Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Returning float("inf") failed with MOTPE #5374

Closed
gen740 opened this issue Mar 29, 2024 · 1 comment · Fixed by #5435
Closed

Returning float("inf") failed with MOTPE #5374

gen740 opened this issue Mar 29, 2024 · 1 comment · Fixed by #5435
Labels
bug Issue/PR about behavior that is broken. Not for typos/examples/CI/test but for Optuna itself.

Comments

@gen740
Copy link
Collaborator

gen740 commented Mar 29, 2024 •
edited

Expected behavior

No errors occur.

Environment

  • Optuna version:4.0.0.dev
  • Python version:3.12.2
  • OS:macOS-14.4.1-arm64-arm-64bit

Error messages, stack traces, or logs

[I 2024-03-29 17:20:44,217] A new study created in memory with name: no-name-3289a723-a36c-4729-a37e-6b440b830668
[I 2024-03-29 17:20:44,218] Trial 0 finished with values: [-0.497867971784556, inf] and parameters: {'x': -0.497867971784556}. 
[I 2024-03-29 17:20:44,218] Trial 1 finished with values: [1.3219469606529486, 0.0] and parameters: {'x': 1.3219469606529486}. 
[I 2024-03-29 17:20:44,218] Trial 2 finished with values: [-2.9993137510959307, inf] and parameters: {'x': -2.9993137510959307}. 
[I 2024-03-29 17:20:44,218] Trial 3 finished with values: [-1.1860045642089614, inf] and parameters: {'x': -1.1860045642089614}. 
[I 2024-03-29 17:20:44,218] Trial 4 finished with values: [-2.119464655097322, inf] and parameters: {'x': -2.119464655097322}. 
[I 2024-03-29 17:20:44,218] Trial 5 finished with values: [-2.445968431387213, inf] and parameters: {'x': -2.445968431387213}. 
[I 2024-03-29 17:20:44,218] Trial 6 finished with values: [-1.8824387317339746, inf] and parameters: {'x': -1.8824387317339746}. 
[I 2024-03-29 17:20:44,218] Trial 7 finished with values: [-0.9266356377417135, inf] and parameters: {'x': -0.9266356377417135}. 
[I 2024-03-29 17:20:44,218] Trial 8 finished with values: [-0.6193951546159804, inf] and parameters: {'x': -0.6193951546159804}. 
[I 2024-03-29 17:20:44,218] Trial 9 finished with values: [0.23290040402014167, 0.0] and parameters: {'x': 0.23290040402014167}. 
/Users/gen740/home/PFN/optuna/optuna/_hypervolume/hssp.py:62: RuntimeWarning: invalid value encountered in subtract
  assert not np.any(reference_point - rank_i_loss_vals < 0)
/Users/gen740/home/PFN/optuna/optuna/_hypervolume/hssp.py:64: RuntimeWarning: invalid value encountered in subtract
  contribs = np.prod(reference_point - rank_i_loss_vals, axis=-1)
[I 2024-03-29 17:20:44,222] Trial 10 finished with values: [2.7299172994598835, 0.0] and parameters: {'x': 2.7299172994598835}. 
/Users/gen740/home/PFN/optuna/optuna/_hypervolume/utils.py:39: RuntimeWarning: invalid value encountered in subtract
  edge_length_y = reference_points_y_cummin[mask] - used_solution[:, 1]
[W 2024-03-29 17:20:44,223] Trial 11 failed with parameters: {} because of the following error: ValueError('probabilities contain NaN').
Traceback (most recent call last):
  File "/Users/gen740/home/PFN/optuna/optuna/study/_optimize.py", line 196, in _run_trial
    value_or_values = func(trial)
                      ^^^^^^^^^^^
  File "/Users/gen740/home/PFN/workspace/pr/5355/motpe_inf.py", line 5, in objective
    x = trial.suggest_float("x", -3, 3)
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/gen740/home/PFN/optuna/optuna/trial/_trial.py", line 163, in suggest_float
    suggested_value = self._suggest(name, distribution)
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/gen740/home/PFN/optuna/optuna/trial/_trial.py", line 629, in _suggest
    param_value = self.study.sampler.sample_independent(
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/gen740/home/PFN/optuna/optuna/samplers/_tpe/sampler.py", line 447, in sample_independent
    return self._sample(study, trial, {param_name: param_distribution})[param_name]
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/gen740/home/PFN/optuna/optuna/samplers/_tpe/sampler.py", line 487, in _sample
    samples_below = mpe_below.sample(self._rng.rng, self._n_ei_candidates)
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/gen740/home/PFN/optuna/optuna/samplers/_tpe/parzen_estimator.py", line 81, in sample
    sampled = self._mixture_distribution.sample(rng, size)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/gen740/home/PFN/optuna/optuna/samplers/_tpe/probability_distributions.py", line 41, in sample
    active_indices = rng.choice(len(self.weights), p=self.weights, size=batch_size)
                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "numpy/random/mtrand.pyx", line 971, in numpy.random.mtrand.RandomState.choice
ValueError: probabilities contain NaN
[W 2024-03-29 17:20:44,225] Trial 11 failed with value None.
Traceback (most recent call last):
  File "/Users/gen740/home/PFN/workspace/pr/5355/motpe_inf.py", line 11, in <module>
    study.optimize(objective, n_trials=20)
  File "/Users/gen740/home/PFN/optuna/optuna/study/study.py", line 451, in optimize
    _optimize(
  File "/Users/gen740/home/PFN/optuna/optuna/study/_optimize.py", line 62, in _optimize
    _optimize_sequential(
  File "/Users/gen740/home/PFN/optuna/optuna/study/_optimize.py", line 159, in _optimize_sequential
    frozen_trial = _run_trial(study, func, catch)
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/gen740/home/PFN/optuna/optuna/study/_optimize.py", line 247, in _run_trial
    raise func_err
  File "/Users/gen740/home/PFN/optuna/optuna/study/_optimize.py", line 196, in _run_trial
    value_or_values = func(trial)
                      ^^^^^^^^^^^
  File "/Users/gen740/home/PFN/workspace/pr/5355/motpe_inf.py", line 5, in objective
    x = trial.suggest_float("x", -3, 3)
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/gen740/home/PFN/optuna/optuna/trial/_trial.py", line 163, in suggest_float
    suggested_value = self._suggest(name, distribution)
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/gen740/home/PFN/optuna/optuna/trial/_trial.py", line 629, in _suggest
    param_value = self.study.sampler.sample_independent(
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/gen740/home/PFN/optuna/optuna/samplers/_tpe/sampler.py", line 447, in sample_independent
    return self._sample(study, trial, {param_name: param_distribution})[param_name]
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/gen740/home/PFN/optuna/optuna/samplers/_tpe/sampler.py", line 487, in _sample
    samples_below = mpe_below.sample(self._rng.rng, self._n_ei_candidates)
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/gen740/home/PFN/optuna/optuna/samplers/_tpe/parzen_estimator.py", line 81, in sample
    sampled = self._mixture_distribution.sample(rng, size)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/gen740/home/PFN/optuna/optuna/samplers/_tpe/probability_distributions.py", line 41, in sample
    active_indices = rng.choice(len(self.weights), p=self.weights, size=batch_size)
                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "numpy/random/mtrand.pyx", line 971, in numpy.random.mtrand.RandomState.choice
ValueError: probabilities contain NaN

Steps to reproduce

import optuna


def objective(trial):
    x = trial.suggest_float("x", -3, 3)
    return x, float("inf") if x < 0 else 0


sampler = optuna.samplers.TPESampler(seed=1)
study = optuna.create_study(directions=["minimize"] * 2, sampler=sampler)
study.optimize(objective, n_trials=20)

Run this code.

Additional context (optional)

No response
@gen740 gen740 added the bug Issue/PR about behavior that is broken. Not for typos/examples/CI/test but for Optuna itself. label Mar 29, 2024
@nabenabe0928
Copy link
Collaborator

nabenabe0928 commented Mar 30, 2024 •
edited

This error happens because the WFG calculates inf - inf, leading to nan.

In principle, we just need to return inf by definition if either the reference point or some of non-dominated points include inf or -inf.

@nabenabe0928 nabenabe0928 changed the title Returning float("inf") failed with MOTPE. Returning float("inf") failed with MOTPE Apr 15, 2024
@nabenabe0928 nabenabe0928 mentioned this issue May 8, 2024
Merged
@eukaryo eukaryo mentioned this issue May 8, 2024
Merged
@y0z y0z closed this as completed in #5435 May 27, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
bug Issue/PR about behavior that is broken. Not for typos/examples/CI/test but for Optuna itself.
Projects
None yet
Milestone
No milestone
Development

Successfully merging a pull request may close this issue.

Guarantee weights_below to be finite in MOTPE eukaryo/optuna
2 participants
@nabenabe0928 @gen740