FrameFlow

PaperCodeNeed >5GB

Motif scaffolding protein design

Readiness

Readiness: Need >5GB
Smoke: passed
Artifacts: passed
Production: need_gt5gb_frameflow_gpu_assets_runtime_pending
Storage: 5-20GB
Strategy: large shared SIF/database/model bundle phase
Next: Rebuild FrameFlow in Apptainer/SIF or ext4-backed micromamba, stage weights/preprocessed assets, then run one-sample single-GPU Slurm inference.

No obvious large model/database dependency identified; validate upstream docs before install. Worker evidence 2026-07-01: source https://github.com/microsoft/protein-frame-flow revision f50d8dbbdae827be291e9f73d732b61b195f8816 staged at /media/nik/seagate_nik/bio_server/tools/frameflow/protein-frame-flow, clean. Micromamba 2.8.1 extracted; attempted upstream fm.yml env with prefix via /home/nik/frameflow_tool symlink to Seagate. Environment creation failed at CUDA package extraction: critical libmamba cannot create lib/libnppc.so.12. No Slurm job submitted. Upstream assets include weights.tar.gz ~0.6GB, preprocessed_pdb.tar.gz ~2.7GB, preprocessed_scope.tar.gz ~0.3GB. Blocker is exFAT/symlink/filesystem/env setup. 2026-07-07 SOP reclassification: agent audit confirmed FrameFlow needs the larger shared GPU/runtime/assets lane. Existing upstream assets are already ~3.6GB before CUDA/PyTorch/conda runtime, and prior env creation was blocked by filesystem/symlink extraction issues; moved to Need >5GB. 2026-07-17T01:36:00+08:00 SOP Need >5GB next-10 audit: remains Need >5GB. Backend has /home/nik/frameflow_runtime_ext4 with only partial mamba root/logs and an old /home/nik/frameflow_tool symlink pointing at the obsolete seagate_nik path; no working FrameFlow Python env was found. Needs SIF/ext4 micromamba rebuild, weights/preprocessed assets, and one-sample GPU inference.