Open kopyl opened 1 year ago
This helped:
pip uninstall xformers
--xformers
from the train commandMaybe just removing --xformers
from the command could have helped, but i did not test it.
But still need a better fix...
Maybe it really doesn't work for python3.10.10
Re-building image with python3.10.6 -___–
Fuck, i spent so much time rebuilding it with python 3.10.6 instead of 3.10.10 and still getting the same error:
No such operator xformers::efficient_attention_forward_cutlass - did you forget to build xformers with `python setup.py develop`
I also tried:
apt-get install libgl1 -y
apt-get install libglib2.0-0 -y
python setup.py develop
Nothing works
Holly shit.
I just ran pip uninstall xformers
and pip install xformers
and it worked. Why?
And removed in ubuntu_setup.sh
this:
pip install -U -I --no-deps https://github.com/C43H66N12O12S2/stable-diffusion-webui/releases/download/linux/xformers-0.0.14.dev0-cp310-cp310-linux_x86_64.whl
And also why you don't include apt-get install libgl1 -y
and apt-get install libglib2.0-0 -y
in the ubuntu_setup.sh
script if it's required?
I had same issue, seems like it's not possible to run without xformers.
I see you mention also ubuntu_setup.sh so you probably refering to different repo (old version, ubuntu_setup.sh is no longer present) with GUI version:
@kgonia seems like I am using https://github.com/bmaltais/kohya_ss . Just checked, they removed ubuntu_setup.sh yesterday for some reason...
@kohya-ss sorry
@kopyl bmaltais/kohya_ss now have new script for installtion setup.sh
@kopyl bmaltais/kohya_ss now have new script for installtion setup.sh
Yeah, but it does not work :(
same issue here, can I really run without xformers?
You can use --mem_eff_attn
option, instead of --xformers
. This option will be slow, but reduce the memory usage same as xformers.
I also meet this question under ubuntu 20.04, finally I fix it by using the command pip install xformers==0.0.19
How to fix it?
Here is the full code: https://pastebin.com/d3xEGEhC
Running Python 3.10.10 and don't want to re-build and re-push and re-download a docker container just to fix this issue.