[GitHub] anirudh2290 closed pull request #12091: [MXNET-792] Fix for issue #9816 with dropout operator and RNG

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

anirudh2290 closed pull request #12091: [MXNET-792] Fix for issue #9816 with dropout operator and RNG
URL: https://github.com/apache/incubator-mxnet/pull/12091
 
 
   

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/src/operator/mshadow_op.h b/src/operator/mshadow_op.h
index 7a2032df758..14f34be5bf3 100644
--- a/src/operator/mshadow_op.h
+++ b/src/operator/mshadow_op.h
@@ -265,6 +265,7 @@ MXNET_UNARY_MATH_OP(square_grad, 2.0f * math::id(a));
 
 /*! \brief used for generate Bernoulli mask */
 MXNET_BINARY_MATH_OP_NC(threshold, a < b ? DType(1) : DType(0));
+MXNET_BINARY_MATH_OP_NC(threshold_eq, a <= b ? DType(1) : DType(0));
 
 /*! \brief used for generate element of abs */
 MXNET_UNARY_MATH_OP(abs, math::fabs(a)); // NOLINT(*)
diff --git a/src/operator/nn/dropout-inl.h b/src/operator/nn/dropout-inl.h
index 8e4aac61354..b7c40fbdf52 100644
--- a/src/operator/nn/dropout-inl.h
+++ b/src/operator/nn/dropout-inl.h
@@ -206,7 +206,7 @@ class DropoutOp {
                                     const real_t pkeep) {
       RNG_KERNEL_LOOP(xpu, DType, id, gen, N, step, {
         const real_t rand_num = static_cast<real_t>(genImpl.uniform());
-        mask_out[i] = mshadow_op::threshold::Map<real_t>(rand_num, pkeep) * (1.0f / pkeep);
+        mask_out[i] = mshadow_op::threshold_eq::Map<real_t>(rand_num, pkeep) * (1.0f / pkeep);
         dropout_out[i] = input_data[i] * mask_out[i];
       });
     }
@@ -258,6 +258,7 @@ class DropoutOp {
                                         this->pkeep_);
             return;
           }
+
           // initialize the mask
           LaunchRNG<BernoulliKernel, xpu>(s, pgen, mask.Size(),
                                           mask.dptr<DType>(),
diff --git a/tests/python/unittest/test_operator.py b/tests/python/unittest/test_operator.py
index 90e85d123d5..b1408455206 100644
--- a/tests/python/unittest/test_operator.py
+++ b/tests/python/unittest/test_operator.py
@@ -5513,8 +5513,7 @@ def test_stack():
         check_numeric_gradient(out, inputs)
 
 
-# test fails with seed 990952066: 0 output seen with dropout ratio=0. See issue #9816
-@with_seed(1234)
+@with_seed()
 def test_dropout():
     def zero_count(array, ratio):
         zeros = 0
@@ -5566,6 +5565,7 @@ def check_dropout_ratio(ratio, shape):
 
         exe.arg_arrays[0][:] = 1
         exe.forward(is_train=True)
+
         if not math.isnan(max_value):
             assert exe.outputs[0].asnumpy().max() > 0
         else:


 

----------------------------------------------------------------
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