pytorch

cce32c6f - functionalization: handle models that resize their program inputs (#83542)

functionalization: handle models that resize their program inputs (#83542) Context: When turning on functionalization + fake tensor mode and running `resnet50_quantized_qat` from torchbench, the model fails (and should be fixed by this PR). Before landing this PR, we need ProxyTensors (soon to be just fake tensors after https://github.com/pytorch/pytorch/pull/83330) to be resizable Pull Request resolved: https://github.com/pytorch/pytorch/pull/83542 Approved by: https://github.com/ezyang