2-D convolution. Args: inputs: A 4-D tensor of shape `[batch, in_height, in_width, in_channels]`. kernel: A 4-D

Evaluate the ops with the given input. When this function is called, the default session will have the graph defined by a previous cal

Like concat except that first tiles the broadcastable dimensions if necessary

Pads a 4-D tensor according to the convolution padding algorithm. Convolution with a padding scheme conv2d(..., padding=padding)

Args: image: A 4-D tensor of shape `[batch, in_height, in_width, in_channels]`. kernels: A 4-D of shape `

Args: image: A 4-D tensor of shape `[batch, in_height, in_width, in_channels]`. kernels: A 4-D or 6-D tensor of shape

Caffe-style flatten. Args: inputs: An N-D tensor. axis: The first axis to flatten: all preceding axes are retained in the ou

Structural SIMilarity (SSIM) index between two images Args: X: A tensor of shape `[..., in_height, in_width, in_channels]`.

2-D transposed convolution. Notes on padding: The equivalent of transposed convolution with full padding is a convolution with valid

Computes the paddings for a 4-D tensor according to the convolution padding algorithm. See pad2d. Reference: https://www.tensor

swap time-batch major :param ten: npy tenosr :return:

:param metadata: :return: train_metadata, val_metadata

Leaky ReLU activation function Reference: https://github.com/tensorflow/tensorflow/blob/master/tensorflow/python/ops/nn_ops.py

Grabs and (optionally) modifies hparams

Grabs and (optionally) modifies hparams