def validate_parameters(self):
"""Ensure the provided parameters are valid and properly formatted"""
#################### metrics_params/metrics ####################
if (self.metrics is not None) and ("metrics"
in self.metrics_params.keys()):
raise ValueError(
"`metrics` may be provided as a kwarg, or as a `metrics_params` key, but NOT BOTH. Received: "
+
f"\n `metrics`={self.metrics}\n `metrics_params`={self.metrics_params}"
)
else:
_metrics_alias = "metrics"
if self.metrics is None:
try:
self.metrics = self.metrics_params["metrics"]
except KeyError:
self.metrics = self.metrics_params["metrics_map"]
_metrics_alias = "metrics_map"
self.metrics = format_metrics(self.metrics)
self.metrics_params = {
**{
_metrics_alias: self.metrics
},
**self.metrics_params
}
def test_get_formatted_target_metric_value_error(target_metric):
with pytest.raises(ValueError):
get_formatted_target_metric(target_metric,
format_metrics(["roc_auc_score"]))
def validate_parameters(self):
"""Ensure the provided parameters are valid and properly formatted"""
#################### results_path ####################
if self.results_path is None:
G.warn(
"Received results_path=None. Results will not be stored at all."
)
elif isinstance(self.results_path, str):
if not self.results_path.endswith(ASSETS_DIRNAME):
self.results_path = os.path.join(self.results_path,
ASSETS_DIRNAME)
self.result_paths["root"] = self.results_path
if not os.path.exists(self.results_path):
make_dirs(self.results_path, exist_ok=True)
else:
raise TypeError(
f"results_path must be None or str, not {self.results_path}")
#################### target_column ####################
if isinstance(self.target_column, str):
self.target_column = [self.target_column]
#################### file_blacklist ####################
self.file_blacklist = validate_file_blacklist(self.file_blacklist)
if self.results_path is None:
self.file_blacklist = "ALL"
#################### Train/Test Datasets ####################
if isinstance(self.train_dataset, str):
self.train_dataset = pd.read_csv(self.train_dataset)
if isinstance(self.test_dataset, str):
self.test_dataset = pd.read_csv(self.test_dataset)
#################### metrics_params/metrics ####################
if (self.metrics is not None) and ("metrics"
in self.metrics_params.keys()):
raise ValueError(
"`metrics` may be provided as a kwarg, or as a `metrics_params` key, but NOT BOTH. Received: "
+
f"\n `metrics`={self.metrics}\n `metrics_params`={self.metrics_params}"
)
else:
_metrics_alias = "metrics"
if self.metrics is None:
try:
self.metrics = self.metrics_params["metrics"]
except KeyError:
self.metrics = self.metrics_params["metrics_map"]
_metrics_alias = "metrics_map"
self.metrics = format_metrics(self.metrics)
self.metrics_params = {
**{
_metrics_alias: self.metrics
},
**self.metrics_params
}
#################### cv_type ####################
if isinstance(self.cv_type, str):
try:
self.cv_type = sk_cv.__getattribute__(self.cv_type)
except AttributeError:
raise AttributeError(
f"'{self.cv_type}' not in `sklearn.model_selection._split`"
)
#################### to_csv_params ####################
self.to_csv_params = {
k: v
for k, v in self.to_csv_params.items() if k != "path_or_buf"
}
#################### cross_experiment_params ####################
self.cross_experiment_params = dict(
cv_type=self.cv_type,
runs=self.runs,
global_random_seed=self.global_random_seed,
random_seeds=self.random_seeds,
random_seed_bounds=self.random_seed_bounds,
)
#################### experiment_callbacks ####################
if not isinstance(self.experiment_callbacks, list):
self.experiment_callbacks = [self.experiment_callbacks]
for cb in self.experiment_callbacks:
if not isclass(cb):
raise TypeError(
f"experiment_callbacks must be classes, not {type(cb)}: {cb}"
)
if cb.__name__ != "LambdaCallback":
raise ValueError(
f"experiment_callbacks must be LambdaCallback instances, not {cb}"
)