api.md

API

In HandyRL, main.py is an entry-point for training and evaluation. This page explains the API interfaces of main.py.

Also see the usage for more details:

• training
• training (server version)

Training API

python main.py [--train|-t]

Evaluation API

python main.py [--eval|-e] MODEL_PATH NUM_GAMES NUM_PROCESSES

• MODEL_PATH
  • model path to evaluate
• NUM_GAMES
  • the number of games to evaluate
• NUM_PROCESSES
  • the number of processes to evaluate model in parallel

Training Server API

python main.py [--train-server|-ts]

Worker API

python main.py [--worker|-w]

Evaluation Server API

python main.py [--eval-server|-es] NUM_GAMES NUM_PROCESSES

• NUM_GAMES
  • the number of games to evaluate
• NUM_PROCESSES
  • the number of processes to evaluate model in parallel

Evaluation Client API

python main.py [--eval-client|-ec] MODEL_PATH

• MODEL_PATH
  • model path to evaluate