shard `linux-xenial-cuda11.3-py3.7-gcc7 / test (default` from 2 to 4 (#77083)
Notes
- high overhead
- even ish?
- but its long (~2.5 hr tts on master)
sharding spreadsheet: https://docs.google.com/spreadsheets/d/1BdtVsjRr0Is9LXMNilR02FEdPXNq7zEWl8AmR3ArsLQ/edit#gid=1242752407
Pull Request resolved: https://github.com/pytorch/pytorch/pull/77083
Approved by: https://github.com/suo
