Views
No views yet
web-xpu-opsTongyi-MAI/Z-Image,
Apache-2.0, for the WebGPU demo in
web-xpu-ops.| file | what | size |
|---|---|---|
dit.manifest.json, dit.q8.bin, dit.q8scales.bin, dit.f32.bin | the diffusion transformer, per-row absmax int8 | 6.17 GB |
manifest.json, decoder.bin | the VAE decoder, repacked flat | 0.20 GB |
config.json, model.safetensors.index.json, model-0000{1,2,3}-of-00003.safetensors | the Qwen3-4B text encoder, unmodified | 8.04 GB |
Tongyi-MAI/Z-Image — it was left there in a first version,
on the argument that copying it duplicated what its own publisher already
served. That argument is sound and the trade is not: two hosts is two things
that can move, and a demo that fills a folder from one place is one thing to
explain. Apache-2.0 permits the copy; the copy is what is here.[-127, 127], four codes per 32-bit word, least-significant byte first,
with one f32 scale per output row applied after the dot product;dit.f32.bin — norms, biases, the patch
embedder — because quantizing a [3840] norm weight buys nothing and costs
resolution;Tongyi-MAI/Z-Image publishes.adaLN_modulation.0.weight is the worst at
4.78% relative RMS on layers.0's output, attention.to_q 1.21%, the rest
below that — which is why a mixed format exists that keeps adaLN at q8 while
taking everything else to q4. This repository is the straight q8 conversion; the
numbers above are in examples/zimage/README.md.examples/zimage-web.
Nothing else reads this layout; for any other purpose take the model from
Tongyi-MAI/Z-Image and get f32
weights rather than a lossy conversion.