mirror of
https://github.com/hanzoai/ml.git
synced 2026-11-19 23:50:51 +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-wuerstchen: Efficient Pretraining of Text-to-Image Models
The wuerstchen example is a port of the diffusers
implementation for Würstchen v2.
The hanzo implementation reproduces the same structure/files for models and
pipelines. Useful resources:
- Official implementation.
- Arxiv paper.
- Blog post: Introducing Würstchen: Fast Diffusion for Image Generation.
Getting the weights
The weights are automatically downloaded for you from the HuggingFace
Hub on the first run. There are various command line
flags to use local files instead, run with --help to learn about them.
Running some example.
cargo run --example wuerstchen --release --features cuda,cudnn -- \
--prompt "Anthropomorphic cat dressed as a fire fighter"
The final image is named sd_final.png by default.
