Instructions to use jnanliu/LiveMath-Judge with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use jnanliu/LiveMath-Judge with Transformers:

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

pipe = pipeline("text-generation", model="jnanliu/LiveMath-Judge")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

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

tokenizer = AutoTokenizer.from_pretrained("jnanliu/LiveMath-Judge")
model = AutoModel.from_pretrained("jnanliu/LiveMath-Judge", 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 jnanliu/LiveMath-Judge with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/jnanliu/LiveMath-Judge

SGLang

How to use jnanliu/LiveMath-Judge 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 "jnanliu/LiveMath-Judge" \
    --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": "jnanliu/LiveMath-Judge",
		"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 "jnanliu/LiveMath-Judge" \
        --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": "jnanliu/LiveMath-Judge",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use jnanliu/LiveMath-Judge with Docker Model Runner:
```
docker model run hf.co/jnanliu/LiveMath-Judge
```

jnanliu commited on Jan 9, 2025

Commit

7b5f5ba

verified ·

1 Parent(s): 82bb9b5

Update README.md

Browse files

Files changed (1) hide show

README.md +6 -6

README.md CHANGED Viewed

@@ -25,7 +25,7 @@ from transformers import AutoModelForCausalLM, AutoTokenizer
 # load the model and tokenizer
 model = AutoModelForCausalLM.from_pretrained(
     'jnanliu/LiveMath-Judge',
-    device_map="auto",
     torch_dtype=torch.bfloat16,
 )
 tokenizer = AutoTokenizer.from_pretrained(
@@ -66,12 +66,12 @@ Analysis:
 conversations = [
   {'role': 'user', 'content': prompt.format(question=question, gold_answer=golden_answer, answer=generated_answer)}
 ]
-inputs = tokenizer.apply_chat_template(conversations, return_tensors="pt")
 # do inference
 pred = model.generate(
-    input_ids=inputs["input_ids"].to(model.device),
-    attention_mask=inputs["attention_mask"].to(model.device),
     num_return_sequences=1,
 )[0].cpu().tolist()
 response = tokenizer.decode(pred, skip_special_tokens=True)
@@ -84,8 +84,8 @@ response = tokenizer.decode(pred, skip_special_tokens=True)
 Following is the mG-Pass@16 results of Qwen2.5-72B-Instruct-as-Judge and LiveMath-Judge-as-Judge of 10 models.
 |model| Qwen2.5-72B-Instruct | LiveMath-Judge |
 | -- | -- | -- |
-| Qwen2.5-7B-Instruct | 26.45 | 26.32 |
-| Qwen2.5-Math-7B-Instruct | 37.91 | 38.01 |
 | Llama-3.1-8B-Instruct | 10.43 | 10.41 |
 | Llama-3.1-70B-Instruct | 21.37 | 22.12 |
 | Llama-3.3-70B-Instruct | 27.36 | 27.23 |

 # load the model and tokenizer
 model = AutoModelForCausalLM.from_pretrained(
     'jnanliu/LiveMath-Judge',
+    device_map='auto',
     torch_dtype=torch.bfloat16,
 )
 tokenizer = AutoTokenizer.from_pretrained(
 conversations = [
   {'role': 'user', 'content': prompt.format(question=question, gold_answer=golden_answer, answer=generated_answer)}
 ]
+inputs = tokenizer.apply_chat_template(conversations, return_tensors='pt')
 # do inference
 pred = model.generate(
+    input_ids=inputs['input_ids'].to(model.device),
+    attention_mask=inputs['attention_mask'].to(model.device),
     num_return_sequences=1,
 )[0].cpu().tolist()
 response = tokenizer.decode(pred, skip_special_tokens=True)
 Following is the mG-Pass@16 results of Qwen2.5-72B-Instruct-as-Judge and LiveMath-Judge-as-Judge of 10 models.
 |model| Qwen2.5-72B-Instruct | LiveMath-Judge |
 | -- | -- | -- |
+| Qwen2.5-7B-Instruct | 26.45 | 28.17 |
+| Qwen2.5-Math-7B-Instruct | 37.91 | 39.54 |
 | Llama-3.1-8B-Instruct | 10.43 | 10.41 |
 | Llama-3.1-70B-Instruct | 21.37 | 22.12 |
 | Llama-3.3-70B-Instruct | 27.36 | 27.23 |