[bug]:

[jintoku]

Is there an existing issue for this problem?

• [X] I have searched the existing issues

Operating system

Windows

GPU vendor

Nvidia (CUDA)

GPU model

RTX 3070

GPU VRAM

No response

Version number

4.2.5

Browser

Chrome

Python dependencies

No response

What happened

Trying to convert safetensors to diffusion, or generating using the checkpoint gives an error; on the latter "'C:\Users\Me\invokeai\configs\stable-diffusion\stable-diffusion\v1-inference.yaml'" I've checked and the file is there.

What you expected to happen

I expected conversion or generation

How to reproduce the problem

Install a checkpoint from the Models Manager starter set, then try to convert to diffusers; or try to generate an image using the checkpoint.

Additional context

Happens every time.

Discord username

No response

[jintoku]

It looks like a duplication of folder names in the instruction.

[lstein]

Do you happen to have convert_cache_dir set in your invokeai.yaml file?

Here's a possible fix for your specific problem:

1. Open the model manager tab
2. Select the model that is giving you issues
3. Select the yellow Edit button
4. Find the field that says "Path to Config"
5. It should read stable-diffusion\v1-inference.yaml. If it has an extra stable-diffusion in it (which I suspect it will), remove the extra directory.
6. Press "Save"

If on step 4 you get empty fields (which is a bug I've seen pop up recently), reload the page and select the model again.

[jintoku]

Thanks for the help. I've not changed my invokeai.yaml since version 2.x so I don't think that's it. I removed the "stable-diffusion" from the Model Edit field, leaving just "v1-inference.yaml" by itself, and that now allows me to generate an image using the checkpoint version of CyberRealistic, but I still get an error when attempting conversion to diffusion. The error message reads "Model Conversion Failed: CyberRealistic_V4.1_FP16 (2)". Thanks again for your help!

[jintoku]

I finally had the chance to inspect my invokai.yaml file and find it doesn't have the setting for the cache directory. I wasn't clear if I should add that config or not, as the docs page says it is a deprecated command. I did find that the conversion effort created some odd folders in my inkokeai\models\cache folder with names that were basically the full directory string without the punctuation, and in each folder there seemed to be a set of diffuser folders. I'll try setting the cache paramater and see what happens.