FLUX

• FLUX.1-dev
  • https://huggingface.co/spaces/jasperai/Flux.1-dev-Controlnet-Upscaler
  • https://huggingface.co/showlab/OmniConsistency
• FLUX.2-schnell
  • Apache-2.0
  • https://huggingface.co/lodestones/Chroma
    • 训练过程 https://training.lodestone-rock.com/
      • large 1024x1024
      • fast 512x512
• 使用 Transformer 而不是 传统 CNN UNet
• 参考
  • https://comfyui-wiki.com/en/tutorial/advanced/flux1-comfyui-guide-workflow-and-examples
  • https://huggingface.co/models?other=base_model:adapter:black-forest-labs/FLUX.1-dev

# 同时下载了 flux1-dev.safetensors 和 transformers - 有点重复，有点大
huggingface-cli download black-forest-labs/FLUX.1-dev \
  --local-dir . --local-dir-use-symlinks False

model	size	notes
flux1-schnell.safetensors	23.8GB	VRAM 16GB+
flux1-dev.safetensors	23.8GB
Flux.1 Dev FP8	17.2GB	VRAM 8GB+
clip_l.safetensors	246 MB
t5xxl_fp8_e4m3fn.safetensors	4.89 GB	lower memory usage, 8-12GB
t5xxl_fp16.safetensors	9.79 GB	better results, 32GB+
ae.safetensors	335 MB

https://huggingface.co/lodestones/Chroma/resolve/main/ae.safetensors
https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors
https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors

https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors