Execute

Execute main.py

$ python main.py

Read an mjai-log

You can read an mjai-log in log directory by

File > Open

Play a game

You need to copy following from akochan(https://github.com/critter-mj/akochan)

ai.dll (windows)
libai.so (linux)
system.exe
setup_mjai.json
params

Then you can play a game by

AI > Start Game

and input random seed (non-negative integer).

Convert Tenhou-log to mjai-log

1. Create a directory named 'tenhou_rawlog'
2. download scraw20XX.zip from https://tenhou.net/sc/raw/ and put extracted directory('scraw20XX') in 'tenhou_rawlog' directory
3. Clone and build akochan-reviewer in root directory

    git clone https://github.com/Equim-chan/akochan-reviewer.git
    cd akochan-reviewer
    cargo build

4. Run. Output will be saved in tenhou_mjailog directory.

    python main.py --tenhou_convlog --year 20XX

Dump feature for supervised learning (from converted Tenhou-log)

    python main.py --dump_feature_tenhou --prefix <prefix_str>

<prefix_str> is prefix of json filename in tenhou_mjailog (for example 2018, 20180104, 2018010400gm-00a9). Note that system.exe from akochan, and converted mjai-log of Tenhou is required. When you run with option --update, old output is updated. Output files are saved in tenhou_npz directory.

Dump feature for supervised learning (from own log)

    python main.py --dump_feature --out_dir <outdir_str> --file_path <filepath_str>

Note that system.exe from akochan is required. Output files are saved in <outdir_str> directory.

Used materials

Images : https://majandofu.com/mahjong-images