🚀 Universal Z-Image Quantizer

Convert sharded Z-Image models directly on Hugging Face to floating-point precisions (FP8/FP16/BF16) or dynamically trigger symmetric integer quantization (INT8).

Hugging Face Token (Write Access)

Hugging Face Username

Source Repository

Components to Quantize

text_encoder transformer vae

Target Precision

⚠️ INT8 Warning: Modifies layer keys (weight_int8, weight_scale). Requires the custom NativeInt8Linear XPU inference code to run.

Target Repository (Auto-generated)

Operation Logs