emotion detection

from kutilities.callbacks import MetricsCallback, PlottingCallback
from sklearn.metrics import f1_score, precision_score, recall_score
from keras.callbacks import ModelCheckpoint, TensorBoard

metrics = {
    "f1_e": (lambda y_test, y_pred:
             f1_score(y_test, y_pred, average='micro',
                      labels=[emotion2label['happy'],
                              emotion2label['sad'],
                              emotion2label['angry']
                              ])),
    "precision_e": (lambda y_test, y_pred:
                    precision_score(y_test, y_pred, average='micro',
                                    labels=[emotion2label['happy'],
                                            emotion2label['sad'],
                                            emotion2label['angry']
                                            ])),
    "recoll_e": (lambda y_test, y_pred:
                 recall_score(y_test, y_pred, average='micro',
                              labels=[emotion2label['happy'],
                                      emotion2label['sad'],
                                      emotion2label['angry']
                                      ]))
}

_datasets = {}
_datasets["dev"] = [[message_first_message_dev, message_second_message_dev, message_third_message_dev],
                    np.array(labels_categorical_dev)]
_datasets["val"] = [[message_first_message_val, message_second_message_val, message_third_message_val],
                    np.array(labels_categorical_val)]

metrics_callback = MetricsCallback(datasets=_datasets, metrics=metrics)
y_pred = model.predict([message_first_message_dev, message_second_message_dev, message_third_message_dev])
from sklearn.metrics import classification_report

for title, metric in metrics.items():
    print(title, metric(labels_categorical_dev.argmax(axis=1), y_pred.argmax(axis=1)))
print(classification_report(labels_categorical_dev.argmax(axis=1), y_pred.argmax(axis=1)))