tf.contrib.kfac.fisher_blocks.num_conv_locations

tf.contrib.kfac.fisher_blocks.num_conv_locations(
    input_shape,
    strides
)

Defined in tensorflow/contrib/kfac/python/ops/fisher_blocks.py.

Returns the number of spatial locations a 2D Conv kernel is applied to.

Args:

  • input_shape: List of ints representing shape of inputs to tf.nn.convolution().
  • strides: List of ints representing strides along spatial dimensions as passed in to tf.nn.convolution().

Returns:

A scalar |T| denoting the number of spatial locations for the Conv layer.