def testConvertTensors(self):
a = tf.placeholder(tf.int32, [None])
protomap = _make_signature({"a": a}, {}).inputs
targets = tensor_info.parse_tensor_info_map(protomap)
# convert constant
in0 = [1, 2, 3]
output = tensor_info.convert_dict_to_compatible_tensor({"a": in0}, targets)
self.assertEquals(output["a"].dtype, a.dtype)
# check sparsity
in1 = tf.sparse_placeholder(tf.int32, [])
with self.assertRaisesRegexp(TypeError, "dense"):
tensor_info.convert_dict_to_compatible_tensor({"a": in1}, targets)
def _convert_dict_inputs(inputs, tensor_info_map):
"""Converts from inputs into dict of input tensors.
This handles:
- putting inputs into a dict, per _prepare_dict_inputs(),
- converting all input values into tensors compatible with the
expected input tensor (dtype, shape).
- check sparse/non-sparse tensor types.
Args:
inputs: inputs fed to Module.__call__().
tensor_info_map: A map from string to `tensor_info.ParsedTensorInfo`
describing the signature inputs.
Returns:
A dict of tensors to feed to the signature instantiation.
Raises:
TypeError: If it fails to convert the input values into a dict of tensors
to feed to the signature instantiation.
"""
dict_inputs = _prepare_dict_inputs(inputs, tensor_info_map)
return tensor_info.convert_dict_to_compatible_tensor(
dict_inputs, tensor_info_map)
