:py:mod:`neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils`
==============================================================================

.. py:module:: neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils

.. autoapi-nested-parse::

   Utils for Tensorflow model converting to ONNX model.



Module Contents
---------------

Classes
~~~~~~~

.. autoapisummary::

   neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils.SeqType



Functions
~~~~~~~~~

.. autoapisummary::

   neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils.set_name
   neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils.find_opset
   neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils.assert_error
   neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils.map_numpy_to_onnx_dtype
   neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils.map_onnx_to_numpy_type
   neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils.add_port_to_name
   neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils.get_tensorflow_node_attr
   neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils.get_tensorflow_tensor_shape
   neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils.get_tensorflow_node_shape_attr
   neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils.map_tensorflow_dtype
   neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils.get_tensorflow_tensor_data
   neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils.convert_tensorflow_tensor_to_onnx
   neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils.read_tensorflow_node_attrs
   neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils.infer_onnx_shape_dtype
   neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils.make_onnx_shape
   neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils.make_onnx_inputs_outputs
   neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils.save_protobuf
   neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils.is_onnx_domain
   neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils.is_list_or_tuple
   neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils.are_shapes_equal
   neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils.get_subgraphs_from_onnx
   neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils.initialize_name_counter
   neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils.get_index_from_strided_slice_of_shape
   neural_compressor.adaptor.tf_utils.graph_rewriter.onnx.tf2onnx_utils.compute_const_folding_using_tf



.. py:function:: set_name(name)

   Set op name for inserted ops.


.. py:function:: find_opset(opset)

   Find opset.


.. py:function:: assert_error(bool_val, error_msg, *args)

   Raise error message.


.. py:function:: map_numpy_to_onnx_dtype(np_dtype)

   Map numpy dtype to ONNX dtype.


.. py:function:: map_onnx_to_numpy_type(onnx_type)

   Map ONNX dtype to numpy dtype.


.. py:function:: add_port_to_name(name, nr=0)

   Map node output number to name.


.. py:function:: get_tensorflow_node_attr(node, name)

   Parse tensorflow node attribute.


.. py:function:: get_tensorflow_tensor_shape(tensor)

   Get shape from tensorflow tensor.


.. py:function:: get_tensorflow_node_shape_attr(node)

   Get shape from tensorflow attr "shape".


.. py:function:: map_tensorflow_dtype(dtype)

   Convert tensorflow dtype to ONNX.


.. py:function:: get_tensorflow_tensor_data(tensor)

   Get data from tensorflow tensor.


.. py:function:: convert_tensorflow_tensor_to_onnx(tensor, name='')

   Convert tensorflow tensor to onnx tensor.


.. py:function:: read_tensorflow_node_attrs(node)

   Read tensorflow node attribute names.


.. py:function:: infer_onnx_shape_dtype(node, opset_version, input_shapes, input_dtypes, initializers=None)

   Infer shapes and dtypes for outputs of the node.

   Sometimes, shape inference needs the values of node's inputs, so initializers are used.


.. py:function:: make_onnx_shape(shape)

   Shape with -1 is not valid in onnx ...

   make it a name.


.. py:class:: SeqType(tensor_dtype)


   Wrap around TensorProto.* to signify a tensor sequence of a given type.


.. py:function:: make_onnx_inputs_outputs(name, elem_type, shape, **kwargs)

   Wrapper for creating onnx graph inputs or outputs.

   :param name: Text
   :param elem_type: TensorProto.DataType
   :param shape: Optional[Sequence[int]]


.. py:function:: save_protobuf(path, message, as_text=False)

   Save ONNX protobuf file.


.. py:function:: is_onnx_domain(domain)

   Check if it's onnx domain.


.. py:function:: is_list_or_tuple(obj)

   Check the object is list or tuple.


.. py:function:: are_shapes_equal(src, dest)

   Check whether 2 shapes are equal.


.. py:function:: get_subgraphs_from_onnx(model_proto)

   Returns an iterator over the graphs/subgraphs of a model (using dfs).


.. py:function:: initialize_name_counter(model_proto)

   Avoid name conflicts by initializing the counter used by make_name based on the provided model.


.. py:function:: get_index_from_strided_slice_of_shape(node, outputs_to_values)

   Returns the index of the dimension that the strided slice is reading from the shape node or None.


.. py:function:: compute_const_folding_using_tf(g, const_node_values, graph_outputs)

   Find nodes with constant inputs and compute their values using TF.