Add warning for batched mode when a sub objective has rev mode

[dpanici]

Does not resolve #2243 but at least warns users of the situation occuring

I am not sure if we can do batched as not fwd mode: maybe we can do it as like batched all at once but in reverse mode if all sub-objectives are rev mode.
Either way this will just throw a warning in the case that batched is set as ObjectiveFunction deriv_mode but any sub objective has rev mode, since that indicates a probable inefficiency.

[github-actions]

Memory benchmark result

|               Test Name                |      %Δ      |    Master (MB)     |      PR (MB)       |    Δ (MB)    |    Time PR (s)     |  Time Master (s)   |
| -------------------------------------- | ------------ | ------------------ | ------------------ | ------------ | ------------------ | ------------------ |
  test_objective_jac_w7x                 |   -0.32 %    |     4.253e+03      |     4.239e+03      |    -13.72    |       30.32        |       28.24        |
  test_proximal_jac_w7x_with_eq_update   |   -0.20 %    |     6.876e+03      |     6.863e+03      |    -13.47    |       154.28       |       150.54       |
  test_proximal_freeb_jac                |   -0.07 %    |     1.355e+04      |     1.354e+04      |    -10.04    |       79.85        |       79.08        |
  test_proximal_freeb_jac_blocked        |   -0.20 %    |     7.899e+03      |     7.883e+03      |    -15.69    |       69.18        |       68.73        |
  test_proximal_freeb_jac_batched        |    0.09 %    |     7.881e+03      |     7.889e+03      |     7.11     |       67.98        |       69.27        |
  test_proximal_jac_ripple               |   -3.14 %    |     3.808e+03      |     3.688e+03      |   -119.69    |       52.79        |       53.30        |
  test_proximal_jac_ripple_bounce1d      |    1.27 %    |     3.962e+03      |     4.013e+03      |    50.41     |       66.54        |       67.18        |
  test_eq_solve                          |    1.28 %    |     1.819e+03      |     1.842e+03      |    23.24     |       51.46        |       52.01        |
  test_objective_quadratic_flux_jac      |   -0.09 %    |     2.712e+03      |     2.710e+03      |    -2.39     |       52.36        |       52.89        |

For the memory plots, go to the summary of Memory Benchmarks workflow and download the artifact.

[YigitElma]

The message was not very clear, and there was a typo or may not even support reverse mode at all should be forward mode.

Also, while we are at it, would we like to print something like this at the end?

if verbose > 1:
    print(f"{self.name} deriv_mode : {self._deriv_mode}")
    if self._deriv_mode == "batched":
        print(f"{self.name} jac_chunk_size: {self._jac_chunk_size}")
    else:
        for o in self.objectives:
             print(f"{o.name} jac_chunk_size: {o._jac_chunk_size}")
    timer.disp("Objective build")

Should also fix the failing test

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 94.38%. Comparing base (33076c8) to head (20fcec1).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #2284   +/-   ##
=======================================
  Coverage   94.37%   94.38%           
=======================================
  Files         101      101           
  Lines       28953    28960    +7     
=======================================
+ Hits        27325    27333    +8     
+ Misses       1628     1627    -1     

Files with missing lines	Coverage Δ
desc/objectives/objective_funs.py	95.41% <100.00%> (+0.03%)	⬆️

... and 1 file with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[github-actions]

|             benchmark_name             |         dt(%)          |         dt(s)          |        t_new(s)        |        t_old(s)        | 
| -------------------------------------- | ---------------------- | ---------------------- | ---------------------- | ---------------------- |
 test_build_transform_fft_lowres         |     -0.40 +/- 8.49     | -3.42e-03 +/- 7.34e-02 |  8.61e-01 +/- 3.4e-02  |  8.65e-01 +/- 6.5e-02  |
 test_equilibrium_init_lowres            |     -0.23 +/- 4.01     | -1.53e-02 +/- 2.63e-01 |  6.53e+00 +/- 1.3e-01  |  6.55e+00 +/- 2.3e-01  |
 test_objective_compile_atf              |     -0.15 +/- 3.95     | -9.43e-03 +/- 2.44e-01 |  6.17e+00 +/- 1.5e-01  |  6.18e+00 +/- 1.9e-01  |
 test_objective_compute_atf              |     -3.82 +/- 18.25    | -8.31e-05 +/- 3.97e-04 |  2.09e-03 +/- 2.4e-04  |  2.18e-03 +/- 3.2e-04  |
 test_objective_jac_atf                  |     -2.11 +/- 4.65     | -3.37e-02 +/- 7.43e-02 |  1.56e+00 +/- 6.1e-02  |  1.60e+00 +/- 4.3e-02  |
 test_perturb_1                          |     +2.56 +/- 2.98     | +3.04e-01 +/- 3.54e-01 |  1.22e+01 +/- 3.1e-01  |  1.19e+01 +/- 1.7e-01  |
 test_proximal_jac_atf                   |     -0.81 +/- 1.16     | -4.28e-02 +/- 6.13e-02 |  5.26e+00 +/- 4.7e-02  |  5.30e+00 +/- 4.0e-02  |
 test_proximal_freeb_compute             |     -2.04 +/- 2.84     | -3.42e-03 +/- 4.77e-03 |  1.64e-01 +/- 3.7e-03  |  1.68e-01 +/- 3.1e-03  |
 test_solve_fixed_iter                   |     +3.20 +/- 3.54     | +7.83e-01 +/- 8.66e-01 |  2.53e+01 +/- 6.7e-01  |  2.45e+01 +/- 5.5e-01  |
 test_LinearConstraintProjection_build   |     +4.52 +/- 3.74     | +3.15e-01 +/- 2.61e-01 |  7.29e+00 +/- 2.4e-01  |  6.98e+00 +/- 9.4e-02  |
 test_objective_compute_ripple           |     -1.70 +/- 4.55     | -3.98e-03 +/- 1.07e-02 |  2.30e-01 +/- 7.3e-03  |  2.34e-01 +/- 7.8e-03  |
 test_objective_grad_ripple              |     -0.60 +/- 3.96     | -5.42e-03 +/- 3.56e-02 |  8.93e-01 +/- 1.9e-02  |  8.98e-01 +/- 3.0e-02  |
 test_objective_quadratic_flux_compute   |     +2.17 +/- 15.09    | +1.12e-03 +/- 7.81e-03 |  5.29e-02 +/- 6.8e-03  |  5.18e-02 +/- 3.8e-03  |
 test_build_transform_fft_midres         |     -4.59 +/- 2.76     | -4.22e-02 +/- 2.53e-02 |  8.76e-01 +/- 1.1e-02  |  9.18e-01 +/- 2.3e-02  |
 test_build_transform_fft_highres        |     -3.60 +/- 2.63     | -4.35e-02 +/- 3.18e-02 |  1.17e+00 +/- 2.8e-02  |  1.21e+00 +/- 1.6e-02  |
 test_equilibrium_init_medres            |     -5.01 +/- 2.89     | -3.59e-01 +/- 2.07e-01 |  6.81e+00 +/- 1.5e-01  |  7.17e+00 +/- 1.4e-01  |
 test_objective_compile_dshape_current   |     -1.28 +/- 2.73     | -5.39e-02 +/- 1.15e-01 |  4.17e+00 +/- 1.1e-01  |  4.23e+00 +/- 2.5e-02  |
 test_objective_compute_dshape_current   |     +4.88 +/- 13.76    | +3.27e-05 +/- 9.21e-05 |  7.02e-04 +/- 7.9e-05  |  6.70e-04 +/- 4.8e-05  |
 test_objective_jac_dshape_current       |     -1.45 +/- 24.25    | -3.50e-04 +/- 5.85e-03 |  2.38e-02 +/- 4.2e-03  |  2.41e-02 +/- 4.1e-03  |
 test_perturb_2                          |     -3.63 +/- 1.62     | -5.78e-01 +/- 2.58e-01 |  1.53e+01 +/- 1.6e-01  |  1.59e+01 +/- 2.0e-01  |
 test_proximal_jac_atf_with_eq_update    |     -0.30 +/- 1.22     | -3.62e-02 +/- 1.48e-01 |  1.21e+01 +/- 6.8e-02  |  1.21e+01 +/- 1.3e-01  |
 test_proximal_freeb_jac                 |     -1.63 +/- 3.03     | -7.90e-02 +/- 1.47e-01 |  4.77e+00 +/- 1.2e-01  |  4.85e+00 +/- 8.3e-02  |
 test_solve_fixed_iter_compiled          |     -2.76 +/- 2.16     | -1.69e-01 +/- 1.32e-01 |  5.94e+00 +/- 8.9e-02  |  6.11e+00 +/- 9.8e-02  |
 test_objective_compute_ripple_bounce1d  |     -2.73 +/- 3.72     | -8.47e-03 +/- 1.15e-02 |  3.02e-01 +/- 3.1e-03  |  3.10e-01 +/- 1.1e-02  |
 test_objective_grad_ripple_bounce1d     |     +0.91 +/- 2.11     | +8.62e-03 +/- 1.99e-02 |  9.52e-01 +/- 1.7e-02  |  9.43e-01 +/- 1.1e-02  |
 test_objective_quadratic_flux_jac       |     -0.32 +/- 0.50     | -2.72e-02 +/- 4.23e-02 |  8.43e+00 +/- 3.9e-02  |  8.46e+00 +/- 1.6e-02  |

Github CI performance can be noisy. When evaluating the benchmarks, developers should take this into account.

[YigitElma]

Other then the changelog issue it is fine.

[ddudt]

I would prefer o.name, since it would be more clear if the user has multiple of the same objectives and gave them unique names. Also because I think we typically use obj.name instead of the class name when printing objectives.

[YigitElma]

Suggested change

	o.__class__.__name__ for o in self.objectives if o._deriv_mode == "rev"
	o.name for o in self.objectives if o._deriv_mode == "rev"