Skip to content

ColabFold 1.6.2

Choose a tag to compare

View all tags
@milot-mirdita milot-mirdita released this 14 Jul 11:16
· 25 commits to main since this release
v1.6.2
c7d1772

ColabFold 1.6.2

Bug fixes

  • Fix colabfold_search --pair-mode paired and --pair-mode unpaired crashing on complexes (4de9759)

New features and enhancements

  • Fast Pallas/Triton Evoformer kernels New --use-pallas flag for colabfold_batch runs the Evoformer with fused Pallas/Triton kernels for ~2.5x faster and slightly lower-memory predictions. Requires NVIDIA GPUs with compute capability 8.0 (Ampere) or newer. (e809493, 0ff4794, 5a83d23, d04bf9c)
  • Model compile-time control New --compile-mode flag (fast, tuned (default), full) trades kernel autotuning/compile time against inference speed; the difference is small (1-5%) but can matter for very large batches or very long sequences (dfac598)
  • TensorFlow removed Removed TensorFlow dependency reducing install size and startup overhead (4744396)
  • Docker improvements Docker images are now built for CUDA 12/13 and ARM64 (9489783, a6ed4c7)
Assets 2
Loading