[TODO] Re-release flatc with some customizations of our own to address the lack of arithmetic precision of the quantization parameters

[PINTO0309]

Issue Type

Others

onnx2tf version number

1.7.2

onnx version number

1.13.1

tensorflow version number

2.12.0rc0

Download URL for ONNX

N/A

Parameter Replacement JSON

N/A

Description

1. Personal
2. [TODO] Re-release flatc with some customizations of our own to address the lack of arithmetic precision of the quantization parameters
   • https://github.com/PINTO0309/tflite2json2tflite/tree/main#5-flatbuffers-flatc
   • tflite to JSON to tflite quantization error tflite2json2tflite#1
   • Stop using the following official binaries
     • https://packages.ubuntu.com/search?keywords=flatbuffers-compiler&searchon=names&suite=focal&section=all
3. Source code
   https://github.com/PINTO0309/flatbuffers

• flatbuffers/include/flatbuffers/util.h

  • From:

    template<> inline std::string NumToString<double>(double t) {
      return FloatToString(t, 12);
    }
    template<> inline std::string NumToString<float>(float t) {
      return FloatToString(t, 6);
    }

  • To:

    template<> inline std::string NumToString<double>(double t) {
      return FloatToString(t, 12);
    }
    template<> inline std::string NumToString<float>(float t) {
      return FloatToString(t, 17);
    }

• build

  cd flatbuffers && mkdir build && cd build
  cmake -G "Unix Makefiles" -DCMAKE_BUILD_TYPE=Release ..
  make -j$(nproc)