prithivMLmods commited on
Commit
a61cc38
·
verified ·
1 Parent(s): 415342a

Upload folder using huggingface_hub

Browse files
checkpoint-1280/config.json ADDED
@@ -0,0 +1,284 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "SiglipForImageClassification"
4
+ ],
5
+ "id2label": {
6
+ "0": "affenpinscher",
7
+ "1": "afghan_hound",
8
+ "2": "african_hunting_dog",
9
+ "3": "airedale",
10
+ "4": "american_staffordshire_terrier",
11
+ "5": "appenzeller",
12
+ "6": "australian_terrier",
13
+ "7": "basenji",
14
+ "8": "basset",
15
+ "9": "beagle",
16
+ "10": "bedlington_terrier",
17
+ "11": "bernese_mountain_dog",
18
+ "12": "black-and-tan_coonhound",
19
+ "13": "blenheim_spaniel",
20
+ "14": "bloodhound",
21
+ "15": "bluetick",
22
+ "16": "border_collie",
23
+ "17": "border_terrier",
24
+ "18": "borzoi",
25
+ "19": "boston_bull",
26
+ "20": "bouvier_des_flandres",
27
+ "21": "boxer",
28
+ "22": "brabancon_griffon",
29
+ "23": "briard",
30
+ "24": "brittany_spaniel",
31
+ "25": "bull_mastiff",
32
+ "26": "cairn",
33
+ "27": "cardigan",
34
+ "28": "chesapeake_bay_retriever",
35
+ "29": "chihuahua",
36
+ "30": "chow",
37
+ "31": "clumber",
38
+ "32": "cocker_spaniel",
39
+ "33": "collie",
40
+ "34": "curly-coated_retriever",
41
+ "35": "dandie_dinmont",
42
+ "36": "dhole",
43
+ "37": "dingo",
44
+ "38": "doberman",
45
+ "39": "english_foxhound",
46
+ "40": "english_setter",
47
+ "41": "english_springer",
48
+ "42": "entlebucher",
49
+ "43": "eskimo_dog",
50
+ "44": "flat-coated_retriever",
51
+ "45": "french_bulldog",
52
+ "46": "german_shepherd",
53
+ "47": "german_short-haired_pointer",
54
+ "48": "giant_schnauzer",
55
+ "49": "golden_retriever",
56
+ "50": "gordon_setter",
57
+ "51": "great_dane",
58
+ "52": "great_pyrenees",
59
+ "53": "greater_swiss_mountain_dog",
60
+ "54": "groenendael",
61
+ "55": "ibizan_hound",
62
+ "56": "irish_setter",
63
+ "57": "irish_terrier",
64
+ "58": "irish_water_spaniel",
65
+ "59": "irish_wolfhound",
66
+ "60": "italian_greyhound",
67
+ "61": "japanese_spaniel",
68
+ "62": "keeshond",
69
+ "63": "kelpie",
70
+ "64": "kerry_blue_terrier",
71
+ "65": "komondor",
72
+ "66": "kuvasz",
73
+ "67": "labrador_retriever",
74
+ "68": "lakeland_terrier",
75
+ "69": "leonberg",
76
+ "70": "lhasa",
77
+ "71": "malamute",
78
+ "72": "malinois",
79
+ "73": "maltese_dog",
80
+ "74": "mexican_hairless",
81
+ "75": "miniature_pinscher",
82
+ "76": "miniature_poodle",
83
+ "77": "miniature_schnauzer",
84
+ "78": "newfoundland",
85
+ "79": "norfolk_terrier",
86
+ "80": "norwegian_elkhound",
87
+ "81": "norwich_terrier",
88
+ "82": "old_english_sheepdog",
89
+ "83": "otterhound",
90
+ "84": "papillon",
91
+ "85": "pekinese",
92
+ "86": "pembroke",
93
+ "87": "pomeranian",
94
+ "88": "pug",
95
+ "89": "redbone",
96
+ "90": "rhodesian_ridgeback",
97
+ "91": "rottweiler",
98
+ "92": "saint_bernard",
99
+ "93": "saluki",
100
+ "94": "samoyed",
101
+ "95": "schipperke",
102
+ "96": "scotch_terrier",
103
+ "97": "scottish_deerhound",
104
+ "98": "sealyham_terrier",
105
+ "99": "shetland_sheepdog",
106
+ "100": "shih-tzu",
107
+ "101": "siberian_husky",
108
+ "102": "silky_terrier",
109
+ "103": "soft-coated_wheaten_terrier",
110
+ "104": "staffordshire_bullterrier",
111
+ "105": "standard_poodle",
112
+ "106": "standard_schnauzer",
113
+ "107": "sussex_spaniel",
114
+ "108": "test",
115
+ "109": "tibetan_mastiff",
116
+ "110": "tibetan_terrier",
117
+ "111": "toy_poodle",
118
+ "112": "toy_terrier",
119
+ "113": "vizsla",
120
+ "114": "walker_hound",
121
+ "115": "weimaraner",
122
+ "116": "welsh_springer_spaniel",
123
+ "117": "west_highland_white_terrier",
124
+ "118": "whippet",
125
+ "119": "wire-haired_fox_terrier",
126
+ "120": "yorkshire_terrier"
127
+ },
128
+ "initializer_factor": 1.0,
129
+ "label2id": {
130
+ "affenpinscher": 0,
131
+ "afghan_hound": 1,
132
+ "african_hunting_dog": 2,
133
+ "airedale": 3,
134
+ "american_staffordshire_terrier": 4,
135
+ "appenzeller": 5,
136
+ "australian_terrier": 6,
137
+ "basenji": 7,
138
+ "basset": 8,
139
+ "beagle": 9,
140
+ "bedlington_terrier": 10,
141
+ "bernese_mountain_dog": 11,
142
+ "black-and-tan_coonhound": 12,
143
+ "blenheim_spaniel": 13,
144
+ "bloodhound": 14,
145
+ "bluetick": 15,
146
+ "border_collie": 16,
147
+ "border_terrier": 17,
148
+ "borzoi": 18,
149
+ "boston_bull": 19,
150
+ "bouvier_des_flandres": 20,
151
+ "boxer": 21,
152
+ "brabancon_griffon": 22,
153
+ "briard": 23,
154
+ "brittany_spaniel": 24,
155
+ "bull_mastiff": 25,
156
+ "cairn": 26,
157
+ "cardigan": 27,
158
+ "chesapeake_bay_retriever": 28,
159
+ "chihuahua": 29,
160
+ "chow": 30,
161
+ "clumber": 31,
162
+ "cocker_spaniel": 32,
163
+ "collie": 33,
164
+ "curly-coated_retriever": 34,
165
+ "dandie_dinmont": 35,
166
+ "dhole": 36,
167
+ "dingo": 37,
168
+ "doberman": 38,
169
+ "english_foxhound": 39,
170
+ "english_setter": 40,
171
+ "english_springer": 41,
172
+ "entlebucher": 42,
173
+ "eskimo_dog": 43,
174
+ "flat-coated_retriever": 44,
175
+ "french_bulldog": 45,
176
+ "german_shepherd": 46,
177
+ "german_short-haired_pointer": 47,
178
+ "giant_schnauzer": 48,
179
+ "golden_retriever": 49,
180
+ "gordon_setter": 50,
181
+ "great_dane": 51,
182
+ "great_pyrenees": 52,
183
+ "greater_swiss_mountain_dog": 53,
184
+ "groenendael": 54,
185
+ "ibizan_hound": 55,
186
+ "irish_setter": 56,
187
+ "irish_terrier": 57,
188
+ "irish_water_spaniel": 58,
189
+ "irish_wolfhound": 59,
190
+ "italian_greyhound": 60,
191
+ "japanese_spaniel": 61,
192
+ "keeshond": 62,
193
+ "kelpie": 63,
194
+ "kerry_blue_terrier": 64,
195
+ "komondor": 65,
196
+ "kuvasz": 66,
197
+ "labrador_retriever": 67,
198
+ "lakeland_terrier": 68,
199
+ "leonberg": 69,
200
+ "lhasa": 70,
201
+ "malamute": 71,
202
+ "malinois": 72,
203
+ "maltese_dog": 73,
204
+ "mexican_hairless": 74,
205
+ "miniature_pinscher": 75,
206
+ "miniature_poodle": 76,
207
+ "miniature_schnauzer": 77,
208
+ "newfoundland": 78,
209
+ "norfolk_terrier": 79,
210
+ "norwegian_elkhound": 80,
211
+ "norwich_terrier": 81,
212
+ "old_english_sheepdog": 82,
213
+ "otterhound": 83,
214
+ "papillon": 84,
215
+ "pekinese": 85,
216
+ "pembroke": 86,
217
+ "pomeranian": 87,
218
+ "pug": 88,
219
+ "redbone": 89,
220
+ "rhodesian_ridgeback": 90,
221
+ "rottweiler": 91,
222
+ "saint_bernard": 92,
223
+ "saluki": 93,
224
+ "samoyed": 94,
225
+ "schipperke": 95,
226
+ "scotch_terrier": 96,
227
+ "scottish_deerhound": 97,
228
+ "sealyham_terrier": 98,
229
+ "shetland_sheepdog": 99,
230
+ "shih-tzu": 100,
231
+ "siberian_husky": 101,
232
+ "silky_terrier": 102,
233
+ "soft-coated_wheaten_terrier": 103,
234
+ "staffordshire_bullterrier": 104,
235
+ "standard_poodle": 105,
236
+ "standard_schnauzer": 106,
237
+ "sussex_spaniel": 107,
238
+ "test": 108,
239
+ "tibetan_mastiff": 109,
240
+ "tibetan_terrier": 110,
241
+ "toy_poodle": 111,
242
+ "toy_terrier": 112,
243
+ "vizsla": 113,
244
+ "walker_hound": 114,
245
+ "weimaraner": 115,
246
+ "welsh_springer_spaniel": 116,
247
+ "west_highland_white_terrier": 117,
248
+ "whippet": 118,
249
+ "wire-haired_fox_terrier": 119,
250
+ "yorkshire_terrier": 120
251
+ },
252
+ "model_type": "siglip",
253
+ "problem_type": "single_label_classification",
254
+ "text_config": {
255
+ "attention_dropout": 0.0,
256
+ "hidden_act": "gelu_pytorch_tanh",
257
+ "hidden_size": 768,
258
+ "intermediate_size": 3072,
259
+ "layer_norm_eps": 1e-06,
260
+ "max_position_embeddings": 64,
261
+ "model_type": "siglip_text_model",
262
+ "num_attention_heads": 12,
263
+ "num_hidden_layers": 12,
264
+ "projection_size": 768,
265
+ "torch_dtype": "float32",
266
+ "vocab_size": 256000
267
+ },
268
+ "torch_dtype": "float32",
269
+ "transformers_version": "4.51.0.dev0",
270
+ "vision_config": {
271
+ "attention_dropout": 0.0,
272
+ "hidden_act": "gelu_pytorch_tanh",
273
+ "hidden_size": 768,
274
+ "image_size": 224,
275
+ "intermediate_size": 3072,
276
+ "layer_norm_eps": 1e-06,
277
+ "model_type": "siglip_vision_model",
278
+ "num_attention_heads": 12,
279
+ "num_channels": 3,
280
+ "num_hidden_layers": 12,
281
+ "patch_size": 16,
282
+ "torch_dtype": "float32"
283
+ }
284
+ }
checkpoint-1280/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:58bc382c9023d8d384bb82edf5250ccd9a6c6fb7343c81cfd4fade5bfc36765e
3
+ size 371934052
checkpoint-1280/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4e6248c7be613c06e08c102bd6ab943fa192b669e64dc0e841bdd8c7b98484bd
3
+ size 687287802
checkpoint-1280/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-1280/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:00fc94c2b413c97788ced4b8b606954bf28c08afdd427b96fed3821eaae67af3
3
+ size 14244
checkpoint-1280/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:989fbbe8318c30f4bab9329be55bf089439ca639ad5ba06940eb9f0624b0e371
3
+ size 1064
checkpoint-1280/trainer_state.json ADDED
@@ -0,0 +1,88 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": 1280,
3
+ "best_metric": 0.6800340414047241,
4
+ "best_model_checkpoint": "siglip2-finetune-full/checkpoint-1280",
5
+ "epoch": 4.0,
6
+ "eval_steps": 500,
7
+ "global_step": 1280,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 1.0,
14
+ "eval_accuracy": 0.48082567012326355,
15
+ "eval_loss": 2.3700547218322754,
16
+ "eval_model_preparation_time": 0.0042,
17
+ "eval_runtime": 153.6453,
18
+ "eval_samples_per_second": 66.53,
19
+ "eval_steps_per_second": 8.318,
20
+ "step": 320
21
+ },
22
+ {
23
+ "epoch": 1.5625,
24
+ "grad_norm": 61.34077453613281,
25
+ "learning_rate": 3.972602739726027e-06,
26
+ "loss": 3.3214,
27
+ "step": 500
28
+ },
29
+ {
30
+ "epoch": 2.0,
31
+ "eval_accuracy": 0.7256896889062806,
32
+ "eval_loss": 1.1354633569717407,
33
+ "eval_model_preparation_time": 0.0042,
34
+ "eval_runtime": 146.0009,
35
+ "eval_samples_per_second": 70.013,
36
+ "eval_steps_per_second": 8.753,
37
+ "step": 640
38
+ },
39
+ {
40
+ "epoch": 3.0,
41
+ "eval_accuracy": 0.7894736842105263,
42
+ "eval_loss": 0.8310916423797607,
43
+ "eval_model_preparation_time": 0.0042,
44
+ "eval_runtime": 146.2788,
45
+ "eval_samples_per_second": 69.88,
46
+ "eval_steps_per_second": 8.737,
47
+ "step": 960
48
+ },
49
+ {
50
+ "epoch": 3.125,
51
+ "grad_norm": 65.03152465820312,
52
+ "learning_rate": 2.8310502283105025e-06,
53
+ "loss": 1.2343,
54
+ "step": 1000
55
+ },
56
+ {
57
+ "epoch": 4.0,
58
+ "eval_accuracy": 0.8205830561533947,
59
+ "eval_loss": 0.6800340414047241,
60
+ "eval_model_preparation_time": 0.0042,
61
+ "eval_runtime": 145.9883,
62
+ "eval_samples_per_second": 70.019,
63
+ "eval_steps_per_second": 8.754,
64
+ "step": 1280
65
+ }
66
+ ],
67
+ "logging_steps": 500,
68
+ "max_steps": 2240,
69
+ "num_input_tokens_seen": 0,
70
+ "num_train_epochs": 7,
71
+ "save_steps": 500,
72
+ "stateful_callbacks": {
73
+ "TrainerControl": {
74
+ "args": {
75
+ "should_epoch_stop": false,
76
+ "should_evaluate": false,
77
+ "should_log": false,
78
+ "should_save": true,
79
+ "should_training_stop": false
80
+ },
81
+ "attributes": {}
82
+ }
83
+ },
84
+ "total_flos": 3.427974098444206e+18,
85
+ "train_batch_size": 32,
86
+ "trial_name": null,
87
+ "trial_params": null
88
+ }
checkpoint-1280/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:28a1d0358eb842d6a8e2b23ac90eef5d6694c7d23b3f07aaeb6d419b85372614
3
+ size 5304
checkpoint-1600/config.json ADDED
@@ -0,0 +1,284 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "SiglipForImageClassification"
4
+ ],
5
+ "id2label": {
6
+ "0": "affenpinscher",
7
+ "1": "afghan_hound",
8
+ "2": "african_hunting_dog",
9
+ "3": "airedale",
10
+ "4": "american_staffordshire_terrier",
11
+ "5": "appenzeller",
12
+ "6": "australian_terrier",
13
+ "7": "basenji",
14
+ "8": "basset",
15
+ "9": "beagle",
16
+ "10": "bedlington_terrier",
17
+ "11": "bernese_mountain_dog",
18
+ "12": "black-and-tan_coonhound",
19
+ "13": "blenheim_spaniel",
20
+ "14": "bloodhound",
21
+ "15": "bluetick",
22
+ "16": "border_collie",
23
+ "17": "border_terrier",
24
+ "18": "borzoi",
25
+ "19": "boston_bull",
26
+ "20": "bouvier_des_flandres",
27
+ "21": "boxer",
28
+ "22": "brabancon_griffon",
29
+ "23": "briard",
30
+ "24": "brittany_spaniel",
31
+ "25": "bull_mastiff",
32
+ "26": "cairn",
33
+ "27": "cardigan",
34
+ "28": "chesapeake_bay_retriever",
35
+ "29": "chihuahua",
36
+ "30": "chow",
37
+ "31": "clumber",
38
+ "32": "cocker_spaniel",
39
+ "33": "collie",
40
+ "34": "curly-coated_retriever",
41
+ "35": "dandie_dinmont",
42
+ "36": "dhole",
43
+ "37": "dingo",
44
+ "38": "doberman",
45
+ "39": "english_foxhound",
46
+ "40": "english_setter",
47
+ "41": "english_springer",
48
+ "42": "entlebucher",
49
+ "43": "eskimo_dog",
50
+ "44": "flat-coated_retriever",
51
+ "45": "french_bulldog",
52
+ "46": "german_shepherd",
53
+ "47": "german_short-haired_pointer",
54
+ "48": "giant_schnauzer",
55
+ "49": "golden_retriever",
56
+ "50": "gordon_setter",
57
+ "51": "great_dane",
58
+ "52": "great_pyrenees",
59
+ "53": "greater_swiss_mountain_dog",
60
+ "54": "groenendael",
61
+ "55": "ibizan_hound",
62
+ "56": "irish_setter",
63
+ "57": "irish_terrier",
64
+ "58": "irish_water_spaniel",
65
+ "59": "irish_wolfhound",
66
+ "60": "italian_greyhound",
67
+ "61": "japanese_spaniel",
68
+ "62": "keeshond",
69
+ "63": "kelpie",
70
+ "64": "kerry_blue_terrier",
71
+ "65": "komondor",
72
+ "66": "kuvasz",
73
+ "67": "labrador_retriever",
74
+ "68": "lakeland_terrier",
75
+ "69": "leonberg",
76
+ "70": "lhasa",
77
+ "71": "malamute",
78
+ "72": "malinois",
79
+ "73": "maltese_dog",
80
+ "74": "mexican_hairless",
81
+ "75": "miniature_pinscher",
82
+ "76": "miniature_poodle",
83
+ "77": "miniature_schnauzer",
84
+ "78": "newfoundland",
85
+ "79": "norfolk_terrier",
86
+ "80": "norwegian_elkhound",
87
+ "81": "norwich_terrier",
88
+ "82": "old_english_sheepdog",
89
+ "83": "otterhound",
90
+ "84": "papillon",
91
+ "85": "pekinese",
92
+ "86": "pembroke",
93
+ "87": "pomeranian",
94
+ "88": "pug",
95
+ "89": "redbone",
96
+ "90": "rhodesian_ridgeback",
97
+ "91": "rottweiler",
98
+ "92": "saint_bernard",
99
+ "93": "saluki",
100
+ "94": "samoyed",
101
+ "95": "schipperke",
102
+ "96": "scotch_terrier",
103
+ "97": "scottish_deerhound",
104
+ "98": "sealyham_terrier",
105
+ "99": "shetland_sheepdog",
106
+ "100": "shih-tzu",
107
+ "101": "siberian_husky",
108
+ "102": "silky_terrier",
109
+ "103": "soft-coated_wheaten_terrier",
110
+ "104": "staffordshire_bullterrier",
111
+ "105": "standard_poodle",
112
+ "106": "standard_schnauzer",
113
+ "107": "sussex_spaniel",
114
+ "108": "test",
115
+ "109": "tibetan_mastiff",
116
+ "110": "tibetan_terrier",
117
+ "111": "toy_poodle",
118
+ "112": "toy_terrier",
119
+ "113": "vizsla",
120
+ "114": "walker_hound",
121
+ "115": "weimaraner",
122
+ "116": "welsh_springer_spaniel",
123
+ "117": "west_highland_white_terrier",
124
+ "118": "whippet",
125
+ "119": "wire-haired_fox_terrier",
126
+ "120": "yorkshire_terrier"
127
+ },
128
+ "initializer_factor": 1.0,
129
+ "label2id": {
130
+ "affenpinscher": 0,
131
+ "afghan_hound": 1,
132
+ "african_hunting_dog": 2,
133
+ "airedale": 3,
134
+ "american_staffordshire_terrier": 4,
135
+ "appenzeller": 5,
136
+ "australian_terrier": 6,
137
+ "basenji": 7,
138
+ "basset": 8,
139
+ "beagle": 9,
140
+ "bedlington_terrier": 10,
141
+ "bernese_mountain_dog": 11,
142
+ "black-and-tan_coonhound": 12,
143
+ "blenheim_spaniel": 13,
144
+ "bloodhound": 14,
145
+ "bluetick": 15,
146
+ "border_collie": 16,
147
+ "border_terrier": 17,
148
+ "borzoi": 18,
149
+ "boston_bull": 19,
150
+ "bouvier_des_flandres": 20,
151
+ "boxer": 21,
152
+ "brabancon_griffon": 22,
153
+ "briard": 23,
154
+ "brittany_spaniel": 24,
155
+ "bull_mastiff": 25,
156
+ "cairn": 26,
157
+ "cardigan": 27,
158
+ "chesapeake_bay_retriever": 28,
159
+ "chihuahua": 29,
160
+ "chow": 30,
161
+ "clumber": 31,
162
+ "cocker_spaniel": 32,
163
+ "collie": 33,
164
+ "curly-coated_retriever": 34,
165
+ "dandie_dinmont": 35,
166
+ "dhole": 36,
167
+ "dingo": 37,
168
+ "doberman": 38,
169
+ "english_foxhound": 39,
170
+ "english_setter": 40,
171
+ "english_springer": 41,
172
+ "entlebucher": 42,
173
+ "eskimo_dog": 43,
174
+ "flat-coated_retriever": 44,
175
+ "french_bulldog": 45,
176
+ "german_shepherd": 46,
177
+ "german_short-haired_pointer": 47,
178
+ "giant_schnauzer": 48,
179
+ "golden_retriever": 49,
180
+ "gordon_setter": 50,
181
+ "great_dane": 51,
182
+ "great_pyrenees": 52,
183
+ "greater_swiss_mountain_dog": 53,
184
+ "groenendael": 54,
185
+ "ibizan_hound": 55,
186
+ "irish_setter": 56,
187
+ "irish_terrier": 57,
188
+ "irish_water_spaniel": 58,
189
+ "irish_wolfhound": 59,
190
+ "italian_greyhound": 60,
191
+ "japanese_spaniel": 61,
192
+ "keeshond": 62,
193
+ "kelpie": 63,
194
+ "kerry_blue_terrier": 64,
195
+ "komondor": 65,
196
+ "kuvasz": 66,
197
+ "labrador_retriever": 67,
198
+ "lakeland_terrier": 68,
199
+ "leonberg": 69,
200
+ "lhasa": 70,
201
+ "malamute": 71,
202
+ "malinois": 72,
203
+ "maltese_dog": 73,
204
+ "mexican_hairless": 74,
205
+ "miniature_pinscher": 75,
206
+ "miniature_poodle": 76,
207
+ "miniature_schnauzer": 77,
208
+ "newfoundland": 78,
209
+ "norfolk_terrier": 79,
210
+ "norwegian_elkhound": 80,
211
+ "norwich_terrier": 81,
212
+ "old_english_sheepdog": 82,
213
+ "otterhound": 83,
214
+ "papillon": 84,
215
+ "pekinese": 85,
216
+ "pembroke": 86,
217
+ "pomeranian": 87,
218
+ "pug": 88,
219
+ "redbone": 89,
220
+ "rhodesian_ridgeback": 90,
221
+ "rottweiler": 91,
222
+ "saint_bernard": 92,
223
+ "saluki": 93,
224
+ "samoyed": 94,
225
+ "schipperke": 95,
226
+ "scotch_terrier": 96,
227
+ "scottish_deerhound": 97,
228
+ "sealyham_terrier": 98,
229
+ "shetland_sheepdog": 99,
230
+ "shih-tzu": 100,
231
+ "siberian_husky": 101,
232
+ "silky_terrier": 102,
233
+ "soft-coated_wheaten_terrier": 103,
234
+ "staffordshire_bullterrier": 104,
235
+ "standard_poodle": 105,
236
+ "standard_schnauzer": 106,
237
+ "sussex_spaniel": 107,
238
+ "test": 108,
239
+ "tibetan_mastiff": 109,
240
+ "tibetan_terrier": 110,
241
+ "toy_poodle": 111,
242
+ "toy_terrier": 112,
243
+ "vizsla": 113,
244
+ "walker_hound": 114,
245
+ "weimaraner": 115,
246
+ "welsh_springer_spaniel": 116,
247
+ "west_highland_white_terrier": 117,
248
+ "whippet": 118,
249
+ "wire-haired_fox_terrier": 119,
250
+ "yorkshire_terrier": 120
251
+ },
252
+ "model_type": "siglip",
253
+ "problem_type": "single_label_classification",
254
+ "text_config": {
255
+ "attention_dropout": 0.0,
256
+ "hidden_act": "gelu_pytorch_tanh",
257
+ "hidden_size": 768,
258
+ "intermediate_size": 3072,
259
+ "layer_norm_eps": 1e-06,
260
+ "max_position_embeddings": 64,
261
+ "model_type": "siglip_text_model",
262
+ "num_attention_heads": 12,
263
+ "num_hidden_layers": 12,
264
+ "projection_size": 768,
265
+ "torch_dtype": "float32",
266
+ "vocab_size": 256000
267
+ },
268
+ "torch_dtype": "float32",
269
+ "transformers_version": "4.51.0.dev0",
270
+ "vision_config": {
271
+ "attention_dropout": 0.0,
272
+ "hidden_act": "gelu_pytorch_tanh",
273
+ "hidden_size": 768,
274
+ "image_size": 224,
275
+ "intermediate_size": 3072,
276
+ "layer_norm_eps": 1e-06,
277
+ "model_type": "siglip_vision_model",
278
+ "num_attention_heads": 12,
279
+ "num_channels": 3,
280
+ "num_hidden_layers": 12,
281
+ "patch_size": 16,
282
+ "torch_dtype": "float32"
283
+ }
284
+ }
checkpoint-1600/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d966fe5d9ebba13a658ef4b116979df8ba9c10f0fe02623a71d65cda1ed8cab1
3
+ size 371934052
checkpoint-1600/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:38a145107c15ff479093862b5e962385ac71967ad40493cae6b27cecabe377c1
3
+ size 687287802
checkpoint-1600/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-1600/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0e007c17f52225af31edd04874ab5ff6d15ecc6a71d42abcf3e335fee9804ac9
3
+ size 14244
checkpoint-1600/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0b8131cb631cf75deb13c2ca37a0539721d2a685753b42a69c304126ca25760e
3
+ size 1064
checkpoint-1600/trainer_state.json ADDED
@@ -0,0 +1,105 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": 1600,
3
+ "best_metric": 0.5670026540756226,
4
+ "best_model_checkpoint": "siglip2-finetune-full/checkpoint-1600",
5
+ "epoch": 5.0,
6
+ "eval_steps": 500,
7
+ "global_step": 1600,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 1.0,
14
+ "eval_accuracy": 0.48082567012326355,
15
+ "eval_loss": 2.3700547218322754,
16
+ "eval_model_preparation_time": 0.0042,
17
+ "eval_runtime": 153.6453,
18
+ "eval_samples_per_second": 66.53,
19
+ "eval_steps_per_second": 8.318,
20
+ "step": 320
21
+ },
22
+ {
23
+ "epoch": 1.5625,
24
+ "grad_norm": 61.34077453613281,
25
+ "learning_rate": 3.972602739726027e-06,
26
+ "loss": 3.3214,
27
+ "step": 500
28
+ },
29
+ {
30
+ "epoch": 2.0,
31
+ "eval_accuracy": 0.7256896889062806,
32
+ "eval_loss": 1.1354633569717407,
33
+ "eval_model_preparation_time": 0.0042,
34
+ "eval_runtime": 146.0009,
35
+ "eval_samples_per_second": 70.013,
36
+ "eval_steps_per_second": 8.753,
37
+ "step": 640
38
+ },
39
+ {
40
+ "epoch": 3.0,
41
+ "eval_accuracy": 0.7894736842105263,
42
+ "eval_loss": 0.8310916423797607,
43
+ "eval_model_preparation_time": 0.0042,
44
+ "eval_runtime": 146.2788,
45
+ "eval_samples_per_second": 69.88,
46
+ "eval_steps_per_second": 8.737,
47
+ "step": 960
48
+ },
49
+ {
50
+ "epoch": 3.125,
51
+ "grad_norm": 65.03152465820312,
52
+ "learning_rate": 2.8310502283105025e-06,
53
+ "loss": 1.2343,
54
+ "step": 1000
55
+ },
56
+ {
57
+ "epoch": 4.0,
58
+ "eval_accuracy": 0.8205830561533947,
59
+ "eval_loss": 0.6800340414047241,
60
+ "eval_model_preparation_time": 0.0042,
61
+ "eval_runtime": 145.9883,
62
+ "eval_samples_per_second": 70.019,
63
+ "eval_steps_per_second": 8.754,
64
+ "step": 1280
65
+ },
66
+ {
67
+ "epoch": 4.6875,
68
+ "grad_norm": 42.12885665893555,
69
+ "learning_rate": 1.6894977168949772e-06,
70
+ "loss": 0.8261,
71
+ "step": 1500
72
+ },
73
+ {
74
+ "epoch": 5.0,
75
+ "eval_accuracy": 0.8485619252592448,
76
+ "eval_loss": 0.5670026540756226,
77
+ "eval_model_preparation_time": 0.0042,
78
+ "eval_runtime": 146.844,
79
+ "eval_samples_per_second": 69.611,
80
+ "eval_steps_per_second": 8.703,
81
+ "step": 1600
82
+ }
83
+ ],
84
+ "logging_steps": 500,
85
+ "max_steps": 2240,
86
+ "num_input_tokens_seen": 0,
87
+ "num_train_epochs": 7,
88
+ "save_steps": 500,
89
+ "stateful_callbacks": {
90
+ "TrainerControl": {
91
+ "args": {
92
+ "should_epoch_stop": false,
93
+ "should_evaluate": false,
94
+ "should_log": false,
95
+ "should_save": true,
96
+ "should_training_stop": false
97
+ },
98
+ "attributes": {}
99
+ }
100
+ },
101
+ "total_flos": 4.2849676230552576e+18,
102
+ "train_batch_size": 32,
103
+ "trial_name": null,
104
+ "trial_params": null
105
+ }
checkpoint-1600/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:28a1d0358eb842d6a8e2b23ac90eef5d6694c7d23b3f07aaeb6d419b85372614
3
+ size 5304
checkpoint-1920/config.json ADDED
@@ -0,0 +1,284 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "SiglipForImageClassification"
4
+ ],
5
+ "id2label": {
6
+ "0": "affenpinscher",
7
+ "1": "afghan_hound",
8
+ "2": "african_hunting_dog",
9
+ "3": "airedale",
10
+ "4": "american_staffordshire_terrier",
11
+ "5": "appenzeller",
12
+ "6": "australian_terrier",
13
+ "7": "basenji",
14
+ "8": "basset",
15
+ "9": "beagle",
16
+ "10": "bedlington_terrier",
17
+ "11": "bernese_mountain_dog",
18
+ "12": "black-and-tan_coonhound",
19
+ "13": "blenheim_spaniel",
20
+ "14": "bloodhound",
21
+ "15": "bluetick",
22
+ "16": "border_collie",
23
+ "17": "border_terrier",
24
+ "18": "borzoi",
25
+ "19": "boston_bull",
26
+ "20": "bouvier_des_flandres",
27
+ "21": "boxer",
28
+ "22": "brabancon_griffon",
29
+ "23": "briard",
30
+ "24": "brittany_spaniel",
31
+ "25": "bull_mastiff",
32
+ "26": "cairn",
33
+ "27": "cardigan",
34
+ "28": "chesapeake_bay_retriever",
35
+ "29": "chihuahua",
36
+ "30": "chow",
37
+ "31": "clumber",
38
+ "32": "cocker_spaniel",
39
+ "33": "collie",
40
+ "34": "curly-coated_retriever",
41
+ "35": "dandie_dinmont",
42
+ "36": "dhole",
43
+ "37": "dingo",
44
+ "38": "doberman",
45
+ "39": "english_foxhound",
46
+ "40": "english_setter",
47
+ "41": "english_springer",
48
+ "42": "entlebucher",
49
+ "43": "eskimo_dog",
50
+ "44": "flat-coated_retriever",
51
+ "45": "french_bulldog",
52
+ "46": "german_shepherd",
53
+ "47": "german_short-haired_pointer",
54
+ "48": "giant_schnauzer",
55
+ "49": "golden_retriever",
56
+ "50": "gordon_setter",
57
+ "51": "great_dane",
58
+ "52": "great_pyrenees",
59
+ "53": "greater_swiss_mountain_dog",
60
+ "54": "groenendael",
61
+ "55": "ibizan_hound",
62
+ "56": "irish_setter",
63
+ "57": "irish_terrier",
64
+ "58": "irish_water_spaniel",
65
+ "59": "irish_wolfhound",
66
+ "60": "italian_greyhound",
67
+ "61": "japanese_spaniel",
68
+ "62": "keeshond",
69
+ "63": "kelpie",
70
+ "64": "kerry_blue_terrier",
71
+ "65": "komondor",
72
+ "66": "kuvasz",
73
+ "67": "labrador_retriever",
74
+ "68": "lakeland_terrier",
75
+ "69": "leonberg",
76
+ "70": "lhasa",
77
+ "71": "malamute",
78
+ "72": "malinois",
79
+ "73": "maltese_dog",
80
+ "74": "mexican_hairless",
81
+ "75": "miniature_pinscher",
82
+ "76": "miniature_poodle",
83
+ "77": "miniature_schnauzer",
84
+ "78": "newfoundland",
85
+ "79": "norfolk_terrier",
86
+ "80": "norwegian_elkhound",
87
+ "81": "norwich_terrier",
88
+ "82": "old_english_sheepdog",
89
+ "83": "otterhound",
90
+ "84": "papillon",
91
+ "85": "pekinese",
92
+ "86": "pembroke",
93
+ "87": "pomeranian",
94
+ "88": "pug",
95
+ "89": "redbone",
96
+ "90": "rhodesian_ridgeback",
97
+ "91": "rottweiler",
98
+ "92": "saint_bernard",
99
+ "93": "saluki",
100
+ "94": "samoyed",
101
+ "95": "schipperke",
102
+ "96": "scotch_terrier",
103
+ "97": "scottish_deerhound",
104
+ "98": "sealyham_terrier",
105
+ "99": "shetland_sheepdog",
106
+ "100": "shih-tzu",
107
+ "101": "siberian_husky",
108
+ "102": "silky_terrier",
109
+ "103": "soft-coated_wheaten_terrier",
110
+ "104": "staffordshire_bullterrier",
111
+ "105": "standard_poodle",
112
+ "106": "standard_schnauzer",
113
+ "107": "sussex_spaniel",
114
+ "108": "test",
115
+ "109": "tibetan_mastiff",
116
+ "110": "tibetan_terrier",
117
+ "111": "toy_poodle",
118
+ "112": "toy_terrier",
119
+ "113": "vizsla",
120
+ "114": "walker_hound",
121
+ "115": "weimaraner",
122
+ "116": "welsh_springer_spaniel",
123
+ "117": "west_highland_white_terrier",
124
+ "118": "whippet",
125
+ "119": "wire-haired_fox_terrier",
126
+ "120": "yorkshire_terrier"
127
+ },
128
+ "initializer_factor": 1.0,
129
+ "label2id": {
130
+ "affenpinscher": 0,
131
+ "afghan_hound": 1,
132
+ "african_hunting_dog": 2,
133
+ "airedale": 3,
134
+ "american_staffordshire_terrier": 4,
135
+ "appenzeller": 5,
136
+ "australian_terrier": 6,
137
+ "basenji": 7,
138
+ "basset": 8,
139
+ "beagle": 9,
140
+ "bedlington_terrier": 10,
141
+ "bernese_mountain_dog": 11,
142
+ "black-and-tan_coonhound": 12,
143
+ "blenheim_spaniel": 13,
144
+ "bloodhound": 14,
145
+ "bluetick": 15,
146
+ "border_collie": 16,
147
+ "border_terrier": 17,
148
+ "borzoi": 18,
149
+ "boston_bull": 19,
150
+ "bouvier_des_flandres": 20,
151
+ "boxer": 21,
152
+ "brabancon_griffon": 22,
153
+ "briard": 23,
154
+ "brittany_spaniel": 24,
155
+ "bull_mastiff": 25,
156
+ "cairn": 26,
157
+ "cardigan": 27,
158
+ "chesapeake_bay_retriever": 28,
159
+ "chihuahua": 29,
160
+ "chow": 30,
161
+ "clumber": 31,
162
+ "cocker_spaniel": 32,
163
+ "collie": 33,
164
+ "curly-coated_retriever": 34,
165
+ "dandie_dinmont": 35,
166
+ "dhole": 36,
167
+ "dingo": 37,
168
+ "doberman": 38,
169
+ "english_foxhound": 39,
170
+ "english_setter": 40,
171
+ "english_springer": 41,
172
+ "entlebucher": 42,
173
+ "eskimo_dog": 43,
174
+ "flat-coated_retriever": 44,
175
+ "french_bulldog": 45,
176
+ "german_shepherd": 46,
177
+ "german_short-haired_pointer": 47,
178
+ "giant_schnauzer": 48,
179
+ "golden_retriever": 49,
180
+ "gordon_setter": 50,
181
+ "great_dane": 51,
182
+ "great_pyrenees": 52,
183
+ "greater_swiss_mountain_dog": 53,
184
+ "groenendael": 54,
185
+ "ibizan_hound": 55,
186
+ "irish_setter": 56,
187
+ "irish_terrier": 57,
188
+ "irish_water_spaniel": 58,
189
+ "irish_wolfhound": 59,
190
+ "italian_greyhound": 60,
191
+ "japanese_spaniel": 61,
192
+ "keeshond": 62,
193
+ "kelpie": 63,
194
+ "kerry_blue_terrier": 64,
195
+ "komondor": 65,
196
+ "kuvasz": 66,
197
+ "labrador_retriever": 67,
198
+ "lakeland_terrier": 68,
199
+ "leonberg": 69,
200
+ "lhasa": 70,
201
+ "malamute": 71,
202
+ "malinois": 72,
203
+ "maltese_dog": 73,
204
+ "mexican_hairless": 74,
205
+ "miniature_pinscher": 75,
206
+ "miniature_poodle": 76,
207
+ "miniature_schnauzer": 77,
208
+ "newfoundland": 78,
209
+ "norfolk_terrier": 79,
210
+ "norwegian_elkhound": 80,
211
+ "norwich_terrier": 81,
212
+ "old_english_sheepdog": 82,
213
+ "otterhound": 83,
214
+ "papillon": 84,
215
+ "pekinese": 85,
216
+ "pembroke": 86,
217
+ "pomeranian": 87,
218
+ "pug": 88,
219
+ "redbone": 89,
220
+ "rhodesian_ridgeback": 90,
221
+ "rottweiler": 91,
222
+ "saint_bernard": 92,
223
+ "saluki": 93,
224
+ "samoyed": 94,
225
+ "schipperke": 95,
226
+ "scotch_terrier": 96,
227
+ "scottish_deerhound": 97,
228
+ "sealyham_terrier": 98,
229
+ "shetland_sheepdog": 99,
230
+ "shih-tzu": 100,
231
+ "siberian_husky": 101,
232
+ "silky_terrier": 102,
233
+ "soft-coated_wheaten_terrier": 103,
234
+ "staffordshire_bullterrier": 104,
235
+ "standard_poodle": 105,
236
+ "standard_schnauzer": 106,
237
+ "sussex_spaniel": 107,
238
+ "test": 108,
239
+ "tibetan_mastiff": 109,
240
+ "tibetan_terrier": 110,
241
+ "toy_poodle": 111,
242
+ "toy_terrier": 112,
243
+ "vizsla": 113,
244
+ "walker_hound": 114,
245
+ "weimaraner": 115,
246
+ "welsh_springer_spaniel": 116,
247
+ "west_highland_white_terrier": 117,
248
+ "whippet": 118,
249
+ "wire-haired_fox_terrier": 119,
250
+ "yorkshire_terrier": 120
251
+ },
252
+ "model_type": "siglip",
253
+ "problem_type": "single_label_classification",
254
+ "text_config": {
255
+ "attention_dropout": 0.0,
256
+ "hidden_act": "gelu_pytorch_tanh",
257
+ "hidden_size": 768,
258
+ "intermediate_size": 3072,
259
+ "layer_norm_eps": 1e-06,
260
+ "max_position_embeddings": 64,
261
+ "model_type": "siglip_text_model",
262
+ "num_attention_heads": 12,
263
+ "num_hidden_layers": 12,
264
+ "projection_size": 768,
265
+ "torch_dtype": "float32",
266
+ "vocab_size": 256000
267
+ },
268
+ "torch_dtype": "float32",
269
+ "transformers_version": "4.51.0.dev0",
270
+ "vision_config": {
271
+ "attention_dropout": 0.0,
272
+ "hidden_act": "gelu_pytorch_tanh",
273
+ "hidden_size": 768,
274
+ "image_size": 224,
275
+ "intermediate_size": 3072,
276
+ "layer_norm_eps": 1e-06,
277
+ "model_type": "siglip_vision_model",
278
+ "num_attention_heads": 12,
279
+ "num_channels": 3,
280
+ "num_hidden_layers": 12,
281
+ "patch_size": 16,
282
+ "torch_dtype": "float32"
283
+ }
284
+ }
checkpoint-1920/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ba409d4bc6bf15310c245cf971b83e9d8d2c5d087f4f60f89e62d9d9adc7d559
3
+ size 371934052
checkpoint-1920/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f4e1b777478e0ae99b01d2f0a56033cf3d087aa73ea476f7b118da3a0f1c6f58
3
+ size 687287802
checkpoint-1920/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-1920/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:56a410e020d00f25376c714e736e0d62129c1fcc3a5919307d1990d360e95934
3
+ size 14244
checkpoint-1920/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e29d46ef43bc2436d9d0117e5ebabf36714577caed4ef85cbf8e7db3943dbbc9
3
+ size 1064
checkpoint-1920/trainer_state.json ADDED
@@ -0,0 +1,115 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": 1920,
3
+ "best_metric": 0.5231612324714661,
4
+ "best_model_checkpoint": "siglip2-finetune-full/checkpoint-1920",
5
+ "epoch": 6.0,
6
+ "eval_steps": 500,
7
+ "global_step": 1920,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 1.0,
14
+ "eval_accuracy": 0.48082567012326355,
15
+ "eval_loss": 2.3700547218322754,
16
+ "eval_model_preparation_time": 0.0042,
17
+ "eval_runtime": 153.6453,
18
+ "eval_samples_per_second": 66.53,
19
+ "eval_steps_per_second": 8.318,
20
+ "step": 320
21
+ },
22
+ {
23
+ "epoch": 1.5625,
24
+ "grad_norm": 61.34077453613281,
25
+ "learning_rate": 3.972602739726027e-06,
26
+ "loss": 3.3214,
27
+ "step": 500
28
+ },
29
+ {
30
+ "epoch": 2.0,
31
+ "eval_accuracy": 0.7256896889062806,
32
+ "eval_loss": 1.1354633569717407,
33
+ "eval_model_preparation_time": 0.0042,
34
+ "eval_runtime": 146.0009,
35
+ "eval_samples_per_second": 70.013,
36
+ "eval_steps_per_second": 8.753,
37
+ "step": 640
38
+ },
39
+ {
40
+ "epoch": 3.0,
41
+ "eval_accuracy": 0.7894736842105263,
42
+ "eval_loss": 0.8310916423797607,
43
+ "eval_model_preparation_time": 0.0042,
44
+ "eval_runtime": 146.2788,
45
+ "eval_samples_per_second": 69.88,
46
+ "eval_steps_per_second": 8.737,
47
+ "step": 960
48
+ },
49
+ {
50
+ "epoch": 3.125,
51
+ "grad_norm": 65.03152465820312,
52
+ "learning_rate": 2.8310502283105025e-06,
53
+ "loss": 1.2343,
54
+ "step": 1000
55
+ },
56
+ {
57
+ "epoch": 4.0,
58
+ "eval_accuracy": 0.8205830561533947,
59
+ "eval_loss": 0.6800340414047241,
60
+ "eval_model_preparation_time": 0.0042,
61
+ "eval_runtime": 145.9883,
62
+ "eval_samples_per_second": 70.019,
63
+ "eval_steps_per_second": 8.754,
64
+ "step": 1280
65
+ },
66
+ {
67
+ "epoch": 4.6875,
68
+ "grad_norm": 42.12885665893555,
69
+ "learning_rate": 1.6894977168949772e-06,
70
+ "loss": 0.8261,
71
+ "step": 1500
72
+ },
73
+ {
74
+ "epoch": 5.0,
75
+ "eval_accuracy": 0.8485619252592448,
76
+ "eval_loss": 0.5670026540756226,
77
+ "eval_model_preparation_time": 0.0042,
78
+ "eval_runtime": 146.844,
79
+ "eval_samples_per_second": 69.611,
80
+ "eval_steps_per_second": 8.703,
81
+ "step": 1600
82
+ },
83
+ {
84
+ "epoch": 6.0,
85
+ "eval_accuracy": 0.8628448444531402,
86
+ "eval_loss": 0.5231612324714661,
87
+ "eval_model_preparation_time": 0.0042,
88
+ "eval_runtime": 146.0726,
89
+ "eval_samples_per_second": 69.979,
90
+ "eval_steps_per_second": 8.749,
91
+ "step": 1920
92
+ }
93
+ ],
94
+ "logging_steps": 500,
95
+ "max_steps": 2240,
96
+ "num_input_tokens_seen": 0,
97
+ "num_train_epochs": 7,
98
+ "save_steps": 500,
99
+ "stateful_callbacks": {
100
+ "TrainerControl": {
101
+ "args": {
102
+ "should_epoch_stop": false,
103
+ "should_evaluate": false,
104
+ "should_log": false,
105
+ "should_save": true,
106
+ "should_training_stop": false
107
+ },
108
+ "attributes": {}
109
+ }
110
+ },
111
+ "total_flos": 5.141961147666309e+18,
112
+ "train_batch_size": 32,
113
+ "trial_name": null,
114
+ "trial_params": null
115
+ }
checkpoint-1920/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:28a1d0358eb842d6a8e2b23ac90eef5d6694c7d23b3f07aaeb6d419b85372614
3
+ size 5304
checkpoint-2240/config.json ADDED
@@ -0,0 +1,284 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "SiglipForImageClassification"
4
+ ],
5
+ "id2label": {
6
+ "0": "affenpinscher",
7
+ "1": "afghan_hound",
8
+ "2": "african_hunting_dog",
9
+ "3": "airedale",
10
+ "4": "american_staffordshire_terrier",
11
+ "5": "appenzeller",
12
+ "6": "australian_terrier",
13
+ "7": "basenji",
14
+ "8": "basset",
15
+ "9": "beagle",
16
+ "10": "bedlington_terrier",
17
+ "11": "bernese_mountain_dog",
18
+ "12": "black-and-tan_coonhound",
19
+ "13": "blenheim_spaniel",
20
+ "14": "bloodhound",
21
+ "15": "bluetick",
22
+ "16": "border_collie",
23
+ "17": "border_terrier",
24
+ "18": "borzoi",
25
+ "19": "boston_bull",
26
+ "20": "bouvier_des_flandres",
27
+ "21": "boxer",
28
+ "22": "brabancon_griffon",
29
+ "23": "briard",
30
+ "24": "brittany_spaniel",
31
+ "25": "bull_mastiff",
32
+ "26": "cairn",
33
+ "27": "cardigan",
34
+ "28": "chesapeake_bay_retriever",
35
+ "29": "chihuahua",
36
+ "30": "chow",
37
+ "31": "clumber",
38
+ "32": "cocker_spaniel",
39
+ "33": "collie",
40
+ "34": "curly-coated_retriever",
41
+ "35": "dandie_dinmont",
42
+ "36": "dhole",
43
+ "37": "dingo",
44
+ "38": "doberman",
45
+ "39": "english_foxhound",
46
+ "40": "english_setter",
47
+ "41": "english_springer",
48
+ "42": "entlebucher",
49
+ "43": "eskimo_dog",
50
+ "44": "flat-coated_retriever",
51
+ "45": "french_bulldog",
52
+ "46": "german_shepherd",
53
+ "47": "german_short-haired_pointer",
54
+ "48": "giant_schnauzer",
55
+ "49": "golden_retriever",
56
+ "50": "gordon_setter",
57
+ "51": "great_dane",
58
+ "52": "great_pyrenees",
59
+ "53": "greater_swiss_mountain_dog",
60
+ "54": "groenendael",
61
+ "55": "ibizan_hound",
62
+ "56": "irish_setter",
63
+ "57": "irish_terrier",
64
+ "58": "irish_water_spaniel",
65
+ "59": "irish_wolfhound",
66
+ "60": "italian_greyhound",
67
+ "61": "japanese_spaniel",
68
+ "62": "keeshond",
69
+ "63": "kelpie",
70
+ "64": "kerry_blue_terrier",
71
+ "65": "komondor",
72
+ "66": "kuvasz",
73
+ "67": "labrador_retriever",
74
+ "68": "lakeland_terrier",
75
+ "69": "leonberg",
76
+ "70": "lhasa",
77
+ "71": "malamute",
78
+ "72": "malinois",
79
+ "73": "maltese_dog",
80
+ "74": "mexican_hairless",
81
+ "75": "miniature_pinscher",
82
+ "76": "miniature_poodle",
83
+ "77": "miniature_schnauzer",
84
+ "78": "newfoundland",
85
+ "79": "norfolk_terrier",
86
+ "80": "norwegian_elkhound",
87
+ "81": "norwich_terrier",
88
+ "82": "old_english_sheepdog",
89
+ "83": "otterhound",
90
+ "84": "papillon",
91
+ "85": "pekinese",
92
+ "86": "pembroke",
93
+ "87": "pomeranian",
94
+ "88": "pug",
95
+ "89": "redbone",
96
+ "90": "rhodesian_ridgeback",
97
+ "91": "rottweiler",
98
+ "92": "saint_bernard",
99
+ "93": "saluki",
100
+ "94": "samoyed",
101
+ "95": "schipperke",
102
+ "96": "scotch_terrier",
103
+ "97": "scottish_deerhound",
104
+ "98": "sealyham_terrier",
105
+ "99": "shetland_sheepdog",
106
+ "100": "shih-tzu",
107
+ "101": "siberian_husky",
108
+ "102": "silky_terrier",
109
+ "103": "soft-coated_wheaten_terrier",
110
+ "104": "staffordshire_bullterrier",
111
+ "105": "standard_poodle",
112
+ "106": "standard_schnauzer",
113
+ "107": "sussex_spaniel",
114
+ "108": "test",
115
+ "109": "tibetan_mastiff",
116
+ "110": "tibetan_terrier",
117
+ "111": "toy_poodle",
118
+ "112": "toy_terrier",
119
+ "113": "vizsla",
120
+ "114": "walker_hound",
121
+ "115": "weimaraner",
122
+ "116": "welsh_springer_spaniel",
123
+ "117": "west_highland_white_terrier",
124
+ "118": "whippet",
125
+ "119": "wire-haired_fox_terrier",
126
+ "120": "yorkshire_terrier"
127
+ },
128
+ "initializer_factor": 1.0,
129
+ "label2id": {
130
+ "affenpinscher": 0,
131
+ "afghan_hound": 1,
132
+ "african_hunting_dog": 2,
133
+ "airedale": 3,
134
+ "american_staffordshire_terrier": 4,
135
+ "appenzeller": 5,
136
+ "australian_terrier": 6,
137
+ "basenji": 7,
138
+ "basset": 8,
139
+ "beagle": 9,
140
+ "bedlington_terrier": 10,
141
+ "bernese_mountain_dog": 11,
142
+ "black-and-tan_coonhound": 12,
143
+ "blenheim_spaniel": 13,
144
+ "bloodhound": 14,
145
+ "bluetick": 15,
146
+ "border_collie": 16,
147
+ "border_terrier": 17,
148
+ "borzoi": 18,
149
+ "boston_bull": 19,
150
+ "bouvier_des_flandres": 20,
151
+ "boxer": 21,
152
+ "brabancon_griffon": 22,
153
+ "briard": 23,
154
+ "brittany_spaniel": 24,
155
+ "bull_mastiff": 25,
156
+ "cairn": 26,
157
+ "cardigan": 27,
158
+ "chesapeake_bay_retriever": 28,
159
+ "chihuahua": 29,
160
+ "chow": 30,
161
+ "clumber": 31,
162
+ "cocker_spaniel": 32,
163
+ "collie": 33,
164
+ "curly-coated_retriever": 34,
165
+ "dandie_dinmont": 35,
166
+ "dhole": 36,
167
+ "dingo": 37,
168
+ "doberman": 38,
169
+ "english_foxhound": 39,
170
+ "english_setter": 40,
171
+ "english_springer": 41,
172
+ "entlebucher": 42,
173
+ "eskimo_dog": 43,
174
+ "flat-coated_retriever": 44,
175
+ "french_bulldog": 45,
176
+ "german_shepherd": 46,
177
+ "german_short-haired_pointer": 47,
178
+ "giant_schnauzer": 48,
179
+ "golden_retriever": 49,
180
+ "gordon_setter": 50,
181
+ "great_dane": 51,
182
+ "great_pyrenees": 52,
183
+ "greater_swiss_mountain_dog": 53,
184
+ "groenendael": 54,
185
+ "ibizan_hound": 55,
186
+ "irish_setter": 56,
187
+ "irish_terrier": 57,
188
+ "irish_water_spaniel": 58,
189
+ "irish_wolfhound": 59,
190
+ "italian_greyhound": 60,
191
+ "japanese_spaniel": 61,
192
+ "keeshond": 62,
193
+ "kelpie": 63,
194
+ "kerry_blue_terrier": 64,
195
+ "komondor": 65,
196
+ "kuvasz": 66,
197
+ "labrador_retriever": 67,
198
+ "lakeland_terrier": 68,
199
+ "leonberg": 69,
200
+ "lhasa": 70,
201
+ "malamute": 71,
202
+ "malinois": 72,
203
+ "maltese_dog": 73,
204
+ "mexican_hairless": 74,
205
+ "miniature_pinscher": 75,
206
+ "miniature_poodle": 76,
207
+ "miniature_schnauzer": 77,
208
+ "newfoundland": 78,
209
+ "norfolk_terrier": 79,
210
+ "norwegian_elkhound": 80,
211
+ "norwich_terrier": 81,
212
+ "old_english_sheepdog": 82,
213
+ "otterhound": 83,
214
+ "papillon": 84,
215
+ "pekinese": 85,
216
+ "pembroke": 86,
217
+ "pomeranian": 87,
218
+ "pug": 88,
219
+ "redbone": 89,
220
+ "rhodesian_ridgeback": 90,
221
+ "rottweiler": 91,
222
+ "saint_bernard": 92,
223
+ "saluki": 93,
224
+ "samoyed": 94,
225
+ "schipperke": 95,
226
+ "scotch_terrier": 96,
227
+ "scottish_deerhound": 97,
228
+ "sealyham_terrier": 98,
229
+ "shetland_sheepdog": 99,
230
+ "shih-tzu": 100,
231
+ "siberian_husky": 101,
232
+ "silky_terrier": 102,
233
+ "soft-coated_wheaten_terrier": 103,
234
+ "staffordshire_bullterrier": 104,
235
+ "standard_poodle": 105,
236
+ "standard_schnauzer": 106,
237
+ "sussex_spaniel": 107,
238
+ "test": 108,
239
+ "tibetan_mastiff": 109,
240
+ "tibetan_terrier": 110,
241
+ "toy_poodle": 111,
242
+ "toy_terrier": 112,
243
+ "vizsla": 113,
244
+ "walker_hound": 114,
245
+ "weimaraner": 115,
246
+ "welsh_springer_spaniel": 116,
247
+ "west_highland_white_terrier": 117,
248
+ "whippet": 118,
249
+ "wire-haired_fox_terrier": 119,
250
+ "yorkshire_terrier": 120
251
+ },
252
+ "model_type": "siglip",
253
+ "problem_type": "single_label_classification",
254
+ "text_config": {
255
+ "attention_dropout": 0.0,
256
+ "hidden_act": "gelu_pytorch_tanh",
257
+ "hidden_size": 768,
258
+ "intermediate_size": 3072,
259
+ "layer_norm_eps": 1e-06,
260
+ "max_position_embeddings": 64,
261
+ "model_type": "siglip_text_model",
262
+ "num_attention_heads": 12,
263
+ "num_hidden_layers": 12,
264
+ "projection_size": 768,
265
+ "torch_dtype": "float32",
266
+ "vocab_size": 256000
267
+ },
268
+ "torch_dtype": "float32",
269
+ "transformers_version": "4.51.0.dev0",
270
+ "vision_config": {
271
+ "attention_dropout": 0.0,
272
+ "hidden_act": "gelu_pytorch_tanh",
273
+ "hidden_size": 768,
274
+ "image_size": 224,
275
+ "intermediate_size": 3072,
276
+ "layer_norm_eps": 1e-06,
277
+ "model_type": "siglip_vision_model",
278
+ "num_attention_heads": 12,
279
+ "num_channels": 3,
280
+ "num_hidden_layers": 12,
281
+ "patch_size": 16,
282
+ "torch_dtype": "float32"
283
+ }
284
+ }
checkpoint-2240/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6814b4780e2822e69ee2c1a33cd522fa7ba74a36789843aeb365254b79112c10
3
+ size 371934052
checkpoint-2240/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0ccd3a2bfd69147b275d6edf925a2d10a2086492fe79c139d1d85f40d27d58b3
3
+ size 687287802
checkpoint-2240/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-2240/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3fcd1cba215b586151b7b134576d34afe7270fd1ce028b4b0cd2ba93b7dffb3f
3
+ size 14244
checkpoint-2240/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:52640a825c516d22c7bb2eb03b975892ded9bd8c2fb25d2ebe168844535587a1
3
+ size 1064
checkpoint-2240/trainer_state.json ADDED
@@ -0,0 +1,132 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": 2240,
3
+ "best_metric": 0.49717578291893005,
4
+ "best_model_checkpoint": "siglip2-finetune-full/checkpoint-2240",
5
+ "epoch": 7.0,
6
+ "eval_steps": 500,
7
+ "global_step": 2240,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 1.0,
14
+ "eval_accuracy": 0.48082567012326355,
15
+ "eval_loss": 2.3700547218322754,
16
+ "eval_model_preparation_time": 0.0042,
17
+ "eval_runtime": 153.6453,
18
+ "eval_samples_per_second": 66.53,
19
+ "eval_steps_per_second": 8.318,
20
+ "step": 320
21
+ },
22
+ {
23
+ "epoch": 1.5625,
24
+ "grad_norm": 61.34077453613281,
25
+ "learning_rate": 3.972602739726027e-06,
26
+ "loss": 3.3214,
27
+ "step": 500
28
+ },
29
+ {
30
+ "epoch": 2.0,
31
+ "eval_accuracy": 0.7256896889062806,
32
+ "eval_loss": 1.1354633569717407,
33
+ "eval_model_preparation_time": 0.0042,
34
+ "eval_runtime": 146.0009,
35
+ "eval_samples_per_second": 70.013,
36
+ "eval_steps_per_second": 8.753,
37
+ "step": 640
38
+ },
39
+ {
40
+ "epoch": 3.0,
41
+ "eval_accuracy": 0.7894736842105263,
42
+ "eval_loss": 0.8310916423797607,
43
+ "eval_model_preparation_time": 0.0042,
44
+ "eval_runtime": 146.2788,
45
+ "eval_samples_per_second": 69.88,
46
+ "eval_steps_per_second": 8.737,
47
+ "step": 960
48
+ },
49
+ {
50
+ "epoch": 3.125,
51
+ "grad_norm": 65.03152465820312,
52
+ "learning_rate": 2.8310502283105025e-06,
53
+ "loss": 1.2343,
54
+ "step": 1000
55
+ },
56
+ {
57
+ "epoch": 4.0,
58
+ "eval_accuracy": 0.8205830561533947,
59
+ "eval_loss": 0.6800340414047241,
60
+ "eval_model_preparation_time": 0.0042,
61
+ "eval_runtime": 145.9883,
62
+ "eval_samples_per_second": 70.019,
63
+ "eval_steps_per_second": 8.754,
64
+ "step": 1280
65
+ },
66
+ {
67
+ "epoch": 4.6875,
68
+ "grad_norm": 42.12885665893555,
69
+ "learning_rate": 1.6894977168949772e-06,
70
+ "loss": 0.8261,
71
+ "step": 1500
72
+ },
73
+ {
74
+ "epoch": 5.0,
75
+ "eval_accuracy": 0.8485619252592448,
76
+ "eval_loss": 0.5670026540756226,
77
+ "eval_model_preparation_time": 0.0042,
78
+ "eval_runtime": 146.844,
79
+ "eval_samples_per_second": 69.611,
80
+ "eval_steps_per_second": 8.703,
81
+ "step": 1600
82
+ },
83
+ {
84
+ "epoch": 6.0,
85
+ "eval_accuracy": 0.8628448444531402,
86
+ "eval_loss": 0.5231612324714661,
87
+ "eval_model_preparation_time": 0.0042,
88
+ "eval_runtime": 146.0726,
89
+ "eval_samples_per_second": 69.979,
90
+ "eval_steps_per_second": 8.749,
91
+ "step": 1920
92
+ },
93
+ {
94
+ "epoch": 6.25,
95
+ "grad_norm": 56.17435836791992,
96
+ "learning_rate": 5.47945205479452e-07,
97
+ "loss": 0.6627,
98
+ "step": 2000
99
+ },
100
+ {
101
+ "epoch": 7.0,
102
+ "eval_accuracy": 0.8681275679906085,
103
+ "eval_loss": 0.49717578291893005,
104
+ "eval_model_preparation_time": 0.0042,
105
+ "eval_runtime": 145.7279,
106
+ "eval_samples_per_second": 70.144,
107
+ "eval_steps_per_second": 8.77,
108
+ "step": 2240
109
+ }
110
+ ],
111
+ "logging_steps": 500,
112
+ "max_steps": 2240,
113
+ "num_input_tokens_seen": 0,
114
+ "num_train_epochs": 7,
115
+ "save_steps": 500,
116
+ "stateful_callbacks": {
117
+ "TrainerControl": {
118
+ "args": {
119
+ "should_epoch_stop": false,
120
+ "should_evaluate": false,
121
+ "should_log": false,
122
+ "should_save": true,
123
+ "should_training_stop": true
124
+ },
125
+ "attributes": {}
126
+ }
127
+ },
128
+ "total_flos": 5.998954672277361e+18,
129
+ "train_batch_size": 32,
130
+ "trial_name": null,
131
+ "trial_params": null
132
+ }
checkpoint-2240/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:28a1d0358eb842d6a8e2b23ac90eef5d6694c7d23b3f07aaeb6d419b85372614
3
+ size 5304
config.json ADDED
@@ -0,0 +1,284 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "SiglipForImageClassification"
4
+ ],
5
+ "id2label": {
6
+ "0": "affenpinscher",
7
+ "1": "afghan_hound",
8
+ "2": "african_hunting_dog",
9
+ "3": "airedale",
10
+ "4": "american_staffordshire_terrier",
11
+ "5": "appenzeller",
12
+ "6": "australian_terrier",
13
+ "7": "basenji",
14
+ "8": "basset",
15
+ "9": "beagle",
16
+ "10": "bedlington_terrier",
17
+ "11": "bernese_mountain_dog",
18
+ "12": "black-and-tan_coonhound",
19
+ "13": "blenheim_spaniel",
20
+ "14": "bloodhound",
21
+ "15": "bluetick",
22
+ "16": "border_collie",
23
+ "17": "border_terrier",
24
+ "18": "borzoi",
25
+ "19": "boston_bull",
26
+ "20": "bouvier_des_flandres",
27
+ "21": "boxer",
28
+ "22": "brabancon_griffon",
29
+ "23": "briard",
30
+ "24": "brittany_spaniel",
31
+ "25": "bull_mastiff",
32
+ "26": "cairn",
33
+ "27": "cardigan",
34
+ "28": "chesapeake_bay_retriever",
35
+ "29": "chihuahua",
36
+ "30": "chow",
37
+ "31": "clumber",
38
+ "32": "cocker_spaniel",
39
+ "33": "collie",
40
+ "34": "curly-coated_retriever",
41
+ "35": "dandie_dinmont",
42
+ "36": "dhole",
43
+ "37": "dingo",
44
+ "38": "doberman",
45
+ "39": "english_foxhound",
46
+ "40": "english_setter",
47
+ "41": "english_springer",
48
+ "42": "entlebucher",
49
+ "43": "eskimo_dog",
50
+ "44": "flat-coated_retriever",
51
+ "45": "french_bulldog",
52
+ "46": "german_shepherd",
53
+ "47": "german_short-haired_pointer",
54
+ "48": "giant_schnauzer",
55
+ "49": "golden_retriever",
56
+ "50": "gordon_setter",
57
+ "51": "great_dane",
58
+ "52": "great_pyrenees",
59
+ "53": "greater_swiss_mountain_dog",
60
+ "54": "groenendael",
61
+ "55": "ibizan_hound",
62
+ "56": "irish_setter",
63
+ "57": "irish_terrier",
64
+ "58": "irish_water_spaniel",
65
+ "59": "irish_wolfhound",
66
+ "60": "italian_greyhound",
67
+ "61": "japanese_spaniel",
68
+ "62": "keeshond",
69
+ "63": "kelpie",
70
+ "64": "kerry_blue_terrier",
71
+ "65": "komondor",
72
+ "66": "kuvasz",
73
+ "67": "labrador_retriever",
74
+ "68": "lakeland_terrier",
75
+ "69": "leonberg",
76
+ "70": "lhasa",
77
+ "71": "malamute",
78
+ "72": "malinois",
79
+ "73": "maltese_dog",
80
+ "74": "mexican_hairless",
81
+ "75": "miniature_pinscher",
82
+ "76": "miniature_poodle",
83
+ "77": "miniature_schnauzer",
84
+ "78": "newfoundland",
85
+ "79": "norfolk_terrier",
86
+ "80": "norwegian_elkhound",
87
+ "81": "norwich_terrier",
88
+ "82": "old_english_sheepdog",
89
+ "83": "otterhound",
90
+ "84": "papillon",
91
+ "85": "pekinese",
92
+ "86": "pembroke",
93
+ "87": "pomeranian",
94
+ "88": "pug",
95
+ "89": "redbone",
96
+ "90": "rhodesian_ridgeback",
97
+ "91": "rottweiler",
98
+ "92": "saint_bernard",
99
+ "93": "saluki",
100
+ "94": "samoyed",
101
+ "95": "schipperke",
102
+ "96": "scotch_terrier",
103
+ "97": "scottish_deerhound",
104
+ "98": "sealyham_terrier",
105
+ "99": "shetland_sheepdog",
106
+ "100": "shih-tzu",
107
+ "101": "siberian_husky",
108
+ "102": "silky_terrier",
109
+ "103": "soft-coated_wheaten_terrier",
110
+ "104": "staffordshire_bullterrier",
111
+ "105": "standard_poodle",
112
+ "106": "standard_schnauzer",
113
+ "107": "sussex_spaniel",
114
+ "108": "test",
115
+ "109": "tibetan_mastiff",
116
+ "110": "tibetan_terrier",
117
+ "111": "toy_poodle",
118
+ "112": "toy_terrier",
119
+ "113": "vizsla",
120
+ "114": "walker_hound",
121
+ "115": "weimaraner",
122
+ "116": "welsh_springer_spaniel",
123
+ "117": "west_highland_white_terrier",
124
+ "118": "whippet",
125
+ "119": "wire-haired_fox_terrier",
126
+ "120": "yorkshire_terrier"
127
+ },
128
+ "initializer_factor": 1.0,
129
+ "label2id": {
130
+ "affenpinscher": 0,
131
+ "afghan_hound": 1,
132
+ "african_hunting_dog": 2,
133
+ "airedale": 3,
134
+ "american_staffordshire_terrier": 4,
135
+ "appenzeller": 5,
136
+ "australian_terrier": 6,
137
+ "basenji": 7,
138
+ "basset": 8,
139
+ "beagle": 9,
140
+ "bedlington_terrier": 10,
141
+ "bernese_mountain_dog": 11,
142
+ "black-and-tan_coonhound": 12,
143
+ "blenheim_spaniel": 13,
144
+ "bloodhound": 14,
145
+ "bluetick": 15,
146
+ "border_collie": 16,
147
+ "border_terrier": 17,
148
+ "borzoi": 18,
149
+ "boston_bull": 19,
150
+ "bouvier_des_flandres": 20,
151
+ "boxer": 21,
152
+ "brabancon_griffon": 22,
153
+ "briard": 23,
154
+ "brittany_spaniel": 24,
155
+ "bull_mastiff": 25,
156
+ "cairn": 26,
157
+ "cardigan": 27,
158
+ "chesapeake_bay_retriever": 28,
159
+ "chihuahua": 29,
160
+ "chow": 30,
161
+ "clumber": 31,
162
+ "cocker_spaniel": 32,
163
+ "collie": 33,
164
+ "curly-coated_retriever": 34,
165
+ "dandie_dinmont": 35,
166
+ "dhole": 36,
167
+ "dingo": 37,
168
+ "doberman": 38,
169
+ "english_foxhound": 39,
170
+ "english_setter": 40,
171
+ "english_springer": 41,
172
+ "entlebucher": 42,
173
+ "eskimo_dog": 43,
174
+ "flat-coated_retriever": 44,
175
+ "french_bulldog": 45,
176
+ "german_shepherd": 46,
177
+ "german_short-haired_pointer": 47,
178
+ "giant_schnauzer": 48,
179
+ "golden_retriever": 49,
180
+ "gordon_setter": 50,
181
+ "great_dane": 51,
182
+ "great_pyrenees": 52,
183
+ "greater_swiss_mountain_dog": 53,
184
+ "groenendael": 54,
185
+ "ibizan_hound": 55,
186
+ "irish_setter": 56,
187
+ "irish_terrier": 57,
188
+ "irish_water_spaniel": 58,
189
+ "irish_wolfhound": 59,
190
+ "italian_greyhound": 60,
191
+ "japanese_spaniel": 61,
192
+ "keeshond": 62,
193
+ "kelpie": 63,
194
+ "kerry_blue_terrier": 64,
195
+ "komondor": 65,
196
+ "kuvasz": 66,
197
+ "labrador_retriever": 67,
198
+ "lakeland_terrier": 68,
199
+ "leonberg": 69,
200
+ "lhasa": 70,
201
+ "malamute": 71,
202
+ "malinois": 72,
203
+ "maltese_dog": 73,
204
+ "mexican_hairless": 74,
205
+ "miniature_pinscher": 75,
206
+ "miniature_poodle": 76,
207
+ "miniature_schnauzer": 77,
208
+ "newfoundland": 78,
209
+ "norfolk_terrier": 79,
210
+ "norwegian_elkhound": 80,
211
+ "norwich_terrier": 81,
212
+ "old_english_sheepdog": 82,
213
+ "otterhound": 83,
214
+ "papillon": 84,
215
+ "pekinese": 85,
216
+ "pembroke": 86,
217
+ "pomeranian": 87,
218
+ "pug": 88,
219
+ "redbone": 89,
220
+ "rhodesian_ridgeback": 90,
221
+ "rottweiler": 91,
222
+ "saint_bernard": 92,
223
+ "saluki": 93,
224
+ "samoyed": 94,
225
+ "schipperke": 95,
226
+ "scotch_terrier": 96,
227
+ "scottish_deerhound": 97,
228
+ "sealyham_terrier": 98,
229
+ "shetland_sheepdog": 99,
230
+ "shih-tzu": 100,
231
+ "siberian_husky": 101,
232
+ "silky_terrier": 102,
233
+ "soft-coated_wheaten_terrier": 103,
234
+ "staffordshire_bullterrier": 104,
235
+ "standard_poodle": 105,
236
+ "standard_schnauzer": 106,
237
+ "sussex_spaniel": 107,
238
+ "test": 108,
239
+ "tibetan_mastiff": 109,
240
+ "tibetan_terrier": 110,
241
+ "toy_poodle": 111,
242
+ "toy_terrier": 112,
243
+ "vizsla": 113,
244
+ "walker_hound": 114,
245
+ "weimaraner": 115,
246
+ "welsh_springer_spaniel": 116,
247
+ "west_highland_white_terrier": 117,
248
+ "whippet": 118,
249
+ "wire-haired_fox_terrier": 119,
250
+ "yorkshire_terrier": 120
251
+ },
252
+ "model_type": "siglip",
253
+ "problem_type": "single_label_classification",
254
+ "text_config": {
255
+ "attention_dropout": 0.0,
256
+ "hidden_act": "gelu_pytorch_tanh",
257
+ "hidden_size": 768,
258
+ "intermediate_size": 3072,
259
+ "layer_norm_eps": 1e-06,
260
+ "max_position_embeddings": 64,
261
+ "model_type": "siglip_text_model",
262
+ "num_attention_heads": 12,
263
+ "num_hidden_layers": 12,
264
+ "projection_size": 768,
265
+ "torch_dtype": "float32",
266
+ "vocab_size": 256000
267
+ },
268
+ "torch_dtype": "float32",
269
+ "transformers_version": "4.51.0.dev0",
270
+ "vision_config": {
271
+ "attention_dropout": 0.0,
272
+ "hidden_act": "gelu_pytorch_tanh",
273
+ "hidden_size": 768,
274
+ "image_size": 224,
275
+ "intermediate_size": 3072,
276
+ "layer_norm_eps": 1e-06,
277
+ "model_type": "siglip_vision_model",
278
+ "num_attention_heads": 12,
279
+ "num_channels": 3,
280
+ "num_hidden_layers": 12,
281
+ "patch_size": 16,
282
+ "torch_dtype": "float32"
283
+ }
284
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6814b4780e2822e69ee2c1a33cd522fa7ba74a36789843aeb365254b79112c10
3
+ size 371934052
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 sha256:28a1d0358eb842d6a8e2b23ac90eef5d6694c7d23b3f07aaeb6d419b85372614
3
+ size 5304