numpy and tensor data into each other:
*Numpy2Tensor
Although TensorFlow network will be automatically converted when the input data for the Tensor Numpy to deal with, but we can also go to their own explicit conversion:
data_tensor = tf.convert_to_tensor (data_numpy )
*Tensor2Numpy
Network output result is still Tensor, when we use these results to perform the operation can only be performed by the Numpy data will appear inexplicable error. Solution:
with tf.Session() as sess:
data_numpy = data_tensor.eval()