dragon

Classes

class EagerTensor : Tensor abstraction for eager executing.

class GradientTape : Record the operations for auto differentiation.

class Tensor : Tensor abstraction for graph executing.

class Workspace : Sandbox to isolate the resources and computations.

Functions

assign(…) : Assign the value to input.

broadcast_to(…) : Broadcast input according to a given shape.

cast(…) : Cast the data type of input.

channel_affine(…) : Apply affine transformation along the channels.

channel_normalize(…) : Normalize channels with mean and standard deviation.

channel_shuffle(…) : Shuffle channels between a given number of groups.

concat(…) : Concatenate the inputs along the given axis.

constant(…) : Return a tensor initialized from the value.

copy(…) : Copy the input.

create_function(…) : Create a callable graph from the specified outputs.

device(…) : Context-manager to nest the the device spec.

eager_mode(…) : Context-manager set the eager execution mode.

eager_scope(…) : Context-manager to nest the name for eager resources.

expand_dims(…) : Expand the dimensions of input with size 1.

eye(…) : Return a tensor constructed as the identity matrix.

eye_like(…) :Return a tensor of identity matrix with shape as the other.

fill(…) : Return a tensor filled with the scalar value.

flatten(…) : Flatten the input along the given axes.

function(…) : Compile a function and return an executable.

get_workspace(…) : Return the current default workspace.

gradients(…) : Compute the symbolic derivatives of ys w.r.t. xs .

graph_mode(…) : Context-manager set the graph execution mode.

index_select(…) : Select the elements according to the index along the given axis.

load_library(…) : Load a shared library.

masked_assign(…) : Assign the value to input where mask is 1.

masked_select(…) : Select the elements of input where mask is 1.

name_scope(…) : Context-manager to nest the name as prefix for operations.

nonzero(…) : Return the index of non-zero elements.

ones(…) : Return a tensor filled with ones.

ones_like(…) : Return a tensor of ones with shape as the other.

one_hot(…) : Return the one-hot representation for input.

pad(…) : Pad the input according to the given sizes.

python_plugin(…) : Create a plugin operator from the python class.

range(…) : Return a tensor of evenly spaced values within a interval.

repeat(…) : Repeat the elements along the given axis.

reset_workspace(…) : Reset the current default workspace.

reshape(…) : Change the dimensions of input.

shape(…) : Return the shape of input.

slice(…) : Select the elements according to the given sections.

split(…) : Split the input into chunks along the given axis.

squeeze(…) : Remove the dimensions of input with size 1.

stack(…) : Stack the inputs along the given axis.

stop_gradient(…) : Return the identity of input with truncated gradient-flow.

tile(…) : Tile the input according to the given repeats.

transpose(…) : Permute the dimensions of input.

unique(…) : Return the unique elements of input.

where(…) : Select the elements from two branches under the condition.

zeros(…) : Return a tensor filled with zeros.

zeros_like(…) : Return a tensor of zeros with shape as the other.