Codecov Report

❌ Patch coverage is 94.59459% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 89.18%. Comparing base (255fc3d) to head (9638811).
⚠️ Report is 26 commits behind head on master.

@@            Coverage Diff             @@
##           master    #6243      +/-   ##
==========================================
+ Coverage   89.11%   89.18%   +0.07%     
==========================================
  Files         208      208              
  Lines       13796    13818      +22     
==========================================
+ Hits        12294    12324      +30     
+ Misses       1502     1494       -8     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

@gen740 @kAIto47802 Could you review this PR?

This pull request has not seen any recent activity.

I used the following benchmark code (in #6244):

import optuna


def objective(trial: optuna.Trial) -> float:
    return sum(trial.suggest_float(f"x{i}", -5, 5)**2 for i in range(5))


sampler = optuna.samplers.GPSampler(seed=0)
study = optuna.create_study(sampler=sampler)
study.optimize(objective, n_trials=300)
print((study.trials[-1].datetime_complete - study.trials[0].datetime_start).total_seconds())

And I found that the master branch is much faster:

• Master: 39.261927s
• This PR: 55.45938s

I tested several times, but the results were consistent.

@gen740
Is it probably because I haven't merged the upstream master?
I pushed the latest version, so could you please benchmark again?

@nabenabe0928
Thank you! I reran the benchmark and confirmed that the performance looks identical.
I also used wfg4 and ran the multi-objective optimization benchmark (see the code below), but the performance is almost the same as on the master branch.

• master: 143.247278
• this pr: 142.377885

import optuna
import optunahub

wfg = optunahub.load_module("benchmarks/wfg")
wfg4 = wfg.Problem(function_id=4, n_objectives=4, dimension=8, k=6)

sampler = optuna.samplers.GPSampler(seed=0)
study = optuna.create_study(
    sampler=sampler,
    directions=wfg4.directions,
)

study.optimize(wfg4, n_trials=100)

print(
    (
        study.trials[-1].datetime_complete - study.trials[0].datetime_start
    ).total_seconds()
)