support 0 size in any of the tensor dimensions in mkldnn (#15295)

Summary:
support 0 size in any of the tensor dimensions in mkldnn
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15295

Differential Revision: D13573747

Pulled By: yinghai

fbshipit-source-id: 5bf7a0b9e2567e80f44981a7823be5407fc94e53

diff --git a/caffe2/ideep/operators/operator_fallback_ideep.h b/caffe2/ideep/operators/operator_fallback_ideep.h
index 87b0b80..2d31489 100644 (file)
--- a/caffe2/ideep/operators/operator_fallback_ideep.h
+++ b/caffe2/ideep/operators/operator_fallback_ideep.h
@@ -126,8 +126,7 @@ class C10_EXPORT IDEEPFallbackOp final : public IDEEPOperator {
           "output type who needs copying.");
       const auto& src = local_output_blobs_[i]->template Get<TensorCPU>();
       auto src_dims = src.sizes().vec();
-      if (src.template IsType<float>() && src.sizes().size() != 0 &&
-          src.numel() != 0 && base_op_->type() != "Python") {
+      if (src.template IsType<float>() && src.dim() != 0 && base_op_->type() != "Python") {
         Blob* dst = OperatorBase::OutputBlob(i);
         // The output tensor must be ideep tensor with public format.
         // If reusing ideep tensor with non-public format, the tensor buffer

diff --git a/caffe2/python/ideep/copy_op_test.py b/caffe2/python/ideep/copy_op_test.py
index 12a6d46..4b0a15b 100644 (file)
--- a/caffe2/python/ideep/copy_op_test.py
+++ b/caffe2/python/ideep/copy_op_test.py
@@ -9,6 +9,7 @@ from random import randint
 from caffe2.proto import caffe2_pb2
 from caffe2.python import core, workspace
 
+
 @unittest.skipIf(not workspace.C.use_mkldnn, "No MKLDNN support.")
 class CopyTest(unittest.TestCase):
     def _get_deep_device(self):
@@ -19,7 +20,7 @@ class CopyTest(unittest.TestCase):
             "CopyCPUToIDEEP",
             ["X"],
             ["X_ideep"],
-            )
+        )
         op.device_option.CopyFrom(self._get_deep_device())
         n = randint(1, 128)
         c = randint(1, 64)
@@ -31,13 +32,12 @@ class CopyTest(unittest.TestCase):
         X_ideep = workspace.FetchBlob("X_ideep")
         np.testing.assert_allclose(X, X_ideep)
 
-    @unittest.skipIf(True, "zero dim is NOT supported for now.")
     def test_copy_to_ideep_zero_dim(self):
         op = core.CreateOperator(
-                "CopyCPUToIDEEP",
-                ["X"],
-                ["X_ideep"],
-            )
+            "CopyCPUToIDEEP",
+            ["X"],
+            ["X_ideep"],
+        )
         op.device_option.CopyFrom(self._get_deep_device())
         n = 0
         c = randint(1, 128)
@@ -52,7 +52,7 @@ class CopyTest(unittest.TestCase):
             "CopyIDEEPToCPU",
             ["X_ideep"],
             ["X"],
-            )
+        )
         op.device_option.CopyFrom(self._get_deep_device())
         n = randint(1, 128)
         c = randint(1, 64)
@@ -64,13 +64,12 @@ class CopyTest(unittest.TestCase):
         X_ideep = workspace.FetchBlob("X")
         np.testing.assert_allclose(X, X_ideep)
 
-    @unittest.skipIf(True, "zero dim is NOT supported for now.")
     def test_copy_from_ideep_zero_dim(self):
         op = core.CreateOperator(
-                "CopyIDEEPToCPU",
-                ["X_ideep"],
-                ["X"],
-            )
+            "CopyIDEEPToCPU",
+            ["X_ideep"],
+            ["X"],
+        )
         op.device_option.CopyFrom(self._get_deep_device())
         n = 0
         c = randint(1, 64)

diff --git a/caffe2/python/pybind_state_ideep.cc b/caffe2/python/pybind_state_ideep.cc
index 17d5d8f..38df5e1 100644 (file)
--- a/caffe2/python/pybind_state_ideep.cc
+++ b/caffe2/python/pybind_state_ideep.cc
@@ -55,7 +55,9 @@ public:
   FetchedBlob FetchTensor(const itensor &atensor, bool force_copy) {
 #ifdef USE_NUMPY
     FetchedBlob result;
-    CAFFE_ENFORCE(atensor.materialized(),
+    CAFFE_ENFORCE((atensor.ndims() != 0) &&
+                  (atensor.get_nelems() == 0 ||
+                   atensor.get_data_handle() != nullptr),
                   "Trying to fetch uninitialized tensor");
     const int numpy_type = CaffeToNumpyType(type_transform(atensor));
     CAFFE_ENFORCE(
@@ -152,9 +154,7 @@ public:
   bool ZeroDim(PyArrayObject *array) {
 #ifdef USE_NUMPY
     int ndim = PyArray_NDIM(array);
-    npy_intp *npy_dims = PyArray_DIMS(array);
-    return ndim == 0 ||
-      std::find(npy_dims, npy_dims + ndim, 0) != npy_dims + ndim;
+    return ndim == 0;
 #else
     CAFFE_THROW("Caffe2 was compiled without NumPy support.");
 #endif