prithivMLmods commited on
Commit
c44fc50
·
verified ·
1 Parent(s): 139ce92

Upload folder using huggingface_hub

Browse files
checkpoint-2368/config.json ADDED
@@ -0,0 +1,244 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "SiglipForImageClassification"
4
+ ],
5
+ "id2label": {
6
+ "0": "apple_pie",
7
+ "1": "baby_back_ribs",
8
+ "2": "baklava",
9
+ "3": "beef_carpaccio",
10
+ "4": "beef_tartare",
11
+ "5": "beet_salad",
12
+ "6": "beignets",
13
+ "7": "bibimbap",
14
+ "8": "bread_pudding",
15
+ "9": "breakfast_burrito",
16
+ "10": "bruschetta",
17
+ "11": "caesar_salad",
18
+ "12": "cannoli",
19
+ "13": "caprese_salad",
20
+ "14": "carrot_cake",
21
+ "15": "ceviche",
22
+ "16": "cheesecake",
23
+ "17": "cheese_plate",
24
+ "18": "chicken_curry",
25
+ "19": "chicken_quesadilla",
26
+ "20": "chicken_wings",
27
+ "21": "chocolate_cake",
28
+ "22": "chocolate_mousse",
29
+ "23": "churros",
30
+ "24": "clam_chowder",
31
+ "25": "club_sandwich",
32
+ "26": "crab_cakes",
33
+ "27": "creme_brulee",
34
+ "28": "croque_madame",
35
+ "29": "cup_cakes",
36
+ "30": "deviled_eggs",
37
+ "31": "donuts",
38
+ "32": "dumplings",
39
+ "33": "edamame",
40
+ "34": "eggs_benedict",
41
+ "35": "escargots",
42
+ "36": "falafel",
43
+ "37": "filet_mignon",
44
+ "38": "fish_and_chips",
45
+ "39": "foie_gras",
46
+ "40": "french_fries",
47
+ "41": "french_onion_soup",
48
+ "42": "french_toast",
49
+ "43": "fried_calamari",
50
+ "44": "fried_rice",
51
+ "45": "frozen_yogurt",
52
+ "46": "garlic_bread",
53
+ "47": "gnocchi",
54
+ "48": "greek_salad",
55
+ "49": "grilled_cheese_sandwich",
56
+ "50": "grilled_salmon",
57
+ "51": "guacamole",
58
+ "52": "gyoza",
59
+ "53": "hamburger",
60
+ "54": "hot_and_sour_soup",
61
+ "55": "hot_dog",
62
+ "56": "huevos_rancheros",
63
+ "57": "hummus",
64
+ "58": "ice_cream",
65
+ "59": "lasagna",
66
+ "60": "lobster_bisque",
67
+ "61": "lobster_roll_sandwich",
68
+ "62": "macaroni_and_cheese",
69
+ "63": "macarons",
70
+ "64": "miso_soup",
71
+ "65": "mussels",
72
+ "66": "nachos",
73
+ "67": "omelette",
74
+ "68": "onion_rings",
75
+ "69": "oysters",
76
+ "70": "pad_thai",
77
+ "71": "paella",
78
+ "72": "pancakes",
79
+ "73": "panna_cotta",
80
+ "74": "peking_duck",
81
+ "75": "pho",
82
+ "76": "pizza",
83
+ "77": "pork_chop",
84
+ "78": "poutine",
85
+ "79": "prime_rib",
86
+ "80": "pulled_pork_sandwich",
87
+ "81": "ramen",
88
+ "82": "ravioli",
89
+ "83": "red_velvet_cake",
90
+ "84": "risotto",
91
+ "85": "samosa",
92
+ "86": "sashimi",
93
+ "87": "scallops",
94
+ "88": "seaweed_salad",
95
+ "89": "shrimp_and_grits",
96
+ "90": "spaghetti_bolognese",
97
+ "91": "spaghetti_carbonara",
98
+ "92": "spring_rolls",
99
+ "93": "steak",
100
+ "94": "strawberry_shortcake",
101
+ "95": "sushi",
102
+ "96": "tacos",
103
+ "97": "takoyaki",
104
+ "98": "tiramisu",
105
+ "99": "tuna_tartare",
106
+ "100": "waffles"
107
+ },
108
+ "initializer_factor": 1.0,
109
+ "label2id": {
110
+ "apple_pie": 0,
111
+ "baby_back_ribs": 1,
112
+ "baklava": 2,
113
+ "beef_carpaccio": 3,
114
+ "beef_tartare": 4,
115
+ "beet_salad": 5,
116
+ "beignets": 6,
117
+ "bibimbap": 7,
118
+ "bread_pudding": 8,
119
+ "breakfast_burrito": 9,
120
+ "bruschetta": 10,
121
+ "caesar_salad": 11,
122
+ "cannoli": 12,
123
+ "caprese_salad": 13,
124
+ "carrot_cake": 14,
125
+ "ceviche": 15,
126
+ "cheese_plate": 17,
127
+ "cheesecake": 16,
128
+ "chicken_curry": 18,
129
+ "chicken_quesadilla": 19,
130
+ "chicken_wings": 20,
131
+ "chocolate_cake": 21,
132
+ "chocolate_mousse": 22,
133
+ "churros": 23,
134
+ "clam_chowder": 24,
135
+ "club_sandwich": 25,
136
+ "crab_cakes": 26,
137
+ "creme_brulee": 27,
138
+ "croque_madame": 28,
139
+ "cup_cakes": 29,
140
+ "deviled_eggs": 30,
141
+ "donuts": 31,
142
+ "dumplings": 32,
143
+ "edamame": 33,
144
+ "eggs_benedict": 34,
145
+ "escargots": 35,
146
+ "falafel": 36,
147
+ "filet_mignon": 37,
148
+ "fish_and_chips": 38,
149
+ "foie_gras": 39,
150
+ "french_fries": 40,
151
+ "french_onion_soup": 41,
152
+ "french_toast": 42,
153
+ "fried_calamari": 43,
154
+ "fried_rice": 44,
155
+ "frozen_yogurt": 45,
156
+ "garlic_bread": 46,
157
+ "gnocchi": 47,
158
+ "greek_salad": 48,
159
+ "grilled_cheese_sandwich": 49,
160
+ "grilled_salmon": 50,
161
+ "guacamole": 51,
162
+ "gyoza": 52,
163
+ "hamburger": 53,
164
+ "hot_and_sour_soup": 54,
165
+ "hot_dog": 55,
166
+ "huevos_rancheros": 56,
167
+ "hummus": 57,
168
+ "ice_cream": 58,
169
+ "lasagna": 59,
170
+ "lobster_bisque": 60,
171
+ "lobster_roll_sandwich": 61,
172
+ "macaroni_and_cheese": 62,
173
+ "macarons": 63,
174
+ "miso_soup": 64,
175
+ "mussels": 65,
176
+ "nachos": 66,
177
+ "omelette": 67,
178
+ "onion_rings": 68,
179
+ "oysters": 69,
180
+ "pad_thai": 70,
181
+ "paella": 71,
182
+ "pancakes": 72,
183
+ "panna_cotta": 73,
184
+ "peking_duck": 74,
185
+ "pho": 75,
186
+ "pizza": 76,
187
+ "pork_chop": 77,
188
+ "poutine": 78,
189
+ "prime_rib": 79,
190
+ "pulled_pork_sandwich": 80,
191
+ "ramen": 81,
192
+ "ravioli": 82,
193
+ "red_velvet_cake": 83,
194
+ "risotto": 84,
195
+ "samosa": 85,
196
+ "sashimi": 86,
197
+ "scallops": 87,
198
+ "seaweed_salad": 88,
199
+ "shrimp_and_grits": 89,
200
+ "spaghetti_bolognese": 90,
201
+ "spaghetti_carbonara": 91,
202
+ "spring_rolls": 92,
203
+ "steak": 93,
204
+ "strawberry_shortcake": 94,
205
+ "sushi": 95,
206
+ "tacos": 96,
207
+ "takoyaki": 97,
208
+ "tiramisu": 98,
209
+ "tuna_tartare": 99,
210
+ "waffles": 100
211
+ },
212
+ "model_type": "siglip",
213
+ "problem_type": "single_label_classification",
214
+ "text_config": {
215
+ "attention_dropout": 0.0,
216
+ "hidden_act": "gelu_pytorch_tanh",
217
+ "hidden_size": 768,
218
+ "intermediate_size": 3072,
219
+ "layer_norm_eps": 1e-06,
220
+ "max_position_embeddings": 64,
221
+ "model_type": "siglip_text_model",
222
+ "num_attention_heads": 12,
223
+ "num_hidden_layers": 12,
224
+ "projection_size": 768,
225
+ "torch_dtype": "float32",
226
+ "vocab_size": 256000
227
+ },
228
+ "torch_dtype": "float32",
229
+ "transformers_version": "4.50.3",
230
+ "vision_config": {
231
+ "attention_dropout": 0.0,
232
+ "hidden_act": "gelu_pytorch_tanh",
233
+ "hidden_size": 768,
234
+ "image_size": 224,
235
+ "intermediate_size": 3072,
236
+ "layer_norm_eps": 1e-06,
237
+ "model_type": "siglip_vision_model",
238
+ "num_attention_heads": 12,
239
+ "num_channels": 3,
240
+ "num_hidden_layers": 12,
241
+ "patch_size": 16,
242
+ "torch_dtype": "float32"
243
+ }
244
+ }
checkpoint-2368/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0aa447b4489d9c1f5db717d82008966d6f8103c1098852ef985ae2d3995653fa
3
+ size 371872532
checkpoint-2368/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:91affd3f043cd403cf52afe66a0afd6be8b09eb9feac83095f20dbf2c32c5bd8
3
+ size 687164794
checkpoint-2368/preprocessor_config.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_convert_rgb": null,
3
+ "do_normalize": true,
4
+ "do_rescale": true,
5
+ "do_resize": true,
6
+ "image_mean": [
7
+ 0.5,
8
+ 0.5,
9
+ 0.5
10
+ ],
11
+ "image_processor_type": "SiglipImageProcessor",
12
+ "image_std": [
13
+ 0.5,
14
+ 0.5,
15
+ 0.5
16
+ ],
17
+ "processor_class": "SiglipProcessor",
18
+ "resample": 2,
19
+ "rescale_factor": 0.00392156862745098,
20
+ "size": {
21
+ "height": 224,
22
+ "width": 224
23
+ }
24
+ }
checkpoint-2368/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:41595153e4598922797ba88b62b653b9c566192aef59376b74d303b0be7696ae
3
+ size 14244
checkpoint-2368/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:824b0c8fb395e6145e807444cf700c691550207341c3bc69b04845ec42ea5a6b
3
+ size 1064
checkpoint-2368/trainer_state.json ADDED
@@ -0,0 +1,72 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": 2368,
3
+ "best_metric": 0.5366769433021545,
4
+ "best_model_checkpoint": "siglip2-finetune-full/checkpoint-2368",
5
+ "epoch": 1.0,
6
+ "eval_steps": 500,
7
+ "global_step": 2368,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.21114864864864866,
14
+ "grad_norm": 43.60944366455078,
15
+ "learning_rate": 4.519846350832267e-06,
16
+ "loss": 3.0569,
17
+ "step": 500
18
+ },
19
+ {
20
+ "epoch": 0.4222972972972973,
21
+ "grad_norm": 62.890960693359375,
22
+ "learning_rate": 3.986342296201451e-06,
23
+ "loss": 1.2464,
24
+ "step": 1000
25
+ },
26
+ {
27
+ "epoch": 0.6334459459459459,
28
+ "grad_norm": 70.39923858642578,
29
+ "learning_rate": 3.4528382415706363e-06,
30
+ "loss": 0.9218,
31
+ "step": 1500
32
+ },
33
+ {
34
+ "epoch": 0.8445945945945946,
35
+ "grad_norm": 54.272945404052734,
36
+ "learning_rate": 2.919334186939821e-06,
37
+ "loss": 0.7906,
38
+ "step": 2000
39
+ },
40
+ {
41
+ "epoch": 1.0,
42
+ "eval_accuracy": 0.868950495049505,
43
+ "eval_loss": 0.5366769433021545,
44
+ "eval_model_preparation_time": 0.0022,
45
+ "eval_runtime": 1066.6007,
46
+ "eval_samples_per_second": 71.02,
47
+ "eval_steps_per_second": 8.878,
48
+ "step": 2368
49
+ }
50
+ ],
51
+ "logging_steps": 500,
52
+ "max_steps": 4736,
53
+ "num_input_tokens_seen": 0,
54
+ "num_train_epochs": 2,
55
+ "save_steps": 500,
56
+ "stateful_callbacks": {
57
+ "TrainerControl": {
58
+ "args": {
59
+ "should_epoch_stop": false,
60
+ "should_evaluate": false,
61
+ "should_log": false,
62
+ "should_save": true,
63
+ "should_training_stop": false
64
+ },
65
+ "attributes": {}
66
+ }
67
+ },
68
+ "total_flos": 6.34968730900224e+18,
69
+ "train_batch_size": 32,
70
+ "trial_name": null,
71
+ "trial_params": null
72
+ }
checkpoint-2368/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1bff73d072beca8c44b79880bc2556032f2ac7f5431580d9d5e02d1c742b4426
3
+ size 5304
checkpoint-4736/config.json ADDED
@@ -0,0 +1,244 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "SiglipForImageClassification"
4
+ ],
5
+ "id2label": {
6
+ "0": "apple_pie",
7
+ "1": "baby_back_ribs",
8
+ "2": "baklava",
9
+ "3": "beef_carpaccio",
10
+ "4": "beef_tartare",
11
+ "5": "beet_salad",
12
+ "6": "beignets",
13
+ "7": "bibimbap",
14
+ "8": "bread_pudding",
15
+ "9": "breakfast_burrito",
16
+ "10": "bruschetta",
17
+ "11": "caesar_salad",
18
+ "12": "cannoli",
19
+ "13": "caprese_salad",
20
+ "14": "carrot_cake",
21
+ "15": "ceviche",
22
+ "16": "cheesecake",
23
+ "17": "cheese_plate",
24
+ "18": "chicken_curry",
25
+ "19": "chicken_quesadilla",
26
+ "20": "chicken_wings",
27
+ "21": "chocolate_cake",
28
+ "22": "chocolate_mousse",
29
+ "23": "churros",
30
+ "24": "clam_chowder",
31
+ "25": "club_sandwich",
32
+ "26": "crab_cakes",
33
+ "27": "creme_brulee",
34
+ "28": "croque_madame",
35
+ "29": "cup_cakes",
36
+ "30": "deviled_eggs",
37
+ "31": "donuts",
38
+ "32": "dumplings",
39
+ "33": "edamame",
40
+ "34": "eggs_benedict",
41
+ "35": "escargots",
42
+ "36": "falafel",
43
+ "37": "filet_mignon",
44
+ "38": "fish_and_chips",
45
+ "39": "foie_gras",
46
+ "40": "french_fries",
47
+ "41": "french_onion_soup",
48
+ "42": "french_toast",
49
+ "43": "fried_calamari",
50
+ "44": "fried_rice",
51
+ "45": "frozen_yogurt",
52
+ "46": "garlic_bread",
53
+ "47": "gnocchi",
54
+ "48": "greek_salad",
55
+ "49": "grilled_cheese_sandwich",
56
+ "50": "grilled_salmon",
57
+ "51": "guacamole",
58
+ "52": "gyoza",
59
+ "53": "hamburger",
60
+ "54": "hot_and_sour_soup",
61
+ "55": "hot_dog",
62
+ "56": "huevos_rancheros",
63
+ "57": "hummus",
64
+ "58": "ice_cream",
65
+ "59": "lasagna",
66
+ "60": "lobster_bisque",
67
+ "61": "lobster_roll_sandwich",
68
+ "62": "macaroni_and_cheese",
69
+ "63": "macarons",
70
+ "64": "miso_soup",
71
+ "65": "mussels",
72
+ "66": "nachos",
73
+ "67": "omelette",
74
+ "68": "onion_rings",
75
+ "69": "oysters",
76
+ "70": "pad_thai",
77
+ "71": "paella",
78
+ "72": "pancakes",
79
+ "73": "panna_cotta",
80
+ "74": "peking_duck",
81
+ "75": "pho",
82
+ "76": "pizza",
83
+ "77": "pork_chop",
84
+ "78": "poutine",
85
+ "79": "prime_rib",
86
+ "80": "pulled_pork_sandwich",
87
+ "81": "ramen",
88
+ "82": "ravioli",
89
+ "83": "red_velvet_cake",
90
+ "84": "risotto",
91
+ "85": "samosa",
92
+ "86": "sashimi",
93
+ "87": "scallops",
94
+ "88": "seaweed_salad",
95
+ "89": "shrimp_and_grits",
96
+ "90": "spaghetti_bolognese",
97
+ "91": "spaghetti_carbonara",
98
+ "92": "spring_rolls",
99
+ "93": "steak",
100
+ "94": "strawberry_shortcake",
101
+ "95": "sushi",
102
+ "96": "tacos",
103
+ "97": "takoyaki",
104
+ "98": "tiramisu",
105
+ "99": "tuna_tartare",
106
+ "100": "waffles"
107
+ },
108
+ "initializer_factor": 1.0,
109
+ "label2id": {
110
+ "apple_pie": 0,
111
+ "baby_back_ribs": 1,
112
+ "baklava": 2,
113
+ "beef_carpaccio": 3,
114
+ "beef_tartare": 4,
115
+ "beet_salad": 5,
116
+ "beignets": 6,
117
+ "bibimbap": 7,
118
+ "bread_pudding": 8,
119
+ "breakfast_burrito": 9,
120
+ "bruschetta": 10,
121
+ "caesar_salad": 11,
122
+ "cannoli": 12,
123
+ "caprese_salad": 13,
124
+ "carrot_cake": 14,
125
+ "ceviche": 15,
126
+ "cheese_plate": 17,
127
+ "cheesecake": 16,
128
+ "chicken_curry": 18,
129
+ "chicken_quesadilla": 19,
130
+ "chicken_wings": 20,
131
+ "chocolate_cake": 21,
132
+ "chocolate_mousse": 22,
133
+ "churros": 23,
134
+ "clam_chowder": 24,
135
+ "club_sandwich": 25,
136
+ "crab_cakes": 26,
137
+ "creme_brulee": 27,
138
+ "croque_madame": 28,
139
+ "cup_cakes": 29,
140
+ "deviled_eggs": 30,
141
+ "donuts": 31,
142
+ "dumplings": 32,
143
+ "edamame": 33,
144
+ "eggs_benedict": 34,
145
+ "escargots": 35,
146
+ "falafel": 36,
147
+ "filet_mignon": 37,
148
+ "fish_and_chips": 38,
149
+ "foie_gras": 39,
150
+ "french_fries": 40,
151
+ "french_onion_soup": 41,
152
+ "french_toast": 42,
153
+ "fried_calamari": 43,
154
+ "fried_rice": 44,
155
+ "frozen_yogurt": 45,
156
+ "garlic_bread": 46,
157
+ "gnocchi": 47,
158
+ "greek_salad": 48,
159
+ "grilled_cheese_sandwich": 49,
160
+ "grilled_salmon": 50,
161
+ "guacamole": 51,
162
+ "gyoza": 52,
163
+ "hamburger": 53,
164
+ "hot_and_sour_soup": 54,
165
+ "hot_dog": 55,
166
+ "huevos_rancheros": 56,
167
+ "hummus": 57,
168
+ "ice_cream": 58,
169
+ "lasagna": 59,
170
+ "lobster_bisque": 60,
171
+ "lobster_roll_sandwich": 61,
172
+ "macaroni_and_cheese": 62,
173
+ "macarons": 63,
174
+ "miso_soup": 64,
175
+ "mussels": 65,
176
+ "nachos": 66,
177
+ "omelette": 67,
178
+ "onion_rings": 68,
179
+ "oysters": 69,
180
+ "pad_thai": 70,
181
+ "paella": 71,
182
+ "pancakes": 72,
183
+ "panna_cotta": 73,
184
+ "peking_duck": 74,
185
+ "pho": 75,
186
+ "pizza": 76,
187
+ "pork_chop": 77,
188
+ "poutine": 78,
189
+ "prime_rib": 79,
190
+ "pulled_pork_sandwich": 80,
191
+ "ramen": 81,
192
+ "ravioli": 82,
193
+ "red_velvet_cake": 83,
194
+ "risotto": 84,
195
+ "samosa": 85,
196
+ "sashimi": 86,
197
+ "scallops": 87,
198
+ "seaweed_salad": 88,
199
+ "shrimp_and_grits": 89,
200
+ "spaghetti_bolognese": 90,
201
+ "spaghetti_carbonara": 91,
202
+ "spring_rolls": 92,
203
+ "steak": 93,
204
+ "strawberry_shortcake": 94,
205
+ "sushi": 95,
206
+ "tacos": 96,
207
+ "takoyaki": 97,
208
+ "tiramisu": 98,
209
+ "tuna_tartare": 99,
210
+ "waffles": 100
211
+ },
212
+ "model_type": "siglip",
213
+ "problem_type": "single_label_classification",
214
+ "text_config": {
215
+ "attention_dropout": 0.0,
216
+ "hidden_act": "gelu_pytorch_tanh",
217
+ "hidden_size": 768,
218
+ "intermediate_size": 3072,
219
+ "layer_norm_eps": 1e-06,
220
+ "max_position_embeddings": 64,
221
+ "model_type": "siglip_text_model",
222
+ "num_attention_heads": 12,
223
+ "num_hidden_layers": 12,
224
+ "projection_size": 768,
225
+ "torch_dtype": "float32",
226
+ "vocab_size": 256000
227
+ },
228
+ "torch_dtype": "float32",
229
+ "transformers_version": "4.50.3",
230
+ "vision_config": {
231
+ "attention_dropout": 0.0,
232
+ "hidden_act": "gelu_pytorch_tanh",
233
+ "hidden_size": 768,
234
+ "image_size": 224,
235
+ "intermediate_size": 3072,
236
+ "layer_norm_eps": 1e-06,
237
+ "model_type": "siglip_vision_model",
238
+ "num_attention_heads": 12,
239
+ "num_channels": 3,
240
+ "num_hidden_layers": 12,
241
+ "patch_size": 16,
242
+ "torch_dtype": "float32"
243
+ }
244
+ }
checkpoint-4736/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d7fae1d0598461e45f85f8f3b2f69cdfc51fec5536eb8f676c097dfb80520766
3
+ size 371872532
checkpoint-4736/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f50d64f683e80f4dfa9d89e21f69dd6ac8f708560bb690b5c062e765199595ef
3
+ size 687164794
checkpoint-4736/preprocessor_config.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_convert_rgb": null,
3
+ "do_normalize": true,
4
+ "do_rescale": true,
5
+ "do_resize": true,
6
+ "image_mean": [
7
+ 0.5,
8
+ 0.5,
9
+ 0.5
10
+ ],
11
+ "image_processor_type": "SiglipImageProcessor",
12
+ "image_std": [
13
+ 0.5,
14
+ 0.5,
15
+ 0.5
16
+ ],
17
+ "processor_class": "SiglipProcessor",
18
+ "resample": 2,
19
+ "rescale_factor": 0.00392156862745098,
20
+ "size": {
21
+ "height": 224,
22
+ "width": 224
23
+ }
24
+ }
checkpoint-4736/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dbd23a562df66feb0bd0116acbaf40f545a693f33ad9a6e4bfad4ee97debb9a7
3
+ size 14244
checkpoint-4736/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2513ae2483818d0c5b5f28cb36018285f158e6678951dd0a9dfe7c8ce1ee8fa6
3
+ size 1064
checkpoint-4736/trainer_state.json ADDED
@@ -0,0 +1,117 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": 4736,
3
+ "best_metric": 0.40901073813438416,
4
+ "best_model_checkpoint": "siglip2-finetune-full/checkpoint-4736",
5
+ "epoch": 2.0,
6
+ "eval_steps": 500,
7
+ "global_step": 4736,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.21114864864864866,
14
+ "grad_norm": 43.60944366455078,
15
+ "learning_rate": 4.519846350832267e-06,
16
+ "loss": 3.0569,
17
+ "step": 500
18
+ },
19
+ {
20
+ "epoch": 0.4222972972972973,
21
+ "grad_norm": 62.890960693359375,
22
+ "learning_rate": 3.986342296201451e-06,
23
+ "loss": 1.2464,
24
+ "step": 1000
25
+ },
26
+ {
27
+ "epoch": 0.6334459459459459,
28
+ "grad_norm": 70.39923858642578,
29
+ "learning_rate": 3.4528382415706363e-06,
30
+ "loss": 0.9218,
31
+ "step": 1500
32
+ },
33
+ {
34
+ "epoch": 0.8445945945945946,
35
+ "grad_norm": 54.272945404052734,
36
+ "learning_rate": 2.919334186939821e-06,
37
+ "loss": 0.7906,
38
+ "step": 2000
39
+ },
40
+ {
41
+ "epoch": 1.0,
42
+ "eval_accuracy": 0.868950495049505,
43
+ "eval_loss": 0.5366769433021545,
44
+ "eval_model_preparation_time": 0.0022,
45
+ "eval_runtime": 1066.6007,
46
+ "eval_samples_per_second": 71.02,
47
+ "eval_steps_per_second": 8.878,
48
+ "step": 2368
49
+ },
50
+ {
51
+ "epoch": 1.0557432432432432,
52
+ "grad_norm": 55.69829559326172,
53
+ "learning_rate": 2.385830132309006e-06,
54
+ "loss": 0.7005,
55
+ "step": 2500
56
+ },
57
+ {
58
+ "epoch": 1.2668918918918919,
59
+ "grad_norm": 57.688602447509766,
60
+ "learning_rate": 1.8523260776781904e-06,
61
+ "loss": 0.5934,
62
+ "step": 3000
63
+ },
64
+ {
65
+ "epoch": 1.4780405405405406,
66
+ "grad_norm": 77.93653869628906,
67
+ "learning_rate": 1.3188220230473753e-06,
68
+ "loss": 0.5654,
69
+ "step": 3500
70
+ },
71
+ {
72
+ "epoch": 1.689189189189189,
73
+ "grad_norm": 48.80607223510742,
74
+ "learning_rate": 7.8531796841656e-07,
75
+ "loss": 0.5406,
76
+ "step": 4000
77
+ },
78
+ {
79
+ "epoch": 1.9003378378378377,
80
+ "grad_norm": 57.90318298339844,
81
+ "learning_rate": 2.518139137857448e-07,
82
+ "loss": 0.5038,
83
+ "step": 4500
84
+ },
85
+ {
86
+ "epoch": 2.0,
87
+ "eval_accuracy": 0.8973333333333333,
88
+ "eval_loss": 0.40901073813438416,
89
+ "eval_model_preparation_time": 0.0022,
90
+ "eval_runtime": 1066.4011,
91
+ "eval_samples_per_second": 71.033,
92
+ "eval_steps_per_second": 8.879,
93
+ "step": 4736
94
+ }
95
+ ],
96
+ "logging_steps": 500,
97
+ "max_steps": 4736,
98
+ "num_input_tokens_seen": 0,
99
+ "num_train_epochs": 2,
100
+ "save_steps": 500,
101
+ "stateful_callbacks": {
102
+ "TrainerControl": {
103
+ "args": {
104
+ "should_epoch_stop": false,
105
+ "should_evaluate": false,
106
+ "should_log": false,
107
+ "should_save": true,
108
+ "should_training_stop": true
109
+ },
110
+ "attributes": {}
111
+ }
112
+ },
113
+ "total_flos": 1.269937461800448e+19,
114
+ "train_batch_size": 32,
115
+ "trial_name": null,
116
+ "trial_params": null
117
+ }
checkpoint-4736/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1bff73d072beca8c44b79880bc2556032f2ac7f5431580d9d5e02d1c742b4426
3
+ size 5304
config.json ADDED
@@ -0,0 +1,244 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "SiglipForImageClassification"
4
+ ],
5
+ "id2label": {
6
+ "0": "apple_pie",
7
+ "1": "baby_back_ribs",
8
+ "2": "baklava",
9
+ "3": "beef_carpaccio",
10
+ "4": "beef_tartare",
11
+ "5": "beet_salad",
12
+ "6": "beignets",
13
+ "7": "bibimbap",
14
+ "8": "bread_pudding",
15
+ "9": "breakfast_burrito",
16
+ "10": "bruschetta",
17
+ "11": "caesar_salad",
18
+ "12": "cannoli",
19
+ "13": "caprese_salad",
20
+ "14": "carrot_cake",
21
+ "15": "ceviche",
22
+ "16": "cheesecake",
23
+ "17": "cheese_plate",
24
+ "18": "chicken_curry",
25
+ "19": "chicken_quesadilla",
26
+ "20": "chicken_wings",
27
+ "21": "chocolate_cake",
28
+ "22": "chocolate_mousse",
29
+ "23": "churros",
30
+ "24": "clam_chowder",
31
+ "25": "club_sandwich",
32
+ "26": "crab_cakes",
33
+ "27": "creme_brulee",
34
+ "28": "croque_madame",
35
+ "29": "cup_cakes",
36
+ "30": "deviled_eggs",
37
+ "31": "donuts",
38
+ "32": "dumplings",
39
+ "33": "edamame",
40
+ "34": "eggs_benedict",
41
+ "35": "escargots",
42
+ "36": "falafel",
43
+ "37": "filet_mignon",
44
+ "38": "fish_and_chips",
45
+ "39": "foie_gras",
46
+ "40": "french_fries",
47
+ "41": "french_onion_soup",
48
+ "42": "french_toast",
49
+ "43": "fried_calamari",
50
+ "44": "fried_rice",
51
+ "45": "frozen_yogurt",
52
+ "46": "garlic_bread",
53
+ "47": "gnocchi",
54
+ "48": "greek_salad",
55
+ "49": "grilled_cheese_sandwich",
56
+ "50": "grilled_salmon",
57
+ "51": "guacamole",
58
+ "52": "gyoza",
59
+ "53": "hamburger",
60
+ "54": "hot_and_sour_soup",
61
+ "55": "hot_dog",
62
+ "56": "huevos_rancheros",
63
+ "57": "hummus",
64
+ "58": "ice_cream",
65
+ "59": "lasagna",
66
+ "60": "lobster_bisque",
67
+ "61": "lobster_roll_sandwich",
68
+ "62": "macaroni_and_cheese",
69
+ "63": "macarons",
70
+ "64": "miso_soup",
71
+ "65": "mussels",
72
+ "66": "nachos",
73
+ "67": "omelette",
74
+ "68": "onion_rings",
75
+ "69": "oysters",
76
+ "70": "pad_thai",
77
+ "71": "paella",
78
+ "72": "pancakes",
79
+ "73": "panna_cotta",
80
+ "74": "peking_duck",
81
+ "75": "pho",
82
+ "76": "pizza",
83
+ "77": "pork_chop",
84
+ "78": "poutine",
85
+ "79": "prime_rib",
86
+ "80": "pulled_pork_sandwich",
87
+ "81": "ramen",
88
+ "82": "ravioli",
89
+ "83": "red_velvet_cake",
90
+ "84": "risotto",
91
+ "85": "samosa",
92
+ "86": "sashimi",
93
+ "87": "scallops",
94
+ "88": "seaweed_salad",
95
+ "89": "shrimp_and_grits",
96
+ "90": "spaghetti_bolognese",
97
+ "91": "spaghetti_carbonara",
98
+ "92": "spring_rolls",
99
+ "93": "steak",
100
+ "94": "strawberry_shortcake",
101
+ "95": "sushi",
102
+ "96": "tacos",
103
+ "97": "takoyaki",
104
+ "98": "tiramisu",
105
+ "99": "tuna_tartare",
106
+ "100": "waffles"
107
+ },
108
+ "initializer_factor": 1.0,
109
+ "label2id": {
110
+ "apple_pie": 0,
111
+ "baby_back_ribs": 1,
112
+ "baklava": 2,
113
+ "beef_carpaccio": 3,
114
+ "beef_tartare": 4,
115
+ "beet_salad": 5,
116
+ "beignets": 6,
117
+ "bibimbap": 7,
118
+ "bread_pudding": 8,
119
+ "breakfast_burrito": 9,
120
+ "bruschetta": 10,
121
+ "caesar_salad": 11,
122
+ "cannoli": 12,
123
+ "caprese_salad": 13,
124
+ "carrot_cake": 14,
125
+ "ceviche": 15,
126
+ "cheese_plate": 17,
127
+ "cheesecake": 16,
128
+ "chicken_curry": 18,
129
+ "chicken_quesadilla": 19,
130
+ "chicken_wings": 20,
131
+ "chocolate_cake": 21,
132
+ "chocolate_mousse": 22,
133
+ "churros": 23,
134
+ "clam_chowder": 24,
135
+ "club_sandwich": 25,
136
+ "crab_cakes": 26,
137
+ "creme_brulee": 27,
138
+ "croque_madame": 28,
139
+ "cup_cakes": 29,
140
+ "deviled_eggs": 30,
141
+ "donuts": 31,
142
+ "dumplings": 32,
143
+ "edamame": 33,
144
+ "eggs_benedict": 34,
145
+ "escargots": 35,
146
+ "falafel": 36,
147
+ "filet_mignon": 37,
148
+ "fish_and_chips": 38,
149
+ "foie_gras": 39,
150
+ "french_fries": 40,
151
+ "french_onion_soup": 41,
152
+ "french_toast": 42,
153
+ "fried_calamari": 43,
154
+ "fried_rice": 44,
155
+ "frozen_yogurt": 45,
156
+ "garlic_bread": 46,
157
+ "gnocchi": 47,
158
+ "greek_salad": 48,
159
+ "grilled_cheese_sandwich": 49,
160
+ "grilled_salmon": 50,
161
+ "guacamole": 51,
162
+ "gyoza": 52,
163
+ "hamburger": 53,
164
+ "hot_and_sour_soup": 54,
165
+ "hot_dog": 55,
166
+ "huevos_rancheros": 56,
167
+ "hummus": 57,
168
+ "ice_cream": 58,
169
+ "lasagna": 59,
170
+ "lobster_bisque": 60,
171
+ "lobster_roll_sandwich": 61,
172
+ "macaroni_and_cheese": 62,
173
+ "macarons": 63,
174
+ "miso_soup": 64,
175
+ "mussels": 65,
176
+ "nachos": 66,
177
+ "omelette": 67,
178
+ "onion_rings": 68,
179
+ "oysters": 69,
180
+ "pad_thai": 70,
181
+ "paella": 71,
182
+ "pancakes": 72,
183
+ "panna_cotta": 73,
184
+ "peking_duck": 74,
185
+ "pho": 75,
186
+ "pizza": 76,
187
+ "pork_chop": 77,
188
+ "poutine": 78,
189
+ "prime_rib": 79,
190
+ "pulled_pork_sandwich": 80,
191
+ "ramen": 81,
192
+ "ravioli": 82,
193
+ "red_velvet_cake": 83,
194
+ "risotto": 84,
195
+ "samosa": 85,
196
+ "sashimi": 86,
197
+ "scallops": 87,
198
+ "seaweed_salad": 88,
199
+ "shrimp_and_grits": 89,
200
+ "spaghetti_bolognese": 90,
201
+ "spaghetti_carbonara": 91,
202
+ "spring_rolls": 92,
203
+ "steak": 93,
204
+ "strawberry_shortcake": 94,
205
+ "sushi": 95,
206
+ "tacos": 96,
207
+ "takoyaki": 97,
208
+ "tiramisu": 98,
209
+ "tuna_tartare": 99,
210
+ "waffles": 100
211
+ },
212
+ "model_type": "siglip",
213
+ "problem_type": "single_label_classification",
214
+ "text_config": {
215
+ "attention_dropout": 0.0,
216
+ "hidden_act": "gelu_pytorch_tanh",
217
+ "hidden_size": 768,
218
+ "intermediate_size": 3072,
219
+ "layer_norm_eps": 1e-06,
220
+ "max_position_embeddings": 64,
221
+ "model_type": "siglip_text_model",
222
+ "num_attention_heads": 12,
223
+ "num_hidden_layers": 12,
224
+ "projection_size": 768,
225
+ "torch_dtype": "float32",
226
+ "vocab_size": 256000
227
+ },
228
+ "torch_dtype": "float32",
229
+ "transformers_version": "4.50.3",
230
+ "vision_config": {
231
+ "attention_dropout": 0.0,
232
+ "hidden_act": "gelu_pytorch_tanh",
233
+ "hidden_size": 768,
234
+ "image_size": 224,
235
+ "intermediate_size": 3072,
236
+ "layer_norm_eps": 1e-06,
237
+ "model_type": "siglip_vision_model",
238
+ "num_attention_heads": 12,
239
+ "num_channels": 3,
240
+ "num_hidden_layers": 12,
241
+ "patch_size": 16,
242
+ "torch_dtype": "float32"
243
+ }
244
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d7fae1d0598461e45f85f8f3b2f69cdfc51fec5536eb8f676c097dfb80520766
3
+ size 371872532
preprocessor_config.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_convert_rgb": null,
3
+ "do_normalize": true,
4
+ "do_rescale": true,
5
+ "do_resize": true,
6
+ "image_mean": [
7
+ 0.5,
8
+ 0.5,
9
+ 0.5
10
+ ],
11
+ "image_processor_type": "SiglipImageProcessor",
12
+ "image_std": [
13
+ 0.5,
14
+ 0.5,
15
+ 0.5
16
+ ],
17
+ "processor_class": "SiglipProcessor",
18
+ "resample": 2,
19
+ "rescale_factor": 0.00392156862745098,
20
+ "size": {
21
+ "height": 224,
22
+ "width": 224
23
+ }
24
+ }
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1bff73d072beca8c44b79880bc2556032f2ac7f5431580d9d5e02d1c742b4426
3
+ size 5304