fix boolean_mask for 0-size output (#15731)

apache · Aug 2, 2019 · 87425d2 · 87425d2
1 parent cf28b46
commit 87425d2
Show file tree

Hide file tree

Showing 6 changed files with 36 additions and 12 deletions.
diff --git a/include/mxnet/ndarray.h b/include/mxnet/ndarray.h
@@ -190,11 +190,7 @@ class NDArray {
   /*!
    * \brief set the correct shape of NDArray directly from the storage_shape of its own chunk.
    */
-  void SetShapeFromChunk() {
-    if (!(ptr_->storage_shape.ndim() == 1 && ptr_->storage_shape[0] == 0)) {
-      shape_ = ptr_->storage_shape;
-    }
-  }
+  void SetShapeFromChunk();
   /*
    * This indicates whether an array is a view of another array (created by
    * reshape or slice). If an array is a view and the data is stored in

diff --git a/src/imperative/imperative.cc b/src/imperative/imperative.cc
@@ -313,7 +313,9 @@ std::vector<NDArray*> Imperative::Backward(
     } else {
       info.outputs.emplace_back(outputs[i]->shape(), outputs[i]->ctx(),
                                 true, outputs[i]->dtype());
-      info.outputs.back() = static_cast<real_t>(1.0);
+      if (info.outputs.back().shape().Size() != 0) {
+        info.outputs.back() = static_cast<real_t>(1.0);
+      }
     }
   }
 

diff --git a/src/ndarray/ndarray.cc b/src/ndarray/ndarray.cc
@@ -96,6 +96,13 @@ NDArray::NDArray(const NDArrayStorageType stype, const mxnet::TShape &shape, Con
         dtype, aux_types, aux_shapes);
 }
 
+void NDArray::SetShapeFromChunk() {
+  if (Imperative::Get()->is_np_shape() ||
+      !(ptr_->storage_shape.ndim() == 1 && ptr_->storage_shape[0] == 0)) {
+    shape_ = ptr_->storage_shape;
+  }
+}
+
 struct ChunkMem {
   Storage::Handle h;
   std::vector<Storage::Handle> aux_h;

diff --git a/src/operator/contrib/boolean_mask.cc b/src/operator/contrib/boolean_mask.cc
@@ -143,6 +143,7 @@ inline void BooleanMaskForward<cpu>(const nnvm::NodeAttrs& attrs,
   // set the output shape forcefully
   mxnet::TShape s = data.shape();
   s[axis] = valid_num;
+
   const_cast<NDArray &>(out).Init(s);
   // do the copy
   MSHADOW_TYPE_SWITCH(data.dtype(), DType, {

diff --git a/src/operator/contrib/boolean_mask.cu b/src/operator/contrib/boolean_mask.cu
@@ -79,7 +79,6 @@ inline void BooleanMaskForward<gpu>(const nnvm::NodeAttrs& attrs,
                                 Stream<gpu>::GetStream(s));
   CUDA_CALL(cudaMemcpy(&valid_num, &prefix_sum[idx_size - 1], sizeof(int32_t),
                        cudaMemcpyDeviceToHost));
-  CHECK(valid_num > 0) << "boolean_mask behavior not defined when all masks are 0";
   // Set the output shape forcefully
   mxnet::TShape data_shape = data.shape();
   data_shape[axis] = valid_num;
@@ -88,8 +87,10 @@ inline void BooleanMaskForward<gpu>(const nnvm::NodeAttrs& attrs,
   size_t col_size = input_size / idx.shape()[0];
   // Do the copy
   MSHADOW_TYPE_SWITCH(out.dtype(), DType, {
-    mxnet_op::Kernel<BooleanMaskForwardKernel, gpu>::Launch(
-      s, input_size, out.data().dptr<DType>(), data.data().dptr<DType>(), prefix_sum, col_size);
+    if (valid_num > 0) {
+      mxnet_op::Kernel<BooleanMaskForwardKernel, gpu>::Launch(
+        s, input_size, out.data().dptr<DType>(), data.data().dptr<DType>(), prefix_sum, col_size);
+    }
   });
 }
 
@@ -143,9 +144,11 @@ inline void BooleanMaskBackward<gpu>(const nnvm::NodeAttrs& attrs,
   size_t col_size = input_size / idx_size;
   // Backward pass
   MSHADOW_TYPE_SWITCH(igrad_data.dtype(), DType, {
-    mxnet_op::Kernel<BooleanMaskBackwardKernel, gpu>::Launch(
-      s, input_size, igrad_data.data().dptr<DType>(), req[0], ograd.data().dptr<DType>(),
-      prefix_sum, col_size);
+    if (input_size > 0) {
+      mxnet_op::Kernel<BooleanMaskBackwardKernel, gpu>::Launch(
+        s, input_size, igrad_data.data().dptr<DType>(), req[0], ograd.data().dptr<DType>(),
+        prefix_sum, col_size);
+    }
   });
 }
 

diff --git a/tests/python/unittest/test_operator.py b/tests/python/unittest/test_operator.py
@@ -5589,6 +5589,21 @@ def test_boolean_mask():
     assert same(out.asnumpy(), expected)
     assert same(data.grad.asnumpy(), expected_grad)
 
+    # test 0-size output
+    mx.set_np_shape(True)
+    data = mx.nd.array([[1, 2, 3],[4, 5, 6],[7, 8, 9]])
+    index = mx.nd.array([0, 0, 0])
+    data.attach_grad()
+    with mx.autograd.record():
+        out = mx.nd.contrib.boolean_mask(data, index)
+    out.backward()
+    data.grad.wait_to_read()
+    expected = np.zeros((0, 3))
+    expected_grad = np.array([[0, 0, 0], [0, 0, 0], [0, 0, 0]])
+    assert same(out.asnumpy(), expected)
+    assert same(data.grad.asnumpy(), expected_grad)
+    mx.set_np_shape(False)
+
     # test gradient
     shape = (100, 30)
     a = mx.nd.random.randint(0, 100, shape=shape)