Loss

This script includes Module class for python users to use Computational Graph in their model.

class singa.module.Module

Bases: object

Base class for your neural network modules.

Example usage:

import numpy as np
from singa import opt
from singa import tensor
from singa import device
from singa import autograd
from singa.module import Module

class Model(Module):
    def __init__(self):
        super(Model, self).__init__()

        self.conv1 = autograd.Conv2d(1, 20, 5, padding=0)
        self.conv2 = autograd.Conv2d(20, 50, 5, padding=0)

        self.sgd = opt.SGD(lr=0.01)

    def forward(self, x):
        y = self.conv1(x)
        y = self.conv2(y)
        return y

    def loss(self, out, y):
        return autograd.softmax_cross_entropy(out, y)

    def optim(self, loss):
        self.sgd.backward_and_update(loss)

eval()

Sets the module in evaluation mode.

forward(*input)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Parameters

*input – the input training data for the module

Returns

the outputs of the forward propagation.

Return type

out

graph(mode=True, sequential=False)

Turn on the computational graph. Specify execution mode.

loss(*args, **kwargs)

Defines the loss function performed when training the module.

on_device(device)

Sets the target device.

The following training will be performed on that device.

Parameters

device (Device) – the target device

optim(*args, **kwargs)

Defines the optim function for backward pass.

train(mode=True)

Set the module in evaluation mode.