V-JEPA 2.1 – HuggingFace ports - a apiantonio Collection

apiantonio 's Collections

V-JEPA 2.1 – HuggingFace ports

V-JEPA 2.1 – HuggingFace ports

updated Jul 27

HF-format conversions of Meta's V-JEPA 2.1 encoders (384px), with numerical validation against the reference implementation.

apiantonio/vjepa2.1-vit-base-384

Video Classification • 0.1B • Updated Jul 26 • 1.22k • 2

Note 80M-class distilled encoder – bit-exact parity"
apiantonio/vjepa2.1-vit-large-384

Video Classification • 0.3B • Updated Jul 26 • 5.63k • 2

Note 300M-class distilled encoder – parity within FP accumulation noise
apiantonio/vjepa2.1-vit-giant-384

Video Classification • 1B • Updated Jul 26 • 205 • 2

Note 1B encoder – first HF port, bit-exact parity
apiantonio/vjepa2.1-vit-gigantic-384

Video Classification • 2B • Updated Jul 26 • 308 • 2

Note 2B encoder – first HF port, bit-exact parity