lunny/keras
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix issue in Theano scan for recurrent layers

Browse Source
This commit is contained in:
fchollet 2015-05-26 13:50:54 -07:00
parent b5fb00d753
commit 5e2b9358d6
1 changed files with 5 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
keras/layers/recurrent.py
View File

@ -60,7 +60,7 @@ class SimpleRNN(Layer):
self._step, # this will be called with arguments (sequences[i], outputs[i-1], non_sequences[i])
sequences=x, # tensors to iterate over, inputs to _step
# initialization of the output. Input to _step with default tap=-1.
outputs_info=alloc_zeros_matrix(X.shape[1], self.output_dim),
outputs_info=T.unbroadcast(alloc_zeros_matrix(X.shape[1], self.output_dim), 1),
non_sequences=self.U, # static inputs to _step
truncate_gradient=self.truncate_gradient
)
@ -233,7 +233,7 @@ class GRU(Layer):
outputs, updates = theano.scan(
self._step,
sequences=[x_z, x_r, x_h],
outputs_info=alloc_zeros_matrix(X.shape[1], self.output_dim),
outputs_info=T.unbroadcast(alloc_zeros_matrix(X.shape[1], self.output_dim), 1),
non_sequences=[self.U_z, self.U_r, self.U_h],
truncate_gradient=self.truncate_gradient
)
@ -346,8 +346,8 @@ class LSTM(Layer):
self._step,
sequences=[xi, xf, xo, xc],
outputs_info=[
alloc_zeros_matrix(X.shape[1], self.output_dim),
alloc_zeros_matrix(X.shape[1], self.output_dim)
T.unbroadcast(alloc_zeros_matrix(X.shape[1], self.output_dim), 1),
T.unbroadcast(alloc_zeros_matrix(X.shape[1], self.output_dim), 1)
],
non_sequences=[self.U_i, self.U_f, self.U_o, self.U_c],
truncate_gradient=self.truncate_gradient