Skip to main content

Add your description here

Project description

Rectified Flow Matching

Cutting-edge and feature-rich implementation of Rectified Flows from Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flows in jax and equinox.

Features

  • Deterministic and stochastic sampling of the associated ODE and SDE respectively,
  • Mixed precision optimisation,
  • Array-typed to-the-teeth for dependable execution with jaxtyping and beartype.

To implement:

  • Guidance by score of conditioning
  • Mixed precision
  • EMA
  • AdaLayerNorm
  • Stochastic sampling
  • ODE Sampling
  • Likelihoods
  • DiT
  • Hyperparameter/model saving

Usage

pip install rectified-flows

Your image description

Citations

@misc{liu2022flowstraightfastlearning,
      title={Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow}, 
      author={Xingchao Liu and Chengyue Gong and Qiang Liu},
      year={2022},
      eprint={2209.03003},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2209.03003}, 
}
@misc{lipman2023flowmatchinggenerativemodeling,
    title={Flow Matching for Generative Modeling}, 
    author={Yaron Lipman and Ricky T. Q. Chen and Heli Ben-Hamu and Maximilian Nickel and Matt Le},
    year={2023},
    eprint={2210.02747},
    archivePrefix={arXiv},
    primaryClass={cs.LG},
    url={https://arxiv.org/abs/2210.02747}, 
}
@misc{singh2024stochasticsamplingdeterministicflow,
      title={Stochastic Sampling from Deterministic Flow Models}, 
      author={Saurabh Singh and Ian Fischer},
      year={2024},
      eprint={2410.02217},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2410.02217}, 
}
@misc{yang2024consistencyflowmatchingdefining,
      title={Consistency Flow Matching: Defining Straight Flows with Velocity Consistency}, 
      author={Ling Yang and Zixiang Zhang and Zhilong Zhang and Xingchao Liu and Minkai Xu and Wentao Zhang and Chenlin Meng and Stefano Ermon and Bin Cui},
      year={2024},
      eprint={2407.02398},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2407.02398}, 
}
@misc{peebles2023scalablediffusionmodelstransformers,
      title={Scalable Diffusion Models with Transformers}, 
      author={William Peebles and Saining Xie},
      year={2023},
      eprint={2212.09748},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2212.09748}, 
}

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

rectified_flows-0.0.1.tar.gz (4.0 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

rectified_flows-0.0.1-py3-none-any.whl (4.6 kB view details)

Uploaded Python 3

File details

Details for the file rectified_flows-0.0.1.tar.gz.

File metadata

  • Download URL: rectified_flows-0.0.1.tar.gz
  • Upload date:
  • Size: 4.0 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: uv/0.7.15

File hashes

Hashes for rectified_flows-0.0.1.tar.gz
Algorithm Hash digest
SHA256 a4a6b2d04f7b095b82922df2485c03ae2150b510064aac61119b2ca418ccba40
MD5 2955c44332149b36e30e297544763ecd
BLAKE2b-256 7aa1e1e1d1bceb74a2d7f29c2a9489b62192fc9e3ab0e30b7bc33d253ef73bd2

See more details on using hashes here.

File details

Details for the file rectified_flows-0.0.1-py3-none-any.whl.

File metadata

File hashes

Hashes for rectified_flows-0.0.1-py3-none-any.whl
Algorithm Hash digest
SHA256 af474fcab18a6ac4a99ba96292244d3c2a35e488bef7a8ed9c52287068bf8d08
MD5 3523cdefe2f7de459c5577d539b9efcd
BLAKE2b-256 41b5eafaf9924480ea20d0b776717cd642ddcf7cc325993b76e7df07f0bf196a

See more details on using hashes here.

Supported by

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