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.
Using the call API now works. This also meant that I was able to add size requirement tests, which found that the size requirements were incorrect.
Another interesting thing is that I changed the scale from 8x to 1x. These models are 1x, they are just trained on 8x (bicubic or whatever) upscaled data.
Note that I only removed a few print calls in the arch code, I didn't touch the forward method, so chainner's FaceUpscale will still work.
Using the call API now works. This also meant that I was able to add size requirement tests, which found that the size requirements were incorrect.
Another interesting thing is that I changed the scale from 8x to 1x. These models are 1x, they are just trained on 8x (bicubic or whatever) upscaled data.
Note that I only removed a few
printcalls in the arch code, I didn't touch theforwardmethod, so chainner's FaceUpscale will still work.