bottleneck/io_handler.py at main · Nebularaid2000/bottleneck · 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
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
import os
from typing import Any, Iterable, Tuple

import numpy as np
import torch

from common import config


def set_args(args):
    args.prefix = config['prefix']
    args.ratios = config['ratios']
    args.stride = config['stride'] # radius of neighborhood for sampling (i,j) pair
    args.datasets_dirname = config['datasets_dirname']
    args.pretrained_models_dirname = config['pretrained_models_dirname']

    # ---- set sample number of ij, S and images ------
    # compute interaction between image pixels/patches (including all channels)
    if args.inter_type == "pixel":
        args.pairs_number = config['pairs_number_pixel']
        args.samples_number_of_s = config['samples_number_of_s_pixel']
        args.selected_img_number = config['selected_img_number_pixel']
        args.output_dirname = args.output_dirname + "_INTER_pixel_CLASS_%s_GRID_%dx%d" % (args.chosen_class, args.grid_size, args.grid_size)
    else:
        raise Exception("Not a valid output_dirname")

    args.output_dirname = args.output_dirname + "_seed%d" % args.seed

    # ======= model checkpoint path ======
    if args.arch == "our_alexnet_cifar10_normal_lr0.01_log1_da_flip_crop_best":
        args.checkpoint_path = os.path.join(args.prefix, "checkpoints/our_alexnet_cifar10/normal_lr0.01_log1_da_flip_crop_seed0/model_best.pth")

    # L+ and L- models
    elif args.arch == "our_alexnet_cifar10_dp_pos0_entropy_deltav_baseline_0.5_0.0_lam1.0_1.0_grid16_lr0.01_log1_da_flip_crop_best":
        args.checkpoint_path = os.path.join(args.prefix, "checkpoints/our_alexnet_cifar10/dp_pos0_entropy_deltav_baseline_0.5_0.0_lam1.0_1.0_grid16_lr0.01_log1_da_flip_crop_seed0/model_best.pth")
    elif args.arch == "our_alexnet_cifar10_dp_pos0_deltav_baseline_0.7_0.3_lam1.0_1.0_grid16_lr0.01_log1_da_flip_crop_best":
        args.checkpoint_path = os.path.join(args.prefix, "checkpoints/our_alexnet_cifar10/dp_pos0_deltav_baseline_0.7_0.3_lam1.0_1.0_grid16_lr0.01_log1_da_flip_crop_seed0/model_best.pth")
    elif args.arch == "our_alexnet_cifar10_dp_pos0_entropy_deltav_baseline_1.0_0.7_lam1.0_1.0_grid16_lr0.01_log1_da_flip_crop_best":
        args.checkpoint_path = os.path.join(args.prefix, "checkpoints/our_alexnet_cifar10/dp_pos0_entropy_deltav_baseline_1.0_0.7_lam1.0_1.0_grid16_lr0.01_log1_da_flip_crop_seed0/model_best.pth")

    else:
        raise Exception(f"Model [{args.arch}] not implemented. Error in set_args.")

    # ------ dataset setting -----
    if args.dataset == "cifar10":
        args.class_number = 10
        args.image_size = 32
    else:
        raise Exception(f"dataset [{args.dataset}] not implemented. Error in set_args.")

    # ----- save paths -----
    args.figures_dirname = config['figures_dirname']
    args.output_dir = os.path.join(args.prefix, "results", args.output_dirname, "MODEL_%s_DATA_%s" % (args.arch, args.dataset))

    args.samples_dir = os.path.join(args.output_dir, config['samples_dirname'])
    args.samples_file = os.path.join(args.samples_dir, config['samples_filename'])

    args.pairs_dir = os.path.join(args.output_dir, config['pairs_dirname'])
    args.pairs_file = os.path.join(args.pairs_dir, config['pairs_filename'])
    args.players_dir = os.path.join(args.pairs_dir, config['players_dirname'])
    args.players_dir_with_ratio_pattern = os.path.join(args.players_dir, 'ratio%d')
    args.players_file_pattern = os.path.join(args.players_dir_with_ratio_pattern, '%s.npy')

    args.interactions_logit_dir = os.path.join(args.output_dir, config['interactions_logit_dirname'])
    args.interactions_logit_dir_with_ratio_pattern = os.path.join(args.interactions_logit_dir, 'ratio%d')
    args.interactions_logit_file_pattern = os.path.join(args.interactions_logit_dir_with_ratio_pattern, '%s.pth')

    if hasattr(args, 'softmax_type') and hasattr(args, 'out_type'):
        args.interactions_dir = os.path.join(args.output_dir, config['interactions_dirname'] + "_out_%s_softmax_%s" % (
            args.out_type, args.softmax_type))
        args.interactions_dir_with_ratio_pattern = os.path.join(args.interactions_dir, 'ratio%d')
        args.interactions_file_pattern = os.path.join(args.interactions_dir_with_ratio_pattern, '%s.npy')


class BaseIoHandler:

    def __init__(self, root: str) -> None:
        self.root = root
        if not os.path.isdir(root):
            os.makedirs(root)

    def save(self, *args, **kwargs) -> None:
        raise NotImplementedError

    def load(self, *args, **kwargs) -> Any:
        raise NotImplementedError


class SampleIoHandler(BaseIoHandler):

    def __init__(self, args):
        super().__init__(args.samples_dir)
        self.file = args.samples_file
        self.dataset = args.dataset

    def save(self, data):
        with open(self.file, 'w', encoding='UTF-8') as f:
            # CIFAR10 format: class_name, img index(in the WHOLE dataset), class_index
            f.write('\n'.join(map(lambda item: f'{item[0]},{item[1]},{item[2]}', data)))

    def load(self):
        data = []
        with open(self.file, 'r', encoding='UTF-8') as f:
            for line in f.readlines():
                item = line.strip().split(',')

                if self.dataset == "cifar10":
                    # CIFAR10 format: class_name, img index(in the WHOLE dataset, 0-based), class_index
                    data.append((item[0], int(item[1]), int(item[2])))
                else:
                    raise Exception(f"dataset [{self.dataset}] not implemented. Error in SampleIoHandler.")

        return data


class PairIoHandler(BaseIoHandler): # load and save a player pair (i,j)

    def __init__(self, args) -> None:
        super().__init__(args.pairs_dir)
        self.file = args.pairs_file

    def save(self, data: np.ndarray) -> None:
        np.save(self.file, data)

    def load(self) -> np.ndarray:
        return np.load(self.file)


class PlayerIoHandler(BaseIoHandler): # load and save a context S

    def __init__(self, args) -> None:
        super().__init__(args.players_dir)
        self.players_dir_with_ratio_pattern = args.players_dir_with_ratio_pattern
        self.players_file_pattern = args.players_file_pattern

    def save(self, ratio, name: str, data: np.ndarray) -> None:
        players_dir_with_ratio = self.players_dir_with_ratio_pattern % ratio
        if not os.path.isdir(players_dir_with_ratio):
            os.makedirs(players_dir_with_ratio)
        players_file = self.players_file_pattern % (ratio, name)
        np.save(players_file, data)

    def load(self, ratio, name: str) -> np.ndarray:
        players_file = self.players_file_pattern % (ratio, name)
        return np.load(players_file)


class InteractionLogitIoHandler(BaseIoHandler): # load and save logits

    def __init__(self, args) -> None:
        super().__init__(args.interactions_logit_dir)
        self.interactions_logit_dir_with_ratio_pattern = args.interactions_logit_dir_with_ratio_pattern
        self.interactions_logit_file_pattern = args.interactions_logit_file_pattern
        self.device = args.device

    def save(self, ratio, name: str, data: torch.Tensor) -> None:
        interaction_logit_dir_with_ratio = self.interactions_logit_dir_with_ratio_pattern % ratio
        if not os.path.isdir(interaction_logit_dir_with_ratio):
            os.makedirs(interaction_logit_dir_with_ratio)
        interaction_logit_file = self.interactions_logit_file_pattern % (ratio, name)
        torch.save(data, interaction_logit_file)

    def load(self, ratio, name: str) -> torch.Tensor:
        interaction_logit_file = self.interactions_logit_file_pattern % (ratio, name)
        return torch.load(interaction_logit_file, map_location=self.device)


class InteractionIoHandler(BaseIoHandler): # load and save interactions

    def __init__(self, args) -> None:
        super().__init__(args.interactions_dir)
        self.interactions_dir_with_ratio_pattern = args.interactions_dir_with_ratio_pattern
        self.interactions_file_pattern = args.interactions_file_pattern

    def save(self, ratio, name: str, data: np.ndarray) -> None:
        interaction_dir_with_ratio = self.interactions_dir_with_ratio_pattern % ratio
        if not os.path.isdir(interaction_dir_with_ratio):
            os.makedirs(interaction_dir_with_ratio)
        interaction_file = self.interactions_file_pattern % (ratio, name)
        np.save(interaction_file, data)

    def load(self, ratio, name: str) -> np.ndarray:
        interaction_file = self.interactions_file_pattern % (ratio, name)
        return np.load(interaction_file)