Fix float related autoquant options (#1562)

* Fix float related autoquant options

Summary:
Forgot to add a test for previous changes, this fixed some
implementations for the quantized model

Test Plan:
python test/integration/test_integration.py -k test_autoquant_float

Reviewers:

Subscribers:

Tasks:

Tags:

* skip non-cuda runs

* update torch version requirement

* typo