Replies: 4 comments 1 reply
-
|
i dont support ip adapter or inpainting training |
Beta Was this translation helpful? Give feedback.
-
|
Hi, Can I use image pairs(raw_image+mask) to train flux-fil-dev model? |
Beta Was this translation helpful? Give feedback.
-
|
Thank you for your help. How I can save the quantized weight files obtained using your framework, specifically in the safetensors format? |
Beta Was this translation helpful? Give feedback.
-
|
do you mean the int8 or fp8 versions created at startup? currently you cannot. |
Beta Was this translation helpful? Give feedback.
-
|
I am very interested in this topic @yyyyyxie did you managed to train loras with flux-fill? |
Beta Was this translation helpful? Give feedback.
-
Hi,
I want to training the flux-fill-dev, but I’m unsure where to start. Specifically, I always seem to get the latent representations using VAE, but I’m unsure how to concatenate the latent and the mask according to the channel concatenation method described in flux-fill-dev.
Beta Was this translation helpful? Give feedback.
All reactions