Releases: lollcat/fab-torch
Releases · lollcat/fab-torch
Demo Alanine Dipeptide and Many Well models
Demo inference with weights from huggingface
- For alanine dipeptide and many well flows trained with FAB, add colab notebooks where weights are pulled from huggingface and simple inference is performed to generate a plot.
What's Changed
- feat: Many Well evaluation, improve prioritized buffer by @lollcat in #75
- fix: naming of metric in Many Well by @lollcat in #79
- feat: small fixes and README update by @lollcat in #80
- Update mention of fab-jax in README.md by @lollcat in #81
- feat: update requirements and improve README clarity by @lollcat in #82
- feat: notebook for aldp hugging face model by @VincentStimper in #83
- feat: add colab for aldp demo by @lollcat in #84
- feat: many well demo by @lollcat in #85
Full Changelog: 0.0.1...v1.0.0
ICLR paper
- Release to match the code used for the experiments in the ICLR paper.
- This is just for reference to the results in the paper. The code is still being actively used and improved, thus any new experiments should use the latest version of the code rather than this release.
FAB arXiv paper first release
- Code for the first release of our paper Flow Annealed Importance Sampling Bootstrap.
- Contains the code for our method (FAB) and other baseline methods (e.g. KL divergence minimisation, training by maximum likelihood)
- Includes scripts for running the experiments from the paper.
- The repo is still under active development, and various improvements to the code will be made in the future.