Commit History
-
Re-convert: store sam2_convs and boxes_pool_project in MLX (channels-last) layout
Verified signatureRe-converted from `facebook/sam3` so the 12 conv weights that were published in PyTorch (channels-first) layout are stored in MLX layout. ### Why `sam3/convert.py` in [Deekshith-Dade/mlx_sam3](https://github.com/Deekshith-Dade/mlx_sam3) transposed the SAM-3 neck convs via a hard-coded key list. Anything outside that list shipped channels-first: the 11 `sam2_convs.*` twins (the second neck, built only when `enable_inst_interactivity=True`) and `geometry_encoder.boxes_pool_project`. The text-prompt path never builds `sam2_convs`, which is why this only showed up on the box-prompt path, where `set_image` crashed in `mx.conv_transpose2d`. Fixed upstream in two steps: - [PR #9](https://github.com/Deekshith-Dade/mlx_sam3/pull/9) (merged) — a shape-guarded, idempotent `_sanitize_conv_layout` that repairs these weights at load time. - [PR #10](https://github.com/Deekshith-Dade/mlx_sam3/pull/10) — makes `convert.py` derive the permutations from the model's module types instead of a hard-…60b0bef
-
model files for image tasks
DeekshithVerified signatureb72a14d -
Update Readme
DeekshithVerified signature4c25caf -
initial commit
DeekshithVerified signatureeb91b3e