YRC-Bench/flags.py at main · bplaut/YRC-Bench · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
import jsonargparse

def make():
    parser = jsonargparse.ArgumentParser()

    parser.add_argument("-c", "--config", type=str,
                        help="path to YAML config file")
    parser.add_argument("-d", "--general.device", type=int,
                        help="device id")
    parser.add_argument("-wandb", "--use_wandb", action="store_true", default=False,
                        help="log to wandb?")
    parser.add_argument("-no_eval", "--algorithm.no_eval", action="store_true", default=False,
                        help="no evaluation")
    parser.add_argument("-log_freq", "--algorithm.log_freq", type=int,
                        help="Frequency of logging")
    parser.add_argument("-clip_vloss", "--algorithm.clip_vloss", type=int,
                        help="Clip value loss (RL)")
    parser.add_argument("-norm_adv", "--algorithm.norm_adv", type=int,
                        help="Normalize advantage (RL)")
    parser.add_argument("-n", "--name", type=str,
                        help="name of this run")
    parser.add_argument("-over", "--overwrite", action="store_true",
                        help="overwrite experiment folder (if exists)")
    parser.add_argument("-query_cost", "--coord_env.strong_query_cost_ratio", type=float,
                        help="Cost of querying strong agent")
    parser.add_argument("-switch_cost", "--coord_env.switch_agent_cost_ratio", type=float,
                        help="Cost of switching agent")
    parser.add_argument("-en", "--environment.common.env_name", type=str,
                        help="name of the environment")
    parser.add_argument("-sim", "--agents.sim_weak", type=str,
                        help="path to the sim weak agent")
    parser.add_argument("-weak", "--agents.weak", type=str,
                        help="path to the weak agent")
    parser.add_argument("-strong", "--agents.strong", type=str,
                        help="path to the strong agent")
    parser.add_argument("-f_n", "--file_name", type=str,
                        help="file name for evaluation")
    parser.add_argument("-agent", "--general.agent", type=str, choices=["weak", "strong"],
                        help="agent to evaluate")
    parser.add_argument("-cp_feature", "--coord_policy.feature_type", type=str,
                        choices=["obs", "hidden", "hidden_obs", "dist", "hidden_dist", "obs_dist", "obs_hidden_dist"],
                        help="Type of features for coordination policy")
    parser.add_argument("-cp_data_agent", "--coord_policy.collect_data_agent", type=str, choices=["weak", "strong"], default="weak",
                        help="agent to collect data")

    # always policy
    parser.add_argument("-cp_agent", "--coord_policy.agent", type=str,
                        choices=["weak", "strong"],
                        help="always choose action of this agent")

    # threshold policy
    parser.add_argument("-cp_metric", "--coord_policy.metric", type=str,
                        choices=["max_logit", "max_prob", "margin", "neg_entropy", "neg_energy"],
                        help="metric for computing scores")

    # ood policy
    parser.add_argument("-cp_method", "--coord_policy.method", type=str,
                        choices=["DeepSVDD"],
                        help="method for detecting OOD samples")

    # random baseline policy
    parser.add_argument("-cp_base", "--coord_policy.baseline", action="store_true",
                        help="baseline policy with random action 0.5 probability")

    # minigrid
    parser.add_argument("-en_tr_suffix", "--environment.train.env_name_suffix", type=str,
                        help="suffix for the train environment name")
    parser.add_argument("-en_te_suffix", "--environment.test.env_name_suffix", type=str,
                        help="suffix for the test environment name")

    # procgen
    parser.add_argument("-use_bg", "--environment.common.use_background", type=bool, default=True,
                        help="use background - only for procgen envs")
    parser.add_argument("-use_mono_asset", "--environment.common.use_monochrome_assets", type=bool, default=False,
                        help="use monochrome assets - only for procgen envs")
    parser.add_argument("-res_theme", "--environment.common.restrict_themes", type=bool, default=False,
                        help="restrict themes - only for procgen envs")

    parser.add_argument("-seed", "--general.seed", type=int,
                        help="random seed")
    args = parser.parse_args()

    return args