Instructions to use l3lab/ntp-mathlib-st-deepseek-coder-1.3b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use l3lab/ntp-mathlib-st-deepseek-coder-1.3b with Transformers:

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

pipe = pipeline("text-generation", model="l3lab/ntp-mathlib-st-deepseek-coder-1.3b")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("l3lab/ntp-mathlib-st-deepseek-coder-1.3b")
model = AutoModelForCausalLM.from_pretrained("l3lab/ntp-mathlib-st-deepseek-coder-1.3b")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use l3lab/ntp-mathlib-st-deepseek-coder-1.3b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "l3lab/ntp-mathlib-st-deepseek-coder-1.3b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "l3lab/ntp-mathlib-st-deepseek-coder-1.3b",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/l3lab/ntp-mathlib-st-deepseek-coder-1.3b

SGLang

How to use l3lab/ntp-mathlib-st-deepseek-coder-1.3b 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 "l3lab/ntp-mathlib-st-deepseek-coder-1.3b" \
    --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": "l3lab/ntp-mathlib-st-deepseek-coder-1.3b",
		"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 "l3lab/ntp-mathlib-st-deepseek-coder-1.3b" \
        --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": "l3lab/ntp-mathlib-st-deepseek-coder-1.3b",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use l3lab/ntp-mathlib-st-deepseek-coder-1.3b with Docker Model Runner:
```
docker model run hf.co/l3lab/ntp-mathlib-st-deepseek-coder-1.3b
```

miniCTX: Neural Theorem Proving with (Long-)Contexts

State-tactic model from miniCTX: Neural Theorem Proving with (Long-)Contexts.

Base language model: deepseek-ai/deepseek-coder-1.3b-base
Data: ntp-mathlib-instruct-st

It is specifically finetuned for Lean 4 tactic prediction given proof states.

Performance

Please see our paper.

Example input

/- You are proving a theorem in Lean 4.
You are given the following information:
- The current proof state, inside [STATE]...[/STATE]

Your task is to generate the next tactic in the proof.
Put the next tactic inside [TAC]...[/TAC]
-/
[STATE]
m n : ℕ
h : Nat.Coprime m n
⊢ Nat.gcd m n = 1
[/STATE]
[TAC]

Example output

rw [Nat.Coprime] at h
[/TAC]

Citation

Please cite:

@misc{hu2024minictx,
  title={miniCTX: Neural Theorem Proving with (Long-)Contexts}, 
  author={Jiewen Hu and Thomas Zhu and Sean Welleck},
  year={2024},
  eprint={2408.03350},
  archivePrefix={arXiv},
  primaryClass={cs.AI},
  url={https://arxiv.org/abs/2408.03350}, 
}

Downloads last month: 7

Collection including l3lab/ntp-mathlib-st-deepseek-coder-1.3b

miniCTX

Collection

miniCTX: Neural Theorem Proving with (Long-)Contexts (ICLR 2025 Oral) • 8 items • Updated Mar 19, 2025 • 2

Paper for l3lab/ntp-mathlib-st-deepseek-coder-1.3b

miniCTX: Neural Theorem Proving with (Long-)Contexts

Paper • 2408.03350 • Published Aug 5, 2024 • 1