Update on "make torch.amp.autocast more generic" #76299

Job	Run time
pytorch-linux-focal-py3-clang9-android-ndk-r21e-build / filter	14s
linux-focal-cuda12.1-py3.10-gcc9 / build	49m 49s
before-test / llm-retrieval	5m 48s
linux-focal-cuda12.1-py3.10-gcc9-no-ops / build	50m 13s
win-vs2019-cpu-py3 / build	42m 41s
macos-12-py3-arm64 / build	11m 38s
linux-focal-rocm6.1-py3.8 / build	1h 47m 38s
win-vs2019-cuda11.8-py3 / build	43m 56s
libtorch-linux-focal-cuda12.1-py3.7-gcc9-debug / build	33m 7s
caffe2-linux-jammy-py3.8-gcc11 / build	40m 22s
pytorch-linux-focal-py3-clang9-android-ndk-r21e-build / build (default, 1, 1, linux.2xlarge)	1h 18m 51s
before-test / target-determination	1m 47s
macos-12-py3-arm64-mps / filter	2m 7s
macos-12-py3-arm64 / test (default, 1, 3, macos-m1-stable)	1h 8m 12s
macos-12-py3-arm64 / test (default, 2, 3, macos-m1-stable)	1h 21m 46s
macos-12-py3-arm64 / test (default, 3, 3, macos-m1-stable)	1h 3m 51s
macos-12-py3-arm64-mps / test (mps, 1, 1, macos-m1-stable)	17m 53s
win-vs2019-cpu-py3 / test (default, 1, 3, windows.4xlarge.nonephemeral)	2h 6m 11s
win-vs2019-cpu-py3 / test (default, 2, 3, windows.4xlarge.nonephemeral)	1h 50m 33s
win-vs2019-cpu-py3 / test (default, 3, 3, windows.4xlarge.nonephemeral)	1h 38m 54s
linux-focal-cuda12.1-py3.10-gcc9 / test (nogpu_AVX512, 1, 1, linux.2xlarge)	2h 7m 39s
linux-focal-cuda12.1-py3.10-gcc9 / test (nogpu_NO_AVX2, 1, 1, linux.2xlarge)	1h 56m 28s
linux-focal-cuda12.1-py3.10-gcc9 / test (jit_legacy, 1, 1, linux.4xlarge.nvidia.gpu)	14m 52s
linux-focal-rocm6.1-py3.8 / test (default, 1, 1, linux.rocm.gpu)	1h 24m 3s
	22h 58m 33s

Provide feedback