#!/usr/bin/env luajit
+--[[
+
+ Copyright (c) 2016 Idiap Research Institute, http://www.idiap.ch/
+ Written by Francois Fleuret <francois.fleuret@idiap.ch>
+
+ This file is free software: you can redistribute it and/or modify
+ it under the terms of the GNU General Public License version 3 as
+ published by the Free Software Foundation.
+
+ It is distributed in the hope that it will be useful, but WITHOUT
+ ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+ or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public
+ License for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with this file. If not, see <http://www.gnu.org/licenses/>.
+
+]]--
+
require 'torch'
require 'nn'
-
require 'dagnn'
-function checkGrad(model, criterion, input, target)
+torch.setdefaulttensortype('torch.DoubleTensor')
+torch.manualSeed(1)
+
+function checkGrad(model, criterion, input, target, epsilon)
local params, gradParams = model:getParameters()
- local epsilon = 1e-5
+ local epsilon = epsilon or 1e-5
local output = model:forward(input)
local loss = criterion:forward(output, target)
model:backward(input, gradOutput)
local analyticalGradParam = gradParams:clone()
+ local err = 0
+
for i = 1, params:size(1) do
local x = params[i]
local ana = analyticalGradParam[i]
local num = (loss1 - loss0) / (2 * epsilon)
- local err = torch.abs(num - ana) / torch.abs(num)
- print(
- err .. ' checkGrad ' .. i
- .. ' analytical ' .. ana
- .. ' numerical ' .. num
- )
+ if num ~= ana then
+ err = math.max(err, math.abs(num - ana) / math.abs(num))
+ end
end
+ return err
end
function printTensorTable(t)
end
end
--- torch.setnumthreads(params.nbThreads)
-torch.setdefaulttensortype('torch.DoubleTensor')
-torch.manualSeed(2)
+-- +-- Linear(10, 10) --> ReLU --> d -->
+-- / /
+-- / /
+-- --> a --> b -----------> c ---------------+
+-- \
+-- \
+-- +--------------- e -->
--- +--> c ----> e --+
--- / / \
--- / / \
--- input --> a --> b ---> d ----+ g --> output
--- \ /
--- \ /
--- +--> f ---+
+dag = nn.DAG()
-a = nn.Linear(10, 10)
+a = nn.Linear(50, 10)
b = nn.ReLU()
-c = nn.Linear(10, 3)
-d = nn.Linear(10, 3)
-e = nn.CMulTable()
-f = nn.Linear(3, 3)
-g = nn.CAddTable()
-
-----------------------------------------------------------------------
-
-model = nn.DAG()
-
-model:addEdge(a, b)
-model:addEdge(b, c)
-model:addEdge(b, d)
-model:addEdge(c, e)
-model:addEdge(d, e)
-model:addEdge(d, f)
-model:addEdge(e, g)
-model:addEdge(f, g)
+c = nn.Linear(10, 15)
+d = nn.CMulTable()
+e = nn.Mul(-1)
-model:setInput(a)
-model:setOutput(g)
+dag:connect(a, b, c)
+dag:connect(b, nn.Linear(10, 15), nn.ReLU(), d)
+dag:connect(c, d)
+dag:connect(c, e)
-input = torch.Tensor(3, 10):uniform()
+dag:setInput(a)
+dag:setOutput({ d, e })
-print('******************************************************************')
-print('** updateOutput **************************************************')
-print('******************************************************************')
+-- Check the output of the dot file
+print('Writing /tmp/graph.dot')
+dag:saveDot('/tmp/graph.dot')
-output = model:updateOutput(input):clone()
+-- Let's make a model where the dag is inside another nn.Container.
+model = nn.Sequential()
+ :add(nn.Linear(50, 50))
+ :add(dag)
+ :add(nn.CAddTable())
-printTensorTable(output)
+criterion = nn.MSECriterion()
-print('******************************************************************')
-print('** updateGradInput ***********************************************')
-print('******************************************************************')
-
-gradInput = model:updateGradInput(input, output)
-
-printTensorTable(gradInput)
-
-print('******************************************************************')
-print('** checkGrad *****************************************************')
-print('******************************************************************')
+-- model:cuda()
+-- criterion:cuda()
+-- torch.setdefaulttensortype('torch.CudaTensor')
+-- epsilon = 1e-4
+local input = torch.Tensor(30, 50):uniform()
+local output = model:updateOutput(input):clone()
output:uniform()
-checkGrad(model, nn.MSECriterion(), input, output)
+-- Check that DAG:accGradParameters and friends work okay
+print('Gradient estimate error ' .. checkGrad(model, criterion, input, output, epsilon))
+
+-- Check that we can save and reload the model
+model:clearState()
+torch.save('/tmp/test.t7', model)
+local otherModel = torch.load('/tmp/test.t7')
+print('Gradient estimate error ' .. checkGrad(otherModel, criterion, input, output, epsilon))