python - Tensorflow - Handling inputs to graph -
background:
i trying multiply tensor of size m tensor of size m. want result mxn tensor
for example:
1.0 0.2 0.2 0.5 0.0 x = 0.0 0.0 0.5 0.2 0.5 1.0 i can numpy:
x_vals = np.array([[1.0, 0.0, 1.0],[0.0, 0.0, 1.0], [0.0, 1.0, 1.0], [1.0, 1.0, 1.0]]) deltas = np.array([0.2, 0.5]) def mult(x): return x*deltas #i can this... x in x_vals: print mult(x.reshape(3,1)) i can't tensorflow?
x_vals = np.array([[1.0, 0.0, 1.0],[0.0, 0.0, 1.0], [0.0, 1.0, 1.0], [1.0, 1.0, 1.0]]) deltas = np.array([0.2, 0.5]) def mult(x): return x*deltas x = tf.placeholder('float', (none,3)) delta = tf.constant(deltas) result = mult(tf.reshape(x, shape=(3,1))) init = tf.global_variables_initializer() # create session , run graph tf.session() sess: sess.run(init) arr = sess.run(result, feed_dict={x: x_vals}) it seems passes in entire x_val array, , not sure looping on session.run each entry how supposed work. can give me pointer?
in numpy can simplified :x_vals[:,:,np.newaxis]*deltas[np.newaxis,:]
this should work in tensorflow well:
x_vals = np.array([[1.0, 0.0, 1.0],[0.0, 0.0, 1.0], [0.0, 1.0, 1.0], [1.0, 1.0, 1.0]]) deltas = np.array([0.2, 0.5]) def mult(x): return x[:,:,tf.newaxis]*deltas[tf.newaxis,:] x = tf.placeholder('float', (none,3)) delta = tf.constant(deltas) result = mult(x) init = tf.global_variables_initializer() # create session , run graph tf.session() sess: sess.run(init) arr = sess.run(result, feed_dict={x: x_vals})
Comments
Post a Comment