Hyeongseok Oh

>How I made this example nnpkg We can add README.md file about this.

@chunseoklee @seanshpark Why don't we use `flatbuffers` latest version (`2.0.0`) for all compiler and runtime modules? Tizen is using `2.0.0` to build both `tensorflow-1.13.1` and `tensorflow-2.7.0`.

### Can we remove environment variable config on runtime? - On most test, possible - Move config reading from environment variable to test binary from runtime - After reading environment...

### Used on CI Test - General test - `DISABLE_COMPILE` - `BACKEND` - `EXECUTOR` - Mixed test - `OP_BACKEND_XXXX` (Operator-Backend mapping) - `ACL_LAYOUT` - `RUY_THREADS` ### Using modules - Frontend...

LGTM

@jyoungyun Can't we close this draft?

As you commented, `CircleGen` is introduced to prevent including model file in this repo and remove dependency with server setting. `Option 1` is used to check nnpackage unittest to check...

Before handling subgraph partitioning, we should consider that `tflite` assume that model input and output are belong to main subgraph (subgraph 0). If we divide model file on main graph,...

How can we meet above message?

> > How can we meet above message? > > in `compiler/luci-micro/CMakeLists.txt` edit `-DTARGET_CPU=armv7em`, then `make luci_interpreter_micro_arm` I checked and found out `TARGET_CPU` is used on cmake toolchain file `infra/nncc/cmake/buildtool/config/arm-none-eabi-gcc.cmake`...