[bug]: fails to load overridden VAE when path is relative

[keturn]

Is there an existing issue for this?

• I have searched the existing issues

OS

Linux

GPU

cuda

VRAM

12GB

What version did you experience this issue on?

4599575

What happened?

sdxl/main/SDXL base 1_0:
  path: sdxl/main/SDXL base 1_0
  description: SDXL base v1.0
  vae: sdxl/vae/sdxl-vae-fp16-fix/
  variant: normal
  format: diffusers

Note the presence of a vae override and the use of a relative path there.

Traceback: ModelNotFoundException

[2023-08-01 16:43:47,222]::[InvokeAI]::ERROR --> Traceback (most recent call last):
  File "src/InvokeAI/invokeai/app/services/processor.py", line 86, in __process
    outputs = invocation.invoke(
              ^^^^^^^^^^^^^^^^^^
  File "lib/python3.11/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
    return func(*args, **kwargs)   
           ^^^^^^^^^^^^^^^^^^^^^
  File "src/InvokeAI/invokeai/app/invocations/latent.py", line 515, in invoke
    vae_info = context.services.model_manager.get_model(     
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "src/InvokeAI/invokeai/app/services/model_manager_service.py", line 364, in get_model
    model_info = self.mgr.get_model(
                 ^^^^^^^^^^^^^^^^^^^
  File "src/InvokeAI/invokeai/backend/model_management/model_manager.py", line 484, in get_model
    model_path = model_class.convert_if_required(
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "src/InvokeAI/invokeai/backend/model_management/models/vae.py", line 103, in convert_if_required
    if cls.detect_format(model_path) == VaeModelFormat.Checkpoint:
       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "src/InvokeAI/invokeai/backend/model_management/models/vae.py", line 83, in detect_format
    raise ModelNotFoundException()                                                              
invokeai.backend.model_management.models.base.ModelNotFoundException        
                                                                                                
[2023-08-01 16:43:47,226]::[InvokeAI]::ERROR --> Error while invoking:     

[lstein]

This is fixed in PR #4099 (despite PR name), as well as in your own #4059