Skip to content

[doc] add potential solution for OOM in llama2 example #4699

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
FrankLeeeee merged 1 commit into from
Sep 13, 2023
Merged

[doc] add potential solution for OOM in llama2 example #4699

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 3 additions & 0 deletions examples/language/llama2/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -149,6 +149,9 @@ Finally, run the following command to start training:
```bash
bash gemini.sh
```

If you encounter out-of-memory(OOM) error during training with script `gemini.sh`, changing to script `gemini_auto.sh` might be a solution, since gemini_auto will set a upper limit on GPU memory usage through offloading part of the model parameters and optimizer states back to CPU memory. But there's a trade-off: `gemini_auto.sh` will be a bit slower, since more data are transmitted between CPU and GPU.

#### c. Results
If you run the above command successfully, you will get the following results:
`max memory usage: 55491.10 MB, throughput: 24.26 samples/s, TFLOPS/GPU: 167.43`.
Expand Down