def main():
options = {
'learning_rate': 0.1,
'beta1': 0.9,
'optimizer': 'gd',
'loss': 'crossentropy'
}
train_x, test_x, train_set_x_orig, train_set_y_orig, test_set_x_orig, test_set_y_orig, classes = load_data(
)
X = np.array([[1, 2], [1, 2], [4, 2]])
Y = np.array([[0], [0], [0]])
print(train_x.shape)
print(test_x.shape)
print(train_set_y_orig.shape)
print(train_set_y_orig[0, 0:10])
layers = [
Dense(32, activation='relu'),
Dense(5, activation='relu'),
Dense(1, activation='sigmoid')
]
print(len(layers))
dnn = DNN(train_x, train_set_y_orig, layers, options)
print(dnn.params.keys())
#for param in sorted(dnn.params):
# print(param, dnn.params[param].shape)
print(dnn)
print(dnn.loss(dnn.predict(test_x), test_set_y_orig))
dnn.train()
def train(data, path):
os.system("mkdir -p " + path)
x_train, y_train = data['x'], data['y']
#pdb.set_trace()
x = tf.placeholder(tf.float32, [None, 11, input_dim], name='x')
y = tf.placeholder(tf.float32, [None, input_dim], name='y')
model = DNN()
loss = model.loss(x, y)
pred = model.predict(x)
tf.add_to_collection('pred', pred)
tf.add_to_collection('loss', loss)
optimize = tf.train.AdamOptimizer(learning_rate=5e-5,
beta1=0.9,
beta2=0.999).minimize(loss)
merged = tf.summary.merge_all()
with tf.Session() as sess:
start = time.time()
init = tf.global_variables_initializer()
sess.run(init)
saver = tf.train.Saver()
writer = tf.summary.FileWriter(path + 'logs', sess.graph)
err_old = 100
for i in range(iter_num):
err_new = 0
count = 0
for j in range(len(x_train)):
idx = random.randint(0, len(x_train) - 1)
#pdb.set_trace()
xt = next_batch(x_train[idx])
yt = y_train[idx % len(y_train)][5:-5, :]
k = 0
for k in range(0, len(xt), batch_num):
xb = xt[k:k + batch_num]
yb = yt[k:k + batch_num]
_ = sess.run([optimize], feed_dict={x: xb, y: yb})
xb = xt[k:]
yb = yt[k:]
err, result = sess.run([loss, merged],
feed_dict={
x: xb,
y: yb
})
err_new += err
count += 1
if j % 100 == 0:
writer.add_summary(result, len(x_train) * i + j)
if err_new / count < err_old:
err_old = err_new / count
saver.save(sess, path + 'test_best_model')
#print('Epoch [%4d] Iter [%4d] Time [%5.4f] \nLoss: [%.4f]' %
# (i+1, j, time.time() - start, err))
print('Epoch [%4d] Time [%10.4f] Loss: [%.4f]: Saved ' %
(i + 1, time.time() - start, err_new / count))
else:
print('Epoch [%4d] Time [%5.4f] Loss: [%.4f]: No save ' %
(i + 1, time.time() - start, err_new / count))