Skip to content

NVIDIA BioNeMo Framework v2.6

Compare
Choose a tag to compare
@trvachov trvachov released this 30 Apr 20:15
· 49 commits to main since this release

New Features

  • Adds support for AMPLIFY doi:10.1101/2024.09.23.614603 pre-training and inference, offering a 70% speedup over the xformers-based attention backend with similar final perplexity values at 1M pre-training steps. (4.23 for 120M, 3.05 for 350M). The model is fully compatible with existing weights on HuggingFace.
  • Adds alpha support for LoRA fine-tuning to for ESM2 models. Inference and fine-tuning are enabled along with resumption from a checkpoint.

Updates & Improvements

  • Blackwell support, tested on B200 systems.
  • Fixed Grace CPU support, released ARM compatible container.

What's Changed

New Contributors

Full Changelog: v2.5...v2.6