Microsoft.ML.OnnxRuntimeGenAI 0.2.0

ONNX Runtime generate() API

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, and KV cache management.

You can call a high level generate() method to generate all of the output at once, or stream the output one token at a time.

See documentation at https://onnxruntime.ai/docs/genai.

Features

• Supported model architectures:
  • Gemma
  • LLaMA
  • Mistral
  • Phi-2
• Supported targets:
  • CPU
  • GPU (CUDA)
  • GPU (DirectML)
• Supported sampling features
  • Beam search
  • Greedy search
  • Top P/Top K
• APIs
  • Python
  • C#
  • C/C++

Coming very soon

• Support for the encoder decoder model architectures, such as whisper, T5 and BART.

Coming soon

• Support for mobile devices (Android and iOS) with Java and Objective-C bindings

Roadmap

• Stable diffusion pipeline
• Automatic model download and cache
• More model architectures

Sample code for phi-2 in Python

params = og.GeneratorParams(model) params.set_search_options(**search_options) params.input_ids = input_tokens generator = og.Generator(model, params)

print("Output: ", end='', flush=True)

1. Build the model

python -m onnxruntime_genai.models.builder -m microsoft/phi-2 -e cpu -p int4 -o ./models/phi2
# You can append --extra_options enable_cuda_graph=1 to build an onnx model that supports using cuda graph in ORT.

2. Run inference

import os
import onnxruntime_genai as og

model_path = os.path.abspath("./models/phi2")

model = og.Model(model_path)

tokenizer = og.Tokenizer(model)

prompt = '''def print_prime(n):
    """
    Print all primes between 1 and n
    """'''

tokens = tokenizer.encode(prompt)

params = og.GeneratorParams(model)
params.set_search_options({"max_length":200})
# Add the following line to enable cuda graph by passing the maximum batch size.
# params.try_use_cuda_graph_with_max_batch_size(16)
params.input_ids = tokens

output_tokens = model.generate(params)

text = tokenizer.decode(output_tokens)

print("Output:")
print(text)

Model download and export

ONNX models are run from a local folder, via a string supplied to the Model() method.

You can bring your own ONNX model or use the model builder utility, included in this package.

Install model builder dependencies.

pip install numpy
pip install transformers
pip install torch
pip install onnx
pip install onnxruntime

Export int4 CPU version

huggingface-cli login --token <your HuggingFace token>
python -m onnxruntime_genai.models.builder -m microsoft/phi-2 -p int4 -e cpu -o <model folder>

Getting the latest nightly Onnxruntime build

By default, onnxruntime-genai uses the latest stable release of onnxruntime. If you want to use the latest nightly build of onnxruntime, you can download the nightly build of onnxruntime from our Azure DevOps Artifacts. nuget package can be uncompressed by renaming the extension to .zip and extracting the contents. The onnxruntime dynamlic libraries and header files are available in the nightly build. You can extract the nuget package and copy the dynamic libraries and header files to the ort/ folder under onnxruntime-genai project root on the same level as this README.md file.

The library files are located in the runtime/$OS-$Arch/native folder and the header files are located in the build/native/include folder in the nuget package.

The final folder structure should look like this:

onnxruntime-genai
│   README.md
│   ... 
│   ort/
│   │   include/
│   │   │   coreml_provider_factory.h
│   │   │   ...
│   │   │   provider_options.h
│   │   lib/
│   │   │   (lib)onnxruntime.(so|dylib|dll)
│   │   │   (lib)onnxruntime_providers_shared.(so|dylib|dll)

Contributing

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 opencode@microsoft.com with any additional questions or comments.

Trademarks

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.