wire issue need help 'AssertionError: conversion failed: lora_unet_conv_in. the model may not be trained by `sd-scripts`.'

[napoleonchn]

wire issue need help 'AssertionError: conversion failed: lora_unet_conv_in. the model may not be trained by `sd-scripts`.'

[jayododds]

also getting this on self trained loras. I'll try some Civitai downloads to confirm

[jayododds]

Civitai downloads seem to work for me - so it would seem there is something in the LORAs that I have created that is causing the issue. Likely a meta data issue or something of that nature - I'm looking into it. I'll let you know if I find anything that might help @napoleonchn, but don't hold your breath :)

[ToolB0xx]

any solution? getting this on all my trained loras

[jayododds]

Hey @ToolB0xx and @napoleonchn what are you training your models in? I had a couple that were trained on a mac with Draw Things - which work fine elsewhere but don't seem to work with this extension at all. I did a couple of new trainings with Kohya and they seem to work flawlessly. I think it comes down to what data is being saved in the .safetensor.