Buckets:

KaisResearch's picture
download
raw
397 Bytes
# Full training run (HF Jobs / cloud GPU).
# Usage: python src/train.py --config configs/full.yaml
# When run via `hf jobs ... -v hf://datasets/cam-kai-ml/ML-Lip-Reader:/data`,
# point data_roots at wherever the preprocessed ROI tree lives in the mount.
data_roots: /data/lrs2,/data/lrs3
epochs: 40
batch_size: 64
num_workers: 4
lr: 3.0e-4
warmup_epochs: 2
out_dir: /checkpoints/kai
log_every: 50

Xet Storage Details

Size:
397 Bytes
·
Xet hash:
8033b06b0e631cc4d62346ef247050ad292dfa5eeb6b939ea4c14357ce5217bd

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.