quantized activations: clean up more unneeded quantizations (#36981)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/36981
Replaces unneeded quantize calls for remaining quantized
activations with empty tensor creation.
Should be a perf win for anyone who uses these.
Test Plan:
python test/quantization/test_quantized.py TestQuantizedOps
Imported from OSS
Differential Revision: D21185969
fbshipit-source-id: 473b2b8aa40046ea3f0665bd45b03f09e8a7d572
