non-interactive default model selection

[abstract ferry]

The tests for the https://github.com/invoke-ai/InvokeAI/pull/1384 diffusers branch want to set the default model to a diffusers-compatible version.

Earlier, the tests worked by copying models.yaml.example to models.yaml, and I added a bit to configure_invokeai so it would download the model in the configure file.

but now, though test-invoke-pip still has that create models.yaml from example step in it, the code I added to configure isn't finding a config file during its stage of the test.

Any pointers for how to get back on track here?

oh @old current has been working on very related things in https://github.com/invoke-ai/InvokeAI/pull/1578

[ancient ruin]

yeah this is all very in flux, i think the best thing for now is to wait for the work that ebr and mauwii are doing to settle into its final form, and only be concerned about making CI/CD pass when it comes time to actually merge

[abstract ferry]

does not compute

[abstract ferry]

ah, the value I was reading for the config file didn't have the root dir on it.

Now it's working.

(though I think we'll be able to clean it up a lot after 1578)

[ancient ruin]

yeah deffo

[old current]

ah, sorry, I did not realize you were also tackling the same issues @abstract ferry . I hope to be able to finish this tonight; the only thing that remains is to move the config file from ~/.invokeai to the runtime dir without breaking backwards compatibility. does this conflict with your work on diffusers?