Model Card for OpenCSG-R1-Qwen2.5-Code-3B-V1

This model is a fine-tuned version of [Qwen/Qwen2.5-Coder-3B-Instruct] (https://huggingface.co/Qwen/Qwen2.5-Coder-3B-Instruct) on the [open-r1/OpenThoughts-114k-Code_decontaminated] datasets. It has been trained using TRL.

Quick start

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch
import pandas as pd

model_name = "/data/project/pj/r1/opencsg-r1/open-r1/train/Qwen2.5-3B-Open-R1-Code-GRPO/checkpoint-150"
model = AutoModelForCausalLM.from_pretrained(
 model_name,
 torch_dtype="auto",
 device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=False)
df = pd.read_parquet('/data/project/pj/r1/opencsg-r1/OpenThoughts-114k-Code_decontaminated/train-00000-of-00006.parquet')
data = df['problem'][0]
messages = [
 {
 "role": "user",
 "content": f"Please help me solve the problem: {data}.Output the thinking process within the <think> </think> tags,and then return the final result within the <answer> </answer> tags.",
 },
 {
 "role": "assistant",
 "content": "Let's solve the problem step by step.\n<think>",
 },
]
text = tokenizer.apply_chat_template(
 messages,
 tokenize=False,
 continue_final_message=True,
 # add_generation_prompt=True
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)

generated_ids = model.generate(
 **model_inputs,
 max_new_tokens=1024,
 temperature=0.6
)
generated_ids = [
 output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]

response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
print(response)

This model was trained with GRPO, a method introduced in DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models.

Framework versions

TRL: 0.15.2
Transformers: 4.49.0
Pytorch: 2.5.1
Datasets: 3.3.2
Tokenizers: 0.21.0

Citations

Cite GRPO as:

@article{zhihong2024deepseekmath,
 title = {{DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models}},
 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},
 year = 2024,
 eprint = {arXiv:2402.03300},
}

Cite TRL as:

@misc{vonwerra2022trl,
 title = {{TRL: Transformer Reinforcement Learning}},
 author = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallouédec},
 year = 2020,
 journal = {GitHub repository},
 publisher = {GitHub},
 howpublished = {\url{https://github.com/huggingface/trl}}
}