Self-host LLMs with MLC-LLM and BentoML

This is a BentoML example project, showing you how to serve and deploy open-source Large Language Models using MLC-LLM, a machine learning compiler and high-performance deployment engine for large language models.

See here for a full list of BentoML example projects.

💡 This example is served as a basis for advanced code customization, such as custom model, inference logic or LMDeploy options. For simple LLM hosting with OpenAI compatible endpoint without writing any code, see OpenLLM.

Prerequisites

• You have installed Python 3.11+ and pip. See the Python downloads page to learn more.
• You have a basic understanding of key concepts in BentoML, such as Services. We recommend you read Quickstart first.
• If you want to test the Service locally, you need a Nvidia GPU with at least 20G VRAM.
• This example uses Llama 3. Make sure you have gained access to the model.
• (Optional) We recommend you create a virtual environment for dependency isolation for this project. See the Conda documentation or the Python documentation for details.

Set up the environment

Clone the repo:

git clone https://github.com/bentoml/BentoMLCLLM.git
cd BentoMLCLLM

Install dependencies:

# for cuda 12.1
python3 -m pip install --pre -U -f https://mlc.ai/wheels mlc-llm-nightly-cu121 mlc-ai-nightly-cu121
pip install -r requirements.txt

Compile and pack the model:

# For 8B model
git clone https://huggingface.co/mlc-ai/Llama-3-8B-Instruct-q0f16-MLC
export MLC_MODEL_PATH=Llama-3-8B-Instruct-q0f16-MLC

# For 70B model
git clone https://huggingface.co/mlc-ai/Llama-3-70B-Instruct-q4f16_1-MLC
export MLC_MODEL_PATH=Llama-3-70B-Instruct-q4f16_1-MLC

export MODEL_LIB=$MLC_MODEL_PATH/lib.so
mlc_llm compile $MLC_MODEL_PATH -o $MODEL_LIB --device cuda

Expected output:

...
[2024-06-06 11:51:49] INFO pipeline.py:52: Compiling external modules
[2024-06-06 11:51:49] INFO pipeline.py:52: Compilation complete! Exporting to disk
[2024-06-06 11:51:58] INFO model_metadata.py:95: Total memory usage without KV cache:: 15628.51 MB (Parameters: 15316.51 MB. Temporary buffer: 312.00 MB)
[2024-06-06 11:51:58] INFO model_metadata.py:103: To reduce memory usage, tweak `prefill_chunk_size`, `context_window_size` and `sliding_window_size`
[2024-06-06 11:51:58] INFO compile.py:193: Generated: Llama-3-8B-Instruct-q0f16-MLC/lib.so

Import the model into the BentoML Model Store:

python3 pack_model.py

Run the BentoML Service

We have defined a BentoML Service in service.py. Run bentoml serve in your project directory to start the Service.

$ bentoml serve .
2024-06-06T12:22:33+0000 [INFO] [cli] Starting production HTTP BentoServer from "service:MLCLLM" listening on http://localhost:3000 (Press CTRL+C to quit)

The server is now active at http://localhost:3000. You can interact with it using the Swagger UI or in other different ways.

CURL

curl -X 'POST' \
  'http://localhost:3000/generate' \
  -H 'accept: text/event-stream' \
  -H 'Content-Type: application/json' \
  -d '{
  "prompt": "Explain superconductors like I'\''m five years old",
  "max_tokens": 1024
}'

Python client

import bentoml

with bentoml.SyncHTTPClient("http://localhost:3000") as client:
    response_generator = client.generate(
        prompt="Explain superconductors like I'm five years old",
        max_tokens=1024
    )
    for response in response_generator:
        print(response, end='')

Deploy to BentoCloud

After the Service is ready, you can deploy the application to BentoCloud for better management and scalability. Sign up if you haven't got a BentoCloud account.

Make sure you have logged in to BentoCloud, then run the following command to deploy it. Note that you need to specify the CUDA version in bentofile.yaml.

bentoml deploy .

Once the application is up and running on BentoCloud, you can access it via the exposed URL.

Note: For custom deployment in your own infrastructure, use BentoML to generate an OCI-compliant image.