Closed EKKIV0K closed 8 months ago
git reset --hard <commit number>
git reset --hard <commit number>
do i need a fresh install of the current version ? @brknsoul
No, you can do that on your current install.
Download the default model? (y/N) n
14:24:28-021354 ERROR codeformer: ModuleNotFoundError
┌───────────────────────────────────────── Traceback (most recent call last) ─────────────────────────────────────────┐
│ C:\SDNEXT\automatic\modules\codeformer_model.py:31 in setup_model │
│ │
│ 30 │ │ from torchvision.transforms.functional import normalize │
│ > 31 │ │ from modules.codeformer.codeformer_arch import CodeFormer │
│ 32 │ │ from basicsr.utils import img2tensor, tensor2img │
│ │
│ C:\SDNEXT\automatic\modules\codeformer\codeformer_arch.py:10 in
@brknsoul
Yeah, that's why old versions aren't supported.
I'd just stick with the new versions, old gens be damned.
and .......
14:28:13-896179 WARNING Setup complete with errors: 1
14:28:13-897190 WARNING See log file for more details: C:\SDNEXT\sdnext.log
14:28:13-914502 INFO Extension preload: {'extensions-builtin': 0.01, 'extensions': 0.0}
14:28:13-914502 INFO Command line args: []
┌───────────────────────────────────────── Traceback (most recent call last) ─────────────────────────────────────────┐
│ C:\SDNEXT\launch.py:213 in
safer way than git reset --hard
is simply git checkout 89ba8e3c
and delete venv
since requirements have changed drastically since then.
but that version is from 09/20/2023 - 5 months ago. if you have performance degradations, document them so they can be solved.
sure, you're allowed to use that commit - but its beyond obsolete and not supported by any means. we're not talking about going back a version or two, we're talking about 5 months.
@vladmandic
Hi Vlad i hope you are fine :) So, the better way is :
1st install the last version, 2nd close the cmd window and type "git checkout 89ba8e3c" in the sdnext folder, 3rd close the git window and then delete the virtual environment folder in sdnext and start ?
you know i don't want to bother any of you, it's crucial for me to get back to this version cause ALL of my last works are based on that specific version.
@vladmandic
Hi Vlad i hope you are fine :) So, the better way is :
1st install the last version, 2nd close the cmd window and type "git checkout 89ba8e3" in the sdnext folder, 3rd close the git window and then delete the virtual environment folder in sdnext and start ?
correct
let's try this :)
@vladmandic
i was thinking about a thing that will be great for people like me, Why don't include in the metadata of a generated images more infos like : Advanced settings used (not the version of sdnext cause it's already in) like the settings in config file after modification if modifications happened into it ?
cause sometimes we can do good settings for sdnext that will works with our purposes, but after reinstallation we loose all of it and can't remember the things we made....
@vladmandic
outch after you tricks here is the problems :
14:45:56-348238 ERROR codeformer: ModuleNotFoundError
┌───────────────────────────────────────────────────────────────────────── Traceback (most recent call last) ─────────────────────────────────────────────────────────────────────────┐
│ C:\SDNEXT\modules\codeformer_model.py:31 in setup_model │
│ │
│ 30 │ │ from torchvision.transforms.functional import normalize │
│ > 31 │ │ from modules.codeformer.codeformer_arch import CodeFormer │
│ 32 │ │ from basicsr.utils import img2tensor, tensor2img │
│ │
│ C:\SDNEXT\modules\codeformer\codeformer_arch.py:10 in
@vladmandic and another one ....
14:48:11-634503 INFO Command line args: []
┌───────────────────────────────────────── Traceback (most recent call last) ─────────────────────────────────────────┐
│ C:\SDNEXT\launch.py:213 in
@vladmandic (LAST UPDATE)
15:27:00-875148 WARNING Setup complete with errors: 1
15:27:00-876144 WARNING See log file for more details: C:\SDNEXT\sdnext.log
15:27:00-901191 INFO Extension preload: {'extensions-builtin': 0.02, 'extensions': 0.0}
15:27:00-904184 INFO Command line args: []
15:27:11-937292 INFO Engine: backend=Backend.ORIGINAL compute=cuda mode=no_grad device=cuda
15:27:11-979055 INFO Device: device=NVIDIA GeForce RTX 2080 SUPER n=1 arch=compute_37 cap=(7, 5) cuda=11.8
cudnn=8700 driver=551.23
15:27:12-489872 ERROR Diffusers import error: version=0.16.1 error: cannot import name 'DPMSolverSDEScheduler' from
'diffusers' (C:\SDNEXT\venv\lib\site-packages\diffusers__init__.py)
15:27:12-605906 INFO Available VAEs: models\VAE items=0
15:27:12-616017 INFO Available models: models\Stable-diffusion items=1 time=0.01s
15:27:13-197243 ERROR codeformer: ModuleNotFoundError
┌───────────────────────────────────────── Traceback (most recent call last) ─────────────────────────────────────────┐
│ C:\SDNEXT\modules\codeformer_model.py:31 in setup_model │
│ │
│ 30 │ │ from torchvision.transforms.functional import normalize │
│ > 31 │ │ from modules.codeformer.codeformer_arch import CodeFormer │
│ 32 │ │ from basicsr.utils import img2tensor, tensor2img │
│ │
│ C:\SDNEXT\modules\codeformer\codeformer_arch.py:10 in
Feature description
Hello, i have a problem, i moved recently to another house and i formated my computer, i wanted to use the last version of SD.NEXT and continuing my professional works on this version : "[89ba8e3]" -> (commit/89ba8e3cf6f4e7697ffe0298b49bed7ea108b039)
But 1st : the new version added some weird samplers modification that i can't figure it out : (brownian noise & discard penultimate sigma) 2nd : Impossible to reproduce old seeds (and that's a real problem) 3rd : Generation speed seems to impact a lot since the update than the version mentioned above.
I tried to roll back version to that one : "[89ba8e3]" -> (commit/89ba8e3cf6f4e7697ffe0298b49bed7ea108b039)
But, i cannot install it cause a Git repo problem, cannot find it .... Is there a way to get that version ? cause my works depend on it, thanks :)
oh and lycoris models seems to not showing too even with the extension (unmaintained) activated.
Version Platform Description
Windows 10 21h2 Pro / Opera GX