def fit(self, X_train, Y_train, X_val=None, y_val=None):
Y_train = Y_train.astype('uint8')
Y_train = np_utils.to_categorical(Y_train, self.nb_classes)
y_val = np_utils.to_categorical(y_val, self.nb_classes)
early_stop = MyEarlyStopping(patience=10, verbose=0)
filename = "weights_%s_%s_%s.h5" % (self.target, self.mode,
self.modelname)
checkpointer = MyModelCheckpoint(filename,
verbose=0,
save_best_only=True)
if (y_val is None):
self.model.fit(X_train,
Y_train,
batch_size=self.batch_size,
epochs=self.epochs,
validation_split=0.2,
callbacks=[early_stop, checkpointer],
verbose=2)
else:
self.model.fit(X_train,
Y_train,
batch_size=self.batch_size,
epochs=self.epochs,
validation_data=(X_val, y_val),
callbacks=[early_stop, checkpointer],
verbose=2)
self.model.load_weights(filename)
if self.mode == 'cv':
os.remove(filename)
return self
def fit(self,X_train,Y_train,X_val=None, y_val=None):
Y_train = Y_train.astype('uint8')
Y_train = np_utils.to_categorical(Y_train, self.nb_classes)
y_val = np_utils.to_categorical(y_val, self.nb_classes)
early_stop = MyEarlyStopping(patience=10, verbose=0)
filename = "weights_%s_%s_%s.h5" %(self.target, self.mode, self.modelname)
checkpointer = MyModelCheckpoint(filename,
verbose=0, save_best_only=True)
if (y_val is None):
self.model.fit(X_train, Y_train, batch_size=self.batch_size,
epochs=self.epochs,validation_split=0.2,
callbacks=[early_stop,checkpointer], verbose=2
)
else:
self.model.fit(X_train, Y_train, batch_size=self.batch_size,
epochs=self.epochs,validation_data=(X_val,y_val),
callbacks=[early_stop,checkpointer], verbose=2
)
self.model.load_weights(filename)
# print out some weights to see if they are binary
# print ('DEBUG: View weights.............')
# for layer in self.model.layers:
# config = layer.get_config()
# if config['name'] == 'conv1':
# print (layer.get_config())
# print (layer.get_weights())
if self.mode == 'cv':
os.remove(filename)
return self
def fit_7(self, training_generator, validation_generator,
class_weights_train):
early_stop = MyEarlyStopping(patience=20, verbose=0)
checkpointer = MyModelCheckpoint(filepath="2channel_sftf_7_train.h5",
verbose=0,
save_best_only=True)
#class_weight = {0: 20.,
#1: 1,
#}
self.model.fit_generator(generator=training_generator,
validation_data=validation_generator,
epochs=self.epochs,
class_weight=class_weights_train,
use_multiprocessing=False,
workers=4,
callbacks=[early_stop, checkpointer])
#self.model.load_weights("weights_best.h5")
# if self.mode == 'cv':
# os.remove("weights_%s_%s.h5" %(self.target, self.mode))
return self
def fit(self,
training_generator,
validation_generator,
class_weights,
weightpath="weights_best.h5"):
early_stop = MyEarlyStopping(patience=10, verbose=0)
checkpointer = MyModelCheckpoint(filepath=weightpath,
verbose=0,
save_best_only=True)
self.model.fit_generator(
generator=training_generator,
validation_data=validation_generator,
epochs=self.epochs,
class_weight=class_weights,
use_multiprocessing=False, #workers=2,
callbacks=[early_stop, checkpointer])
# self.model.load_weights(weightpath)
#if self.mode == 'cv':
#os.remove("weights_%s_%s.h5" %(self.target, self.mode))
return self