informed-methods/load_model_for_interactive.py at master · Security-FIT/informed-methods · 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
#!/usr/bin/env python3
from argparse import Namespace
import torch

from classifiers.FFBase import FFBase
from common import build_model, EXTRACTORS, CLASSIFIERS

from torchaudio import load


def load_model_for_interactive():
    args: Namespace = Namespace()
    args.extractor = "XLSR_300M"
    args.classifier = "FFAttn1"

    args.processor = "MHFA"
    model_mhfa, _ = build_model(args)
    assert isinstance(model_mhfa, FFBase)
    # model_mhfa.load_state_dict(torch.load("FFConcat3_MHFA_finetune_7.pt", map_location=torch.device('cpu'), weights_only=True))

    # args.processor = "AASIST"
    # model_aasist, _ = build_model(args)
    # assert isinstance(model_aasist, FFBase)
    # model_aasist.load_state_dict(torch.load("FF_AASIST_finetune_5.pt", map_location=torch.device('cpu'), weights_only=True))

    # print("Models loaded successfully")
    # print(model_mhfa, model_aasist)
    return model_mhfa.eval() #, model_aasist.eval()

def model_params():
    params = {}
    args: Namespace = Namespace()
    for extractor in EXTRACTORS.keys():
        for pooling in ("MHFA", "AASIST", "SLS"):
            # for classifier in CLASSIFIERS.keys():
            #     if classifier in ["FFLSTM", "FFLSTM2", "GMMDiff", "LDAGaussianDiff", "SVMDiff"]:
            #         continue
            args.extractor = extractor
            args.processor = pooling
            args.classifier = "FF"
            model, _ = build_model(args)
            assert isinstance(model, FFBase)

            # compute the number of parameters in the model
            num_params = sum(p.numel() for p in model.parameters())
            params[f"{extractor}_{pooling}_scores.txt"] = num_params

    # args.processor = "AASIST"
    # model_aasist, _ = build_model(args)
    # assert isinstance(model_aasist, FFBase)
    # model_aasist.load_state_dict(torch.load("FF_AASIST_finetune_5.pt", map_location=torch.device('cpu'), weights_only=True))

    print(params)
    # return model_mhfa.parameters() #, model_aasist.parameters()

if __name__ == "__main__":
    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
    f_wf, sr1 = load("fake.flac")
    r_wf, sr2 = load("fake.flac")
    # wf, sr = load("babis-zeman.mp3")

    model = load_model_for_interactive()
    print(model(torch.vstack([r_wf, r_wf]).to(device), torch.vstack([f_wf, f_wf]).to(device)))
    # model_params()