Open archimedesinstitute opened 1 year ago
Ive gone ahead and reinstalled everything so we'll see how that goes!
copy the whole error log
Training the unet...
Traceback (most recent call last):
File "/content/diffusers/examples/dreambooth/train_dreambooth.py", line 785, in
"/content/models/Fortuna_Lofi does not appear to have a file named text_encoder/config.json"
make sure there is a folder named "Fortuna_Lofi" in "/content/models/"
In the sd folder rather than the fast-dreambooth, I assume? fast-dreambooth only has a Sessions folder, which leads to the individual named session folders.
or I should say fast-dreambooth>sessions>[folder name] is how it currently installs automatically.
This is how it's automatically setting up.
you can't resume training, there is no model in the output folder
I am not trying to resume training, I set up a new model.
I just ran a backup version of the fast-dreambooth model saved before yesterday and it's running fine, so it seems as though perhaps if there have been any updates, it is no longer automatically setting things correctly
This seems still to be an issue. Any clarity on other fixes that can be applied?
I think it's fixed now
Unfortunately when I tried it on a new model the same result. Let me know if it's something I'm doing? -
File "/content/diffusers/examples/dreambooth/train_dreambooth.py", line 787, in
1024 resolution is too much, use 704 or 768 (and checking the box reduce memory usage)
Ok cool, thanks!
Hello!
I have been using your dreambooth colab for a couple weeks, and it has always worked incredibly well.
I was launching a new training set last night, and although I was following the steps exactly as i always have, and using instances I have used before on another version of the finetune I have been working on, I have consistently gotten the following error after the instance images are attempted to be pulled from the the folder they have been uploaded to:
--max_train_steps=4000']' returned non-zero exit status 1. Something went wrong
Please let me know if there is something i have overlooked. I did confirm they have been uploaded correctly as I can see them in the Instances folder.