Geek Answers Handbook

Equivalent to Anano Dimshuffle on Google TensorFlow?

I have seen that transposing and changing together can help, but I don't know how to use it.

Eg. dimshuffle (0, 'x')

What is its equivalent with transpose and change? or is there a better way? Thanks.

+7
python numpy theano tensorflow
source share
3 answers

TensorFlow has three matching ops for implementing Theano dimshuffle :

  • tf.transpose() used to rearrange the dimensions of the tensor. If the pattern specified in the dimshuffle arguments is a permutation of the input tensor sizes (i.e., No 'x' or missing dimension), you can use tf.transpose() to implement dimshuffle() .

  • tf.expand_dims() used to add one or more dimensions of size-1 to the tensor. This handles the case where 'x' is specified as part of the dimshuffle() template, but does not reorder existing dimensions.

  • tf.squeeze() used to remove one or more dimensions of size 1 from the tensor. This handles the case when the dimension is not in the dimshuffle() template, but it does not change the order of the existing dimensions.

Assuming the input is a vector, your example ( dimshuffle(0, 'x') ) can only be expressed with tf.expand_dims() :

 input = tf.placeholder(tf.float32, [None]) # Defines an arbitrary-sized vector. result = tf.expand_dims(input, 1) print result.get_shape() # ==> TensorShape([Dimension(None), Dimension(1)])

Taking a more complex example, dimshuffle(1, 'x', 0) , applied to the matrix, would be:

 input = tf.placeholder(tf.float32, [128, 32]) # Defines a matrix. output = tf.expand_dims(tf.transpose(input, [1, 0]), 1) print output.get_shape() # ==> TensorShape([Dimension(32), Dimension(1), Dimension(128)])
+9
source share

I implemented a dimshuffle for TensorFlow in our Returnn framework ( here ). The code looks like this:

 def expand_multiple_dims(x, axes, name="expand_multiple_dims"): """ :param tf.Tensor x: :param list[int]|tuple[int] axes: after completion, tf.shape(y)[axis] == 1 for axis in axes :param str name: scope name :return: y where we have a new broadcast axis for each axis in axes :rtype: tf.Tensor """ with tf.name_scope(name): for i in sorted(axes): x = tf.expand_dims(x, axis=i, name="expand_axis_%i" % i) return x def dimshuffle(x, axes, name="dimshuffle"): """ Like Theanos dimshuffle. Combines tf.transpose, tf.expand_dims and tf.squeeze. :param tf.Tensor x: :param list[int|str]|tuple[int|str] axes: :param str name: scope name :rtype: tf.Tensor """ with tf.name_scope(name): assert all([i == "x" or isinstance(i, int) for i in axes]) real_axes = [i for i in axes if isinstance(i, int)] bc_axes = [i for (i, j) in enumerate(axes) if j == "x"] if x.get_shape().ndims is None: x_shape = tf.shape(x) x = tf.reshape(x, [x_shape[i] for i in range(max(real_axes) + 1)]) # will have static ndims assert x.get_shape().ndims is not None # First squeeze missing axes. i = 0 while i < x.get_shape().ndims: if i not in real_axes: x = tf.squeeze(x, axis=i) real_axes = [(j if (j < i) else (j - 1)) for j in real_axes] else: i += 1 # Now permute. assert list(sorted(real_axes)) == list(range(x.get_shape().ndims)) if real_axes != list(range(x.get_shape().ndims)): x = tf.transpose(x, real_axes) # Now add broadcast dimensions. if bc_axes: x = expand_multiple_dims(x, bc_axes) assert len(axes) == x.get_shape().ndims return x
0
source share

If the tensor flow is your backend

 from keras import baskend as K K.permute_dimension should do
0
source share

More articles:


All Articles