Readiness
- Readiness
- Needs Token/License
- Smoke
- passed
- Artifacts
- passed
- Production
- needs_authenticated_or_supplied_p2pxml_model_weights
- Storage
- <5
- Strategy
- shared python312_slim.sif + tool ext3 overlay
- Next
- Obtain/supply P2PXML model_weights.pth or verified public checkpoint; then finish Torch/PyG runtime and validate inference through Slurm/API using the staged Zenodo dataset.
SOP recovery agent found staged p2pxml source at /media/nik/seagate_nik/bio_server/tools/p2pxml/source, upstream Drug-Discovery-ENTC/p2pxml commit 9b5c952, source about 26M. Main code src/integrated_model_v1.py passed Python AST syntax parse. Required dataset is not staged: ./P2PXML_Structure/P2PXML_structure.csv, antibodies/, antigens/, labels.npy, and output paths. Default Python lacks pandas, torch, torch_geometric, torchmetrics, biopandas, periodictable, networkx, tqdm, transformers. No p2pxml-specific env found. src/p2pxml.sh hardcodes nonportable module/venv/user paths. No Slurm benchmark submitted. 2026-07-10T21:33:01+08:00 SOP Production Blocked retry: confirmed upstream GitHub/Colab inference path and Zenodo record 11531319. The Colab demo requires model_weights.pth from a University of Moratuwa SharePoint link, but backend curl to the notebook URL returned HTTP 401, so the trained model checkpoint is not a reproducible/public asset from the current backend. A clean Zenodo dataset download was also attempted via the official API content URL; transfer repeatedly crawled and closed early before the 118,339,320-byte archive completed, so no verified dataset archive was staged. Keep Production Blocked until model_weights.pth is supplied by the user/upstream or a verified public checkpoint is found; then build the isolated Torch/PyG runtime and run bundled example-PDB inference through Slurm. 2026-07-10T22:01:41+08:00 continuation update: Zenodo P2PXML_dataset.zip was downloaded locally and uploaded to /media/nik/seagate_nik/bio_server/tools/p2pxml/dataset/P2PXML_dataset.zip, then extracted under /media/nik/seagate_nik/bio_server/tools/p2pxml/dataset/extracted/P2PXML_dataset. Verified P2PXML_Structure/P2PXML_structure.csv and 1,615 antibody plus 948 antigen PDB files. No .pth/.pt/.tar checkpoint is present in the source or dataset tree. Upstream Colab/documented inference expects model_weights.pth from an authenticated SharePoint URL, so production inference remains blocked pending a supplied/public checkpoint. Aborted partial runtime install attempts after checkpoint absence was confirmed. 2026-07-10T22:37:32+08:00 continuation recheck: P2PXML dataset remains staged/extracted, but no model_weights.pth/.pt/.tar checkpoint exists in source or dataset. Runtime install attempts were stopped because upstream inference checkpoint is unavailable/authenticated. 2026-07-10T22:55:53+08:00 SOP reclassification: moved from Production Blocked to Needs Token/License because dataset is staged, but upstream checkpoint is behind authenticated SharePoint/no public model artifact found.