feat: Inline ColabFold and AlphaFold2 into BioEmu#206
Merged
josejimenezluna merged 2 commits intomainfrom Mar 30, 2026
Merged
Conversation
sarahnlewis
force-pushed
the
sarahlewis/inline-colabfold
branch
from
f715a1c to
5f8a7ca
Compare
Replace the subprocess-based ColabFold integration (separate venv, patched site-packages, setup.sh) with code inlined directly into the BioEmu package. Everything now runs in a single Python environment. ## What changed ### Removed - src/bioemu/colabfold_setup/ (setup.sh, batch.patch, modules.patch) - Subprocess calls to colabfold_batch - Separate ColabFold venv (~/.bioemu_colabfold) is no longer needed ### Added: src/bioemu/colabfold_inline/ - msa_client.py: MMseqs2 API client (from colabfold.colabfold, MIT) - input_parsing.py: FASTA/A3M parser (from colabfold.batch, MIT) - features.py: Monomer feature pipeline wrapping vendored alphafold - model_runner.py: AF2 forward pass orchestration, weight downloading - LICENSES/: ColabFold MIT + AlphaFold2 Apache 2.0 license texts ### Added: src/_vendor/alphafold/ Vendored, patched subset of AlphaFold2 v2.3.2 (Apache 2.0): - Evoformer and model runner (the forward pass) - Patched modules.py to expose representations_evo - Removed: structure module, multimer, relax, templates, data tools (~13,000 lines of unused code deleted) - Registered via sys.modules aliasing (no sys.path manipulation) ### Modified - get_embeds.py: Calls inlined code directly instead of subprocess - pyproject.toml: JAX, Haiku, ml-collections, TF now required deps; [cuda] extra for GPU-specific packages - README.md: Updated install instructions, removed Python 3.12 cap, clarified conda only needed for optional hpacker - NOTICE.md, cgmanifest.json: Added ColabFold + AlphaFold2 attribution ### Tests - 96 tests passing (54 new tests for inlined code) - GPU regression tests verify embeddings match main branch (correlation >0.9999, per-residue cosine similarity >0.999) - Mocks target _run_model (JAX forward pass) only; feature building runs for real in unit tests ## License compliance - Vendored AF2 files retain original DeepMind copyright headers - Modified files carry Apache 2.0 Section 4(b) change notices - ColabFold-derived files carry MIT attribution headers - Full license texts in LICENSES/ and src/_vendor/alphafold/LICENSE Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
sarahnlewis
force-pushed
the
sarahlewis/inline-colabfold
branch
from
5f8a7ca to
aefe217
Compare
SummarySummary
Coveragesrc.bioemu - 88.8%
src.bioemu.colabfold_inline - 62%
src.bioemu.hpacker_setup - 58.8%
src.bioemu.training - 100%
|
SummarySummary
Coveragesrc.bioemu - 88.8%
src.bioemu.colabfold_inline - 62%
src.bioemu.hpacker_setup - 58.8%
src.bioemu.training - 100%
|
josejimenezluna
approved these changes
Mar 30, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Replace the subprocess-based ColabFold integration (separate venv, patched site-packages, setup.sh) with code inlined directly into the BioEmu package. Everything now runs in a single Python environment.
What changed
Removed
src/bioemu/colabfold_setup/(setup.sh, batch.patch, modules.patch)colabfold_batch~/.bioemu_colabfold) is no longer neededAdded:
src/bioemu/colabfold_inline/Added:
src/_vendor/modules.pyto exposerepresentations_evo. Removed ~13,000 lines of unused code (structure module, multimer, relax, templates, data tools).src/bioemu/openfold/with samesys.modulesaliasing.CI
uv(astral-sh/setup-uv@v4)JAX_PLATFORMS=cpufor CPU-only CI runnersDependencies
[cuda]extra for GPU-specific packages (jax[cuda12], nvcc)Tests
_run_modelonly; feature building runs for realLicense compliance
NOTICE.mdandcgmanifest.jsonupdatedE2E verified on GPU
Produces correct PDB + XTC output with fresh cache (no precomputed embeds or MSAs).