This go module is auxiliary and serves to link onnx profiler and visualizer. It analyzes the data of two utilities and compares them with each other.

In the root of the project, run:

go build
./profile_parser *args*

The module supports different startup modes and support for command line arguments has been added for this. To find out information about all the arguments, you need to type in the command line after compilation:

./profile_parser --help

Detailed description of each argument:

Usage of ./profile_parser:
  -input_json_path string
        input path (json file with nodes info)
  -input_txt_path string
        input path (txt file with nodes)
  -output_path string
        output path (csv file)

To run, you need to have two files:

1. the json file received when launching the Inference Session (onnx python module) with the setting

enable_profiling = True;

2. the txt file received when running the visualizer with the parameters

--info_mode=nodes --write_mode=file

It is necessary to specify the paths to these files at startup, and also specify where to put the resulting csv file.