我试图在这里修改基本教程模型（http://rnduja.github.io/2015/10/13/torch-mnist/ 并在火炬上制作更深层次的模型。这是我的代码（当我在上面的url中尝试相同的模型时代码工作）：
加载数据
require 'torch'
require 'nn'
require 'optim'
mnist = require 'mnist'
fullset = mnist.traindataset()
testset = mnist.testdataset()
trainset = {
size = 50000,
data = fullset.data[{{1,50000}}]:double(),
label = fullset.label[{{1,50000}}]
}
validationset = {
size = 10000,
data = fullset.data[{{50001,60000}}]:double(),
label = fullset.label[{{50001,60000}}]
}
定义网络
model = nn.Sequential()
model:add(nn.Reshape(28*28))
model:add(nn.SpatialConvolution(1, 32, 3, 3))
model:add(nn.ReLU())
model:add(nn.SpatialConvolution(32, 32, 3, 3))
model:add(nn.ReLU())
model:add(nn.SpatialMaxPooling(2, 2, 2, 2))
-- model:add(nn.SpatialDropout(0.25)) -- Ignored for simplicity.
-- model:add(nn.Reshape(12*12*32))
model:add(nn.View(12*12*32))
model:add(nn.Linear(12*12*32, 128))
model:add(nn.ReLU())
-- model:add(nn.SpatialDropout(0.5)) -- Ignored for simplicity.
model:add(nn.Linear(128, 10))
model:add(nn.SoftMax())
criterion = nn.ClassNLLCriterion()
定义下降算法
sgd_params = {
learningRate = 1e-2,
learningRateDecay = 1e-4,
weightDecay = 1e-3,
momentum = 1e-4
}
x, dl_dx = model:getParameters()
step = function(batch_size)
local current_loss = 0
local count = 0
local shuffle = torch.randperm(trainset.size)
batch_size = batch_size or 200
for t = 1,trainset.size,batch_size do
-- setup inputs and targets for this mini-batch
local size = math.min(t + batch_size - 1, trainset.size) - t
local inputs = torch.Tensor(size, 28, 28)
local targets = torch.Tensor(size)
for i = 1,size do
local input = trainset.data[shuffle[i+t]]
local target = trainset.label[shuffle[i+t]]
-- if target == 0 then target = 10 end
inputs[i] = input
targets[i] = target
end
targets:add(1)
local feval = function(x_new)
-- reset data
if x ~= x_new then x:copy(x_new) end
dl_dx:zero()
-- perform mini-batch gradient descent
local loss = criterion:forward(model:forward(inputs), targets)
model:backward(inputs, criterion:backward(model.output, targets))
return loss, dl_dx
end
_, fs = optim.sgd(feval, x, sgd_params)
-- fs is a table containing value of the loss function
-- (just 1 value for the SGD optimization)
count = count + 1
current_loss = current_loss + fs[1]
end
-- normalize loss
return current_loss / count
end
eval = function(dataset, batch_size)
local count = 0
batch_size = batch_size or 200
for i = 1,dataset.size,batch_size do
local size = math.min(i + batch_size - 1, dataset.size) - i
local inputs = dataset.data[{{i,i+size-1}}]
local targets = dataset.label[{{i,i+size-1}}]:long()
local outputs = model:forward(inputs)
local _, indices = torch.max(outputs, 2)
indices:add(-1)
local guessed_right = indices:eq(targets):sum()
count = count + guessed_right
end
return count / dataset.size
end
列车
max_iters = 30
do
local last_accuracy = 0
local decreasing = 0
local threshold = 1 -- how many deacreasing epochs we allow
for i = 1,max_iters do
local loss = step()
print(string.format('Epoch: %d Current loss: %4f', i, loss))
local accuracy = eval(validationset)
print(string.format('Accuracy on the validation set: %4f', accuracy))
if accuracy < last_accuracy then
if decreasing > threshold then break end
decreasing = decreasing + 1
else
decreasing = 0
end
last_accuracy = accuracy
end
end
现在我在运行Train部分时收到以下错误消息。
...s/username/torch/install/share/lua/5.1/nn/Container.lua:67:
In 2 module of nn.Sequential:
/Users/username/torch/install/share/lua/5.1/nn/THNN.lua:110: bad argument #2 to 'v' (3D or 4D input tensor expected but got: [199 x 784] at /tmp/luarocks_nn-scm-1-2325/nn/lib/THNN/generic/SpatialConvolutionMM.c:33)
stack traceback:
[C]: in function 'v'
/Users/username/torch/install/share/lua/5.1/nn/THNN.lua:110: in function 'SpatialConvolutionMM_updateOutput'
...go/torch/install/share/lua/5.1/nn/SpatialConvolution.lua:79: in function <...go/torch/install/share/lua/5.1/nn/SpatialConvolution.lua:76>
[C]: in function 'xpcall'
...s/username/torch/install/share/lua/5.1/nn/Container.lua:63: in function 'rethrowErrors'
.../username/torch/install/share/lua/5.1/nn/Sequential.lua:44: in function 'forward'
[string "step = function(batch_size)..."]:27: in function 'opfunc'
/Users/username/torch/install/share/lua/5.1/optim/sgd.lua:44: in function 'sgd'
[string "step = function(batch_size)..."]:33: in function 'step'
[string "do..."]:6: in main chunk
[C]: in function 'xpcall'
...rs/username/torch/install/share/lua/5.1/itorch/main.lua:210: in function <...rs/username/torch/install/share/lua/5.1/itorch/main.lua:174>
...rs/username/torch/install/share/lua/5.1/lzmq/poller.lua:75: in function 'poll'
...username/torch/install/share/lua/5.1/lzmq/impl/loop.lua:307: in function 'poll'
...username/torch/install/share/lua/5.1/lzmq/impl/loop.lua:325: in function 'sleep_ex'
...username/torch/install/share/lua/5.1/lzmq/impl/loop.lua:370: in function 'start'
...rs/username/torch/install/share/lua/5.1/itorch/main.lua:389: in main chunk
[C]: in function 'require'
(command line):1: in main chunk
[C]: at 0x0100c03350
WARNING: If you see a stack trace below, it doesn't point to the place where this error occurred. Please use only the one above.
stack traceback:
[C]: in function 'error'
...s/username/torch/install/share/lua/5.1/nn/Container.lua:67: in function 'rethrowErrors'
.../username/torch/install/share/lua/5.1/nn/Sequential.lua:44: in function 'forward'
[string "step = function(batch_size)..."]:27: in function 'opfunc'
/Users/username/torch/install/share/lua/5.1/optim/sgd.lua:44: in function 'sgd'
[string "step = function(batch_size)..."]:33: in function 'step'
[string "do..."]:6: in main chunk
[C]: in function 'xpcall'
...rs/username/torch/install/share/lua/5.1/itorch/main.lua:210: in function <...rs/username/torch/install/share/lua/5.1/itorch/main.lua:174>
...rs/username/torch/install/share/lua/5.1/lzmq/poller.lua:75: in function 'poll'
...username/torch/install/share/lua/5.1/lzmq/impl/loop.lua:307: in function 'poll'
...username/torch/install/share/lua/5.1/lzmq/impl/loop.lua:325: in function 'sleep_ex'
...username/torch/install/share/lua/5.1/lzmq/impl/loop.lua:370: in function 'start'
...rs/username/torch/install/share/lua/5.1/itorch/main.lua:389: in main chunk
[C]: in function 'require'
(command line):1: in main chunk
[C]: at 0x0100c03350
我尝试根据此处的解决方案（Torch mnist simple）进行修复，在输入数据上添加:view(1, 28, 28)或从inputs = torch.Tensor(size, 28, 28)更改为inputs = torch.Tensor(size, 1, 28, 28)，但是，没有解决问题。
我不明白我还能尝试什么，而且我还不知道如何调试这种类型的错误。谢谢你的帮助。
答案 0 :(得分：0)
在模型中发现错误，不需要这句话：model:add(nn.Reshape(28*28))。在注释掉之后，又出现了另一个错误。
……./torch/install/share/lua/5.1/nn/THNN.lua:110: Need input of dimension 3 and input.size[0] == 1 but got input to be of shape: [199 x 28 x 28] at /tmp/luarocks_nn-scm-1-2325/nn/lib/THNN/generic/SpatialConvolutionMM.c:47
然后，发现Torch的目标不接受0进行分类，因此需要10来代替。以下有帮助。