Skip to content

Add V-JEPA2 ViT-G#2321

Open
KartikP wants to merge 2 commits into
masterfrom
kp/vjepa2-backbone
Open

Add V-JEPA2 ViT-G#2321
KartikP wants to merge 2 commits into
masterfrom
kp/vjepa2-backbone

Conversation

@KartikP
Copy link
Copy Markdown
Collaborator

@KartikP KartikP commented Mar 26, 2026

  • Add V-JEPA2 ViT-G (facebook/vjepa2-vitg-fpc64-256) as a standalone brain-score vision model plugin
  • Loads directly from HuggingFace, no dependency on TRIBEv2 package
  • Wraps the video backbone for single-image input (1-frame video) using standard PytorchWrapper

This model is added because it is the backpoint of TRIBEv2 (d'Ascoli et al., 2026) for brain encoding. Makes it a good candidate for neural vision benchmarking.

@KartikP KartikP added the OOM When unit tests run out of memory label Mar 26, 2026
@KartikP KartikP closed this Mar 26, 2026
@KartikP KartikP reopened this Mar 26, 2026
@KartikP
Copy link
Copy Markdown
Collaborator Author

KartikP commented Mar 26, 2026

Starting layer mapping. Unit tests will re-run after completion.

1 similar comment
@KartikP
Copy link
Copy Markdown
Collaborator Author

KartikP commented Mar 26, 2026

Starting layer mapping. Unit tests will re-run after completion.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

OOM When unit tests run out of memory Skip trigger-mapping

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant