hollowstrawberry / kohya-colab

Accessible Google Colab notebooks for Stable Diffusion Lora training, based on the work of kohya-ss and Linaqruf
GNU General Public License v3.0
599 stars 86 forks source link

Error in trainer XL #156

Closed gwhitez closed 1 month ago

gwhitez commented 4 months ago

When training on the xl trainer (any model) this error occurs.

image

kwapdt commented 4 months ago

Getting the same error, please help

kyle215ps3 commented 4 months ago

Getting the same error, please help

I found the fix, go to the main cell and double click any empty space, scroll down and look for something like this on the left, it's lower than the ui settings, now you will see the packages with their versions. look for torch==2.2.1+cu121 (It's in the beginning of the group of packages with their versions like below, mine is changed but it's the same spot) now change the ==2.2.1 to 2.3.0, I can now train with it. and it's working

image

torch==2.3.0+cu121 you can copy and paste this and make sure the spacing isn't changed.

hollowstrawberry commented 4 months ago

Must have been another colab update, if torch==2.3.0+cu121 fixes it then I'll update it later today, thanks!

kyle215ps3 commented 4 months ago

Must have been another colab update, if torch==2.3.0+cu121 fixes it then I'll update it later today, thanks!

yep it's fixed, I'm currently training XL lora, also I think huggingface_hub was changed too I had to change it to not get this pop up error I had. so changed to huggingface-hub==0.23.0

kyle215ps3 commented 4 months ago

Must have been another colab update, if torch==2.3.0+cu121 fixes it then I'll update it later today, thanks!

oh and you're welcome :) thank you for making an awesome training for us to use :D

gwhitez commented 4 months ago

Getting the same error, please help

I found the fix, go to the main cell and double click any empty space, scroll down and look for something like this on the left, it's lower than the ui settings, now you will see the packages with their versions. look for torch==2.2.1+cu121 (It's in the beginning of the group of packages with their versions like below, mine is changed but it's the same spot) now change the ==2.2.1 to 2.3.0, I can now train with it. and it's working

image

torch==2.3.0+cu121 you can copy and paste this and make sure the spacing isn't changed.

thanks, in fact with the help of bing I also fixed it and with the same parameters, I came to comment that :)

kyle215ps3 commented 4 months ago

Getting the same error, please help

I found the fix, go to the main cell and double click any empty space, scroll down and look for something like this on the left, it's lower than the ui settings, now you will see the packages with their versions. look for torch==2.2.1+cu121 (It's in the beginning of the group of packages with their versions like below, mine is changed but it's the same spot) now change the ==2.2.1 to 2.3.0, I can now train with it. and it's working image torch==2.3.0+cu121 you can copy and paste this and make sure the spacing isn't changed.

thanks, in fact with the help of bing I also fixed it and with the same parameters, I came to comment that :)

haha I keep forgetting how advanced bing has become hahaha xD probably could have went there too xD

tomokiya0627 commented 2 months ago

Hello. I need your help with this issue again. Even when I use the updated TrainerXL, the same error happens again. I would appreciate it if you could provide a solution. I appreciate your help. スクリーンショット 2024-07-17 133600

gwhitez commented 2 months ago

Hello. I need your help with this issue again. Even when I use the updated TrainerXL, the same error happens again. I would appreciate it if you could provide a solution. I appreciate your help. スクリーンショット 2024-07-17 133600

replace torch==2.3.0 by torch==2.3.1

hollowstrawberry commented 2 months ago

replace torch==2.3.0 by torch==2.3.1

Just pushed this change, let me know if it works now

gwhitez commented 2 months ago

Just pushed this change, let me know if it works now

if it works, I'm training now image

tomokiya0627 commented 2 months ago

replace torch==2.3.0 by torch==2.3.1

Just pushed this change, let me know if it works now

I appreciate your help. It works and I am training now, too.