oliat commited on
Commit
6b46ae3
1 Parent(s): 5f2a65f

uploading model files

Browse files
.gitattributes CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ tokenizer.json filter=lfs diff=lfs merge=lfs -text
config.json ADDED
@@ -0,0 +1,919 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "xlm-roberta-large",
3
+ "architectures": [
4
+ "XLMRobertaForTokenClassification"
5
+ ],
6
+ "attention_probs_dropout_prob": 0.1,
7
+ "bos_token_id": 0,
8
+ "classifier_dropout": null,
9
+ "eos_token_id": 2,
10
+ "finetuning_task": "ner",
11
+ "hidden_act": "gelu",
12
+ "hidden_dropout_prob": 0.1,
13
+ "hidden_size": 1024,
14
+ "id2label": {
15
+ "0": "LABEL_0",
16
+ "1": "LABEL_1",
17
+ "2": "LABEL_2",
18
+ "3": "LABEL_3",
19
+ "4": "LABEL_4",
20
+ "5": "LABEL_5",
21
+ "6": "LABEL_6",
22
+ "7": "LABEL_7",
23
+ "8": "LABEL_8",
24
+ "9": "LABEL_9",
25
+ "10": "LABEL_10",
26
+ "11": "LABEL_11",
27
+ "12": "LABEL_12",
28
+ "13": "LABEL_13",
29
+ "14": "LABEL_14",
30
+ "15": "LABEL_15",
31
+ "16": "LABEL_16",
32
+ "17": "LABEL_17",
33
+ "18": "LABEL_18",
34
+ "19": "LABEL_19",
35
+ "20": "LABEL_20",
36
+ "21": "LABEL_21",
37
+ "22": "LABEL_22",
38
+ "23": "LABEL_23",
39
+ "24": "LABEL_24",
40
+ "25": "LABEL_25",
41
+ "26": "LABEL_26",
42
+ "27": "LABEL_27",
43
+ "28": "LABEL_28",
44
+ "29": "LABEL_29",
45
+ "30": "LABEL_30",
46
+ "31": "LABEL_31",
47
+ "32": "LABEL_32",
48
+ "33": "LABEL_33",
49
+ "34": "LABEL_34",
50
+ "35": "LABEL_35",
51
+ "36": "LABEL_36",
52
+ "37": "LABEL_37",
53
+ "38": "LABEL_38",
54
+ "39": "LABEL_39",
55
+ "40": "LABEL_40",
56
+ "41": "LABEL_41",
57
+ "42": "LABEL_42",
58
+ "43": "LABEL_43",
59
+ "44": "LABEL_44",
60
+ "45": "LABEL_45",
61
+ "46": "LABEL_46",
62
+ "47": "LABEL_47",
63
+ "48": "LABEL_48",
64
+ "49": "LABEL_49",
65
+ "50": "LABEL_50",
66
+ "51": "LABEL_51",
67
+ "52": "LABEL_52",
68
+ "53": "LABEL_53",
69
+ "54": "LABEL_54",
70
+ "55": "LABEL_55",
71
+ "56": "LABEL_56",
72
+ "57": "LABEL_57",
73
+ "58": "LABEL_58",
74
+ "59": "LABEL_59",
75
+ "60": "LABEL_60",
76
+ "61": "LABEL_61",
77
+ "62": "LABEL_62",
78
+ "63": "LABEL_63",
79
+ "64": "LABEL_64",
80
+ "65": "LABEL_65",
81
+ "66": "LABEL_66",
82
+ "67": "LABEL_67",
83
+ "68": "LABEL_68",
84
+ "69": "LABEL_69",
85
+ "70": "LABEL_70",
86
+ "71": "LABEL_71",
87
+ "72": "LABEL_72",
88
+ "73": "LABEL_73",
89
+ "74": "LABEL_74",
90
+ "75": "LABEL_75",
91
+ "76": "LABEL_76",
92
+ "77": "LABEL_77",
93
+ "78": "LABEL_78",
94
+ "79": "LABEL_79",
95
+ "80": "LABEL_80",
96
+ "81": "LABEL_81",
97
+ "82": "LABEL_82",
98
+ "83": "LABEL_83",
99
+ "84": "LABEL_84",
100
+ "85": "LABEL_85",
101
+ "86": "LABEL_86",
102
+ "87": "LABEL_87",
103
+ "88": "LABEL_88",
104
+ "89": "LABEL_89",
105
+ "90": "LABEL_90",
106
+ "91": "LABEL_91",
107
+ "92": "LABEL_92",
108
+ "93": "LABEL_93",
109
+ "94": "LABEL_94",
110
+ "95": "LABEL_95",
111
+ "96": "LABEL_96",
112
+ "97": "LABEL_97",
113
+ "98": "LABEL_98",
114
+ "99": "LABEL_99",
115
+ "100": "LABEL_100",
116
+ "101": "LABEL_101",
117
+ "102": "LABEL_102",
118
+ "103": "LABEL_103",
119
+ "104": "LABEL_104",
120
+ "105": "LABEL_105",
121
+ "106": "LABEL_106",
122
+ "107": "LABEL_107",
123
+ "108": "LABEL_108",
124
+ "109": "LABEL_109",
125
+ "110": "LABEL_110",
126
+ "111": "LABEL_111",
127
+ "112": "LABEL_112",
128
+ "113": "LABEL_113",
129
+ "114": "LABEL_114",
130
+ "115": "LABEL_115",
131
+ "116": "LABEL_116",
132
+ "117": "LABEL_117",
133
+ "118": "LABEL_118",
134
+ "119": "LABEL_119",
135
+ "120": "LABEL_120",
136
+ "121": "LABEL_121",
137
+ "122": "LABEL_122",
138
+ "123": "LABEL_123",
139
+ "124": "LABEL_124",
140
+ "125": "LABEL_125",
141
+ "126": "LABEL_126",
142
+ "127": "LABEL_127",
143
+ "128": "LABEL_128",
144
+ "129": "LABEL_129",
145
+ "130": "LABEL_130",
146
+ "131": "LABEL_131",
147
+ "132": "LABEL_132",
148
+ "133": "LABEL_133",
149
+ "134": "LABEL_134",
150
+ "135": "LABEL_135",
151
+ "136": "LABEL_136",
152
+ "137": "LABEL_137",
153
+ "138": "LABEL_138",
154
+ "139": "LABEL_139",
155
+ "140": "LABEL_140",
156
+ "141": "LABEL_141",
157
+ "142": "LABEL_142",
158
+ "143": "LABEL_143",
159
+ "144": "LABEL_144",
160
+ "145": "LABEL_145",
161
+ "146": "LABEL_146",
162
+ "147": "LABEL_147",
163
+ "148": "LABEL_148",
164
+ "149": "LABEL_149",
165
+ "150": "LABEL_150",
166
+ "151": "LABEL_151",
167
+ "152": "LABEL_152",
168
+ "153": "LABEL_153",
169
+ "154": "LABEL_154",
170
+ "155": "LABEL_155",
171
+ "156": "LABEL_156",
172
+ "157": "LABEL_157",
173
+ "158": "LABEL_158",
174
+ "159": "LABEL_159",
175
+ "160": "LABEL_160",
176
+ "161": "LABEL_161",
177
+ "162": "LABEL_162",
178
+ "163": "LABEL_163",
179
+ "164": "LABEL_164",
180
+ "165": "LABEL_165",
181
+ "166": "LABEL_166",
182
+ "167": "LABEL_167",
183
+ "168": "LABEL_168",
184
+ "169": "LABEL_169",
185
+ "170": "LABEL_170",
186
+ "171": "LABEL_171",
187
+ "172": "LABEL_172",
188
+ "173": "LABEL_173",
189
+ "174": "LABEL_174",
190
+ "175": "LABEL_175",
191
+ "176": "LABEL_176",
192
+ "177": "LABEL_177",
193
+ "178": "LABEL_178",
194
+ "179": "LABEL_179",
195
+ "180": "LABEL_180",
196
+ "181": "LABEL_181",
197
+ "182": "LABEL_182",
198
+ "183": "LABEL_183",
199
+ "184": "LABEL_184",
200
+ "185": "LABEL_185",
201
+ "186": "LABEL_186",
202
+ "187": "LABEL_187",
203
+ "188": "LABEL_188",
204
+ "189": "LABEL_189",
205
+ "190": "LABEL_190",
206
+ "191": "LABEL_191",
207
+ "192": "LABEL_192",
208
+ "193": "LABEL_193",
209
+ "194": "LABEL_194",
210
+ "195": "LABEL_195",
211
+ "196": "LABEL_196",
212
+ "197": "LABEL_197",
213
+ "198": "LABEL_198",
214
+ "199": "LABEL_199",
215
+ "200": "LABEL_200",
216
+ "201": "LABEL_201",
217
+ "202": "LABEL_202",
218
+ "203": "LABEL_203",
219
+ "204": "LABEL_204",
220
+ "205": "LABEL_205",
221
+ "206": "LABEL_206",
222
+ "207": "LABEL_207",
223
+ "208": "LABEL_208",
224
+ "209": "LABEL_209",
225
+ "210": "LABEL_210",
226
+ "211": "LABEL_211",
227
+ "212": "LABEL_212",
228
+ "213": "LABEL_213",
229
+ "214": "LABEL_214",
230
+ "215": "LABEL_215",
231
+ "216": "LABEL_216",
232
+ "217": "LABEL_217",
233
+ "218": "LABEL_218",
234
+ "219": "LABEL_219",
235
+ "220": "LABEL_220",
236
+ "221": "LABEL_221",
237
+ "222": "LABEL_222",
238
+ "223": "LABEL_223",
239
+ "224": "LABEL_224",
240
+ "225": "LABEL_225",
241
+ "226": "LABEL_226",
242
+ "227": "LABEL_227",
243
+ "228": "LABEL_228",
244
+ "229": "LABEL_229",
245
+ "230": "LABEL_230",
246
+ "231": "LABEL_231",
247
+ "232": "LABEL_232",
248
+ "233": "LABEL_233",
249
+ "234": "LABEL_234",
250
+ "235": "LABEL_235",
251
+ "236": "LABEL_236",
252
+ "237": "LABEL_237",
253
+ "238": "LABEL_238",
254
+ "239": "LABEL_239",
255
+ "240": "LABEL_240",
256
+ "241": "LABEL_241",
257
+ "242": "LABEL_242",
258
+ "243": "LABEL_243",
259
+ "244": "LABEL_244",
260
+ "245": "LABEL_245",
261
+ "246": "LABEL_246",
262
+ "247": "LABEL_247",
263
+ "248": "LABEL_248",
264
+ "249": "LABEL_249",
265
+ "250": "LABEL_250",
266
+ "251": "LABEL_251",
267
+ "252": "LABEL_252",
268
+ "253": "LABEL_253",
269
+ "254": "LABEL_254",
270
+ "255": "LABEL_255",
271
+ "256": "LABEL_256",
272
+ "257": "LABEL_257",
273
+ "258": "LABEL_258",
274
+ "259": "LABEL_259",
275
+ "260": "LABEL_260",
276
+ "261": "LABEL_261",
277
+ "262": "LABEL_262",
278
+ "263": "LABEL_263",
279
+ "264": "LABEL_264",
280
+ "265": "LABEL_265",
281
+ "266": "LABEL_266",
282
+ "267": "LABEL_267",
283
+ "268": "LABEL_268",
284
+ "269": "LABEL_269",
285
+ "270": "LABEL_270",
286
+ "271": "LABEL_271",
287
+ "272": "LABEL_272",
288
+ "273": "LABEL_273",
289
+ "274": "LABEL_274",
290
+ "275": "LABEL_275",
291
+ "276": "LABEL_276",
292
+ "277": "LABEL_277",
293
+ "278": "LABEL_278",
294
+ "279": "LABEL_279",
295
+ "280": "LABEL_280",
296
+ "281": "LABEL_281",
297
+ "282": "LABEL_282",
298
+ "283": "LABEL_283",
299
+ "284": "LABEL_284",
300
+ "285": "LABEL_285",
301
+ "286": "LABEL_286",
302
+ "287": "LABEL_287",
303
+ "288": "LABEL_288",
304
+ "289": "LABEL_289",
305
+ "290": "LABEL_290",
306
+ "291": "LABEL_291",
307
+ "292": "LABEL_292",
308
+ "293": "LABEL_293",
309
+ "294": "LABEL_294",
310
+ "295": "LABEL_295",
311
+ "296": "LABEL_296",
312
+ "297": "LABEL_297",
313
+ "298": "LABEL_298",
314
+ "299": "LABEL_299",
315
+ "300": "LABEL_300",
316
+ "301": "LABEL_301",
317
+ "302": "LABEL_302",
318
+ "303": "LABEL_303",
319
+ "304": "LABEL_304",
320
+ "305": "LABEL_305",
321
+ "306": "LABEL_306",
322
+ "307": "LABEL_307",
323
+ "308": "LABEL_308",
324
+ "309": "LABEL_309",
325
+ "310": "LABEL_310",
326
+ "311": "LABEL_311",
327
+ "312": "LABEL_312",
328
+ "313": "LABEL_313",
329
+ "314": "LABEL_314",
330
+ "315": "LABEL_315",
331
+ "316": "LABEL_316",
332
+ "317": "LABEL_317",
333
+ "318": "LABEL_318",
334
+ "319": "LABEL_319",
335
+ "320": "LABEL_320",
336
+ "321": "LABEL_321",
337
+ "322": "LABEL_322",
338
+ "323": "LABEL_323",
339
+ "324": "LABEL_324",
340
+ "325": "LABEL_325",
341
+ "326": "LABEL_326",
342
+ "327": "LABEL_327",
343
+ "328": "LABEL_328",
344
+ "329": "LABEL_329",
345
+ "330": "LABEL_330",
346
+ "331": "LABEL_331",
347
+ "332": "LABEL_332",
348
+ "333": "LABEL_333",
349
+ "334": "LABEL_334",
350
+ "335": "LABEL_335",
351
+ "336": "LABEL_336",
352
+ "337": "LABEL_337",
353
+ "338": "LABEL_338",
354
+ "339": "LABEL_339",
355
+ "340": "LABEL_340",
356
+ "341": "LABEL_341",
357
+ "342": "LABEL_342",
358
+ "343": "LABEL_343",
359
+ "344": "LABEL_344",
360
+ "345": "LABEL_345",
361
+ "346": "LABEL_346",
362
+ "347": "LABEL_347",
363
+ "348": "LABEL_348",
364
+ "349": "LABEL_349",
365
+ "350": "LABEL_350",
366
+ "351": "LABEL_351",
367
+ "352": "LABEL_352",
368
+ "353": "LABEL_353",
369
+ "354": "LABEL_354",
370
+ "355": "LABEL_355",
371
+ "356": "LABEL_356",
372
+ "357": "LABEL_357",
373
+ "358": "LABEL_358",
374
+ "359": "LABEL_359",
375
+ "360": "LABEL_360",
376
+ "361": "LABEL_361",
377
+ "362": "LABEL_362",
378
+ "363": "LABEL_363",
379
+ "364": "LABEL_364",
380
+ "365": "LABEL_365",
381
+ "366": "LABEL_366",
382
+ "367": "LABEL_367",
383
+ "368": "LABEL_368",
384
+ "369": "LABEL_369",
385
+ "370": "LABEL_370",
386
+ "371": "LABEL_371",
387
+ "372": "LABEL_372",
388
+ "373": "LABEL_373",
389
+ "374": "LABEL_374",
390
+ "375": "LABEL_375",
391
+ "376": "LABEL_376",
392
+ "377": "LABEL_377",
393
+ "378": "LABEL_378",
394
+ "379": "LABEL_379",
395
+ "380": "LABEL_380",
396
+ "381": "LABEL_381",
397
+ "382": "LABEL_382",
398
+ "383": "LABEL_383",
399
+ "384": "LABEL_384",
400
+ "385": "LABEL_385",
401
+ "386": "LABEL_386",
402
+ "387": "LABEL_387",
403
+ "388": "LABEL_388",
404
+ "389": "LABEL_389",
405
+ "390": "LABEL_390",
406
+ "391": "LABEL_391",
407
+ "392": "LABEL_392",
408
+ "393": "LABEL_393",
409
+ "394": "LABEL_394",
410
+ "395": "LABEL_395",
411
+ "396": "LABEL_396",
412
+ "397": "LABEL_397",
413
+ "398": "LABEL_398",
414
+ "399": "LABEL_399",
415
+ "400": "LABEL_400",
416
+ "401": "LABEL_401",
417
+ "402": "LABEL_402",
418
+ "403": "LABEL_403",
419
+ "404": "LABEL_404",
420
+ "405": "LABEL_405",
421
+ "406": "LABEL_406",
422
+ "407": "LABEL_407",
423
+ "408": "LABEL_408",
424
+ "409": "LABEL_409",
425
+ "410": "LABEL_410",
426
+ "411": "LABEL_411",
427
+ "412": "LABEL_412",
428
+ "413": "LABEL_413",
429
+ "414": "LABEL_414",
430
+ "415": "LABEL_415",
431
+ "416": "LABEL_416",
432
+ "417": "LABEL_417",
433
+ "418": "LABEL_418",
434
+ "419": "LABEL_419",
435
+ "420": "LABEL_420",
436
+ "421": "LABEL_421",
437
+ "422": "LABEL_422",
438
+ "423": "LABEL_423",
439
+ "424": "LABEL_424",
440
+ "425": "LABEL_425",
441
+ "426": "LABEL_426",
442
+ "427": "LABEL_427",
443
+ "428": "LABEL_428",
444
+ "429": "LABEL_429",
445
+ "430": "LABEL_430",
446
+ "431": "LABEL_431",
447
+ "432": "LABEL_432",
448
+ "433": "LABEL_433",
449
+ "434": "LABEL_434",
450
+ "435": "LABEL_435",
451
+ "436": "LABEL_436",
452
+ "437": "LABEL_437",
453
+ "438": "LABEL_438",
454
+ "439": "LABEL_439",
455
+ "440": "LABEL_440",
456
+ "441": "LABEL_441",
457
+ "442": "LABEL_442"
458
+ },
459
+ "initializer_range": 0.02,
460
+ "intermediate_size": 4096,
461
+ "label2id": {
462
+ "LABEL_0": 0,
463
+ "LABEL_1": 1,
464
+ "LABEL_10": 10,
465
+ "LABEL_100": 100,
466
+ "LABEL_101": 101,
467
+ "LABEL_102": 102,
468
+ "LABEL_103": 103,
469
+ "LABEL_104": 104,
470
+ "LABEL_105": 105,
471
+ "LABEL_106": 106,
472
+ "LABEL_107": 107,
473
+ "LABEL_108": 108,
474
+ "LABEL_109": 109,
475
+ "LABEL_11": 11,
476
+ "LABEL_110": 110,
477
+ "LABEL_111": 111,
478
+ "LABEL_112": 112,
479
+ "LABEL_113": 113,
480
+ "LABEL_114": 114,
481
+ "LABEL_115": 115,
482
+ "LABEL_116": 116,
483
+ "LABEL_117": 117,
484
+ "LABEL_118": 118,
485
+ "LABEL_119": 119,
486
+ "LABEL_12": 12,
487
+ "LABEL_120": 120,
488
+ "LABEL_121": 121,
489
+ "LABEL_122": 122,
490
+ "LABEL_123": 123,
491
+ "LABEL_124": 124,
492
+ "LABEL_125": 125,
493
+ "LABEL_126": 126,
494
+ "LABEL_127": 127,
495
+ "LABEL_128": 128,
496
+ "LABEL_129": 129,
497
+ "LABEL_13": 13,
498
+ "LABEL_130": 130,
499
+ "LABEL_131": 131,
500
+ "LABEL_132": 132,
501
+ "LABEL_133": 133,
502
+ "LABEL_134": 134,
503
+ "LABEL_135": 135,
504
+ "LABEL_136": 136,
505
+ "LABEL_137": 137,
506
+ "LABEL_138": 138,
507
+ "LABEL_139": 139,
508
+ "LABEL_14": 14,
509
+ "LABEL_140": 140,
510
+ "LABEL_141": 141,
511
+ "LABEL_142": 142,
512
+ "LABEL_143": 143,
513
+ "LABEL_144": 144,
514
+ "LABEL_145": 145,
515
+ "LABEL_146": 146,
516
+ "LABEL_147": 147,
517
+ "LABEL_148": 148,
518
+ "LABEL_149": 149,
519
+ "LABEL_15": 15,
520
+ "LABEL_150": 150,
521
+ "LABEL_151": 151,
522
+ "LABEL_152": 152,
523
+ "LABEL_153": 153,
524
+ "LABEL_154": 154,
525
+ "LABEL_155": 155,
526
+ "LABEL_156": 156,
527
+ "LABEL_157": 157,
528
+ "LABEL_158": 158,
529
+ "LABEL_159": 159,
530
+ "LABEL_16": 16,
531
+ "LABEL_160": 160,
532
+ "LABEL_161": 161,
533
+ "LABEL_162": 162,
534
+ "LABEL_163": 163,
535
+ "LABEL_164": 164,
536
+ "LABEL_165": 165,
537
+ "LABEL_166": 166,
538
+ "LABEL_167": 167,
539
+ "LABEL_168": 168,
540
+ "LABEL_169": 169,
541
+ "LABEL_17": 17,
542
+ "LABEL_170": 170,
543
+ "LABEL_171": 171,
544
+ "LABEL_172": 172,
545
+ "LABEL_173": 173,
546
+ "LABEL_174": 174,
547
+ "LABEL_175": 175,
548
+ "LABEL_176": 176,
549
+ "LABEL_177": 177,
550
+ "LABEL_178": 178,
551
+ "LABEL_179": 179,
552
+ "LABEL_18": 18,
553
+ "LABEL_180": 180,
554
+ "LABEL_181": 181,
555
+ "LABEL_182": 182,
556
+ "LABEL_183": 183,
557
+ "LABEL_184": 184,
558
+ "LABEL_185": 185,
559
+ "LABEL_186": 186,
560
+ "LABEL_187": 187,
561
+ "LABEL_188": 188,
562
+ "LABEL_189": 189,
563
+ "LABEL_19": 19,
564
+ "LABEL_190": 190,
565
+ "LABEL_191": 191,
566
+ "LABEL_192": 192,
567
+ "LABEL_193": 193,
568
+ "LABEL_194": 194,
569
+ "LABEL_195": 195,
570
+ "LABEL_196": 196,
571
+ "LABEL_197": 197,
572
+ "LABEL_198": 198,
573
+ "LABEL_199": 199,
574
+ "LABEL_2": 2,
575
+ "LABEL_20": 20,
576
+ "LABEL_200": 200,
577
+ "LABEL_201": 201,
578
+ "LABEL_202": 202,
579
+ "LABEL_203": 203,
580
+ "LABEL_204": 204,
581
+ "LABEL_205": 205,
582
+ "LABEL_206": 206,
583
+ "LABEL_207": 207,
584
+ "LABEL_208": 208,
585
+ "LABEL_209": 209,
586
+ "LABEL_21": 21,
587
+ "LABEL_210": 210,
588
+ "LABEL_211": 211,
589
+ "LABEL_212": 212,
590
+ "LABEL_213": 213,
591
+ "LABEL_214": 214,
592
+ "LABEL_215": 215,
593
+ "LABEL_216": 216,
594
+ "LABEL_217": 217,
595
+ "LABEL_218": 218,
596
+ "LABEL_219": 219,
597
+ "LABEL_22": 22,
598
+ "LABEL_220": 220,
599
+ "LABEL_221": 221,
600
+ "LABEL_222": 222,
601
+ "LABEL_223": 223,
602
+ "LABEL_224": 224,
603
+ "LABEL_225": 225,
604
+ "LABEL_226": 226,
605
+ "LABEL_227": 227,
606
+ "LABEL_228": 228,
607
+ "LABEL_229": 229,
608
+ "LABEL_23": 23,
609
+ "LABEL_230": 230,
610
+ "LABEL_231": 231,
611
+ "LABEL_232": 232,
612
+ "LABEL_233": 233,
613
+ "LABEL_234": 234,
614
+ "LABEL_235": 235,
615
+ "LABEL_236": 236,
616
+ "LABEL_237": 237,
617
+ "LABEL_238": 238,
618
+ "LABEL_239": 239,
619
+ "LABEL_24": 24,
620
+ "LABEL_240": 240,
621
+ "LABEL_241": 241,
622
+ "LABEL_242": 242,
623
+ "LABEL_243": 243,
624
+ "LABEL_244": 244,
625
+ "LABEL_245": 245,
626
+ "LABEL_246": 246,
627
+ "LABEL_247": 247,
628
+ "LABEL_248": 248,
629
+ "LABEL_249": 249,
630
+ "LABEL_25": 25,
631
+ "LABEL_250": 250,
632
+ "LABEL_251": 251,
633
+ "LABEL_252": 252,
634
+ "LABEL_253": 253,
635
+ "LABEL_254": 254,
636
+ "LABEL_255": 255,
637
+ "LABEL_256": 256,
638
+ "LABEL_257": 257,
639
+ "LABEL_258": 258,
640
+ "LABEL_259": 259,
641
+ "LABEL_26": 26,
642
+ "LABEL_260": 260,
643
+ "LABEL_261": 261,
644
+ "LABEL_262": 262,
645
+ "LABEL_263": 263,
646
+ "LABEL_264": 264,
647
+ "LABEL_265": 265,
648
+ "LABEL_266": 266,
649
+ "LABEL_267": 267,
650
+ "LABEL_268": 268,
651
+ "LABEL_269": 269,
652
+ "LABEL_27": 27,
653
+ "LABEL_270": 270,
654
+ "LABEL_271": 271,
655
+ "LABEL_272": 272,
656
+ "LABEL_273": 273,
657
+ "LABEL_274": 274,
658
+ "LABEL_275": 275,
659
+ "LABEL_276": 276,
660
+ "LABEL_277": 277,
661
+ "LABEL_278": 278,
662
+ "LABEL_279": 279,
663
+ "LABEL_28": 28,
664
+ "LABEL_280": 280,
665
+ "LABEL_281": 281,
666
+ "LABEL_282": 282,
667
+ "LABEL_283": 283,
668
+ "LABEL_284": 284,
669
+ "LABEL_285": 285,
670
+ "LABEL_286": 286,
671
+ "LABEL_287": 287,
672
+ "LABEL_288": 288,
673
+ "LABEL_289": 289,
674
+ "LABEL_29": 29,
675
+ "LABEL_290": 290,
676
+ "LABEL_291": 291,
677
+ "LABEL_292": 292,
678
+ "LABEL_293": 293,
679
+ "LABEL_294": 294,
680
+ "LABEL_295": 295,
681
+ "LABEL_296": 296,
682
+ "LABEL_297": 297,
683
+ "LABEL_298": 298,
684
+ "LABEL_299": 299,
685
+ "LABEL_3": 3,
686
+ "LABEL_30": 30,
687
+ "LABEL_300": 300,
688
+ "LABEL_301": 301,
689
+ "LABEL_302": 302,
690
+ "LABEL_303": 303,
691
+ "LABEL_304": 304,
692
+ "LABEL_305": 305,
693
+ "LABEL_306": 306,
694
+ "LABEL_307": 307,
695
+ "LABEL_308": 308,
696
+ "LABEL_309": 309,
697
+ "LABEL_31": 31,
698
+ "LABEL_310": 310,
699
+ "LABEL_311": 311,
700
+ "LABEL_312": 312,
701
+ "LABEL_313": 313,
702
+ "LABEL_314": 314,
703
+ "LABEL_315": 315,
704
+ "LABEL_316": 316,
705
+ "LABEL_317": 317,
706
+ "LABEL_318": 318,
707
+ "LABEL_319": 319,
708
+ "LABEL_32": 32,
709
+ "LABEL_320": 320,
710
+ "LABEL_321": 321,
711
+ "LABEL_322": 322,
712
+ "LABEL_323": 323,
713
+ "LABEL_324": 324,
714
+ "LABEL_325": 325,
715
+ "LABEL_326": 326,
716
+ "LABEL_327": 327,
717
+ "LABEL_328": 328,
718
+ "LABEL_329": 329,
719
+ "LABEL_33": 33,
720
+ "LABEL_330": 330,
721
+ "LABEL_331": 331,
722
+ "LABEL_332": 332,
723
+ "LABEL_333": 333,
724
+ "LABEL_334": 334,
725
+ "LABEL_335": 335,
726
+ "LABEL_336": 336,
727
+ "LABEL_337": 337,
728
+ "LABEL_338": 338,
729
+ "LABEL_339": 339,
730
+ "LABEL_34": 34,
731
+ "LABEL_340": 340,
732
+ "LABEL_341": 341,
733
+ "LABEL_342": 342,
734
+ "LABEL_343": 343,
735
+ "LABEL_344": 344,
736
+ "LABEL_345": 345,
737
+ "LABEL_346": 346,
738
+ "LABEL_347": 347,
739
+ "LABEL_348": 348,
740
+ "LABEL_349": 349,
741
+ "LABEL_35": 35,
742
+ "LABEL_350": 350,
743
+ "LABEL_351": 351,
744
+ "LABEL_352": 352,
745
+ "LABEL_353": 353,
746
+ "LABEL_354": 354,
747
+ "LABEL_355": 355,
748
+ "LABEL_356": 356,
749
+ "LABEL_357": 357,
750
+ "LABEL_358": 358,
751
+ "LABEL_359": 359,
752
+ "LABEL_36": 36,
753
+ "LABEL_360": 360,
754
+ "LABEL_361": 361,
755
+ "LABEL_362": 362,
756
+ "LABEL_363": 363,
757
+ "LABEL_364": 364,
758
+ "LABEL_365": 365,
759
+ "LABEL_366": 366,
760
+ "LABEL_367": 367,
761
+ "LABEL_368": 368,
762
+ "LABEL_369": 369,
763
+ "LABEL_37": 37,
764
+ "LABEL_370": 370,
765
+ "LABEL_371": 371,
766
+ "LABEL_372": 372,
767
+ "LABEL_373": 373,
768
+ "LABEL_374": 374,
769
+ "LABEL_375": 375,
770
+ "LABEL_376": 376,
771
+ "LABEL_377": 377,
772
+ "LABEL_378": 378,
773
+ "LABEL_379": 379,
774
+ "LABEL_38": 38,
775
+ "LABEL_380": 380,
776
+ "LABEL_381": 381,
777
+ "LABEL_382": 382,
778
+ "LABEL_383": 383,
779
+ "LABEL_384": 384,
780
+ "LABEL_385": 385,
781
+ "LABEL_386": 386,
782
+ "LABEL_387": 387,
783
+ "LABEL_388": 388,
784
+ "LABEL_389": 389,
785
+ "LABEL_39": 39,
786
+ "LABEL_390": 390,
787
+ "LABEL_391": 391,
788
+ "LABEL_392": 392,
789
+ "LABEL_393": 393,
790
+ "LABEL_394": 394,
791
+ "LABEL_395": 395,
792
+ "LABEL_396": 396,
793
+ "LABEL_397": 397,
794
+ "LABEL_398": 398,
795
+ "LABEL_399": 399,
796
+ "LABEL_4": 4,
797
+ "LABEL_40": 40,
798
+ "LABEL_400": 400,
799
+ "LABEL_401": 401,
800
+ "LABEL_402": 402,
801
+ "LABEL_403": 403,
802
+ "LABEL_404": 404,
803
+ "LABEL_405": 405,
804
+ "LABEL_406": 406,
805
+ "LABEL_407": 407,
806
+ "LABEL_408": 408,
807
+ "LABEL_409": 409,
808
+ "LABEL_41": 41,
809
+ "LABEL_410": 410,
810
+ "LABEL_411": 411,
811
+ "LABEL_412": 412,
812
+ "LABEL_413": 413,
813
+ "LABEL_414": 414,
814
+ "LABEL_415": 415,
815
+ "LABEL_416": 416,
816
+ "LABEL_417": 417,
817
+ "LABEL_418": 418,
818
+ "LABEL_419": 419,
819
+ "LABEL_42": 42,
820
+ "LABEL_420": 420,
821
+ "LABEL_421": 421,
822
+ "LABEL_422": 422,
823
+ "LABEL_423": 423,
824
+ "LABEL_424": 424,
825
+ "LABEL_425": 425,
826
+ "LABEL_426": 426,
827
+ "LABEL_427": 427,
828
+ "LABEL_428": 428,
829
+ "LABEL_429": 429,
830
+ "LABEL_43": 43,
831
+ "LABEL_430": 430,
832
+ "LABEL_431": 431,
833
+ "LABEL_432": 432,
834
+ "LABEL_433": 433,
835
+ "LABEL_434": 434,
836
+ "LABEL_435": 435,
837
+ "LABEL_436": 436,
838
+ "LABEL_437": 437,
839
+ "LABEL_438": 438,
840
+ "LABEL_439": 439,
841
+ "LABEL_44": 44,
842
+ "LABEL_440": 440,
843
+ "LABEL_441": 441,
844
+ "LABEL_442": 442,
845
+ "LABEL_45": 45,
846
+ "LABEL_46": 46,
847
+ "LABEL_47": 47,
848
+ "LABEL_48": 48,
849
+ "LABEL_49": 49,
850
+ "LABEL_5": 5,
851
+ "LABEL_50": 50,
852
+ "LABEL_51": 51,
853
+ "LABEL_52": 52,
854
+ "LABEL_53": 53,
855
+ "LABEL_54": 54,
856
+ "LABEL_55": 55,
857
+ "LABEL_56": 56,
858
+ "LABEL_57": 57,
859
+ "LABEL_58": 58,
860
+ "LABEL_59": 59,
861
+ "LABEL_6": 6,
862
+ "LABEL_60": 60,
863
+ "LABEL_61": 61,
864
+ "LABEL_62": 62,
865
+ "LABEL_63": 63,
866
+ "LABEL_64": 64,
867
+ "LABEL_65": 65,
868
+ "LABEL_66": 66,
869
+ "LABEL_67": 67,
870
+ "LABEL_68": 68,
871
+ "LABEL_69": 69,
872
+ "LABEL_7": 7,
873
+ "LABEL_70": 70,
874
+ "LABEL_71": 71,
875
+ "LABEL_72": 72,
876
+ "LABEL_73": 73,
877
+ "LABEL_74": 74,
878
+ "LABEL_75": 75,
879
+ "LABEL_76": 76,
880
+ "LABEL_77": 77,
881
+ "LABEL_78": 78,
882
+ "LABEL_79": 79,
883
+ "LABEL_8": 8,
884
+ "LABEL_80": 80,
885
+ "LABEL_81": 81,
886
+ "LABEL_82": 82,
887
+ "LABEL_83": 83,
888
+ "LABEL_84": 84,
889
+ "LABEL_85": 85,
890
+ "LABEL_86": 86,
891
+ "LABEL_87": 87,
892
+ "LABEL_88": 88,
893
+ "LABEL_89": 89,
894
+ "LABEL_9": 9,
895
+ "LABEL_90": 90,
896
+ "LABEL_91": 91,
897
+ "LABEL_92": 92,
898
+ "LABEL_93": 93,
899
+ "LABEL_94": 94,
900
+ "LABEL_95": 95,
901
+ "LABEL_96": 96,
902
+ "LABEL_97": 97,
903
+ "LABEL_98": 98,
904
+ "LABEL_99": 99
905
+ },
906
+ "layer_norm_eps": 1e-05,
907
+ "max_position_embeddings": 514,
908
+ "model_type": "xlm-roberta",
909
+ "num_attention_heads": 16,
910
+ "num_hidden_layers": 24,
911
+ "output_past": true,
912
+ "pad_token_id": 1,
913
+ "position_embedding_type": "absolute",
914
+ "torch_dtype": "float32",
915
+ "transformers_version": "4.20.0",
916
+ "type_vocab_size": 1,
917
+ "use_cache": true,
918
+ "vocab_size": 250002
919
+ }
pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:56632af50a13edb9377e20d4f8aeba63bd1e43bad46c488cee04dd61c7498408
3
+ size 2237319402
special_tokens_map.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": "<s>",
3
+ "cls_token": "<s>",
4
+ "eos_token": "</s>",
5
+ "mask_token": {
6
+ "content": "<mask>",
7
+ "lstrip": true,
8
+ "normalized": false,
9
+ "rstrip": false,
10
+ "single_word": false
11
+ },
12
+ "pad_token": "<pad>",
13
+ "sep_token": "</s>",
14
+ "unk_token": "<unk>"
15
+ }
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f2c509a525eb51aebb33fb59c24ee923c1d4c1db23c3ae81fe05ccf354084f7b
3
+ size 17082758
tokenizer_config.json ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": true,
3
+ "bos_token": "<s>",
4
+ "cls_token": "<s>",
5
+ "eos_token": "</s>",
6
+ "mask_token": {
7
+ "__type": "AddedToken",
8
+ "content": "<mask>",
9
+ "lstrip": true,
10
+ "normalized": true,
11
+ "rstrip": false,
12
+ "single_word": false
13
+ },
14
+ "model_max_length": 512,
15
+ "name_or_path": "xlm-roberta-large",
16
+ "pad_token": "<pad>",
17
+ "sep_token": "</s>",
18
+ "special_tokens_map_file": null,
19
+ "tokenizer_class": "XLMRobertaTokenizer",
20
+ "unk_token": "<unk>"
21
+ }
trainer_state.json ADDED
@@ -0,0 +1,385 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 0.9590698631546456,
3
+ "best_model_checkpoint": ".//debugged_es_gsd_ses_udpipe_8_0.1_0.00002_20_04-22-24_22-38/checkpoint-32020",
4
+ "epoch": 19.99968779269435,
5
+ "global_step": 32020,
6
+ "is_hyper_param_search": false,
7
+ "is_local_process_zero": true,
8
+ "is_world_process_zero": true,
9
+ "log_history": [
10
+ {
11
+ "epoch": 1.0,
12
+ "learning_rate": 1.6659729448491154e-05,
13
+ "loss": 0.9658,
14
+ "step": 1601
15
+ },
16
+ {
17
+ "epoch": 1.0,
18
+ "eval_accuracy": 0.9632624280173934,
19
+ "eval_f1": 0.9191161314259801,
20
+ "eval_loss": 0.153659850358963,
21
+ "eval_precision": 0.9176029962546817,
22
+ "eval_recall": 0.9206342651856134,
23
+ "eval_runtime": 12.5305,
24
+ "eval_samples_per_second": 127.848,
25
+ "eval_steps_per_second": 16.041,
26
+ "step": 1601
27
+ },
28
+ {
29
+ "epoch": 2.0,
30
+ "learning_rate": 1.9149445145856868e-05,
31
+ "loss": 0.118,
32
+ "step": 3202
33
+ },
34
+ {
35
+ "epoch": 2.0,
36
+ "eval_accuracy": 0.9749441767540251,
37
+ "eval_f1": 0.9426544669476715,
38
+ "eval_loss": 0.09417378902435303,
39
+ "eval_precision": 0.9387746744205094,
40
+ "eval_recall": 0.9465664615765784,
41
+ "eval_runtime": 12.5537,
42
+ "eval_samples_per_second": 127.612,
43
+ "eval_steps_per_second": 16.011,
44
+ "step": 3202
45
+ },
46
+ {
47
+ "epoch": 3.0,
48
+ "learning_rate": 1.8085587082198154e-05,
49
+ "loss": 0.0727,
50
+ "step": 4803
51
+ },
52
+ {
53
+ "epoch": 3.0,
54
+ "eval_accuracy": 0.9784228463979316,
55
+ "eval_f1": 0.9501459044017919,
56
+ "eval_loss": 0.07907041162252426,
57
+ "eval_precision": 0.9457148701165883,
58
+ "eval_recall": 0.9546186563158112,
59
+ "eval_runtime": 12.5508,
60
+ "eval_samples_per_second": 127.642,
61
+ "eval_steps_per_second": 16.015,
62
+ "step": 4803
63
+ },
64
+ {
65
+ "epoch": 4.0,
66
+ "learning_rate": 1.702172901853944e-05,
67
+ "loss": 0.0479,
68
+ "step": 6404
69
+ },
70
+ {
71
+ "epoch": 4.0,
72
+ "eval_accuracy": 0.9803267128922317,
73
+ "eval_f1": 0.9543441998393706,
74
+ "eval_loss": 0.07787933200597763,
75
+ "eval_precision": 0.9518938460274423,
76
+ "eval_recall": 0.9568072015526283,
77
+ "eval_runtime": 12.5524,
78
+ "eval_samples_per_second": 127.625,
79
+ "eval_steps_per_second": 16.013,
80
+ "step": 6404
81
+ },
82
+ {
83
+ "epoch": 5.0,
84
+ "learning_rate": 1.5957870954880726e-05,
85
+ "loss": 0.0342,
86
+ "step": 8005
87
+ },
88
+ {
89
+ "epoch": 5.0,
90
+ "eval_accuracy": 0.9791984957104243,
91
+ "eval_f1": 0.951995205720072,
92
+ "eval_loss": 0.08329462260007858,
93
+ "eval_precision": 0.9528418962521717,
94
+ "eval_recall": 0.9511500185819879,
95
+ "eval_runtime": 12.5712,
96
+ "eval_samples_per_second": 127.435,
97
+ "eval_steps_per_second": 15.989,
98
+ "step": 8005
99
+ },
100
+ {
101
+ "epoch": 6.0,
102
+ "learning_rate": 1.489401289122201e-05,
103
+ "loss": 0.0233,
104
+ "step": 9606
105
+ },
106
+ {
107
+ "epoch": 6.0,
108
+ "eval_accuracy": 0.9811023622047244,
109
+ "eval_f1": 0.956315778633658,
110
+ "eval_loss": 0.08429370075464249,
111
+ "eval_precision": 0.9539776462853385,
112
+ "eval_recall": 0.9586654003386051,
113
+ "eval_runtime": 12.5589,
114
+ "eval_samples_per_second": 127.559,
115
+ "eval_steps_per_second": 16.005,
116
+ "step": 9606
117
+ },
118
+ {
119
+ "epoch": 7.0,
120
+ "learning_rate": 1.3830154827563293e-05,
121
+ "loss": 0.0168,
122
+ "step": 11207
123
+ },
124
+ {
125
+ "epoch": 7.0,
126
+ "eval_accuracy": 0.9807262898107886,
127
+ "eval_f1": 0.9542881840821619,
128
+ "eval_loss": 0.09856697916984558,
129
+ "eval_precision": 0.9512925728354534,
130
+ "eval_recall": 0.9573027212288888,
131
+ "eval_runtime": 12.7018,
132
+ "eval_samples_per_second": 126.124,
133
+ "eval_steps_per_second": 15.825,
134
+ "step": 11207
135
+ },
136
+ {
137
+ "epoch": 8.0,
138
+ "learning_rate": 1.276629676390458e-05,
139
+ "loss": 0.0121,
140
+ "step": 12808
141
+ },
142
+ {
143
+ "epoch": 8.0,
144
+ "eval_accuracy": 0.9815254436479022,
145
+ "eval_f1": 0.9566986580077682,
146
+ "eval_loss": 0.10770849883556366,
147
+ "eval_precision": 0.9522952295229523,
148
+ "eval_recall": 0.9611429987199075,
149
+ "eval_runtime": 12.5218,
150
+ "eval_samples_per_second": 127.937,
151
+ "eval_steps_per_second": 16.052,
152
+ "step": 12808
153
+ },
154
+ {
155
+ "epoch": 9.0,
156
+ "learning_rate": 1.1702438700245863e-05,
157
+ "loss": 0.0102,
158
+ "step": 14409
159
+ },
160
+ {
161
+ "epoch": 9.0,
162
+ "eval_accuracy": 0.9800916676460218,
163
+ "eval_f1": 0.9545660129106535,
164
+ "eval_loss": 0.12106840312480927,
165
+ "eval_precision": 0.9505015353121802,
166
+ "eval_recall": 0.9586654003386051,
167
+ "eval_runtime": 12.5371,
168
+ "eval_samples_per_second": 127.781,
169
+ "eval_steps_per_second": 16.032,
170
+ "step": 14409
171
+ },
172
+ {
173
+ "epoch": 10.0,
174
+ "learning_rate": 1.063858063658715e-05,
175
+ "loss": 0.0073,
176
+ "step": 16010
177
+ },
178
+ {
179
+ "epoch": 10.0,
180
+ "eval_accuracy": 0.9807732988600305,
181
+ "eval_f1": 0.9554837514663813,
182
+ "eval_loss": 0.12773701548576355,
183
+ "eval_precision": 0.9524454291810274,
184
+ "eval_recall": 0.95854152041954,
185
+ "eval_runtime": 12.7247,
186
+ "eval_samples_per_second": 125.897,
187
+ "eval_steps_per_second": 15.796,
188
+ "step": 16010
189
+ },
190
+ {
191
+ "epoch": 11.0,
192
+ "learning_rate": 9.574722572928434e-06,
193
+ "loss": 0.0068,
194
+ "step": 17611
195
+ },
196
+ {
197
+ "epoch": 11.0,
198
+ "eval_accuracy": 0.9810083441062405,
199
+ "eval_f1": 0.9555354647817867,
200
+ "eval_loss": 0.13228633999824524,
201
+ "eval_precision": 0.9509631507913787,
202
+ "eval_recall": 0.9601519593673865,
203
+ "eval_runtime": 12.7411,
204
+ "eval_samples_per_second": 125.735,
205
+ "eval_steps_per_second": 15.776,
206
+ "step": 17611
207
+ },
208
+ {
209
+ "epoch": 12.0,
210
+ "learning_rate": 8.51086450926972e-06,
211
+ "loss": 0.0051,
212
+ "step": 19212
213
+ },
214
+ {
215
+ "epoch": 12.0,
216
+ "eval_accuracy": 0.9815724526971442,
217
+ "eval_f1": 0.95762694407251,
218
+ "eval_loss": 0.12687553465366364,
219
+ "eval_precision": 0.9554422887208155,
220
+ "eval_recall": 0.9598216129165462,
221
+ "eval_runtime": 12.7033,
222
+ "eval_samples_per_second": 126.109,
223
+ "eval_steps_per_second": 15.823,
224
+ "step": 19212
225
+ },
226
+ {
227
+ "epoch": 13.0,
228
+ "learning_rate": 7.447006445611005e-06,
229
+ "loss": 0.0035,
230
+ "step": 20813
231
+ },
232
+ {
233
+ "epoch": 13.0,
234
+ "eval_accuracy": 0.982089552238806,
235
+ "eval_f1": 0.9584635630860339,
236
+ "eval_loss": 0.13741779327392578,
237
+ "eval_precision": 0.9545771042391972,
238
+ "eval_recall": 0.9623817979105587,
239
+ "eval_runtime": 12.6726,
240
+ "eval_samples_per_second": 126.414,
241
+ "eval_steps_per_second": 15.861,
242
+ "step": 20813
243
+ },
244
+ {
245
+ "epoch": 14.0,
246
+ "learning_rate": 6.38314838195229e-06,
247
+ "loss": 0.0027,
248
+ "step": 22414
249
+ },
250
+ {
251
+ "epoch": 14.0,
252
+ "eval_accuracy": 0.9823481020096368,
253
+ "eval_f1": 0.9589226973684211,
254
+ "eval_loss": 0.14472806453704834,
255
+ "eval_precision": 0.9548785980428285,
256
+ "eval_recall": 0.9630011975058843,
257
+ "eval_runtime": 12.6834,
258
+ "eval_samples_per_second": 126.306,
259
+ "eval_steps_per_second": 15.847,
260
+ "step": 22414
261
+ },
262
+ {
263
+ "epoch": 15.0,
264
+ "learning_rate": 5.319290318293575e-06,
265
+ "loss": 0.003,
266
+ "step": 24015
267
+ },
268
+ {
269
+ "epoch": 15.0,
270
+ "eval_accuracy": 0.9819015160418381,
271
+ "eval_f1": 0.9580811200329422,
272
+ "eval_loss": 0.1390347182750702,
273
+ "eval_precision": 0.955405904816655,
274
+ "eval_recall": 0.9607713589627122,
275
+ "eval_runtime": 13.4492,
276
+ "eval_samples_per_second": 119.115,
277
+ "eval_steps_per_second": 14.945,
278
+ "step": 24015
279
+ },
280
+ {
281
+ "epoch": 16.0,
282
+ "learning_rate": 4.25543225463486e-06,
283
+ "loss": 0.0025,
284
+ "step": 25616
285
+ },
286
+ {
287
+ "epoch": 16.0,
288
+ "eval_accuracy": 0.9812433893524504,
289
+ "eval_f1": 0.9562028758600799,
290
+ "eval_loss": 0.1537328064441681,
291
+ "eval_precision": 0.9540801644398766,
292
+ "eval_recall": 0.9583350538877647,
293
+ "eval_runtime": 12.9219,
294
+ "eval_samples_per_second": 123.976,
295
+ "eval_steps_per_second": 15.555,
296
+ "step": 25616
297
+ },
298
+ {
299
+ "epoch": 17.0,
300
+ "learning_rate": 3.191574190976145e-06,
301
+ "loss": 0.0029,
302
+ "step": 27217
303
+ },
304
+ {
305
+ "epoch": 17.0,
306
+ "eval_accuracy": 0.982207074861911,
307
+ "eval_f1": 0.9580219463487947,
308
+ "eval_loss": 0.15050023794174194,
309
+ "eval_precision": 0.9552882246674331,
310
+ "eval_recall": 0.9607713589627122,
311
+ "eval_runtime": 12.7344,
312
+ "eval_samples_per_second": 125.801,
313
+ "eval_steps_per_second": 15.784,
314
+ "step": 27217
315
+ },
316
+ {
317
+ "epoch": 18.0,
318
+ "learning_rate": 2.12771612731743e-06,
319
+ "loss": 0.0022,
320
+ "step": 28818
321
+ },
322
+ {
323
+ "epoch": 18.0,
324
+ "eval_accuracy": 0.9823245974850159,
325
+ "eval_f1": 0.9589283509609449,
326
+ "eval_loss": 0.14898641407489777,
327
+ "eval_precision": 0.955703211517165,
328
+ "eval_recall": 0.9621753313787835,
329
+ "eval_runtime": 12.8609,
330
+ "eval_samples_per_second": 124.563,
331
+ "eval_steps_per_second": 15.629,
332
+ "step": 28818
333
+ },
334
+ {
335
+ "epoch": 19.0,
336
+ "learning_rate": 1.063858063658715e-06,
337
+ "loss": 0.0021,
338
+ "step": 30419
339
+ },
340
+ {
341
+ "epoch": 19.0,
342
+ "eval_accuracy": 0.982019038664943,
343
+ "eval_f1": 0.9582288369033425,
344
+ "eval_loss": 0.15266422927379608,
345
+ "eval_precision": 0.954070981210856,
346
+ "eval_recall": 0.9624230912169137,
347
+ "eval_runtime": 12.7956,
348
+ "eval_samples_per_second": 125.199,
349
+ "eval_steps_per_second": 15.709,
350
+ "step": 30419
351
+ },
352
+ {
353
+ "epoch": 20.0,
354
+ "learning_rate": 0.0,
355
+ "loss": 0.0018,
356
+ "step": 32020
357
+ },
358
+ {
359
+ "epoch": 20.0,
360
+ "eval_accuracy": 0.9824421201081208,
361
+ "eval_f1": 0.9590698631546456,
362
+ "eval_loss": 0.15378834307193756,
363
+ "eval_precision": 0.9559028632373452,
364
+ "eval_recall": 0.9622579179914936,
365
+ "eval_runtime": 12.9229,
366
+ "eval_samples_per_second": 123.966,
367
+ "eval_steps_per_second": 15.554,
368
+ "step": 32020
369
+ },
370
+ {
371
+ "epoch": 20.0,
372
+ "step": 32020,
373
+ "total_flos": 2.8281562629741216e+16,
374
+ "train_loss": 0.06704426395528247,
375
+ "train_runtime": 9262.2202,
376
+ "train_samples_per_second": 27.663,
377
+ "train_steps_per_second": 3.457
378
+ }
379
+ ],
380
+ "max_steps": 32020,
381
+ "num_train_epochs": 20,
382
+ "total_flos": 2.8281562629741216e+16,
383
+ "trial_name": null,
384
+ "trial_params": null
385
+ }
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c87232ed0e7c8494b8f9fa76ed2a5797df3debe277f89cc496a4a28e9ce0d74c
3
+ size 3832