mirror of
https://github.com/vale981/ray
synced 2025-03-10 13:26:39 -04:00
24 lines
630 B
YAML
24 lines
630 B
YAML
|
two-step-game-qmix-with-qmix-mixer:
|
||
|
|
env: ray.rllib.examples.env.two_step_game.TwoStepGameWithGroupedAgents
|
||
|
|
run: QMIX
|
||
|
|
stop:
|
||
|
|
episode_reward_mean: 8.0
|
||
|
|
timesteps_total: 70000
|
||
|
|
config:
|
||
|
|
# QMIX only supports torch for now.
|
||
|
|
framework: torch
|
||
|
|
|
||
|
|
env_config:
|
||
|
|
env_config:
|
||
|
|
separate_state_space: true
|
||
|
|
one_hot_state_encoding: true
|
||
|
|
|
||
|
|
# W/o this setting, won't get to 8.0 reward.
|
||
|
|
exploration_config:
|
||
|
|
final_epsilon: 0.0
|
||
|
|
|
||
|
|
rollout_fragment_length: 4
|
||
|
|
train_batch_size: 32
|
||
|
|
num_workers: 0
|
||
|
|
mixer: qmix
|