mirror of
https://github.com/hanzoai/ml.git
synced 2026-11-20 22:54:34 +00:00
Three-way per file through the rename: base 39355c6c91, theirs the upstream
merge of main with 35d7ae7c (via 7ef6a694d), ours origin/main.
Lands the barrier thread pool, runtime CPU feature dispatch, aarch64 and x86
repacked quantized kernels (indexed_gemv, gemv_fused_shared_lhs), quantized
embedding on CPU/Metal/CUDA, paged flash-attn kernels, cutile fused MoE, the
in-place op aliasing API, parking_lot storage locks, Metal buffer builders and
debug labels, safetensors 0.8, hf-hub 1.0, tokenizers 0.23.
Ours keeps ROCm, Vulkan, wgpu, the IQ/MXFP4/NVFP4/TQ dtypes, and the i-quant
weight/activation block split in matmul. Our backends ride InplaceOpN and the
quantized embedding; Metal get_rows gains the IQ kernels. ug leaves, as upstream.
Co-authored-by: Hanzo Dev <[email protected]>
hanzo-ml-convmixer
A lightweight CNN architecture that processes image patches similar to a vision transformer, with separate spatial and channel convolutions.
ConvMixer from Patches Are All You Need? and ConvMixer.
Running an example
$ cargo run --example convmixer --release -- --image hanzo-ml-examples/examples/yolo-v8/assets/bike.jpg
> mountain bike, all-terrain bike, off-roader: 61.75%
> unicycle, monocycle : 5.73%
> moped : 3.66%
> bicycle-built-for-two, tandem bicycle, tandem: 3.51%
> crash helmet : 0.85%