FLUX.1-dev-fp8

lastly work on 12G -GPU

work flow for this model

https://civitai.com/models/622932?modelVersionId=696399

work flow for ( fp8 comfyui editions)

https://civitai.com/models/628210/advanced-ai-art-remix-workflow-for-flux

Utilizes Flux1-dev and Flux1-schnell models

what do you need

1- t5xxl_fp8_e4m3fn.safetensors https://huggingface.co/stabilityai/stable-diffusion-3-medium/resolve/main/text_encoders/t5xxl_fp8_e4m3fn.safetensors?download=true

clip_l.safetensors https://huggingface.co/stabilityai/stable-diffusion-3-medium/resolve/main/text_encoders/clip_l.safetensors?download=true

2- ae.sft https://huggingface.co/black-forest-labs/FLUX.1-dev/resolve/main/ae.sft?download=true

https://huggingface.co/black-forest-labs/FLUX.1-dev/resolve/main/vae/diffusion_pytorch_model.safetensors?download=true (168 MB) big thanks for @OliviaRossi

3- flux1-dev-fp8.safetensors which is this file

here quick tutorial of Pixart sigma model quantization

big thanks for comfyui team for this update

big thanks for

Kijai

you can find schnell version here also

https://huggingface.co/Kijai/flux-fp8