Readiness
- Readiness
- Production Ready
- Smoke
- passed
- Artifacts
- passed
- Production
- ready
- Storage
- <5
- Strategy
- dedicated DeepFRI TensorFlow 2.3 SIF plus staged Python 3.6 deps and official trained models
- Next
- Production Ready for validated contact-map MF prediction. Optional hardening: add uploaded contact-map/PDB conversion flow and ontology selection for BP/CC/EC after separate validation.
DeepFRI source cloned from flatironinstitute/DeepFRI rev 3bd6c7eae67894ccedbe1117e33b3abe544bca71. Pretrained URLs are live (~872-904 MB). Slurm job 394 used python312_slim SIF and reached predict.py but failed on missing numpy; container also lacks TensorFlow/Bio/sklearn and DeepFRI expects Python 3.7-era TensorFlow 2.3.1. 2026-07-05 SOP triage: backend inventory still lacks a DeepFRI TensorFlow/Python 3.7-era SIF and extracted trained_models; previous job 394 reached predict.py but python312_slim.sif lacked numpy/TensorFlow/Bio/sklearn. Kept Smoke Pending pending compatible <5GB TensorFlow runtime. 2026-07-05 SOP recheck: still no DeepFRI TensorFlow/Python 3.7-era SIF or extracted trained_models on the backend; python312_slim remains insufficient for numpy/TensorFlow/Bio/sklearn. Kept Smoke Pending. 2026-07-05 SOP batch recheck: no DeepFRI TensorFlow/Python 3.7-era SIF or extracted trained_models; prior job 394 failed in python312_slim due missing numpy/TensorFlow/Bio/sklearn. Kept Smoke Pending. 2026-07-05 SOP recheck: backend inventory still lacks DeepFRI TensorFlow/Python 3.7-era SIF and extracted trained_models; prior job 394 failed in python312_slim. Kept Smoke Pending. 2026-07-05 SOP batch recheck: backend still lacks a compatible DeepFRI TensorFlow/Python 3.7-era SIF and extracted trained_models; prior Slurm job 394 reached predict.py but python312_slim lacked numpy/TensorFlow/Bio/sklearn. Kept Smoke Pending. 2026-07-05 18:19 +08:00 backend re-audit: source remains staged at /media/nik/seagate_nik/bio_server/repos/deepfri (~322 MB) and DeepFRI test sbatch/evidence exists, but shared containers still only include python312_slim.sif, ubuntu2404.sif, and proteinmpnn_ddg base; no DeepFRI TensorFlow/Python 3.7-era SIF or extracted trained_models found. Agents agreed Smoke Pending remains correct. 2026-07-06 21:34 +08:00 recheck: sidecar review confirmed prior job 394 reached predict.py but Python 3.12 slim lacks TensorFlow/Bio/sklearn and no DeepFRI TensorFlow 2.3/Python 3.7 SIF or extracted trained_models exists. 2026-07-06 23:06 +08:00 promotion: materialized TensorFlow 2.3.1 SIF at /media/nik/seagate_nik/bio_server/containers/shared/deepfri_tf23.sif (555 MB), downloaded official CPU newest_trained_models.tar.gz (872,246,438 bytes), extracted trained_models under /media/nik/seagate_nik/bio_server/repos/deepfri, and staged Python 3.6 deps at /media/nik/seagate_nik/bio_server/envs/deepfri_tf23_py36_site. Direct Slurm job 475 passed using upstream examples/pdb_cmaps/1S3P-A.npz with MF ontology. Run /media/nik/seagate_nik/bio_server/runs/deepfri/smoke_20260706_225800 produced summary.json, results.csv, deepfri_1s3p_MF_predictions.csv, deepfri_1s3p_MF_pred_scores.json, job.log, and Slurm logs. Scientific sanity: top verbose prediction was GO:0005509 calcium ion binding with score 0.99999. Runner/API wiring remains pending. 2026-07-07 19:59 +08:00 production validation: real DeepFRI runner adapter wired and deployed to Ubuntu backend. Runner API remote job cf66a6f19e0e / Slurm job 508 completed using deepfri_tf23.sif, deepfri_tf23_py36_site, official trained_models, and upstream examples/pdb_cmaps/1S3P-A.npz with MF ontology. Produced deepfri_1s3p_MF_predictions.csv, deepfri_1s3p_MF_pred_scores.json, results.csv, summary.json, deepfri stdout/stderr, job.log, slurm.out, and slurm.err. Runtime 9.191s; top prediction GO:0005509 calcium ion binding score 0.99999.