relayrl_framework 0.4.52

A system-oriented, distributed reinforcement learning framework using a Rust backend with Python interfaces.
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

{
    "algorithms": {
        "REINFORCE": {
            "discrete": true,
            "with_vf_baseline": false,
            "seed": 1,
            "traj_per_epoch": 8,
            "gamma": 0.98,
            "lam": 0.97,
            "pi_lr": 3e-4,
            "vf_lr": 1e-3,
            "train_vf_iters": 80
        }
    },
    "grpc_idle_timeout": 30,
    "max_traj_length": 1000,
    "model_paths": {
        "client_model": "client_model.pt",
        "server_model": "server_model.pt"
    },
    "server": {
        "_comment": "gRPC uses only this address (prefix is unused).",
        "training_server": {
            "prefix": "tcp://",
            "host": "127.0.0.1",
            "port": "50051"
        },
        "trajectory_server": {
            "prefix": "tcp://",
            "host": "127.0.0.1",
            "port": "7776"
        },
        "agent_listener": {
            "prefix": "tcp://",
            "host": "127.0.0.1",
            "port": "7777"
        }
    },
    "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.",
        "scalar_tags": "AverageEpRet;LossQ",
        "global_step_tag": "Epoch"
    }
}