On Windows, torch doesn't have the inductor package, thus the FakeTensorUpdater is not available. However, we want to use the FakeTensorUpdater after the lowering pass across all platforms. Thus, we may need to copy the Torch's FakeTensorUpdater over, until torch has resolved this.
On Windows, torch doesn't have the inductor package, thus the FakeTensorUpdater is not available. However, we want to use the FakeTensorUpdater after the lowering pass across all platforms. Thus, we may need to copy the Torch's FakeTensorUpdater over, until torch has resolved this.