Instructions to use infly/Universal-PRM-7B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use infly/Universal-PRM-7B with Transformers:

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

pipe = pipeline("text-generation", model="infly/Universal-PRM-7B", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModel

tokenizer = AutoTokenizer.from_pretrained("infly/Universal-PRM-7B", trust_remote_code=True)
model = AutoModel.from_pretrained("infly/Universal-PRM-7B", trust_remote_code=True)
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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use infly/Universal-PRM-7B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "infly/Universal-PRM-7B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "infly/Universal-PRM-7B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/infly/Universal-PRM-7B

SGLang

How to use infly/Universal-PRM-7B 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 "infly/Universal-PRM-7B" \
    --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": "infly/Universal-PRM-7B",
		"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 "infly/Universal-PRM-7B" \
        --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": "infly/Universal-PRM-7B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use infly/Universal-PRM-7B with Docker Model Runner:
```
docker model run hf.co/infly/Universal-PRM-7B
```

MinghaoYang

nielsr HF Staff commited on Feb 19, 2025

Commit

cdc97f5

verified ·

1 Parent(s): 3b068e5

Add library_name, pipeline_tag, and project page link (#2)

Browse files

- Add library_name, pipeline_tag, and project page link (8ef09e4e5b3ac2915c49bd4415f9fab73c2a0b7b)

Co-authored-by: Niels Rogge <nielsr@users.noreply.huggingface.co>

Files changed (1) hide show

README.md +7 -2

README.md CHANGED Viewed

@@ -1,7 +1,13 @@
 ---
 license: apache-2.0
 ---
 # Universal-PRM-7B
 ## 1. Overview
 Universal-PRM is trained using Qwen2.5-Math-7B-Instruct as the base. The training process incorporates diverse policy distributions, ensemble prompting, and reverse verification to enhance generalization and robustness. It achieves state-of-the-art performance on ProcessBench and the internally developed UniversalBench.
 ## 2. Experiments
@@ -75,5 +81,4 @@ with torch.no_grad():
         judge_list_infer.append(reward)
 print(judge_list_infer)     # [0.73828125, 0.7265625, 0.73046875, 0.73828125, 0.734375]
-```

 ---
 license: apache-2.0
+library_name: transformers
+pipeline_tag: text-generation
 ---
 # Universal-PRM-7B
+Project page: https://auroraprm.github.io/
 ## 1. Overview
 Universal-PRM is trained using Qwen2.5-Math-7B-Instruct as the base. The training process incorporates diverse policy distributions, ensemble prompting, and reverse verification to enhance generalization and robustness. It achieves state-of-the-art performance on ProcessBench and the internally developed UniversalBench.
 ## 2. Experiments
         judge_list_infer.append(reward)
 print(judge_list_infer)     # [0.73828125, 0.7265625, 0.73046875, 0.73828125, 0.734375]
+```