tf vs theano (simple example)

tf.py

import tensorflow as tf
import numpy as np

trX = np.linspace(-1, 1, 1001, dtype=np.float32)
trY = 2 * trX + np.random.randn(*trX.shape).astype(np.float32) * 0.33

EPOCHS = 1000
CAP = 1000
queue_xy = tf.train.input_producer(tf.pack([trX, trY], axis=1), shuffle=False, capacity=CAP, num_epochs=EPOCHS)

W = tf.Variable(0.0, name="weights", dtype=tf.float32)

opt = tf.train.GradientDescentOptimizer(0.01)
gs = tf.Variable(0)

def body(i):
    x,y = tf.unpack(queue_xy.dequeue())

    cost = tf.square(y - tf.mul(x,W))
    train_op = opt.minimize(cost, global_step=gs)
    return tf.tuple([tf.add(i,1)], control_inputs=[train_op])

loop = tf.while_loop(lambda _: True, body, [tf.constant(0)])

with tf.Session() as sess:
  tf.initialize_all_variables().run()
  tf.initialize_local_variables().run()

  coord = tf.train.Coordinator()
  threads = tf.train.start_queue_runners(sess=sess, coord=coord)
  i = 0
  try:
    while not coord.should_stop():
      sess.run(loop)

  except tf.errors.OutOfRangeError:
    print('Done training')
  finally:
    coord.request_stop()
  coord.join(threads)

  print sess.run(W)

theano.py

import theano
from theano import tensor as T
import numpy as np

trX = np.linspace(-1, 1, 1001)
trY = 2 * trX + np.random.randn(*trX.shape) * 0.33

X = T.scalar()
Y = T.scalar()

W = theano.shared(np.asarray(0., dtype=theano.config.floatX))
Y_p = X * W

cost = T.mean(T.sqr(Y - Y_p))
gradient = T.grad(cost=cost, wrt=W)
updates = [[W, W - gradient * 0.01]]

train = theano.function(inputs=[X, Y], outputs=cost, updates=updates, allow_input_downcast=True)

for i in range(1000):
  for x, y in zip(trX, trY):
    train(x, y)

print W.get_value()