File size: 5,187 Bytes
32b542e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
187
188
189
190
191
192
193
194
195
196
197
198
199
## data structure

* imagenet 1k 

```  
data = {
    'input_sample_list': [
        {
            'data':
            torch.rand(bs, 3, 224, 224, dtype=torch.float32),
            'invalid_mask':
            None,
            'modality':
            'image',
            'data_type': 'input',
            'sample_info': {
                'id': list(range(bs)),
                'path': ['hah' for _ in range(bs)]
            }
        },
    ],
    'target_sample_list': [],
    'target_idx_list': [torch.randint(0, 1000, (bs, ))],
    'target_set_list': ['ImageNet22k'],
    'shared_target_sets': {
        'ImageNet22k': [{
            'data':
            torch.randint(0, 49411, (1000, 11)),
            'invalid_mask':
            torch.zeros(1000, 11, dtype=torch.bool),
            'modality':
            'text',
            'data_type': 'target',
            'sample_info': {
                'distributed': True,
                'total_num': 1000,
            }
        }]
    },
    'task_info': {
        'task_name': 'imagenet',
        'task_type': 'image_classification',
        'dataset_name': 'ImageNet22k',
        'batchsize': None,
        'sampling_ratio': None
    }
}
```
* mscoco caption
```           data = {
    'input_sample_list': [
        {
            'data':
            torch.rand(bs, 3, 224, 224, dtype=torch.float32),
            'invalid_mask':
            None,
            'modality':
            'image',
            'data_type': 'input',
            'sample_info': [{
                'id': id,
                'path': 'hahah',
                'bs': bs
            } for _ in range(bs)]
        },
        {
            'data':
            torch.randint(0, 49411, (bs, 31 * 2)),
            'invalid_mask':
            torch.zeros(bs, 31 * 2, dtype=torch.bool),
            'modality':
            'text',
            'data_type': 'input',
            'sample_info': [{
                'pe_index':
                torch.cat([torch.arange(31),
                            torch.arange(31)],
                            dim=0)
            } for _ in range(bs)]
        },
    ],
    'target_sample_list': [],
    'target_idx_list': [torch.randint(0, 49411, (bs, 31))],
    'target_set_list': ['Vocab_Word'],
    'shared_target_sets': {
        'Vocab_Word': [{
            'data': torch.randint(0, 49411, (49411, 2)),
            'invalid_mask': None,
            'modality': 'text',
            'data_type': 'target',
            'sample_info': {
                'distributed': True,
                'total_num': 49411,
            }
        }]
    },
    'task_info': {
        'task_name': 'mscoco_caption',
        'task_type': 'image_caption',
        'dataset_name': 'MSCOCO',
        'batchsize': None,
        'sampling_ratio': None
    }
}
```


*  text_mlm
```
data = {
    'input_sample_list': [
        {
            'data': torch.randint(0, 49411, (bs, 128)),
            'invalid_mask': torch.zeros(bs, 128, dtype=torch.bool),
            'modality': 'text',
            'data_type': 'input',
            'sample_info': {
                'seq_length': 128
            }
        },
    ],
    'target_sample_list': [],
    'target_idx_list': [torch.randint(0, 49411,
                                        (bs, 128))],  # most are -1,
    'target_set_list': ['Vocab_Word'],
    'shared_target_sets': {
        'Vocab_Word': [{
            'data': torch.randint(0, 49411, (49411, 2)),
            'invalid_mask': None,
            'modality': 'text',
            'data_type': 'target',
            'sample_info': {
                'distributed': True,
                'total_num': 49411,
            }
        }]
    },
    'task_info': {
        'task_name':  'bookswiki_pretrain',
        'task_type': 'text_mlm',
        'dataset_name': 'BooksWiki',
        'batchsize': None,
        'sampling_ratio': None
    }
}
```


 * mscoco retrieval
 ```
data = {
    'input_sample_list': [
        {
            'data':
            torch.rand(bs, 3, 224, 224, dtype=torch.float32),
            'invalid_mask':
            None,
            'modality':
            'image',
            'sample_info': {
                'id': list(range(bs)),
                'path': ['hah' for _ in range(bs)]
            }
        },
    ],
    'target_sample_list': [
        {
            'data': torch.randint(0, 49411, (bs, 30)),
            'invalid_mask': torch.zeros(bs, 30,
                                        dtype=torch.bool),
            'modality': 'text',
            'sample_info': {}
        },
    ],
    'target_idx_list': [],
    'target_set_list': [],
    'shared_target_sets': {
        'ImageNet22k': [{
            'data':
            torch.randint(0, 49411, (1000, 11)),
            'invalid_mask':
            torch.zeros(1000, 11, dtype=torch.bool),
            'modality':
            'text',
            'sample_info': {
                'distributed': True,
                'total_num': 1000,
            }
        }]
    },
    'task_info': {
        'task_name': 'mscoco_retrieve',
        'task_type': 'image_retrieval',
        'dataset_name': 'MSCOCO',
        'batchsize': None,
        'sampling_ratio': None
    }
}
```