YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

SigLIP2 So400m temporal action-chunk robotics VLA

Self-contained policy produced by train_siglip2_action_chunk_vla.py.

  • Frozen SigLIP2 So400m 384/14 visual encoder
  • Official global pooler token + pooled 8x8 spatial patch tokens (no CLS)
  • Frozen SigLIP2 text encoder and normalized 1152D instruction embeddings
  • Lowercased SentencePiece instructions with a 64-token context
  • Raw-RGB spatial color tokens
  • Four-step visual/proprio/phase history
  • Three-head, eight-step 7D action chunks
  • Overlapping temporal ensemble and gripper hysteresis at inference
  • Parameters: 1,138,041,878
  • Selected/refit epochs: 31

Upload every file in this directory. The adapter loads model.safetensors strictly and performs no model-hub or network access.

The SigLIP2 checkpoint is distributed under Apache License 2.0. See LICENSE and ATTRIBUTION.md.

Downloads last month

-

Downloads are not tracked for this model. How to track
Safetensors
Model size
1B params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support