ayshajavd commited on
Commit
b329c79
·
verified ·
1 Parent(s): 82b783c

v2: GraphCodeBERT + ASL + two-phase + per-class thresholds + calibration

Browse files
Files changed (5) hide show
  1. config.json +68 -66
  2. label_config.json +307 -46
  3. model.safetensors +2 -2
  4. tokenizer.json +0 -0
  5. tokenizer_config.json +0 -2
config.json CHANGED
@@ -8,88 +8,90 @@
8
  "classifier_dropout": null,
9
  "dtype": "float32",
10
  "eos_token_id": 2,
 
11
  "hidden_act": "gelu",
12
  "hidden_dropout_prob": 0.1,
13
  "hidden_size": 768,
14
  "id2label": {
15
- "0": "safe",
16
- "1": "CWE-20",
17
- "2": "CWE-22",
18
- "3": "CWE-78",
19
- "4": "CWE-79",
20
- "5": "CWE-89",
21
- "6": "CWE-94",
22
- "7": "CWE-119",
23
- "8": "CWE-125",
24
- "9": "CWE-190",
25
- "10": "CWE-200",
26
- "11": "CWE-264",
27
- "12": "CWE-269",
28
- "13": "CWE-276",
29
- "14": "CWE-284",
30
- "15": "CWE-287",
31
- "16": "CWE-310",
32
- "17": "CWE-327",
33
- "18": "CWE-330",
34
- "19": "CWE-352",
35
- "20": "CWE-362",
36
- "21": "CWE-399",
37
- "22": "CWE-401",
38
- "23": "CWE-416",
39
- "24": "CWE-434",
40
- "25": "CWE-476",
41
- "26": "CWE-502",
42
- "27": "CWE-601",
43
- "28": "CWE-787",
44
- "29": "CWE-798",
45
- "30": "CWE-918"
46
  },
47
  "initializer_range": 0.02,
48
  "intermediate_size": 3072,
49
  "is_decoder": false,
50
  "label2id": {
51
- "safe": 0,
52
- "CWE-20": 1,
53
- "CWE-22": 2,
54
- "CWE-78": 3,
55
- "CWE-79": 4,
56
- "CWE-89": 5,
57
- "CWE-94": 6,
58
- "CWE-119": 7,
59
- "CWE-125": 8,
60
- "CWE-190": 9,
61
- "CWE-200": 10,
62
- "CWE-264": 11,
63
- "CWE-269": 12,
64
- "CWE-276": 13,
65
- "CWE-284": 14,
66
- "CWE-287": 15,
67
- "CWE-310": 16,
68
- "CWE-327": 17,
69
- "CWE-330": 18,
70
- "CWE-352": 19,
71
- "CWE-362": 20,
72
- "CWE-399": 21,
73
- "CWE-401": 22,
74
- "CWE-416": 23,
75
- "CWE-434": 24,
76
- "CWE-476": 25,
77
- "CWE-502": 26,
78
- "CWE-601": 27,
79
- "CWE-787": 28,
80
- "CWE-798": 29,
81
- "CWE-918": 30
82
  },
83
  "layer_norm_eps": 1e-05,
84
  "max_position_embeddings": 514,
85
  "model_type": "roberta",
86
  "num_attention_heads": 12,
87
- "num_hidden_layers": 6,
 
88
  "pad_token_id": 1,
89
  "problem_type": "multi_label_classification",
90
  "tie_word_embeddings": true,
91
- "transformers_version": "5.6.1",
92
  "type_vocab_size": 1,
93
  "use_cache": false,
94
- "vocab_size": 52000
95
- }
 
8
  "classifier_dropout": null,
9
  "dtype": "float32",
10
  "eos_token_id": 2,
11
+ "gradient_checkpointing": false,
12
  "hidden_act": "gelu",
13
  "hidden_dropout_prob": 0.1,
14
  "hidden_size": 768,
15
  "id2label": {
16
+ "0": "LABEL_0",
17
+ "1": "LABEL_1",
18
+ "2": "LABEL_2",
19
+ "3": "LABEL_3",
20
+ "4": "LABEL_4",
21
+ "5": "LABEL_5",
22
+ "6": "LABEL_6",
23
+ "7": "LABEL_7",
24
+ "8": "LABEL_8",
25
+ "9": "LABEL_9",
26
+ "10": "LABEL_10",
27
+ "11": "LABEL_11",
28
+ "12": "LABEL_12",
29
+ "13": "LABEL_13",
30
+ "14": "LABEL_14",
31
+ "15": "LABEL_15",
32
+ "16": "LABEL_16",
33
+ "17": "LABEL_17",
34
+ "18": "LABEL_18",
35
+ "19": "LABEL_19",
36
+ "20": "LABEL_20",
37
+ "21": "LABEL_21",
38
+ "22": "LABEL_22",
39
+ "23": "LABEL_23",
40
+ "24": "LABEL_24",
41
+ "25": "LABEL_25",
42
+ "26": "LABEL_26",
43
+ "27": "LABEL_27",
44
+ "28": "LABEL_28",
45
+ "29": "LABEL_29",
46
+ "30": "LABEL_30"
47
  },
48
  "initializer_range": 0.02,
49
  "intermediate_size": 3072,
50
  "is_decoder": false,
51
  "label2id": {
52
+ "LABEL_0": 0,
53
+ "LABEL_1": 1,
54
+ "LABEL_10": 10,
55
+ "LABEL_11": 11,
56
+ "LABEL_12": 12,
57
+ "LABEL_13": 13,
58
+ "LABEL_14": 14,
59
+ "LABEL_15": 15,
60
+ "LABEL_16": 16,
61
+ "LABEL_17": 17,
62
+ "LABEL_18": 18,
63
+ "LABEL_19": 19,
64
+ "LABEL_2": 2,
65
+ "LABEL_20": 20,
66
+ "LABEL_21": 21,
67
+ "LABEL_22": 22,
68
+ "LABEL_23": 23,
69
+ "LABEL_24": 24,
70
+ "LABEL_25": 25,
71
+ "LABEL_26": 26,
72
+ "LABEL_27": 27,
73
+ "LABEL_28": 28,
74
+ "LABEL_29": 29,
75
+ "LABEL_3": 3,
76
+ "LABEL_30": 30,
77
+ "LABEL_4": 4,
78
+ "LABEL_5": 5,
79
+ "LABEL_6": 6,
80
+ "LABEL_7": 7,
81
+ "LABEL_8": 8,
82
+ "LABEL_9": 9
83
  },
84
  "layer_norm_eps": 1e-05,
85
  "max_position_embeddings": 514,
86
  "model_type": "roberta",
87
  "num_attention_heads": 12,
88
+ "num_hidden_layers": 12,
89
+ "output_past": true,
90
  "pad_token_id": 1,
91
  "problem_type": "multi_label_classification",
92
  "tie_word_embeddings": true,
93
+ "transformers_version": "5.0.0",
94
  "type_vocab_size": 1,
95
  "use_cache": false,
96
+ "vocab_size": 50265
97
+ }
label_config.json CHANGED
@@ -37,28 +37,28 @@
37
  "CWE-20": "Improper Input Validation",
38
  "CWE-22": "Path Traversal",
39
  "CWE-78": "OS Command Injection",
40
- "CWE-79": "Cross-Site Scripting (XSS)",
41
  "CWE-89": "SQL Injection",
42
  "CWE-94": "Code Injection",
43
  "CWE-119": "Buffer Overflow",
44
  "CWE-125": "Out-of-bounds Read",
45
  "CWE-190": "Integer Overflow",
46
  "CWE-200": "Information Exposure",
47
- "CWE-264": "Permissions/Privileges",
48
- "CWE-269": "Improper Privilege Management",
49
  "CWE-276": "Incorrect Permissions",
50
- "CWE-284": "Improper Access Control",
51
- "CWE-287": "Improper Authentication",
52
- "CWE-310": "Cryptographic Issues",
53
- "CWE-327": "Broken Crypto Algorithm",
54
  "CWE-330": "Insufficient Randomness",
55
  "CWE-352": "CSRF",
56
  "CWE-362": "Race Condition",
57
- "CWE-399": "Resource Management Errors",
58
  "CWE-401": "Memory Leak",
59
  "CWE-416": "Use After Free",
60
- "CWE-434": "Unrestricted File Upload",
61
- "CWE-476": "NULL Pointer Dereference",
62
  "CWE-502": "Insecure Deserialization",
63
  "CWE-601": "Open Redirect",
64
  "CWE-787": "Out-of-bounds Write",
@@ -98,48 +98,309 @@
98
  "CWE-502": "A08",
99
  "CWE-918": "A10"
100
  },
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
101
  "eval_metrics": {
102
- "eval_loss": 0.18841227889060974,
103
- "eval_macro_f1": 0.11565881718569303,
104
- "eval_micro_f1": 0.7043137254901961,
105
- "eval_weighted_f1": 0.8779236820172528,
106
- "eval_macro_precision": 0.08711028841020804,
107
- "eval_macro_recall": 0.2759197005459791,
108
- "eval_f1_safe": 0.9464082098061574,
109
- "eval_f1_CWE-20": 0.03125,
110
- "eval_f1_CWE-22": 0.0,
111
  "eval_f1_CWE-78": 0.0,
112
- "eval_f1_CWE-79": 0.0,
113
- "eval_f1_CWE-89": 0.6,
114
- "eval_f1_CWE-94": 0.43478260869565216,
115
- "eval_f1_CWE-119": 0.12903225806451613,
116
- "eval_f1_CWE-125": 0.13333333333333333,
117
- "eval_f1_CWE-190": 0.4,
118
- "eval_f1_CWE-200": 0.0,
119
- "eval_f1_CWE-264": 0.0,
120
  "eval_f1_CWE-269": 0.0,
121
  "eval_f1_CWE-276": 0.0,
122
- "eval_f1_CWE-284": 0.0,
123
- "eval_f1_CWE-287": 0.0,
124
- "eval_f1_CWE-310": 0.0,
125
  "eval_f1_CWE-327": 0.0,
126
- "eval_f1_CWE-330": 0.0,
127
  "eval_f1_CWE-352": 0.0,
128
- "eval_f1_CWE-362": 0.0,
129
- "eval_f1_CWE-399": 0.18181818181818182,
130
- "eval_f1_CWE-401": 0.0,
131
- "eval_f1_CWE-416": 0.0,
132
  "eval_f1_CWE-434": 0.0,
133
- "eval_f1_CWE-476": 0.21052631578947367,
134
- "eval_f1_CWE-502": 0.2857142857142857,
135
- "eval_f1_CWE-601": 0.0,
136
- "eval_f1_CWE-787": 0.23255813953488372,
137
- "eval_f1_CWE-798": 0.0,
138
- "eval_f1_CWE-918": 0.0,
139
- "eval_runtime": 29.0452,
140
- "eval_samples_per_second": 17.215,
141
- "eval_steps_per_second": 1.102,
142
- "epoch": 2.0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
143
  },
144
- "classification_report": " precision recall f1-score support\n\n safe 0.9742 0.9202 0.9464 451\n CWE-20 0.0169 0.2000 0.0312 5\n CWE-22 0.0000 0.0000 0.0000 0\n CWE-78 0.0000 0.0000 0.0000 0\n CWE-79 0.0000 0.0000 0.0000 1\n CWE-89 0.4286 1.0000 0.6000 6\n CWE-94 0.2778 1.0000 0.4348 5\n CWE-119 0.0723 0.6000 0.1290 10\n CWE-125 0.0741 0.6667 0.1333 3\n CWE-190 0.3333 0.5000 0.4000 4\n CWE-200 0.0000 0.0000 0.0000 1\n CWE-264 0.0000 0.0000 0.0000 0\n CWE-269 0.0000 0.0000 0.0000 0\n CWE-276 0.0000 0.0000 0.0000 0\n CWE-284 0.0000 0.0000 0.0000 0\n CWE-287 0.0000 0.0000 0.0000 0\n CWE-310 0.0000 0.0000 0.0000 0\n CWE-327 0.0000 0.0000 0.0000 0\n CWE-330 0.0000 0.0000 0.0000 0\n CWE-352 0.0000 0.0000 0.0000 0\n CWE-362 0.0000 0.0000 0.0000 0\n CWE-399 0.1000 1.0000 0.1818 3\n CWE-401 0.0000 0.0000 0.0000 0\n CWE-416 0.0000 0.0000 0.0000 1\n CWE-434 0.0000 0.0000 0.0000 0\n CWE-476 0.1250 0.6667 0.2105 3\n CWE-502 0.1667 1.0000 0.2857 2\n CWE-601 0.0000 0.0000 0.0000 0\n CWE-787 0.1316 1.0000 0.2326 5\n CWE-798 0.0000 0.0000 0.0000 0\n CWE-918 0.0000 0.0000 0.0000 0\n\n micro avg 0.5794 0.8980 0.7043 500\n macro avg 0.0871 0.2759 0.1157 500\nweighted avg 0.8947 0.8980 0.8779 500\n samples avg 0.7754 0.8980 0.8074 500\n"
 
 
 
 
 
 
 
 
 
 
 
 
 
145
  }
 
37
  "CWE-20": "Improper Input Validation",
38
  "CWE-22": "Path Traversal",
39
  "CWE-78": "OS Command Injection",
40
+ "CWE-79": "XSS",
41
  "CWE-89": "SQL Injection",
42
  "CWE-94": "Code Injection",
43
  "CWE-119": "Buffer Overflow",
44
  "CWE-125": "Out-of-bounds Read",
45
  "CWE-190": "Integer Overflow",
46
  "CWE-200": "Information Exposure",
47
+ "CWE-264": "Permissions",
48
+ "CWE-269": "Privilege Management",
49
  "CWE-276": "Incorrect Permissions",
50
+ "CWE-284": "Access Control",
51
+ "CWE-287": "Authentication",
52
+ "CWE-310": "Crypto Issues",
53
+ "CWE-327": "Broken Crypto",
54
  "CWE-330": "Insufficient Randomness",
55
  "CWE-352": "CSRF",
56
  "CWE-362": "Race Condition",
57
+ "CWE-399": "Resource Management",
58
  "CWE-401": "Memory Leak",
59
  "CWE-416": "Use After Free",
60
+ "CWE-434": "File Upload",
61
+ "CWE-476": "NULL Pointer Deref",
62
  "CWE-502": "Insecure Deserialization",
63
  "CWE-601": "Open Redirect",
64
  "CWE-787": "Out-of-bounds Write",
 
98
  "CWE-502": "A08",
99
  "CWE-918": "A10"
100
  },
101
+ "optimized_thresholds": {
102
+ "safe": 0.5750000000000002,
103
+ "CWE-20": 0.8750000000000003,
104
+ "CWE-22": 0.7750000000000002,
105
+ "CWE-78": 0.7500000000000002,
106
+ "CWE-79": 0.5750000000000002,
107
+ "CWE-89": 0.5250000000000001,
108
+ "CWE-94": 0.7750000000000002,
109
+ "CWE-119": 0.6000000000000002,
110
+ "CWE-125": 0.7000000000000003,
111
+ "CWE-190": 0.8750000000000003,
112
+ "CWE-200": 0.7500000000000002,
113
+ "CWE-264": 0.8250000000000003,
114
+ "CWE-269": 0.5500000000000003,
115
+ "CWE-276": 0.5,
116
+ "CWE-284": 0.6000000000000002,
117
+ "CWE-287": 0.4250000000000001,
118
+ "CWE-310": 0.8000000000000003,
119
+ "CWE-327": 0.5,
120
+ "CWE-330": 0.7250000000000003,
121
+ "CWE-352": 0.5,
122
+ "CWE-362": 0.6250000000000002,
123
+ "CWE-399": 0.9250000000000004,
124
+ "CWE-401": 0.4250000000000001,
125
+ "CWE-416": 0.6000000000000002,
126
+ "CWE-434": 0.3000000000000001,
127
+ "CWE-476": 0.8000000000000003,
128
+ "CWE-502": 0.8750000000000003,
129
+ "CWE-601": 0.47500000000000014,
130
+ "CWE-787": 0.6000000000000002,
131
+ "CWE-798": 0.25000000000000006,
132
+ "CWE-918": 0.5
133
+ },
134
+ "temperature": 0.6163493394851685,
135
  "eval_metrics": {
136
+ "eval_macro_f1": 0.47583485129569136,
137
+ "eval_micro_f1": 0.9426379247351114,
138
+ "eval_weighted_f1": 0.945370829629837,
139
+ "eval_macro_precision": 0.4824943179864961,
140
+ "eval_macro_recall": 0.5278332334756789,
141
+ "eval_f1_safe": 0.9820306051177853,
142
+ "eval_f1_CWE-20": 0.5721153846153846,
143
+ "eval_f1_CWE-22": 0.5,
 
144
  "eval_f1_CWE-78": 0.0,
145
+ "eval_f1_CWE-79": 0.5932203389830508,
146
+ "eval_f1_CWE-89": 0.7916666666666666,
147
+ "eval_f1_CWE-94": 0.7068273092369478,
148
+ "eval_f1_CWE-119": 0.7054610564010743,
149
+ "eval_f1_CWE-125": 0.5844748858447488,
150
+ "eval_f1_CWE-190": 0.7704918032786885,
151
+ "eval_f1_CWE-200": 0.7046632124352331,
152
+ "eval_f1_CWE-264": 0.697986577181208,
153
  "eval_f1_CWE-269": 0.0,
154
  "eval_f1_CWE-276": 0.0,
155
+ "eval_f1_CWE-284": 0.5128205128205128,
156
+ "eval_f1_CWE-287": 0.25,
157
+ "eval_f1_CWE-310": 0.5,
158
  "eval_f1_CWE-327": 0.0,
159
+ "eval_f1_CWE-330": 0.5,
160
  "eval_f1_CWE-352": 0.0,
161
+ "eval_f1_CWE-362": 0.7010309278350515,
162
+ "eval_f1_CWE-399": 0.6783625730994152,
163
+ "eval_f1_CWE-401": 0.5909090909090909,
164
+ "eval_f1_CWE-416": 0.48484848484848486,
165
  "eval_f1_CWE-434": 0.0,
166
+ "eval_f1_CWE-476": 0.5483870967741935,
167
+ "eval_f1_CWE-502": 0.9,
168
+ "eval_f1_CWE-601": 0.8571428571428571,
169
+ "eval_f1_CWE-787": 0.47558386411889597,
170
+ "eval_f1_CWE-798": 0.14285714285714285,
171
+ "eval_f1_CWE-918": 0.0
172
+ },
173
+ "per_class_metrics": {
174
+ "safe": {
175
+ "f1": 0.9820306051177853,
176
+ "precision": 0.976204001767342,
177
+ "recall": 0.9879271798147556,
178
+ "threshold": 0.5750000000000002,
179
+ "support": 15655
180
+ },
181
+ "CWE-20": {
182
+ "f1": 0.5721153846153846,
183
+ "precision": 0.7,
184
+ "recall": 0.483739837398374,
185
+ "threshold": 0.8750000000000003,
186
+ "support": 246
187
+ },
188
+ "CWE-22": {
189
+ "f1": 0.5,
190
+ "precision": 1.0,
191
+ "recall": 0.3333333333333333,
192
+ "threshold": 0.7750000000000002,
193
+ "support": 6
194
+ },
195
+ "CWE-78": {
196
+ "f1": 0.0,
197
+ "precision": 0.0,
198
+ "recall": 0.0,
199
+ "threshold": 0.7500000000000002,
200
+ "support": 7
201
+ },
202
+ "CWE-79": {
203
+ "f1": 0.5932203389830508,
204
+ "precision": 0.4861111111111111,
205
+ "recall": 0.7608695652173914,
206
+ "threshold": 0.5750000000000002,
207
+ "support": 46
208
+ },
209
+ "CWE-89": {
210
+ "f1": 0.7916666666666666,
211
+ "precision": 0.6846846846846847,
212
+ "recall": 0.9382716049382716,
213
+ "threshold": 0.5250000000000001,
214
+ "support": 81
215
+ },
216
+ "CWE-94": {
217
+ "f1": 0.7068273092369478,
218
+ "precision": 0.676923076923077,
219
+ "recall": 0.7394957983193278,
220
+ "threshold": 0.7750000000000002,
221
+ "support": 119
222
+ },
223
+ "CWE-119": {
224
+ "f1": 0.7054610564010743,
225
+ "precision": 0.6127527216174183,
226
+ "recall": 0.8312236286919831,
227
+ "threshold": 0.6000000000000002,
228
+ "support": 474
229
+ },
230
+ "CWE-125": {
231
+ "f1": 0.5844748858447488,
232
+ "precision": 0.5245901639344263,
233
+ "recall": 0.6597938144329897,
234
+ "threshold": 0.7000000000000003,
235
+ "support": 97
236
+ },
237
+ "CWE-190": {
238
+ "f1": 0.7704918032786885,
239
+ "precision": 0.9038461538461539,
240
+ "recall": 0.6714285714285714,
241
+ "threshold": 0.8750000000000003,
242
+ "support": 70
243
+ },
244
+ "CWE-200": {
245
+ "f1": 0.7046632124352331,
246
+ "precision": 0.6938775510204082,
247
+ "recall": 0.7157894736842105,
248
+ "threshold": 0.7500000000000002,
249
+ "support": 95
250
+ },
251
+ "CWE-264": {
252
+ "f1": 0.697986577181208,
253
+ "precision": 0.7027027027027027,
254
+ "recall": 0.6933333333333334,
255
+ "threshold": 0.8250000000000003,
256
+ "support": 75
257
+ },
258
+ "CWE-269": {
259
+ "f1": 0.0,
260
+ "precision": 0.0,
261
+ "recall": 0.0,
262
+ "threshold": 0.5500000000000003,
263
+ "support": 2
264
+ },
265
+ "CWE-276": {
266
+ "f1": 0.0,
267
+ "precision": 0.0,
268
+ "recall": 0.0,
269
+ "threshold": 0.5,
270
+ "support": 0
271
+ },
272
+ "CWE-284": {
273
+ "f1": 0.5128205128205128,
274
+ "precision": 0.43478260869565216,
275
+ "recall": 0.625,
276
+ "threshold": 0.6000000000000002,
277
+ "support": 16
278
+ },
279
+ "CWE-287": {
280
+ "f1": 0.25,
281
+ "precision": 0.25,
282
+ "recall": 0.25,
283
+ "threshold": 0.4250000000000001,
284
+ "support": 4
285
+ },
286
+ "CWE-310": {
287
+ "f1": 0.5,
288
+ "precision": 0.7142857142857143,
289
+ "recall": 0.38461538461538464,
290
+ "threshold": 0.8000000000000003,
291
+ "support": 13
292
+ },
293
+ "CWE-327": {
294
+ "f1": 0.0,
295
+ "precision": 0.0,
296
+ "recall": 0.0,
297
+ "threshold": 0.5,
298
+ "support": 2
299
+ },
300
+ "CWE-330": {
301
+ "f1": 0.5,
302
+ "precision": 0.6,
303
+ "recall": 0.42857142857142855,
304
+ "threshold": 0.7250000000000003,
305
+ "support": 7
306
+ },
307
+ "CWE-352": {
308
+ "f1": 0.0,
309
+ "precision": 0.0,
310
+ "recall": 0.0,
311
+ "threshold": 0.5,
312
+ "support": 1
313
+ },
314
+ "CWE-362": {
315
+ "f1": 0.7010309278350515,
316
+ "precision": 0.6538461538461539,
317
+ "recall": 0.7555555555555555,
318
+ "threshold": 0.6250000000000002,
319
+ "support": 45
320
+ },
321
+ "CWE-399": {
322
+ "f1": 0.6783625730994152,
323
+ "precision": 0.7837837837837838,
324
+ "recall": 0.5979381443298969,
325
+ "threshold": 0.9250000000000004,
326
+ "support": 97
327
+ },
328
+ "CWE-401": {
329
+ "f1": 0.5909090909090909,
330
+ "precision": 0.4482758620689655,
331
+ "recall": 0.8666666666666667,
332
+ "threshold": 0.4250000000000001,
333
+ "support": 15
334
+ },
335
+ "CWE-416": {
336
+ "f1": 0.48484848484848486,
337
+ "precision": 0.4266666666666667,
338
+ "recall": 0.5614035087719298,
339
+ "threshold": 0.6000000000000002,
340
+ "support": 57
341
+ },
342
+ "CWE-434": {
343
+ "f1": 0.0,
344
+ "precision": 0.0,
345
+ "recall": 0.0,
346
+ "threshold": 0.3000000000000001,
347
+ "support": 0
348
+ },
349
+ "CWE-476": {
350
+ "f1": 0.5483870967741935,
351
+ "precision": 0.5862068965517241,
352
+ "recall": 0.5151515151515151,
353
+ "threshold": 0.8000000000000003,
354
+ "support": 99
355
+ },
356
+ "CWE-502": {
357
+ "f1": 0.9,
358
+ "precision": 0.9,
359
+ "recall": 0.9,
360
+ "threshold": 0.8750000000000003,
361
+ "support": 40
362
+ },
363
+ "CWE-601": {
364
+ "f1": 0.8571428571428571,
365
+ "precision": 0.75,
366
+ "recall": 1.0,
367
+ "threshold": 0.47500000000000014,
368
+ "support": 3
369
+ },
370
+ "CWE-787": {
371
+ "f1": 0.47558386411889597,
372
+ "precision": 0.3708609271523179,
373
+ "recall": 0.6627218934911243,
374
+ "threshold": 0.6000000000000002,
375
+ "support": 169
376
+ },
377
+ "CWE-798": {
378
+ "f1": 0.14285714285714285,
379
+ "precision": 0.07692307692307693,
380
+ "recall": 1.0,
381
+ "threshold": 0.25000000000000006,
382
+ "support": 1
383
+ },
384
+ "CWE-918": {
385
+ "f1": 0.0,
386
+ "precision": 0.0,
387
+ "recall": 0.0,
388
+ "threshold": 0.5,
389
+ "support": 0
390
+ }
391
  },
392
+ "classification_report": " precision recall f1-score support\n\n safe 0.9762 0.9879 0.9820 15655\n CWE-20 0.7000 0.4837 0.5721 246\n CWE-22 1.0000 0.3333 0.5000 6\n CWE-78 0.0000 0.0000 0.0000 7\n CWE-79 0.4861 0.7609 0.5932 46\n CWE-89 0.6847 0.9383 0.7917 81\n CWE-94 0.6769 0.7395 0.7068 119\n CWE-119 0.6128 0.8312 0.7055 474\n CWE-125 0.5246 0.6598 0.5845 97\n CWE-190 0.9038 0.6714 0.7705 70\n CWE-200 0.6939 0.7158 0.7047 95\n CWE-264 0.7027 0.6933 0.6980 75\n CWE-269 0.0000 0.0000 0.0000 2\n CWE-276 0.0000 0.0000 0.0000 0\n CWE-284 0.4348 0.6250 0.5128 16\n CWE-287 0.2500 0.2500 0.2500 4\n CWE-310 0.7143 0.3846 0.5000 13\n CWE-327 0.0000 0.0000 0.0000 2\n CWE-330 0.6000 0.4286 0.5000 7\n CWE-352 0.0000 0.0000 0.0000 1\n CWE-362 0.6538 0.7556 0.7010 45\n CWE-399 0.7838 0.5979 0.6784 97\n CWE-401 0.4483 0.8667 0.5909 15\n CWE-416 0.4267 0.5614 0.4848 57\n CWE-434 0.0000 0.0000 0.0000 0\n CWE-476 0.5862 0.5152 0.5484 99\n CWE-502 0.9000 0.9000 0.9000 40\n CWE-601 0.7500 1.0000 0.8571 3\n CWE-787 0.3709 0.6627 0.4756 169\n CWE-798 0.0769 1.0000 0.1429 1\n CWE-918 0.0000 0.0000 0.0000 0\n\n micro avg 0.9297 0.9560 0.9426 17542\n macro avg 0.4825 0.5278 0.4758 17542\nweighted avg 0.9383 0.9560 0.9454 17542\n samples avg 0.9437 0.9560 0.9477 17542\n",
393
+ "improvements": [
394
+ "GraphCodeBERT-base (125M, 12 layers) vs CodeBERTa-small (83M, 6 layers)",
395
+ "Asymmetric Loss (ASL) gamma_neg=4, gamma_pos=0, clip=0.05",
396
+ "Two-phase training: 4 epochs frozen + up to 9 epochs full fine-tune",
397
+ "Per-class threshold optimization on validation set",
398
+ "Temperature scaling calibration (T=0.6163)",
399
+ "Classification head bias initialization for imbalanced classes"
400
+ ],
401
+ "baseline_comparison": {
402
+ "baseline_macro_f1": 0.1157,
403
+ "new_macro_f1": 0.47583485129569136,
404
+ "improvement_pct": 311.2660771786442
405
+ }
406
  }
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:7c2863b59f345fa82b44d21d25483aa4ddd11665213b069e302f9a815985add4
3
- size 333911404
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:27e6820a40e1881e3570e986e2fd0239019e209183c4707a3bb349f88523bdad
3
+ size 498702004
tokenizer.json CHANGED
The diff for this file is too large to render. See raw diff
 
tokenizer_config.json CHANGED
@@ -6,9 +6,7 @@
6
  "eos_token": "</s>",
7
  "errors": "replace",
8
  "is_local": false,
9
- "local_files_only": false,
10
  "mask_token": "<mask>",
11
- "max_len": 512,
12
  "model_max_length": 512,
13
  "pad_token": "<pad>",
14
  "sep_token": "</s>",
 
6
  "eos_token": "</s>",
7
  "errors": "replace",
8
  "is_local": false,
 
9
  "mask_token": "<mask>",
 
10
  "model_max_length": 512,
11
  "pad_token": "<pad>",
12
  "sep_token": "</s>",