[GitHub] piiswrong closed pull request #10982: fix rnn layer kernel forward

[GitBox <gi...@apache.org>]

piiswrong closed pull request #10982: fix rnn layer kernel forward
URL: https://github.com/apache/incubator-mxnet/pull/10982
 
 
   

This is a PR merged from a forked repository.
As GitHub hides the original diff on merge, it is displayed below for
the sake of provenance:

As this is a foreign pull request (from a fork), the diff is supplied
below (as it won't show otherwise due to GitHub magic):

diff --git a/python/mxnet/gluon/rnn/rnn_layer.py b/python/mxnet/gluon/rnn/rnn_layer.py
index 89224cf6f9b..2beae96f949 100644
--- a/python/mxnet/gluon/rnn/rnn_layer.py
+++ b/python/mxnet/gluon/rnn/rnn_layer.py
@@ -23,7 +23,7 @@
 from __future__ import print_function
 __all__ = ['RNN', 'LSTM', 'GRU']
 
-from ... import ndarray, autograd
+from ... import ndarray
 from .. import Block
 from . import rnn_cell
 
@@ -186,7 +186,7 @@ def forward(self, inputs, states=None):
                 self.i2h_weight[i].shape = (self._gates*self._hidden_size, inputs.shape[2])
                 self.i2h_weight[i]._finish_deferred_init()
         if inputs.context.device_type == 'gpu' or \
-           self._mode == 'lstm' and not (self._dropout and autograd.is_training()):
+           self._mode == 'lstm' and not self._dropout:
             out = self._forward_kernel(inputs, states)
         else:
             out = self._forward(inputs, states)
diff --git a/tests/python/unittest/test_gluon_rnn.py b/tests/python/unittest/test_gluon_rnn.py
index 24d5a932d7b..9dbcb3b3be8 100644
--- a/tests/python/unittest/test_gluon_rnn.py
+++ b/tests/python/unittest/test_gluon_rnn.py
@@ -268,6 +268,8 @@ def check_rnn_layer_forward(layer, inputs, states=None, run_only=False):
             assert isinstance(out, mx.nd.NDArray)
         out.backward()
 
+    layer(inputs, states) # test is_training = false
+
     if not run_only:
         mx.test_utils.assert_almost_equal(np_out, out.asnumpy(), rtol=1e-3, atol=1e-5)
         mx.test_utils.assert_almost_equal(np_dx, inputs.grad.asnumpy(), rtol=1e-3, atol=1e-5)


 

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services