grain

library structure

Modules

autograd
module grain.autograd

A module for a variable used as a node in autograd computation graph

chain
module grain.chain

Chain means autograd operators in grain that is equivalent to - pytorch: torch.nn.Module - chainer: chainer.Chain or chainer.Link

config
module grain.config

global configurations

cublas
module grain.cublas

cuBLAS wrappers

cuda
module grain.cuda

CUDA wrapper module

cudnn
module grain.cudnn

cuDNN high level wrapper for grain.autograd.Variable

dataset
module grain.dataset

dataset module TODO: define abstract class/interface for dataset

functions
module grain.functions

A type defined in grain.functions is equivalent to chainer.Function or torch.autograd.Function. Function is a set of core autograd algorithms for grain.chain that compose some functions in computation graph.

hdf5
module grain.hdf5

HDF5 bindings

metric
module grain.metric

Metric (e.g., accuracy)

optim
module grain.optim

A module for gradient descent optimizer

serializer
module grain.serializer

A module for serialization. HDF5 read/write is supported

testing
module grain.testing

A module for testing

utility
module grain.utility

A module of utility functions

warpctc
module grain.warpctc
Undocumented in source.

Public Imports

grain.autograd
public import grain.autograd;
grain.chain
public import grain.chain;
grain.config
public import grain.config;
grain.serializer
public import grain.serializer;
grain.optim
public import grain.optim;
grain.metric
public import grain.metric;
grain.cuda
public import grain.cuda;

Meta