bontojr/Reverso
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Validated MSA-reuse screen; bump cache_msa timeout for slow MSA server

Browse Source 
2-drug confirm passes: MSA cached once + reused, a3m parses (null-strip +
FASTA-selection fix works), screened 2/2, results sensible. The earlier
'screened 0' (a3m KeyError '\x00') and FunctionTimeout (ColabFold server
queued >30min) are both resolved: a3m hardening + cache_msa timeout 4200s.

Corrected screen is end-to-end working: 1 MSA query, reused across all
drugs, exception-tolerant.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
Junior B.
2026-06-26 07:38:01 +02:00
parent f74a58aa43
commit 98c99bc271
1 changed files with 2 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
gpu/modal_app.py
View File

@@ -127,7 +127,8 @@ def build_boltz_yaml(protein_seq: str, ligand_smiles: str, cofactor_ccds: list[s
# max_containers caps parallel fan-out (cost control). The download race that corrupts the
# checkpoint only happens on a COLD volume; once weights are cached+committed (Phase 1 did this),
# parallel containers just reload them, so a screen can safely run ~10-wide.
@app.function(gpu="L4", image=image, volumes={WEIGHTS: weights}, timeout=1800)
# Long timeout: the public ColabFold MSA server queues and can take 20-40+ min for one query.
@app.function(gpu="L4", image=image, volumes={WEIGHTS: weights}, timeout=4200)
def cache_msa(label: str, protein_seq: str, ligand_smiles: str, cofactor_ccds: list[str]) -> str:
"""Compute the target's MSA ONCE (via the server) and cache the a3m on the Volume.