Results 726 comments of bmaltais

I have implemented the feature via https://github.com/StartHua/Comfyui-Mine-PhotoMaker/pull/4

Does this image exist? the sd-script complain it can't find it: `D:\github\lora\images\10_level\1.jpg` You may have to open an issue with kohya directly at https://github.com/kohya-ss/sd-scripts as this appear to be something...

The CLI is still there. The GUI is just a wrapper on top. I just refocussed the README around the GUI... but the clio based code is well and alive.

Talking with a few model builders and devs we generally find that the 3 files produced by the trainer will make it difficult to share. One suggestion... would it be...

The MVP for training is ready for use: ![image](https://github.com/replicate/cog-sdxl/assets/7474674/2d0fb3ab-522e-4595-899e-827964defbfd)

The cog-sdxl-webui will now generate a proper embeddings.safetensors that can be used by ComfyUI to generate outputs from the lora.safetensors + embeddings.safetensors files. Image generated by ComFY UI: ![ComfyUI_00884_](https://github.com/replicate/cog-sdxl/assets/7474674/75593e48-5069-41e2-9dde-652b53f15501)

Hi, this is an issue with the training scripts themselves. I don't maintain them. You need to open an issue directly on kohya's sd-scripts repo to get this looked into:...

I am not sure why they are not created in runpod... is there an error message being displayed in the output when it can't create the file? Might point to...

Good catch. I will see where the root cause of this is.

I have fixed the issue in the `dev` branch