phonghoccode commited on
Commit
07b4a5a
·
verified ·
1 Parent(s): e9d9ebf

Upload CustomViltForVQA

Browse files
Files changed (2) hide show
  1. config.json +216 -216
  2. model.safetensors +1 -1
config.json CHANGED
@@ -8,227 +8,227 @@
8
  "hidden_dropout_prob": 0.0,
9
  "hidden_size": 768,
10
  "id2label": {
11
- "0": "sandwich",
12
- "1": "6",
13
- "2": "vase",
14
- "3": "5",
15
- "4": "truck",
16
- "5": "carrot",
17
- "6": "potted plant",
18
- "7": "car",
19
- "8": "apple",
20
- "9": "donut",
21
- "10": "fork",
22
- "11": "handbag",
23
- "12": "tv",
24
- "13": "motorcycle",
25
- "14": "baseball glove",
26
- "15": "surfboard",
27
- "16": "toilet",
28
- "17": "person",
29
- "18": "bottle",
30
- "19": "bear",
31
- "20": "dining table",
32
- "21": "knife",
33
- "22": "cell phone",
34
- "23": "No",
35
- "24": "Blue",
36
- "25": "Green",
37
- "26": "bowl",
38
- "27": "spoon",
39
- "28": "broccoli",
40
- "29": "0",
41
- "30": "toothbrush",
42
- "31": "sheep",
43
- "32": "elephant",
44
- "33": "giraffe",
45
- "34": "cake",
46
- "35": "parking meter",
47
- "36": "cat",
48
- "37": "traffic light",
49
- "38": "Behind",
50
- "39": "Yes",
51
- "40": "wine glass",
52
- "41": "Right",
53
- "42": "Brown",
54
- "43": "stop sign",
55
- "44": "right",
56
- "45": "sink",
57
- "46": "sports ball",
58
- "47": "book",
59
- "48": "horse",
60
- "49": "7",
61
- "50": "bench",
62
- "51": "tie",
63
- "52": "cup",
64
- "53": "fire hydrant",
65
- "54": "hot dog",
66
- "55": "skateboard",
67
- "56": "Orange",
68
- "57": "remote",
69
- "58": "baseball bat",
70
- "59": "oven",
71
- "60": "train",
72
- "61": "Below",
73
- "62": "8",
74
- "63": "Red",
75
- "64": "mouse",
76
- "65": "teddy bear",
77
- "66": "chair",
78
- "67": "keyboard",
79
- "68": "2",
80
- "69": "zebra",
81
- "70": "laptop",
82
- "71": "Front",
83
- "72": "1",
84
- "73": "clock",
85
- "74": "Grey",
86
- "75": "pizza",
87
- "76": "umbrella",
88
- "77": "above",
89
- "78": "kite",
90
- "79": "bus",
91
- "80": "Black",
92
- "81": "Yellow",
93
- "82": "frisbee",
94
- "83": "airplane",
95
- "84": "White",
96
- "85": "left",
97
- "86": "bird",
98
- "87": "Pink",
99
- "88": "orange",
100
- "89": "bicycle",
101
- "90": "couch",
102
- "91": "Purple",
103
- "92": "skis",
104
- "93": "3",
105
- "94": "9",
106
- "95": "refrigerator",
107
- "96": "snowboard",
108
- "97": "scissors",
109
- "98": "tennis racket",
110
- "99": "suitcase",
111
- "100": "cow",
112
- "101": "bed",
113
- "102": "dog",
114
- "103": "boat",
115
- "104": "4",
116
- "105": "Above",
117
- "106": "banana",
118
- "107": "Left"
119
  },
120
  "image_size": 384,
121
  "initializer_range": 0.02,
122
  "intermediate_size": 3072,
123
  "label2id": {
124
- "0": 29,
125
- "1": 72,
126
- "2": 68,
127
- "3": 93,
128
- "4": 104,
129
- "5": 3,
130
- "6": 1,
131
- "7": 49,
132
- "8": 62,
133
- "9": 94,
134
- "Above": 105,
135
- "Behind": 38,
136
- "Below": 61,
137
- "Black": 80,
138
- "Blue": 24,
139
- "Brown": 42,
140
- "Front": 71,
141
- "Green": 25,
142
- "Grey": 74,
143
- "Left": 107,
144
- "No": 23,
145
- "Orange": 56,
146
- "Pink": 87,
147
- "Purple": 91,
148
- "Red": 63,
149
- "Right": 41,
150
- "White": 84,
151
- "Yellow": 81,
152
- "Yes": 39,
153
- "above": 77,
154
- "airplane": 83,
155
- "apple": 8,
156
- "banana": 106,
157
- "baseball bat": 58,
158
- "baseball glove": 14,
159
- "bear": 19,
160
- "bed": 101,
161
- "bench": 50,
162
- "bicycle": 89,
163
- "bird": 86,
164
- "boat": 103,
165
- "book": 47,
166
- "bottle": 18,
167
- "bowl": 26,
168
- "broccoli": 28,
169
- "bus": 79,
170
- "cake": 34,
171
- "car": 7,
172
- "carrot": 5,
173
- "cat": 36,
174
- "cell phone": 22,
175
- "chair": 66,
176
- "clock": 73,
177
- "couch": 90,
178
- "cow": 100,
179
- "cup": 52,
180
- "dining table": 20,
181
- "dog": 102,
182
- "donut": 9,
183
- "elephant": 32,
184
- "fire hydrant": 53,
185
- "fork": 10,
186
- "frisbee": 82,
187
- "giraffe": 33,
188
- "handbag": 11,
189
- "horse": 48,
190
- "hot dog": 54,
191
- "keyboard": 67,
192
- "kite": 78,
193
- "knife": 21,
194
- "laptop": 70,
195
- "left": 85,
196
- "motorcycle": 13,
197
- "mouse": 64,
198
- "orange": 88,
199
- "oven": 59,
200
- "parking meter": 35,
201
- "person": 17,
202
- "pizza": 75,
203
- "potted plant": 6,
204
- "refrigerator": 95,
205
- "remote": 57,
206
- "right": 44,
207
- "sandwich": 0,
208
- "scissors": 97,
209
- "sheep": 31,
210
- "sink": 45,
211
- "skateboard": 55,
212
- "skis": 92,
213
- "snowboard": 96,
214
- "spoon": 27,
215
- "sports ball": 46,
216
- "stop sign": 43,
217
- "suitcase": 99,
218
- "surfboard": 15,
219
- "teddy bear": 65,
220
- "tennis racket": 98,
221
- "tie": 51,
222
- "toilet": 16,
223
- "toothbrush": 30,
224
- "traffic light": 37,
225
- "train": 60,
226
- "truck": 4,
227
- "tv": 12,
228
- "umbrella": 76,
229
- "vase": 2,
230
- "wine glass": 40,
231
- "zebra": 69
232
  },
233
  "layer_norm_eps": 1e-12,
234
  "max_image_length": -1,
 
8
  "hidden_dropout_prob": 0.0,
9
  "hidden_size": 768,
10
  "id2label": {
11
+ "0": "5",
12
+ "1": "zebra",
13
+ "2": "couch",
14
+ "3": "sink",
15
+ "4": "apple",
16
+ "5": "umbrella",
17
+ "6": "bus",
18
+ "7": "dog",
19
+ "8": "boat",
20
+ "9": "Grey",
21
+ "10": "suitcase",
22
+ "11": "2",
23
+ "12": "bottle",
24
+ "13": "spoon",
25
+ "14": "3",
26
+ "15": "toilet",
27
+ "16": "8",
28
+ "17": "7",
29
+ "18": "knife",
30
+ "19": "Pink",
31
+ "20": "skis",
32
+ "21": "4",
33
+ "22": "toothbrush",
34
+ "23": "surfboard",
35
+ "24": "Behind",
36
+ "25": "person",
37
+ "26": "orange",
38
+ "27": "chair",
39
+ "28": "handbag",
40
+ "29": "cow",
41
+ "30": "1",
42
+ "31": "fire hydrant",
43
+ "32": "oven",
44
+ "33": "tennis racket",
45
+ "34": "Yellow",
46
+ "35": "remote",
47
+ "36": "9",
48
+ "37": "No",
49
+ "38": "0",
50
+ "39": "carrot",
51
+ "40": "Above",
52
+ "41": "sports ball",
53
+ "42": "Purple",
54
+ "43": "snowboard",
55
+ "44": "parking meter",
56
+ "45": "mouse",
57
+ "46": "White",
58
+ "47": "clock",
59
+ "48": "dining table",
60
+ "49": "wine glass",
61
+ "50": "car",
62
+ "51": "teddy bear",
63
+ "52": "bicycle",
64
+ "53": "scissors",
65
+ "54": "keyboard",
66
+ "55": "Green",
67
+ "56": "Below",
68
+ "57": "Left",
69
+ "58": "book",
70
+ "59": "Front",
71
+ "60": "fork",
72
+ "61": "broccoli",
73
+ "62": "giraffe",
74
+ "63": "baseball glove",
75
+ "64": "Red",
76
+ "65": "Orange",
77
+ "66": "Black",
78
+ "67": "frisbee",
79
+ "68": "potted plant",
80
+ "69": "sandwich",
81
+ "70": "cup",
82
+ "71": "right",
83
+ "72": "cake",
84
+ "73": "Yes",
85
+ "74": "Right",
86
+ "75": "bed",
87
+ "76": "cell phone",
88
+ "77": "skateboard",
89
+ "78": "bowl",
90
+ "79": "truck",
91
+ "80": "donut",
92
+ "81": "above",
93
+ "82": "Brown",
94
+ "83": "kite",
95
+ "84": "cat",
96
+ "85": "traffic light",
97
+ "86": "pizza",
98
+ "87": "sheep",
99
+ "88": "elephant",
100
+ "89": "laptop",
101
+ "90": "refrigerator",
102
+ "91": "6",
103
+ "92": "banana",
104
+ "93": "Blue",
105
+ "94": "hot dog",
106
+ "95": "bear",
107
+ "96": "bird",
108
+ "97": "motorcycle",
109
+ "98": "horse",
110
+ "99": "tv",
111
+ "100": "tie",
112
+ "101": "left",
113
+ "102": "vase",
114
+ "103": "train",
115
+ "104": "baseball bat",
116
+ "105": "stop sign",
117
+ "106": "airplane",
118
+ "107": "bench"
119
  },
120
  "image_size": 384,
121
  "initializer_range": 0.02,
122
  "intermediate_size": 3072,
123
  "label2id": {
124
+ "0": 38,
125
+ "1": 30,
126
+ "2": 11,
127
+ "3": 14,
128
+ "4": 21,
129
+ "5": 0,
130
+ "6": 91,
131
+ "7": 17,
132
+ "8": 16,
133
+ "9": 36,
134
+ "Above": 40,
135
+ "Behind": 24,
136
+ "Below": 56,
137
+ "Black": 66,
138
+ "Blue": 93,
139
+ "Brown": 82,
140
+ "Front": 59,
141
+ "Green": 55,
142
+ "Grey": 9,
143
+ "Left": 57,
144
+ "No": 37,
145
+ "Orange": 65,
146
+ "Pink": 19,
147
+ "Purple": 42,
148
+ "Red": 64,
149
+ "Right": 74,
150
+ "White": 46,
151
+ "Yellow": 34,
152
+ "Yes": 73,
153
+ "above": 81,
154
+ "airplane": 106,
155
+ "apple": 4,
156
+ "banana": 92,
157
+ "baseball bat": 104,
158
+ "baseball glove": 63,
159
+ "bear": 95,
160
+ "bed": 75,
161
+ "bench": 107,
162
+ "bicycle": 52,
163
+ "bird": 96,
164
+ "boat": 8,
165
+ "book": 58,
166
+ "bottle": 12,
167
+ "bowl": 78,
168
+ "broccoli": 61,
169
+ "bus": 6,
170
+ "cake": 72,
171
+ "car": 50,
172
+ "carrot": 39,
173
+ "cat": 84,
174
+ "cell phone": 76,
175
+ "chair": 27,
176
+ "clock": 47,
177
+ "couch": 2,
178
+ "cow": 29,
179
+ "cup": 70,
180
+ "dining table": 48,
181
+ "dog": 7,
182
+ "donut": 80,
183
+ "elephant": 88,
184
+ "fire hydrant": 31,
185
+ "fork": 60,
186
+ "frisbee": 67,
187
+ "giraffe": 62,
188
+ "handbag": 28,
189
+ "horse": 98,
190
+ "hot dog": 94,
191
+ "keyboard": 54,
192
+ "kite": 83,
193
+ "knife": 18,
194
+ "laptop": 89,
195
+ "left": 101,
196
+ "motorcycle": 97,
197
+ "mouse": 45,
198
+ "orange": 26,
199
+ "oven": 32,
200
+ "parking meter": 44,
201
+ "person": 25,
202
+ "pizza": 86,
203
+ "potted plant": 68,
204
+ "refrigerator": 90,
205
+ "remote": 35,
206
+ "right": 71,
207
+ "sandwich": 69,
208
+ "scissors": 53,
209
+ "sheep": 87,
210
+ "sink": 3,
211
+ "skateboard": 77,
212
+ "skis": 20,
213
+ "snowboard": 43,
214
+ "spoon": 13,
215
+ "sports ball": 41,
216
+ "stop sign": 105,
217
+ "suitcase": 10,
218
+ "surfboard": 23,
219
+ "teddy bear": 51,
220
+ "tennis racket": 33,
221
+ "tie": 100,
222
+ "toilet": 15,
223
+ "toothbrush": 22,
224
+ "traffic light": 85,
225
+ "train": 103,
226
+ "truck": 79,
227
+ "tv": 99,
228
+ "umbrella": 5,
229
+ "vase": 102,
230
+ "wine glass": 49,
231
+ "zebra": 1
232
  },
233
  "layer_norm_eps": 1e-12,
234
  "max_image_length": -1,
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:e3bb26ca2b050db776191062c079d1e51f24c68974255e0069216db2b96a384b
3
  size 446736704
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8c1fef7cdefb7b719d44d7f77a38afcc57b092e5a2a0e6ac7da5c8eebbbd43a4
3
  size 446736704