Error while deserializing header: MetadataIncompleteBuffer
Getting the following error trying to run my own LoRA model for the first time and wondered if anyone can help. It's for the model "ComfyUI with Flux.1 dev one-click". I have my own LoRA's that I used jupyter to put into the files. The models are visible in the workflow, but immediately gives the attached error when I start to generate. Any advice?
Thank you!
Error:
16 Replies
It might be because of the lora is corrupted
How did you transfer it
To the pod
Placed it into a Google Drive from my Azure VM, opened from Google Drive into the jupyter system and then opened it in the LoRA
Okay, how do you "open Google drive into jupyter"?
Within my Mac's Google Chrome and then copied them by dragging.
Dragging into jupyter
Ah there
Try downloading using wget or something else from maybe Azure blob storage
Or rclone ✨ if it works with Google drive
So don't use jupyter drag and drop
Ok. I saw the rclone instructions but they made no sense
I'm not very technical as you can see
Mmm don't use that tool then
Yes, try some and if you experience any difficulties, can ask here too
I'll try to help
Ok thanks. Is wget the connection service between storage?
Wget is like a command in linux to download things I guess, Google is cool too btw
Thanks. Trying to use the following that I found. Not working so far...
Yeah I don't think it's working with googledrive
If u use gdrive use rclone
Finally worked. After about a million errors. Thanks again.
Well, I say worked. One of the Lora's is just generating random images totally unrelated to the trained images. No idea why. The rest are working.
Ooh how did you download it
Maybe the lora itself, try on another generator to confirm
Webui*
Sure enough. I loaded my mult-lora workflow into it and the LoRA that wasn’t working now works…. I used the gdown to download the model files
Ahh
Nc
Yeah if the generated imgs are broken then it's the lora
Ok