Skip to content

Conversation

@bonneted
Copy link
Contributor

No description provided.

@lululxvi lululxvi requested a review from mitchelldaneker June 17, 2024 15:28
@bonneted
Copy link
Contributor Author

I've got the same final parameters as the ones in the paper :

Epoch E tp ti td k Rm a1 C1 C2 C4 C5 Ub U0 Um Rg alpha beta
1010000 0.20086897812616378 5.984280323717105 101.07973763650493 11.9777607693349 0.008332535739606049 208.50174547489897 6.593664003041278 301.0031941572703 34.69381362708656 79.07846369570885 25.952992080032526 71.5795274689672 3.99920640001627 89.31458070648316 179.61120094478127 7.538939453290313 1.778194872565763

@lululxvi
Copy link
Member

Is this ready to merge?

@bonneted
Copy link
Contributor Author

The code relies on this DeepXDE PR that isn't merged yet: lululxvi/deepxde#1775
I've just updated the code of DeepXDE PR, please let me know if you'd like any more changes.

I've tested the code again and it's working as intended.
So whenever the other PR is accepted, this one should be ready to merge.

@lululxvi
Copy link
Member

Is this ready to merge?

@bonneted
Copy link
Contributor Author

Yes, I just changed epochs to iterations.

I've run the code again with dde master branch, it works as expected. It took 20min on a RTX 4070 GPU with jax backend.
Here are the final values:

Epoch E tp ti td k Rm a1 C1 C2 C4 C5 Ub U0 Um Rg alpha beta
1010000 0.2212 5.882 140.0 11.72 0.008384 376.2 6.405 428.2 33.55 43.03 22.20 60.63 0.9840 77.95 155.9 13.50 2.273

Some values are worse than the original paper, but this is probably due to the stochasticity of the run.

@lululxvi lululxvi merged commit 2d2639a into lu-group:main Mar 20, 2025
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.

3 participants