Dropout enabled, but it is not deterministic for bidirectional rnn. See PR #198

2 files changed, 7 insertions, 7 deletions

diff --git a/RNN.lua b/RNN.lua
index daf3b19..f22c9ef 100644
--- a/RNN.lua
+++ b/RNN.lua
@@ -44,7 +44,7 @@ function RNN:reset(stdv)
    errcheck('cudnnGetRNNParamsSize',
             cudnn.getHandle(),
             self.rnnDesc[0],
-            self.xDescs[0],	    
+            self.xDescs[0],
             weightSize:data(),
 	    self.datatype)
    weightSize[1] = (weightSize[1] + 3) / 4 -- sizeof(float)
@@ -117,7 +117,7 @@ end
 function RNN:resetRNNDescriptor()
    if not self.rnnDesc then
       self.rnnDesc = self:createRNNDescriptors(1)
-   end   
+   end
    errcheck('cudnnSetRNNDescriptor',
             self.rnnDesc[0],
             self.hiddenSize,
@@ -236,7 +236,7 @@ function RNN:updateOutput(input)
         input = input:transpose(1, 2)
     end
    assert(input:dim() == 3, 'input must have 3 dimensions: seqLength, miniBatch, inputSize')
-   assert(self.dropout == 0, 'dropout currently not supported')
+   -- assert(self.dropout == 0, 'dropout currently not supported')
    -- Decide which descriptors/tensors need to be updated.
    local resetRNN = not self.dropoutDesc or not self.rnnDesc
    local resetIO = not self.xDescs or not self.yDescs
@@ -364,7 +364,7 @@ function RNN:updateGradInput(input, gradOutput)
         gradOutput = gradOutput:transpose(1, 2)
         self.output = self.output:transpose(1, 2)
     end
-   assert(self.dropout == 0, 'dropout currently not supported')
+   -- assert(self.dropout == 0, 'dropout currently not supported')
    assert(input:dim() == 3, 'input should have 3 dimensions: seqLength, miniBatch, inputSize')
    assert(input:size(1) == self.seqLength, 'input has incorrect sequence length!')
    assert(input:size(2) == self.miniBatch, 'input has incorrect minibatch size!')
@@ -448,7 +448,7 @@ function RNN:accGradParameters(input, gradOutput, scale)
     end
    scale = scale or 1
    if scale == 0 then return end
-   assert(self.dropout == 0, 'dropout currently not supported')
+   -- assert(self.dropout == 0, 'dropout currently not supported')
    assert(input:dim() == 3, 'input should have 3 dimensions: seqLength, miniBatch, inputSize')
    assert(input:size(1) == self.seqLength, 'input has incorrect sequence length!')
    assert(input:size(2) == self.miniBatch, 'input has incorrect minibatch size!')
diff --git a/cudnn-scm-1.rockspec b/cudnn-scm-1.rockspec
index bc36117..463eb20 100644
--- a/cudnn-scm-1.rockspec
+++ b/cudnn-scm-1.rockspec
@@ -2,13 +2,13 @@ package = "cudnn"
 version = "scm-1"
 
 source = {
-   url = "git://github.com/soumith/cudnn.torch.git",
+   url = "git://github.com/jpuigcerver/cudnn.torch.git",
 }
 
 description = {
    summary = "Torch7 FFI bindings for NVIDIA CuDNN kernels!",
    detailed = [[
-   All CuDNN modules exposed as nn.Module derivatives so 
+   All CuDNN modules exposed as nn.Module derivatives so
    that they can be used with torch's neural network package
    ]],
    homepage = "https://github.com/soumith/cudnn.torch",