BentoML: The Unified Model Serving Framework

BentoML is an open-source model serving library for building performant and scalable AI applications with Python. It comes with everything you need for serving optimization, model packaging, and production deployment.

👉 Join our Slack community!

• Open standard and SDK for AI apps, pack your code, inference pipelines, model files, dependencies, and runtime configurations in a Bento.
• Auto-generate API servers, supporting REST API, gRPC, and long-running inference jobs.
• Auto-generate Docker container images.

• Import from any model hub or bring your own models built with frameworks like PyTorch, TensorFlow, Keras, Scikit-Learn, XGBoost and many more.
• Native support for LLM inference, generative AI, embedding creation, and multi-modal AI apps.
• Run and debug your BentoML apps locally on Mac, Windows, or Linux.

• Integrate with high-performance runtimes such as ONNX-runtime and TorchScript to boost response time and throughput.
• Support parallel processing of model inferences for improved speed and efficiency.
• Implement adaptive batching to optimize processing.
• Built-in optimization for specific model architectures (like OpenLLM for LLMs).

• Python-first! Effortlessly scale complex AI workloads.
• Enable GPU inference without the headache.
• Compose multiple models to run concurrently or sequentially, over multiple GPUs or on a Kubernetes Cluster.
• Natively integrates with MLFlow, LangChain, Kubeflow, Triton, Spark, Ray, and many more to complete your production AI stack.

• One-click deployment to ☁️ BentoCloud, the Serverless platform made for hosting and operating AI apps.
• Scalable BentoML deployment with 🦄️ Yatai on Kubernetes.
• Deploy auto-generated container images anywhere Docker runs.

• Installation: pip install bentoml
• Documentation: docs.bentoml.com
• Tutorial: Quickstart

• OpenLLM - Run any open-source LLMs, such as Llama 2 and Mistral, as OpenAI compatible API endpoints, locally and in the cloud.
• BentoXTTS - Convert text to speech based on your custom audio data.
• BentoSDXLTurbo - Create an image generation application and run inference with a single step.
• BentoSD2Upscaler - Build an image generation application with upscaling capability.
• BentoControlNet - Influence image composition, adjust specific elements, and ensure spatial consistency by integrating ControlNet with your image generation process.
• BentoWhisperX - Convert spoken words into text for AI scenarios like virtual assistants, voice-controlled devices, and automated transcription services.
• Sentence Transformer - Transform text into numerical vectors for a variety of natural language processing (NLP) tasks.
• BentoCLIP - Build a CLIP (Contrastive Language-Image Pre-training) application for tasks like zero-shot learning, image classification, and image-text matching.
• BentoBLIP - Leverage BLIP (Bootstrapping Language Image Pre-training) to improve the way AI models understand and process the relationship between images and textual descriptions.
• BentoLCM - Deploy a REST API server for Stable Diffusion with Latent Consistency LoRAs.
• BentoSVD - Create a video generation application powered by Stable Video Diffusion (SVD).
• BentoVLLM - Accelerate your model inference and improve serving throughput by using vLLM as your LLM backend.
• BentoBark - Generate highly realistic audio like music, background noise and simple sound effects with Bark.
• BentoYolo - Build an object detection inference API server with YOLO.
• RAG - Self-host a RAG web service with BentoML step by step, including an embedding model, a large language model, and a vector database.

This example demonstrates how to serve and deploy a simple text summarization application.

Install dependencies:

pip install torch transformers "bentoml>=1.2.0a0"

Define the serving logic of your model in a service.py file.

from __future__ import annotations
import bentoml
from transformers import pipeline


@bentoml.service(
    resources={"cpu": "2"},
    traffic={"timeout": 10},
)
class Summarization:
    def __init__(self) -> None:
        # Load model into pipeline
        self.pipeline = pipeline('summarization')

    @bentoml.api
    def summarize(self, text: str) -> str:
        result = self.pipeline(text)
        return result[0]['summary_text']

Run this BentoML Service locally, which is accessible at http://localhost:3000.

bentoml serve service:Summarization

Send a request to summarize a short news paragraph:

curl -X 'POST' \
  'http://localhost:3000/summarize' \
  -H 'accept: text/plain' \
  -H 'Content-Type: application/json' \
  -d '{
  "text": "Breaking News: In an astonishing turn of events, the small town of Willow Creek has been taken by storm as local resident Jerry Thompson'\''s cat, Whiskers, performed what witnesses are calling a '\''miraculous and gravity-defying leap.'\'' Eyewitnesses report that Whiskers, an otherwise unremarkable tabby cat, jumped a record-breaking 20 feet into the air to catch a fly. The event, which took place in Thompson'\''s backyard, is now being investigated by scientists for potential breaches in the laws of physics. Local authorities are considering a town festival to celebrate what is being hailed as '\''The Leap of the Century."
}'

After your Service is ready, you can deploy it to BentoCloud or as a Docker image.

First, create a bentofile.yaml file for building a Bento.

service: "service:Summarization"
labels:
  owner: bentoml-team
  project: gallery
include:
  - "*.py"
python:
  packages:
  - torch
  - transformers

Then, choose one of the following ways for deployment:

bentoml deploy .

bentoml build

bentoml containerize summarization:latest

docker run --rm -p 3000:3000 summarization:latest

For detailed explanations, read Quickstart.

BentoML supports billions of model runs per day and is used by thousands of organizations around the globe.

Join our Community Slack 💬, where thousands of AI application developers contribute to the project and help each other.

To report a bug or suggest a feature request, use GitHub Issues.

There are many ways to contribute to the project:

• Report bugs and "Thumbs up" on issues that are relevant to you.
• Investigate issues and review other developers' pull requests.
• Contribute code or documentation to the project by submitting a GitHub pull request.
• Check out the Contributing Guide and Development Guide to learn more
• Share your feedback and discuss roadmap plans in the #bentoml-contributors channel here.

Thanks to all of our amazing contributors!

BentoML collects usage data that helps our team to improve the product. Only BentoML's internal API calls are being reported. We strip out as much potentially sensitive information as possible, and we will never collect user code, model data, model names, or stack traces. Here's the code for usage tracking. You can opt-out of usage tracking by the --do-not-track CLI option:

bentoml [command] --do-not-track

Or by setting environment variable BENTOML_DO_NOT_TRACK=True:

export BENTOML_DO_NOT_TRACK=True

Apache License 2.0

If you use BentoML in your research, please cite using the following citation:

@software{Yang_BentoML_The_framework,
author = {Yang, Chaoyu and Sheng, Sean and Pham, Aaron and  Zhao, Shenyang and Lee, Sauyon and Jiang, Bo and Dong, Fog and Guan, Xipeng and Ming, Frost},
license = {Apache-2.0},
title = {{BentoML: The framework for building reliable, scalable and cost-efficient AI application}},
url = {https://github.com/bentoml/bentoml}
}