mickr777

found the issue (I think) grab the updated start.sh file info from main post and delete container (but remove delete image tick box) recreate the container, as the main post or used the Simplified Install for main, as I pushed an update to it (use either way but make sure to delete the appdata/invokeai/venv/pyvenv.cfg and if you have it appdata/invokeai/userfiles/invokeai.init files)

looks like they changed the way the invokeai-configure script works I will have to do some testing

there shouldn't be any config folder in the invokeai folder but you can add the folder and models.yaml there and see if that helps

models.yaml can you do a test for me copy the attached models.yaml into your appdata/invokeai/userfiles/configs/ folder and see if the same error happens

added the Main docker for anyone wanting to migrate from prenodes to v3.0 beta, (this docker will be the default one for updates, i did it this way to not force everyone on prenodes to do the update) if you wish to do this migration, go to main post for instructions DO NOT do this if you dont want to do the migration to v3.0 beta yet

just a warning to others, going to 6.12.1 half of my dockers disappeared and had to be reinstalled, now I had to reinstalled all my dockers as they all disappeared after this update to 6.12.2, lucky the appdata and xml are all there.

Ok thats why, Yes all the legacy web server has been removed from main branch on get hub, however the nodes web server isnt fully ready for public use (still lots of changes happening), currently the front end needs to be yarn bulit on each start, runs on port 5173, the backend needs to be run separately and the whole user configuration file has changed. So You need to git clone -b prenodes, if you still want a working install, (or use the main dockerhub image in first post) I have docker image that loads nodes ui, but this will all change before the public release of invokeai v3.0 I can post when im back in the office what i had to do to get nodes ui to currently work if you wanted to see

Something has got messed up somewhere, you might need to delete all files but the userfiles folder, make sure your using the main image option in the first post and the my-invokeai.xml file is this updated on your flash drive from the first post and let it rebuild. This image is locked to the prenodes version of main, once node migtaion has been finalised and working, i will release an update. To that image

What image are you on the main or v2.3, also what errors are showing for you? Currently while the migration to nodes i locked the main image to the prenodes tag, so it shouldnt be getting any updates at the moment till the migration has finished, so anything they do should not effect it. V2.3 did get some updates for xformers, you may need to delete the file /invokeai/venv/pyvenv.cfg and rerun the docker to fix the issue. As for the current main branch on git hub I have a working image for it but lots still dont work, the ui needs to be used in dev mode and things change so much with every update, I didnt think anyone on main wants that yet.

yes this is the breaking issue, just uploaded a quick fix to docker hub, locked the current main branch to the pre-nodes tag till main migration to nodes is done. however if there is files under /invokeai/invokeai/ they will need to be deleted before starting the docker so it can pull the pre-nodes tag. Side note ignore these errors for now since pre-nodes is a tag not a branch: fatal: HEAD does not point to a branch fatal: HEAD does not point to a branch start.sh: line 38: [: fdad62e88b77478ef7016e447aeed6a221149f6b: unary operator expected start.sh: line 40: [: fdad62e88b77478ef7016e447aeed6a221149f6b: unary operator expected

If you can dont restart your docker, lastest updates will break it. Looking into a fix

I cant seem to replicate this error, is this the main or v2.3 docker? some things to try is delete the files in the invokeai/invokeai folder and let it repull from the repo and update the venv, if that doesn't work try this, in invokeai/venv folder delete the pyvenv.cfg file and rerun the docker

Thank you very much, but no need to tip me, the guys at invokeai are the ones doing all the hard work making there app, i just spent the time working out how to put it in a docker for unraid. I am happy it is working now. Just an fyi there is a major migration to nodes right now in the background of invokeai, so some updates might break the docker, so far it hasnt effected it, but I am keeping an eye on it

Sounds like the ptyhon venv may not have been setup correctly during the first run, if you have Dynamix File Manager plugin installed, in unraid go to it then go to the folder cache, appdata, invokeai then venv and delete the pyvenv.cfg file and rerun the docker. The file Could also be deleted via the unriad terminal if you dont have that plugin

Trying building it without the --gpu 0 (also probably best to remove your hugging face token from the post above, for your security)

Pretty sure at the moment invokai is single threaded, so will only utilise one gpu. This should be changing in the upcoming migration to nodes

You can add new models in the model manager, move them in to a folder under userfiles, I added mine in to a folder called mymodels in the models folder, next you will need to click "add new" under model manager, select the type of model it is and then type in the folder they are in eg. /home/invokeuser/userfiles/models/mymodels/ then click find models Then it should show any models it found, to add them. as for loras its only just been added into the v2.3 and this docker is based on main branch as it will be based on nodes and loras hasnt been added to nodes yet. I have a image on docker hub that i use for the v2.3 branch, but main branch has had a massive file layout change, so this image is will be only compatible with v2.3.* only, so you would need to make it a sparate docker container for it, as it wont get updates once v3.0 is out

The version will stay the same till the final version of 3.0.0 is released. Since the docker is pulling from main branch it gets all the updates, think of it as alpha testing new releases

What errors are you getting? as Mine updated fine to the latest changes after I restarted it

It appears to be a error in the invokeai code I have reported the bug https://github.com/invoke-ai/InvokeAI/issues/2872 Looks like there is a pr https://github.com/invoke-ai/InvokeAI/pull/2876 Once the pr is merged just start your invokeai docker and it should retrieve the change edit: Pr Merged, all should be working again

If you get this and the docker stops >> ALERT: >> The location of your previously-installed diffusers models needs to move from >> invokeai/models/diffusers to invokeai/models/hub due to a change introduced by >> diffusers version 0.14. InvokeAI will now move all models from the "diffusers" directory >> into "hub" and then remove the diffusers directory. This is a quick, safe, one-time >> operation. However if you have customized either of these directories and need to >> make adjustments, please press ctrl-C now to abort and relaunch InvokeAI when you are ready. >> Otherwise press <enter> to continue. you will need to manually move all files in /userfiles/models/diffusers/ in to /userfiles/models/hub/ and delete the diffusers folder, now it should load. once in the webui, use the model manger to change any models that was pointing to models/diffusers/ to now use models/hub/

Make sure you have rebuild the container with the new chages in the main post, and Trying delete everything in the invokeai/invokeai (leave userfiles and venv alone) folder and allow it git clone that folder again There is another issue that I am looking into, on start its asking for user input to move the model from diffusers folder to hub, but this is stopping the container starting.

Due to a major source tree restructure on the invokeai git the main start.sh script had to be changed, if you did a manual install the changes are in the main post under start.sh code, if you are using the docker repository install I have pushed the changes already you should get a update notice.

Yes I ran into this the other day, the config folder is being made in the root folder in the docker and not in the userfiles folder, (ebr from invokeai is looking into it) temp fix was to add export INVOKEAI_ROOT=/userfiles to the start.sh before "invokeai-configure --root="/userfiles/" --yes" line I have fixed the start.sh script in main post, if you delete the /venv/pyvenv.cfg file and rebuild the docker container with the new start.sh change, it should work and rebuild the enviroment

I'm very sorry I just realised I had changed the guide by mistake to a test build of Dockerfile, start.sh and my-invoke.xml I was playing with I have fixed this, please retry following the guide now