search

chaiNNer-org / spandrel

Spandrel gives your project support for various PyTorch architectures meant for AI Super-Resolution, restoration, and inpainting. Based on the model support implemented in chaiNNer.
MIT License
103 stars 7 forks source link

Add support for ATD #200

Closed RunDevelopment closed 3 months ago

RunDevelopment commented 3 months ago

This adds support for ATD.

RunDevelopment commented 3 months ago

This is so weird. Why do I get vastly different outputs in CI?

joeyballentine commented 3 months ago

CPU vs GPU?

RunDevelopment commented 3 months ago

I use CPU for tests locally as well.

joeyballentine commented 3 months ago

maybe its linux thats the problem? i have no idea what else it could be

RunDevelopment commented 3 months ago

Okay, so I just did test runs on CPU vs GPU. It seems like ATD is highly sensitive to floating point impression. When I compare the output of my CPU vs GPU, I get the same type of large error as we see with CI. Despite the large error, the two outputs are visually the same, so I think I'm going to add another error metric to account for this.

RunDevelopment commented 3 months ago

Finally.