An easy cmake project for tflite-micro,
aims to develope or debug tflm on multi platforms.
The structure is also compatible with arduino library.
- win64 clang14
cmake .. -DCMAKE_C_COMPILER=clang -G "Unix Makefiles"
- win64 gcc 12.2
cmake .. -DCMAKE_C_COMPILER=gcc -G "Unix Makefiles"
- win64 msvc 14
cmake .. -G "Visual Studio 17 2022"
- linux64 gcc 9.4
cmake .. -DCMAKE_C_COMPILER=gcc -G "Unix Makefiles"
- esp32s3 gcc 8.4
cmake .. -DCMAKE_C_COMPILER=xtensa-esp32s3-elf-gcc -DCMAKE_SYSTEM_NAME=Linux -G "Unix Makefiles"
In platformio enviroment, put into lib/tflm
and config ini :
[env:esp32s3]
platform = espressif32
framework = arduino
board = um_pros3
monitor_speed = 115200
board_build.partitions = no_ota.csv
build_flags =
-DTF_LITE_STATIC_MEMORY
-Ilib/tflm/src
-Ilib/tflm/src/third_party
-Ilib/tflm/src/third_party/flatbuffers/include
-Ilib/tflm/src/third_party/gemmlowp
-Ilib/tflm/src/third_party/kissfft
-Ilib/tflm/src/third_party/ruy
sudo apt-get install unzip
sudo pip3 install numpy pillow
make -f tensorflow/lite/micro/tools/make/Makefile third_party_downloads
python3 tensorflow/lite/micro/tools/project_generation/create_tflm_tree.py -e hello_world /tmp/tflm
Then modify some details to fix errors on compile, such as reporter TF_LITE_REMOVE_VIRTUAL_DELETE
to public function. And make a CmakeList.txt
file to for multi platforms.