Expected behavior
LightGBMTunerCV should work without problems when using a custom objective function. LightGBM cv() docs specify that a custom function can be passed to the "objective" parameter.
Environment
- Optuna version: 4.1.0
- Optuna Integration version: 4.1.0
- Python version: 3.11.5
- OS: macOS-13.4-arm64-arm-64bit
Error messages, stack traces, or logs
Traceback (most recent call last):
File "/Users/pmandiola/Documents/Lidz/data-exploration/.venv/lib/python3.11/site-packages/marimo/_runtime/executor.py", line 157, in execute_cell
exec(cell.body, glbls)
Cell marimo://1_train_model.py#cell=cell-10, line 24, in <module>
tuner.run()
File "/Users/pmandiola/Documents/Lidz/data-exploration/.venv/lib/python3.11/site-packages/optuna_integration/_lightgbm_tuner/optimize.py", line 475, in run
self.tune_feature_fraction()
File "/Users/pmandiola/Documents/Lidz/data-exploration/.venv/lib/python3.11/site-packages/optuna_integration/_lightgbm_tuner/optimize.py", line 500, in tune_feature_fraction
self._tune_params([param_name], len(param_values), sampler, "feature_fraction")
File "/Users/pmandiola/Documents/Lidz/data-exploration/.venv/lib/python3.11/site-packages/optuna_integration/_lightgbm_tuner/optimize.py", line 583, in _tune_params
study.optimize(
File "/Users/pmandiola/Documents/Lidz/data-exploration/.venv/lib/python3.11/site-packages/optuna/study/study.py", line 475, in optimize
_optimize(
File "/Users/pmandiola/Documents/Lidz/data-exploration/.venv/lib/python3.11/site-packages/optuna/study/_optimize.py", line 63, in _optimize
_optimize_sequential(
File "/Users/pmandiola/Documents/Lidz/data-exploration/.venv/lib/python3.11/site-packages/optuna/study/_optimize.py", line 160, in _optimize_sequential
frozen_trial = _run_trial(study, func, catch)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/pmandiola/Documents/Lidz/data-exploration/.venv/lib/python3.11/site-packages/optuna/study/_optimize.py", line 248, in _run_trial
raise func_err
File "/Users/pmandiola/Documents/Lidz/data-exploration/.venv/lib/python3.11/site-packages/optuna/study/_optimize.py", line 197, in _run_trial
value_or_values = func(trial)
^^^^^^^^^^^
File "/Users/pmandiola/Documents/Lidz/data-exploration/.venv/lib/python3.11/site-packages/optuna_integration/_lightgbm_tuner/optimize.py", line 340, in __call__
self._postprocess(trial, elapsed_secs, average_iteration_time)
File "/Users/pmandiola/Documents/Lidz/data-exploration/.venv/lib/python3.11/site-packages/optuna_integration/_lightgbm_tuner/optimize.py", line 271, in _postprocess
trial._trial_id, _LGBM_PARAMS_KEY, json.dumps(self.lgbm_params)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/opt/homebrew/Cellar/python@3.11/3.11.5/Frameworks/Python.framework/Versions/3.11/lib/python3.11/json/__init__.py", line 231, in dumps
return _default_encoder.encode(obj)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/opt/homebrew/Cellar/python@3.11/3.11.5/Frameworks/Python.framework/Versions/3.11/lib/python3.11/json/encoder.py", line 200, in encode
chunks = self.iterencode(o, _one_shot=True)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/opt/homebrew/Cellar/python@3.11/3.11.5/Frameworks/Python.framework/Versions/3.11/lib/python3.11/json/encoder.py", line 258, in iterencode
return _iterencode(o, 0)
^^^^^^^^^^^^^^^^^
File "/opt/homebrew/Cellar/python@3.11/3.11.5/Frameworks/Python.framework/Versions/3.11/lib/python3.11/json/encoder.py", line 180, in default
raise TypeError(f'Object of type {o.__class__.__name__} '
TypeError: Object of type function is not JSON serializableSteps to reproduce
- Create a custom objective and eval function (I'm using focal loss as defined here)
- Pass the custom function in the "objective" parameter for LightGBMTunerCV
- Run the tuner
dtrain = lgb.Dataset(X_train, y_train)
params = {
"objective": focal_loss,
"metric": "focal_loss",
"verbosity": -1,
"boosting_type": "gbdt"
}
tuner = lgbo.LightGBMTunerCV(
params,
dtrain,
callbacks=[early_stopping(100), log_evaluation(100)],
feval = eval_error
)
tuner.run()Additional context (optional)
The error seems to occur when the lgbm_params are stored as json in the Trial object after running each trial, as the custom function is not json serializable. Happy to contribute with a fix if possible.
Expected behavior
LightGBMTunerCV should work without problems when using a custom objective function. LightGBM cv() docs specify that a custom function can be passed to the "objective" parameter.
Environment
Error messages, stack traces, or logs
Steps to reproduce
Additional context (optional)
The error seems to occur when the lgbm_params are stored as json in the Trial object after running each trial, as the custom function is not json serializable. Happy to contribute with a fix if possible.