Instructions to use Pythagoras-LM/Pythagoras-Prover-SFT-32B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Pythagoras-LM/Pythagoras-Prover-SFT-32B with Transformers:

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

pipe = pipeline("text-generation", model="Pythagoras-LM/Pythagoras-Prover-SFT-32B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Pythagoras-LM/Pythagoras-Prover-SFT-32B")
model = AutoModelForCausalLM.from_pretrained("Pythagoras-LM/Pythagoras-Prover-SFT-32B")
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 Pythagoras-LM/Pythagoras-Prover-SFT-32B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Pythagoras-LM/Pythagoras-Prover-SFT-32B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Pythagoras-LM/Pythagoras-Prover-SFT-32B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Pythagoras-LM/Pythagoras-Prover-SFT-32B

SGLang

How to use Pythagoras-LM/Pythagoras-Prover-SFT-32B 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 "Pythagoras-LM/Pythagoras-Prover-SFT-32B" \
    --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": "Pythagoras-LM/Pythagoras-Prover-SFT-32B",
		"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 "Pythagoras-LM/Pythagoras-Prover-SFT-32B" \
        --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": "Pythagoras-LM/Pythagoras-Prover-SFT-32B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Pythagoras-LM/Pythagoras-Prover-SFT-32B with Docker Model Runner:
```
docker model run hf.co/Pythagoras-LM/Pythagoras-Prover-SFT-32B
```

Pythagoras-Prover-SFT-32B

Pythagoras-Prover-SFT-32B is a Qwen3-32B based causal language model fine-tuned for formal mathematical reasoning and theorem proving.

This repository contains the merged model weights. The LoRA adapter from checkpoint-6242 was merged into the Qwen3-32B base model with LLaMA-Factory.

Model Details

Base model: Qwen/Qwen3-32B
Fine-tuning method: LoRA SFT, merged into the base weights for this release
Training framework: LLaMA-Factory
Source checkpoint: saves/qwen3-32b/lora_v2/sft_data_updated_adaptive_clean_lora/checkpoint-6242
Export precision: follows the source Qwen3-32B checkpoint dtype

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "Pythagoras-LM/Pythagoras-Prover-SFT-32B"

tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype="auto",
    device_map="auto",
    trust_remote_code=True,
)

Intended Use

This model is intended for research and development workflows involving formal reasoning, mathematical problem solving, and Lean-style theorem proving.

Downloads last month: 46

Safetensors

Model size

33B params

Tensor type

BF16

Model tree for Pythagoras-LM/Pythagoras-Prover-SFT-32B

Base model

Qwen/Qwen3-32B

Finetuned

(508)

this model