stable-diffusion
stable-diffusion copied to clipboard
ModuleNotFoundError: no module named 'samplers'
after having a blast with CompVis stable-diffusion paired with v0.8 of optimisedSD, decided to give the latest version a go. ive grabed the latest version of stable-diffusion from CompVis as well as the latest OptimisedSD version and now im getting this error.
Global seed set to 27
Loading model from models/ldm/stable-diffusion-v1/model.ckpt
Global Step: 470000
Traceback (most recent call last):
File "optimizedSD/optimized_txt2img.py", line 204, in
any help would be great. im sure there is an easy solution to this. but ive had no luck at the moment.
kind regards.
I just made a duplicate issue a few seconds after yours, I'll go delete mine.
I just made a duplicate issue a few seconds after yours, I'll go delete mine.
i'll be sure to update this thread if i find the solution. i hope you do the same :)
I have no idea how any of this works, but this is a quick fix I came up with
-
pip install git+https://github.com/crowsonkb/k-diffusion.git
- inside optimizedSD/ddpm.py I replaced
from samplers import CompVisDenoiser, get_ancestral_step, to_d, append_dims,linear_multistep_coeff
with
from k_diffusion.external import CompVisDenoiser
from k_diffusion.utils import append_dims
from k_diffusion.sampling import get_ancestral_step, to_d, linear_multistep_coeff
Getting the same error. Looks like it was recently introduced, so a workaround could be to checkout commit 7ecf962 and run that until the issue is properly patched.
I have no idea how any of this works, but this is a quick fix I came up with
1. `pip install git+https://github.com/crowsonkb/k-diffusion.git` 2. inside optimizedSD/ddpm.py I replaced
from samplers import CompVisDenoiser, get_ancestral_step, to_d, append_dims,linear_multistep_coeff
with
from k_diffusion.external import CompVisDenoiser from k_diffusion.utils import append_dims from k_diffusion.sampling import get_ancestral_step, to_d, linear_multistep_coeff
cheers mate, this worked for me.
Hi, thank you all for reporting the bug and the workaround; I have fixed the issue now. Thanks!