search

yenchenlin / DeepLearningFlappyBird

Flappy Bird hack using Deep Reinforcement Learning (Deep Q-learning).
MIT License
6.62k stars 2.04k forks source link

Add argparse for configuring parameters at runtime and wandb for logging #65

Closed gabesmed closed 2 years ago

gabesmed commented 4 years ago

This PR adds a few additions that make it easier to configure parameters for training and logging.

I added argparse and moved a few of the parameters to be settable with command line arguments. This makes it easier to run a training run with different settings and compare.

I also added wandb conditionally. W&B a free tool that anyone can use anonymously to get hosted TensorBoard and graphs for key metrics like q_max. It makes it easy to compare runs too.

Example: https://app.wandb.ai/gabesmed/flappy/runs/cxll4uss

Screen Shot 2019-09-27 at 4 46 39 PM

Cheers!