NVIDIA BioNeMo Framework v2.2

New Features

• Small Molecule Featurization
  • Implemented elementary and advanced atom, bond, and full molecule featurizers.
• GH200 Support for BioNeMo
  • Added a Dockerfile.arm that builds a BioNeMo container that runs on GH200 machines.
  • Publish a version of the BioNeMo container that supports multiple architectures to NGC.

Updates & Improvements

• Single-Cell Dataloader (SCDL)
  • Changed metadata storage to parquet files, which creates a 30x speed up when iterating over a large dataset.
  • Added functionality to concatenate several anndata files without doubling disk memory usage.
• ESM2
  • Added support for SIGTERM preemption checkpoint saving.
  • Moved ESM-2 and Geneformer training scripts to new executables, train_esm2 and train_geneformer, respectively.
  • Moved inference script to a new executable infer_esm2, and deprecated the inference example in the fine-tuning tutorial.
  • Added new Jupyter notebook tutorials for inference and zero-shot protein design. These notebooks can be deployed on the cloud resources as a brev.dev launchable.

Known Issues

• Loading a checkpoint for Geneformer inference on H100 has a known regression in accuracy. Work is in progress to resolve by next release.

Changes

• Move ESM2 scripts to sub-packages by @farhadrgh in #406
• WAR: sets checkpoint filename to be more unique by @skothenhill-nv in #429
• Update NeMo and Megatron to TOT by @pstjohn in #424
• re-enable merge groups to trigger blossom-ci by @pstjohn in #431
• Revert "re-enable merge groups to trigger blossom-ci" by @pstjohn in #434
• Updated notebook, and nemo2 checkpoint with geneformer by @jstjohn in #430
• add pre-emption callback to esm2 train by @pstjohn in #433
• add rdkit dependency to bionemo-geometric by @sveccham in #432
• eliminate the need for NGC login - bionemo2 by @dorotat-nv in #440
• Add documentation and release info to README by @sirelkhatim in #447
• Bump 3rdparty/Megatron-LM from aded519 to 5438d15 by @dependabot in #444
• Launchable notebooks in docs! by @jstjohn in #451
• Cache dev build from our nightly public container by @jstjohn in #462
• set num_workers to 1 for esm2 tests by @pstjohn in #461
• ESM2 Tutorial Updates by @farhadrgh in #426
• BugFix: fix bugs on bionemo-size-aware-batching by @guoqing-zhou in #449
• Fix typos in geneformer benchmark description by @jstjohn in #470
• Pillow version bump into main by @polinabinder1 in #465
• Refactor SCDL Row Feature Index for Performance Improvement (Rebased) by @savitha-eng in #466
• pin correct tornado requirement by @polinabinder1 in #474
• Updating Brev.Dev documentation by @polinabinder1 in #483
• Add release notes for v2.1 by @tshimko-nv in #468
• Update VERSION by @polinabinder1 in #488
• Atom and bond features by @sveccham in #453
• Molecule featurizer and molecule graph by @sveccham in #484
• hillst/bionemo noodles by @skothenhill-nv in #458
• update collate mask_value by @pstjohn in #485
• override checkpoint precision by @farhadrgh in #475
• JSON -> YAML for CLI by @skothenhill-nv in #436
• [QA Bug] Remove NGC dependency by @farhadrgh in #494
• Bump 3rdparty/NeMo from e2b0f0e to 06e6703 by @dependabot in #486
• Bump 3rdparty/Megatron-LM from 5438d15 to 844119f by @dependabot in #496
• change source for coverage report by @pstjohn in #495
• Pstjohn/stop and go test non validation by @pstjohn in #476
• Add support on num steps for learning rate scheduler by @sichu2023 in #489
• Initial compatibility testing images by @malcolmgreaves in #438
• Conda-Based Compatibility Test Images by @malcolmgreaves in #507
• Instructions on compatibility image build by @malcolmgreaves in #512
• Formatting by @malcolmgreaves in #513
• Pstjohn/fix ci by @pstjohn in #515
• [FEA][webdatamodule]: support webdataset invocable by @DejunL in #501
• GH200 support by @gagank1 in #369
• Remove quotes for Jupyter command on startup in init guide by @tshimko-nv in #523
• Reduce esm2 and geneformer test burden by @sichu2023 in #499
• [v2.2] Publish release notes for BioNeMo FW v2.2. by @cspades in #522
• Disable validation/test stages in ESM-2 and Geneformer by @sichu2023 in #492
• CI HOTFIX: ignore inrun_pytest.sh a notebook by @dorotat-nv in #526
• added NeMoLogger unit tests by @dorotat-nv in #511
• Bump 3rdparty/Megatron-LM from 844119f to 99f23d2 by @dependabot in #528
• [cye/wandb-fix] Fix WandB issue. by @cspades in #530
• xFail known bad tests on H100 and fix CVEs by @gagank1 in #547

New Contributors

• @sveccham made their first contribution in #432
• @sirelkhatim made their first contribution in #447

Full Changelog: v2.1...v2.2