Instructions to use TingchenFu/coldrl_qwen-2.5-math-7b_04252230 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use TingchenFu/coldrl_qwen-2.5-math-7b_04252230 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="TingchenFu/coldrl_qwen-2.5-math-7b_04252230")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("TingchenFu/coldrl_qwen-2.5-math-7b_04252230")
model = AutoModelForCausalLM.from_pretrained("TingchenFu/coldrl_qwen-2.5-math-7b_04252230")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use TingchenFu/coldrl_qwen-2.5-math-7b_04252230 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "TingchenFu/coldrl_qwen-2.5-math-7b_04252230"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "TingchenFu/coldrl_qwen-2.5-math-7b_04252230",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/TingchenFu/coldrl_qwen-2.5-math-7b_04252230

SGLang

How to use TingchenFu/coldrl_qwen-2.5-math-7b_04252230 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "TingchenFu/coldrl_qwen-2.5-math-7b_04252230" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "TingchenFu/coldrl_qwen-2.5-math-7b_04252230",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "TingchenFu/coldrl_qwen-2.5-math-7b_04252230" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "TingchenFu/coldrl_qwen-2.5-math-7b_04252230",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use TingchenFu/coldrl_qwen-2.5-math-7b_04252230 with Docker Model Runner:
```
docker model run hf.co/TingchenFu/coldrl_qwen-2.5-math-7b_04252230
```

Model Card

cold-RL for mathematical reasoning in our MathIF project.

Github Repository: https://github.com/TingchenFu/MathIF

Training Details

We base our experiments on the DeepScaler dataset, which contains approximately 40k math reasoning samples. The training is conducted using 16 NVIDIA H100 GPUs. For reinforcement learning, we adopt the GRPO framework and use verifiable outcome-based rewards. The model is trained with VeRL framework with most hyper-parameters following the default setting.

Evaluation

We use nucleus sampling (T=1.0, p=0.95) with a maximum generation length of 16,384 tokens for decoding and vLLM engine for efficient inference.

Citation

BibTeX:

@article{fu2025scaling,
  title={Scaling Reasoning, Losing Control: Evaluating Instruction Following in Large Reasoning Models},
  author={Fu, Tingchen and Gu, Jiawei and Li, Yafu and Qu, Xiaoye and Cheng, Yu},
  journal={arXiv preprint arXiv:2505.14810},
  year={2025}
}