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.
Turns out, calling .eval isn't a noop. It takes a very small amount of time to do what it does (under 1ms), but it still does it every time. We can save this op by just checking if the model is in training mode first. If it is, then we can .eval it. If it isn't, we can save this.
Turns out, calling .eval isn't a noop. It takes a very small amount of time to do what it does (under 1ms), but it still does it every time. We can save this op by just checking if the model is in training mode first. If it is, then we can .eval it. If it isn't, we can save this.