mirror of
https://github.com/hanzoai/ml.git
synced 2026-11-20 23:08:29 +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]>
SigLIP
SigLIP is multi-modal text-vision model that improves over CLIP by using a sigmoid based loss, HuggingFace.
Running an example
$ cargo run --features cuda -r --example siglip
softmax_image_vec: [2.1912122e-14, 2.3624872e-14, 1.0, 1.0, 2.4787932e-8, 3.2784535e-12]
Results for image: hanzo-ml-examples/examples/stable-diffusion/assets/stable-diffusion-xl.jpg
Probability: 0.0000% Text: a cycling race
Probability: 0.0000% Text: a photo of two cats
Probability: 100.0000% Text: a robot holding a hanzo
Results for image: hanzo-ml-examples/examples/yolo-v8/assets/bike.jpg
Probability: 100.0000% Text: a cycling race
Probability: 0.0000% Text: a photo of two cats
Probability: 0.0000% Text: a robot holding a hanzo