[Feature] convert np.ndarray to backend tensor when setting ndata/edata(#850)

python/dgl/backend/pytorch/tensor.py

@@ -270,7 +270,7 @@ def zerocopy_to_numpy(input):
     return asnumpy(input)
 
 def zerocopy_from_numpy(np_array):
-    return th.from_numpy(np_array)
+    return th.as_tensor(np_array)
 
 def zerocopy_to_dgl_ndarray(input):
     return nd.from_dlpack(dlpack.to_dlpack(input.contiguous()))

python/dgl/view.py

@@ -4,6 +4,8 @@ from __future__ import absolute_import
 from collections import namedtuple
 from collections.abc import MutableMapping
 
+import numpy as np
+
 from .base import ALL, is_all, DGLError
 from . import backend as F
 
@@ -57,6 +59,8 @@ class NodeDataView(MutableMapping):
         return self._graph.get_n_repr(self._nodes)[key]
 
     def __setitem__(self, key, val):
+        if isinstance(val, np.ndarray):
+            val = F.zerocopy_from_numpy(val)
         self._graph.set_n_repr({key : val}, self._nodes)
 
     def __delitem__(self, key):
@@ -125,6 +129,8 @@ class EdgeDataView(MutableMapping):
         return self._graph.get_e_repr(self._edges)[key]
 
     def __setitem__(self, key, val):
+        if isinstance(val, np.ndarray):
+            val = F.zerocopy_from_numpy(val)
         self._graph.set_e_repr({key : val}, self._edges)
 
     def __delitem__(self, key):