print(
'valid: f1: %.5f, precision: %.5f, recall: %.5f, best f1: %.5f\n'
% (f1, precision, recall, self.best_val_f1))
f1, precision, recall = evaluate(test_data, self.model)
print('test: f1: %.5f, precision: %.5f, recall: %.5f\n' %
(f1, precision, recall))
if __name__ == '__main__':
train_generator = data_generator(train_data, batch_size)
predecessor_model_name = 'predecessor_best.weights'
predecessor_evaluator = Evaluator(predecessor_model_name)
predecessor_model.fit(train_generator.generator(),
steps_per_epoch=len(train_generator),
epochs=epochs,
callbacks=[predecessor_evaluator])
theseus_model_name = 'theseus_best.weights'
theseus_evaluator = Evaluator(theseus_model_name)
theseus_model.fit_generator(train_generator.generator(),
steps_per_epoch=len(train_generator),
epochs=epochs * 2,
callbacks=[theseus_evaluator])
theseus_model.load_weights(theseus_model_name)
successor_model_name = 'successor_best.weights'
successor_evaluator = Evaluator(successor_model_name)
successor_model.fit_generator(train_generator.generator(),
steps_per_epoch=len(train_generator),
% (filename, filename + '.pred.json')).read().strip())
return metrics
class Evaluator(keras.callbacks.Callback):
"""评估和保存模型
"""
def __init__(self):
self.best_val_f1 = 0.
def on_epoch_end(self, epoch, logs=None):
metrics = evaluate(
'/home/mingming.xu/datasets/NLP/qa/dureader_robust-data/dev.json')
if float(metrics['F1']) >= float(self.best_val_f1):
self.best_val_f1 = metrics['F1']
model.save_weights('best_model.weights')
metrics['BEST F1'] = self.best_val_f1
print(metrics)
if __name__ == '__main__':
train_generator = data_generator(train_data, batch_size)
evaluator = Evaluator()
model.fit(train_generator.generator(),
steps_per_epoch=len(train_generator),
epochs=5,
callbacks=[evaluator])
else:
model.load_weights('best_model.weights')