Usage

NOTE all commands ran in the parent directory of dlgo

Training

Amend the main_network.py file as desired. Training checkpoints are created at each epoch and stored in checkpoints/.

Editable Hyperparams/things

• num_games
• epochs
• batch_size
• model
• optimizer
• encoder

Run

python -m dlgo.nn.main_network

Run locally

python -m dlgo.gtp.play_local

Run reinforcement agent self-play

python -m dlgo.reinforcement.simulate_game

options

• --num_games -n: number of games to play; default = 1
• --in_file -in: hdf5 file containing a Keras 'model' group; default = ./dlgo/agent/latest.hdf5