yolo26 and yolo26-seg use a top level size directory because they publish
five sizes. This repo publishes one, so it stays flat, like blazeface and
rfdetr-nano-detector. The corrected configs from the previous commit are
unaffected.
Move the size-n exports under n/ and rewrite the backend configs
Matches the yolo26 and yolo26-seg layout, where the model size is the top
level directory and the input resolution stays in the filename.
The xnnpack config described a single yolo26_pose_n_xnnpack_fp32.pte with
forward_384/512/640 methods. That file was never published here: the repo
ships three files, each with a single forward. Core ML had no config at all.
The previous removal commit said earlier tags still contain the MLX artifacts. That is wrong: MLX for this model was only ever published on main / v0.10.0, and no tag at v0.9.0 or below ever carried it. The files remain recoverable from git history. v0.9.0 and below are, and have been, untouched.
Remove MLX artifacts: 7.5-19.6x slower than Core ML on device
MLX measured substantially slower than Core ML on an iPhone 16 for every convolutional vision model tested (efficientnet-v2-s 19.6x, yolo26-s 16.0x, yolo26-seg-s 7.5x, yolo26-pose-n 10.0x). MLX weight quantization also cannot shrink a conv model, since it only rewrites nn.Linear/embedding modules.
The .pte and config.json are removed; each mlx/ dir keeps a NOTES.md with the measurements. The exporter is retained in export-scripts so this can be re-tested against a future ExecuTorch/MLX release.
Earlier tags (v0.9.0 and below) are untouched and still contain the MLX artifacts.
Rebuild mlx export with fork toolchain (ExecuTorch 1.4.1)
