Skip to content

Conversation

@zoemcc
Copy link
Contributor

@zoemcc zoemcc commented Apr 13, 2021

Basic structure is here for adaptive losses.

TODO:

  • Finish performance debugging (highest priority), currently NonAdaptiveLoss performs better than the adaptive loss even on very simple cases
  • Add more strategies for reweighting (i.e. from Automatic weighting between equations #155 )
  • Split up PDE and BC loss terms and add one loss weight for each term
  • Integrate the tests into the main test script/expand test cases
  • Thorough performance benchmarking
  • Make sure GPU support isn't hindered
  • Clean up the manual instrumentation, add documentation & example script in the docs

zoemcc added 2 commits April 10, 2021 18:19
… over satisfies the BC loss compared to the PDE loss in the test case, this may be because it is a point BC constraint.
…ed efficient NonAdaptiveLoss and made it default.
@zoemcc
Copy link
Contributor Author

zoemcc commented Jan 20, 2022

Using this PR with a different branch instead: #482

@zoemcc zoemcc closed this Jan 20, 2022
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.

1 participant