Composes several augmentations together. Args: transforms (List[Transform]): list of transforms to compose. Example: >>> augme

Composes several augmentations together. Args: transforms (List[Transform]): list of transforms to compose. Example: >>> augme

Takes onnx Constant nodes and puts the tensor into graph initializers instead.

Fuses Add layer into parent convolution layer.

To handle errors and addition of custom layers

A Function specifies a layer, its parameters, and its inputs (which are Tops from other layers).

Crop Arguments: img (Image): the image being input during training boxes (Tensor): the original bounding boxes in pt form

Crop Arguments: img (Image): the image being input during training boxes (Tensor): the original bounding boxes in pt form

Transforms a tensorized image by swapping the channels in the order specified in the swap tuple. Args: swaps (int triple): final orde

Transforms a tensorized image by swapping the channels in the order specified in the swap tuple. Args: swaps (int triple): final orde

Expose hidden states of recurrent layers as model inputs and outputs

Fuses Add into BatchNorm

Fuses Mul into BatchNorm

Removes Dropout layer

Removes ImageScaler layer if connected to a model input and single parent child nodes

Applies a lambda as a transform.

Applies a lambda as a transform.

A Layers object is a pseudo-module which generates functions that specify layers; e.g., Layers().Convolution(bottom, kernel_size=3) will produce a

An abstract helper for merging nodes

Rename outputs according to mapping

Fuses 3 operators reshape->transpose->reshape which is equivalent to pytorch's pixel_shuffle layer

Crop Arguments: img (Image): the image being input during training boxes (Tensor): the original bounding boxes in pt form

Crop Arguments: img (Image): the image being input during training boxes (Tensor): the original bounding boxes in pt form

Fuses Reshape operator if it is used only to reshape blob in graph initializer. We can reshape here instead of runtime.