relayrl_framework 0.5.0-alpha.1

A distributed, system-oriented multi-agent reinforcement learning framework.
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
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100

{
    "training_server_config": {
        "config_update_polling_seconds": 10.0,
        "default_hyperparameters": {
            "DDPG": {
                "seed": 1,
                "gamma": 0.99,
                "tau": 1e-2,
                "learning_rate": 3e-3,
                "batch_size": 128,
                "buffer_size": 50000,
                "learning_starts": 128,
                "policy_frequency": 1,  
                "noise_scale": 0.1,
                "train_iters": 50
            },
            "PPO": {
                "discrete": true,
                "seed": 0,
                "traj_per_epoch": 1,
                "clip_ratio": 0.1,
                "gamma": 0.99,
                "lam": 0.97,
                "pi_lr": 3e-4,
                "vf_lr": 3e-4,
                "train_pi_iters": 40,
                "train_v_iters": 40,
                "target_kl": 0.01
            },
            "REINFORCE": {
                "discrete": true,
                "with_vf_baseline": true,
                "seed": 1,
                "traj_per_epoch": 8,
                "gamma": 0.98,
                "lam": 0.97,
                "pi_lr": 3e-4,
                "vf_lr": 1e-3,
                "train_vf_iters": 80
            },
            "TD3": {
                "seed": 1,
                "gamma": 0.99,
                "tau": 0.005,
                "learning_rate": 3e-4,
                "batch_size": 128,
                "buffer_size": 50000,
                "exploration_noise": 0.1,
                "policy_noise": 0.2,
                "noise_clip": 0.5,
                "learning_starts": 25000,
                "policy_frequency": 2
            }
        },
        "training_tensorboard": {
            "_comment1": "Runs `tensorboard --logdir /logs` in cwd on start up of server.",
            "launch_tb_on_startup": true,
            "_comment2": "scalar tags can be any column header from `progress.txt` files.",
            "_comment3": "For more than one tag, separate by semi-colon (;)",
            "scalar_tags": "AverageEpRet;LossQ",
            "global_step_tag": "Epoch"
        }
    },
    "transport_config": {
        "inference_addresses": {
            "inference_server": {
                "host": "127.0.0.1",
                "port": "7800"
            },
            "inference_scaling_server": {
                "host": "127.0.0.1",
                "port": "7801"
            }
        },
        "training_addresses": {
            "model_server": {
                "host": "127.0.0.1",
                "port": "50051"
            },
            "trajectory_server": {
                "host": "127.0.0.1",
                "port": "7776"
            },
            "agent_listener": {
                "host": "127.0.0.1",
                "port": "7777"
            },
            "training_scaling_server": {
                "host": "127.0.0.1",
                "port": "7778"
            }
        },
        "local_model_module": {
            "directory": "model_module",
            "model_name": "client_model",
            "format": "pt"
        },
        "max_traj_length": 100000000
    }
}