Dataset Viewer
The dataset viewer is not available for this dataset.
Unexpected token '<', "<html>
<h"... is not valid JSON
Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
0511 Backup
Created on 2026-05-11 in /data/home/azanette/DSGRPO.
Layout
artifacts/hf_models/: physical copy of checkpoint directories fromhf_models/.artifacts/wandb/: physical copy of remainingwandb/offline-run-*directories.artifacts/logs/: training logs.artifacts/repo_files/: launch/train scripts plus the sourcerun_inventory.csv.index/backup_run_index.csv: main run index, one checkpoint run per row, with original and backup paths.index/source_run_inventory.csv: original inventory generated before backup.index/project_summary.csv: per-project count and size summary.index/unmapped_wandb_runs.csv: W&B offline dirs with no matching checkpoint run.index/copy_units.csv: top-level copy plan.
Counts
- Checkpoint runs: 41
- Checkpoint directories: 257
- Checkpoint bytes in indexed runs: 2172570681345 (2.0 TiB)
- Mapped W&B offline dirs: 47
- Total W&B offline dirs copied: 56
- Unmapped W&B offline dirs copied but separated in index: 9
Notes
The two stopped entropy seed102 runs and their W&B dirs were removed before this backup. This folder is intended as a staging tree for Hugging Face upload.
Upload Helper
After index/COPY_DONE exists, upload with:
HF_REPO_TYPE=dataset HF_PRIVATE=0 HF_UPLOAD_NUM_WORKERS=2 bash 0511_backup/scripts/upload_to_hf.sh <namespace/repo_name>
The script refuses to upload before the backup copy has completed.
- Downloads last month
- 23