[docs] DeepSpeed (#28542)

* config

* optim

* pre deploy

* deploy

* save weights, memory, troubleshoot, non-Trainer

* done
This commit is contained in:
Steven Liu
2024-01-24 08:31:28 -08:00
committed by GitHub
parent bb6aa8bc5f
commit 738ec75c90
5 changed files with 1312 additions and 2286 deletions

View File

@@ -1312,3 +1312,19 @@ You can vote for this feature and see where it is at these CI-specific threads:
- [Github Actions:](https://github.com/actions/toolkit/issues/399)
- [CircleCI:](https://ideas.circleci.com/ideas/CCI-I-344)
## DeepSpeed integration
For a PR that involves the DeepSpeed integration, keep in mind our CircleCI PR CI setup doesn't have GPUs. Tests requiring GPUs are run on a different CI nightly. This means if you get a passing CI report in your PR, it doesnt mean the DeepSpeed tests pass.
To run DeepSpeed tests:
```bash
RUN_SLOW=1 pytest tests/deepspeed/test_deepspeed.py
```
Any changes to the modeling or PyTorch examples code requires running the model zoo tests as well.
```bash
RUN_SLOW=1 pytest tests/deepspeed
```