Instructions to use XingYing-stack/TIPS-Qwen3-4B-Instruct-2507-Math with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use XingYing-stack/TIPS-Qwen3-4B-Instruct-2507-Math with Transformers:

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

pipe = pipeline("text-generation", model="XingYing-stack/TIPS-Qwen3-4B-Instruct-2507-Math")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("XingYing-stack/TIPS-Qwen3-4B-Instruct-2507-Math")
model = AutoModelForCausalLM.from_pretrained("XingYing-stack/TIPS-Qwen3-4B-Instruct-2507-Math", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use XingYing-stack/TIPS-Qwen3-4B-Instruct-2507-Math with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "XingYing-stack/TIPS-Qwen3-4B-Instruct-2507-Math"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "XingYing-stack/TIPS-Qwen3-4B-Instruct-2507-Math",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/XingYing-stack/TIPS-Qwen3-4B-Instruct-2507-Math

SGLang

How to use XingYing-stack/TIPS-Qwen3-4B-Instruct-2507-Math 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 "XingYing-stack/TIPS-Qwen3-4B-Instruct-2507-Math" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "XingYing-stack/TIPS-Qwen3-4B-Instruct-2507-Math",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "XingYing-stack/TIPS-Qwen3-4B-Instruct-2507-Math" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "XingYing-stack/TIPS-Qwen3-4B-Instruct-2507-Math",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use XingYing-stack/TIPS-Qwen3-4B-Instruct-2507-Math with Docker Model Runner:
```
docker model run hf.co/XingYing-stack/TIPS-Qwen3-4B-Instruct-2507-Math
```

Add link to paper

by nielsr HF Staff - opened 10 days ago

base: refs/heads/main

←

from: refs/pr/1

Discussion Files changed

-2

Files changed (1) hide show

README.md +4 -2

README.md CHANGED Viewed

@@ -1,9 +1,9 @@
 ---
-license: apache-2.0
 base_model: Qwen/Qwen3-4B-Instruct-2507
 datasets:
 - XingYing-stack/TIPS-Training-Data
 library_name: transformers
 pipeline_tag: text-generation
 tags:
 - reward-model
@@ -16,8 +16,10 @@ tags:
 This TIPS math checkpoint is initialized from `Qwen/Qwen3-4B-Instruct-2507` and trained with outcome-only GRPO. It is a generative reward model that reasons over a mathematical solution before producing step-level and outcome labels.
 Training data and code are available at https://huggingface.co/datasets/XingYing-stack/TIPS-Training-Data and https://github.com/RUCBM/TIPS.
 Use the prompt templates and evaluation scripts in the TIPS repository. This checkpoint is intended for reward modeling and process verification rather than general-purpose chat.
-Built upon [verl](https://github.com/volcengine/verl) and released under Apache-2.0.

 ---
 base_model: Qwen/Qwen3-4B-Instruct-2507
 datasets:
 - XingYing-stack/TIPS-Training-Data
 library_name: transformers
+license: apache-2.0
 pipeline_tag: text-generation
 tags:
 - reward-model
 This TIPS math checkpoint is initialized from `Qwen/Qwen3-4B-Instruct-2507` and trained with outcome-only GRPO. It is a generative reward model that reasons over a mathematical solution before producing step-level and outcome labels.
+This model is presented in the paper [Inducing Process Supervision from Outcome-Only Reinforcement Learning](https://huggingface.co/papers/2609.36641).
 Training data and code are available at https://huggingface.co/datasets/XingYing-stack/TIPS-Training-Data and https://github.com/RUCBM/TIPS.
 Use the prompt templates and evaluation scripts in the TIPS repository. This checkpoint is intended for reward modeling and process verification rather than general-purpose chat.
+Built upon [verl](https://github.com/volcengine/verl) and released under Apache-2.0.