dontBackward
montyhall opened this issue · comments
peyman faratin commented
Running a dontBackward()
on FB's Resnet-18 model, from layers 8 and below using
model.modules[1]:get(i):dontBackward()
results in the following error when I run model:backward(inputs,dE_dy)
In 1 module of nn.Sequential:
In 2 module of nn.Sequential:
/home/faratin/torch/install/share/lua/5.1/cudnn/init.lua:118: Error in CuDNN: CUDNN_STATUS_BAD_PARAM (cudnnBatchNormalizationBackward)
stack traceback:
[C]: in function 'error'
/home/faratin/torch/install/share/lua/5.1/cudnn/init.lua:118: in function 'errcheck'
...torch/install/share/lua/5.1/cudnn/BatchNormalization.lua:95: in function <...torch/install/share/lua/5.1/cudnn/BatchNormalization.lua:83>
[C]: in function 'xpcall'
/home/faratin/torch/install/share/lua/5.1/nn/Container.lua:63: in function 'rethrowErrors'
/home/faratin/torch/install/share/lua/5.1/nn/Sequential.lua:84: in function </home/faratin/torch/install/share/lua/5.1/nn/Sequential.lua:78>
[C]: in function 'xpcall'
/home/faratin/torch/install/share/lua/5.1/nn/Container.lua:63: in function 'rethrowErrors'
/home/faratin/torch/install/share/lua/5.1/nn/Sequential.lua:88: in function 'backward'
dev.lua:125: in function 'dryRun'
[string "_RESULT={dryRun(model,loss,10)}"]:1: in main chunk
[C]: in function 'xpcall'
/home/faratin/torch/install/share/lua/5.1/trepl/init.lua:652: in function 'repl'
...atin/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:199: in main chunk
[C]: at 0x004064f0
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'
/home/faratin/torch/install/share/lua/5.1/nn/Container.lua:67: in function 'rethrowErrors'
/home/faratin/torch/install/share/lua/5.1/nn/Sequential.lua:88: in function 'backward'
dev.lua:125: in function 'dryRun'
[string "_RESULT={dryRun(model,loss,10)}"]:1: in main chunk
[C]: in function 'xpcall'
/home/faratin/torch/install/share/lua/5.1/trepl/init.lua:652: in function 'repl'
...atin/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:199: in main chunk
[C]: at 0x004064f0```
If I do not run `dontBackward()` the model:backward() works fine