The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.
YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
PARC2026 RunPod artifacts — 2026-09-17
Public backup of retained Track 1 artifacts from the retired 150 GB network volume and 80 GB Pod disk. No training or evaluation was performed as part of this backup. Prior experimental outcomes remain unchanged.
| Archive | Preserved paths | Unique tar bytes | Parts |
|---|---|---|---|
track1 |
2670 | 17112360960 | 16 |
track1-volume80 |
58063 | 54961408000 | 52 |
All archive parts were read back without authentication and checked against SHA-256. Manifests map original paths to deduplicated objects/<sha256> entries. The archives contain models, code, and evaluation evidence, not a single ready-to-train dataset. Caches, environments, and Git internals are excluded and listed in the manifests. Git internals from the former GitHub backup are retained locally outside this public repository.
Restore
Pin the repository revision before downloading. Install huggingface_hub for the hf CLI, then:
hf download ekunish/parc2026-runpod-artifacts-20260917 --repo-type dataset --revision REVISION --include 'track1/*' --local-dir backup
python restore_public.py backup/track1/PUBLIC_BACKUP_COMPLETE.json backup/track1 restored-track1
hf download ekunish/parc2026-runpod-artifacts-20260917 --repo-type dataset --revision REVISION --include 'track1-volume80/*' --local-dir backup
python restore_public.py backup/track1-volume80/BACKUP80_COMPLETE.json backup/track1-volume80 restored-volume80
Use a new destination directory and enough free disk for downloaded tar parts plus restored logical files. No archive is loaded fully into RAM. track1 also records 89 original upstream files by immutable Hugging Face revision; pass --include-upstream to restore those as well. track1-volume80 stores all inventoried working files directly.
Internal absolute symlinks under the original workspace are relocated into the restore destination. External symlinks are retained as manifest information and reported in RESTORE_RECEIPT.json; the restore script does not create links outside the destination. It does not claim that external targets lost before this backup were recovered. Environment/cache exclusions require reinstallation.
Existing upstream model, code, and data licenses continue to apply; this backup grants no new license. Inspect each retained package and its upstream repository before reuse.
- Downloads last month
- 481