def create_model():
if args.model == 'myVGG':
model = myMethod.create_myVGG()
else:
model = myMethod.create_myModel()
model.compile(optimizer='adam',
loss='categorical_crossentropy',
metrics=["accuracy"])
load_path = args.load_path
model.load_weights(load_path)
model.summary()
return model
def singleGPU():
if args.model == 'myVGG':
model = myMethod.create_myVGG()
else:
model = myMethod.create_myModel()
model.compile(optimizer='adam',
loss='categorical_crossentropy',
metrics=["accuracy"])
model.summary()
model.fit(train_data, epochs=TOTAL_EPOCHS, steps_per_epoch=TOTAL_TRAIN // BATCH_SIZE_TRAIN,
callbacks=[tensorboard_callback, cp_callback],
validation_data=public_test_data,
validation_steps=TOTAL_TEST // BATCH_SIZE_TRAIN)
model.evaluate(private_test_data, steps=TOTAL_TEST // BATCH_SIZE_TRAIN)
def multiGPUs():
GPUS = args.gpus
strategy = tf.distribute.MirroredStrategy()
with strategy.scope():
if args.model == 'myVGG':
model = myMethod.create_myVGG()
else:
model = myMethod.create_myModel()
model.compile(optimizer='adam',
loss='categorical_crossentropy',
metrics=["accuracy"])
model.fit(train_data, epochs=TOTAL_EPOCHS, steps_per_epoch=TOTAL_TRAIN // BATCH_SIZE_TRAIN // GPUS,
callbacks=[tensorboard_callback, cp_callback],
validation_data=public_test_data,
validation_steps=TOTAL_TEST // BATCH_SIZE_TRAIN // GPUS)
model.evaluate(private_test_data, steps=TOTAL_TEST // BATCH_SIZE_TRAIN // GPUS)
private_test_data = private_test_data.map(myMethod.preprocess_DAtestdata)
# get standard result
correct_answer = np.loadtxt(test_private_path,
dtype=np.int,
delimiter=',',
skiprows=1,
usecols=(0),
encoding='utf-8')
# correct_answer = correct_answer.repeat(10)
if args.model == 'myVGG':
model = myMethod.create_myVGG()
else:
model = myMethod.create_myModel()
model.compile(optimizer='adam',
loss='categorical_crossentropy',
metrics=["accuracy"])
def get_acc_predict(load_path):
model.load_weights(load_path)
x = model.predict(private_test_data,
steps=TOTAL_TEST // BATCH_SIZE_TEST_DA)
predict_result = np.zeros(shape=(3589, 7))
for i in range(0, 3589):
sum = np.zeros(shape=(1, 7))
for j in range(0, 10):
sum += x[10 * i + j]