Run generative AI models with ONNX Runtime. This API gives you an easy, flexible and performant way of running LLMs on device. It implements the generative AI loop for ONNX models, including pre and post processing, inference with ONNX Runtime, logits processing, search and sampling, KV cache management, and grammar specification for tool calling.

ONNX Runtime GenAI powers Foundry Local, Windows ML, and the Visual Studio Code AI Toolkit.

See documentation at the ONNX Runtime website for more details.

^ Requires build from source

See installation instructions or build from source

1. Download the model

   huggingface-cli download microsoft/Phi-3-mini-4k-instruct-onnx --include cpu_and_mobile/cpu-int4-rtn-block-32-acc-level-4/* --local-dir .

2. Install the API

   pip install numpy
   pip install --pre onnxruntime-genai

3. Run the model

   import onnxruntime_genai as og
   
   model = og.Model('cpu_and_mobile/cpu-int4-rtn-block-32-acc-level-4')
   tokenizer = og.Tokenizer(model)
   tokenizer_stream = tokenizer.create_stream()
    
   # Set the max length to something sensible by default,
   # since otherwise it will be set to the entire context length
   search_options = {}
   search_options['max_length'] = 2048
   search_options['batch_size'] = 1
   
   chat_template = '<|user|>\n{input} <|end|>\n<|assistant|>'
   
   text = input("Input: ")
   if not text:
      print("Error, input cannot be empty")
      exit()
   
   prompt = f'{chat_template.format(input=text)}'
   
   input_tokens = tokenizer.encode(prompt)
   
   params = og.GeneratorParams(model)
   params.set_search_options(**search_options)
   generator = og.Generator(model, params)
   
   print("Output: ", end='', flush=True)
   
   try:
      generator.append_tokens(input_tokens)
      while True:
         generator.generate_next_token()
         if generator.is_done():
            break
         new_token = generator.get_next_tokens()[0]
         print(tokenizer_stream.decode(new_token), end='', flush=True)
   except KeyboardInterrupt:
         print("  --control+c pressed, aborting generation--")
   
   print()
   del generator

Due to the evolving nature of this project and ongoing feature additions, examples in the main branch may not always align with the latest stable release. This section outlines how to ensure compatibility between the examples and the corresponding version.

Install the package according to the installation instructions. For example, install the Python package.

pip install onnxruntime-genai

Get the version of the package

Linux/Mac:

pip list | grep onnxruntime-genai

Windows:

pip list | findstr "onnxruntime-genai"

Checkout the version of the examples that correspond to that release.

# Clone the repo
git clone https://github.com/microsoft/onnxruntime-genai.git && cd onnxruntime-genai
# Checkout the branch for the version you are using
git checkout v0.11.4
cd examples

Checkout the main branch of the repo

git clone https://github.com/microsoft/onnxruntime-genai.git && cd onnxruntime-genai

Build from source, using these instructions. For example, to build the Python wheel:

python build.py

Navigate to the examples folder in the main branch.

cd examples

Between v0.11.0 and v0.10.1, there is a breaking API usage change to improve model quality during multi-turn conversations.

Previously, the decoding loop could be written as follows.

while not IsDone():
    GenerateToken()
    GetLastToken()
    PrintLastToken()

In 0.11.0, the decoding loop should now be written as follows.

while True:
    GenerateToken()
    if IsDone():
        break
    GetLastToken()
    PrintLastToken()

See the Discussions to request new features and up-vote existing requests.

This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.opensource.microsoft.com.

When you submit a pull request, a CLA bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., status check, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA.

This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact [email protected] with any additional questions or comments.

This project enables lintrunner for linting. You can install the dependencies and initialize with

pip install -r requirements-lintrunner.txt
lintrunner init

This will install lintrunner on your system and download all the necessary dependencies to run linters locally.

To format local changes:

lintrunner -a

To format all files:

lintrunner -a --all-files

This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow Microsoft's Trademark & Brand Guidelines. Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos are subject to those third-party's policies.