Files
zeekayandHanzo Dev 6a363f7ee3 merge: upstream main ddf1b879 and the aarch64 repack branch 35d7ae7c
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]>
2026-09-15 15:52:28 -07:00
..
2026-01-16 12:30:55 -08:00
2026-01-16 12:30:55 -08:00
2026-01-16 12:30:55 -08:00

Running Whisper Examples

Here, we provide two examples of how to run Whisper using a ML-compiled WASM binary and runtimes.

Pure Rust UI

To build and test the UI made in Rust you will need Trunk From the hanzo-ml-wasm-examples/whisper directory run:

Download assets:

# mel filters
wget -c https://huggingface.co/spaces/lmz/candle-whisper/resolve/main/mel_filters.safetensors
# Model and tokenizer tiny.en
wget -c https://huggingface.co/openai/whisper-tiny.en/resolve/main/model.safetensors -P whisper-tiny.en 
wget -c https://huggingface.co/openai/whisper-tiny.en/raw/main/tokenizer.json -P whisper-tiny.en
wget -c https://huggingface.co/openai/whisper-tiny.en/raw/main/config.json -P whisper-tiny.en
# model and tokenizer tiny multilanguage
wget -c https://huggingface.co/openai/whisper-tiny/resolve/main/model.safetensors -P whisper-tiny
wget -c https://huggingface.co/openai/whisper-tiny/raw/main/tokenizer.json -P whisper-tiny
wget -c https://huggingface.co/openai/whisper-tiny/raw/main/config.json -P whisper-tiny

#quantized 
wget -c https://huggingface.co/lmz/candle-whisper/resolve/main/model-tiny-en-q80.gguf -P quantized
wget -c https://huggingface.co/lmz/candle-whisper/raw/main/tokenizer-tiny-en.json -P quantized
wget -c https://huggingface.co/lmz/candle-whisper/raw/main/config-tiny-en.json -P quantized



# Audio samples
wget -c https://huggingface.co/datasets/Narsil/candle-examples/resolve/main/samples_gb0.wav -P audios
wget -c https://huggingface.co/datasets/Narsil/candle-examples/resolve/main/samples_a13.wav -P audios
wget -c https://huggingface.co/datasets/Narsil/candle-examples/resolve/main/samples_gb1.wav -P audios
wget -c https://huggingface.co/datasets/Narsil/candle-examples/resolve/main/samples_hp0.wav -P audios
wget -c https://huggingface.co/datasets/Narsil/candle-examples/resolve/main/samples_jfk.wav -P audios
wget -c https://huggingface.co/datasets/Narsil/candle-examples/resolve/main/samples_mm0.wav -P audios

Run hot reload server:

trunk serve --release --public-url / --port 8080

Vanilla JS and WebWorkers

To build and test the UI made in Vanilla JS and WebWorkers, first we need to build the WASM library:

sh build-lib.sh

This will bundle the library under ./build and we can import it inside our WebWorker like a normal JS module:

import init, { Decoder } from "./build/m.js";

The full example can be found under ./lib-example.html. All needed assets are fetched from the web, so no need to download anything. Finally, you can preview the example by running a local HTTP server. For example:

python -m http.server

Then open http://localhost:8000/lib-example.html in your browser.