Training in progress, epoch 1
Browse files
config.json
ADDED
|
@@ -0,0 +1,307 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "answerdotai/ModernBERT-large",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"ModernBertForSequenceClassification"
|
| 5 |
+
],
|
| 6 |
+
"attention_bias": false,
|
| 7 |
+
"attention_dropout": 0.0,
|
| 8 |
+
"bos_token_id": 50281,
|
| 9 |
+
"classifier_activation": "gelu",
|
| 10 |
+
"classifier_bias": false,
|
| 11 |
+
"classifier_dropout": 0.0,
|
| 12 |
+
"classifier_pooling": "mean",
|
| 13 |
+
"cls_token_id": 50281,
|
| 14 |
+
"decoder_bias": true,
|
| 15 |
+
"deterministic_flash_attn": false,
|
| 16 |
+
"embedding_dropout": 0.0,
|
| 17 |
+
"eos_token_id": 50282,
|
| 18 |
+
"global_attn_every_n_layers": 3,
|
| 19 |
+
"global_rope_theta": 160000.0,
|
| 20 |
+
"gradient_checkpointing": false,
|
| 21 |
+
"hidden_activation": "gelu",
|
| 22 |
+
"hidden_size": 1024,
|
| 23 |
+
"id2label": {
|
| 24 |
+
"0": "Aflatoxin",
|
| 25 |
+
"1": "abnormal smell",
|
| 26 |
+
"10": "bad smell / off odor",
|
| 27 |
+
"100": "soybeans and products thereof",
|
| 28 |
+
"101": "specified risk material (srm)",
|
| 29 |
+
"102": "spoilage",
|
| 30 |
+
"103": "staphylococcal enterotoxin",
|
| 31 |
+
"104": "staphylococcus",
|
| 32 |
+
"105": "stones",
|
| 33 |
+
"106": "sulfamethazine unauthorised",
|
| 34 |
+
"107": "sulphates/sulphites",
|
| 35 |
+
"108": "sulphur dioxide",
|
| 36 |
+
"109": "sulphur dioxide and sulphites",
|
| 37 |
+
"11": "bone fragment",
|
| 38 |
+
"110": "tampering",
|
| 39 |
+
"111": "taste disturbance",
|
| 40 |
+
"112": "too high content of fluorinated organic chemicals",
|
| 41 |
+
"113": "too high content of tetrahydrocannabinol (THC)",
|
| 42 |
+
"114": "toxin",
|
| 43 |
+
"115": "unauthorised colour sudan 4",
|
| 44 |
+
"116": "unauthorised import",
|
| 45 |
+
"117": "unauthorised operator",
|
| 46 |
+
"118": "unauthorised placing on the market",
|
| 47 |
+
"119": "unauthorised substance ethylene oxide",
|
| 48 |
+
"12": "brazil nut",
|
| 49 |
+
"120": "unauthorised substance sildenafil",
|
| 50 |
+
"121": "undeclared additive",
|
| 51 |
+
"122": "undeclared constituent",
|
| 52 |
+
"123": "unfit for human consumption",
|
| 53 |
+
"124": "vibrio spp",
|
| 54 |
+
"125": "virus",
|
| 55 |
+
"126": "walnut",
|
| 56 |
+
"127": "yeasts",
|
| 57 |
+
"13": "bulging packaging",
|
| 58 |
+
"14": "cashew",
|
| 59 |
+
"15": "celery and products thereof",
|
| 60 |
+
"16": "cereals",
|
| 61 |
+
"17": "cereals containing gluten and products thereof",
|
| 62 |
+
"18": "chemical",
|
| 63 |
+
"19": "chemical compound (high content)",
|
| 64 |
+
"2": "alcohol content",
|
| 65 |
+
"20": "chemical compound (unauthorised)",
|
| 66 |
+
"21": "chlorine",
|
| 67 |
+
"22": "clostridium botulinum",
|
| 68 |
+
"23": "coconut",
|
| 69 |
+
"24": "coliforms",
|
| 70 |
+
"25": "compositional deviation",
|
| 71 |
+
"26": "contaminant",
|
| 72 |
+
"27": "cronobacter spp",
|
| 73 |
+
"28": "crustaceans and products thereof",
|
| 74 |
+
"29": "cyclospora",
|
| 75 |
+
"3": "alkaloids",
|
| 76 |
+
"30": "dairy products",
|
| 77 |
+
"31": "damaged packaging",
|
| 78 |
+
"32": "dioxins",
|
| 79 |
+
"33": "e 425 - konjac unauthorised",
|
| 80 |
+
"34": "eggs and products thereof",
|
| 81 |
+
"35": "enteroviruses",
|
| 82 |
+
"36": "escherichia coli",
|
| 83 |
+
"37": "false marketing claims",
|
| 84 |
+
"38": "fish and products thereof",
|
| 85 |
+
"39": "foreign bodies",
|
| 86 |
+
"4": "allergens",
|
| 87 |
+
"40": "glass fragment",
|
| 88 |
+
"41": "haloxyfop",
|
| 89 |
+
"42": "hazelnut",
|
| 90 |
+
"43": "heavy metals",
|
| 91 |
+
"44": "high content of copper",
|
| 92 |
+
"45": "high content of cyanide",
|
| 93 |
+
"46": "high content of vitamins",
|
| 94 |
+
"47": "improper conditions",
|
| 95 |
+
"48": "improper packaging",
|
| 96 |
+
"49": "incorrect labeling",
|
| 97 |
+
"5": "almond",
|
| 98 |
+
"50": "incorrect use by dates",
|
| 99 |
+
"51": "insects",
|
| 100 |
+
"52": "inspection issues",
|
| 101 |
+
"53": "insufficient labelling/documentation",
|
| 102 |
+
"54": "labelling/misdescription",
|
| 103 |
+
"55": "listeria monocytogenes",
|
| 104 |
+
"56": "listeria spp",
|
| 105 |
+
"57": "metal fragment",
|
| 106 |
+
"58": "microbiological contamination",
|
| 107 |
+
"59": "milk and products thereof",
|
| 108 |
+
"6": "altered organoleptic characteristics",
|
| 109 |
+
"60": "misbranding",
|
| 110 |
+
"61": "misdescription",
|
| 111 |
+
"62": "mislabelled",
|
| 112 |
+
"63": "molluscs and products thereof",
|
| 113 |
+
"64": "moulds",
|
| 114 |
+
"65": "mustard and products thereof",
|
| 115 |
+
"66": "nitrogen",
|
| 116 |
+
"67": "non compliance with HACCP",
|
| 117 |
+
"68": "norovirus",
|
| 118 |
+
"69": "novel food",
|
| 119 |
+
"7": "amygdalin",
|
| 120 |
+
"70": "nuts",
|
| 121 |
+
"71": "ochratoxin",
|
| 122 |
+
"72": "organochlorine",
|
| 123 |
+
"73": "other",
|
| 124 |
+
"74": "other misbranding issues",
|
| 125 |
+
"75": "other not classified",
|
| 126 |
+
"76": "other not classified allergen hazards",
|
| 127 |
+
"77": "other not classified biological hazards",
|
| 128 |
+
"78": "other not classified chemical hazards",
|
| 129 |
+
"79": "other not classified hazards for foreign bodies",
|
| 130 |
+
"8": "antibiotics, vet drugs",
|
| 131 |
+
"80": "packaging defect",
|
| 132 |
+
"81": "pahs",
|
| 133 |
+
"82": "paralytic shellfish poisoning (psp) toxins",
|
| 134 |
+
"83": "pathogenic bacteria",
|
| 135 |
+
"84": "patulin",
|
| 136 |
+
"85": "peanuts and products thereof",
|
| 137 |
+
"86": "pecan nut",
|
| 138 |
+
"87": "phenylpyrazole",
|
| 139 |
+
"88": "pieces of wood",
|
| 140 |
+
"89": "pistachio nut",
|
| 141 |
+
"9": "bacillus spp.",
|
| 142 |
+
"90": "plastic fragment",
|
| 143 |
+
"91": "polycyclic aromatic hydrocarbons",
|
| 144 |
+
"92": "poor hygienic state",
|
| 145 |
+
"93": "poor or insufficient controls",
|
| 146 |
+
"94": "processing",
|
| 147 |
+
"95": "product category/characteristics",
|
| 148 |
+
"96": "pseudomonas spp.",
|
| 149 |
+
"97": "rubber fragments",
|
| 150 |
+
"98": "salmonella",
|
| 151 |
+
"99": "sesame seeds and products thereof"
|
| 152 |
+
},
|
| 153 |
+
"initializer_cutoff_factor": 2.0,
|
| 154 |
+
"initializer_range": 0.02,
|
| 155 |
+
"intermediate_size": 2624,
|
| 156 |
+
"label2id": {
|
| 157 |
+
"0": "0",
|
| 158 |
+
"1": "1",
|
| 159 |
+
"10": "10",
|
| 160 |
+
"100": "100",
|
| 161 |
+
"101": "101",
|
| 162 |
+
"102": "102",
|
| 163 |
+
"103": "103",
|
| 164 |
+
"104": "104",
|
| 165 |
+
"105": "105",
|
| 166 |
+
"106": "106",
|
| 167 |
+
"107": "107",
|
| 168 |
+
"108": "108",
|
| 169 |
+
"109": "109",
|
| 170 |
+
"11": "11",
|
| 171 |
+
"110": "110",
|
| 172 |
+
"111": "111",
|
| 173 |
+
"112": "112",
|
| 174 |
+
"113": "113",
|
| 175 |
+
"114": "114",
|
| 176 |
+
"115": "115",
|
| 177 |
+
"116": "116",
|
| 178 |
+
"117": "117",
|
| 179 |
+
"118": "118",
|
| 180 |
+
"119": "119",
|
| 181 |
+
"12": "12",
|
| 182 |
+
"120": "120",
|
| 183 |
+
"121": "121",
|
| 184 |
+
"122": "122",
|
| 185 |
+
"123": "123",
|
| 186 |
+
"124": "124",
|
| 187 |
+
"125": "125",
|
| 188 |
+
"126": "126",
|
| 189 |
+
"127": "127",
|
| 190 |
+
"13": "13",
|
| 191 |
+
"14": "14",
|
| 192 |
+
"15": "15",
|
| 193 |
+
"16": "16",
|
| 194 |
+
"17": "17",
|
| 195 |
+
"18": "18",
|
| 196 |
+
"19": "19",
|
| 197 |
+
"2": "2",
|
| 198 |
+
"20": "20",
|
| 199 |
+
"21": "21",
|
| 200 |
+
"22": "22",
|
| 201 |
+
"23": "23",
|
| 202 |
+
"24": "24",
|
| 203 |
+
"25": "25",
|
| 204 |
+
"26": "26",
|
| 205 |
+
"27": "27",
|
| 206 |
+
"28": "28",
|
| 207 |
+
"29": "29",
|
| 208 |
+
"3": "3",
|
| 209 |
+
"30": "30",
|
| 210 |
+
"31": "31",
|
| 211 |
+
"32": "32",
|
| 212 |
+
"33": "33",
|
| 213 |
+
"34": "34",
|
| 214 |
+
"35": "35",
|
| 215 |
+
"36": "36",
|
| 216 |
+
"37": "37",
|
| 217 |
+
"38": "38",
|
| 218 |
+
"39": "39",
|
| 219 |
+
"4": "4",
|
| 220 |
+
"40": "40",
|
| 221 |
+
"41": "41",
|
| 222 |
+
"42": "42",
|
| 223 |
+
"43": "43",
|
| 224 |
+
"44": "44",
|
| 225 |
+
"45": "45",
|
| 226 |
+
"46": "46",
|
| 227 |
+
"47": "47",
|
| 228 |
+
"48": "48",
|
| 229 |
+
"49": "49",
|
| 230 |
+
"5": "5",
|
| 231 |
+
"50": "50",
|
| 232 |
+
"51": "51",
|
| 233 |
+
"52": "52",
|
| 234 |
+
"53": "53",
|
| 235 |
+
"54": "54",
|
| 236 |
+
"55": "55",
|
| 237 |
+
"56": "56",
|
| 238 |
+
"57": "57",
|
| 239 |
+
"58": "58",
|
| 240 |
+
"59": "59",
|
| 241 |
+
"6": "6",
|
| 242 |
+
"60": "60",
|
| 243 |
+
"61": "61",
|
| 244 |
+
"62": "62",
|
| 245 |
+
"63": "63",
|
| 246 |
+
"64": "64",
|
| 247 |
+
"65": "65",
|
| 248 |
+
"66": "66",
|
| 249 |
+
"67": "67",
|
| 250 |
+
"68": "68",
|
| 251 |
+
"69": "69",
|
| 252 |
+
"7": "7",
|
| 253 |
+
"70": "70",
|
| 254 |
+
"71": "71",
|
| 255 |
+
"72": "72",
|
| 256 |
+
"73": "73",
|
| 257 |
+
"74": "74",
|
| 258 |
+
"75": "75",
|
| 259 |
+
"76": "76",
|
| 260 |
+
"77": "77",
|
| 261 |
+
"78": "78",
|
| 262 |
+
"79": "79",
|
| 263 |
+
"8": "8",
|
| 264 |
+
"80": "80",
|
| 265 |
+
"81": "81",
|
| 266 |
+
"82": "82",
|
| 267 |
+
"83": "83",
|
| 268 |
+
"84": "84",
|
| 269 |
+
"85": "85",
|
| 270 |
+
"86": "86",
|
| 271 |
+
"87": "87",
|
| 272 |
+
"88": "88",
|
| 273 |
+
"89": "89",
|
| 274 |
+
"9": "9",
|
| 275 |
+
"90": "90",
|
| 276 |
+
"91": "91",
|
| 277 |
+
"92": "92",
|
| 278 |
+
"93": "93",
|
| 279 |
+
"94": "94",
|
| 280 |
+
"95": "95",
|
| 281 |
+
"96": "96",
|
| 282 |
+
"97": "97",
|
| 283 |
+
"98": "98",
|
| 284 |
+
"99": "99"
|
| 285 |
+
},
|
| 286 |
+
"layer_norm_eps": 1e-05,
|
| 287 |
+
"local_attention": 128,
|
| 288 |
+
"local_rope_theta": 10000.0,
|
| 289 |
+
"max_position_embeddings": 8192,
|
| 290 |
+
"mlp_bias": false,
|
| 291 |
+
"mlp_dropout": 0.0,
|
| 292 |
+
"model_type": "modernbert",
|
| 293 |
+
"norm_bias": false,
|
| 294 |
+
"norm_eps": 1e-05,
|
| 295 |
+
"num_attention_heads": 16,
|
| 296 |
+
"num_hidden_layers": 28,
|
| 297 |
+
"pad_token_id": 50283,
|
| 298 |
+
"position_embedding_type": "absolute",
|
| 299 |
+
"problem_type": "single_label_classification",
|
| 300 |
+
"reference_compile": true,
|
| 301 |
+
"sep_token_id": 50282,
|
| 302 |
+
"sparse_pred_ignore_index": -100,
|
| 303 |
+
"sparse_prediction": false,
|
| 304 |
+
"torch_dtype": "float32",
|
| 305 |
+
"transformers_version": "4.48.0.dev0",
|
| 306 |
+
"vocab_size": 50368
|
| 307 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a2fbdd9932eeca825b42959ef4cab035a3202bfda5545c3c4d995dcda3877d32
|
| 3 |
+
size 1583868248
|
runs/Jan13_10-55-53_ben-iir/events.out.tfevents.1736765755.ben-iir.2241653.0
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:be2872e98d09510d3da63ed52fc345c40c51b66602df5feb6f1566e3411f30a6
|
| 3 |
+
size 12577
|
training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:924637a765b13958aa5929394a79758ce569e4b66dd0d3309ba09583a4b897fc
|
| 3 |
+
size 5368
|