---
license: other
license_name: voxceleb2-derived
pretty_name: bulk_v1 progress sample
tags:
  - deepfake-detection
  - voxceleb2
configs:
  - config_name: ff77
    default: true
    data_files:
      - split: train
        path: data/ff77/train-*
  - config_name: ff153
    data_files:
      - split: train
        path: data/ff153/train-*
  - config_name: s2v77
    data_files:
      - split: train
        path: data/s2v77/train-*
  - config_name: wan77
    data_files:
      - split: train
        path: data/wan77/train-*
  - config_name: s2v153
    data_files:
      - split: train
        path: data/s2v153/train-*
---

# bulk_v1 progress sample

A **random** look at the clips still on disk (withdrawn rejects are excluded). Not a
release: the whole dataset is replaced on every run. `selection` is kept/rejected
once a slice is gated, and `not gated yet` only for slices that never ran the gate.

- drawn: 2026-09-25 10:45, seed 20260925, 3 pair-set(s) per family x owner x slice, slices 2 and up
- `selection`: `kept` / `rejected` after `GATE_DONE` or the drop ledger (or refill kept lists); `not gated yet` only if the slice was never gated
- one config per family; `video` is a browser-playable remux of the shipped `.mov`
- the same mp4s sit under `preview/<config>/` as plain files, playable from the Files tab

| config | clips |
|---|---|
| ff77 | 518 |
| ff153 | 525 |
| s2v77 | 500 |
| wan77 | 498 |
| s2v153 | 468 |

Refresh: `python src/deepfakes/generation/slurm/hf_progress_sample.py` from the repo (`Deep-Fake/shaina-progress-sample`).