diff --git a/aten/src/ATen/Formatting.cpp b/aten/src/ATen/Formatting.cpp
index d6ae4453ff0d4..aab224f0d9d39 100644
--- a/aten/src/ATen/Formatting.cpp
+++ b/aten/src/ATen/Formatting.cpp
@@ -260,10 +260,7 @@ std::ostream& print(std::ostream& stream, const Tensor & tensor_, int64_t linesi
       stream << defaultfloat << tensor.data<double>()[0] << std::endl;
       stream << "[ " << tensor_.pImpl->toString() << "{} ]";
     } else if(tensor.ndimension() == 1) {
-      if (tensor.numel() == 0) {
-        stream << "[ Tensor (empty) ]";
-      }
-      else {
+      if (tensor.numel() > 0) {
         double scale;
         int64_t sz;
         std::tie(scale, sz) =  __printFormat(stream, tensor);
@@ -274,18 +271,22 @@ std::ostream& print(std::ostream& stream, const Tensor & tensor_, int64_t linesi
         for(int64_t i = 0; i < tensor.size(0); i++) {
           stream << std::setw(sz) << tensor_p[i]/scale << std::endl;
         }
-        stream << "[ " << tensor_.pImpl->toString() << "{" << tensor.size(0) << "} ]";
       }
+      stream << "[ " << tensor_.pImpl->toString() << "{" << tensor.size(0) << "} ]";
     } else if(tensor.ndimension() == 2) {
-      __printMatrix(stream, tensor, linesize, 0);
+      if (tensor.numel() > 0) {
+        __printMatrix(stream, tensor, linesize, 0);
+      }
       stream << "[ " << tensor_.pImpl->toString() << "{" << tensor.size(0) << "," <<  tensor.size(1) << "} ]";
     } else {
+      if (tensor.numel() > 0) {
         __printTensor(stream, tensor, linesize);
-        stream << "[ " << tensor_.pImpl->toString() << "{" << tensor.size(0);
-        for(int64_t i = 1; i < tensor.ndimension(); i++) {
-          stream << "," << tensor.size(i);
-        }
-        stream << "} ]";
+      }
+      stream << "[ " << tensor_.pImpl->toString() << "{" << tensor.size(0);
+      for(int64_t i = 1; i < tensor.ndimension(); i++) {
+        stream << "," << tensor.size(i);
+      }
+      stream << "} ]";
     }
   }
   return stream;
diff --git a/aten/src/TH/generic/THTensorMath.cpp b/aten/src/TH/generic/THTensorMath.cpp
index 19633cc357813..ad2389f6ee935 100644
--- a/aten/src/TH/generic/THTensorMath.cpp
+++ b/aten/src/TH/generic/THTensorMath.cpp
@@ -288,7 +288,7 @@ void THTensor_(nonzero)(THLongTensor *subscript, THTensor *tensor)
 #ifdef DEBUG
   THAssert(numel <= LONG_MAX);
 #endif
-  THLongTensor_resize2d(subscript, numel, tensor->_dim());
+  THLongTensor_resize2d(subscript, numel, tensor->dim());
 
   /* Second pass populates subscripts */
   subscript_data = THLongTensor_data(subscript);
@@ -296,12 +296,12 @@ void THTensor_(nonzero)(THLongTensor *subscript, THTensor *tensor)
                   if IS_NONZERO(*tensor_data) {
                     div = 1;
 
-                    for (dim = tensor->_dim() - 1; dim >= 0; dim--) {
+                    for (dim = tensor->dim() - 1; dim >= 0; dim--) {
                       *(subscript_data + dim) = (i/div) % tensor->size[dim];
                       div *= tensor->size[dim];
                     }
 
-                    subscript_data += tensor->_dim();
+                    subscript_data += tensor->dim();
                   }
                   ++i;);
 }
diff --git a/aten/src/THC/generic/THCTensorMath.cu b/aten/src/THC/generic/THCTensorMath.cu
index 83019ec101456..368cdebbeff6a 100644
--- a/aten/src/THC/generic/THCTensorMath.cu
+++ b/aten/src/THC/generic/THCTensorMath.cu
@@ -271,7 +271,7 @@ void THCTensor_(nonzero)(THCState* state, THCudaLongTensor *tensor,
   self = THCTensor_(newContiguous)(state, self);
   thrust::device_ptr<real> self_data(THCTensor_(data)(state, self));
 
-  int num_dim = THCTensor_(_nDimension)(state, self);
+  int num_dim = THCTensor_(nDimension)(state, self);
   int64_t N = THCTensor_(nElement)(state, self);
 
   THCudaLongTensor_resize2d(state, tensor, N, num_dim);
diff --git a/test/test_torch.py b/test/test_torch.py
index 20b8a9ded7517..a1c84c8b0818d 100644
--- a/test/test_torch.py
+++ b/test/test_torch.py
@@ -6208,6 +6208,17 @@ def test_nonzero(self):
                     for i in range(dst1.size(0)):
                         self.assertNotEqual(tensor[dst1[i, 0], dst1[i, 1], dst1[i, 2]].item(), 0)
 
+    def test_nonzero_empty(self):
+        if not torch._C._use_zero_size_dim():
+            return
+
+        devices = ['cpu'] if not torch.cuda.is_available() else ['cpu', 'cuda']
+        for device in devices:
+            x = torch.randn(0, 2, 0, 5, 0, device=device)
+            y = torch.nonzero(x)
+            self.assertEqual(0, y.numel())
+            self.assertEqual(torch.Size([0, 5]), y.shape)
+
     def test_deepcopy(self):
         from copy import deepcopy
         a = torch.randn(5, 5)