratio/low_mean": 0.0032552082557231188,
|
| 190 |
+
"clip_ratio/low_min": 0.0032552082557231188,
|
| 191 |
+
"clip_ratio/region_mean": 0.013671875,
|
| 192 |
+
"entropy": 1.1474007368087769,
|
| 193 |
+
"epoch": 0.4166666666666667,
|
| 194 |
+
"grad_norm": 2.1202971935272217,
|
| 195 |
+
"kl": 0.03631855919957161,
|
| 196 |
+
"learning_rate": 4.1e-06,
|
| 197 |
+
"loss": 0.03980601206421852,
|
| 198 |
+
"step": 10,
|
| 199 |
+
"step_time": 0.22497136300000875
|
| 200 |
+
},
|
| 201 |
+
{
|
| 202 |
+
"clip_ratio/high_max": 0.008463541977107525,
|
| 203 |
+
"clip_ratio/high_mean": 0.008463541977107525,
|
| 204 |
+
"clip_ratio/low_mean": 0.00390625,
|
| 205 |
+
"clip_ratio/low_min": 0.00390625,
|
| 206 |
+
"clip_ratio/region_mean": 0.012369791977107525,
|
| 207 |
+
"entropy": 0.9981658458709717,
|
| 208 |
+
"epoch": 0.4583333333333333,
|
| 209 |
+
"grad_norm": 2.094111680984497,
|
| 210 |
+
"kl": 0.049915943294763565,
|
| 211 |
+
"learning_rate": 4.000000000000001e-06,
|
| 212 |
+
"loss": 0.06331142038106918,
|
| 213 |
+
"step": 11,
|
| 214 |
+
"step_time": 0.22518257699999822
|
| 215 |
+
},
|
| 216 |
+
{
|
| 217 |
+
"clip_ratio/high_max": 0.0403645820915699,
|
| 218 |
+
"clip_ratio/high_mean": 0.0403645820915699,
|
| 219 |
+
"clip_ratio/low_mean": 0.0065104165114462376,
|
| 220 |
+
"clip_ratio/low_min": 0.0065104165114462376,
|
| 221 |
+
"clip_ratio/region_mean": 0.046875,
|
| 222 |
+
"entropy": 1.3572144508361816,
|
| 223 |
+
"epoch": 0.5,
|
| 224 |
+
"grad_norm": 2.6413395404815674,
|
| 225 |
+
"kl": 0.053241848945617676,
|
| 226 |
+
"learning_rate": 3.900000000000001e-06,
|
| 227 |
+
"loss": -0.022430941462516785,
|
| 228 |
+
"step": 12,
|
| 229 |
+
"step_time": 0.2254562319999991
|
| 230 |
+
},
|
| 231 |
+
{
|
| 232 |
+
"clip_ratio/high_max": 0.0,
|
| 233 |
+
"clip_ratio/high_mean": 0.0,
|
| 234 |
+
"clip_ratio/low_mean": 0.001953125,
|
| 235 |
+
"clip_ratio/low_min": 0.001953125,
|
| 236 |
+
"clip_ratio/region_mean": 0.001953125,
|
| 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": 1.629288673400879,
|
| 245 |
+
"epoch": 0.5416666666666666,
|
| 246 |
+
"frac_reward_zero_std": 0.25,
|
| 247 |
+
"grad_norm": 2.107161283493042,
|
| 248 |
+
"kl": 0.0636429563164711,
|
| 249 |
+
"learning_rate": 3.8000000000000005e-06,
|
| 250 |
+
"loss": 0.11612584441900253,
|
| 251 |
+
"num_tokens": 101112.0,
|
| 252 |
+
"reward": -0.11937499791383743,
|
| 253 |
+
"reward_std": 0.20747588574886322,
|
| 254 |
+
"rewards/GeneratorRewardFunction/mean": -0.11937499791383743,
|
| 255 |
+
"rewards/GeneratorRewardFunction/std": 0.20747590065002441,
|
| 256 |
+
"step": 13,
|
| 257 |
+
"step_time": 12.014855219999987
|
| 258 |
+
},
|
| 259 |
+
{
|
| 260 |
+
"clip_ratio/high_max": 0.0032552082557231188,
|
| 261 |
+
"clip_ratio/high_mean": 0.0032552082557231188,
|
| 262 |
+
"clip_ratio/low_mean": 0.0006510416860692203,
|
| 263 |
+
"clip_ratio/low_min": 0.0006510416860692203,
|
| 264 |
+
"clip_ratio/region_mean": 0.00390625,
|
| 265 |
+
"entropy": 1.4905215501785278,
|
| 266 |
+
"epoch": 0.5833333333333334,
|
| 267 |
+
"grad_norm": 1.75613272190094,
|
| 268 |
+
"kl": 0.060588542371988297,
|
| 269 |
+
"learning_rate": 3.7e-06,
|
| 270 |
+
"loss": 0.0006357845850288868,
|
| 271 |
+
"step": 14,
|
| 272 |
+
"step_time": 0.21892069699998729
|
| 273 |
+
},
|
| 274 |
+
{
|
| 275 |
+
"clip_ratio/high_max": 0.0071614584885537624,
|
| 276 |
+
"clip_ratio/high_mean": 0.0071614584885537624,
|
| 277 |
+
"clip_ratio/low_mean": 0.009114583022892475,
|
| 278 |
+
"clip_ratio/low_min": 0.009114583022892475,
|
| 279 |
+
"clip_ratio/region_mean": 0.01627604104578495,
|
| 280 |
+
"entropy": 1.2682157754898071,
|
| 281 |
+
"epoch": 0.625,
|
| 282 |
+
"grad_norm": 2.94674015045166,
|
| 283 |
+
"kl": 0.08301883935928345,
|
| 284 |
+
"learning_rate": 3.6000000000000003e-06,
|
| 285 |
+
"loss": -0.0923055037856102,
|
| 286 |
+
"step": 15,
|
| 287 |
+
"step_time": 0.21824655300000018
|
| 288 |
+
},
|
| 289 |
+
{
|
| 290 |
+
"clip_ratio/high_max": 0.008463541977107525,
|
| 291 |
+
"clip_ratio/high_mean": 0.008463541977107525,
|
| 292 |
+
"clip_ratio/low_mean": 0.01627604104578495,
|
| 293 |
+
"clip_ratio/low_min": 0.01627604104578495,
|
| 294 |
+
"clip_ratio/region_mean": 0.02473958395421505,
|
| 295 |
+
"entropy": 0.9931669235229492,
|
| 296 |
+
"epoch": 0.6666666666666666,
|
| 297 |
+
"grad_norm": 2.1109514236450195,
|
| 298 |
+
"kl": 0.09274417906999588,
|
| 299 |
+
"learning_rate": 3.5e-06,
|
| 300 |
+
"loss": -0.0218100156635046,
|
| 301 |
+
"step": 16,
|
| 302 |
+
"step_time": 0.21831970400000955
|
| 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": 1.376638412475586,
|
| 318 |
+
"epoch": 0.7083333333333334,
|
| 319 |
+
"frac_reward_zero_std": 0.0,
|
| 320 |
+
"grad_norm": 2.625627279281616,
|
| 321 |
+
"kl": 0.07451707124710083,
|
| 322 |
+
"learning_rate": 3.4000000000000005e-06,
|
| 323 |
+
"loss": 0.02886168472468853,
|
| 324 |
+
"num_tokens": 126436.0,
|
| 325 |
+
"reward": -0.09125000238418579,
|
| 326 |
+
"reward_std": 0.1655445545911789,
|
| 327 |
+
"rewards/GeneratorRewardFunction/mean": -0.09125000238418579,
|
| 328 |
+
"rewards/GeneratorRewardFunction/std": 0.1655445545911789,
|
| 329 |
+
"step": 17,
|
| 330 |
+
"step_time": 12.074089973000014
|
| 331 |
+
},
|
| 332 |
+
{
|
| 333 |
+
"clip_ratio/high_max": 0.0006510416860692203,
|
| 334 |
+
"clip_ratio/high_mean": 0.0006510416860692203,
|
| 335 |
+
"clip_ratio/low_mean": 0.001953125,
|
| 336 |
+
"clip_ratio/low_min": 0.001953125,
|
| 337 |
+
"clip_ratio/region_mean": 0.0026041667442768812,
|
| 338 |
+
"entropy": 0.8447733521461487,
|
| 339 |
+
"epoch": 0.75,
|
| 340 |
+
"grad_norm": 2.2611021995544434,
|
| 341 |
+
"kl": 0.08117184042930603,
|
| 342 |
+
"learning_rate": 3.3000000000000006e-06,
|
| 343 |
+
"loss": -0.003659568028524518,
|
| 344 |
+
"step": 18,
|
| 345 |
+
"step_time": 0.22418350800001008
|
| 346 |
+
},
|
| 347 |
+
{
|
| 348 |
+
"clip_ratio/high_max": 0.0065104165114462376,
|
| 349 |
+
"clip_ratio/high_mean": 0.0065104165114462376,
|
| 350 |
+
"clip_ratio/low_mean": 0.0006510416860692203,
|
| 351 |
+
"clip_ratio/low_min": 0.0006510416860692203,
|
| 352 |
+
"clip_ratio/region_mean": 0.0071614584885537624,
|
| 353 |
+
"entropy": 0.9943304061889648,
|
| 354 |
+
"epoch": 0.7916666666666666,
|
| 355 |
+
"grad_norm": 1.5852197408676147,
|
| 356 |
+
"kl": 0.08660884946584702,
|
| 357 |
+
"learning_rate": 3.2000000000000003e-06,
|
| 358 |
+
"loss": -0.10765092819929123,
|
| 359 |
+
"step": 19,
|
| 360 |
+
"step_time": 0.224853281999998
|
| 361 |
+
},
|
| 362 |
+
{
|
| 363 |
+
"clip_ratio/high_max": 0.008463541977107525,
|
| 364 |
+
"clip_ratio/high_mean": 0.008463541977107525,
|
| 365 |
+
"clip_ratio/low_mean": 0.02213541604578495,
|
| 366 |
+
"clip_ratio/low_min": 0.02213541604578495,
|
| 367 |
+
"clip_ratio/region_mean": 0.03059895895421505,
|
| 368 |
+
"entropy": 1.2907896041870117,
|
| 369 |
+
"epoch": 0.8333333333333334,
|
| 370 |
+
"grad_norm": 2.97239089012146,
|
| 371 |
+
"kl": 0.08734595775604248,
|
| 372 |
+
"learning_rate": 3.1000000000000004e-06,
|
| 373 |
+
"loss": 0.08410018682479858,
|
| 374 |
+
"step": 20,
|
| 375 |
+
"step_time": 0.22594529400001306
|
| 376 |
+
},
|
| 377 |
+
{
|
| 378 |
+
"clip_ratio/high_max": 0.0026041667442768812,
|
| 379 |
+
"clip_ratio/high_mean": 0.0026041667442768812,
|
| 380 |
+
"clip_ratio/low_mean": 0.0,
|
| 381 |
+
"clip_ratio/low_min": 0.0,
|
| 382 |
+
"clip_ratio/region_mean": 0.0026041667442768812,
|
| 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": 1.3083521127700806,
|
| 391 |
+
"epoch": 0.875,
|
| 392 |
+
"frac_reward_zero_std": 0.25,
|
| 393 |
+
"grad_norm": 2.6607460975646973,
|
| 394 |
+
"kl": 0.08000912517309189,
|
| 395 |
+
"learning_rate": 3e-06,
|
| 396 |
+
"loss": -0.11818201094865799,
|
| 397 |
+
"num_tokens": 151596.0,
|
| 398 |
+
"reward": -0.22624999284744263,
|
| 399 |
+
"reward_std": 0.2758048474788666,
|
| 400 |
+
"rewards/GeneratorRewardFunction/mean": -0.22624999284744263,
|
| 401 |
+
"rewards/GeneratorRewardFunction/std": 0.27580487728118896,
|
| 402 |
+
"step": 21,
|
| 403 |
+
"step_time": 11.957008280999986
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"clip_ratio/high_max": 0.0013020833721384406,
|
| 407 |
+
"clip_ratio/high_mean": 0.0013020833721384406,
|
| 408 |
+
"clip_ratio/low_mean": 0.0013020833721384406,
|
| 409 |
+
"clip_ratio/low_min": 0.0013020833721384406,
|
| 410 |
+
"clip_ratio/region_mean": 0.0026041667442768812,
|
| 411 |
+
"entropy": 1.2017608880996704,
|
| 412 |
+
"epoch": 0.9166666666666666,
|
| 413 |
+
"grad_norm": 0.8747857809066772,
|
| 414 |
+
"kl": 0.11337386816740036,
|
| 415 |
+
"learning_rate": 2.9e-06,
|
| 416 |
+
"loss": -0.003786882385611534,
|
| 417 |
+
"step": 22,
|
| 418 |
+
"step_time": 0.221026849999987
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"clip_ratio/high_max": 0.0013020833721384406,
|
| 422 |
+
"clip_ratio/high_mean": 0.0013020833721384406,
|
| 423 |
+
"clip_ratio/low_mean": 0.001953125,
|
| 424 |
+
"clip_ratio/low_min": 0.001953125,
|
| 425 |
+
"clip_ratio/region_mean": 0.0032552082557231188,
|
| 426 |
+
"entropy": 1.107405662536621,
|
| 427 |
+
"epoch": 0.9583333333333334,
|
| 428 |
+
"grad_norm": 2.115562915802002,
|
| 429 |
+
"kl": 0.09759091585874557,
|
| 430 |
+
"learning_rate": 2.8000000000000003e-06,
|
| 431 |
+
"loss": 0.08772162348031998,
|
| 432 |
+
"step": 23,
|
| 433 |
+
"step_time": 0.22019210400000588
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"clip_ratio/high_max": 0.005859375,
|
| 437 |
+
"clip_ratio/high_mean": 0.005859375,
|
| 438 |
+
"clip_ratio/low_mean": 0.00390625,
|
| 439 |
+
"clip_ratio/low_min": 0.00390625,
|
| 440 |
+
"clip_ratio/region_mean": 0.009765625,
|
| 441 |
+
"entropy": 0.7833542823791504,
|
| 442 |
+
"epoch": 1.0,
|
| 443 |
+
"grad_norm": 1.686574101448059,
|
| 444 |
+
"kl": 0.10800782591104507,
|
| 445 |
+
"learning_rate": 2.7000000000000004e-06,
|
| 446 |
+
"loss": 0.03493640199303627,
|
| 447 |
+
"step": 24,
|
| 448 |
+
"step_time": 0.22060076099998582
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"clip_ratio/high_max": 0.0013020833721384406,
|
| 452 |
+
"clip_ratio/high_mean": 0.0013020833721384406,
|
| 453 |
+
"clip_ratio/low_mean": 0.0,
|
| 454 |
+
"clip_ratio/low_min": 0.0,
|
| 455 |
+
"clip_ratio/region_mean": 0.0013020833721384406,
|
| 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": 1.0690312385559082,
|
| 464 |
+
"epoch": 1.0416666666666667,
|
| 465 |
+
"frac_reward_zero_std": 0.0,
|
| 466 |
+
"grad_norm": 2.565182685852051,
|
| 467 |
+
"kl": 0.12982706725597382,
|
| 468 |
+
"learning_rate": 2.6e-06,
|
| 469 |
+
"loss": 0.12511824071407318,
|
| 470 |
+
"num_tokens": 177212.0,
|
| 471 |
+
"reward": -0.12406250089406967,
|
| 472 |
+
"reward_std": 0.19458690285682678,
|
| 473 |
+
"rewards/GeneratorRewardFunction/mean": -0.12406250089406967,
|
| 474 |
+
"rewards/GeneratorRewardFunction/std": 0.19458691775798798,
|
| 475 |
+
"step": 25,
|
| 476 |
+
"step_time": 12.125360838999995
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"clip_ratio/high_max": 0.00390625,
|
| 480 |
+
"clip_ratio/high_mean": 0.00390625,
|
| 481 |
+
"clip_ratio/low_mean": 0.0013020833721384406,
|
| 482 |
+
"clip_ratio/low_min": 0.0013020833721384406,
|
| 483 |
+
"clip_ratio/region_mean": 0.0052083334885537624,
|
| 484 |
+
"entropy": 0.8722183108329773,
|
| 485 |
+
"epoch": 1.0833333333333333,
|
| 486 |
+
"grad_norm": 2.401808261871338,
|
| 487 |
+
"kl": 0.12285982817411423,
|
| 488 |
+
"learning_rate": 2.5e-06,
|
| 489 |
+
"loss": -0.13922104239463806,
|
| 490 |
+
"step": 26,
|
| 491 |
+
"step_time": 0.22810240500001555
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"clip_ratio/high_max": 0.0052083334885537624,
|
| 495 |
+
"clip_ratio/high_mean": 0.0052083334885537624,
|
| 496 |
+
"clip_ratio/low_mean": 0.005859375,
|
| 497 |
+
"clip_ratio/low_min": 0.005859375,
|
| 498 |
+
"clip_ratio/region_mean": 0.011067708022892475,
|
| 499 |
+
"entropy": 1.3027639389038086,
|
| 500 |
+
"epoch": 1.125,
|
| 501 |
+
"grad_norm": 1.7678114175796509,
|
| 502 |
+
"kl": 0.10112806409597397,
|
| 503 |
+
"learning_rate": 2.4000000000000003e-06,
|
| 504 |
+
"loss": -0.0586722195148468,
|
| 505 |
+
"step": 27,
|
| 506 |
+
"step_time": 0.22764332500003093
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"clip_ratio/high_max": 0.0065104165114462376,
|
| 510 |
+
"clip_ratio/high_mean": 0.0065104165114462376,
|
| 511 |
+
"clip_ratio/low_mean": 0.0071614584885537624,
|
| 512 |
+
"clip_ratio/low_min": 0.0071614584885537624,
|
| 513 |
+
"clip_ratio/region_mean": 0.013671875,
|
| 514 |
+
"entropy": 0.9790509343147278,
|
| 515 |
+
"epoch": 1.1666666666666667,
|
| 516 |
+
"grad_norm": 1.9319959878921509,
|
| 517 |
+
"kl": 0.11484679579734802,
|
| 518 |
+
"learning_rate": 2.3000000000000004e-06,
|
| 519 |
+
"loss": 0.07509768754243851,
|
| 520 |
+
"step": 28,
|
| 521 |
+
"step_time": 0.22808939600002986
|
| 522 |
+
},
|
| 523 |
+
{
|
| 524 |
+
"clip_ratio/high_max": 0.0006510416860692203,
|
| 525 |
+
"clip_ratio/high_mean": 0.0006510416860692203,
|
| 526 |
+
"clip_ratio/low_mean": 0.0,
|
| 527 |
+
"clip_ratio/low_min": 0.0,
|
| 528 |
+
"clip_ratio/region_mean": 0.0006510416860692203,
|
| 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": 0.9768911004066467,
|
| 537 |
+
"epoch": 1.2083333333333333,
|
| 538 |
+
"frac_reward_zero_std": 0.0,
|
| 539 |
+
"grad_norm": 2.2773685455322266,
|
| 540 |
+
"kl": 0.10255210846662521,
|
| 541 |
+
"learning_rate": 2.2e-06,
|
| 542 |
+
"loss": -0.01104909647256136,
|
| 543 |
+
"num_tokens": 202200.0,
|
| 544 |
+
"reward": -0.13343749940395355,
|
| 545 |
+
"reward_std": 0.2115633636713028,
|
| 546 |
+
"rewards/GeneratorRewardFunction/mean": -0.13343749940395355,
|
| 547 |
+
"rewards/GeneratorRewardFunction/std": 0.211563378572464,
|
| 548 |
+
"step": 29,
|
| 549 |
+
"step_time": 11.981290445000013
|
| 550 |
+
},
|
| 551 |
+
{
|
| 552 |
+
"clip_ratio/high_max": 0.0013020833721384406,
|
| 553 |
+
"clip_ratio/high_mean": 0.0013020833721384406,
|
| 554 |
+
"clip_ratio/low_mean": 0.0006510416860692203,
|
| 555 |
+
"clip_ratio/low_min": 0.0006510416860692203,
|
| 556 |
+
"clip_ratio/region_mean": 0.001953125,
|
| 557 |
+
"entropy": 0.9071128368377686,
|
| 558 |
+
"epoch": 1.25,
|
| 559 |
+
"grad_norm": 2.377110004425049,
|
| 560 |
+
"kl": 0.11365322023630142,
|
| 561 |
+
"learning_rate": 2.1000000000000002e-06,
|
| 562 |
+
"loss": 0.07630521804094315,
|
| 563 |
+
"step": 30,
|
| 564 |
+
"step_time": 0.22243629600001213
|
| 565 |
+
},
|
| 566 |
+
{
|
| 567 |
+
"clip_ratio/high_max": 0.0065104165114462376,
|
| 568 |
+
"clip_ratio/high_mean": 0.0065104165114462376,
|
| 569 |
+
"clip_ratio/low_mean": 0.0026041667442768812,
|
| 570 |
+
"clip_ratio/low_min": 0.0026041667442768812,
|
| 571 |
+
"clip_ratio/region_mean": 0.009114583022892475,
|
| 572 |
+
"entropy": 1.3066421747207642,
|
| 573 |
+
"epoch": 1.2916666666666667,
|
| 574 |
+
"grad_norm": 2.6143717765808105,
|
| 575 |
+
"kl": 0.09395217150449753,
|
| 576 |
+
"learning_rate": 2.0000000000000003e-06,
|
| 577 |
+
"loss": -0.023749127984046936,
|
| 578 |
+
"step": 31,
|
| 579 |
+
"step_time": 0.22260347799999636
|
| 580 |
+
},
|
| 581 |
+
{
|
| 582 |
+
"clip_ratio/high_max": 0.005859375,
|
| 583 |
+
"clip_ratio/high_mean": 0.005859375,
|
| 584 |
+
"clip_ratio/low_mean": 0.001953125,
|
| 585 |
+
"clip_ratio/low_min": 0.001953125,
|
| 586 |
+
"clip_ratio/region_mean": 0.0078125,
|
| 587 |
+
"entropy": 1.2596086263656616,
|
| 588 |
+
"epoch": 1.3333333333333333,
|
| 589 |
+
"grad_norm": 1.4453171491622925,
|
| 590 |
+
"kl": 0.09631065279245377,
|
| 591 |
+
"learning_rate": 1.9000000000000002e-06,
|
| 592 |
+
"loss": -0.04094076156616211,
|
| 593 |
+
"step": 32,
|
| 594 |
+
"step_time": 0.2226373689999832
|
| 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": 0.9096196293830872,
|
| 610 |
+
"epoch": 1.375,
|
| 611 |
+
"frac_reward_zero_std": 0.0,
|
| 612 |
+
"grad_norm": 2.4287660121917725,
|
| 613 |
+
"kl": 0.15610061585903168,
|
| 614 |
+
"learning_rate": 1.8000000000000001e-06,
|
| 615 |
+
"loss": 0.060271572321653366,
|
| 616 |
+
"num_tokens": 227556.0,
|
| 617 |
+
"reward": -0.13343749940395355,
|
| 618 |
+
"reward_std": 0.19603331387043,
|
| 619 |
+
"rewards/GeneratorRewardFunction/mean": -0.13343749940395355,
|
| 620 |
+
"rewards/GeneratorRewardFunction/std": 0.19603331387043,
|
| 621 |
+
"step": 33,
|
| 622 |
+
"step_time": 12.116691536000019
|
| 623 |
+
},
|
| 624 |
+
{
|
| 625 |
+
"clip_ratio/high_max": 0.0032552082557231188,
|
| 626 |
+
"clip_ratio/high_mean": 0.0032552082557231188,
|
| 627 |
+
"clip_ratio/low_mean": 0.0006510416860692203,
|
| 628 |
+
"clip_ratio/low_min": 0.0006510416860692203,
|
| 629 |
+
"clip_ratio/region_mean": 0.00390625,
|
| 630 |
+
"entropy": 1.2461239099502563,
|
| 631 |
+
"epoch": 1.4166666666666667,
|
| 632 |
+
"grad_norm": 2.0638527870178223,
|
| 633 |
+
"kl": 0.11391329020261765,
|
| 634 |
+
"learning_rate": 1.7000000000000002e-06,
|
| 635 |
+
"loss": -0.012484799139201641,
|
| 636 |
+
"step": 34,
|
| 637 |
+
"step_time": 0.22859811600000057
|
| 638 |
+
},
|
| 639 |
+
{
|
| 640 |
+
"clip_ratio/high_max": 0.0065104165114462376,
|
| 641 |
+
"clip_ratio/high_mean": 0.0065104165114462376,
|
| 642 |
+
"clip_ratio/low_mean": 0.0013020833721384406,
|
| 643 |
+
"clip_ratio/low_min": 0.0013020833721384406,
|
| 644 |
+
"clip_ratio/region_mean": 0.0078125,
|
| 645 |
+
"entropy": 0.9673511385917664,
|
| 646 |
+
"epoch": 1.4583333333333333,
|
| 647 |
+
"grad_norm": 2.4296762943267822,
|
| 648 |
+
"kl": 0.1084410771727562,
|
| 649 |
+
"learning_rate": 1.6000000000000001e-06,
|
| 650 |
+
"loss": -0.05158400535583496,
|
| 651 |
+
"step": 35,
|
| 652 |
+
"step_time": 0.23024993899997526
|
| 653 |
+
},
|
| 654 |
+
{
|
| 655 |
+
"clip_ratio/high_max": 0.00390625,
|
| 656 |
+
"clip_ratio/high_mean": 0.00390625,
|
| 657 |
+
"clip_ratio/low_mean": 0.0032552082557231188,
|
| 658 |
+
"clip_ratio/low_min": 0.0032552082557231188,
|
| 659 |
+
"clip_ratio/region_mean": 0.0071614584885537624,
|
| 660 |
+
"entropy": 0.983039915561676,
|
| 661 |
+
"epoch": 1.5,
|
| 662 |
+
"grad_norm": 1.957944631576538,
|
| 663 |
+
"kl": 0.13104547560214996,
|
| 664 |
+
"learning_rate": 1.5e-06,
|
| 665 |
+
"loss": 0.004632837139070034,
|
| 666 |
+
"step": 36,
|
| 667 |
+
"step_time": 0.2287184080000202
|
| 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": 1.426942229270935,
|
| 683 |
+
"epoch": 1.5416666666666665,
|
| 684 |
+
"frac_reward_zero_std": 0.25,
|
| 685 |
+
"grad_norm": 1.2855005264282227,
|
| 686 |
+
"kl": 0.14382179081439972,
|
| 687 |
+
"learning_rate": 1.4000000000000001e-06,
|
| 688 |
+
"loss": -0.0295367781072855,
|
| 689 |
+
"num_tokens": 252640.0,
|
| 690 |
+
"reward": -0.07735294103622437,
|
| 691 |
+
"reward_std": 0.3284520208835602,
|
| 692 |
+
"rewards/GeneratorRewardFunction/mean": -0.07735294103622437,
|
| 693 |
+
"rewards/GeneratorRewardFunction/std": 0.32845205068588257,
|
| 694 |
+
"step": 37,
|
| 695 |
+
"step_time": 13.249790346999987
|
| 696 |
+
},
|
| 697 |
+
{
|
| 698 |
+
"clip_ratio/high_max": 0.0,
|
| 699 |
+
"clip_ratio/high_mean": 0.0,
|
| 700 |
+
"clip_ratio/low_mean": 0.0013020833721384406,
|
| 701 |
+
"clip_ratio/low_min": 0.0013020833721384406,
|
| 702 |
+
"clip_ratio/region_mean": 0.0013020833721384406,
|
| 703 |
+
"entropy": 1.1062594652175903,
|
| 704 |
+
"epoch": 1.5833333333333335,
|
| 705 |
+
"grad_norm": 1.2463343143463135,
|
| 706 |
+
"kl": 0.14914868772029877,
|
| 707 |
+
"learning_rate": 1.3e-06,
|
| 708 |
+
"loss": 0.06158822774887085,
|
| 709 |
+
"step": 38,
|
| 710 |
+
"step_time": 0.22136241700002302
|
| 711 |
+
},
|
| 712 |
+
{
|
| 713 |
+
"clip_ratio/high_max": 0.0013020833721384406,
|
| 714 |
+
"clip_ratio/high_mean": 0.0013020833721384406,
|
| 715 |
+
"clip_ratio/low_mean": 0.0006510416860692203,
|
| 716 |
+
"clip_ratio/low_min": 0.0006510416860692203,
|
| 717 |
+
"clip_ratio/region_mean": 0.001953125,
|
| 718 |
+
"entropy": 1.3345317840576172,
|
| 719 |
+
"epoch": 1.625,
|
| 720 |
+
"grad_norm": 2.388456106185913,
|
| 721 |
+
"kl": 0.1212289109826088,
|
| 722 |
+
"learning_rate": 1.2000000000000002e-06,
|
| 723 |
+
"loss": 0.013628202490508556,
|
| 724 |
+
"step": 39,
|
| 725 |
+
"step_time": 0.22244895000000042
|
| 726 |
+
},
|
| 727 |
+
{
|
| 728 |
+
"clip_ratio/high_max": 0.0026041667442768812,
|
| 729 |
+
"clip_ratio/high_mean": 0.0026041667442768812,
|
| 730 |
+
"clip_ratio/low_mean": 0.0006510416860692203,
|
| 731 |
+
"clip_ratio/low_min": 0.0006510416860692203,
|
| 732 |
+
"clip_ratio/region_mean": 0.0032552082557231188,
|
| 733 |
+
"entropy": 0.9387586712837219,
|
| 734 |
+
"epoch": 1.6666666666666665,
|
| 735 |
+
"grad_norm": 2.4696860313415527,
|
| 736 |
+
"kl": 0.1326405256986618,
|
| 737 |
+
"learning_rate": 1.1e-06,
|
| 738 |
+
"loss": -0.0440821647644043,
|
| 739 |
+
"step": 40,
|
| 740 |
+
"step_time": 0.22124087500003498
|
| 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 sha256:7f1bb3fea31b76835f54fffde7e1eeacafdd13f1ca40601af302caf5d8275af4
|
| 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:eeb016f10d80583c7030f8924276f5af074ee87a36483989ee09deeb02394767
|
| 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": 3.343350648880005,
|
| 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": 12.520110770000002
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"clip_ratio/high_max": 0.0045572915114462376,
|
| 42 |
+
"clip_ratio/high_mean": 0.0045572915114462376,
|
| 43 |
+
"clip_ratio/low_mean": 0.0013020833721384406,
|
| 44 |
+
"clip_ratio/low_min": 0.0013020833721384406,
|
| 45 |
+
"clip_ratio/region_mean": 0.005859375,
|
| 46 |
+
"entropy": 1.2710224390029907,
|
| 47 |
+
"epoch": 0.08333333333333333,
|
| 48 |
+
"grad_norm": 2.8392181396484375,
|
| 49 |
+
"kl": 0.001467077643610537,
|
| 50 |
+
"learning_rate": 4.9000000000000005e-06,
|
| 51 |
+
"loss": -0.06676606088876724,
|
| 52 |
+
"step": 2,
|
| 53 |
+
"step_time": 0.22065511100001345
|
| 54 |
+
},
|
| 55 |
+
{
|
| 56 |
+
"clip_ratio/high_max": 0.013671875,
|
| 57 |
+
"clip_ratio/high_mean": 0.013671875,
|
| 58 |
+
"clip_ratio/low_mean": 0.014322916977107525,
|
| 59 |
+
"clip_ratio/low_min": 0.014322916977107525,
|
| 60 |
+
"clip_ratio/region_mean": 0.02799479104578495,
|
| 61 |
+
"entropy": 1.871756911277771,
|
| 62 |
+
"epoch": 0.125,
|
| 63 |
+
"grad_norm": 2.4508721828460693,
|
| 64 |
+
"kl": 0.004840313456952572,
|
| 65 |
+
"learning_rate": 4.800000000000001e-06,
|
| 66 |
+
"loss": 0.010330882854759693,
|
| 67 |
+
"step": 3,
|
| 68 |
+
"step_time": 0.21905603899998027
|
| 69 |
+
},
|
| 70 |
+
{
|
| 71 |
+
"clip_ratio/high_max": 0.01822916604578495,
|
| 72 |
+
"clip_ratio/high_mean": 0.01822916604578495,
|
| 73 |
+
"clip_ratio/low_mean": 0.010416666977107525,
|
| 74 |
+
"clip_ratio/low_min": 0.010416666977107525,
|
| 75 |
+
"clip_ratio/region_mean": 0.02864583395421505,
|
| 76 |
+
"entropy": 1.1871482133865356,
|
| 77 |
+
"epoch": 0.16666666666666666,
|
| 78 |
+
"grad_norm": 1.5818687677383423,
|
| 79 |
+
"kl": 0.005701068323105574,
|
| 80 |
+
"learning_rate": 4.7e-06,
|
| 81 |
+
"loss": -0.08211664110422134,
|
| 82 |
+
"step": 4,
|
| 83 |
+
"step_time": 0.21982950500000698
|
| 84 |
+
},
|
| 85 |
+
{
|
| 86 |
+
"clip_ratio/high_max": 0.0006510416860692203,
|
| 87 |
+
"clip_ratio/high_mean": 0.0006510416860692203,
|
| 88 |
+
"clip_ratio/low_mean": 0.0006510416860692203,
|
| 89 |
+
"clip_ratio/low_min": 0.0006510416860692203,
|
| 90 |
+
"clip_ratio/region_mean": 0.0013020833721384406,
|
| 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": 1.0733331441879272,
|
| 99 |
+
"epoch": 0.20833333333333334,
|
| 100 |
+
"frac_reward_zero_std": 0.0,
|
| 101 |
+
"grad_norm": 3.4188177585601807,
|
| 102 |
+
"kl": 0.018397731706500053,
|
| 103 |
+
"learning_rate": 4.600000000000001e-06,
|
| 104 |
+
"loss": 0.14518485963344574,
|
| 105 |
+
"num_tokens": 50440.0,
|
| 106 |
+
"reward": -0.2228125035762787,
|
| 107 |
+
"reward_std": 0.274566113948822,
|
| 108 |
+
"rewards/GeneratorRewardFunction/mean": -0.2228125035762787,
|
| 109 |
+
"rewards/GeneratorRewardFunction/std": 0.2745661437511444,
|
| 110 |
+
"step": 5,
|
| 111 |
+
"step_time": 12.121955963999994
|
| 112 |
+
},
|
| 113 |
+
{
|
| 114 |
+
"clip_ratio/high_max": 0.0052083334885537624,
|
| 115 |
+
"clip_ratio/high_mean": 0.0052083334885537624,
|
| 116 |
+
"clip_ratio/low_mean": 0.001953125,
|
| 117 |
+
"clip_ratio/low_min": 0.001953125,
|
| 118 |
+
"clip_ratio/region_mean": 0.0071614584885537624,
|
| 119 |
+
"entropy": 1.327884316444397,
|
| 120 |
+
"epoch": 0.25,
|
| 121 |
+
"grad_norm": 2.6934618949890137,
|
| 122 |
+
"kl": 0.01829482428729534,
|
| 123 |
+
"learning_rate": 4.5e-06,
|
| 124 |
+
"loss": -0.037107061594724655,
|
| 125 |
+
"step": 6,
|
| 126 |
+
"step_time": 0.22655425900001092
|
| 127 |
+
},
|
| 128 |
+
{
|
| 129 |
+
"clip_ratio/high_max": 0.0065104165114462376,
|
| 130 |
+
"clip_ratio/high_mean": 0.0065104165114462376,
|
| 131 |
+
"clip_ratio/low_mean": 0.0052083334885537624,
|
| 132 |
+
"clip_ratio/low_min": 0.0052083334885537624,
|
| 133 |
+
"clip_ratio/region_mean": 0.01171875,
|
| 134 |
+
"entropy": 1.2937031984329224,
|
| 135 |
+
"epoch": 0.2916666666666667,
|
| 136 |
+
"grad_norm": 2.8983969688415527,
|
| 137 |
+
"kl": 0.021993428468704224,
|
| 138 |
+
"learning_rate": 4.4e-06,
|
| 139 |
+
"loss": 0.013194209896028042,
|
| 140 |
+
"step": 7,
|
| 141 |
+
"step_time": 0.22578505299998142
|
| 142 |
+
},
|
| 143 |
+
{
|
| 144 |
+
"clip_ratio/high_max": 0.029296875,
|
| 145 |
+
"clip_ratio/high_mean": 0.029296875,
|
| 146 |
+
"clip_ratio/low_mean": 0.0,
|
| 147 |
+
"clip_ratio/low_min": 0.0,
|
| 148 |
+
"clip_ratio/region_mean": 0.029296875,
|
| 149 |
+
"entropy": 0.7918106913566589,
|
| 150 |
+
"epoch": 0.3333333333333333,
|
| 151 |
+
"grad_norm": 1.268328309059143,
|
| 152 |
+
"kl": 0.03330208733677864,
|
| 153 |
+
"learning_rate": 4.3e-06,
|
| 154 |
+
"loss": -0.12033451348543167,
|
| 155 |
+
"step": 8,
|
| 156 |
+
"step_time": 0.2253496229999996
|
| 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": 1.0043877363204956,
|
| 172 |
+
"epoch": 0.375,
|
| 173 |
+
"frac_reward_zero_std": 0.25,
|
| 174 |
+
"grad_norm": 1.3302299976348877,
|
| 175 |
+
"kl": 0.02948068641126156,
|
| 176 |
+
"learning_rate": 4.2000000000000004e-06,
|
| 177 |
+
"loss": -0.07878098636865616,
|
| 178 |
+
"num_tokens": 75884.0,
|
| 179 |
+
"reward": -0.12250000238418579,
|
| 180 |
+
"reward_std": 0.21038061380386353,
|
| 181 |
+
"rewards/GeneratorRewardFunction/mean": -0.12250000238418579,
|
| 182 |
+
"rewards/GeneratorRewardFunction/std": 0.21038061380386353,
|
| 183 |
+
"step": 9,
|
| 184 |
+
"step_time": 11.949294435000013
|
| 185 |
+
},
|
| 186 |
+
{
|
| 187 |
+
"clip_ratio/high_max": 0.010416666977107525,
|
| 188 |
+
"clip_ratio/high_mean": 0.010416666977107525,
|
| 189 |
+
"clip_ratio/low_mean": 0.0032552082557231188,
|
| 190 |
+
"clip_ratio/low_min": 0.0032552082557231188,
|
| 191 |
+
"clip_ratio/region_mean": 0.013671875,
|
| 192 |
+
"entropy": 1.1474007368087769,
|
| 193 |
+
"epoch": 0.4166666666666667,
|
| 194 |
+
"grad_norm": 2.1202971935272217,
|
| 195 |
+
"kl": 0.03631855919957161,
|
| 196 |
+
"learning_rate": 4.1e-06,
|
| 197 |
+
"loss": 0.03980601206421852,
|
| 198 |
+
"step": 10,
|
| 199 |
+
"step_time": 0.22497136300000875
|
| 200 |
+
},
|
| 201 |
+
{
|
| 202 |
+
"clip_ratio/high_max": 0.008463541977107525,
|
| 203 |
+
"clip_ratio/high_mean": 0.008463541977107525,
|
| 204 |
+
"clip_ratio/low_mean": 0.00390625,
|
| 205 |
+
"clip_ratio/low_min": 0.00390625,
|
| 206 |
+
"clip_ratio/region_mean": 0.012369791977107525,
|
| 207 |
+
"entropy": 0.9981658458709717,
|
| 208 |
+
"epoch": 0.4583333333333333,
|
| 209 |
+
"grad_norm": 2.094111680984497,
|
| 210 |
+
"kl": 0.049915943294763565,
|
| 211 |
+
"learning_rate": 4.000000000000001e-06,
|
| 212 |
+
"loss": 0.06331142038106918,
|
| 213 |
+
"step": 11,
|
| 214 |
+
"step_time": 0.22518257699999822
|
| 215 |
+
},
|
| 216 |
+
{
|
| 217 |
+
"clip_ratio/high_max": 0.0403645820915699,
|
| 218 |
+
"clip_ratio/high_mean": 0.0403645820915699,
|
| 219 |
+
"clip_ratio/low_mean": 0.0065104165114462376,
|
| 220 |
+
"clip_ratio/low_min": 0.0065104165114462376,
|
| 221 |
+
"clip_ratio/region_mean": 0.046875,
|
| 222 |
+
"entropy": 1.3572144508361816,
|
| 223 |
+
"epoch": 0.5,
|
| 224 |
+
"grad_norm": 2.6413395404815674,
|
| 225 |
+
"kl": 0.053241848945617676,
|
| 226 |
+
"learning_rate": 3.900000000000001e-06,
|
| 227 |
+
"loss": -0.022430941462516785,
|
| 228 |
+
"step": 12,
|
| 229 |
+
"step_time": 0.2254562319999991
|
| 230 |
+
},
|
| 231 |
+
{
|
| 232 |
+
"clip_ratio/high_max": 0.0,
|
| 233 |
+
"clip_ratio/high_mean": 0.0,
|
| 234 |
+
"clip_ratio/low_mean": 0.001953125,
|
| 235 |
+
"clip_ratio/low_min": 0.001953125,
|
| 236 |
+
"clip_ratio/region_mean": 0.001953125,
|
| 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": 1.629288673400879,
|
| 245 |
+
"epoch": 0.5416666666666666,
|
| 246 |
+
"frac_reward_zero_std": 0.25,
|
| 247 |
+
"grad_norm": 2.107161283493042,
|
| 248 |
+
"kl": 0.0636429563164711,
|
| 249 |
+
"learning_rate": 3.8000000000000005e-06,
|
| 250 |
+
"loss": 0.11612584441900253,
|
| 251 |
+
"num_tokens": 101112.0,
|
| 252 |
+
"reward": -0.11937499791383743,
|
| 253 |
+
"reward_std": 0.20747588574886322,
|
| 254 |
+
"rewards/GeneratorRewardFunction/mean": -0.11937499791383743,
|
| 255 |
+
"rewards/GeneratorRewardFunction/std": 0.20747590065002441,
|
| 256 |
+
"step": 13,
|
| 257 |
+
"step_time": 12.014855219999987
|
| 258 |
+
},
|
| 259 |
+
{
|
| 260 |
+
"clip_ratio/high_max": 0.0032552082557231188,
|
| 261 |
+
"clip_ratio/high_mean": 0.0032552082557231188,
|
| 262 |
+
"clip_ratio/low_mean": 0.0006510416860692203,
|
| 263 |
+
"clip_ratio/low_min": 0.0006510416860692203,
|
| 264 |
+
"clip_ratio/region_mean": 0.00390625,
|
| 265 |
+
"entropy": 1.4905215501785278,
|
| 266 |
+
"epoch": 0.5833333333333334,
|
| 267 |
+
"grad_norm": 1.75613272190094,
|
| 268 |
+
"kl": 0.060588542371988297,
|
| 269 |
+
"learning_rate": 3.7e-06,
|
| 270 |
+
"loss": 0.0006357845850288868,
|
| 271 |
+
"step": 14,
|
| 272 |
+
"step_time": 0.21892069699998729
|
| 273 |
+
},
|
| 274 |
+
{
|
| 275 |
+
"clip_ratio/high_max": 0.0071614584885537624,
|
| 276 |
+
"clip_ratio/high_mean": 0.0071614584885537624,
|
| 277 |
+
"clip_ratio/low_mean": 0.009114583022892475,
|
| 278 |
+
"clip_ratio/low_min": 0.009114583022892475,
|
| 279 |
+
"clip_ratio/region_mean": 0.01627604104578495,
|
| 280 |
+
"entropy": 1.2682157754898071,
|
| 281 |
+
"epoch": 0.625,
|
| 282 |
+
"grad_norm": 2.94674015045166,
|
| 283 |
+
"kl": 0.08301883935928345,
|
| 284 |
+
"learning_rate": 3.6000000000000003e-06,
|
| 285 |
+
"loss": -0.0923055037856102,
|
| 286 |
+
"step": 15,
|
| 287 |
+
"step_time": 0.21824655300000018
|
| 288 |
+
},
|
| 289 |
+
{
|
| 290 |
+
"clip_ratio/high_max": 0.008463541977107525,
|
| 291 |
+
"clip_ratio/high_mean": 0.008463541977107525,
|
| 292 |
+
"clip_ratio/low_mean": 0.01627604104578495,
|
| 293 |
+
"clip_ratio/low_min": 0.01627604104578495,
|
| 294 |
+
"clip_ratio/region_mean": 0.02473958395421505,
|
| 295 |
+
"entropy": 0.9931669235229492,
|
| 296 |
+
"epoch": 0.6666666666666666,
|
| 297 |
+
"grad_norm": 2.1109514236450195,
|
| 298 |
+
"kl": 0.09274417906999588,
|
| 299 |
+
"learning_rate": 3.5e-06,
|
| 300 |
+
"loss": -0.0218100156635046,
|
| 301 |
+
"step": 16,
|
| 302 |
+
"step_time": 0.21831970400000955
|
| 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": 1.376638412475586,
|
| 318 |
+
"epoch": 0.7083333333333334,
|
| 319 |
+
"frac_reward_zero_std": 0.0,
|
| 320 |
+
"grad_norm": 2.625627279281616,
|
| 321 |
+
"kl": 0.07451707124710083,
|
| 322 |
+
"learning_rate": 3.4000000000000005e-06,
|
| 323 |
+
"loss": 0.02886168472468853,
|
| 324 |
+
"num_tokens": 126436.0,
|
| 325 |
+
"reward": -0.09125000238418579,
|
| 326 |
+
"reward_std": 0.1655445545911789,
|
| 327 |
+
"rewards/GeneratorRewardFunction/mean": -0.09125000238418579,
|
| 328 |
+
"rewards/GeneratorRewardFunction/std": 0.1655445545911789,
|
| 329 |
+
"step": 17,
|
| 330 |
+
"step_time": 12.074089973000014
|
| 331 |
+
},
|
| 332 |
+
{
|
| 333 |
+
"clip_ratio/high_max": 0.0006510416860692203,
|
| 334 |
+
"clip_ratio/high_mean": 0.0006510416860692203,
|
| 335 |
+
"clip_ratio/low_mean": 0.001953125,
|
| 336 |
+
"clip_ratio/low_min": 0.001953125,
|
| 337 |
+
"clip_ratio/region_mean": 0.0026041667442768812,
|
| 338 |
+
"entropy": 0.8447733521461487,
|
| 339 |
+
"epoch": 0.75,
|
| 340 |
+
"grad_norm": 2.2611021995544434,
|
| 341 |
+
"kl": 0.08117184042930603,
|
| 342 |
+
"learning_rate": 3.3000000000000006e-06,
|
| 343 |
+
"loss": -0.003659568028524518,
|
| 344 |
+
"step": 18,
|
| 345 |
+
"step_time": 0.22418350800001008
|
| 346 |
+
},
|
| 347 |
+
{
|
| 348 |
+
"clip_ratio/high_max": 0.0065104165114462376,
|
| 349 |
+
"clip_ratio/high_mean": 0.0065104165114462376,
|
| 350 |
+
"clip_ratio/low_mean": 0.0006510416860692203,
|
| 351 |
+
"clip_ratio/low_min": 0.0006510416860692203,
|
| 352 |
+
"clip_ratio/region_mean": 0.0071614584885537624,
|
| 353 |
+
"entropy": 0.9943304061889648,
|
| 354 |
+
"epoch": 0.7916666666666666,
|
| 355 |
+
"grad_norm": 1.5852197408676147,
|
| 356 |
+
"kl": 0.08660884946584702,
|
| 357 |
+
"learning_rate": 3.2000000000000003e-06,
|
| 358 |
+
"loss": -0.10765092819929123,
|
| 359 |
+
"step": 19,
|
| 360 |
+
"step_time": 0.224853281999998
|
| 361 |
+
},
|
| 362 |
+
{
|
| 363 |
+
"clip_ratio/high_max": 0.008463541977107525,
|
| 364 |
+
"clip_ratio/high_mean": 0.008463541977107525,
|
| 365 |
+
"clip_ratio/low_mean": 0.02213541604578495,
|
| 366 |
+
"clip_ratio/low_min": 0.02213541604578495,
|
| 367 |
+
"clip_ratio/region_mean": 0.03059895895421505,
|
| 368 |
+
"entropy": 1.2907896041870117,
|
| 369 |
+
"epoch": 0.8333333333333334,
|
| 370 |
+
"grad_norm": 2.97239089012146,
|
| 371 |
+
"kl": 0.08734595775604248,
|
| 372 |
+
"learning_rate": 3.1000000000000004e-06,
|
| 373 |
+
"loss": 0.08410018682479858,
|
| 374 |
+
"step": 20,
|
| 375 |
+
"step_time": 0.22594529400001306
|
| 376 |
+
},
|
| 377 |
+
{
|
| 378 |
+
"clip_ratio/high_max": 0.0026041667442768812,
|
| 379 |
+
"clip_ratio/high_mean": 0.0026041667442768812,
|
| 380 |
+
"clip_ratio/low_mean": 0.0,
|
| 381 |
+
"clip_ratio/low_min": 0.0,
|
| 382 |
+
"clip_ratio/region_mean": 0.0026041667442768812,
|
| 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": 1.3083521127700806,
|
| 391 |
+
"epoch": 0.875,
|
| 392 |
+
"frac_reward_zero_std": 0.25,
|
| 393 |
+
"grad_norm": 2.6607460975646973,
|
| 394 |
+
"kl": 0.08000912517309189,
|
| 395 |
+
"learning_rate": 3e-06,
|
| 396 |
+
"loss": -0.11818201094865799,
|
| 397 |
+
"num_tokens": 151596.0,
|
| 398 |
+
"reward": -0.22624999284744263,
|
| 399 |
+
"reward_std": 0.2758048474788666,
|
| 400 |
+
"rewards/GeneratorRewardFunction/mean": -0.22624999284744263,
|
| 401 |
+
"rewards/GeneratorRewardFunction/std": 0.27580487728118896,
|
| 402 |
+
"step": 21,
|
| 403 |
+
"step_time": 11.957008280999986
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"clip_ratio/high_max": 0.0013020833721384406,
|
| 407 |
+
"clip_ratio/high_mean": 0.0013020833721384406,
|
| 408 |
+
"clip_ratio/low_mean": 0.0013020833721384406,
|
| 409 |
+
"clip_ratio/low_min": 0.0013020833721384406,
|
| 410 |
+
"clip_ratio/region_mean": 0.0026041667442768812,
|
| 411 |
+
"entropy": 1.2017608880996704,
|
| 412 |
+
"epoch": 0.9166666666666666,
|
| 413 |
+
"grad_norm": 0.8747857809066772,
|
| 414 |
+
"kl": 0.11337386816740036,
|
| 415 |
+
"learning_rate": 2.9e-06,
|
| 416 |
+
"loss": -0.003786882385611534,
|
| 417 |
+
"step": 22,
|
| 418 |
+
"step_time": 0.221026849999987
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"clip_ratio/high_max": 0.0013020833721384406,
|
| 422 |
+
"clip_ratio/high_mean": 0.0013020833721384406,
|
| 423 |
+
"clip_ratio/low_mean": 0.001953125,
|
| 424 |
+
"clip_ratio/low_min": 0.001953125,
|
| 425 |
+
"clip_ratio/region_mean": 0.0032552082557231188,
|
| 426 |
+
"entropy": 1.107405662536621,
|
| 427 |
+
"epoch": 0.9583333333333334,
|
| 428 |
+
"grad_norm": 2.115562915802002,
|
| 429 |
+
"kl": 0.09759091585874557,
|
| 430 |
+
"learning_rate": 2.8000000000000003e-06,
|
| 431 |
+
"loss": 0.08772162348031998,
|
| 432 |
+
"step": 23,
|
| 433 |
+
"step_time": 0.22019210400000588
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"clip_ratio/high_max": 0.005859375,
|
| 437 |
+
"clip_ratio/high_mean": 0.005859375,
|
| 438 |
+
"clip_ratio/low_mean": 0.00390625,
|
| 439 |
+
"clip_ratio/low_min": 0.00390625,
|
| 440 |
+
"clip_ratio/region_mean": 0.009765625,
|
| 441 |
+
"entropy": 0.7833542823791504,
|
| 442 |
+
"epoch": 1.0,
|
| 443 |
+
"grad_norm": 1.686574101448059,
|
| 444 |
+
"kl": 0.10800782591104507,
|
| 445 |
+
"learning_rate": 2.7000000000000004e-06,
|
| 446 |
+
"loss": 0.03493640199303627,
|
| 447 |
+
"step": 24,
|
| 448 |
+
"step_time": 0.22060076099998582
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"clip_ratio/high_max": 0.0013020833721384406,
|
| 452 |
+
"clip_ratio/high_mean": 0.0013020833721384406,
|
| 453 |
+
"clip_ratio/low_mean": 0.0,
|
| 454 |
+
"clip_ratio/low_min": 0.0,
|
| 455 |
+
"clip_ratio/region_mean": 0.0013020833721384406,
|
| 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": 1.0690312385559082,
|
| 464 |
+
"epoch": 1.0416666666666667,
|
| 465 |
+
"frac_reward_zero_std": 0.0,
|
| 466 |
+
"grad_norm": 2.565182685852051,
|
| 467 |
+
"kl": 0.12982706725597382,
|
| 468 |
+
"learning_rate": 2.6e-06,
|
| 469 |
+
"loss": 0.12511824071407318,
|
| 470 |
+
"num_tokens": 177212.0,
|
| 471 |
+
"reward": -0.12406250089406967,
|
| 472 |
+
"reward_std": 0.19458690285682678,
|
| 473 |
+
"rewards/GeneratorRewardFunction/mean": -0.12406250089406967,
|
| 474 |
+
"rewards/GeneratorRewardFunction/std": 0.19458691775798798,
|
| 475 |
+
"step": 25,
|
| 476 |
+
"step_time": 12.125360838999995
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"clip_ratio/high_max": 0.00390625,
|
| 480 |
+
"clip_ratio/high_mean": 0.00390625,
|
| 481 |
+
"clip_ratio/low_mean": 0.0013020833721384406,
|
| 482 |
+
"clip_ratio/low_min": 0.0013020833721384406,
|
| 483 |
+
"clip_ratio/region_mean": 0.0052083334885537624,
|
| 484 |
+
"entropy": 0.8722183108329773,
|
| 485 |
+
"epoch": 1.0833333333333333,
|
| 486 |
+
"grad_norm": 2.401808261871338,
|
| 487 |
+
"kl": 0.12285982817411423,
|
| 488 |
+
"learning_rate": 2.5e-06,
|
| 489 |
+
"loss": -0.13922104239463806,
|
| 490 |
+
"step": 26,
|
| 491 |
+
"step_time": 0.22810240500001555
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"clip_ratio/high_max": 0.0052083334885537624,
|
| 495 |
+
"clip_ratio/high_mean": 0.0052083334885537624,
|
| 496 |
+
"clip_ratio/low_mean": 0.005859375,
|
| 497 |
+
"clip_ratio/low_min": 0.005859375,
|
| 498 |
+
"clip_ratio/region_mean": 0.011067708022892475,
|
| 499 |
+
"entropy": 1.3027639389038086,
|
| 500 |
+
"epoch": 1.125,
|
| 501 |
+
"grad_norm": 1.7678114175796509,
|
| 502 |
+
"kl": 0.10112806409597397,
|
| 503 |
+
"learning_rate": 2.4000000000000003e-06,
|
| 504 |
+
"loss": -0.0586722195148468,
|
| 505 |
+
"step": 27,
|
| 506 |
+
"step_time": 0.22764332500003093
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"clip_ratio/high_max": 0.0065104165114462376,
|
| 510 |
+
"clip_ratio/high_mean": 0.0065104165114462376,
|
| 511 |
+
"clip_ratio/low_mean": 0.0071614584885537624,
|
| 512 |
+
"clip_ratio/low_min": 0.0071614584885537624,
|
| 513 |
+
"clip_ratio/region_mean": 0.013671875,
|
| 514 |
+
"entropy": 0.9790509343147278,
|
| 515 |
+
"epoch": 1.1666666666666667,
|
| 516 |
+
"grad_norm": 1.9319959878921509,
|
| 517 |
+
"kl": 0.11484679579734802,
|
| 518 |
+
"learning_rate": 2.3000000000000004e-06,
|
| 519 |
+
"loss": 0.07509768754243851,
|
| 520 |
+
"step": 28,
|
| 521 |
+
"step_time": 0.22808939600002986
|
| 522 |
+
},
|
| 523 |
+
{
|
| 524 |
+
"clip_ratio/high_max": 0.0006510416860692203,
|
| 525 |
+
"clip_ratio/high_mean": 0.0006510416860692203,
|
| 526 |
+
"clip_ratio/low_mean": 0.0,
|
| 527 |
+
"clip_ratio/low_min": 0.0,
|
| 528 |
+
"clip_ratio/region_mean": 0.0006510416860692203,
|
| 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": 0.9768911004066467,
|
| 537 |
+
"epoch": 1.2083333333333333,
|
| 538 |
+
"frac_reward_zero_std": 0.0,
|
| 539 |
+
"grad_norm": 2.2773685455322266,
|
| 540 |
+
"kl": 0.10255210846662521,
|
| 541 |
+
"learning_rate": 2.2e-06,
|
| 542 |
+
"loss": -0.01104909647256136,
|
| 543 |
+
"num_tokens": 202200.0,
|
| 544 |
+
"reward": -0.13343749940395355,
|
| 545 |
+
"reward_std": 0.2115633636713028,
|
| 546 |
+
"rewards/GeneratorRewardFunction/mean": -0.13343749940395355,
|
| 547 |
+
"rewards/GeneratorRewardFunction/std": 0.211563378572464,
|
| 548 |
+
"step": 29,
|
| 549 |
+
"step_time": 11.981290445000013
|
| 550 |
+
},
|
| 551 |
+
{
|
| 552 |
+
"clip_ratio/high_max": 0.0013020833721384406,
|
| 553 |
+
"clip_ratio/high_mean": 0.0013020833721384406,
|
| 554 |
+
"clip_ratio/low_mean": 0.0006510416860692203,
|
| 555 |
+
"clip_ratio/low_min": 0.0006510416860692203,
|
| 556 |
+
"clip_ratio/region_mean": 0.001953125,
|
| 557 |
+
"entropy": 0.9071128368377686,
|
| 558 |
+
"epoch": 1.25,
|
| 559 |
+
"grad_norm": 2.377110004425049,
|
| 560 |
+
"kl": 0.11365322023630142,
|
| 561 |
+
"learning_rate": 2.1000000000000002e-06,
|
| 562 |
+
"loss": 0.07630521804094315,
|
| 563 |
+
"step": 30,
|
| 564 |
+
"step_time": 0.22243629600001213
|
| 565 |
+
},
|
| 566 |
+
{
|
| 567 |
+
"clip_ratio/high_max": 0.0065104165114462376,
|
| 568 |
+
"clip_ratio/high_mean": 0.0065104165114462376,
|
| 569 |
+
"clip_ratio/low_mean": 0.0026041667442768812,
|
| 570 |
+
"clip_ratio/low_min": 0.0026041667442768812,
|
| 571 |
+
"clip_ratio/region_mean": 0.009114583022892475,
|
| 572 |
+
"entropy": 1.3066421747207642,
|
| 573 |
+
"epoch": 1.2916666666666667,
|
| 574 |
+
"grad_norm": 2.6143717765808105,
|
| 575 |
+
"kl": 0.09395217150449753,
|
| 576 |
+
"learning_rate": 2.0000000000000003e-06,
|
| 577 |
+
"loss": -0.023749127984046936,
|
| 578 |
+
"step": 31,
|
| 579 |
+
"step_time": 0.22260347799999636
|
| 580 |
+
},
|
| 581 |
+
{
|
| 582 |
+
"clip_ratio/high_max": 0.005859375,
|
| 583 |
+
"clip_ratio/high_mean": 0.005859375,
|
| 584 |
+
"clip_ratio/low_mean": 0.001953125,
|
| 585 |
+
"clip_ratio/low_min": 0.001953125,
|
| 586 |
+
"clip_ratio/region_mean": 0.0078125,
|
| 587 |
+
"entropy": 1.2596086263656616,
|
| 588 |
+
"epoch": 1.3333333333333333,
|
| 589 |
+
"grad_norm": 1.4453171491622925,
|
| 590 |
+
"kl": 0.09631065279245377,
|
| 591 |
+
"learning_rate": 1.9000000000000002e-06,
|
| 592 |
+
"loss": -0.04094076156616211,
|
| 593 |
+
"step": 32,
|
| 594 |
+
"step_time": 0.2226373689999832
|
| 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": 0.9096196293830872,
|
| 610 |
+
"epoch": 1.375,
|
| 611 |
+
"frac_reward_zero_std": 0.0,
|
| 612 |
+
"grad_norm": 2.4287660121917725,
|
| 613 |
+
"kl": 0.15610061585903168,
|
| 614 |
+
"learning_rate": 1.8000000000000001e-06,
|
| 615 |
+
"loss": 0.060271572321653366,
|
| 616 |
+
"num_tokens": 227556.0,
|
| 617 |
+
"reward": -0.13343749940395355,
|
| 618 |
+
"reward_std": 0.19603331387043,
|
| 619 |
+
"rewards/GeneratorRewardFunction/mean": -0.13343749940395355,
|
| 620 |
+
"rewards/GeneratorRewardFunction/std": 0.19603331387043,
|
| 621 |
+
"step": 33,
|
| 622 |
+
"step_time": 12.116691536000019
|
| 623 |
+
},
|
| 624 |
+
{
|
| 625 |
+
"clip_ratio/high_max": 0.0032552082557231188,
|
| 626 |
+
"clip_ratio/high_mean": 0.0032552082557231188,
|
| 627 |
+
"clip_ratio/low_mean": 0.0006510416860692203,
|
| 628 |
+
"clip_ratio/low_min": 0.0006510416860692203,
|
| 629 |
+
"clip_ratio/region_mean": 0.00390625,
|
| 630 |
+
"entropy": 1.2461239099502563,
|
| 631 |
+
"epoch": 1.4166666666666667,
|
| 632 |
+
"grad_norm": 2.0638527870178223,
|
| 633 |
+
"kl": 0.11391329020261765,
|
| 634 |
+
"learning_rate": 1.7000000000000002e-06,
|
| 635 |
+
"loss": -0.012484799139201641,
|
| 636 |
+
"step": 34,
|
| 637 |
+
"step_time": 0.22859811600000057
|
| 638 |
+
},
|
| 639 |
+
{
|
| 640 |
+
"clip_ratio/high_max": 0.0065104165114462376,
|
| 641 |
+
"clip_ratio/high_mean": 0.0065104165114462376,
|
| 642 |
+
"clip_ratio/low_mean": 0.0013020833721384406,
|
| 643 |
+
"clip_ratio/low_min": 0.0013020833721384406,
|
| 644 |
+
"clip_ratio/region_mean": 0.0078125,
|
| 645 |
+
"entropy": 0.9673511385917664,
|
| 646 |
+
"epoch": 1.4583333333333333,
|
| 647 |
+
"grad_norm": 2.4296762943267822,
|
| 648 |
+
"kl": 0.1084410771727562,
|
| 649 |
+
"learning_rate": 1.6000000000000001e-06,
|
| 650 |
+
"loss": -0.05158400535583496,
|
| 651 |
+
"step": 35,
|
| 652 |
+
"step_time": 0.23024993899997526
|
| 653 |
+
},
|
| 654 |
+
{
|
| 655 |
+
"clip_ratio/high_max": 0.00390625,
|
| 656 |
+
"clip_ratio/high_mean": 0.00390625,
|
| 657 |
+
"clip_ratio/low_mean": 0.0032552082557231188,
|
| 658 |
+
"clip_ratio/low_min": 0.0032552082557231188,
|
| 659 |
+
"clip_ratio/region_mean": 0.0071614584885537624,
|
| 660 |
+
"entropy": 0.983039915561676,
|
| 661 |
+
"epoch": 1.5,
|
| 662 |
+
"grad_norm": 1.957944631576538,
|
| 663 |
+
"kl": 0.13104547560214996,
|
| 664 |
+
"learning_rate": 1.5e-06,
|
| 665 |
+
"loss": 0.004632837139070034,
|
| 666 |
+
"step": 36,
|
| 667 |
+
"step_time": 0.2287184080000202
|
| 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": 1.426942229270935,
|
| 683 |
+
"epoch": 1.5416666666666665,
|
| 684 |
+
"frac_reward_zero_std": 0.25,
|
| 685 |
+
"grad_norm": 1.2855005264282227,
|
| 686 |
+
"kl": 0.14382179081439972,
|
| 687 |
+
"learning_rate": 1.4000000000000001e-06,
|
| 688 |
+
"loss": -0.0295367781072855,
|
| 689 |
+
"num_tokens": 252640.0,
|
| 690 |
+
"reward": -0.07735294103622437,
|
| 691 |
+
"reward_std": 0.3284520208835602,
|
| 692 |
+
"rewards/GeneratorRewardFunction/mean": -0.07735294103622437,
|
| 693 |
+
"rewards/GeneratorRewardFunction/std": 0.32845205068588257,
|
| 694 |
+
"step": 37,
|
| 695 |
+
"step_time": 13.249790346999987
|
| 696 |
+
},
|
| 697 |
+
{
|
| 698 |
+
"clip_ratio/high_max": 0.0,
|
| 699 |
+
"clip_ratio/high_mean": 0.0,
|
| 700 |
+
"clip_ratio/low_mean": 0.0013020833721384406,
|
| 701 |
+
"clip_ratio/low_min": 0.0013020833721384406,
|
| 702 |
+
"clip_ratio/region_mean": 0.0013020833721384406,
|
| 703 |
+
"entropy": 1.1062594652175903,
|
| 704 |
+
"epoch": 1.5833333333333335,
|
| 705 |
+
"grad_norm": 1.2463343143463135,
|
| 706 |
+
"kl": 0.14914868772029877,
|
| 707 |
+
"learning_rate": 1.3e-06,
|
| 708 |
+
"loss": 0.06158822774887085,
|
| 709 |
+
"step": 38,
|
| 710 |
+
"step_time": 0.22136241700002302
|
| 711 |
+
},
|
| 712 |
+
{
|
| 713 |
+
"clip_ratio/high_max": 0.0013020833721384406,
|
| 714 |
+
"clip_ratio/high_mean": 0.0013020833721384406,
|
| 715 |
+
"clip_ratio/low_mean": 0.0006510416860692203,
|
| 716 |
+
"clip_ratio/low_min": 0.0006510416860692203,
|
| 717 |
+
"clip_ratio/region_mean": 0.001953125,
|
| 718 |
+
"entropy": 1.3345317840576172,
|
| 719 |
+
"epoch": 1.625,
|
| 720 |
+
"grad_norm": 2.388456106185913,
|
| 721 |
+
"kl": 0.1212289109826088,
|
| 722 |
+
"learning_rate": 1.2000000000000002e-06,
|
| 723 |
+
"loss": 0.013628202490508556,
|
| 724 |
+
"step": 39,
|
| 725 |
+
"step_time": 0.22244895000000042
|
| 726 |
+
},
|
| 727 |
+
{
|
| 728 |
+
"clip_ratio/high_max": 0.0026041667442768812,
|
| 729 |
+
"clip_ratio/high_mean": 0.0026041667442768812,
|
| 730 |
+
"clip_ratio/low_mean": 0.0006510416860692203,
|
| 731 |
+
"clip_ratio/low_min": 0.0006510416860692203,
|
| 732 |
+
"clip_ratio/region_mean": 0.0032552082557231188,
|
| 733 |
+
"entropy": 0.9387586712837219,
|
| 734 |
+
"epoch": 1.6666666666666665,
|
| 735 |
+
"grad_norm": 2.4696860313415527,
|
| 736 |
+
"kl": 0.1326405256986618,
|
| 737 |
+
"learning_rate": 1.1e-06,
|
| 738 |
+
"loss": -0.0440821647644043,
|
| 739 |
+
"step": 40,
|
| 740 |
+
"step_time": 0.22124087500003498
|
| 741 |
+
},
|
| 742 |
+
{
|
| 743 |
+
"clip_ratio/high_max": 0.0,
|
| 744 |
+
"clip_ratio/high_mean": 0.0,
|
| 745 |
+
"clip_ratio/low_mean": 0.0,
|
| 746 |
+
"clip_ratio/low_min": 0.0,
|
| 747 |
+
"clip_ratio/region_mean": 0.0,
|
| 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": 1.2471710443496704,
|
| 756 |
+
"epoch": 1.7083333333333335,
|
| 757 |
+
"frac_reward_zero_std": 0.5,
|
| 758 |
+
"grad_norm": 2.290759801864624,
|
| 759 |
+
"kl": 0.12401092797517776,
|
| 760 |
+
"learning_rate": 1.0000000000000002e-06,
|
| 761 |
+
"loss": 0.05721667408943176,
|
| 762 |
+
"num_tokens": 277896.0,
|
| 763 |
+
"reward": -0.11687500029802322,
|
| 764 |
+
"reward_std": 0.1675596982240677,
|
| 765 |
+
"rewards/GeneratorRewardFunction/mean": -0.11687500029802322,
|
| 766 |
+
"rewards/GeneratorRewardFunction/std": 0.1675596833229065,
|
| 767 |
+
"step": 41,
|
| 768 |
+
"step_time": 12.131979827999999
|
| 769 |
+
},
|
| 770 |
+
{
|
| 771 |
+
"clip_ratio/high_max": 0.0,
|
| 772 |
+
"clip_ratio/high_mean": 0.0,
|
| 773 |
+
"clip_ratio/low_mean": 0.0,
|
| 774 |
+
"clip_ratio/low_min": 0.0,
|
| 775 |
+
"clip_ratio/region_mean": 0.0,
|
| 776 |
+
"entropy": 1.2584272623062134,
|
| 777 |
+
"epoch": 1.75,
|
| 778 |
+
"grad_norm": 0.04703531414270401,
|
| 779 |
+
"kl": 0.11258962005376816,
|
| 780 |
+
"learning_rate": 9.000000000000001e-07,
|
| 781 |
+
"loss": 0.001895441091619432,
|
| 782 |
+
"step": 42,
|
| 783 |
+
"step_time": 0.2273904870000365
|
| 784 |
+
},
|
| 785 |
+
{
|
| 786 |
+
"clip_ratio/high_max": 0.0,
|
| 787 |
+
"clip_ratio/high_mean": 0.0,
|
| 788 |
+
"clip_ratio/low_mean": 0.0,
|
| 789 |
+
"clip_ratio/low_min": 0.0,
|
| 790 |
+
"clip_ratio/region_mean": 0.0,
|
| 791 |
+
"entropy": 1.1508691310882568,
|
| 792 |
+
"epoch": 1.7916666666666665,
|
| 793 |
+
"grad_norm": 1.4680578708648682,
|
| 794 |
+
"kl": 0.1344568133354187,
|
| 795 |
+
"learning_rate": 8.000000000000001e-07,
|
| 796 |
+
"loss": -0.10274048894643784,
|
| 797 |
+
"step": 43,
|
| 798 |
+
"step_time": 0.22648300899999185
|
| 799 |
+
},
|
| 800 |
+
{
|
| 801 |
+
"clip_ratio/high_max": 0.0006510416860692203,
|
| 802 |
+
"clip_ratio/high_mean": 0.0006510416860692203,
|
| 803 |
+
"clip_ratio/low_mean": 0.0013020833721384406,
|
| 804 |
+
"clip_ratio/low_min": 0.0013020833721384406,
|
| 805 |
+
"clip_ratio/region_mean": 0.001953125,
|
| 806 |
+
"entropy": 1.0707014799118042,
|
| 807 |
+
"epoch": 1.8333333333333335,
|
| 808 |
+
"grad_norm": 2.05830717086792,
|
| 809 |
+
"kl": 0.14121703803539276,
|
| 810 |
+
"learning_rate": 7.000000000000001e-07,
|
| 811 |
+
"loss": 0.045286085456609726,
|
| 812 |
+
"step": 44,
|
| 813 |
+
"step_time": 0.22805016099999875
|
| 814 |
+
},
|
| 815 |
+
{
|
| 816 |
+
"clip_ratio/high_max": 0.0013020833721384406,
|
| 817 |
+
"clip_ratio/high_mean": 0.0013020833721384406,
|
| 818 |
+
"clip_ratio/low_mean": 0.0,
|
| 819 |
+
"clip_ratio/low_min": 0.0,
|
| 820 |
+
"clip_ratio/region_mean": 0.0013020833721384406,
|
| 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": 1.4754749536514282,
|
| 829 |
+
"epoch": 1.875,
|
| 830 |
+
"frac_reward_zero_std": 0.25,
|
| 831 |
+
"grad_norm": 2.129523277282715,
|
| 832 |
+
"kl": 0.15608109533786774,
|
| 833 |
+
"learning_rate": 6.000000000000001e-07,
|
| 834 |
+
"loss": 0.06181947514414787,
|
| 835 |
+
"num_tokens": 303192.0,
|
| 836 |
+
"reward": -0.04625000059604645,
|
| 837 |
+
"reward_std": 0.013964240439236164,
|
| 838 |
+
"rewards/GeneratorRewardFunction/mean": -0.04625000059604645,
|
| 839 |
+
"rewards/GeneratorRewardFunction/std": 0.013964240439236164,
|
| 840 |
+
"step": 45,
|
| 841 |
+
"step_time": 12.076911616000018
|
| 842 |
+
},
|
| 843 |
+
{
|
| 844 |
+
"clip_ratio/high_max": 0.0026041667442768812,
|
| 845 |
+
"clip_ratio/high_mean": 0.0026041667442768812,
|
| 846 |
+
"clip_ratio/low_mean": 0.0,
|
| 847 |
+
"clip_ratio/low_min": 0.0,
|
| 848 |
+
"clip_ratio/region_mean": 0.0026041667442768812,
|
| 849 |
+
"entropy": 1.0166983604431152,
|
| 850 |
+
"epoch": 1.9166666666666665,
|
| 851 |
+
"grad_norm": 2.7428877353668213,
|
| 852 |
+
"kl": 0.16378425061702728,
|
| 853 |
+
"learning_rate": 5.000000000000001e-07,
|
| 854 |
+
"loss": -0.17563432455062866,
|
| 855 |
+
"step": 46,
|
| 856 |
+
"step_time": 0.2192206379999675
|
| 857 |
+
},
|
| 858 |
+
{
|
| 859 |
+
"clip_ratio/high_max": 0.0006510416860692203,
|
| 860 |
+
"clip_ratio/high_mean": 0.0006510416860692203,
|
| 861 |
+
"clip_ratio/low_mean": 0.001953125,
|
| 862 |
+
"clip_ratio/low_min": 0.001953125,
|
| 863 |
+
"clip_ratio/region_mean": 0.0026041667442768812,
|
| 864 |
+
"entropy": 1.2661925554275513,
|
| 865 |
+
"epoch": 1.9583333333333335,
|
| 866 |
+
"grad_norm": 2.3470659255981445,
|
| 867 |
+
"kl": 0.13440369069576263,
|
| 868 |
+
"learning_rate": 4.0000000000000003e-07,
|
| 869 |
+
"loss": 0.08357550948858261,
|
| 870 |
+
"step": 47,
|
| 871 |
+
"step_time": 0.21930876000004673
|
| 872 |
+
},
|
| 873 |
+
{
|
| 874 |
+
"clip_ratio/high_max": 0.0032552082557231188,
|
| 875 |
+
"clip_ratio/high_mean": 0.0032552082557231188,
|
| 876 |
+
"clip_ratio/low_mean": 0.0006510416860692203,
|
| 877 |
+
"clip_ratio/low_min": 0.0006510416860692203,
|
| 878 |
+
"clip_ratio/region_mean": 0.00390625,
|
| 879 |
+
"entropy": 0.8933156132698059,
|
| 880 |
+
"epoch": 2.0,
|
| 881 |
+
"grad_norm": 1.1481467485427856,
|
| 882 |
+
"kl": 0.17724938690662384,
|
| 883 |
+
"learning_rate": 3.0000000000000004e-07,
|
| 884 |
+
"loss": 0.031690943986177444,
|
| 885 |
+
"step": 48,
|
| 886 |
+
"step_time": 0.21845309399998314
|
| 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": 1.071245551109314,
|
| 902 |
+
"epoch": 2.0416666666666665,
|
| 903 |
+
"frac_reward_zero_std": 0.25,
|
| 904 |
+
"grad_norm": 1.4133681058883667,
|
| 905 |
+
"kl": 0.16416768729686737,
|
| 906 |
+
"learning_rate": 2.0000000000000002e-07,
|
| 907 |
+
"loss": -0.07582488656044006,
|
| 908 |
+
"num_tokens": 328804.0,
|
| 909 |
+
"reward": -0.05125000327825546,
|
| 910 |
+
"reward_std": 0.015329709276556969,
|
| 911 |
+
"rewards/GeneratorRewardFunction/mean": -0.05125000327825546,
|
| 912 |
+
"rewards/GeneratorRewardFunction/std": 0.015329709276556969,
|
| 913 |
+
"step": 49,
|
| 914 |
+
"step_time": 12.023586194000018
|
| 915 |
+
},
|
| 916 |
+
{
|
| 917 |
+
"clip_ratio/high_max": 0.0032552082557231188,
|
| 918 |
+
"clip_ratio/high_mean": 0.0032552082557231188,
|
| 919 |
+
"clip_ratio/low_mean": 0.0,
|
| 920 |
+
"clip_ratio/low_min": 0.0,
|
| 921 |
+
"clip_ratio/region_mean": 0.0032552082557231188,
|
| 922 |
+
"entropy": 1.2276629209518433,
|
| 923 |
+
"epoch": 2.0833333333333335,
|
| 924 |
+
"grad_norm": 2.9151129722595215,
|
| 925 |
+
"kl": 0.150254487991333,
|
| 926 |
+
"learning_rate": 1.0000000000000001e-07,
|
| 927 |
+
"loss": -0.024596773087978363,
|
| 928 |
+
"step": 50,
|
| 929 |
+
"step_time": 0.22950664599994752
|
| 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 sha256:7f1bb3fea31b76835f54fffde7e1eeacafdd13f1ca40601af302caf5d8275af4
|
| 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
|