JaxComplexPolySlab support
#1698
Conversation
There was a problem hiding this comment.
Wow, this is looking really good! Thanks @yaugenst-flex . I think this turned out to be a lot more changes than I anticipated. So this pretty much makes JaxPolySlab dilation and sidewall operations traceable?
Well, at least those that are used in the polyslab splitting routine, yeah. Although |
yaugenst-flex
force-pushed
the
yaugenst-flex/issue1675
branch
3 times, most recently
from
8a01af0
to
b64717f
Compare
yaugenst-flex
force-pushed
the
yaugenst-flex/issue1675
branch
from
0ea06dc
to
6edf491
Compare
| @@ -1913,6 +1930,10 @@ def test_sidewall_angle_validator(log_capture, sidewall_angle, log_expected): # | |||
|
|
|||
| jax_polyslab1 = JaxPolySlab(axis=POLYSLAB_AXIS, vertices=VERTICES, slab_bounds=(-1, 1)) | |||
|
|
|||
| msg = "'JaxPolySlab' does not yet perform the full adjoint gradient treatment " | |||
| for entry in td.log._static_cache: | |||
There was a problem hiding this comment.
we have an assert_log_level and AssertLogLevel (context manager version) in tests/utils. But this is probably fine.
Closes #1675
Initial version of
JaxComplexPolySlab. Works in principle, but still needs tests. Supports differentiation w.r.t. bothverticesandslab_bounds.