Skip to main content

Machine learning in FPGAs using HLS

Project description

hls4ml

DOI License Documentation Status PyPI version Downloads conda-forge

A package for machine learning inference in FPGAs. We create firmware implementations of machine learning algorithms using high level synthesis language (HLS). We translate traditional open-source machine learning package models into HLS that can be configured for your use-case!

If you have any questions, comments, or ideas regarding hls4ml or just want to show us how you use hls4ml, don't hesitate to reach us through the discussions tab.

Documentation & Tutorial

For more information visit the webpage: https://fastmachinelearning.org/hls4ml/.

For introductory material on FPGAs, HLS and ML inferences using hls4ml, check out the video.

Detailed tutorials on how to use hls4ml's various functionalities can be found here.

Installation

pip install hls4ml

To install the extra dependencies for profiling:

pip install hls4ml[profiling]

Getting Started

Creating an HLS project

import hls4ml

# Fetch a keras model from our example repository
# This will download our example model to your working directory and return an example configuration file
config = hls4ml.utils.fetch_example_model('KERAS_3layer.json')

# You can print the configuration to see some default parameters
print(config)

# Convert it to a hls project
hls_model = hls4ml.converters.keras_to_hls(config)

# Print full list of example models if you want to explore more
hls4ml.utils.fetch_example_list()

Building a project.

We will build the project using Xilinx Vivado HLS, which can be downloaded and installed from here. Alongside Vivado HLS, hls4ml also supports Vitis HLS, Intel HLS, Catapult HLS and has some experimental support dor Intel oneAPI. The target back-end can be changed using the argument backend when building the model.

# Use Vivado HLS to synthesize the model
# This might take several minutes
hls_model.build()

# Print out the report if you want
hls4ml.report.read_vivado_report('my-hls-test')

FAQ

List of frequently asked questions and common HLS synthesis can be found here

Citation

If you use this software in a publication, please cite the software

@software{fastml_hls4ml,
  author       = {{FastML Team}},
  title        = {fastmachinelearning/hls4ml},
  year         = 2024,
  publisher    = {Zenodo},
  version      = {v1.0.0},
  doi          = {10.5281/zenodo.1201549},
  url          = {https://github.com/fastmachinelearning/hls4ml}
}

and first publication:

@article{Duarte:2018ite,
    author = "Duarte, Javier and others",
    title = "{Fast inference of deep neural networks in FPGAs for particle physics}",
    eprint = "1804.06913",
    archivePrefix = "arXiv",
    primaryClass = "physics.ins-det",
    reportNumber = "FERMILAB-PUB-18-089-E",
    doi = "10.1088/1748-0221/13/07/P07027",
    journal = "JINST",
    volume = "13",
    number = "07",
    pages = "P07027",
    year = "2018"
}

Additionally, if you use specific features developed in later papers, please cite those as well. For example, CNNs:

@article{Aarrestad:2021zos,
    author = "Aarrestad, Thea and others",
    title = "{Fast convolutional neural networks on FPGAs with hls4ml}",
    eprint = "2101.05108",
    archivePrefix = "arXiv",
    primaryClass = "cs.LG",
    reportNumber = "FERMILAB-PUB-21-130-SCD",
    doi = "10.1088/2632-2153/ac0ea1",
    journal = "Mach. Learn. Sci. Tech.",
    volume = "2",
    number = "4",
    pages = "045015",
    year = "2021"
}
@article{Ghielmetti:2022ndm,
    author = "Ghielmetti, Nicol\`{o} and others",
    title = "{Real-time semantic segmentation on FPGAs for autonomous vehicles with hls4ml}",
    eprint = "2205.07690",
    archivePrefix = "arXiv",
    primaryClass = "cs.CV",
    reportNumber = "FERMILAB-PUB-22-435-PPD",
    doi = "10.1088/2632-2153/ac9cb5",
    journal ="Mach. Learn. Sci. Tech.",
    year = "2022"
}

binary/ternary networks:

@article{Loncar:2020hqp,
    author = "Ngadiuba, Jennifer and others",
    title = "{Compressing deep neural networks on FPGAs to binary and ternary precision with HLS4ML}",
    eprint = "2003.06308",
    archivePrefix = "arXiv",
    primaryClass = "cs.LG",
    reportNumber = "FERMILAB-PUB-20-167-PPD-SCD",
    doi = "10.1088/2632-2153/aba042",
    journal = "Mach. Learn. Sci. Tech.",
    volume = "2",
    pages = "015001",
    year = "2021"
}

Acknowledgments

If you benefited from participating in our community, we ask that you please acknowledge the Fast Machine Learning collaboration, and particular individuals who helped you, in any publications. Please use the following text for this acknowledgment:

We acknowledge the Fast Machine Learning collective as an open community of multi-domain experts and collaborators. This community and <names of individuals>, in particular, were important for the development of this project.

Funding

We gratefully acknowledge previous and current support from the U.S. National Science Foundation (NSF) Harnessing the Data Revolution (HDR) Institute for Accelerating AI Algorithms for Data Driven Discovery (A3D3) under Cooperative Agreement No. PHY-2117997, U.S. Department of Energy (DOE) Office of Science, Office of Advanced Scientific Computing Research under the Real‐time Data Reduction Codesign at the Extreme Edge for Science (XDR) Project (DE-FOA-0002501), DOE Office of Science, Office of High Energy Physics Early Career Research Program (DE-SC0021187, DE-0000247070), and the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation program (Grant No. 772369).

A3D3 NSF DOE ERC

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

hls4ml-1.1.0.tar.gz (10.0 MB view details)

Uploaded Source

Built Distribution

hls4ml-1.1.0-py3-none-any.whl (3.1 MB view details)

Uploaded Python 3

File details

Details for the file hls4ml-1.1.0.tar.gz.

File metadata

  • Download URL: hls4ml-1.1.0.tar.gz
  • Upload date:
  • Size: 10.0 MB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.1.0 CPython/3.12.9

File hashes

Hashes for hls4ml-1.1.0.tar.gz
Algorithm Hash digest
SHA256 660ec1f70c33db1540b17eee4ff5888f1716a60c26402d6f2497c96b966a8735
MD5 0a4ca7aeeb3b50bd9c1f512bca29794f
BLAKE2b-256 62380547db70cb6a370d19800a755012b9476716948c0ce4e50213c9656f5ee2

See more details on using hashes here.

File details

Details for the file hls4ml-1.1.0-py3-none-any.whl.

File metadata

  • Download URL: hls4ml-1.1.0-py3-none-any.whl
  • Upload date:
  • Size: 3.1 MB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.1.0 CPython/3.12.9

File hashes

Hashes for hls4ml-1.1.0-py3-none-any.whl
Algorithm Hash digest
SHA256 f598b51c76932317e1c82a2c4dca7076bfd02d43e0210ba79b0c8abae5862ac4
MD5 5aca16895b8e31ec225fd21ab5c6e90e
BLAKE2b-256 61d016d5e01b1cf080058f67ddf34c9bcead3c67e3e0986c29e1cff447a24f5e

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page