[0-size Tensor No.237] Add 0-size Tensor support for paddle.nn.utils.vector_to_parameters

python/paddle/nn/utils/transform_parameters.py

@@ -165,20 +165,34 @@ def vector_to_parameters(
  Tensor(shape=[], dtype=bool, place=Place(cpu), stop_gradient=True,
  True)
  """
+ assert len(vec.shape) == 1
  origin_shapes = []
  sections = []
+ total_elements = 0
  for param in parameters:
  shape = param.shape
  origin_shapes.append(shape)
  numel = reduce(lambda x, y: x * y, shape, 1)
+ total_elements += numel
  sections.append(numel)
 
  if len(sections) == 1:
  sections.append(0)
 
  if in_dygraph_mode():
  with paddle.base.dygraph.no_grad():
- res = _C_ops.split(vec, sections, 0)
+ res = []
+ if total_elements == vec.shape[0]:
+ res = _C_ops.split(vec, sections, 0)
+ elif total_elements < vec.shape[0]:
+ pointer = 0
+ for section in sections:
+ res.append(vec[pointer : pointer + section])
+ pointer += section
+ else:
+ raise ValueError(
+ "The total_elements of vec should be equal to or larger than the number of elements in parameters."
+ )
  for i in range(0, len(parameters)):
  res[i]._share_underline_tensor_to(parameters[i])
  else:

test/legacy_test/test_parameter.py

@@ -150,5 +150,47 @@ def test_parambase_to_vector_zero(self):
  self.assertEqual(vec.shape, [15])
 
 
+class TestVectorToParam(unittest.TestCase):
+ def test_vector_to_param_zerosize(self):
+ # test the case that the parameters contains zero size tensor
+ with guard():
+ vec = paddle.randn([18], dtype='float32')
+ param1 = paddle.empty([5], dtype='float32')
+ param2 = paddle.empty([5], dtype='float32')
+ param3 = paddle.empty([8], dtype='float32')
+ param4 = paddle.empty([0], dtype='float32')
+ params = [param1, param2, param3, param4]
+ paddle.nn.utils.vector_to_parameters(vec, params)
+ # concat the parameters and get the original vector
+ vec_ = paddle.concat(params, axis=0)
+ np.testing.assert_array_equal(vec_.numpy(), vec.numpy())
+
+ def test_vector_to_param1(self):
+ # test the case that the sum of parameter's elements less than vector elements
+ with guard():
+ vec = paddle.randn([18], dtype='float32')
+ param1 = paddle.empty([5], dtype='float32')
+ param2 = paddle.empty([5], dtype='float32')
+ param3 = paddle.empty([7], dtype='float32')
+ params = [param1, param2, param3]
+ paddle.nn.utils.vector_to_parameters(vec, params)
+ # concat the parameters and get the original vector
+ vec_ = paddle.concat(params, axis=0)
+ np.testing.assert_array_equal(vec_.numpy(), vec[:17].numpy())
+
+ def test_vector_to_param2(self):
+ # test the case that the sum of parameter's elements grater than vector elements
+ def _test_vector_to_param():
+ with guard():
+ vec = paddle.randn([18], dtype='float32')
+ param1 = paddle.empty([5], dtype='float32')
+ param2 = paddle.empty([5], dtype='float32')
+ param3 = paddle.empty([9], dtype='float32')
+ params = [param1, param2, param3]
+ paddle.nn.utils.vector_to_parameters(vec, params)
+
+ self.assertRaises(ValueError, _test_vector_to_param)
+
+
 if __name__ == '__main__':
  unittest.main()