langchain.md

# LangChain

vLLM is also available via [LangChain](https://github.com/langchain-ai/langchain) .

To install LangChain, run

```bash
pip install langchain langchain_community -q
```

To run inference on a single or multiple GPUs, use `VLLM` class from `langchain`.

??? code

    ```python
    from langchain_community.llms import VLLM

    llm = VLLM(model="mosaicml/mpt-7b",
            trust_remote_code=True,  # mandatory for hf models
            max_new_tokens=128,
            top_k=10,
            top_p=0.95,
            temperature=0.8,
            # tensor_parallel_size=... # for distributed inference
    )

    print(llm("What is the capital of France ?"))
    ```

Please refer to this [Tutorial](https://python.langchain.com/docs/integrations/llms/vllm) for more details.