Skip to content
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.

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

Curriculum Learning Support #695

Merged
merged 48 commits into from
Mar 9, 2023
Merged

Curriculum Learning Support #695

merged 48 commits into from
Mar 9, 2023

Conversation

Quentin-Anthony
Copy link
Member

@Quentin-Anthony Quentin-Anthony commented Sep 25, 2022

Adds support for DeepSpeed's curriculum learning: https://www.deepspeed.ai/tutorials/curriculum-learning/, based on the code in https://github.com/microsoft/Megatron-DeepSpeed

This PR requires latest DeepSpeed, and is therefore based on deepspeed_main from #663

Experimental results are tracked at: https://wandb.ai/eleutherai/curriculum?workspace=user-

  • Add non-PP support
  • Test/verify with various model sizes and curriculum schedules (in-progress)
  • Add PP support

Dashiell Stander and others added 21 commits November 11, 2022 19:58
Signed-off-by: Dashiell Stander <[email protected]>
Signed-off-by: Dashiell Stander <[email protected]>
dashstander and others added 19 commits November 14, 2022 19:15
Signed-off-by: Dashiell Stander <[email protected]>
Signed-off-by: Dashiell Stander <[email protected]>
Signed-off-by: Dashiell Stander <[email protected]>
Signed-off-by: Dashiell Stander <[email protected]>
@dashstander dashstander marked this pull request as ready for review November 16, 2022 00:12
Copy link
Member Author

@Quentin-Anthony Quentin-Anthony left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM. This PR depends on two other PRs that should be merged first:

  • Latest DS (deepspeed_main) should be merged into main
  • The slurm runner updates should be separated into a new PR

I'll merge deepspeed_main into main by the end of this week. @dashstander -- Do these slurm changes need their own PR or are they specific to your setup?

@StellaAthena StellaAthena added this to the Release V2 milestone Dec 20, 2022
@Quentin-Anthony Quentin-Anthony enabled auto-merge (squash) March 9, 2023 17:17
@Quentin-Anthony Quentin-Anthony merged commit 68d223c into main Mar 9, 2023
@Quentin-Anthony Quentin-Anthony deleted the curriculum_learning branch March 9, 2023 17:18
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

4 participants