Hugging Face's logo

prithivMLmods
/
ImageShield-SUPER-90M

Image Classification
Transformers
Safetensors
English
siglip
SigLIP2
ImageShield
90M
Guardrail
Model card Files
xet
 Community
1
prithivMLmods commited on
Commit
ee806e6
·
verified ·
1 Parent(s): 055ff88

Add files using upload-large-folder tool

Browse files
Files changed (44) hide show
  1. .gitattributes +1 -0
  2. assets/class_distribution_pie.png +0 -0
  3. assets/classification_report_bar.png +0 -0
  4. assets/confusion_matrix.png +0 -0
  5. assets/misalignment_distribution_pie.png +0 -0
  6. assets/prediction_accuracy_pie.png +0 -0
  7. assets/training_eval_graph.png +3 -0
  8. checkpoint-1500/config.json +50 -0
  9. checkpoint-1500/model.safetensors +3 -0
  10. checkpoint-1500/optimizer.pt +3 -0
  11. checkpoint-1500/preprocessor_config.json +23 -0
  12. checkpoint-1500/rng_state.pth +3 -0
  13. checkpoint-1500/scheduler.pt +3 -0
  14. checkpoint-1500/trainer_state.json +65 -0
  15. checkpoint-1500/training_args.bin +3 -0
  16. checkpoint-3000/config.json +50 -0
  17. checkpoint-3000/model.safetensors +3 -0
  18. checkpoint-3000/optimizer.pt +3 -0
  19. checkpoint-3000/preprocessor_config.json +23 -0
  20. checkpoint-3000/rng_state.pth +3 -0
  21. checkpoint-3000/scheduler.pt +3 -0
  22. checkpoint-3000/trainer_state.json +96 -0
  23. checkpoint-3000/training_args.bin +3 -0
  24. checkpoint-4500/config.json +50 -0
  25. checkpoint-4500/model.safetensors +3 -0
  26. checkpoint-4500/optimizer.pt +3 -0
  27. checkpoint-4500/preprocessor_config.json +23 -0
  28. checkpoint-4500/rng_state.pth +3 -0
  29. checkpoint-4500/scheduler.pt +3 -0
  30. checkpoint-4500/trainer_state.json +127 -0
  31. checkpoint-4500/training_args.bin +3 -0
  32. checkpoint-6000/config.json +50 -0
  33. checkpoint-6000/model.safetensors +3 -0
  34. checkpoint-6000/optimizer.pt +3 -0
  35. checkpoint-6000/preprocessor_config.json +23 -0
  36. checkpoint-6000/rng_state.pth +3 -0
  37. checkpoint-6000/scheduler.pt +3 -0
  38. checkpoint-6000/trainer_state.json +158 -0
  39. checkpoint-6000/training_args.bin +3 -0
  40. config.json +50 -0
  41. model.safetensors +3 -0
  42. preprocessor_config.json +23 -0
  43. results/report.txt +11 -0
  44. training_args.bin +3 -0
.gitattributes CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ assets/training_eval_graph.png filter=lfs diff=lfs merge=lfs -text
assets/class_distribution_pie.png ADDED
assets/classification_report_bar.png ADDED
assets/confusion_matrix.png ADDED
assets/misalignment_distribution_pie.png ADDED
assets/prediction_accuracy_pie.png ADDED
assets/training_eval_graph.png ADDED

Git LFS Details

  • SHA256: 632bf795814638de7a8aa0064880ec9608c2619ee15f2e7418e9066846b16ffd
  • Pointer size: 131 Bytes
  • Size of remote file: 108 kB
checkpoint-1500/config.json ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "SiglipForImageClassification"
4
+ ],
5
+ "dtype": "float32",
6
+ "id2label": {
7
+ "0": "Safe",
8
+ "1": "Unsafe"
9
+ },
10
+ "initializer_factor": 1.0,
11
+ "label2id": {
12
+ "Safe": 0,
13
+ "Unsafe": 1
14
+ },
15
+ "model_type": "siglip",
16
+ "problem_type": "single_label_classification",
17
+ "text_config": {
18
+ "attention_dropout": 0.0,
19
+ "bos_token_id": 49406,
20
+ "dtype": "float32",
21
+ "eos_token_id": 49407,
22
+ "hidden_act": "gelu_pytorch_tanh",
23
+ "hidden_size": 768,
24
+ "intermediate_size": 3072,
25
+ "layer_norm_eps": 1e-06,
26
+ "max_position_embeddings": 64,
27
+ "model_type": "siglip_text_model",
28
+ "num_attention_heads": 12,
29
+ "num_hidden_layers": 12,
30
+ "pad_token_id": 1,
31
+ "projection_size": 768,
32
+ "vocab_size": 256000
33
+ },
34
+ "transformers_version": "5.15.0",
35
+ "use_cache": false,
36
+ "vision_config": {
37
+ "attention_dropout": 0.0,
38
+ "dtype": "float32",
39
+ "hidden_act": "gelu_pytorch_tanh",
40
+ "hidden_size": 768,
41
+ "image_size": 224,
42
+ "intermediate_size": 3072,
43
+ "layer_norm_eps": 1e-06,
44
+ "model_type": "siglip_vision_model",
45
+ "num_attention_heads": 12,
46
+ "num_channels": 3,
47
+ "num_hidden_layers": 12,
48
+ "patch_size": 16
49
+ }
50
+ }
checkpoint-1500/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1a4bf26b53cbcff0b6f4a3bec770a10efabbaa2423c2cdac63b425073c5d90e9
3
+ size 371567992
checkpoint-1500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:596e83388c7003d528a9bb0536d31f7cd9e15b6a10522598513ab8b4941f323f
3
+ size 686558987
checkpoint-1500/preprocessor_config.json ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ "resample": 2,
18
+ "rescale_factor": 0.00392156862745098,
19
+ "size": {
20
+ "height": 224,
21
+ "width": 224
22
+ }
23
+ }
checkpoint-1500/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1fec776a01b7ecf149a6b1fa67bccfb1d6b77988fc3a05eb7372052d21fffb59
3
+ size 14645
checkpoint-1500/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7f80ce5b2896740197905819b11f7b8511b8342b145922d0782dfccc2e7d6ffb
3
+ size 1465
checkpoint-1500/trainer_state.json ADDED
@@ -0,0 +1,65 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": 1500,
3
+ "best_metric": 0.37709423899650574,
4
+ "best_model_checkpoint": "siglip2-image-classification/checkpoint-1500",
5
+ "epoch": 1.0,
6
+ "eval_steps": 500,
7
+ "global_step": 1500,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.3333333333333333,
14
+ "grad_norm": 3.479196786880493,
15
+ "learning_rate": 0.00018490756302521008,
16
+ "loss": 0.5047473754882813,
17
+ "step": 500
18
+ },
19
+ {
20
+ "epoch": 0.6666666666666666,
21
+ "grad_norm": 4.285048007965088,
22
+ "learning_rate": 0.00016810084033613447,
23
+ "loss": 0.42015097045898436,
24
+ "step": 1000
25
+ },
26
+ {
27
+ "epoch": 1.0,
28
+ "grad_norm": 2.655521869659424,
29
+ "learning_rate": 0.00015129411764705882,
30
+ "loss": 0.38898434448242186,
31
+ "step": 1500
32
+ },
33
+ {
34
+ "epoch": 1.0,
35
+ "eval_accuracy": 0.8312134754211069,
36
+ "eval_loss": 0.37709423899650574,
37
+ "eval_model_preparation_time": 0.0019,
38
+ "eval_runtime": 283.3202,
39
+ "eval_samples_per_second": 112.943,
40
+ "eval_steps_per_second": 14.118,
41
+ "step": 1500
42
+ }
43
+ ],
44
+ "logging_steps": 500,
45
+ "max_steps": 6000,
46
+ "num_input_tokens_seen": 0,
47
+ "num_train_epochs": 4,
48
+ "save_steps": 500,
49
+ "stateful_callbacks": {
50
+ "TrainerControl": {
51
+ "args": {
52
+ "should_epoch_stop": false,
53
+ "should_evaluate": false,
54
+ "should_log": false,
55
+ "should_save": true,
56
+ "should_training_stop": false
57
+ },
58
+ "attributes": {}
59
+ }
60
+ },
61
+ "total_flos": 4.0200962884313334e+18,
62
+ "train_batch_size": 32,
63
+ "trial_name": null,
64
+ "trial_params": null
65
+ }
checkpoint-1500/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe9e466d18e2bd615281c80583abd8151d5c8bc84623dad56b5e15f0d4c899ee
3
+ size 5201
checkpoint-3000/config.json ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "SiglipForImageClassification"
4
+ ],
5
+ "dtype": "float32",
6
+ "id2label": {
7
+ "0": "Safe",
8
+ "1": "Unsafe"
9
+ },
10
+ "initializer_factor": 1.0,
11
+ "label2id": {
12
+ "Safe": 0,
13
+ "Unsafe": 1
14
+ },
15
+ "model_type": "siglip",
16
+ "problem_type": "single_label_classification",
17
+ "text_config": {
18
+ "attention_dropout": 0.0,
19
+ "bos_token_id": 49406,
20
+ "dtype": "float32",
21
+ "eos_token_id": 49407,
22
+ "hidden_act": "gelu_pytorch_tanh",
23
+ "hidden_size": 768,
24
+ "intermediate_size": 3072,
25
+ "layer_norm_eps": 1e-06,
26
+ "max_position_embeddings": 64,
27
+ "model_type": "siglip_text_model",
28
+ "num_attention_heads": 12,
29
+ "num_hidden_layers": 12,
30
+ "pad_token_id": 1,
31
+ "projection_size": 768,
32
+ "vocab_size": 256000
33
+ },
34
+ "transformers_version": "5.15.0",
35
+ "use_cache": false,
36
+ "vision_config": {
37
+ "attention_dropout": 0.0,
38
+ "dtype": "float32",
39
+ "hidden_act": "gelu_pytorch_tanh",
40
+ "hidden_size": 768,
41
+ "image_size": 224,
42
+ "intermediate_size": 3072,
43
+ "layer_norm_eps": 1e-06,
44
+ "model_type": "siglip_vision_model",
45
+ "num_attention_heads": 12,
46
+ "num_channels": 3,
47
+ "num_hidden_layers": 12,
48
+ "patch_size": 16
49
+ }
50
+ }
checkpoint-3000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:84c3d8ec730897768c694c4c0cdc5854cdb146532987a0dde4d6ec503a37565a
3
+ size 371567992
checkpoint-3000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ac81a435b16596ccca147f836058a98f63cb5fbdcf6ae54287d348b317d99e67
3
+ size 686558987
checkpoint-3000/preprocessor_config.json ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ "resample": 2,
18
+ "rescale_factor": 0.00392156862745098,
19
+ "size": {
20
+ "height": 224,
21
+ "width": 224
22
+ }
23
+ }
checkpoint-3000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6fc2d9ff41fb99781f8df34958e6a2b2f9e812af596cecf3dbc71061de56ddd7
3
+ size 14645
checkpoint-3000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:22571f75ee2018c3049a09458bbe6c6c1231f3f666682ee6fdc5481ec5a4aacb
3
+ size 1465
checkpoint-3000/trainer_state.json ADDED
@@ -0,0 +1,96 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": 3000,
3
+ "best_metric": 0.3550371527671814,
4
+ "best_model_checkpoint": "siglip2-image-classification/checkpoint-3000",
5
+ "epoch": 2.0,
6
+ "eval_steps": 500,
7
+ "global_step": 3000,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.3333333333333333,
14
+ "grad_norm": 3.479196786880493,
15
+ "learning_rate": 0.00018490756302521008,
16
+ "loss": 0.5047473754882813,
17
+ "step": 500
18
+ },
19
+ {
20
+ "epoch": 0.6666666666666666,
21
+ "grad_norm": 4.285048007965088,
22
+ "learning_rate": 0.00016810084033613447,
23
+ "loss": 0.42015097045898436,
24
+ "step": 1000
25
+ },
26
+ {
27
+ "epoch": 1.0,
28
+ "grad_norm": 2.655521869659424,
29
+ "learning_rate": 0.00015129411764705882,
30
+ "loss": 0.38898434448242186,
31
+ "step": 1500
32
+ },
33
+ {
34
+ "epoch": 1.0,
35
+ "eval_accuracy": 0.8312134754211069,
36
+ "eval_loss": 0.37709423899650574,
37
+ "eval_model_preparation_time": 0.0019,
38
+ "eval_runtime": 283.3202,
39
+ "eval_samples_per_second": 112.943,
40
+ "eval_steps_per_second": 14.118,
41
+ "step": 1500
42
+ },
43
+ {
44
+ "epoch": 1.3333333333333333,
45
+ "grad_norm": 1.3074887990951538,
46
+ "learning_rate": 0.0001344873949579832,
47
+ "loss": 0.3693521728515625,
48
+ "step": 2000
49
+ },
50
+ {
51
+ "epoch": 1.6666666666666665,
52
+ "grad_norm": 1.2879441976547241,
53
+ "learning_rate": 0.00011768067226890757,
54
+ "loss": 0.3484758605957031,
55
+ "step": 2500
56
+ },
57
+ {
58
+ "epoch": 2.0,
59
+ "grad_norm": 1.5275274515151978,
60
+ "learning_rate": 0.00010087394957983194,
61
+ "loss": 0.34061029052734376,
62
+ "step": 3000
63
+ },
64
+ {
65
+ "epoch": 2.0,
66
+ "eval_accuracy": 0.8569642801337541,
67
+ "eval_loss": 0.3550371527671814,
68
+ "eval_model_preparation_time": 0.0019,
69
+ "eval_runtime": 282.8579,
70
+ "eval_samples_per_second": 113.127,
71
+ "eval_steps_per_second": 14.141,
72
+ "step": 3000
73
+ }
74
+ ],
75
+ "logging_steps": 500,
76
+ "max_steps": 6000,
77
+ "num_input_tokens_seen": 0,
78
+ "num_train_epochs": 4,
79
+ "save_steps": 500,
80
+ "stateful_callbacks": {
81
+ "TrainerControl": {
82
+ "args": {
83
+ "should_epoch_stop": false,
84
+ "should_evaluate": false,
85
+ "should_log": false,
86
+ "should_save": true,
87
+ "should_training_stop": false
88
+ },
89
+ "attributes": {}
90
+ }
91
+ },
92
+ "total_flos": 8.040192576862667e+18,
93
+ "train_batch_size": 32,
94
+ "trial_name": null,
95
+ "trial_params": null
96
+ }
checkpoint-3000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe9e466d18e2bd615281c80583abd8151d5c8bc84623dad56b5e15f0d4c899ee
3
+ size 5201
checkpoint-4500/config.json ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "SiglipForImageClassification"
4
+ ],
5
+ "dtype": "float32",
6
+ "id2label": {
7
+ "0": "Safe",
8
+ "1": "Unsafe"
9
+ },
10
+ "initializer_factor": 1.0,
11
+ "label2id": {
12
+ "Safe": 0,
13
+ "Unsafe": 1
14
+ },
15
+ "model_type": "siglip",
16
+ "problem_type": "single_label_classification",
17
+ "text_config": {
18
+ "attention_dropout": 0.0,
19
+ "bos_token_id": 49406,
20
+ "dtype": "float32",
21
+ "eos_token_id": 49407,
22
+ "hidden_act": "gelu_pytorch_tanh",
23
+ "hidden_size": 768,
24
+ "intermediate_size": 3072,
25
+ "layer_norm_eps": 1e-06,
26
+ "max_position_embeddings": 64,
27
+ "model_type": "siglip_text_model",
28
+ "num_attention_heads": 12,
29
+ "num_hidden_layers": 12,
30
+ "pad_token_id": 1,
31
+ "projection_size": 768,
32
+ "vocab_size": 256000
33
+ },
34
+ "transformers_version": "5.15.0",
35
+ "use_cache": false,
36
+ "vision_config": {
37
+ "attention_dropout": 0.0,
38
+ "dtype": "float32",
39
+ "hidden_act": "gelu_pytorch_tanh",
40
+ "hidden_size": 768,
41
+ "image_size": 224,
42
+ "intermediate_size": 3072,
43
+ "layer_norm_eps": 1e-06,
44
+ "model_type": "siglip_vision_model",
45
+ "num_attention_heads": 12,
46
+ "num_channels": 3,
47
+ "num_hidden_layers": 12,
48
+ "patch_size": 16
49
+ }
50
+ }
checkpoint-4500/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f08a49c02e63f8ed0272e48f0425cfa98e203356333b19cdf52b77265f081278
3
+ size 371567992
checkpoint-4500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:baf9fd75acfa623d1a667b661965dd586163ebd551a0b6081177457eb1b75307
3
+ size 686558987
checkpoint-4500/preprocessor_config.json ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ "resample": 2,
18
+ "rescale_factor": 0.00392156862745098,
19
+ "size": {
20
+ "height": 224,
21
+ "width": 224
22
+ }
23
+ }
checkpoint-4500/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8a229ed0ae2d399ab1847c127a24a2743b2b12149859e0cb5c90c1d8fa8ece68
3
+ size 14645
checkpoint-4500/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fda5005268f2e2d020512b396c9904946a7fa4a8525e5969867d90876d389218
3
+ size 1465
checkpoint-4500/trainer_state.json ADDED
@@ -0,0 +1,127 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": 4500,
3
+ "best_metric": 0.2799703776836395,
4
+ "best_model_checkpoint": "siglip2-image-classification/checkpoint-4500",
5
+ "epoch": 3.0,
6
+ "eval_steps": 500,
7
+ "global_step": 4500,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.3333333333333333,
14
+ "grad_norm": 3.479196786880493,
15
+ "learning_rate": 0.00018490756302521008,
16
+ "loss": 0.5047473754882813,
17
+ "step": 500
18
+ },
19
+ {
20
+ "epoch": 0.6666666666666666,
21
+ "grad_norm": 4.285048007965088,
22
+ "learning_rate": 0.00016810084033613447,
23
+ "loss": 0.42015097045898436,
24
+ "step": 1000
25
+ },
26
+ {
27
+ "epoch": 1.0,
28
+ "grad_norm": 2.655521869659424,
29
+ "learning_rate": 0.00015129411764705882,
30
+ "loss": 0.38898434448242186,
31
+ "step": 1500
32
+ },
33
+ {
34
+ "epoch": 1.0,
35
+ "eval_accuracy": 0.8312134754211069,
36
+ "eval_loss": 0.37709423899650574,
37
+ "eval_model_preparation_time": 0.0019,
38
+ "eval_runtime": 283.3202,
39
+ "eval_samples_per_second": 112.943,
40
+ "eval_steps_per_second": 14.118,
41
+ "step": 1500
42
+ },
43
+ {
44
+ "epoch": 1.3333333333333333,
45
+ "grad_norm": 1.3074887990951538,
46
+ "learning_rate": 0.0001344873949579832,
47
+ "loss": 0.3693521728515625,
48
+ "step": 2000
49
+ },
50
+ {
51
+ "epoch": 1.6666666666666665,
52
+ "grad_norm": 1.2879441976547241,
53
+ "learning_rate": 0.00011768067226890757,
54
+ "loss": 0.3484758605957031,
55
+ "step": 2500
56
+ },
57
+ {
58
+ "epoch": 2.0,
59
+ "grad_norm": 1.5275274515151978,
60
+ "learning_rate": 0.00010087394957983194,
61
+ "loss": 0.34061029052734376,
62
+ "step": 3000
63
+ },
64
+ {
65
+ "epoch": 2.0,
66
+ "eval_accuracy": 0.8569642801337541,
67
+ "eval_loss": 0.3550371527671814,
68
+ "eval_model_preparation_time": 0.0019,
69
+ "eval_runtime": 282.8579,
70
+ "eval_samples_per_second": 113.127,
71
+ "eval_steps_per_second": 14.141,
72
+ "step": 3000
73
+ },
74
+ {
75
+ "epoch": 2.3333333333333335,
76
+ "grad_norm": 4.5269551277160645,
77
+ "learning_rate": 8.406722689075631e-05,
78
+ "loss": 0.31870977783203125,
79
+ "step": 3500
80
+ },
81
+ {
82
+ "epoch": 2.6666666666666665,
83
+ "grad_norm": 1.9202643632888794,
84
+ "learning_rate": 6.726050420168067e-05,
85
+ "loss": 0.30601947021484377,
86
+ "step": 4000
87
+ },
88
+ {
89
+ "epoch": 3.0,
90
+ "grad_norm": 2.203441858291626,
91
+ "learning_rate": 5.045378151260505e-05,
92
+ "loss": 0.2938027648925781,
93
+ "step": 4500
94
+ },
95
+ {
96
+ "epoch": 3.0,
97
+ "eval_accuracy": 0.8811525360167505,
98
+ "eval_loss": 0.2799703776836395,
99
+ "eval_model_preparation_time": 0.0019,
100
+ "eval_runtime": 287.094,
101
+ "eval_samples_per_second": 111.458,
102
+ "eval_steps_per_second": 13.933,
103
+ "step": 4500
104
+ }
105
+ ],
106
+ "logging_steps": 500,
107
+ "max_steps": 6000,
108
+ "num_input_tokens_seen": 0,
109
+ "num_train_epochs": 4,
110
+ "save_steps": 500,
111
+ "stateful_callbacks": {
112
+ "TrainerControl": {
113
+ "args": {
114
+ "should_epoch_stop": false,
115
+ "should_evaluate": false,
116
+ "should_log": false,
117
+ "should_save": true,
118
+ "should_training_stop": false
119
+ },
120
+ "attributes": {}
121
+ }
122
+ },
123
+ "total_flos": 1.2060288865294e+19,
124
+ "train_batch_size": 32,
125
+ "trial_name": null,
126
+ "trial_params": null
127
+ }
checkpoint-4500/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe9e466d18e2bd615281c80583abd8151d5c8bc84623dad56b5e15f0d4c899ee
3
+ size 5201
checkpoint-6000/config.json ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "SiglipForImageClassification"
4
+ ],
5
+ "dtype": "float32",
6
+ "id2label": {
7
+ "0": "Safe",
8
+ "1": "Unsafe"
9
+ },
10
+ "initializer_factor": 1.0,
11
+ "label2id": {
12
+ "Safe": 0,
13
+ "Unsafe": 1
14
+ },
15
+ "model_type": "siglip",
16
+ "problem_type": "single_label_classification",
17
+ "text_config": {
18
+ "attention_dropout": 0.0,
19
+ "bos_token_id": 49406,
20
+ "dtype": "float32",
21
+ "eos_token_id": 49407,
22
+ "hidden_act": "gelu_pytorch_tanh",
23
+ "hidden_size": 768,
24
+ "intermediate_size": 3072,
25
+ "layer_norm_eps": 1e-06,
26
+ "max_position_embeddings": 64,
27
+ "model_type": "siglip_text_model",
28
+ "num_attention_heads": 12,
29
+ "num_hidden_layers": 12,
30
+ "pad_token_id": 1,
31
+ "projection_size": 768,
32
+ "vocab_size": 256000
33
+ },
34
+ "transformers_version": "5.15.0",
35
+ "use_cache": false,
36
+ "vision_config": {
37
+ "attention_dropout": 0.0,
38
+ "dtype": "float32",
39
+ "hidden_act": "gelu_pytorch_tanh",
40
+ "hidden_size": 768,
41
+ "image_size": 224,
42
+ "intermediate_size": 3072,
43
+ "layer_norm_eps": 1e-06,
44
+ "model_type": "siglip_vision_model",
45
+ "num_attention_heads": 12,
46
+ "num_channels": 3,
47
+ "num_hidden_layers": 12,
48
+ "patch_size": 16
49
+ }
50
+ }
checkpoint-6000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:25339db142dbeb2ed734696f6d099da305318d044f83f45145fc778c2f59bc17
3
+ size 371567992
checkpoint-6000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9d9f44539117d3865deea4a332d2112b86f25ac942a8e27c501e43447ad02f53
3
+ size 686558987
checkpoint-6000/preprocessor_config.json ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ "resample": 2,
18
+ "rescale_factor": 0.00392156862745098,
19
+ "size": {
20
+ "height": 224,
21
+ "width": 224
22
+ }
23
+ }
checkpoint-6000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6325ce64fa1116d0b2384d58a7d2de831adbfb9513edd5786abb3ac15c7867f2
3
+ size 14645
checkpoint-6000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1f0326c04dcdb07f5852da1801a6efbb41ba882da2637d7d10449bb43a8b585a
3
+ size 1465
checkpoint-6000/trainer_state.json ADDED
@@ -0,0 +1,158 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": 6000,
3
+ "best_metric": 0.25916826725006104,
4
+ "best_model_checkpoint": "siglip2-image-classification/checkpoint-6000",
5
+ "epoch": 4.0,
6
+ "eval_steps": 500,
7
+ "global_step": 6000,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.3333333333333333,
14
+ "grad_norm": 3.479196786880493,
15
+ "learning_rate": 0.00018490756302521008,
16
+ "loss": 0.5047473754882813,
17
+ "step": 500
18
+ },
19
+ {
20
+ "epoch": 0.6666666666666666,
21
+ "grad_norm": 4.285048007965088,
22
+ "learning_rate": 0.00016810084033613447,
23
+ "loss": 0.42015097045898436,
24
+ "step": 1000
25
+ },
26
+ {
27
+ "epoch": 1.0,
28
+ "grad_norm": 2.655521869659424,
29
+ "learning_rate": 0.00015129411764705882,
30
+ "loss": 0.38898434448242186,
31
+ "step": 1500
32
+ },
33
+ {
34
+ "epoch": 1.0,
35
+ "eval_accuracy": 0.8312134754211069,
36
+ "eval_loss": 0.37709423899650574,
37
+ "eval_model_preparation_time": 0.0019,
38
+ "eval_runtime": 283.3202,
39
+ "eval_samples_per_second": 112.943,
40
+ "eval_steps_per_second": 14.118,
41
+ "step": 1500
42
+ },
43
+ {
44
+ "epoch": 1.3333333333333333,
45
+ "grad_norm": 1.3074887990951538,
46
+ "learning_rate": 0.0001344873949579832,
47
+ "loss": 0.3693521728515625,
48
+ "step": 2000
49
+ },
50
+ {
51
+ "epoch": 1.6666666666666665,
52
+ "grad_norm": 1.2879441976547241,
53
+ "learning_rate": 0.00011768067226890757,
54
+ "loss": 0.3484758605957031,
55
+ "step": 2500
56
+ },
57
+ {
58
+ "epoch": 2.0,
59
+ "grad_norm": 1.5275274515151978,
60
+ "learning_rate": 0.00010087394957983194,
61
+ "loss": 0.34061029052734376,
62
+ "step": 3000
63
+ },
64
+ {
65
+ "epoch": 2.0,
66
+ "eval_accuracy": 0.8569642801337541,
67
+ "eval_loss": 0.3550371527671814,
68
+ "eval_model_preparation_time": 0.0019,
69
+ "eval_runtime": 282.8579,
70
+ "eval_samples_per_second": 113.127,
71
+ "eval_steps_per_second": 14.141,
72
+ "step": 3000
73
+ },
74
+ {
75
+ "epoch": 2.3333333333333335,
76
+ "grad_norm": 4.5269551277160645,
77
+ "learning_rate": 8.406722689075631e-05,
78
+ "loss": 0.31870977783203125,
79
+ "step": 3500
80
+ },
81
+ {
82
+ "epoch": 2.6666666666666665,
83
+ "grad_norm": 1.9202643632888794,
84
+ "learning_rate": 6.726050420168067e-05,
85
+ "loss": 0.30601947021484377,
86
+ "step": 4000
87
+ },
88
+ {
89
+ "epoch": 3.0,
90
+ "grad_norm": 2.203441858291626,
91
+ "learning_rate": 5.045378151260505e-05,
92
+ "loss": 0.2938027648925781,
93
+ "step": 4500
94
+ },
95
+ {
96
+ "epoch": 3.0,
97
+ "eval_accuracy": 0.8811525360167505,
98
+ "eval_loss": 0.2799703776836395,
99
+ "eval_model_preparation_time": 0.0019,
100
+ "eval_runtime": 287.094,
101
+ "eval_samples_per_second": 111.458,
102
+ "eval_steps_per_second": 13.933,
103
+ "step": 4500
104
+ },
105
+ {
106
+ "epoch": 3.3333333333333335,
107
+ "grad_norm": 3.2569119930267334,
108
+ "learning_rate": 3.364705882352941e-05,
109
+ "loss": 0.2793130798339844,
110
+ "step": 5000
111
+ },
112
+ {
113
+ "epoch": 3.6666666666666665,
114
+ "grad_norm": 3.2401018142700195,
115
+ "learning_rate": 1.6840336134453782e-05,
116
+ "loss": 0.26365960693359375,
117
+ "step": 5500
118
+ },
119
+ {
120
+ "epoch": 4.0,
121
+ "grad_norm": 1.6921696662902832,
122
+ "learning_rate": 3.3613445378151266e-08,
123
+ "loss": 0.2508096923828125,
124
+ "step": 6000
125
+ },
126
+ {
127
+ "epoch": 4.0,
128
+ "eval_accuracy": 0.8935591737241789,
129
+ "eval_loss": 0.25916826725006104,
130
+ "eval_model_preparation_time": 0.0019,
131
+ "eval_runtime": 286.6404,
132
+ "eval_samples_per_second": 111.635,
133
+ "eval_steps_per_second": 13.955,
134
+ "step": 6000
135
+ }
136
+ ],
137
+ "logging_steps": 500,
138
+ "max_steps": 6000,
139
+ "num_input_tokens_seen": 0,
140
+ "num_train_epochs": 4,
141
+ "save_steps": 500,
142
+ "stateful_callbacks": {
143
+ "TrainerControl": {
144
+ "args": {
145
+ "should_epoch_stop": false,
146
+ "should_evaluate": false,
147
+ "should_log": false,
148
+ "should_save": true,
149
+ "should_training_stop": true
150
+ },
151
+ "attributes": {}
152
+ }
153
+ },
154
+ "total_flos": 1.6080385153725334e+19,
155
+ "train_batch_size": 32,
156
+ "trial_name": null,
157
+ "trial_params": null
158
+ }
checkpoint-6000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe9e466d18e2bd615281c80583abd8151d5c8bc84623dad56b5e15f0d4c899ee
3
+ size 5201
config.json ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "SiglipForImageClassification"
4
+ ],
5
+ "dtype": "float32",
6
+ "id2label": {
7
+ "0": "Safe",
8
+ "1": "Unsafe"
9
+ },
10
+ "initializer_factor": 1.0,
11
+ "label2id": {
12
+ "Safe": 0,
13
+ "Unsafe": 1
14
+ },
15
+ "model_type": "siglip",
16
+ "problem_type": "single_label_classification",
17
+ "text_config": {
18
+ "attention_dropout": 0.0,
19
+ "bos_token_id": 49406,
20
+ "dtype": "float32",
21
+ "eos_token_id": 49407,
22
+ "hidden_act": "gelu_pytorch_tanh",
23
+ "hidden_size": 768,
24
+ "intermediate_size": 3072,
25
+ "layer_norm_eps": 1e-06,
26
+ "max_position_embeddings": 64,
27
+ "model_type": "siglip_text_model",
28
+ "num_attention_heads": 12,
29
+ "num_hidden_layers": 12,
30
+ "pad_token_id": 1,
31
+ "projection_size": 768,
32
+ "vocab_size": 256000
33
+ },
34
+ "transformers_version": "5.15.0",
35
+ "use_cache": false,
36
+ "vision_config": {
37
+ "attention_dropout": 0.0,
38
+ "dtype": "float32",
39
+ "hidden_act": "gelu_pytorch_tanh",
40
+ "hidden_size": 768,
41
+ "image_size": 224,
42
+ "intermediate_size": 3072,
43
+ "layer_norm_eps": 1e-06,
44
+ "model_type": "siglip_vision_model",
45
+ "num_attention_heads": 12,
46
+ "num_channels": 3,
47
+ "num_hidden_layers": 12,
48
+ "patch_size": 16
49
+ }
50
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:25339db142dbeb2ed734696f6d099da305318d044f83f45145fc778c2f59bc17
3
+ size 371567992
preprocessor_config.json ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ "resample": 2,
18
+ "rescale_factor": 0.00392156862745098,
19
+ "size": {
20
+ "height": 224,
21
+ "width": 224
22
+ }
23
+ }
results/report.txt ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Classification Report
2
+ ============================================================
3
+
4
+ precision recall f1-score support
5
+
6
+ Safe 0.9128 0.8702 0.8910 16000
7
+ Unsafe 0.8760 0.9169 0.8960 15999
8
+
9
+ accuracy 0.8936 31999
10
+ macro avg 0.8944 0.8936 0.8935 31999
11
+ weighted avg 0.8944 0.8936 0.8935 31999
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe9e466d18e2bd615281c80583abd8151d5c8bc84623dad56b5e15f0d4c899ee
3
+ size 5201