transformer-flows 0.0.13

Implementation of Transformer Flows (Apple ML) in JAX and Equinox.

Transformer flows

Implementation of Apple ML's Transformer Flow (or TARFlow) from Normalising flows are capable generative models in jax and equinox.

Features:

• jax.vmap & jax.lax.scan construction & forward-pass, for layers respectively for fast compilation and execution,
• multi-device training, inference and sampling,
• score-based denoising step (see paper),
• conditioning via class embedding (for discrete class labels) or adaptive layer-normalisation (for continuous variables, like in DiT),
• array-typed to-the-teeth for dependable execution with jaxtyping and beartype.

To implement:

• Guidance
• Denoising
• Mixed precision
• EMA
• AdaLayerNorm
• Class embedding
• Hyperparameter/model saving
• Uniform and Gaussian noise for dequantisation

Usage

pip install transformer-flows

uv run --all-extras python examples/main.py

Samples

I haven't optimised anything here (the authors mention varying the variance of noise used to dequantise the images), nor have I trained for very long. You can see slight artifacts due to the dequantisation noise.

Citation

@misc{zhai2024normalizingflowscapablegenerative,
      title={Normalizing Flows are Capable Generative Models}, 
      author={Shuangfei Zhai and Ruixiang Zhang and Preetum Nakkiran and David Berthelot and Jiatao Gu and Huangjie Zheng and Tianrong Chen and Miguel Angel Bautista and Navdeep Jaitly and Josh Susskind},
      year={2024},
      eprint={2412.06329},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2412.06329}, 
}