Jane (Yuan) Xu

Add explicit APIs in BenchmarkModel

4

This is part 1 of adding ways to more granularly benchmark. Also explicitly document what methods we expect BenchmarkModels to have.

We should update HUD wiki

1

https://github.com/pytorch/pytorch/wiki/Using-hud.pytorch.org is outdated. We should update this and have https://github.com/pytorch/pytorch/wiki/PyTorch-CI-Metrics-Dashboards:-the-HUD point to this wiki This would be better engineering + improving devX

Done: added what's needed to test out GaLore Next step: add SVD strategy

Allow snapshot.pickle to be saved for all recipes with profiler support

1

#### Context What is the purpose of this PR? Is it to - [x] add a new feature - [ ] fix a bug - [ ] update tests and/or...

[pytorchbot] retest command

5

Several peeps have asked for a `@pytorchbot retest` command that would retrigger tests on a commit. Even cooler could be a @pytorchbot retest -f that would rerun only failed jobs.

[WIP] Add tests for lora finetune single device act offload

1

Need to rebase this afer #1443