def _fit_ranker(self, xtrain, ytrain, xtest, ytest, next_point):
start = datetime.now()
self._set_new_parameters(next_point)
try:
self.learner.fit(xtrain, ytrain, **self._fit_params)
ypred = self.learner(xtest)
loss = get_mean_loss(self.validation_loss, ytest, ypred)
time_taken = duration_till_now(start)
except:
self.logger.error(traceback.format_exc())
self.logger.info(
"For current parameter error occurred so taking loss as maximum value"
)
loss = 1.00
time_taken = duration_till_now(start)
return loss, time_taken
'cluster_id': str(cluster_id)
}
for name, evaluation_metric in lp_metric_dict[
learning_problem].items():
predictions = s_pred
if evaluation_metric in metrics_on_predictions:
logger.info("Metric on predictions")
predictions = y_pred
if "NDCG" in name:
evaluation_metric = make_ndcg_at_k_loss(k=n_objects)
predictions = y_pred
if "CategoricalTopK{}" == name:
k = int(n_objects / 2) + 1
evaluation_metric = topk_categorical_accuracy_np(k=k)
name = name.format(k)
metric_loss = get_mean_loss(evaluation_metric, Y_test,
predictions)
logger.info(ERROR_OUTPUT_STRING % (name, metric_loss))
if np.isnan(metric_loss):
results[name] = "\'Infinity\'"
else:
results[name] = "{0:.4f}".format(metric_loss)
dbConnector.insert_results(experiment_schema=experiment_schema,
experiment_table=experiment_table,
results=results)
if fold_id != 0:
dbConnector.mark_running_job_finished(current_job_id)
dbConnector.mark_running_job_finished(job_id)
except Exception as e:
if hasattr(e, 'message'):
message = e.message
else: