text-classification/torchutil.py at master · silknow/text-classification · 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
139
140
141
142
143
144
145
146
147
148
149
"""
Utility functions for using with pytorch
"""

import torch
import torch.nn.functional as F
import numpy as np
from tqdm import tqdm


def output_to_multiclass(out, dim=1, return_scores=False):
    """Converts output of an output layer to a class (id) via softmax
    and argmax."""
    softmax_output = F.softmax(out, dim=dim)
    class_ids = torch.argmax(softmax_output, dim=dim)

    if return_scores:
        scores = torch.max(softmax_output, dim=dim).values
        return class_ids, scores

    return class_ids


def output_to_multiclass_pair(out, dim=1):
    """Converts output of an output layer to a class (id) via softmax
    and argmax. Returns (class, score)."""
    out = F.softmax(out, dim=dim)

    out = torch.max(out, dim=dim)
    score = out.values.item()
    class_id = out.indices.item()

    return (class_id, score)


def output_to_multilabel(out, dim=None):
    """Converts output of an output layer to a class (id) via softmax
    and argmax."""
    out = torch.sigmoid(out)
    out = out > 0.5
    return out


def predict(model, data_iter):
    model.eval()
    y_prd = []
    with torch.no_grad():
        for x, y in tqdm(data_iter):
            if torch.cuda.is_available():
                x = x.cuda()
                y = y.cuda()

            # forward
            output = model(x)

            output = output_to_multiclass(output, dim=1)
            output = output.item()
            y_prd.append(output)

    y_prd = np.array(y_prd)
    return y_prd


def predict_sko(model, data_iter, mode, return_scores=False):
    model.eval()
    y_prd = []
    y_scr = []
    with torch.no_grad():
        for x, y in tqdm(data_iter):
            if torch.cuda.is_available():
                x = x.cuda()
                y = y.cuda()

            # forward
            output = model(x)

            # is hidden included
            if type(output) == tuple:
                output, _ = output

            if mode == "multiclass" and return_scores:
                output, scores = output_to_multiclass(
                    output, dim=1, return_scores=True
                )
                output = output.item()
                y_prd.append(output)
                scores = scores.item()
                y_scr.append(scores)

            elif mode == "multiclass":
                output = output_to_multiclass(output, dim=1)
                output = output.item()
                y_prd.append(output)
            elif mode == "multilabel":
                output = output_to_multilabel(output, dim=1)
                y_prd.append(output.cpu())
            else:
                raise ValueError(f"Invalid mode: {mode}")

    if mode == "multiclass":
        y_prd = np.array(y_prd)
    elif mode == "multilabel":
        y_prd = torch.cat(y_prd, dim=0).numpy()

    if return_scores:
        y_scr = np.array(y_scr)
        return y_prd, y_scr

    return y_prd


def predict_simple(model, data_iter):
    model.eval()
    y_prd = []
    with torch.no_grad():
        for x, y in tqdm(data_iter):
            if torch.cuda.is_available():
                x = x.cuda()
                y = y.cuda()

            # forward
            output = model(x)

            # is hidden included
            if type(output) == tuple:
                output, _ = output

            output = output_to_multiclass_pair(output, dim=1)
            y_prd.append(output)

    return y_prd


def generate_hidden_vectors(model, data_iter):
    model.eval()
    hidden_vectors = []
    with torch.no_grad():
        for x, y in tqdm(data_iter):
            if torch.cuda.is_available():
                x = x.cuda()
                y = y.cuda()

            # forward
            _, hidden = model(x)
            hidden_vectors.append(hidden)

    hidden_vectors = [h.cpu().numpy() for h in hidden_vectors]

    return hidden_vectors