MCPcopy Create free account
hub / github.com/YeWR/EfficientZero / BaseConfig

Class BaseConfig

core/config.py:20–422  ·  view source on GitHub ↗

Source from the content-addressed store, hash-verified

18
19
20class BaseConfig(object):
21
22 def __init__(self,
23 training_steps: int,
24 last_steps: int,
25 test_interval: int,
26 test_episodes: int,
27 checkpoint_interval: int,
28 target_model_interval: int,
29 save_ckpt_interval: int,
30 log_interval: int,
31 vis_interval: int,
32 max_moves: int,
33 test_max_moves: int,
34 history_length: int,
35 discount: float,
36 dirichlet_alpha: float,
37 value_delta_max: float,
38 num_simulations: int,
39 batch_size: int,
40 td_steps: int,
41 num_actors: int,
42 lr_warm_up: float,
43 lr_init: float,
44 lr_decay_rate: float,
45 lr_decay_steps: float,
46 start_transitions: int,
47 auto_td_steps_ratio: float = 0.3,
48 total_transitions: int = 100 * 1000,
49 transition_num: float = 25,
50 do_consistency: bool = True,
51 use_value_prefix: bool = True,
52 off_correction: bool = True,
53 gray_scale: bool = False,
54 episode_life: bool = False,
55 change_temperature: bool = True,
56 init_zero: bool = False,
57 state_norm: bool = False,
58 clip_reward: bool = False,
59 random_start: bool = True,
60 cvt_string: bool = False,
61 image_based: bool = False,
62 frame_skip: int = 1,
63 stacked_observations: int = 16,
64 lstm_hidden_size: int = 64,
65 lstm_horizon_len: int = 1,
66 reward_loss_coeff: float = 1,
67 value_loss_coeff: float = 1,
68 policy_loss_coeff: float = 1,
69 consistency_coeff: float = 1,
70 proj_hid: int = 256,
71 proj_out: int = 256,
72 pred_hid: int = 64,
73 pred_out: int = 256,
74 value_support: DiscreteSupport = DiscreteSupport(-300, 300, delta=1),
75 reward_support: DiscreteSupport = DiscreteSupport(-300, 300, delta=1)):
76 """Base Config for EfficietnZero
77 Parameters

Callers

nothing calls this directly

Calls

no outgoing calls

Tested by

no test coverage detected