ding.config.example.PPOOffPolicy.gym_qbertnoframeskip_v4

ding.config.example.PPOOffPolicy.gym_qbertnoframeskip_v4

Full Source Code

../ding/config/example/PPOOffPolicy/gym_qbertnoframeskip_v4.py

1from easydict import EasyDict 2import ding.envs.gym_env 3 4cfg = dict( 5 exp_name='QbertNoFrameskip-v4-PPOOffPolicy', 6 env=dict( 7 collector_env_num=16, 8 evaluator_env_num=8, 9 n_evaluator_episode=8, 10 stop_value=10000000000, 11 env_id='QbertNoFrameskip-v4', 12 frame_stack=4, 13 env_wrapper='atari_default', 14 ), 15 policy=dict( 16 cuda=True, 17 model=dict( 18 obs_shape=[4, 84, 84], 19 action_shape=6, 20 encoder_hidden_size_list=[32, 64, 64, 128], 21 actor_head_hidden_size=128, 22 critic_head_hidden_size=128, 23 critic_head_layer_num=2, 24 ), 25 learn=dict( 26 update_per_collect=18, 27 batch_size=128, 28 learning_rate=0.0001, 29 value_weight=1.0, 30 entropy_weight=0.005, 31 clip_ratio=0.1, 32 adv_norm=False, 33 ), 34 collect=dict( 35 n_sample=1024, 36 unroll_len=1, 37 discount_factor=0.99, 38 gae_lambda=0.95, 39 ), 40 ), 41 wandb_logger=dict( 42 gradient_logger=True, video_logger=True, plot_logger=True, action_logger=True, return_logger=False 43 ), 44) 45 46cfg = EasyDict(cfg) 47 48env = ding.envs.gym_env.env