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-endocec
EnCodec is a high-quality audio compression model using an encoder/decoder architecture with residual vector quantization.
Running one example
cargo run --example encodec --features encodec --release -- code-to-audio \
hanzo-ml-examples/examples/encodec/jfk-codes.safetensors \
jfk.wav
This decodes the EnCodec tokens stored in jfk-codes.safetensors and generates
an output wav file containing the audio data.
Instead of code-to-audio one can use:
audio-to-audio in.mp3 out.wav: encodes the input audio file then decodes it to a wav file.audio-to-code in.mp3 out.safetensors: generates a safetensors file containing EnCodec tokens for the input audio file.
If the audio output file name is set to -, the audio content directly gets
played on default audio output device. If the audio input file is set to -, the audio
gets recorded from the default audio input.