Tune.run exploration and exploitation

Ray Libraries (Data, Train, Tune, Serve) Ray Tune
1

Explore and Exploit in tune.run

I’m training DQN agent with tune.run and I haven’t been able to find documentation about exploration and exploitation, if there is or not - please let me know.

Thanks in advance,
Amit & Guy.

2

For example on DQN:

exploration_config:
    type: "EpsilonGreedy"
    epsilon_schedule:
      type: "ExponentialSchedule"
      initial_p: 1.0
      schedule_timesteps: 100_000
      decay_rate: 0.01

For apex-DQN, each worker has its own epsilon