[bug]: /config/models.yaml not being created in right folder

[mickr777]

Is there an existing issue for this?

• [X] I have searched the existing issues

OS

Linux

GPU

cuda

VRAM

12GB

What happened?

During fresh manual install and running - invokeai-configure --root="/userfiles/" --yes and after diffusers are download the configs/models.yaml file is not created

Screenshots

No response

Additional context

No response

Contact Details

No response

[lstein]

Is this with RC6?

[lstein]

I have been unable to reproduce this on a Linux/CUDA system. Did the script run to completion or was it interrupted during the downloads? The models file doesn't get created until after all models are downloaded correctly. (This is probably a misfeature, actually.)

[mickr777]

Is this with RC6?

Latest git clone from main

I have been unable to reproduce this on a Linux/CUDA system. Did the script run to completion or was it interrupted during the downloads? The models file doesn't get created until after all models are downloaded correctly. (This is probably a misfeature, actually.)

Yes it is completing the diffuser download and the other models and I am getting the ** Model Installation Successful **

You're all set!

---

[mickr777]

@lstein ok just noticed they are getting created but under a folder /configs/ at root level, not under my --root="/userfiles/"

[ebr]

@mickr777 :

• when you ran the installation, did you also select /userfiles/ as the destination directory?
• what is the location of your virtual environment directory? I see a /venv in your screenshot - is there another one in /userfiles? how about the models - did they end up in /userfiles/models?
• could you please see if this still happens if you export INVOKEAI_ROOT=/userfiles before running invokeai-configure

@lstein i suspect this to be another case of #2064

[mickr777]

• when you ran the installation, did you also select /userfiles/ as the destination directory?

I used invokeai-configure --root="/userfiles/" --yes to set the root folder and skip prompts

• what is the location of your virtual environment directory? I see a /venv in your screenshot - is there another one in /userfiles? how about the models - did they end up in /userfiles/models?

I setup the venv outside the userfiles folder so there is no venv folder in /userfiles and yes all the models are in the /userfiles/models/

• could you please see if this still happens if you export INVOKEAI_ROOT=/userfiles before running invokeai-configure

using export INVOKEAI_ROOT=/userfiles worked, the config files where created in the right place

[ebr]

Thanks for confirming. The --root argument should be taking precedence in this case, but we are only considering the env vars instead. Definitely should be fixed.

In the meantime, are you able to add the INVOKEAI_ROOT variable to your .bashrc file as a workaround?

[lstein]

Thanks for confirming. The --root argument should be taking precedence in this case, but we are only considering the env vars instead. Definitely should be fixed.

In the meantime, are you able to add the INVOKEAI_ROOT variable to your .bashrc file as a workaround?

Do you think you can commit a quick fix tonight? I am done in from writing documentation.

[mickr777]

More info if it helps: When not using export INVOKEAI_ROOT=/userfiles only invokeai-configure has issues with the config folder location when using --root argument.

invoke.py follows the --root argument correctly and brings up error it cant find the models.yaml in the correct folder eg /userfile/config/,

if i manually copy from /config/ to /userfiles/config/ invoke.py loads fine

So i looks like it is only the creation location of the config folder is a issue

[ebr]

sorry for the delay on this, it fell off my radar. I'll try to get it sorted this week

[lstein]

@ebr we may have a related issue in #2779 .

[lstein]

@mickr777 Following up on this old issue, could you give the install process a try using the 2.3.1 release candidate? The instructions in manual install should still work, but you need to provide a specific version to PIP, viz:

pip install InvokeAI[xformers]==2.3.1-rc4 --use-pep517 --extra-index-url https://download.pytorch.org/whl/cu117

Thanks in advance! -Lincoln

[mickr777]

For me it appears to place it in the right folder, but I used get clone main

@mickr777 Following up on this old issue, could you give the install process a try using the 2.3.1 release candidate? The instructions in manual install should still work, but you need to provide a specific version to PIP, viz:

pip install InvokeAI[xformers]==2.3.1-rc4 --use-pep517 --extra-index-url https://download.pytorch.org/whl/cu117

Thanks in advance! -Lincoln

@lstein

I did fresh manual install using my unraid docker image that uses git clone main and the config folder was created in the --root folder I set now without the need to use "export INVOKEAI_ROOT="

I can try to install it via the pip install 2.3.1-rc4 if you would like

[lstein]

I think all should be good now. Thanks for your help.

Lincoln

On Thu, Feb 23, 2023 at 7:32 PM mickr777 @.***> wrote:

For me it appears to place it in the right folder, but I used get clone main

@mickr777 https://github.com/mickr777 Following up on this old issue, could you give the install process a try using the 2.3.1 release candidate? The instructions in manual install should still work, but you need to provide a specific version to PIP, viz:

pip install InvokeAI[xformers]==2.3.1-rc4 --use-pep517 --extra-index-url https://download.pytorch.org/whl/cu117

Thanks in advance! -Lincoln

@lstein https://github.com/lstein

I did fresh manual install using my unraid docker image that uses git clone main and the config folder was created in the --root folder I set

I can try to install it via the pip install 2.3.1-rc4 if you would like

— Reply to this email directly, view it on GitHub https://github.com/invoke-ai/InvokeAI/issues/2577#issuecomment-1442620719, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAA3EVMYUZ5S5A56NRBD4DTWY76QJANCNFSM6AAAAAAUUVOXSU . You are receiving this because you were mentioned.Message ID: @.***>