Kiran Ryali comments

Repositories
Issues
Comments

Results 1 comments of


                                            Kiran Ryali

flux fill fp8 load failed

@Suprhimp you could try the gguf quantize method that was added very recently ```python transformer = FluxTransformer2DModel.from_single_file( "https://huggingface.co/YarvixPA/FLUX.1-Fill-dev-gguf/blob/main/flux1-fill-dev-Q4_0.gguf", quantization_config=GGUFQuantizationConfig(compute_dtype=torch.bfloat16), torch_dtype=torch.bfloat16, ) pipe = FluxFillPipeline.from_pretrained( "black-forest-labs/FLUX.1-Fill-dev", transformer=transformer, torch_dtype=torch.bfloat16, ).to(device) pipe.enable_model_cpu_offload() mask_image...