sharp remnant
A new release candidate, which fixes bugs and improves the install experience, can be found here: https://github.com/invoke-ai/InvokeAI/releases/tag/2.2.4-rc2
Please feel free to check it out and let the team know of any bugs or other issues you encounter.
bleak pewter
I had to reinstall again today, used the 2.2.4 version. Works like a charm.
sharp remnant
Great to hear
low sapphire
After this "update" 2.2.4, when one wants to update to latest version by clicking update.bat, it tries to install everything again. Models, output folder locations etc. Is this supposed to work like this?
sharp remnant
That's correct. But it should pick up the existing invokeai directory and not download any of the existing models. Are you seeing a different behavior?
low sapphire
Yes, I see different behavior. My InvokeAI is installed to D:\invokeai. When I go to that folder and click update.bat I expect to see something similar to "git pull" to happen. I am not coder, so i don't know how to properly explain this, but when update.bat has fetched updates, it should exit from update.bat, but it does not do it. It continues to ask where is my image output folder, and instead of offering the folder (D:\IAI_outputs) I gave already when I made first install, it offers me new folder from my C-drive (see attachment). And next it asks again what weight's I would like to install. I have already installed the weight's I want in the first time I made install. Why this update.bat works like this?
sharp remnant
So sorry. I see what’s happening. I think the best thing is to run the new installer, and when you are asked to locate the Invokeai directory please give it the path to the directory that contains your existing directory. That way it’ll recognize the weights you already downloaded
tender ether
Hey, FYI it seems like the update.sh file doesn't add the launch scripts to the invoke-ai folder in the home directory
Just updated to 2.2.4 and it didn't add it there like the instructions now say is supposed to happen
tender ether
Uh oh... I tried running the installer again, and now I'm getting this error? I definitely have pip installed
sharp remnant
I may have fixed the issue with pip. See here: https://github.com/invoke-ai/InvokeAI/files/10212350/InvokeAI-installer-2.2.4-p1-mac.zip or here: https://github.com/invoke-ai/InvokeAI/files/10212351/InvokeAI-installer-2.2.4-p1-linux.zip
sharp remnant
Hey, FYI it seems like the update.sh file doesn't add the launch scripts to the ...
Joshua, is this when you try to use the update.sh script that was part of the 2.2.3 release? I'm afraid that script was not fully functional and the recommendation is to run one of the 2.2.4 installers. When the installer asks for the location of the invokeai directory, point it at the folder that the current invokeai lives in so that you don't get bothered about downloading model weights again.
tender ether
Joshua, is this when you try to use the `update.sh` script that was part of the ...
Yeah, that was what I first tried to do last night before I moved on to using the new 2.2.4 installer you helped me work with last night
Now everything seems to be working as expected. 🙂
sharp remnant
I've just updated the release notes to warn people about this.
unkempt vessel
Never had an app so deterimated not to be installed! just kidding developers own
echo vine
It's crashing when I start the web server
>> Point your browser at http://127.0.0.1:9090
./invoke.sh: line 24: 61366 Segmentation fault: 11 .venv/bin/python .venv/bin/invoke.py --web $*
twin eagle
invoke.sh isn't in the home dir after pulling the latest changes and running install.sh
exporting the INVOKEAI_ROOT to the previous environment worked for having the script not downloading all the models again but there is still no launcher script
twin eagle
alright, just pulling from main doesn't get me the latest install.sh it seems, I will later try to do what the release notes actually say X-]
twin eagle
Can you avoid the installer to redownload pytorch ?
sharp remnant
Those who are getting a segfault error and are on the Macintosh platform, the cause of this error has been identified and should be corrected in this version of the installer: https://github.com/invoke-ai/InvokeAI/files/10221812/InvokeAI-installer-2.2.4-p2-mac.zip Please try it and report your results.
sharp remnant
Can you avoid the installer to redownload pytorch ?
All the launcher script does is to activate the virtual environment in the runtime directory and then run the invoke.py python script. You can find the launcher script hidden under installer/templates/install.sh.in. Just copy it somewhere and remove the .in part. However be aware that it assumes that both .venv and the launcher script itself are located inside the invokeai directory, next to models, etc. If you've been running invokeai from the git source repository, then this is not necessarily the case, which is why we only create install.sh when InvokeAI has been installed using the install.sh script.