JashVora7 commited on
Commit
a04291b
·
verified ·
1 Parent(s): f6a3457

Upload hybrid-guardrails checkpoint (seed 42) with model card

Browse files
README.md ADDED
@@ -0,0 +1,72 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: mit
3
+ base_model: distilbert-base-uncased
4
+ tags:
5
+ - text-classification
6
+ - content-moderation
7
+ - safety
8
+ - llm-guardrails
9
+ pipeline_tag: text-classification
10
+ ---
11
+
12
+ # hybrid-guardrails-distilbert-moderation
13
+
14
+ A DistilBERT binary safe/unsafe content classifier, fine-tuned as a lightweight
15
+ neural baseline in the **hybrid-guardrails** neuro-symbolic moderation pipeline
16
+ project. Evaluated standalone as the "Distilbert Only" baseline.
17
+
18
+ ## Architecture
19
+
20
+ `DistilBertForSequenceClassification`, fine-tuned from `distilbert-base-uncased`
21
+ (6 layers, hidden size 768), 2-way classification (`safe` / `unsafe`).
22
+
23
+ ## Training data
24
+
25
+ Same combined corpus as the project's DeBERTa checkpoint (see
26
+ `JashVora7/hybrid-guardrails-deberta-moderation`): AdvBench, ToxiGen, and
27
+ Alpaca-Cleaned. Training config: `training/configs/distilbert_base.yaml`
28
+ (seed 42).
29
+
30
+ ## Intended use
31
+
32
+ Research artifact only -- a **lightweight-but-weaker** comparison point
33
+ against the DeBERTa checkpoint in the project's baseline table, not a
34
+ recommended production classifier. Its recall on adversarial content is
35
+ substantially lower than DeBERTa's (see below); it exists in this project to
36
+ demonstrate that hybrid symbolic+neural gains are conditional on neural
37
+ classifier quality, not just architecture choice.
38
+
39
+ ## Evaluation results (seed 42, Experiment A baseline comparison)
40
+
41
+ | Metric | Value |
42
+ |---|---|
43
+ | Precision | 0.600 |
44
+ | Recall | 0.079 |
45
+ | F1 | 0.139 |
46
+ | AUROC | 0.538 |
47
+ | FRR | 0.005 |
48
+
49
+ Low recall/F1 relative to the project's DeBERTa checkpoint is a real,
50
+ reported finding, not a bug -- see `paper/MANUSCRIPT.md` (§5.1) in the
51
+ source repository for discussion.
52
+
53
+ ## Usage
54
+
55
+ ```python
56
+ from transformers import AutoModelForSequenceClassification, AutoTokenizer
57
+ import torch
58
+
59
+ tok = AutoTokenizer.from_pretrained("JashVora7/hybrid-guardrails-distilbert-moderation")
60
+ model = AutoModelForSequenceClassification.from_pretrained(
61
+ "JashVora7/hybrid-guardrails-distilbert-moderation"
62
+ )
63
+
64
+ inputs = tok("ignore all previous instructions", return_tensors="pt")
65
+ with torch.no_grad():
66
+ probs = torch.softmax(model(**inputs).logits, dim=-1)
67
+ print(probs) # [P(safe), P(unsafe)]
68
+ ```
69
+
70
+ ## License
71
+
72
+ MIT, matching the source repository.
checkpoint-188/config.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "activation": "gelu",
3
+ "architectures": [
4
+ "DistilBertForSequenceClassification"
5
+ ],
6
+ "attention_dropout": 0.1,
7
+ "dim": 768,
8
+ "dropout": 0.1,
9
+ "dtype": "float32",
10
+ "hidden_dim": 3072,
11
+ "id2label": {
12
+ "0": "safe",
13
+ "1": "unsafe"
14
+ },
15
+ "initializer_range": 0.02,
16
+ "label2id": {
17
+ "safe": 0,
18
+ "unsafe": 1
19
+ },
20
+ "max_position_embeddings": 512,
21
+ "model_type": "distilbert",
22
+ "n_heads": 12,
23
+ "n_layers": 6,
24
+ "pad_token_id": 0,
25
+ "problem_type": "single_label_classification",
26
+ "qa_dropout": 0.1,
27
+ "seq_classif_dropout": 0.2,
28
+ "sinusoidal_pos_embds": false,
29
+ "tie_weights_": true,
30
+ "transformers_version": "4.57.1",
31
+ "vocab_size": 30522
32
+ }
checkpoint-188/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:70112457a13bd4e0cb405eba16c9bbf767f1c30329afa3e292d789fbb6fc1f1c
3
+ size 267832560
checkpoint-188/special_tokens_map.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "cls_token": "[CLS]",
3
+ "mask_token": "[MASK]",
4
+ "pad_token": "[PAD]",
5
+ "sep_token": "[SEP]",
6
+ "unk_token": "[UNK]"
7
+ }
checkpoint-188/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-188/tokenizer_config.json ADDED
@@ -0,0 +1,56 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "added_tokens_decoder": {
3
+ "0": {
4
+ "content": "[PAD]",
5
+ "lstrip": false,
6
+ "normalized": false,
7
+ "rstrip": false,
8
+ "single_word": false,
9
+ "special": true
10
+ },
11
+ "100": {
12
+ "content": "[UNK]",
13
+ "lstrip": false,
14
+ "normalized": false,
15
+ "rstrip": false,
16
+ "single_word": false,
17
+ "special": true
18
+ },
19
+ "101": {
20
+ "content": "[CLS]",
21
+ "lstrip": false,
22
+ "normalized": false,
23
+ "rstrip": false,
24
+ "single_word": false,
25
+ "special": true
26
+ },
27
+ "102": {
28
+ "content": "[SEP]",
29
+ "lstrip": false,
30
+ "normalized": false,
31
+ "rstrip": false,
32
+ "single_word": false,
33
+ "special": true
34
+ },
35
+ "103": {
36
+ "content": "[MASK]",
37
+ "lstrip": false,
38
+ "normalized": false,
39
+ "rstrip": false,
40
+ "single_word": false,
41
+ "special": true
42
+ }
43
+ },
44
+ "clean_up_tokenization_spaces": false,
45
+ "cls_token": "[CLS]",
46
+ "do_lower_case": true,
47
+ "extra_special_tokens": {},
48
+ "mask_token": "[MASK]",
49
+ "model_max_length": 512,
50
+ "pad_token": "[PAD]",
51
+ "sep_token": "[SEP]",
52
+ "strip_accents": null,
53
+ "tokenize_chinese_chars": true,
54
+ "tokenizer_class": "DistilBertTokenizer",
55
+ "unk_token": "[UNK]"
56
+ }
checkpoint-188/trainer_state.json ADDED
@@ -0,0 +1,172 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": 188,
3
+ "best_metric": 0.8062015503875969,
4
+ "best_model_checkpoint": "models/distilbert_moderation_finetuned\\checkpoint-188",
5
+ "epoch": 1.0,
6
+ "eval_steps": 500,
7
+ "global_step": 188,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.05319148936170213,
14
+ "grad_norm": 2.128387689590454,
15
+ "learning_rate": 3.5526315789473683e-06,
16
+ "loss": 0.6783,
17
+ "step": 10
18
+ },
19
+ {
20
+ "epoch": 0.10638297872340426,
21
+ "grad_norm": 2.286456823348999,
22
+ "learning_rate": 7.5e-06,
23
+ "loss": 0.6585,
24
+ "step": 20
25
+ },
26
+ {
27
+ "epoch": 0.1595744680851064,
28
+ "grad_norm": 1.3711037635803223,
29
+ "learning_rate": 1.1447368421052632e-05,
30
+ "loss": 0.6436,
31
+ "step": 30
32
+ },
33
+ {
34
+ "epoch": 0.2127659574468085,
35
+ "grad_norm": 1.3716976642608643,
36
+ "learning_rate": 1.5394736842105264e-05,
37
+ "loss": 0.6011,
38
+ "step": 40
39
+ },
40
+ {
41
+ "epoch": 0.26595744680851063,
42
+ "grad_norm": 1.3208516836166382,
43
+ "learning_rate": 1.9342105263157896e-05,
44
+ "loss": 0.5312,
45
+ "step": 50
46
+ },
47
+ {
48
+ "epoch": 0.3191489361702128,
49
+ "grad_norm": 1.8064684867858887,
50
+ "learning_rate": 2.3289473684210525e-05,
51
+ "loss": 0.4832,
52
+ "step": 60
53
+ },
54
+ {
55
+ "epoch": 0.3723404255319149,
56
+ "grad_norm": 3.465531826019287,
57
+ "learning_rate": 2.723684210526316e-05,
58
+ "loss": 0.4031,
59
+ "step": 70
60
+ },
61
+ {
62
+ "epoch": 0.425531914893617,
63
+ "grad_norm": 6.162868499755859,
64
+ "learning_rate": 2.9866863905325445e-05,
65
+ "loss": 0.3736,
66
+ "step": 80
67
+ },
68
+ {
69
+ "epoch": 0.4787234042553192,
70
+ "grad_norm": 5.821077346801758,
71
+ "learning_rate": 2.9423076923076923e-05,
72
+ "loss": 0.3381,
73
+ "step": 90
74
+ },
75
+ {
76
+ "epoch": 0.5319148936170213,
77
+ "grad_norm": 3.1542975902557373,
78
+ "learning_rate": 2.8979289940828404e-05,
79
+ "loss": 0.332,
80
+ "step": 100
81
+ },
82
+ {
83
+ "epoch": 0.5851063829787234,
84
+ "grad_norm": 3.219808578491211,
85
+ "learning_rate": 2.8535502958579882e-05,
86
+ "loss": 0.3636,
87
+ "step": 110
88
+ },
89
+ {
90
+ "epoch": 0.6382978723404256,
91
+ "grad_norm": 3.52590012550354,
92
+ "learning_rate": 2.809171597633136e-05,
93
+ "loss": 0.3288,
94
+ "step": 120
95
+ },
96
+ {
97
+ "epoch": 0.6914893617021277,
98
+ "grad_norm": 2.8622491359710693,
99
+ "learning_rate": 2.764792899408284e-05,
100
+ "loss": 0.3061,
101
+ "step": 130
102
+ },
103
+ {
104
+ "epoch": 0.7446808510638298,
105
+ "grad_norm": 2.1935667991638184,
106
+ "learning_rate": 2.7204142011834322e-05,
107
+ "loss": 0.3211,
108
+ "step": 140
109
+ },
110
+ {
111
+ "epoch": 0.7978723404255319,
112
+ "grad_norm": 3.4211273193359375,
113
+ "learning_rate": 2.67603550295858e-05,
114
+ "loss": 0.2889,
115
+ "step": 150
116
+ },
117
+ {
118
+ "epoch": 0.851063829787234,
119
+ "grad_norm": 3.5952651500701904,
120
+ "learning_rate": 2.6316568047337277e-05,
121
+ "loss": 0.3066,
122
+ "step": 160
123
+ },
124
+ {
125
+ "epoch": 0.9042553191489362,
126
+ "grad_norm": 4.181521892547607,
127
+ "learning_rate": 2.5872781065088758e-05,
128
+ "loss": 0.3504,
129
+ "step": 170
130
+ },
131
+ {
132
+ "epoch": 0.9574468085106383,
133
+ "grad_norm": 3.022378921508789,
134
+ "learning_rate": 2.542899408284024e-05,
135
+ "loss": 0.2445,
136
+ "step": 180
137
+ },
138
+ {
139
+ "epoch": 1.0,
140
+ "eval_accuracy": 0.8666666666666667,
141
+ "eval_f1": 0.8062015503875969,
142
+ "eval_loss": 0.2918033301830292,
143
+ "eval_precision": 0.7563636363636363,
144
+ "eval_recall": 0.8630705394190872,
145
+ "eval_runtime": 13.7915,
146
+ "eval_samples_per_second": 54.381,
147
+ "eval_steps_per_second": 0.87,
148
+ "step": 188
149
+ }
150
+ ],
151
+ "logging_steps": 10,
152
+ "max_steps": 752,
153
+ "num_input_tokens_seen": 0,
154
+ "num_train_epochs": 4,
155
+ "save_steps": 500,
156
+ "stateful_callbacks": {
157
+ "TrainerControl": {
158
+ "args": {
159
+ "should_epoch_stop": false,
160
+ "should_evaluate": false,
161
+ "should_log": false,
162
+ "should_save": true,
163
+ "should_training_stop": false
164
+ },
165
+ "attributes": {}
166
+ }
167
+ },
168
+ "total_flos": 70761393618000.0,
169
+ "train_batch_size": 32,
170
+ "trial_name": null,
171
+ "trial_params": null
172
+ }
checkpoint-188/vocab.txt ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-376/config.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "activation": "gelu",
3
+ "architectures": [
4
+ "DistilBertForSequenceClassification"
5
+ ],
6
+ "attention_dropout": 0.1,
7
+ "dim": 768,
8
+ "dropout": 0.1,
9
+ "dtype": "float32",
10
+ "hidden_dim": 3072,
11
+ "id2label": {
12
+ "0": "safe",
13
+ "1": "unsafe"
14
+ },
15
+ "initializer_range": 0.02,
16
+ "label2id": {
17
+ "safe": 0,
18
+ "unsafe": 1
19
+ },
20
+ "max_position_embeddings": 512,
21
+ "model_type": "distilbert",
22
+ "n_heads": 12,
23
+ "n_layers": 6,
24
+ "pad_token_id": 0,
25
+ "problem_type": "single_label_classification",
26
+ "qa_dropout": 0.1,
27
+ "seq_classif_dropout": 0.2,
28
+ "sinusoidal_pos_embds": false,
29
+ "tie_weights_": true,
30
+ "transformers_version": "4.57.1",
31
+ "vocab_size": 30522
32
+ }
checkpoint-376/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:989c5ef6b333f37de263aa46ba6c63cb758ca2dd84bcd69165f6746e61bcfe2f
3
+ size 267832560
checkpoint-376/special_tokens_map.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "cls_token": "[CLS]",
3
+ "mask_token": "[MASK]",
4
+ "pad_token": "[PAD]",
5
+ "sep_token": "[SEP]",
6
+ "unk_token": "[UNK]"
7
+ }
checkpoint-376/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-376/tokenizer_config.json ADDED
@@ -0,0 +1,56 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "added_tokens_decoder": {
3
+ "0": {
4
+ "content": "[PAD]",
5
+ "lstrip": false,
6
+ "normalized": false,
7
+ "rstrip": false,
8
+ "single_word": false,
9
+ "special": true
10
+ },
11
+ "100": {
12
+ "content": "[UNK]",
13
+ "lstrip": false,
14
+ "normalized": false,
15
+ "rstrip": false,
16
+ "single_word": false,
17
+ "special": true
18
+ },
19
+ "101": {
20
+ "content": "[CLS]",
21
+ "lstrip": false,
22
+ "normalized": false,
23
+ "rstrip": false,
24
+ "single_word": false,
25
+ "special": true
26
+ },
27
+ "102": {
28
+ "content": "[SEP]",
29
+ "lstrip": false,
30
+ "normalized": false,
31
+ "rstrip": false,
32
+ "single_word": false,
33
+ "special": true
34
+ },
35
+ "103": {
36
+ "content": "[MASK]",
37
+ "lstrip": false,
38
+ "normalized": false,
39
+ "rstrip": false,
40
+ "single_word": false,
41
+ "special": true
42
+ }
43
+ },
44
+ "clean_up_tokenization_spaces": false,
45
+ "cls_token": "[CLS]",
46
+ "do_lower_case": true,
47
+ "extra_special_tokens": {},
48
+ "mask_token": "[MASK]",
49
+ "model_max_length": 512,
50
+ "pad_token": "[PAD]",
51
+ "sep_token": "[SEP]",
52
+ "strip_accents": null,
53
+ "tokenize_chinese_chars": true,
54
+ "tokenizer_class": "DistilBertTokenizer",
55
+ "unk_token": "[UNK]"
56
+ }
checkpoint-376/trainer_state.json ADDED
@@ -0,0 +1,317 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": 188,
3
+ "best_metric": 0.8062015503875969,
4
+ "best_model_checkpoint": "models/distilbert_moderation_finetuned\\checkpoint-188",
5
+ "epoch": 2.0,
6
+ "eval_steps": 500,
7
+ "global_step": 376,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.05319148936170213,
14
+ "grad_norm": 2.128387689590454,
15
+ "learning_rate": 3.5526315789473683e-06,
16
+ "loss": 0.6783,
17
+ "step": 10
18
+ },
19
+ {
20
+ "epoch": 0.10638297872340426,
21
+ "grad_norm": 2.286456823348999,
22
+ "learning_rate": 7.5e-06,
23
+ "loss": 0.6585,
24
+ "step": 20
25
+ },
26
+ {
27
+ "epoch": 0.1595744680851064,
28
+ "grad_norm": 1.3711037635803223,
29
+ "learning_rate": 1.1447368421052632e-05,
30
+ "loss": 0.6436,
31
+ "step": 30
32
+ },
33
+ {
34
+ "epoch": 0.2127659574468085,
35
+ "grad_norm": 1.3716976642608643,
36
+ "learning_rate": 1.5394736842105264e-05,
37
+ "loss": 0.6011,
38
+ "step": 40
39
+ },
40
+ {
41
+ "epoch": 0.26595744680851063,
42
+ "grad_norm": 1.3208516836166382,
43
+ "learning_rate": 1.9342105263157896e-05,
44
+ "loss": 0.5312,
45
+ "step": 50
46
+ },
47
+ {
48
+ "epoch": 0.3191489361702128,
49
+ "grad_norm": 1.8064684867858887,
50
+ "learning_rate": 2.3289473684210525e-05,
51
+ "loss": 0.4832,
52
+ "step": 60
53
+ },
54
+ {
55
+ "epoch": 0.3723404255319149,
56
+ "grad_norm": 3.465531826019287,
57
+ "learning_rate": 2.723684210526316e-05,
58
+ "loss": 0.4031,
59
+ "step": 70
60
+ },
61
+ {
62
+ "epoch": 0.425531914893617,
63
+ "grad_norm": 6.162868499755859,
64
+ "learning_rate": 2.9866863905325445e-05,
65
+ "loss": 0.3736,
66
+ "step": 80
67
+ },
68
+ {
69
+ "epoch": 0.4787234042553192,
70
+ "grad_norm": 5.821077346801758,
71
+ "learning_rate": 2.9423076923076923e-05,
72
+ "loss": 0.3381,
73
+ "step": 90
74
+ },
75
+ {
76
+ "epoch": 0.5319148936170213,
77
+ "grad_norm": 3.1542975902557373,
78
+ "learning_rate": 2.8979289940828404e-05,
79
+ "loss": 0.332,
80
+ "step": 100
81
+ },
82
+ {
83
+ "epoch": 0.5851063829787234,
84
+ "grad_norm": 3.219808578491211,
85
+ "learning_rate": 2.8535502958579882e-05,
86
+ "loss": 0.3636,
87
+ "step": 110
88
+ },
89
+ {
90
+ "epoch": 0.6382978723404256,
91
+ "grad_norm": 3.52590012550354,
92
+ "learning_rate": 2.809171597633136e-05,
93
+ "loss": 0.3288,
94
+ "step": 120
95
+ },
96
+ {
97
+ "epoch": 0.6914893617021277,
98
+ "grad_norm": 2.8622491359710693,
99
+ "learning_rate": 2.764792899408284e-05,
100
+ "loss": 0.3061,
101
+ "step": 130
102
+ },
103
+ {
104
+ "epoch": 0.7446808510638298,
105
+ "grad_norm": 2.1935667991638184,
106
+ "learning_rate": 2.7204142011834322e-05,
107
+ "loss": 0.3211,
108
+ "step": 140
109
+ },
110
+ {
111
+ "epoch": 0.7978723404255319,
112
+ "grad_norm": 3.4211273193359375,
113
+ "learning_rate": 2.67603550295858e-05,
114
+ "loss": 0.2889,
115
+ "step": 150
116
+ },
117
+ {
118
+ "epoch": 0.851063829787234,
119
+ "grad_norm": 3.5952651500701904,
120
+ "learning_rate": 2.6316568047337277e-05,
121
+ "loss": 0.3066,
122
+ "step": 160
123
+ },
124
+ {
125
+ "epoch": 0.9042553191489362,
126
+ "grad_norm": 4.181521892547607,
127
+ "learning_rate": 2.5872781065088758e-05,
128
+ "loss": 0.3504,
129
+ "step": 170
130
+ },
131
+ {
132
+ "epoch": 0.9574468085106383,
133
+ "grad_norm": 3.022378921508789,
134
+ "learning_rate": 2.542899408284024e-05,
135
+ "loss": 0.2445,
136
+ "step": 180
137
+ },
138
+ {
139
+ "epoch": 1.0,
140
+ "eval_accuracy": 0.8666666666666667,
141
+ "eval_f1": 0.8062015503875969,
142
+ "eval_loss": 0.2918033301830292,
143
+ "eval_precision": 0.7563636363636363,
144
+ "eval_recall": 0.8630705394190872,
145
+ "eval_runtime": 13.7915,
146
+ "eval_samples_per_second": 54.381,
147
+ "eval_steps_per_second": 0.87,
148
+ "step": 188
149
+ },
150
+ {
151
+ "epoch": 1.0106382978723405,
152
+ "grad_norm": 4.7592291831970215,
153
+ "learning_rate": 2.4985207100591717e-05,
154
+ "loss": 0.2994,
155
+ "step": 190
156
+ },
157
+ {
158
+ "epoch": 1.0638297872340425,
159
+ "grad_norm": 2.541706085205078,
160
+ "learning_rate": 2.4541420118343198e-05,
161
+ "loss": 0.2576,
162
+ "step": 200
163
+ },
164
+ {
165
+ "epoch": 1.1170212765957448,
166
+ "grad_norm": 2.562579870223999,
167
+ "learning_rate": 2.4097633136094676e-05,
168
+ "loss": 0.1996,
169
+ "step": 210
170
+ },
171
+ {
172
+ "epoch": 1.1702127659574468,
173
+ "grad_norm": 4.691830635070801,
174
+ "learning_rate": 2.3653846153846153e-05,
175
+ "loss": 0.226,
176
+ "step": 220
177
+ },
178
+ {
179
+ "epoch": 1.2234042553191489,
180
+ "grad_norm": 7.327857971191406,
181
+ "learning_rate": 2.3210059171597634e-05,
182
+ "loss": 0.2802,
183
+ "step": 230
184
+ },
185
+ {
186
+ "epoch": 1.2765957446808511,
187
+ "grad_norm": 3.7329373359680176,
188
+ "learning_rate": 2.2766272189349115e-05,
189
+ "loss": 0.1984,
190
+ "step": 240
191
+ },
192
+ {
193
+ "epoch": 1.3297872340425532,
194
+ "grad_norm": 3.0301780700683594,
195
+ "learning_rate": 2.232248520710059e-05,
196
+ "loss": 0.22,
197
+ "step": 250
198
+ },
199
+ {
200
+ "epoch": 1.3829787234042552,
201
+ "grad_norm": 2.9645023345947266,
202
+ "learning_rate": 2.187869822485207e-05,
203
+ "loss": 0.2505,
204
+ "step": 260
205
+ },
206
+ {
207
+ "epoch": 1.4361702127659575,
208
+ "grad_norm": 4.786125659942627,
209
+ "learning_rate": 2.1434911242603552e-05,
210
+ "loss": 0.2214,
211
+ "step": 270
212
+ },
213
+ {
214
+ "epoch": 1.4893617021276595,
215
+ "grad_norm": 4.271234035491943,
216
+ "learning_rate": 2.099112426035503e-05,
217
+ "loss": 0.2128,
218
+ "step": 280
219
+ },
220
+ {
221
+ "epoch": 1.5425531914893615,
222
+ "grad_norm": 4.055238246917725,
223
+ "learning_rate": 2.054733727810651e-05,
224
+ "loss": 0.2123,
225
+ "step": 290
226
+ },
227
+ {
228
+ "epoch": 1.5957446808510638,
229
+ "grad_norm": 2.908184766769409,
230
+ "learning_rate": 2.010355029585799e-05,
231
+ "loss": 0.2081,
232
+ "step": 300
233
+ },
234
+ {
235
+ "epoch": 1.648936170212766,
236
+ "grad_norm": 3.0964977741241455,
237
+ "learning_rate": 1.9659763313609466e-05,
238
+ "loss": 0.2078,
239
+ "step": 310
240
+ },
241
+ {
242
+ "epoch": 1.702127659574468,
243
+ "grad_norm": 3.982306957244873,
244
+ "learning_rate": 1.9215976331360947e-05,
245
+ "loss": 0.182,
246
+ "step": 320
247
+ },
248
+ {
249
+ "epoch": 1.7553191489361701,
250
+ "grad_norm": 6.94441556930542,
251
+ "learning_rate": 1.8772189349112428e-05,
252
+ "loss": 0.1946,
253
+ "step": 330
254
+ },
255
+ {
256
+ "epoch": 1.8085106382978724,
257
+ "grad_norm": 3.193429708480835,
258
+ "learning_rate": 1.8328402366863906e-05,
259
+ "loss": 0.1965,
260
+ "step": 340
261
+ },
262
+ {
263
+ "epoch": 1.8617021276595744,
264
+ "grad_norm": 5.883879661560059,
265
+ "learning_rate": 1.7884615384615384e-05,
266
+ "loss": 0.2246,
267
+ "step": 350
268
+ },
269
+ {
270
+ "epoch": 1.9148936170212765,
271
+ "grad_norm": 3.561206817626953,
272
+ "learning_rate": 1.7440828402366865e-05,
273
+ "loss": 0.2101,
274
+ "step": 360
275
+ },
276
+ {
277
+ "epoch": 1.9680851063829787,
278
+ "grad_norm": 3.8192715644836426,
279
+ "learning_rate": 1.6997041420118342e-05,
280
+ "loss": 0.2027,
281
+ "step": 370
282
+ },
283
+ {
284
+ "epoch": 2.0,
285
+ "eval_accuracy": 0.868,
286
+ "eval_f1": 0.7744874715261959,
287
+ "eval_loss": 0.3263615667819977,
288
+ "eval_precision": 0.8585858585858586,
289
+ "eval_recall": 0.7053941908713693,
290
+ "eval_runtime": 13.525,
291
+ "eval_samples_per_second": 55.453,
292
+ "eval_steps_per_second": 0.887,
293
+ "step": 376
294
+ }
295
+ ],
296
+ "logging_steps": 10,
297
+ "max_steps": 752,
298
+ "num_input_tokens_seen": 0,
299
+ "num_train_epochs": 4,
300
+ "save_steps": 500,
301
+ "stateful_callbacks": {
302
+ "TrainerControl": {
303
+ "args": {
304
+ "should_epoch_stop": false,
305
+ "should_evaluate": false,
306
+ "should_log": false,
307
+ "should_save": true,
308
+ "should_training_stop": false
309
+ },
310
+ "attributes": {}
311
+ }
312
+ },
313
+ "total_flos": 141279067920504.0,
314
+ "train_batch_size": 32,
315
+ "trial_name": null,
316
+ "trial_params": null
317
+ }
checkpoint-376/vocab.txt ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-564/config.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "activation": "gelu",
3
+ "architectures": [
4
+ "DistilBertForSequenceClassification"
5
+ ],
6
+ "attention_dropout": 0.1,
7
+ "dim": 768,
8
+ "dropout": 0.1,
9
+ "dtype": "float32",
10
+ "hidden_dim": 3072,
11
+ "id2label": {
12
+ "0": "safe",
13
+ "1": "unsafe"
14
+ },
15
+ "initializer_range": 0.02,
16
+ "label2id": {
17
+ "safe": 0,
18
+ "unsafe": 1
19
+ },
20
+ "max_position_embeddings": 512,
21
+ "model_type": "distilbert",
22
+ "n_heads": 12,
23
+ "n_layers": 6,
24
+ "pad_token_id": 0,
25
+ "problem_type": "single_label_classification",
26
+ "qa_dropout": 0.1,
27
+ "seq_classif_dropout": 0.2,
28
+ "sinusoidal_pos_embds": false,
29
+ "tie_weights_": true,
30
+ "transformers_version": "4.57.1",
31
+ "vocab_size": 30522
32
+ }
checkpoint-564/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2eae0144156802debdd211b8dd8fb55a3f5c1173259a37d94d31f44d14dc03c8
3
+ size 267832560
checkpoint-564/special_tokens_map.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "cls_token": "[CLS]",
3
+ "mask_token": "[MASK]",
4
+ "pad_token": "[PAD]",
5
+ "sep_token": "[SEP]",
6
+ "unk_token": "[UNK]"
7
+ }
checkpoint-564/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-564/tokenizer_config.json ADDED
@@ -0,0 +1,56 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "added_tokens_decoder": {
3
+ "0": {
4
+ "content": "[PAD]",
5
+ "lstrip": false,
6
+ "normalized": false,
7
+ "rstrip": false,
8
+ "single_word": false,
9
+ "special": true
10
+ },
11
+ "100": {
12
+ "content": "[UNK]",
13
+ "lstrip": false,
14
+ "normalized": false,
15
+ "rstrip": false,
16
+ "single_word": false,
17
+ "special": true
18
+ },
19
+ "101": {
20
+ "content": "[CLS]",
21
+ "lstrip": false,
22
+ "normalized": false,
23
+ "rstrip": false,
24
+ "single_word": false,
25
+ "special": true
26
+ },
27
+ "102": {
28
+ "content": "[SEP]",
29
+ "lstrip": false,
30
+ "normalized": false,
31
+ "rstrip": false,
32
+ "single_word": false,
33
+ "special": true
34
+ },
35
+ "103": {
36
+ "content": "[MASK]",
37
+ "lstrip": false,
38
+ "normalized": false,
39
+ "rstrip": false,
40
+ "single_word": false,
41
+ "special": true
42
+ }
43
+ },
44
+ "clean_up_tokenization_spaces": false,
45
+ "cls_token": "[CLS]",
46
+ "do_lower_case": true,
47
+ "extra_special_tokens": {},
48
+ "mask_token": "[MASK]",
49
+ "model_max_length": 512,
50
+ "pad_token": "[PAD]",
51
+ "sep_token": "[SEP]",
52
+ "strip_accents": null,
53
+ "tokenize_chinese_chars": true,
54
+ "tokenizer_class": "DistilBertTokenizer",
55
+ "unk_token": "[UNK]"
56
+ }
checkpoint-564/trainer_state.json ADDED
@@ -0,0 +1,462 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": 188,
3
+ "best_metric": 0.8062015503875969,
4
+ "best_model_checkpoint": "models/distilbert_moderation_finetuned\\checkpoint-188",
5
+ "epoch": 3.0,
6
+ "eval_steps": 500,
7
+ "global_step": 564,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.05319148936170213,
14
+ "grad_norm": 2.128387689590454,
15
+ "learning_rate": 3.5526315789473683e-06,
16
+ "loss": 0.6783,
17
+ "step": 10
18
+ },
19
+ {
20
+ "epoch": 0.10638297872340426,
21
+ "grad_norm": 2.286456823348999,
22
+ "learning_rate": 7.5e-06,
23
+ "loss": 0.6585,
24
+ "step": 20
25
+ },
26
+ {
27
+ "epoch": 0.1595744680851064,
28
+ "grad_norm": 1.3711037635803223,
29
+ "learning_rate": 1.1447368421052632e-05,
30
+ "loss": 0.6436,
31
+ "step": 30
32
+ },
33
+ {
34
+ "epoch": 0.2127659574468085,
35
+ "grad_norm": 1.3716976642608643,
36
+ "learning_rate": 1.5394736842105264e-05,
37
+ "loss": 0.6011,
38
+ "step": 40
39
+ },
40
+ {
41
+ "epoch": 0.26595744680851063,
42
+ "grad_norm": 1.3208516836166382,
43
+ "learning_rate": 1.9342105263157896e-05,
44
+ "loss": 0.5312,
45
+ "step": 50
46
+ },
47
+ {
48
+ "epoch": 0.3191489361702128,
49
+ "grad_norm": 1.8064684867858887,
50
+ "learning_rate": 2.3289473684210525e-05,
51
+ "loss": 0.4832,
52
+ "step": 60
53
+ },
54
+ {
55
+ "epoch": 0.3723404255319149,
56
+ "grad_norm": 3.465531826019287,
57
+ "learning_rate": 2.723684210526316e-05,
58
+ "loss": 0.4031,
59
+ "step": 70
60
+ },
61
+ {
62
+ "epoch": 0.425531914893617,
63
+ "grad_norm": 6.162868499755859,
64
+ "learning_rate": 2.9866863905325445e-05,
65
+ "loss": 0.3736,
66
+ "step": 80
67
+ },
68
+ {
69
+ "epoch": 0.4787234042553192,
70
+ "grad_norm": 5.821077346801758,
71
+ "learning_rate": 2.9423076923076923e-05,
72
+ "loss": 0.3381,
73
+ "step": 90
74
+ },
75
+ {
76
+ "epoch": 0.5319148936170213,
77
+ "grad_norm": 3.1542975902557373,
78
+ "learning_rate": 2.8979289940828404e-05,
79
+ "loss": 0.332,
80
+ "step": 100
81
+ },
82
+ {
83
+ "epoch": 0.5851063829787234,
84
+ "grad_norm": 3.219808578491211,
85
+ "learning_rate": 2.8535502958579882e-05,
86
+ "loss": 0.3636,
87
+ "step": 110
88
+ },
89
+ {
90
+ "epoch": 0.6382978723404256,
91
+ "grad_norm": 3.52590012550354,
92
+ "learning_rate": 2.809171597633136e-05,
93
+ "loss": 0.3288,
94
+ "step": 120
95
+ },
96
+ {
97
+ "epoch": 0.6914893617021277,
98
+ "grad_norm": 2.8622491359710693,
99
+ "learning_rate": 2.764792899408284e-05,
100
+ "loss": 0.3061,
101
+ "step": 130
102
+ },
103
+ {
104
+ "epoch": 0.7446808510638298,
105
+ "grad_norm": 2.1935667991638184,
106
+ "learning_rate": 2.7204142011834322e-05,
107
+ "loss": 0.3211,
108
+ "step": 140
109
+ },
110
+ {
111
+ "epoch": 0.7978723404255319,
112
+ "grad_norm": 3.4211273193359375,
113
+ "learning_rate": 2.67603550295858e-05,
114
+ "loss": 0.2889,
115
+ "step": 150
116
+ },
117
+ {
118
+ "epoch": 0.851063829787234,
119
+ "grad_norm": 3.5952651500701904,
120
+ "learning_rate": 2.6316568047337277e-05,
121
+ "loss": 0.3066,
122
+ "step": 160
123
+ },
124
+ {
125
+ "epoch": 0.9042553191489362,
126
+ "grad_norm": 4.181521892547607,
127
+ "learning_rate": 2.5872781065088758e-05,
128
+ "loss": 0.3504,
129
+ "step": 170
130
+ },
131
+ {
132
+ "epoch": 0.9574468085106383,
133
+ "grad_norm": 3.022378921508789,
134
+ "learning_rate": 2.542899408284024e-05,
135
+ "loss": 0.2445,
136
+ "step": 180
137
+ },
138
+ {
139
+ "epoch": 1.0,
140
+ "eval_accuracy": 0.8666666666666667,
141
+ "eval_f1": 0.8062015503875969,
142
+ "eval_loss": 0.2918033301830292,
143
+ "eval_precision": 0.7563636363636363,
144
+ "eval_recall": 0.8630705394190872,
145
+ "eval_runtime": 13.7915,
146
+ "eval_samples_per_second": 54.381,
147
+ "eval_steps_per_second": 0.87,
148
+ "step": 188
149
+ },
150
+ {
151
+ "epoch": 1.0106382978723405,
152
+ "grad_norm": 4.7592291831970215,
153
+ "learning_rate": 2.4985207100591717e-05,
154
+ "loss": 0.2994,
155
+ "step": 190
156
+ },
157
+ {
158
+ "epoch": 1.0638297872340425,
159
+ "grad_norm": 2.541706085205078,
160
+ "learning_rate": 2.4541420118343198e-05,
161
+ "loss": 0.2576,
162
+ "step": 200
163
+ },
164
+ {
165
+ "epoch": 1.1170212765957448,
166
+ "grad_norm": 2.562579870223999,
167
+ "learning_rate": 2.4097633136094676e-05,
168
+ "loss": 0.1996,
169
+ "step": 210
170
+ },
171
+ {
172
+ "epoch": 1.1702127659574468,
173
+ "grad_norm": 4.691830635070801,
174
+ "learning_rate": 2.3653846153846153e-05,
175
+ "loss": 0.226,
176
+ "step": 220
177
+ },
178
+ {
179
+ "epoch": 1.2234042553191489,
180
+ "grad_norm": 7.327857971191406,
181
+ "learning_rate": 2.3210059171597634e-05,
182
+ "loss": 0.2802,
183
+ "step": 230
184
+ },
185
+ {
186
+ "epoch": 1.2765957446808511,
187
+ "grad_norm": 3.7329373359680176,
188
+ "learning_rate": 2.2766272189349115e-05,
189
+ "loss": 0.1984,
190
+ "step": 240
191
+ },
192
+ {
193
+ "epoch": 1.3297872340425532,
194
+ "grad_norm": 3.0301780700683594,
195
+ "learning_rate": 2.232248520710059e-05,
196
+ "loss": 0.22,
197
+ "step": 250
198
+ },
199
+ {
200
+ "epoch": 1.3829787234042552,
201
+ "grad_norm": 2.9645023345947266,
202
+ "learning_rate": 2.187869822485207e-05,
203
+ "loss": 0.2505,
204
+ "step": 260
205
+ },
206
+ {
207
+ "epoch": 1.4361702127659575,
208
+ "grad_norm": 4.786125659942627,
209
+ "learning_rate": 2.1434911242603552e-05,
210
+ "loss": 0.2214,
211
+ "step": 270
212
+ },
213
+ {
214
+ "epoch": 1.4893617021276595,
215
+ "grad_norm": 4.271234035491943,
216
+ "learning_rate": 2.099112426035503e-05,
217
+ "loss": 0.2128,
218
+ "step": 280
219
+ },
220
+ {
221
+ "epoch": 1.5425531914893615,
222
+ "grad_norm": 4.055238246917725,
223
+ "learning_rate": 2.054733727810651e-05,
224
+ "loss": 0.2123,
225
+ "step": 290
226
+ },
227
+ {
228
+ "epoch": 1.5957446808510638,
229
+ "grad_norm": 2.908184766769409,
230
+ "learning_rate": 2.010355029585799e-05,
231
+ "loss": 0.2081,
232
+ "step": 300
233
+ },
234
+ {
235
+ "epoch": 1.648936170212766,
236
+ "grad_norm": 3.0964977741241455,
237
+ "learning_rate": 1.9659763313609466e-05,
238
+ "loss": 0.2078,
239
+ "step": 310
240
+ },
241
+ {
242
+ "epoch": 1.702127659574468,
243
+ "grad_norm": 3.982306957244873,
244
+ "learning_rate": 1.9215976331360947e-05,
245
+ "loss": 0.182,
246
+ "step": 320
247
+ },
248
+ {
249
+ "epoch": 1.7553191489361701,
250
+ "grad_norm": 6.94441556930542,
251
+ "learning_rate": 1.8772189349112428e-05,
252
+ "loss": 0.1946,
253
+ "step": 330
254
+ },
255
+ {
256
+ "epoch": 1.8085106382978724,
257
+ "grad_norm": 3.193429708480835,
258
+ "learning_rate": 1.8328402366863906e-05,
259
+ "loss": 0.1965,
260
+ "step": 340
261
+ },
262
+ {
263
+ "epoch": 1.8617021276595744,
264
+ "grad_norm": 5.883879661560059,
265
+ "learning_rate": 1.7884615384615384e-05,
266
+ "loss": 0.2246,
267
+ "step": 350
268
+ },
269
+ {
270
+ "epoch": 1.9148936170212765,
271
+ "grad_norm": 3.561206817626953,
272
+ "learning_rate": 1.7440828402366865e-05,
273
+ "loss": 0.2101,
274
+ "step": 360
275
+ },
276
+ {
277
+ "epoch": 1.9680851063829787,
278
+ "grad_norm": 3.8192715644836426,
279
+ "learning_rate": 1.6997041420118342e-05,
280
+ "loss": 0.2027,
281
+ "step": 370
282
+ },
283
+ {
284
+ "epoch": 2.0,
285
+ "eval_accuracy": 0.868,
286
+ "eval_f1": 0.7744874715261959,
287
+ "eval_loss": 0.3263615667819977,
288
+ "eval_precision": 0.8585858585858586,
289
+ "eval_recall": 0.7053941908713693,
290
+ "eval_runtime": 13.525,
291
+ "eval_samples_per_second": 55.453,
292
+ "eval_steps_per_second": 0.887,
293
+ "step": 376
294
+ },
295
+ {
296
+ "epoch": 2.021276595744681,
297
+ "grad_norm": 4.577354431152344,
298
+ "learning_rate": 1.6553254437869823e-05,
299
+ "loss": 0.2085,
300
+ "step": 380
301
+ },
302
+ {
303
+ "epoch": 2.074468085106383,
304
+ "grad_norm": 3.511462688446045,
305
+ "learning_rate": 1.6109467455621304e-05,
306
+ "loss": 0.1305,
307
+ "step": 390
308
+ },
309
+ {
310
+ "epoch": 2.127659574468085,
311
+ "grad_norm": 5.641916751861572,
312
+ "learning_rate": 1.566568047337278e-05,
313
+ "loss": 0.0863,
314
+ "step": 400
315
+ },
316
+ {
317
+ "epoch": 2.1808510638297873,
318
+ "grad_norm": 4.49448823928833,
319
+ "learning_rate": 1.522189349112426e-05,
320
+ "loss": 0.1099,
321
+ "step": 410
322
+ },
323
+ {
324
+ "epoch": 2.2340425531914896,
325
+ "grad_norm": 5.283958911895752,
326
+ "learning_rate": 1.4778106508875741e-05,
327
+ "loss": 0.1282,
328
+ "step": 420
329
+ },
330
+ {
331
+ "epoch": 2.2872340425531914,
332
+ "grad_norm": 8.828641891479492,
333
+ "learning_rate": 1.4334319526627219e-05,
334
+ "loss": 0.1199,
335
+ "step": 430
336
+ },
337
+ {
338
+ "epoch": 2.3404255319148937,
339
+ "grad_norm": 5.02363395690918,
340
+ "learning_rate": 1.3890532544378698e-05,
341
+ "loss": 0.0994,
342
+ "step": 440
343
+ },
344
+ {
345
+ "epoch": 2.393617021276596,
346
+ "grad_norm": 2.679619073867798,
347
+ "learning_rate": 1.3446745562130179e-05,
348
+ "loss": 0.1052,
349
+ "step": 450
350
+ },
351
+ {
352
+ "epoch": 2.4468085106382977,
353
+ "grad_norm": 4.862517356872559,
354
+ "learning_rate": 1.3002958579881657e-05,
355
+ "loss": 0.1275,
356
+ "step": 460
357
+ },
358
+ {
359
+ "epoch": 2.5,
360
+ "grad_norm": 5.10927677154541,
361
+ "learning_rate": 1.2559171597633136e-05,
362
+ "loss": 0.1023,
363
+ "step": 470
364
+ },
365
+ {
366
+ "epoch": 2.5531914893617023,
367
+ "grad_norm": 0.47581568360328674,
368
+ "learning_rate": 1.2115384615384615e-05,
369
+ "loss": 0.1321,
370
+ "step": 480
371
+ },
372
+ {
373
+ "epoch": 2.6063829787234045,
374
+ "grad_norm": 9.42883586883545,
375
+ "learning_rate": 1.1671597633136095e-05,
376
+ "loss": 0.1094,
377
+ "step": 490
378
+ },
379
+ {
380
+ "epoch": 2.6595744680851063,
381
+ "grad_norm": 7.33889102935791,
382
+ "learning_rate": 1.1227810650887574e-05,
383
+ "loss": 0.0927,
384
+ "step": 500
385
+ },
386
+ {
387
+ "epoch": 2.7127659574468086,
388
+ "grad_norm": 4.663527965545654,
389
+ "learning_rate": 1.0784023668639054e-05,
390
+ "loss": 0.1834,
391
+ "step": 510
392
+ },
393
+ {
394
+ "epoch": 2.7659574468085104,
395
+ "grad_norm": 7.534264087677002,
396
+ "learning_rate": 1.0340236686390533e-05,
397
+ "loss": 0.1423,
398
+ "step": 520
399
+ },
400
+ {
401
+ "epoch": 2.8191489361702127,
402
+ "grad_norm": 7.19485330581665,
403
+ "learning_rate": 9.896449704142012e-06,
404
+ "loss": 0.1361,
405
+ "step": 530
406
+ },
407
+ {
408
+ "epoch": 2.872340425531915,
409
+ "grad_norm": 4.051627159118652,
410
+ "learning_rate": 9.452662721893492e-06,
411
+ "loss": 0.109,
412
+ "step": 540
413
+ },
414
+ {
415
+ "epoch": 2.925531914893617,
416
+ "grad_norm": 4.114497184753418,
417
+ "learning_rate": 9.00887573964497e-06,
418
+ "loss": 0.0761,
419
+ "step": 550
420
+ },
421
+ {
422
+ "epoch": 2.978723404255319,
423
+ "grad_norm": 8.258723258972168,
424
+ "learning_rate": 8.56508875739645e-06,
425
+ "loss": 0.2024,
426
+ "step": 560
427
+ },
428
+ {
429
+ "epoch": 3.0,
430
+ "eval_accuracy": 0.8653333333333333,
431
+ "eval_f1": 0.7900207900207901,
432
+ "eval_loss": 0.3574095368385315,
433
+ "eval_precision": 0.7916666666666666,
434
+ "eval_recall": 0.7883817427385892,
435
+ "eval_runtime": 19.6116,
436
+ "eval_samples_per_second": 38.243,
437
+ "eval_steps_per_second": 0.612,
438
+ "step": 564
439
+ }
440
+ ],
441
+ "logging_steps": 10,
442
+ "max_steps": 752,
443
+ "num_input_tokens_seen": 0,
444
+ "num_train_epochs": 4,
445
+ "save_steps": 500,
446
+ "stateful_callbacks": {
447
+ "TrainerControl": {
448
+ "args": {
449
+ "should_epoch_stop": false,
450
+ "should_evaluate": false,
451
+ "should_log": false,
452
+ "should_save": true,
453
+ "should_training_stop": false
454
+ },
455
+ "attributes": {}
456
+ }
457
+ },
458
+ "total_flos": 211374502389792.0,
459
+ "train_batch_size": 32,
460
+ "trial_name": null,
461
+ "trial_params": null
462
+ }
checkpoint-564/vocab.txt ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-752/config.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "activation": "gelu",
3
+ "architectures": [
4
+ "DistilBertForSequenceClassification"
5
+ ],
6
+ "attention_dropout": 0.1,
7
+ "dim": 768,
8
+ "dropout": 0.1,
9
+ "dtype": "float32",
10
+ "hidden_dim": 3072,
11
+ "id2label": {
12
+ "0": "safe",
13
+ "1": "unsafe"
14
+ },
15
+ "initializer_range": 0.02,
16
+ "label2id": {
17
+ "safe": 0,
18
+ "unsafe": 1
19
+ },
20
+ "max_position_embeddings": 512,
21
+ "model_type": "distilbert",
22
+ "n_heads": 12,
23
+ "n_layers": 6,
24
+ "pad_token_id": 0,
25
+ "problem_type": "single_label_classification",
26
+ "qa_dropout": 0.1,
27
+ "seq_classif_dropout": 0.2,
28
+ "sinusoidal_pos_embds": false,
29
+ "tie_weights_": true,
30
+ "transformers_version": "4.57.1",
31
+ "vocab_size": 30522
32
+ }
checkpoint-752/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a8bdbe8c0966dca2f7eb7faac0cc21ee922977bdd349d10b3967642ef836e259
3
+ size 267832560
checkpoint-752/special_tokens_map.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "cls_token": "[CLS]",
3
+ "mask_token": "[MASK]",
4
+ "pad_token": "[PAD]",
5
+ "sep_token": "[SEP]",
6
+ "unk_token": "[UNK]"
7
+ }
checkpoint-752/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-752/tokenizer_config.json ADDED
@@ -0,0 +1,56 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "added_tokens_decoder": {
3
+ "0": {
4
+ "content": "[PAD]",
5
+ "lstrip": false,
6
+ "normalized": false,
7
+ "rstrip": false,
8
+ "single_word": false,
9
+ "special": true
10
+ },
11
+ "100": {
12
+ "content": "[UNK]",
13
+ "lstrip": false,
14
+ "normalized": false,
15
+ "rstrip": false,
16
+ "single_word": false,
17
+ "special": true
18
+ },
19
+ "101": {
20
+ "content": "[CLS]",
21
+ "lstrip": false,
22
+ "normalized": false,
23
+ "rstrip": false,
24
+ "single_word": false,
25
+ "special": true
26
+ },
27
+ "102": {
28
+ "content": "[SEP]",
29
+ "lstrip": false,
30
+ "normalized": false,
31
+ "rstrip": false,
32
+ "single_word": false,
33
+ "special": true
34
+ },
35
+ "103": {
36
+ "content": "[MASK]",
37
+ "lstrip": false,
38
+ "normalized": false,
39
+ "rstrip": false,
40
+ "single_word": false,
41
+ "special": true
42
+ }
43
+ },
44
+ "clean_up_tokenization_spaces": false,
45
+ "cls_token": "[CLS]",
46
+ "do_lower_case": true,
47
+ "extra_special_tokens": {},
48
+ "mask_token": "[MASK]",
49
+ "model_max_length": 512,
50
+ "pad_token": "[PAD]",
51
+ "sep_token": "[SEP]",
52
+ "strip_accents": null,
53
+ "tokenize_chinese_chars": true,
54
+ "tokenizer_class": "DistilBertTokenizer",
55
+ "unk_token": "[UNK]"
56
+ }
checkpoint-752/trainer_state.json ADDED
@@ -0,0 +1,607 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": 752,
3
+ "best_metric": 0.8179959100204499,
4
+ "best_model_checkpoint": "models/distilbert_moderation_finetuned\\checkpoint-752",
5
+ "epoch": 4.0,
6
+ "eval_steps": 500,
7
+ "global_step": 752,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.05319148936170213,
14
+ "grad_norm": 2.128387689590454,
15
+ "learning_rate": 3.5526315789473683e-06,
16
+ "loss": 0.6783,
17
+ "step": 10
18
+ },
19
+ {
20
+ "epoch": 0.10638297872340426,
21
+ "grad_norm": 2.286456823348999,
22
+ "learning_rate": 7.5e-06,
23
+ "loss": 0.6585,
24
+ "step": 20
25
+ },
26
+ {
27
+ "epoch": 0.1595744680851064,
28
+ "grad_norm": 1.3711037635803223,
29
+ "learning_rate": 1.1447368421052632e-05,
30
+ "loss": 0.6436,
31
+ "step": 30
32
+ },
33
+ {
34
+ "epoch": 0.2127659574468085,
35
+ "grad_norm": 1.3716976642608643,
36
+ "learning_rate": 1.5394736842105264e-05,
37
+ "loss": 0.6011,
38
+ "step": 40
39
+ },
40
+ {
41
+ "epoch": 0.26595744680851063,
42
+ "grad_norm": 1.3208516836166382,
43
+ "learning_rate": 1.9342105263157896e-05,
44
+ "loss": 0.5312,
45
+ "step": 50
46
+ },
47
+ {
48
+ "epoch": 0.3191489361702128,
49
+ "grad_norm": 1.8064684867858887,
50
+ "learning_rate": 2.3289473684210525e-05,
51
+ "loss": 0.4832,
52
+ "step": 60
53
+ },
54
+ {
55
+ "epoch": 0.3723404255319149,
56
+ "grad_norm": 3.465531826019287,
57
+ "learning_rate": 2.723684210526316e-05,
58
+ "loss": 0.4031,
59
+ "step": 70
60
+ },
61
+ {
62
+ "epoch": 0.425531914893617,
63
+ "grad_norm": 6.162868499755859,
64
+ "learning_rate": 2.9866863905325445e-05,
65
+ "loss": 0.3736,
66
+ "step": 80
67
+ },
68
+ {
69
+ "epoch": 0.4787234042553192,
70
+ "grad_norm": 5.821077346801758,
71
+ "learning_rate": 2.9423076923076923e-05,
72
+ "loss": 0.3381,
73
+ "step": 90
74
+ },
75
+ {
76
+ "epoch": 0.5319148936170213,
77
+ "grad_norm": 3.1542975902557373,
78
+ "learning_rate": 2.8979289940828404e-05,
79
+ "loss": 0.332,
80
+ "step": 100
81
+ },
82
+ {
83
+ "epoch": 0.5851063829787234,
84
+ "grad_norm": 3.219808578491211,
85
+ "learning_rate": 2.8535502958579882e-05,
86
+ "loss": 0.3636,
87
+ "step": 110
88
+ },
89
+ {
90
+ "epoch": 0.6382978723404256,
91
+ "grad_norm": 3.52590012550354,
92
+ "learning_rate": 2.809171597633136e-05,
93
+ "loss": 0.3288,
94
+ "step": 120
95
+ },
96
+ {
97
+ "epoch": 0.6914893617021277,
98
+ "grad_norm": 2.8622491359710693,
99
+ "learning_rate": 2.764792899408284e-05,
100
+ "loss": 0.3061,
101
+ "step": 130
102
+ },
103
+ {
104
+ "epoch": 0.7446808510638298,
105
+ "grad_norm": 2.1935667991638184,
106
+ "learning_rate": 2.7204142011834322e-05,
107
+ "loss": 0.3211,
108
+ "step": 140
109
+ },
110
+ {
111
+ "epoch": 0.7978723404255319,
112
+ "grad_norm": 3.4211273193359375,
113
+ "learning_rate": 2.67603550295858e-05,
114
+ "loss": 0.2889,
115
+ "step": 150
116
+ },
117
+ {
118
+ "epoch": 0.851063829787234,
119
+ "grad_norm": 3.5952651500701904,
120
+ "learning_rate": 2.6316568047337277e-05,
121
+ "loss": 0.3066,
122
+ "step": 160
123
+ },
124
+ {
125
+ "epoch": 0.9042553191489362,
126
+ "grad_norm": 4.181521892547607,
127
+ "learning_rate": 2.5872781065088758e-05,
128
+ "loss": 0.3504,
129
+ "step": 170
130
+ },
131
+ {
132
+ "epoch": 0.9574468085106383,
133
+ "grad_norm": 3.022378921508789,
134
+ "learning_rate": 2.542899408284024e-05,
135
+ "loss": 0.2445,
136
+ "step": 180
137
+ },
138
+ {
139
+ "epoch": 1.0,
140
+ "eval_accuracy": 0.8666666666666667,
141
+ "eval_f1": 0.8062015503875969,
142
+ "eval_loss": 0.2918033301830292,
143
+ "eval_precision": 0.7563636363636363,
144
+ "eval_recall": 0.8630705394190872,
145
+ "eval_runtime": 13.7915,
146
+ "eval_samples_per_second": 54.381,
147
+ "eval_steps_per_second": 0.87,
148
+ "step": 188
149
+ },
150
+ {
151
+ "epoch": 1.0106382978723405,
152
+ "grad_norm": 4.7592291831970215,
153
+ "learning_rate": 2.4985207100591717e-05,
154
+ "loss": 0.2994,
155
+ "step": 190
156
+ },
157
+ {
158
+ "epoch": 1.0638297872340425,
159
+ "grad_norm": 2.541706085205078,
160
+ "learning_rate": 2.4541420118343198e-05,
161
+ "loss": 0.2576,
162
+ "step": 200
163
+ },
164
+ {
165
+ "epoch": 1.1170212765957448,
166
+ "grad_norm": 2.562579870223999,
167
+ "learning_rate": 2.4097633136094676e-05,
168
+ "loss": 0.1996,
169
+ "step": 210
170
+ },
171
+ {
172
+ "epoch": 1.1702127659574468,
173
+ "grad_norm": 4.691830635070801,
174
+ "learning_rate": 2.3653846153846153e-05,
175
+ "loss": 0.226,
176
+ "step": 220
177
+ },
178
+ {
179
+ "epoch": 1.2234042553191489,
180
+ "grad_norm": 7.327857971191406,
181
+ "learning_rate": 2.3210059171597634e-05,
182
+ "loss": 0.2802,
183
+ "step": 230
184
+ },
185
+ {
186
+ "epoch": 1.2765957446808511,
187
+ "grad_norm": 3.7329373359680176,
188
+ "learning_rate": 2.2766272189349115e-05,
189
+ "loss": 0.1984,
190
+ "step": 240
191
+ },
192
+ {
193
+ "epoch": 1.3297872340425532,
194
+ "grad_norm": 3.0301780700683594,
195
+ "learning_rate": 2.232248520710059e-05,
196
+ "loss": 0.22,
197
+ "step": 250
198
+ },
199
+ {
200
+ "epoch": 1.3829787234042552,
201
+ "grad_norm": 2.9645023345947266,
202
+ "learning_rate": 2.187869822485207e-05,
203
+ "loss": 0.2505,
204
+ "step": 260
205
+ },
206
+ {
207
+ "epoch": 1.4361702127659575,
208
+ "grad_norm": 4.786125659942627,
209
+ "learning_rate": 2.1434911242603552e-05,
210
+ "loss": 0.2214,
211
+ "step": 270
212
+ },
213
+ {
214
+ "epoch": 1.4893617021276595,
215
+ "grad_norm": 4.271234035491943,
216
+ "learning_rate": 2.099112426035503e-05,
217
+ "loss": 0.2128,
218
+ "step": 280
219
+ },
220
+ {
221
+ "epoch": 1.5425531914893615,
222
+ "grad_norm": 4.055238246917725,
223
+ "learning_rate": 2.054733727810651e-05,
224
+ "loss": 0.2123,
225
+ "step": 290
226
+ },
227
+ {
228
+ "epoch": 1.5957446808510638,
229
+ "grad_norm": 2.908184766769409,
230
+ "learning_rate": 2.010355029585799e-05,
231
+ "loss": 0.2081,
232
+ "step": 300
233
+ },
234
+ {
235
+ "epoch": 1.648936170212766,
236
+ "grad_norm": 3.0964977741241455,
237
+ "learning_rate": 1.9659763313609466e-05,
238
+ "loss": 0.2078,
239
+ "step": 310
240
+ },
241
+ {
242
+ "epoch": 1.702127659574468,
243
+ "grad_norm": 3.982306957244873,
244
+ "learning_rate": 1.9215976331360947e-05,
245
+ "loss": 0.182,
246
+ "step": 320
247
+ },
248
+ {
249
+ "epoch": 1.7553191489361701,
250
+ "grad_norm": 6.94441556930542,
251
+ "learning_rate": 1.8772189349112428e-05,
252
+ "loss": 0.1946,
253
+ "step": 330
254
+ },
255
+ {
256
+ "epoch": 1.8085106382978724,
257
+ "grad_norm": 3.193429708480835,
258
+ "learning_rate": 1.8328402366863906e-05,
259
+ "loss": 0.1965,
260
+ "step": 340
261
+ },
262
+ {
263
+ "epoch": 1.8617021276595744,
264
+ "grad_norm": 5.883879661560059,
265
+ "learning_rate": 1.7884615384615384e-05,
266
+ "loss": 0.2246,
267
+ "step": 350
268
+ },
269
+ {
270
+ "epoch": 1.9148936170212765,
271
+ "grad_norm": 3.561206817626953,
272
+ "learning_rate": 1.7440828402366865e-05,
273
+ "loss": 0.2101,
274
+ "step": 360
275
+ },
276
+ {
277
+ "epoch": 1.9680851063829787,
278
+ "grad_norm": 3.8192715644836426,
279
+ "learning_rate": 1.6997041420118342e-05,
280
+ "loss": 0.2027,
281
+ "step": 370
282
+ },
283
+ {
284
+ "epoch": 2.0,
285
+ "eval_accuracy": 0.868,
286
+ "eval_f1": 0.7744874715261959,
287
+ "eval_loss": 0.3263615667819977,
288
+ "eval_precision": 0.8585858585858586,
289
+ "eval_recall": 0.7053941908713693,
290
+ "eval_runtime": 13.525,
291
+ "eval_samples_per_second": 55.453,
292
+ "eval_steps_per_second": 0.887,
293
+ "step": 376
294
+ },
295
+ {
296
+ "epoch": 2.021276595744681,
297
+ "grad_norm": 4.577354431152344,
298
+ "learning_rate": 1.6553254437869823e-05,
299
+ "loss": 0.2085,
300
+ "step": 380
301
+ },
302
+ {
303
+ "epoch": 2.074468085106383,
304
+ "grad_norm": 3.511462688446045,
305
+ "learning_rate": 1.6109467455621304e-05,
306
+ "loss": 0.1305,
307
+ "step": 390
308
+ },
309
+ {
310
+ "epoch": 2.127659574468085,
311
+ "grad_norm": 5.641916751861572,
312
+ "learning_rate": 1.566568047337278e-05,
313
+ "loss": 0.0863,
314
+ "step": 400
315
+ },
316
+ {
317
+ "epoch": 2.1808510638297873,
318
+ "grad_norm": 4.49448823928833,
319
+ "learning_rate": 1.522189349112426e-05,
320
+ "loss": 0.1099,
321
+ "step": 410
322
+ },
323
+ {
324
+ "epoch": 2.2340425531914896,
325
+ "grad_norm": 5.283958911895752,
326
+ "learning_rate": 1.4778106508875741e-05,
327
+ "loss": 0.1282,
328
+ "step": 420
329
+ },
330
+ {
331
+ "epoch": 2.2872340425531914,
332
+ "grad_norm": 8.828641891479492,
333
+ "learning_rate": 1.4334319526627219e-05,
334
+ "loss": 0.1199,
335
+ "step": 430
336
+ },
337
+ {
338
+ "epoch": 2.3404255319148937,
339
+ "grad_norm": 5.02363395690918,
340
+ "learning_rate": 1.3890532544378698e-05,
341
+ "loss": 0.0994,
342
+ "step": 440
343
+ },
344
+ {
345
+ "epoch": 2.393617021276596,
346
+ "grad_norm": 2.679619073867798,
347
+ "learning_rate": 1.3446745562130179e-05,
348
+ "loss": 0.1052,
349
+ "step": 450
350
+ },
351
+ {
352
+ "epoch": 2.4468085106382977,
353
+ "grad_norm": 4.862517356872559,
354
+ "learning_rate": 1.3002958579881657e-05,
355
+ "loss": 0.1275,
356
+ "step": 460
357
+ },
358
+ {
359
+ "epoch": 2.5,
360
+ "grad_norm": 5.10927677154541,
361
+ "learning_rate": 1.2559171597633136e-05,
362
+ "loss": 0.1023,
363
+ "step": 470
364
+ },
365
+ {
366
+ "epoch": 2.5531914893617023,
367
+ "grad_norm": 0.47581568360328674,
368
+ "learning_rate": 1.2115384615384615e-05,
369
+ "loss": 0.1321,
370
+ "step": 480
371
+ },
372
+ {
373
+ "epoch": 2.6063829787234045,
374
+ "grad_norm": 9.42883586883545,
375
+ "learning_rate": 1.1671597633136095e-05,
376
+ "loss": 0.1094,
377
+ "step": 490
378
+ },
379
+ {
380
+ "epoch": 2.6595744680851063,
381
+ "grad_norm": 7.33889102935791,
382
+ "learning_rate": 1.1227810650887574e-05,
383
+ "loss": 0.0927,
384
+ "step": 500
385
+ },
386
+ {
387
+ "epoch": 2.7127659574468086,
388
+ "grad_norm": 4.663527965545654,
389
+ "learning_rate": 1.0784023668639054e-05,
390
+ "loss": 0.1834,
391
+ "step": 510
392
+ },
393
+ {
394
+ "epoch": 2.7659574468085104,
395
+ "grad_norm": 7.534264087677002,
396
+ "learning_rate": 1.0340236686390533e-05,
397
+ "loss": 0.1423,
398
+ "step": 520
399
+ },
400
+ {
401
+ "epoch": 2.8191489361702127,
402
+ "grad_norm": 7.19485330581665,
403
+ "learning_rate": 9.896449704142012e-06,
404
+ "loss": 0.1361,
405
+ "step": 530
406
+ },
407
+ {
408
+ "epoch": 2.872340425531915,
409
+ "grad_norm": 4.051627159118652,
410
+ "learning_rate": 9.452662721893492e-06,
411
+ "loss": 0.109,
412
+ "step": 540
413
+ },
414
+ {
415
+ "epoch": 2.925531914893617,
416
+ "grad_norm": 4.114497184753418,
417
+ "learning_rate": 9.00887573964497e-06,
418
+ "loss": 0.0761,
419
+ "step": 550
420
+ },
421
+ {
422
+ "epoch": 2.978723404255319,
423
+ "grad_norm": 8.258723258972168,
424
+ "learning_rate": 8.56508875739645e-06,
425
+ "loss": 0.2024,
426
+ "step": 560
427
+ },
428
+ {
429
+ "epoch": 3.0,
430
+ "eval_accuracy": 0.8653333333333333,
431
+ "eval_f1": 0.7900207900207901,
432
+ "eval_loss": 0.3574095368385315,
433
+ "eval_precision": 0.7916666666666666,
434
+ "eval_recall": 0.7883817427385892,
435
+ "eval_runtime": 19.6116,
436
+ "eval_samples_per_second": 38.243,
437
+ "eval_steps_per_second": 0.612,
438
+ "step": 564
439
+ },
440
+ {
441
+ "epoch": 3.0319148936170213,
442
+ "grad_norm": 5.233397483825684,
443
+ "learning_rate": 8.121301775147928e-06,
444
+ "loss": 0.0837,
445
+ "step": 570
446
+ },
447
+ {
448
+ "epoch": 3.0851063829787235,
449
+ "grad_norm": 5.9814348220825195,
450
+ "learning_rate": 7.677514792899408e-06,
451
+ "loss": 0.0645,
452
+ "step": 580
453
+ },
454
+ {
455
+ "epoch": 3.1382978723404253,
456
+ "grad_norm": 3.991489887237549,
457
+ "learning_rate": 7.233727810650888e-06,
458
+ "loss": 0.0817,
459
+ "step": 590
460
+ },
461
+ {
462
+ "epoch": 3.1914893617021276,
463
+ "grad_norm": 8.920732498168945,
464
+ "learning_rate": 6.789940828402366e-06,
465
+ "loss": 0.0489,
466
+ "step": 600
467
+ },
468
+ {
469
+ "epoch": 3.24468085106383,
470
+ "grad_norm": 4.329430103302002,
471
+ "learning_rate": 6.3461538461538466e-06,
472
+ "loss": 0.0924,
473
+ "step": 610
474
+ },
475
+ {
476
+ "epoch": 3.297872340425532,
477
+ "grad_norm": 6.949727535247803,
478
+ "learning_rate": 5.902366863905326e-06,
479
+ "loss": 0.0787,
480
+ "step": 620
481
+ },
482
+ {
483
+ "epoch": 3.351063829787234,
484
+ "grad_norm": 5.294066429138184,
485
+ "learning_rate": 5.458579881656805e-06,
486
+ "loss": 0.0655,
487
+ "step": 630
488
+ },
489
+ {
490
+ "epoch": 3.404255319148936,
491
+ "grad_norm": 0.20922839641571045,
492
+ "learning_rate": 5.014792899408284e-06,
493
+ "loss": 0.0488,
494
+ "step": 640
495
+ },
496
+ {
497
+ "epoch": 3.4574468085106385,
498
+ "grad_norm": 2.275608777999878,
499
+ "learning_rate": 4.571005917159763e-06,
500
+ "loss": 0.0832,
501
+ "step": 650
502
+ },
503
+ {
504
+ "epoch": 3.5106382978723403,
505
+ "grad_norm": 4.606672763824463,
506
+ "learning_rate": 4.127218934911243e-06,
507
+ "loss": 0.0714,
508
+ "step": 660
509
+ },
510
+ {
511
+ "epoch": 3.5638297872340425,
512
+ "grad_norm": 4.710702419281006,
513
+ "learning_rate": 3.683431952662722e-06,
514
+ "loss": 0.0398,
515
+ "step": 670
516
+ },
517
+ {
518
+ "epoch": 3.617021276595745,
519
+ "grad_norm": 0.4108816683292389,
520
+ "learning_rate": 3.239644970414201e-06,
521
+ "loss": 0.076,
522
+ "step": 680
523
+ },
524
+ {
525
+ "epoch": 3.670212765957447,
526
+ "grad_norm": 0.8540160059928894,
527
+ "learning_rate": 2.7958579881656808e-06,
528
+ "loss": 0.0755,
529
+ "step": 690
530
+ },
531
+ {
532
+ "epoch": 3.723404255319149,
533
+ "grad_norm": 0.5908358097076416,
534
+ "learning_rate": 2.3520710059171597e-06,
535
+ "loss": 0.036,
536
+ "step": 700
537
+ },
538
+ {
539
+ "epoch": 3.776595744680851,
540
+ "grad_norm": 3.4919886589050293,
541
+ "learning_rate": 1.908284023668639e-06,
542
+ "loss": 0.063,
543
+ "step": 710
544
+ },
545
+ {
546
+ "epoch": 3.829787234042553,
547
+ "grad_norm": 0.7406045794487,
548
+ "learning_rate": 1.4644970414201183e-06,
549
+ "loss": 0.0929,
550
+ "step": 720
551
+ },
552
+ {
553
+ "epoch": 3.882978723404255,
554
+ "grad_norm": 10.849448204040527,
555
+ "learning_rate": 1.0207100591715976e-06,
556
+ "loss": 0.0642,
557
+ "step": 730
558
+ },
559
+ {
560
+ "epoch": 3.9361702127659575,
561
+ "grad_norm": 8.105487823486328,
562
+ "learning_rate": 5.76923076923077e-07,
563
+ "loss": 0.032,
564
+ "step": 740
565
+ },
566
+ {
567
+ "epoch": 3.9893617021276597,
568
+ "grad_norm": 4.906630039215088,
569
+ "learning_rate": 1.331360946745562e-07,
570
+ "loss": 0.0431,
571
+ "step": 750
572
+ },
573
+ {
574
+ "epoch": 4.0,
575
+ "eval_accuracy": 0.8813333333333333,
576
+ "eval_f1": 0.8179959100204499,
577
+ "eval_loss": 0.3919216990470886,
578
+ "eval_precision": 0.8064516129032258,
579
+ "eval_recall": 0.8298755186721992,
580
+ "eval_runtime": 51.458,
581
+ "eval_samples_per_second": 14.575,
582
+ "eval_steps_per_second": 0.233,
583
+ "step": 752
584
+ }
585
+ ],
586
+ "logging_steps": 10,
587
+ "max_steps": 752,
588
+ "num_input_tokens_seen": 0,
589
+ "num_train_epochs": 4,
590
+ "save_steps": 500,
591
+ "stateful_callbacks": {
592
+ "TrainerControl": {
593
+ "args": {
594
+ "should_epoch_stop": false,
595
+ "should_evaluate": false,
596
+ "should_log": false,
597
+ "should_save": true,
598
+ "should_training_stop": true
599
+ },
600
+ "attributes": {}
601
+ }
602
+ },
603
+ "total_flos": 282008085666120.0,
604
+ "train_batch_size": 32,
605
+ "trial_name": null,
606
+ "trial_params": null
607
+ }
checkpoint-752/vocab.txt ADDED
The diff for this file is too large to render. See raw diff
 
config.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "activation": "gelu",
3
+ "architectures": [
4
+ "DistilBertForSequenceClassification"
5
+ ],
6
+ "attention_dropout": 0.1,
7
+ "dim": 768,
8
+ "dropout": 0.1,
9
+ "dtype": "float32",
10
+ "hidden_dim": 3072,
11
+ "id2label": {
12
+ "0": "safe",
13
+ "1": "unsafe"
14
+ },
15
+ "initializer_range": 0.02,
16
+ "label2id": {
17
+ "safe": 0,
18
+ "unsafe": 1
19
+ },
20
+ "max_position_embeddings": 512,
21
+ "model_type": "distilbert",
22
+ "n_heads": 12,
23
+ "n_layers": 6,
24
+ "pad_token_id": 0,
25
+ "problem_type": "single_label_classification",
26
+ "qa_dropout": 0.1,
27
+ "seq_classif_dropout": 0.2,
28
+ "sinusoidal_pos_embds": false,
29
+ "tie_weights_": true,
30
+ "transformers_version": "4.57.1",
31
+ "vocab_size": 30522
32
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a8bdbe8c0966dca2f7eb7faac0cc21ee922977bdd349d10b3967642ef836e259
3
+ size 267832560
special_tokens_map.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "cls_token": "[CLS]",
3
+ "mask_token": "[MASK]",
4
+ "pad_token": "[PAD]",
5
+ "sep_token": "[SEP]",
6
+ "unk_token": "[UNK]"
7
+ }
tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
tokenizer_config.json ADDED
@@ -0,0 +1,56 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "added_tokens_decoder": {
3
+ "0": {
4
+ "content": "[PAD]",
5
+ "lstrip": false,
6
+ "normalized": false,
7
+ "rstrip": false,
8
+ "single_word": false,
9
+ "special": true
10
+ },
11
+ "100": {
12
+ "content": "[UNK]",
13
+ "lstrip": false,
14
+ "normalized": false,
15
+ "rstrip": false,
16
+ "single_word": false,
17
+ "special": true
18
+ },
19
+ "101": {
20
+ "content": "[CLS]",
21
+ "lstrip": false,
22
+ "normalized": false,
23
+ "rstrip": false,
24
+ "single_word": false,
25
+ "special": true
26
+ },
27
+ "102": {
28
+ "content": "[SEP]",
29
+ "lstrip": false,
30
+ "normalized": false,
31
+ "rstrip": false,
32
+ "single_word": false,
33
+ "special": true
34
+ },
35
+ "103": {
36
+ "content": "[MASK]",
37
+ "lstrip": false,
38
+ "normalized": false,
39
+ "rstrip": false,
40
+ "single_word": false,
41
+ "special": true
42
+ }
43
+ },
44
+ "clean_up_tokenization_spaces": false,
45
+ "cls_token": "[CLS]",
46
+ "do_lower_case": true,
47
+ "extra_special_tokens": {},
48
+ "mask_token": "[MASK]",
49
+ "model_max_length": 512,
50
+ "pad_token": "[PAD]",
51
+ "sep_token": "[SEP]",
52
+ "strip_accents": null,
53
+ "tokenize_chinese_chars": true,
54
+ "tokenizer_class": "DistilBertTokenizer",
55
+ "unk_token": "[UNK]"
56
+ }
vocab.txt ADDED
The diff for this file is too large to render. See raw diff