ε and Γ_c with Fourier Bounce #1290

unalmis · 2024-10-03T05:56:42Z

This PR updates the

Gamma_c, Gamma_c Velasco Gamma_c (Nemov and Velasco et. al) Γ_c #1042
effective ripple Effective ripple ε #1003

objectives to use the numerical methods from #1119

Also includes some incremental performance improvements

github-actions · 2024-10-03T06:31:45Z

|             benchmark_name             |         dt(%)          |         dt(s)          |        t_new(s)        |        t_old(s)        | 
| -------------------------------------- | ---------------------- | ---------------------- | ---------------------- | ---------------------- |
 test_build_transform_fft_lowres         |     +3.05 +/- 9.20     | +1.56e-02 +/- 4.73e-02 |  5.29e-01 +/- 4.7e-02  |  5.14e-01 +/- 6.8e-03  |
 test_equilibrium_init_medres            |     +2.02 +/- 2.52     | +8.18e-02 +/- 1.02e-01 |  4.13e+00 +/- 9.8e-02  |  4.04e+00 +/- 2.9e-02  |
 test_equilibrium_init_highres           |     +1.40 +/- 2.54     | +7.45e-02 +/- 1.35e-01 |  5.39e+00 +/- 1.2e-01  |  5.31e+00 +/- 5.6e-02  |
 test_objective_compile_dshape_current   |     -0.25 +/- 0.82     | -9.62e-03 +/- 3.21e-02 |  3.88e+00 +/- 2.1e-02  |  3.89e+00 +/- 2.5e-02  |
 test_objective_compute_dshape_current   |     +0.88 +/- 2.14     | +4.49e-05 +/- 1.10e-04 |  5.16e-03 +/- 1.0e-04  |  5.11e-03 +/- 3.5e-05  |
 test_objective_jac_dshape_current       |     -1.55 +/- 6.11     | -6.69e-04 +/- 2.63e-03 |  4.24e-02 +/- 1.7e-03  |  4.30e-02 +/- 2.0e-03  |
 test_perturb_2                          |     -0.01 +/- 1.01     | -2.69e-03 +/- 1.95e-01 |  1.94e+01 +/- 1.0e-01  |  1.94e+01 +/- 1.7e-01  |
 test_proximal_freeb_jac                 |     -0.41 +/- 1.13     | -3.02e-02 +/- 8.28e-02 |  7.32e+00 +/- 4.8e-02  |  7.35e+00 +/- 6.8e-02  |
 test_solve_fixed_iter                   |     +0.19 +/- 1.74     | +5.86e-02 +/- 5.45e-01 |  3.13e+01 +/- 4.3e-01  |  3.13e+01 +/- 3.3e-01  |
 test_LinearConstraintProjection_build   |     -0.79 +/- 2.92     | -7.98e-02 +/- 2.96e-01 |  1.00e+01 +/- 2.0e-01  |  1.01e+01 +/- 2.2e-01  |
 test_build_transform_fft_midres         |     +4.68 +/- 3.50     | +2.81e-02 +/- 2.11e-02 |  6.30e-01 +/- 2.0e-02  |  6.02e-01 +/- 7.4e-03  |
 test_build_transform_fft_highres        |     +1.65 +/- 2.35     | +1.60e-02 +/- 2.29e-02 |  9.88e-01 +/- 2.0e-02  |  9.72e-01 +/- 1.0e-02  |
 test_equilibrium_init_lowres            |     +5.18 +/- 3.07     | +1.97e-01 +/- 1.16e-01 |  3.99e+00 +/- 1.1e-01  |  3.79e+00 +/- 3.3e-02  |
 test_objective_compile_atf              |     -0.07 +/- 1.90     | -5.81e-03 +/- 1.55e-01 |  8.12e+00 +/- 1.2e-01  |  8.13e+00 +/- 9.9e-02  |
 test_objective_compute_atf              |     +0.07 +/- 2.36     | +1.17e-05 +/- 3.73e-04 |  1.58e-02 +/- 2.7e-04  |  1.58e-02 +/- 2.6e-04  |
 test_objective_jac_atf                  |     -0.66 +/- 2.43     | -1.33e-02 +/- 4.86e-02 |  1.99e+00 +/- 4.2e-02  |  2.00e+00 +/- 2.4e-02  |
 test_perturb_1                          |     +0.14 +/- 1.33     | +2.08e-02 +/- 1.94e-01 |  1.46e+01 +/- 1.6e-01  |  1.46e+01 +/- 1.1e-01  |
 test_proximal_jac_atf                   |     +0.46 +/- 0.82     | +3.76e-02 +/- 6.78e-02 |  8.27e+00 +/- 4.1e-02  |  8.23e+00 +/- 5.4e-02  |
 test_proximal_freeb_compute             |     -0.42 +/- 1.32     | -8.31e-04 +/- 2.63e-03 |  1.98e-01 +/- 1.5e-03  |  1.99e-01 +/- 2.1e-03  |
 test_solve_fixed_iter_compiled          |     +0.31 +/- 0.57     | +6.31e-02 +/- 1.17e-01 |  2.05e+01 +/- 1.0e-01  |  2.04e+01 +/- 5.3e-02  |

review-notebook-app · 2024-10-19T20:58:04Z

Check out this pull request on

See visual diffs & provide feedback on Jupyter Notebooks.

Powered by ReviewNB

old

codecov · 2024-12-19T10:09:49Z

Codecov Report

Attention: Patch coverage is 98.71795% with 4 lines in your changes missing coverage. Please review.

Project coverage is 95.64%. Comparing base (e357a9b) to head (08b31b2).
Report is 120 commits behind head on master.

Files with missing lines	Patch %	Lines
desc/objectives/_fast_ion.py	96.55%	2 Missing ⚠️
desc/equilibrium/coords.py	66.66%	1 Missing ⚠️
desc/objectives/_neoclassical.py	95.45%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1290      +/-   ##
==========================================
+ Coverage   95.61%   95.64%   +0.03%     
==========================================
  Files          98      101       +3     
  Lines       25425    25542     +117     
==========================================
+ Hits        24309    24429     +120     
+ Misses       1116     1113       -3

Files with missing lines	Coverage Δ
desc/batching.py	`86.78% <100.00%> (+0.15%)`	⬆️
desc/compute/__init__.py	`100.00% <ø> (ø)`
desc/compute/_deprecated.py	`100.00% <100.00%> (ø)`
desc/compute/_fast_ion.py	`100.00% <100.00%> (ø)`
desc/compute/_geometry.py	`99.53% <100.00%> (+0.51%)`	⬆️
desc/compute/_neoclassical.py	`100.00% <100.00%> (+0.91%)`	⬆️
desc/grid.py	`94.58% <100.00%> (ø)`
desc/integrals/_bounce_utils.py	`84.61% <100.00%> (+0.24%)`	⬆️
desc/integrals/bounce_integral.py	`99.48% <100.00%> (+1.54%)`	⬆️
desc/integrals/quad_utils.py	`100.00% <100.00%> (ø)`
... and 10 more

... and 1 file with indirect coverage changes

unalmis · 2024-12-19T10:26:52Z

all requests are now done.

desc/integrals/bounce_integral.py

Set default automorphism to None to simplify for user in case they forget to set to None

rahulgaur104

A bunch of minor points. I'll ask some more questions over the next hour.

rahulgaur104 · 2024-12-24T10:35:45Z

desc/compute/_deprecated.py

+}
+
+
+def _compute(fun, fun_data, data, grid, num_pitch, simp=False, reduce=True):


Does the AD chain go through each iteration of the loop when you do multiple alpha or rho values?

How would I test this

desc/compute/_deprecated.py

desc/compute/_fast_ion.py

rahulgaur104 · 2024-12-24T10:43:57Z

desc/compute/_fast_ion.py

+        "max_tz |B|",
+        "cvdrift0",
+        "gbdrift (periodic)",
+        "gbdrift (secular)/phi",


Should we add a test to ensure that the secular gbdrift gives the same answer as periodic gbdrift as field line length is increased? Otherwise, it can be added later.

It will not because the secular term drives the computation here to zero. We warn this in the objective docstring for now. Without the secular term it converges quickly to basically $\rho * Gamma_c Nemov$

desc/compute/_neoclassical.py

desc/integrals/_bounce_utils.py

rahulgaur104 · 2024-12-24T10:48:58Z

desc/integrals/_bounce_utils.py

-def interp_to_argmin(
-    h, points, knots, g, dg_dz, method="cubic", beta=-100, upper_sentinel=1e2
-):
+def interp_to_argmin(h, points, knots, g, dg_dz, method="cubic"):


No suggestion here but can you explain what this function is doing?

We interpolate the function h to the global minimum of g that lies between each bounce point. In our application $g = \vert B \vert$ and $h$ is either $\nabla \psi$ or $e_{\alpha}$ at fixed $\rho, \phi$.

rahulgaur104 · 2024-12-24T14:06:12Z

tests/test_neoclassical.py

+    """Test effective ripple with W7-X against NEO."""
+    eq = get("W7-X")
+    rho = np.linspace(0, 1, 10)
+    grid = LinearGrid(rho=rho, M=eq.M_grid, N=eq.N_grid, NFP=eq.NFP, sym=False)


How sensitive is the ripple to the grid resolution now?

Merge branch 'Gamma_c' into ku/fourier_bounce_neo

c5d6a99

unalmis added the easy Short and simple to code or review label Oct 3, 2024

unalmis changed the base branch from master to ku/fourier_bounce October 3, 2024 05:56

unalmis changed the title ~~Writing neoclassical functions to use pseudospectral methods~~ Writing ε¹ᐧ⁵ and Γ_c compute funs with new methods Oct 3, 2024

unalmis mentioned this pull request Oct 3, 2024

fourier bounce #1119

Merged

unalmis changed the title ~~Writing ε¹ᐧ⁵ and Γ_c compute funs with new methods~~ Writing ε and Γ_c compute funs with new methods Oct 3, 2024

unalmis added 17 commits October 3, 2024 14:42

write eps_eff compute function with bounce2d

b1a5c81

Merge branch 'ku/fourier_bounce' into ku/fourier_bounce_neo

a81014d

Merge branch 'ku/fourier_bounce' into ku/fourier_bounce_neo

29d24f5

Merge branch 'ku/fourier_bounce' into ku/fourier_bounce_neo

eb448e9

Debugging JAX issues

15875d9

Merge branch 'ku/fourier_bounce' into ku/fourier_bounce_neo

5eb11eb

Merge branch 'ku/fourier_bounce' into ku/fourier_bounce_neo

22500b2

Merge branch 'ku/fourier_bounce' into ku/fourier_bounce_neo

70a3a9e

Merge branch 'ku/fourier_bounce' into ku/fourier_bounce_neo

dc82cfa

Merge commit '70a3a9e' into ku/fourier_bounce_neo

d20ee87

Merge branch 'Gamma_c' into ku/fourier_bounce_neo

fd723a6

Merge branch 'ku/fourier_bounce' into ku/fourier_bounce_neo

8b5cdb8

Merge branch 'ku/fourier_bounce' into ku/fourier_bounce_neo

276f6af

Merge branch 'ku/fourier_bounce' into ku/fourier_bounce_neo

4d89326

Merge branch 'Gamma_c' into ku/fourier_bounce_neo

d6d9407

Merge branch 'ku/fourier_bounce' into ku/fourier_bounce_neo

4ed1701

Merge branch 'ku/fourier_bounce' into ku/fourier_bounce_neo

2f3a03a

Writing new compute funs with Bounce2D

5a58582

unalmis added a commit that referenced this pull request Oct 20, 2024

Merging changes from downstream branch #1290

927ab54

Merge branch 'ku/fourier_bounce' into ku/fourier_bounce_neo

5ea5c02

unalmis changed the title ~~Writing ε and Γ_c compute funs with new methods~~ ε and Γ_c objectives with Bounce2D Oct 20, 2024

unalmis removed the easy Short and simple to code or review label Oct 20, 2024

Fix space

67484f1

unalmis requested a review from f0uriest December 19, 2024 09:39

Fighting with the docs. all of these edits were necesseary apparently

9848354

unalmis requested review from a team and removed request for a team December 19, 2024 10:24

.

548674e

unalmis added the P3 Highest Priority, someone is/should be actively working on this label Dec 19, 2024

Update notebook name so it appears in docs

f3237b1

unalmis commented Dec 19, 2024

View reviewed changes

desc/integrals/bounce_integral.py Outdated Show resolved Hide resolved

unalmis added 6 commits December 19, 2024 06:41

Fix docs

274352c

Fix math comment

4097aa8

Fix math comment

30988de

Fix docs

3fd85a1

adding back method useful for nan debugging

0a4c3eb

Remove unneded min

d3c84d3

unalmis mentioned this pull request Dec 20, 2024

xla memory leak on compiled jax function #1484

Open

unalmis added 6 commits December 21, 2024 02:51

Fix batching

735bf21

Set default automorphism to None to simplify for user

059f217

Simplify api for user (#1487)

d62435c

Set default automorphism to None to simplify for user in case they forget to set to None

Merge branch 'nan_debug' into ku/fourier_bounce_neo

441630d

Remove softargmax interp to argmin now that nan gradient identified

2cb0f3e

Merge branch 'master' into ku/fourier_bounce_neo

08b31b2

rahulgaur104 approved these changes Dec 24, 2024

View reviewed changes

unalmis merged commit 7d378c2 into master Dec 24, 2024
25 checks passed

unalmis deleted the ku/fourier_bounce_neo branch December 24, 2024 22:47

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

ε and Γ_c with Fourier Bounce #1290

ε and Γ_c with Fourier Bounce #1290

unalmis commented Oct 3, 2024 •

edited

Loading

github-actions bot commented Oct 3, 2024 •

edited

Loading

review-notebook-app bot commented Oct 19, 2024

codecov bot commented Dec 19, 2024 •

edited

Loading

unalmis commented Dec 19, 2024 •

edited

Loading

rahulgaur104 left a comment

rahulgaur104 Dec 24, 2024

unalmis Dec 24, 2024

rahulgaur104 Dec 24, 2024

unalmis Dec 24, 2024 •

edited

Loading

rahulgaur104 Dec 24, 2024

unalmis Dec 24, 2024

rahulgaur104 Dec 24, 2024

		}


		def _compute(fun, fun_data, data, grid, num_pitch, simp=False, reduce=True):

ε and Γ_c with Fourier Bounce #1290

ε and Γ_c with Fourier Bounce #1290

Conversation

unalmis commented Oct 3, 2024 • edited Loading

github-actions bot commented Oct 3, 2024 • edited Loading

review-notebook-app bot commented Oct 19, 2024

codecov bot commented Dec 19, 2024 • edited Loading

Codecov Report

unalmis commented Dec 19, 2024 • edited Loading

rahulgaur104 left a comment

Choose a reason for hiding this comment

rahulgaur104 Dec 24, 2024

Choose a reason for hiding this comment

unalmis Dec 24, 2024

Choose a reason for hiding this comment

rahulgaur104 Dec 24, 2024

Choose a reason for hiding this comment

unalmis Dec 24, 2024 • edited Loading

Choose a reason for hiding this comment

rahulgaur104 Dec 24, 2024

Choose a reason for hiding this comment

unalmis Dec 24, 2024

Choose a reason for hiding this comment

rahulgaur104 Dec 24, 2024

Choose a reason for hiding this comment

unalmis commented Oct 3, 2024 •

edited

Loading

github-actions bot commented Oct 3, 2024 •

edited

Loading

codecov bot commented Dec 19, 2024 •

edited

Loading

unalmis commented Dec 19, 2024 •

edited

Loading

unalmis Dec 24, 2024 •

edited

Loading