近期使用torch写神经网络的框架，初次接触，踩了不少坑，在这里总结一下。
tensor type
torch中tensor是有类型的，默认是DoubleTensor：
th>torch.getdefaulttensortype()
torch.DoubleTensor除此之外，还有FloatTensor, CudaTensor, IntTensor, CharTensor等，可以使用setdefaulttensortype设置缺省的tensortype，通常情况下，如果使用cuda进行网络的训练，就可以将缺省的Tensor Type设置为CudaTensor。但是这里有一个坑，就是CudaTensor相比其他的Tensor缺少一些成员函数（我们暂且叫成员函数），比如image。
th> a=torch.FloatTensor():resize(sample_temp:size()):copy(sample_temp)
[0.0001s]
th> a.i
a.image. a.indexCopy( a.isContiguous( a.isSize(
a.index( a.indexFill( a.isSameSizeAs(
a.indexAdd( a.int( a.isSetTo(
th> a=torch.Tensor():resize(sample_temp:size()):copy(sample_temp)
[0.0003s]
th> a.i
a.index( a.indexFill( a.isContiguous( a.isSize(
a.indexAdd( a.int( a.isSameSizeAs(
a.indexCopy( a.inverse( a.isSetTo(
th> a=torch.FloatTensor():resize(sample_temp:size()):copy(sample_temp)
[0.0001s] 在defaulttensortype是CudaTensor时调用image.save函数会出现如下错误：
th> image.save('1.jpg',sample_temp);
.../lijiguo/code/torch/install/share/lua/5.1/image/init.lua:166: attempt to index field 'image' (a nil value)
stack traceback:
.../lijiguo/code/torch/install/share/lua/5.1/image/init.lua:166: in function 'clampImage'
.../lijiguo/code/torch/install/share/lua/5.1/image/init.lua:267: in function 'saver'
.../lijiguo/code/torch/install/share/lua/5.1/image/init.lua:457: in function 'save'
[string "image.save('1.jpg',sample_temp);"]:1: in main chunk
[C]: in function 'xpcall'
.../lijiguo/code/torch/install/share/lua/5.1/trepl/init.lua:679: in function 'repl'
...code/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:204: in main chunk
[C]: at 0x00406620
[0.0004s] 还有就是network输入的TensorType应该与network的TensorType一致，如果network是在GPU上的，那么他就是CudaTensor，这时候应该讲input vector设置为CudaTensor，否则会出现以下错误（github上也有人遇到这个问题：https://github.com/torch/torch7/issues/1087）
th> input_vector=torch.FloatTensor(1,100);
[0.0000s]
th> sample=model.G:forward(input_vector)
...ijiguo/code/torch/install/share/lua/5.1/nn/Container.lua:67:
In 1 module of nn.Sequential:
...1/lijiguo/code/torch/install/share/lua/5.1/nn/Linear.lua:66: invalid arguments: CudaTensor number CudaTensor number FloatTensor CudaTensor
expected arguments: *CudaTensor~2D* [CudaTensor~2D] [float] CudaTensor~2D CudaTensor~2D | *CudaTensor~2D* float [CudaTensor~2D] float CudaTensor~2D CudaTensor~2D
stack traceback:
[C]: in function 'addmm'
...1/lijiguo/code/torch/install/share/lua/5.1/nn/Linear.lua:66: in function <...1/lijiguo/code/torch/install/share/lua/5.1/nn/Linear.lua:53>
[C]: in function 'xpcall'
...ijiguo/code/torch/install/share/lua/5.1/nn/Container.lua:63: in function 'rethrowErrors'
...jiguo/code/torch/install/share/lua/5.1/nn/Sequential.lua:44: in function 'forward'
[string "sample=model.G:forward(input_vector)"]:1: in main chunk
[C]: in function 'xpcall'
.../lijiguo/code/torch/install/share/lua/5.1/trepl/init.lua:679: in function 'repl'
...code/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:204: in main chunk
[C]: at 0x00406620
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'
...ijiguo/code/torch/install/share/lua/5.1/nn/Container.lua:67: in function 'rethrowErrors'
...jiguo/code/torch/install/share/lua/5.1/nn/Sequential.lua:44: in function 'forward'
[string "sample=model.G:forward(input_vector)"]:1: in main chunk
[C]: in function 'xpcall'
.../lijiguo/code/torch/install/share/lua/5.1/trepl/init.lua:679: in function 'repl'
...code/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:204: in main chunk
[C]: at 0x00406620
[0.0009s] lua table index
lua的table索引从1开始，切记切记切记！！！
否则将出现以下错误：
th> image.save('1.jpg',sample[0]);
bad argument #2 to '?' (out of range)
stack traceback:
[C]: at 0x7f1d34138bd0
[C]: in function '__index'
[string "image.save('1.jpg',sample[0]);"]:1: in main chunk
[C]: in function 'xpcall'
.../lijiguo/code/torch/install/share/lua/5.1/trepl/init.lua:679: in function 'repl'
...code/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:204: in main chunk
[C]: at 0x00406620
[0.0001s]