Telegram ML Competition 2023

Boostrap dependencies:

• install python3
• install your favorite compiler (for example gcc/clang)
• install cmake See docker/builder.dockerfile example for Debian-10

Build:

By default build artifacts will be inside build/${target_name} directory:

• library itself will be located at build/libtglang/libtglang.so
• test binary will be located at build/tglang-tester/tglang-tester

Examples:

#all commands must be run from project root directory

#build all targets
./build.py

#remove artifacts/cache/config from previous build and rebuild from scratch
./build.py --clean

#build just required target, for example libtglang
./build.py --target libtglang

Run & export:

Before run or export build all targets, see build section for details how to build and where to find artifacts.

Run example:

# must be run from project root, will detect build script language
./build/tglang-tester/tglang-tester build.py

For export:

• copy artifact libtglang.so from libtglang target to your library path
• add libtglang sources dir to your include path

Docker build & run:

• Install docker
• Build container:

  #build dev container (for build,validate)
  docker build ./docker/ -f docker/builder.dockerfile -t tglang-builder
  
  #build test container (for validation before submit)
  docker build ./docker/ -f docker/tester.dockerfile -t tglang-tester
  

• Run build script from docker:

  #build and link binaries, then run tester on test file
  #use build script as an examples, 
  #but keep free to add your examples to `examples` dir in project root and run test on it.
  docker run -v .:/app -it tglang-builder --test-file /app/build.py
  

• Run tester in production like environment:

  #single file, original tg tester
  docker run -v ./build-docker/bin:/app/bin -v ./examples:/app/examples -it tglang-tester single /app/examples/build.py
  
  #multi file, modified tg tester accept directory as an argument and print stats about each file (name, tag, size, duration)
  #to profile timing, please use '--build-type Relase' argument for builder for maximum optimizations
  #please copy your files into examples dir inside this project before run
  docker run -v ./build-docker/bin:/app/bin -v ./examples:/app/examples -it tglang-tester multi /app/examples/
  

Keep in mind, you need to forward project root directory to container via -v argument. Base command is docker run -v .:/app -it tglang-builder, after it will run build script with next arguments. So you can work with container as with build script (pass same arguments). To run shell in docker container instead of build script:

docker run --entrypoint bash -v .:/app -it tglang-builder

Prepare submission

# create sbmission
docker run -v .:/app -it tglang-builder --clean --target create-submission

Validate submission

• run empty debian docker container:

   docker run -v ./build-docker/bin:/app/bin -v ./examples:/app/examples -it debian:10 bash
  

• execute next command in empty debian docker container:

   # navigate to directory with artifacts
   cd /app/bin && ls -lah .
  
   # check archive library deps (run inside submission root dir with libtglang.so artifact)
   ldd libtglang.so
  
   # validate archive content,
   # need to install unzip before
   unzip submission.zip -d submission && cd submission && ls -lah .
  
   # test src can be build. 
   # reproduce steps from contest description https://contest.com/docs/ML-Competition-2023#development-and-testing
   # need to install g++ cmake python3
   cd src && mkdir -p build && cd build && cmake -DCMAKE_BUILD_TYPE=Release .. && cmake --build . -j 16
  
  
   # check build library deps (run inside build dir)
   ldd libtglang.so
  
   # TODO: check library can be loaded & executed