shard crossref 1->2
Fixes #ISSUE_NUMBER
Regarding sharding crossref:
- Pros:
- Does not result in a lot of additional time
- Balances fairly well across shards
- On linux so it's cheap
- It is on the longer side at 2h
- Cons:
- Not super important b/c its not the longest running job
spreadsheet regarding sharding: https://docs.google.com/spreadsheets/d/1BdtVsjRr0Is9LXMNilR02FEdPXNq7zEWl8AmR3ArsLQ/edit#gid=1153012347
Pull Request resolved: https://github.com/pytorch/pytorch/pull/76450
Approved by: https://github.com/janeyx99, https://github.com/seemethere