Hugging Face's logo

prithivMLmods
/
Fashion-Product-baseColour

Image Classification
Transformers
Safetensors
English
siglip
color
cloth
Model card Files
xet
prithivMLmods commited on
Commit
da51a68
·
verified ·
1 Parent(s): bfa96fe

Upload folder using huggingface_hub

Browse files
Files changed (28) hide show
  1. checkpoint-1378/config.json +134 -0
  2. checkpoint-1378/model.safetensors +3 -0
  3. checkpoint-1378/optimizer.pt +3 -0
  4. checkpoint-1378/preprocessor_config.json +24 -0
  5. checkpoint-1378/rng_state.pth +3 -0
  6. checkpoint-1378/scheduler.pt +3 -0
  7. checkpoint-1378/trainer_state.json +58 -0
  8. checkpoint-1378/training_args.bin +3 -0
  9. checkpoint-2756/config.json +134 -0
  10. checkpoint-2756/model.safetensors +3 -0
  11. checkpoint-2756/optimizer.pt +3 -0
  12. checkpoint-2756/preprocessor_config.json +24 -0
  13. checkpoint-2756/rng_state.pth +3 -0
  14. checkpoint-2756/scheduler.pt +3 -0
  15. checkpoint-2756/trainer_state.json +89 -0
  16. checkpoint-2756/training_args.bin +3 -0
  17. checkpoint-4134/config.json +134 -0
  18. checkpoint-4134/model.safetensors +3 -0
  19. checkpoint-4134/optimizer.pt +3 -0
  20. checkpoint-4134/preprocessor_config.json +24 -0
  21. checkpoint-4134/rng_state.pth +3 -0
  22. checkpoint-4134/scheduler.pt +3 -0
  23. checkpoint-4134/trainer_state.json +120 -0
  24. checkpoint-4134/training_args.bin +3 -0
  25. config.json +134 -0
  26. model.safetensors +3 -0
  27. preprocessor_config.json +24 -0
  28. training_args.bin +3 -0
checkpoint-1378/config.json ADDED
@@ -0,0 +1,134 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "SiglipForImageClassification"
4
+ ],
5
+ "id2label": {
6
+ "0": "Beige",
7
+ "1": "Black",
8
+ "2": "Blue",
9
+ "3": "Bronze",
10
+ "4": "Brown",
11
+ "5": "Burgundy",
12
+ "6": "Charcoal",
13
+ "7": "Coffee Brown",
14
+ "8": "Copper",
15
+ "9": "Cream",
16
+ "10": "Fluorescent Green",
17
+ "11": "Gold",
18
+ "12": "Green",
19
+ "13": "Grey",
20
+ "14": "Grey Melange",
21
+ "15": "Khaki",
22
+ "16": "Lavender",
23
+ "17": "Lime Green",
24
+ "18": "Magenta",
25
+ "19": "Maroon",
26
+ "20": "Mauve",
27
+ "21": "Metallic",
28
+ "22": "Multi",
29
+ "23": "Mushroom Brown",
30
+ "24": "Mustard",
31
+ "25": "Navy Blue",
32
+ "26": "Nude",
33
+ "27": "Off White",
34
+ "28": "Olive",
35
+ "29": "Orange",
36
+ "30": "Peach",
37
+ "31": "Pink",
38
+ "32": "Purple",
39
+ "33": "Red",
40
+ "34": "Rose",
41
+ "35": "Rust",
42
+ "36": "Sea Green",
43
+ "37": "Silver",
44
+ "38": "Skin",
45
+ "39": "Steel",
46
+ "40": "Tan",
47
+ "41": "Taupe",
48
+ "42": "Teal",
49
+ "43": "Turquoise Blue",
50
+ "44": "White",
51
+ "45": "Yellow"
52
+ },
53
+ "initializer_factor": 1.0,
54
+ "label2id": {
55
+ "Beige": 0,
56
+ "Black": 1,
57
+ "Blue": 2,
58
+ "Bronze": 3,
59
+ "Brown": 4,
60
+ "Burgundy": 5,
61
+ "Charcoal": 6,
62
+ "Coffee Brown": 7,
63
+ "Copper": 8,
64
+ "Cream": 9,
65
+ "Fluorescent Green": 10,
66
+ "Gold": 11,
67
+ "Green": 12,
68
+ "Grey": 13,
69
+ "Grey Melange": 14,
70
+ "Khaki": 15,
71
+ "Lavender": 16,
72
+ "Lime Green": 17,
73
+ "Magenta": 18,
74
+ "Maroon": 19,
75
+ "Mauve": 20,
76
+ "Metallic": 21,
77
+ "Multi": 22,
78
+ "Mushroom Brown": 23,
79
+ "Mustard": 24,
80
+ "Navy Blue": 25,
81
+ "Nude": 26,
82
+ "Off White": 27,
83
+ "Olive": 28,
84
+ "Orange": 29,
85
+ "Peach": 30,
86
+ "Pink": 31,
87
+ "Purple": 32,
88
+ "Red": 33,
89
+ "Rose": 34,
90
+ "Rust": 35,
91
+ "Sea Green": 36,
92
+ "Silver": 37,
93
+ "Skin": 38,
94
+ "Steel": 39,
95
+ "Tan": 40,
96
+ "Taupe": 41,
97
+ "Teal": 42,
98
+ "Turquoise Blue": 43,
99
+ "White": 44,
100
+ "Yellow": 45
101
+ },
102
+ "model_type": "siglip",
103
+ "problem_type": "single_label_classification",
104
+ "text_config": {
105
+ "attention_dropout": 0.0,
106
+ "hidden_act": "gelu_pytorch_tanh",
107
+ "hidden_size": 768,
108
+ "intermediate_size": 3072,
109
+ "layer_norm_eps": 1e-06,
110
+ "max_position_embeddings": 64,
111
+ "model_type": "siglip_text_model",
112
+ "num_attention_heads": 12,
113
+ "num_hidden_layers": 12,
114
+ "projection_size": 768,
115
+ "torch_dtype": "float32",
116
+ "vocab_size": 256000
117
+ },
118
+ "torch_dtype": "float32",
119
+ "transformers_version": "4.50.3",
120
+ "vision_config": {
121
+ "attention_dropout": 0.0,
122
+ "hidden_act": "gelu_pytorch_tanh",
123
+ "hidden_size": 768,
124
+ "image_size": 224,
125
+ "intermediate_size": 3072,
126
+ "layer_norm_eps": 1e-06,
127
+ "model_type": "siglip_vision_model",
128
+ "num_attention_heads": 12,
129
+ "num_channels": 3,
130
+ "num_hidden_layers": 12,
131
+ "patch_size": 16,
132
+ "torch_dtype": "float32"
133
+ }
134
+ }
checkpoint-1378/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:325cdf2fefb1bd586e7c7f1b7f7e0d96613c874a15d1bdbaad0918696c040c13
3
+ size 371703344
checkpoint-1378/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e8bcf562a47642cf847e3cd410d3eda33fe697a44c162bbcd32e240fc2019e2
3
+ size 686826362
checkpoint-1378/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-1378/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:be3800323adfbabbfee073f7c32c3d86bc0c9fd7937c41a132afa461c6c74cbe
3
+ size 14244
checkpoint-1378/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9369375317f777819b50f0e173e894c4d4bef7c15bb65b3cf3522f34d2c7ad99
3
+ size 1064
checkpoint-1378/trainer_state.json ADDED
@@ -0,0 +1,58 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": 1378,
3
+ "best_metric": 1.3363316059112549,
4
+ "best_model_checkpoint": "siglip2-finetune-full/checkpoint-1378",
5
+ "epoch": 1.0,
6
+ "eval_steps": 500,
7
+ "global_step": 1378,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.36284470246734396,
14
+ "grad_norm": 5.684016704559326,
15
+ "learning_rate": 0.00017796278158667975,
16
+ "loss": 1.841,
17
+ "step": 500
18
+ },
19
+ {
20
+ "epoch": 0.7256894049346879,
21
+ "grad_norm": 4.77960205078125,
22
+ "learning_rate": 0.0001534769833496572,
23
+ "loss": 1.4093,
24
+ "step": 1000
25
+ },
26
+ {
27
+ "epoch": 1.0,
28
+ "eval_accuracy": 0.6121346886912326,
29
+ "eval_loss": 1.3363316059112549,
30
+ "eval_model_preparation_time": 0.0023,
31
+ "eval_runtime": 540.1365,
32
+ "eval_samples_per_second": 81.594,
33
+ "eval_steps_per_second": 10.199,
34
+ "step": 1378
35
+ }
36
+ ],
37
+ "logging_steps": 500,
38
+ "max_steps": 4134,
39
+ "num_input_tokens_seen": 0,
40
+ "num_train_epochs": 3,
41
+ "save_steps": 500,
42
+ "stateful_callbacks": {
43
+ "TrainerControl": {
44
+ "args": {
45
+ "should_epoch_stop": false,
46
+ "should_evaluate": false,
47
+ "should_log": false,
48
+ "should_save": true,
49
+ "should_training_stop": false
50
+ },
51
+ "attributes": {}
52
+ }
53
+ },
54
+ "total_flos": 3.6926190341952307e+18,
55
+ "train_batch_size": 32,
56
+ "trial_name": null,
57
+ "trial_params": null
58
+ }
checkpoint-1378/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d732915852aa8238abe1f128b3231ab9d60325b25b8e62363150b9e372e24184
3
+ size 5304
checkpoint-2756/config.json ADDED
@@ -0,0 +1,134 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "SiglipForImageClassification"
4
+ ],
5
+ "id2label": {
6
+ "0": "Beige",
7
+ "1": "Black",
8
+ "2": "Blue",
9
+ "3": "Bronze",
10
+ "4": "Brown",
11
+ "5": "Burgundy",
12
+ "6": "Charcoal",
13
+ "7": "Coffee Brown",
14
+ "8": "Copper",
15
+ "9": "Cream",
16
+ "10": "Fluorescent Green",
17
+ "11": "Gold",
18
+ "12": "Green",
19
+ "13": "Grey",
20
+ "14": "Grey Melange",
21
+ "15": "Khaki",
22
+ "16": "Lavender",
23
+ "17": "Lime Green",
24
+ "18": "Magenta",
25
+ "19": "Maroon",
26
+ "20": "Mauve",
27
+ "21": "Metallic",
28
+ "22": "Multi",
29
+ "23": "Mushroom Brown",
30
+ "24": "Mustard",
31
+ "25": "Navy Blue",
32
+ "26": "Nude",
33
+ "27": "Off White",
34
+ "28": "Olive",
35
+ "29": "Orange",
36
+ "30": "Peach",
37
+ "31": "Pink",
38
+ "32": "Purple",
39
+ "33": "Red",
40
+ "34": "Rose",
41
+ "35": "Rust",
42
+ "36": "Sea Green",
43
+ "37": "Silver",
44
+ "38": "Skin",
45
+ "39": "Steel",
46
+ "40": "Tan",
47
+ "41": "Taupe",
48
+ "42": "Teal",
49
+ "43": "Turquoise Blue",
50
+ "44": "White",
51
+ "45": "Yellow"
52
+ },
53
+ "initializer_factor": 1.0,
54
+ "label2id": {
55
+ "Beige": 0,
56
+ "Black": 1,
57
+ "Blue": 2,
58
+ "Bronze": 3,
59
+ "Brown": 4,
60
+ "Burgundy": 5,
61
+ "Charcoal": 6,
62
+ "Coffee Brown": 7,
63
+ "Copper": 8,
64
+ "Cream": 9,
65
+ "Fluorescent Green": 10,
66
+ "Gold": 11,
67
+ "Green": 12,
68
+ "Grey": 13,
69
+ "Grey Melange": 14,
70
+ "Khaki": 15,
71
+ "Lavender": 16,
72
+ "Lime Green": 17,
73
+ "Magenta": 18,
74
+ "Maroon": 19,
75
+ "Mauve": 20,
76
+ "Metallic": 21,
77
+ "Multi": 22,
78
+ "Mushroom Brown": 23,
79
+ "Mustard": 24,
80
+ "Navy Blue": 25,
81
+ "Nude": 26,
82
+ "Off White": 27,
83
+ "Olive": 28,
84
+ "Orange": 29,
85
+ "Peach": 30,
86
+ "Pink": 31,
87
+ "Purple": 32,
88
+ "Red": 33,
89
+ "Rose": 34,
90
+ "Rust": 35,
91
+ "Sea Green": 36,
92
+ "Silver": 37,
93
+ "Skin": 38,
94
+ "Steel": 39,
95
+ "Tan": 40,
96
+ "Taupe": 41,
97
+ "Teal": 42,
98
+ "Turquoise Blue": 43,
99
+ "White": 44,
100
+ "Yellow": 45
101
+ },
102
+ "model_type": "siglip",
103
+ "problem_type": "single_label_classification",
104
+ "text_config": {
105
+ "attention_dropout": 0.0,
106
+ "hidden_act": "gelu_pytorch_tanh",
107
+ "hidden_size": 768,
108
+ "intermediate_size": 3072,
109
+ "layer_norm_eps": 1e-06,
110
+ "max_position_embeddings": 64,
111
+ "model_type": "siglip_text_model",
112
+ "num_attention_heads": 12,
113
+ "num_hidden_layers": 12,
114
+ "projection_size": 768,
115
+ "torch_dtype": "float32",
116
+ "vocab_size": 256000
117
+ },
118
+ "torch_dtype": "float32",
119
+ "transformers_version": "4.50.3",
120
+ "vision_config": {
121
+ "attention_dropout": 0.0,
122
+ "hidden_act": "gelu_pytorch_tanh",
123
+ "hidden_size": 768,
124
+ "image_size": 224,
125
+ "intermediate_size": 3072,
126
+ "layer_norm_eps": 1e-06,
127
+ "model_type": "siglip_vision_model",
128
+ "num_attention_heads": 12,
129
+ "num_channels": 3,
130
+ "num_hidden_layers": 12,
131
+ "patch_size": 16,
132
+ "torch_dtype": "float32"
133
+ }
134
+ }
checkpoint-2756/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:965466928739f5fc5b80340a0eec500128e66791c27ed18a54e1e1701c807f71
3
+ size 371703344
checkpoint-2756/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6a886655f653580ec366a19b67d57b9d403c75a105ce97b68c2ce803c935c812
3
+ size 686826362
checkpoint-2756/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-2756/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f6895266ce7775b656647ddf97f9b07ff3ff78add76255d92106d9353ec64e09
3
+ size 14244
checkpoint-2756/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c79b378e5fed78f4780ae514ecb06d57ec850332911343c9350a774ee53bb850
3
+ size 1064
checkpoint-2756/trainer_state.json ADDED
@@ -0,0 +1,89 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": 2756,
3
+ "best_metric": 1.0952644348144531,
4
+ "best_model_checkpoint": "siglip2-finetune-full/checkpoint-2756",
5
+ "epoch": 2.0,
6
+ "eval_steps": 500,
7
+ "global_step": 2756,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.36284470246734396,
14
+ "grad_norm": 5.684016704559326,
15
+ "learning_rate": 0.00017796278158667975,
16
+ "loss": 1.841,
17
+ "step": 500
18
+ },
19
+ {
20
+ "epoch": 0.7256894049346879,
21
+ "grad_norm": 4.77960205078125,
22
+ "learning_rate": 0.0001534769833496572,
23
+ "loss": 1.4093,
24
+ "step": 1000
25
+ },
26
+ {
27
+ "epoch": 1.0,
28
+ "eval_accuracy": 0.6121346886912326,
29
+ "eval_loss": 1.3363316059112549,
30
+ "eval_model_preparation_time": 0.0023,
31
+ "eval_runtime": 540.1365,
32
+ "eval_samples_per_second": 81.594,
33
+ "eval_steps_per_second": 10.199,
34
+ "step": 1378
35
+ },
36
+ {
37
+ "epoch": 1.0885341074020318,
38
+ "grad_norm": 4.179084300994873,
39
+ "learning_rate": 0.00012899118511263468,
40
+ "loss": 1.3033,
41
+ "step": 1500
42
+ },
43
+ {
44
+ "epoch": 1.4513788098693758,
45
+ "grad_norm": 4.6432366371154785,
46
+ "learning_rate": 0.00010450538687561216,
47
+ "loss": 1.2174,
48
+ "step": 2000
49
+ },
50
+ {
51
+ "epoch": 1.8142235123367199,
52
+ "grad_norm": 2.942708969116211,
53
+ "learning_rate": 8.001958863858962e-05,
54
+ "loss": 1.1675,
55
+ "step": 2500
56
+ },
57
+ {
58
+ "epoch": 2.0,
59
+ "eval_accuracy": 0.6670902160101652,
60
+ "eval_loss": 1.0952644348144531,
61
+ "eval_model_preparation_time": 0.0023,
62
+ "eval_runtime": 535.0051,
63
+ "eval_samples_per_second": 82.377,
64
+ "eval_steps_per_second": 10.297,
65
+ "step": 2756
66
+ }
67
+ ],
68
+ "logging_steps": 500,
69
+ "max_steps": 4134,
70
+ "num_input_tokens_seen": 0,
71
+ "num_train_epochs": 3,
72
+ "save_steps": 500,
73
+ "stateful_callbacks": {
74
+ "TrainerControl": {
75
+ "args": {
76
+ "should_epoch_stop": false,
77
+ "should_evaluate": false,
78
+ "should_log": false,
79
+ "should_save": true,
80
+ "should_training_stop": false
81
+ },
82
+ "attributes": {}
83
+ }
84
+ },
85
+ "total_flos": 7.385238068390461e+18,
86
+ "train_batch_size": 32,
87
+ "trial_name": null,
88
+ "trial_params": null
89
+ }
checkpoint-2756/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d732915852aa8238abe1f128b3231ab9d60325b25b8e62363150b9e372e24184
3
+ size 5304
checkpoint-4134/config.json ADDED
@@ -0,0 +1,134 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "SiglipForImageClassification"
4
+ ],
5
+ "id2label": {
6
+ "0": "Beige",
7
+ "1": "Black",
8
+ "2": "Blue",
9
+ "3": "Bronze",
10
+ "4": "Brown",
11
+ "5": "Burgundy",
12
+ "6": "Charcoal",
13
+ "7": "Coffee Brown",
14
+ "8": "Copper",
15
+ "9": "Cream",
16
+ "10": "Fluorescent Green",
17
+ "11": "Gold",
18
+ "12": "Green",
19
+ "13": "Grey",
20
+ "14": "Grey Melange",
21
+ "15": "Khaki",
22
+ "16": "Lavender",
23
+ "17": "Lime Green",
24
+ "18": "Magenta",
25
+ "19": "Maroon",
26
+ "20": "Mauve",
27
+ "21": "Metallic",
28
+ "22": "Multi",
29
+ "23": "Mushroom Brown",
30
+ "24": "Mustard",
31
+ "25": "Navy Blue",
32
+ "26": "Nude",
33
+ "27": "Off White",
34
+ "28": "Olive",
35
+ "29": "Orange",
36
+ "30": "Peach",
37
+ "31": "Pink",
38
+ "32": "Purple",
39
+ "33": "Red",
40
+ "34": "Rose",
41
+ "35": "Rust",
42
+ "36": "Sea Green",
43
+ "37": "Silver",
44
+ "38": "Skin",
45
+ "39": "Steel",
46
+ "40": "Tan",
47
+ "41": "Taupe",
48
+ "42": "Teal",
49
+ "43": "Turquoise Blue",
50
+ "44": "White",
51
+ "45": "Yellow"
52
+ },
53
+ "initializer_factor": 1.0,
54
+ "label2id": {
55
+ "Beige": 0,
56
+ "Black": 1,
57
+ "Blue": 2,
58
+ "Bronze": 3,
59
+ "Brown": 4,
60
+ "Burgundy": 5,
61
+ "Charcoal": 6,
62
+ "Coffee Brown": 7,
63
+ "Copper": 8,
64
+ "Cream": 9,
65
+ "Fluorescent Green": 10,
66
+ "Gold": 11,
67
+ "Green": 12,
68
+ "Grey": 13,
69
+ "Grey Melange": 14,
70
+ "Khaki": 15,
71
+ "Lavender": 16,
72
+ "Lime Green": 17,
73
+ "Magenta": 18,
74
+ "Maroon": 19,
75
+ "Mauve": 20,
76
+ "Metallic": 21,
77
+ "Multi": 22,
78
+ "Mushroom Brown": 23,
79
+ "Mustard": 24,
80
+ "Navy Blue": 25,
81
+ "Nude": 26,
82
+ "Off White": 27,
83
+ "Olive": 28,
84
+ "Orange": 29,
85
+ "Peach": 30,
86
+ "Pink": 31,
87
+ "Purple": 32,
88
+ "Red": 33,
89
+ "Rose": 34,
90
+ "Rust": 35,
91
+ "Sea Green": 36,
92
+ "Silver": 37,
93
+ "Skin": 38,
94
+ "Steel": 39,
95
+ "Tan": 40,
96
+ "Taupe": 41,
97
+ "Teal": 42,
98
+ "Turquoise Blue": 43,
99
+ "White": 44,
100
+ "Yellow": 45
101
+ },
102
+ "model_type": "siglip",
103
+ "problem_type": "single_label_classification",
104
+ "text_config": {
105
+ "attention_dropout": 0.0,
106
+ "hidden_act": "gelu_pytorch_tanh",
107
+ "hidden_size": 768,
108
+ "intermediate_size": 3072,
109
+ "layer_norm_eps": 1e-06,
110
+ "max_position_embeddings": 64,
111
+ "model_type": "siglip_text_model",
112
+ "num_attention_heads": 12,
113
+ "num_hidden_layers": 12,
114
+ "projection_size": 768,
115
+ "torch_dtype": "float32",
116
+ "vocab_size": 256000
117
+ },
118
+ "torch_dtype": "float32",
119
+ "transformers_version": "4.50.3",
120
+ "vision_config": {
121
+ "attention_dropout": 0.0,
122
+ "hidden_act": "gelu_pytorch_tanh",
123
+ "hidden_size": 768,
124
+ "image_size": 224,
125
+ "intermediate_size": 3072,
126
+ "layer_norm_eps": 1e-06,
127
+ "model_type": "siglip_vision_model",
128
+ "num_attention_heads": 12,
129
+ "num_channels": 3,
130
+ "num_hidden_layers": 12,
131
+ "patch_size": 16,
132
+ "torch_dtype": "float32"
133
+ }
134
+ }
checkpoint-4134/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a00e319522ea985fd7b18a3c0064a602bb96a2a7558ad4dcf9075bca072ce2df
3
+ size 371703344
checkpoint-4134/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:48809d260b6c86e88f59e55f01a91709f56d484f3f3adaffba37f4fa9488fdc6
3
+ size 686826362
checkpoint-4134/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-4134/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5478a919be85a579b7e7075ad64e2b2fdb812bb5f7dc5b68704fe8bb15677b4e
3
+ size 14244
checkpoint-4134/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5763790c9bc4f7cfec8363dc10ae41a0e5ee6799dd6abd8a934646dd824b40b1
3
+ size 1064
checkpoint-4134/trainer_state.json ADDED
@@ -0,0 +1,120 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": 4134,
3
+ "best_metric": 0.9364688992500305,
4
+ "best_model_checkpoint": "siglip2-finetune-full/checkpoint-4134",
5
+ "epoch": 3.0,
6
+ "eval_steps": 500,
7
+ "global_step": 4134,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.36284470246734396,
14
+ "grad_norm": 5.684016704559326,
15
+ "learning_rate": 0.00017796278158667975,
16
+ "loss": 1.841,
17
+ "step": 500
18
+ },
19
+ {
20
+ "epoch": 0.7256894049346879,
21
+ "grad_norm": 4.77960205078125,
22
+ "learning_rate": 0.0001534769833496572,
23
+ "loss": 1.4093,
24
+ "step": 1000
25
+ },
26
+ {
27
+ "epoch": 1.0,
28
+ "eval_accuracy": 0.6121346886912326,
29
+ "eval_loss": 1.3363316059112549,
30
+ "eval_model_preparation_time": 0.0023,
31
+ "eval_runtime": 540.1365,
32
+ "eval_samples_per_second": 81.594,
33
+ "eval_steps_per_second": 10.199,
34
+ "step": 1378
35
+ },
36
+ {
37
+ "epoch": 1.0885341074020318,
38
+ "grad_norm": 4.179084300994873,
39
+ "learning_rate": 0.00012899118511263468,
40
+ "loss": 1.3033,
41
+ "step": 1500
42
+ },
43
+ {
44
+ "epoch": 1.4513788098693758,
45
+ "grad_norm": 4.6432366371154785,
46
+ "learning_rate": 0.00010450538687561216,
47
+ "loss": 1.2174,
48
+ "step": 2000
49
+ },
50
+ {
51
+ "epoch": 1.8142235123367199,
52
+ "grad_norm": 2.942708969116211,
53
+ "learning_rate": 8.001958863858962e-05,
54
+ "loss": 1.1675,
55
+ "step": 2500
56
+ },
57
+ {
58
+ "epoch": 2.0,
59
+ "eval_accuracy": 0.6670902160101652,
60
+ "eval_loss": 1.0952644348144531,
61
+ "eval_model_preparation_time": 0.0023,
62
+ "eval_runtime": 535.0051,
63
+ "eval_samples_per_second": 82.377,
64
+ "eval_steps_per_second": 10.297,
65
+ "step": 2756
66
+ },
67
+ {
68
+ "epoch": 2.1770682148040637,
69
+ "grad_norm": 4.254842281341553,
70
+ "learning_rate": 5.55337904015671e-05,
71
+ "loss": 1.1002,
72
+ "step": 3000
73
+ },
74
+ {
75
+ "epoch": 2.539912917271408,
76
+ "grad_norm": 3.6314480304718018,
77
+ "learning_rate": 3.104799216454456e-05,
78
+ "loss": 1.0287,
79
+ "step": 3500
80
+ },
81
+ {
82
+ "epoch": 2.9027576197387517,
83
+ "grad_norm": 3.816922664642334,
84
+ "learning_rate": 6.562193927522038e-06,
85
+ "loss": 0.9905,
86
+ "step": 4000
87
+ },
88
+ {
89
+ "epoch": 3.0,
90
+ "eval_accuracy": 0.7072290796877836,
91
+ "eval_loss": 0.9364688992500305,
92
+ "eval_model_preparation_time": 0.0023,
93
+ "eval_runtime": 534.5396,
94
+ "eval_samples_per_second": 82.449,
95
+ "eval_steps_per_second": 10.306,
96
+ "step": 4134
97
+ }
98
+ ],
99
+ "logging_steps": 500,
100
+ "max_steps": 4134,
101
+ "num_input_tokens_seen": 0,
102
+ "num_train_epochs": 3,
103
+ "save_steps": 500,
104
+ "stateful_callbacks": {
105
+ "TrainerControl": {
106
+ "args": {
107
+ "should_epoch_stop": false,
108
+ "should_evaluate": false,
109
+ "should_log": false,
110
+ "should_save": true,
111
+ "should_training_stop": true
112
+ },
113
+ "attributes": {}
114
+ }
115
+ },
116
+ "total_flos": 1.1077857102585692e+19,
117
+ "train_batch_size": 32,
118
+ "trial_name": null,
119
+ "trial_params": null
120
+ }
checkpoint-4134/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d732915852aa8238abe1f128b3231ab9d60325b25b8e62363150b9e372e24184
3
+ size 5304
config.json ADDED
@@ -0,0 +1,134 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "SiglipForImageClassification"
4
+ ],
5
+ "id2label": {
6
+ "0": "Beige",
7
+ "1": "Black",
8
+ "2": "Blue",
9
+ "3": "Bronze",
10
+ "4": "Brown",
11
+ "5": "Burgundy",
12
+ "6": "Charcoal",
13
+ "7": "Coffee Brown",
14
+ "8": "Copper",
15
+ "9": "Cream",
16
+ "10": "Fluorescent Green",
17
+ "11": "Gold",
18
+ "12": "Green",
19
+ "13": "Grey",
20
+ "14": "Grey Melange",
21
+ "15": "Khaki",
22
+ "16": "Lavender",
23
+ "17": "Lime Green",
24
+ "18": "Magenta",
25
+ "19": "Maroon",
26
+ "20": "Mauve",
27
+ "21": "Metallic",
28
+ "22": "Multi",
29
+ "23": "Mushroom Brown",
30
+ "24": "Mustard",
31
+ "25": "Navy Blue",
32
+ "26": "Nude",
33
+ "27": "Off White",
34
+ "28": "Olive",
35
+ "29": "Orange",
36
+ "30": "Peach",
37
+ "31": "Pink",
38
+ "32": "Purple",
39
+ "33": "Red",
40
+ "34": "Rose",
41
+ "35": "Rust",
42
+ "36": "Sea Green",
43
+ "37": "Silver",
44
+ "38": "Skin",
45
+ "39": "Steel",
46
+ "40": "Tan",
47
+ "41": "Taupe",
48
+ "42": "Teal",
49
+ "43": "Turquoise Blue",
50
+ "44": "White",
51
+ "45": "Yellow"
52
+ },
53
+ "initializer_factor": 1.0,
54
+ "label2id": {
55
+ "Beige": 0,
56
+ "Black": 1,
57
+ "Blue": 2,
58
+ "Bronze": 3,
59
+ "Brown": 4,
60
+ "Burgundy": 5,
61
+ "Charcoal": 6,
62
+ "Coffee Brown": 7,
63
+ "Copper": 8,
64
+ "Cream": 9,
65
+ "Fluorescent Green": 10,
66
+ "Gold": 11,
67
+ "Green": 12,
68
+ "Grey": 13,
69
+ "Grey Melange": 14,
70
+ "Khaki": 15,
71
+ "Lavender": 16,
72
+ "Lime Green": 17,
73
+ "Magenta": 18,
74
+ "Maroon": 19,
75
+ "Mauve": 20,
76
+ "Metallic": 21,
77
+ "Multi": 22,
78
+ "Mushroom Brown": 23,
79
+ "Mustard": 24,
80
+ "Navy Blue": 25,
81
+ "Nude": 26,
82
+ "Off White": 27,
83
+ "Olive": 28,
84
+ "Orange": 29,
85
+ "Peach": 30,
86
+ "Pink": 31,
87
+ "Purple": 32,
88
+ "Red": 33,
89
+ "Rose": 34,
90
+ "Rust": 35,
91
+ "Sea Green": 36,
92
+ "Silver": 37,
93
+ "Skin": 38,
94
+ "Steel": 39,
95
+ "Tan": 40,
96
+ "Taupe": 41,
97
+ "Teal": 42,
98
+ "Turquoise Blue": 43,
99
+ "White": 44,
100
+ "Yellow": 45
101
+ },
102
+ "model_type": "siglip",
103
+ "problem_type": "single_label_classification",
104
+ "text_config": {
105
+ "attention_dropout": 0.0,
106
+ "hidden_act": "gelu_pytorch_tanh",
107
+ "hidden_size": 768,
108
+ "intermediate_size": 3072,
109
+ "layer_norm_eps": 1e-06,
110
+ "max_position_embeddings": 64,
111
+ "model_type": "siglip_text_model",
112
+ "num_attention_heads": 12,
113
+ "num_hidden_layers": 12,
114
+ "projection_size": 768,
115
+ "torch_dtype": "float32",
116
+ "vocab_size": 256000
117
+ },
118
+ "torch_dtype": "float32",
119
+ "transformers_version": "4.50.3",
120
+ "vision_config": {
121
+ "attention_dropout": 0.0,
122
+ "hidden_act": "gelu_pytorch_tanh",
123
+ "hidden_size": 768,
124
+ "image_size": 224,
125
+ "intermediate_size": 3072,
126
+ "layer_norm_eps": 1e-06,
127
+ "model_type": "siglip_vision_model",
128
+ "num_attention_heads": 12,
129
+ "num_channels": 3,
130
+ "num_hidden_layers": 12,
131
+ "patch_size": 16,
132
+ "torch_dtype": "float32"
133
+ }
134
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a00e319522ea985fd7b18a3c0064a602bb96a2a7558ad4dcf9075bca072ce2df
3
+ size 371703344
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:d732915852aa8238abe1f128b3231ab9d60325b25b8e62363150b9e372e24184
3
+ size 5304