def preTrainModel(net, data_path=None):
D = AutoSpeechDataset(
os.path.join("../sample_data/test_data1", 'data01.data'))
D.read_dataset()
metadata = D.get_metadata()
data_manager = DataManager(metadata, D.get_train())
train_x, train_y, val_x, val_y = data_manager.get_train_data(
train_loop_num=11,
model_num=1,
round_num=2,
use_new_train=True,
use_mfcc=True)
net.init_model(train_x.shape[1:], metadata["class_num"])
net.fit(train_x, train_y, (val_x, val_y), 2)
result = net.predict(val_x)
print('a')
if self._context.is_finished:
log("Finish all stages")
self.done_training = True
self._last_pred = preds
except MemoryError as mem_error:
log("MemoryError has occurred: {}".format(mem_error))
self._has_exception = True
self.done_training = True
except Exception as exp:
log("Exception has occurred: {}".format(exp))
self._has_exception = True
self.done_training = True
return self._last_pred
if __name__ == '__main__':
from ingestion.dataset import AutoSpeechDataset
D = AutoSpeechDataset(os.path.join("../sample_data/DEMO", 'train.data'))
D.read_dataset()
m = Model(D.get_metadata())
m.train(D.get_train())
m.test(D.get_test())
m.train(D.get_train())
m.test(D.get_test())
m.train(D.get_train())
m.test(D.get_test())
# m.train(D.get_train())
# m.test(D.get_test())