FelTris/pjepa_features
P-JEPA feature archives Precomputed input features and aligned annotations for the P-JEPA encoders. These are backbone/pooler outputs, before P-JEPA temporal encoding. No raw videos or P-JEPA output caches are included. All feature tensors are float32. The files use the directory layout expected by the P-JEPA code. That GitHub repository is currently private pending the code release. The plain PyTorch example below works without it; the full training/evaluation recipes require… See the full description on the dataset page: https://huggingface.co/datasets/FelTris/pjepa_features.
0
