Skip to content

Add OpenVino/XPU(Intel) acceleration?  #6

@Universal-Invariant

Description

@Universal-Invariant

I was able to hack the code to get it to use XPU/open vino.

It wasn't that difficult but required a lot of trial and error, specially since I don't know much about using these NN models.

Most of the changes involved messing with inswapper.py and refacer.py:

NeoRefacer.zip

While it now runs on the intel GPU it is still relatively slow. To get it to work I had to use a lock because I didn't feel like trying to figure out how to use the Async code.

Also, I think the BasicSR still uses the cpu(I switched it over the XPU and it partially says so but it

BasicSR Archs: Using XPU backend.
BasicSR: Using CPU backend.
Using Intel GPU (XPU) for acceleration.
Available providers: ['OpenVINOExecutionProvider', 'CPUExecutionProvider']
Using providers: ['OpenVINOExecutionProvider', 'CPUExecutionProvider']
Active provider: OpenVINOExecutionProvider
Mode: RefacerMode.COREML
Face Detector providers: ['OpenVINOExecutionProvider', 'CPUExecutionProvider']
Face Recognizer providers: ['OpenVINOExecutionProvider', 'CPUExecutionProvider']

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions