Adaptive BDF/NDF in Leap #17
Conversation
|
To cut down on noise in my email, I'm unsubscribing from this PR. When it next needs my attention, please @-mention me or hit the "request review" button. Otherwise, I may not see your messages in a timely manner. |
the CI (should that occur)
|
@inducer with the addition of fixed-order convergence tests (I managed to get the "bootstrapping" to work with some finagling), this is now ready to be discussed. Note also the dependence on the forked |
test/test_bdf.py
Outdated
| bs_targ=0.05, show_dag=show_dag, plot=plot, | ||
| implicit=True, solver=newton_solver, | ||
| solver_hook=solver_hook) | ||
| # Use "DEBUG" to trace execution |
There was a problem hiding this comment.
How come so much of this is commented out?
There was a problem hiding this comment.
This was something I did when unifying this convergence test within a utility, and was an erroneous artifact. No longer present as of 6606465.
| pytest.importorskip("scipy") | ||
| pytest.importorskip("cantera") | ||
| pytest.importorskip("pyrometheus") | ||
| pytest.importorskip("jax") |
There was a problem hiding this comment.
We should make it so that this actually runs in CI.
There was a problem hiding this comment.
Yeah, I can take care of this. I know I had to rip out a Scipy importorskip in a previous PR, so I can look there to remember how I did it.
There was a problem hiding this comment.
I'll help. You'll need to set up a Conda-based CI, which I think I can probably do more quickly.
|
Thanks! I'm converting this back to draft pending the two items I added to the description, to avoid me merging this by accident before it's ready. |
|
@cmikida2 Not sure what happened with the CI there. You can use https://github.com/marketplace/actions/debugging-with-tmate to SSH in and see if you figure out what crashed the CI. |
|
@inducer I've just looked this over again and brought it up to speed with the main I'll also note that the previous failing CI issue, in which (by looking at previous workflows) the reactor test was failing deep in the guts of Jax somewhere, appears to have been dealt with via ebfe487. |
2 participants
Aims to implement the order and timestep adaptive backward differentiation formulas/numerical differentiation formulas as Leap timesteppers. The method here matches that of MATLAB ode15s and Scipy's
integrate.bdf. In particular, this paper from Shampine and Reichelt describes the method well.Draft until:
requirements.txtback to main before merge