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

Extend with action points #10

Open
rvandernoort opened this issue Apr 14, 2023 · 0 comments
Open

Extend with action points #10

rvandernoort opened this issue Apr 14, 2023 · 0 comments
Labels
enhancement New feature or request good first issue Good for newcomers help wanted Extra attention is needed

Comments

@rvandernoort
Copy link
Contributor

Summary

Currently, the implementation only reports the energy consumption (EC) measurements visually and leaves it up to the ML developer to act accordingly. We have a few ideas for which GATorch can be extended with certain suggestion/action points. These action points can be reported to the user with a direct suggestion to for example define an early stopping criteria based on the energy consumption per unit of loss. This should provide even more awareness about the energy consumption of ML and possibly reduce wasted energy.

Roadmap

  • Investigate awareness level with the current implementation
  • Investigate possible action points and their criteria
  • Implement action points seamlessly

Possible action points

  • Suggestions for optimizers based on EC.
  • Early stopping criteria based on EC per unit of loss.
  • Model architecture adjustments based on individual layer EC.
  • Hyperparameters adjustment based on EC
  • etc.
@rvandernoort rvandernoort added enhancement New feature or request good first issue Good for newcomers help wanted Extra attention is needed labels Apr 14, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
enhancement New feature or request good first issue Good for newcomers help wanted Extra attention is needed
Projects
None yet
Development

No branches or pull requests

1 participant