make_tensor_proto.md

function make_tensor_proto

make_tensor_proto(values, dtype=None, shape=None)

Creates TensorProto object from values.

Args:

• values: Values to put in the TensorProto. The accepted types are: python scalar, pythons list, numpy scalar and numpy ndarray. Python scalars and lists are internally converted to their numpy counterparts before further processing. Bytes values are placed in TensorProto string_val field.

• dtype (optional): tensor_pb2 DataType value. If not provided, the function will try to infer the data type from values argument.

• shape (optional): The list or tuple of integers defining the shape of the tensor. If not provided, the function will try to infer the shape from values argument.

Returns: TensorProto object filled with values.

Raises:

• TypeError: if unsupported types are provided.
• ValueError: if arguments have inappropriate values.

Examples:

# With python list
data = [[1, 2, 3], [4, 5, 6]]
tensor_proto = make_tensor_proto(data)


# With numpy array:
data = np.array([[1, 2, 3], [4, 5, 6]], np.int32)
tensor_proto = make_tensor_proto(data)


# With binary data:
with open("image.jpg", "rb") as f:
    data = f.read()
tensor_proto = make_tensor_proto(data)

---

Return to the main page