FLUX.2 [klein] 4B for the Snapdragon 8 Elite neural unit
A phone conversion of FLUX.2 [klein] 4B, compiled as QNN context binaries for the Hexagon v79 neural unit (Snapdragon 8 Elite, e.g. iQOO 13). It draws 512x512 pictures in 4 steps and edits a picture from an instruction.
| Files | What they do |
|---|---|
text0.bin ... text2.bin, klein_qwen_q8.raw, qwen_tokenizer.json |
the prompt reader (Qwen3-4B) |
part0.bin ... part3.bin |
the drawing model |
enc.bin, dec.bin |
taef2 picture encoder and decoder |
manifest.json |
file list with sizes and SHA-256, used by the app's download |
The weights were converted to ONNX, quantized and compiled for the neural unit, so pictures may differ slightly from the original model's. Licences: FLUX.2 [klein] 4B (Apache 2.0, LICENSE.md), Qwen3-4B (Apache 2.0), taef2 by madebyollin (MIT).
The app, and building these files yourself
These files are made for the Anima app, which downloads them itself.
The scripts that build them from the original model weights, with a step-by-step guide, are in
build/ in this repo.