amlproject/main.py at main · VitoFe/amlproject · 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
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
"""
python main.py --help
python main.py train --mode centralized --epochs 50
python main.py train --mode federated --num-clients 100
python main.py train --mode sparse --sparsity-ratio 0.9
"""

import argparse
import sys


def main():
    parser = argparse.ArgumentParser(
        description="Federated Learning with Task Arithmetic",
        formatter_class=argparse.RawDescriptionHelpFormatter,
        epilog="""
    # Centralized baseline
    python main.py train --mode centralized --epochs 50

    # Federated learning (IID)
    python main.py train --mode federated --sharding iid

    # Federated learning (non-IID)
    python main.py train --mode federated --sharding non_iid --nc 5

    # Sparse fine-tuning
    python main.py train --mode sparse --sparsity-ratio 0.9

    # Run all experiments
    python main.py run-all --quick
        """,
    )

    subparsers = parser.add_subparsers(dest="command", help="Available commands")

    # Train command
    train_parser = subparsers.add_parser("train", help="Train a model")
    train_parser.add_argument(
        "--mode",
        type=str,
        required=True,
        choices=["centralized", "federated", "sparse"],
        help="Training mode",
    )
    train_parser.add_argument("--config", type=str, help="Path to config file")
    train_parser.add_argument("--epochs", type=int, default=50)
    train_parser.add_argument("--num-clients", type=int, default=100)
    train_parser.add_argument("--num-rounds", type=int, default=500)
    train_parser.add_argument("--local-steps", type=int, default=4)
    train_parser.add_argument(
        "--sharding", type=str, default="iid", choices=["iid", "non_iid"]
    )
    train_parser.add_argument("--nc", type=int, default=10)
    train_parser.add_argument("--sparsity-ratio", type=float, default=0.9)
    train_parser.add_argument("--mask-strategy", type=str, default="least_sensitive")
    train_parser.add_argument("--device", type=str, default="cuda")
    train_parser.add_argument("--seed", type=int, default=42)

    # Run all experiments command
    runall_parser = subparsers.add_parser("run-all", help="Run all experiments")
    runall_parser.add_argument("--quick", action="store_true", help="Quick test run")
    runall_parser.add_argument("--full", action="store_true", help="Full experiments")
    runall_parser.add_argument("--device", type=str, default="cuda")

    args = parser.parse_args()

    if args.command is None:
        parser.print_help()
        return

    if args.command == "train":
        if args.mode == "centralized":
            from scripts.train_centralized import main as train_main

            sys.argv = [
                "train_centralized.py",
                "--epochs",
                str(args.epochs),
                "--device",
                args.device,
                "--seed",
                str(args.seed),
            ]
            train_main()
        elif args.mode == "federated":
            from scripts.train_federated import main as train_main

            sys.argv = [
                "train_federated.py",
                "--num-clients",
                str(args.num_clients),
                "--num-rounds",
                str(args.num_rounds),
                "--local-steps",
                str(args.local_steps),
                "--sharding",
                args.sharding,
                "--nc",
                str(args.nc),
                "--device",
                args.device,
                "--seed",
                str(args.seed),
            ]
            train_main()
        elif args.mode == "sparse":
            from scripts.train_federated_sparse import main as train_main

            sys.argv = [
                "train_federated_sparse.py",
                "--num-clients",
                str(args.num_clients),
                "--num-rounds",
                str(args.num_rounds),
                "--sparsity-ratio",
                str(args.sparsity_ratio),
                "--mask-strategy",
                args.mask_strategy,
                "--device",
                args.device,
                "--seed",
                str(args.seed),
            ]
            train_main()

    elif args.command == "run-all":
        from scripts.run_all_experiments import main as run_all_main

        sys.argv = ["run_all_experiments.py", "--device", args.device]
        if args.quick:
            sys.argv.append("--quick")
        elif args.full:
            sys.argv.append("--full")
        run_all_main()


if __name__ == "__main__":
    main()