snd4onnx

Simple node deletion tool for onnx. I only test very miscellaneous and limited patterns as a hobby. There are probably a large number of bugs. Pull requests are welcome.

PINTO0309/simple-onnx-processing-tools

1. Setup

1-1. HostPC

### option
$ echo export PATH="~/.local/bin:$PATH" >> ~/.bashrc \
&& source ~/.bashrc

### run
$ pip install -U onnx \
&& python3 -m pip install -U onnx_graphsurgeon --index-url https://pypi.ngc.nvidia.com \
&& pip install -U snd4onnx

1-2. Docker

PINTO0309/simple-onnx-processing-tools

2. CLI Usage

$ snd4onnx -h

usage:
    snd4onnx [-h]
    -rn REMOVE_NODE_NAMES [REMOVE_NODE_NAMES ...]
    -if INPUT_ONNX_FILE_PATH
    -of OUTPUT_ONNX_FILE_PATH
    [-n]

optional arguments:
  -h, --help
        show this help message and exit.

  -rn REMOVE_NODE_NAMES [REMOVE_NODE_NAMES ...], --remove_node_names REMOVE_NODE_NAMES [REMOVE_NODE_NAMES ...]
        ONNX node name to be deleted.

  -if INPUT_ONNX_FILE_PATH, --input_onnx_file_path INPUT_ONNX_FILE_PATH
        Input onnx file path.

  -of OUTPUT_ONNX_FILE_PATH, --output_onnx_file_path OUTPUT_ONNX_FILE_PATH
        Output onnx file path.

  -n, --non_verbose
        Do not show all information logs. Only error logs are displayed.

3. In-script Usage

>>> from snd4onnx import remove
>>> help(remove)

Help on function remove in module snd4onnx.onnx_remove_node:

remove(
    remove_node_names: List[str],
    input_onnx_file_path: Union[str, NoneType] = '',
    output_onnx_file_path: Union[str, NoneType] = '',
    onnx_graph: Union[onnx.onnx_ml_pb2.ModelProto, NoneType] = None,
    non_verbose: Union[bool, NoneType] = False
) -> onnx.onnx_ml_pb2.ModelProto

    Parameters
    ----------
    remove_node_names: List[str]
        List of OP names to be deleted.
        e.g. remove_node_names = ['op_name1', 'op_name2', 'op_name3', ...]

    input_onnx_file_path: Optional[str]
        Input onnx file path.
        Either input_onnx_file_path or onnx_graph must be specified.

    output_onnx_file_path: Optional[str]
        Output onnx file path.
        If output_onnx_file_path is not specified, no .onnx file is output.

    onnx_graph: Optional[onnx.ModelProto]
        onnx.ModelProto.
        Either input_onnx_file_path or onnx_graph must be specified.
        onnx_graph If specified, ignore input_onnx_file_path and process onnx_graph.

    non_verbose: Optional[bool]
        Do not show all information logs. Only error logs are displayed.
        Default: False

    Returns
    -------
    removed_graph: onnx.ModelProto
        OP removed onnx ModelProto.

4. CLI Execution

$ snd4onnx \
--remove_node_names node_name_a node_name_b \
--input_onnx_file_path input.onnx \
--output_onnx_file_path output.onnx

5. In-script Execution

from snd4onnx import remove

onnx_graph = remove(
    remove_node_names=['node_name_a', 'node_name_b'],
    input_onnx_file_path='input.onnx',
)

# or

onnx_graph = remove(
    remove_node_names=['node_name_a', 'node_name_b'],
    onnx_graph=graph,
)

6. Sample

6-1. sample.1

Before	After

6-2. sample.2

Before	After

6-3. sample.3

Before	After

6-4. sample.4

Before	After

6-5. sample.5

Before	After

7. Reference

1. https://docs.nvidia.com/deeplearning/tensorrt/onnx-graphsurgeon/docs/index.html
2. https://github.com/NVIDIA/TensorRT/tree/main/tools/onnx-graphsurgeon
3. PINTO0309/scs4onnx
4. PINTO0309/sne4onnx
5. PINTO0309/snc4onnx
6. PINTO0309/sog4onnx
7. PINTO0309/PINTO_model_zoo

8. Issues

https://github.com/PINTO0309/simple-onnx-processing-tools/issues