1miqi1 commited on
Commit
377d4f2
·
verified ·
1 Parent(s): 802d484

Training in progress, step 500

Browse files
Files changed (3) hide show
  1. config.json +2 -2
  2. model.safetensors +2 -2
  3. tokenizer.json +347 -76
config.json CHANGED
@@ -40,7 +40,7 @@
40
  "tie_word_embeddings": true,
41
  "type_vocab_size": 2,
42
  "use_cache": true,
43
- "vocab_size": 53
44
  },
45
  "decoder_start_token_id": 2,
46
  "dtype": "float32",
@@ -82,7 +82,7 @@
82
  "tie_word_embeddings": true,
83
  "type_vocab_size": 2,
84
  "use_cache": true,
85
- "vocab_size": 53
86
  },
87
  "eos_token_id": 0,
88
  "is_encoder_decoder": true,
 
40
  "tie_word_embeddings": true,
41
  "type_vocab_size": 2,
42
  "use_cache": true,
43
+ "vocab_size": 108
44
  },
45
  "decoder_start_token_id": 2,
46
  "dtype": "float32",
 
82
  "tie_word_embeddings": true,
83
  "type_vocab_size": 2,
84
  "use_cache": true,
85
+ "vocab_size": 108
86
  },
87
  "eos_token_id": 0,
88
  "is_encoder_decoder": true,
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:df44ee02992cf3018a2062b84e63124247286bb4c4026951735859786ca8ffef
3
- size 31207604
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dc0eed79d14e94b5efed2bf834ba1389ac093b5f50db8f06eb1a9720c0f6a6b2
3
+ size 31320464
tokenizer.json CHANGED
@@ -101,64 +101,123 @@
101
  "[UNK]": 1,
102
  "[CLS]": 2,
103
  "[PAD]": 3,
104
- "+": 4,
105
- "-": 5,
106
- "0": 6,
107
- "1": 7,
108
- "2": 8,
109
- "3": 9,
110
- "4": 10,
111
- "5": 11,
112
- "6": 12,
113
- "7": 13,
114
- "8": 14,
115
- "9": 15,
116
- "=": 16,
117
- "99": 17,
118
  "10": 18,
119
- "98": 19,
120
- "11": 20,
121
- "12": 21,
122
- "97": 22,
123
- "13": 23,
124
- "96": 24,
125
- "14": 25,
126
- "95": 26,
127
- "94": 27,
128
- "15": 28,
129
- "16": 29,
130
- "93": 30,
131
- "92": 31,
132
- "17": 32,
133
- "18": 33,
134
- "91": 34,
135
- "19": 35,
136
- "90": 36,
137
- "89": 37,
138
- "20": 38,
139
- "88": 39,
140
- "21": 40,
141
- "87": 41,
142
- "22": 42,
143
- "86": 43,
144
- "23": 44,
145
- "24": 45,
146
- "85": 46,
147
- "25": 47,
148
- "84": 48,
149
- "26": 49,
150
- "83": 50,
151
- "27": 51,
152
- "82": 52
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
153
  },
154
  "merges": [
155
  [
156
- "9",
157
- "9"
158
  ],
159
  [
160
  "1",
161
- "0"
 
 
 
 
162
  ],
163
  [
164
  "9",
@@ -170,131 +229,343 @@
170
  ],
171
  [
172
  "1",
173
- "2"
174
  ],
175
  [
176
  "9",
177
  "7"
178
  ],
 
 
 
 
 
 
 
 
 
 
 
 
179
  [
180
  "1",
181
  "3"
182
  ],
183
  [
184
  "9",
185
- "6"
186
  ],
187
  [
188
  "1",
189
- "4"
190
  ],
191
  [
192
  "9",
193
- "5"
194
  ],
195
  [
196
  "9",
 
 
 
 
197
  "4"
198
  ],
199
  [
200
  "1",
201
- "5"
202
  ],
203
  [
204
- "1",
 
 
 
 
205
  "6"
206
  ],
207
  [
208
- "9",
209
- "3"
210
  ],
211
  [
212
  "9",
 
 
 
 
 
 
 
 
213
  "2"
214
  ],
215
  [
216
- "1",
217
- "7"
218
  ],
219
  [
220
- "1",
221
  "8"
222
  ],
223
  [
224
- "9",
 
 
 
 
225
  "1"
226
  ],
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
227
  [
228
  "1",
229
  "9"
230
  ],
 
 
 
 
 
 
 
 
 
 
 
 
231
  [
232
  "9",
233
  "0"
234
  ],
235
  [
236
- "8",
237
  "9"
238
  ],
239
  [
240
  "2",
241
- "0"
242
  ],
243
  [
244
- "8",
245
- "8"
246
  ],
247
  [
248
  "2",
 
 
 
 
249
  "1"
250
  ],
251
  [
252
  "8",
253
- "7"
254
  ],
255
  [
256
- "2",
257
- "2"
 
 
 
 
 
 
 
 
 
 
 
 
258
  ],
259
  [
260
  "8",
261
  "6"
262
  ],
 
 
 
 
 
 
 
 
 
 
 
 
263
  [
264
  "2",
 
 
 
 
265
  "3"
266
  ],
267
  [
268
- "2",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
269
  "4"
270
  ],
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
271
  [
272
  "8",
273
  "5"
274
  ],
275
  [
276
- "2",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
277
  "5"
278
  ],
 
 
 
 
 
 
 
 
 
 
 
 
279
  [
280
  "8",
 
 
 
 
 
 
 
 
 
 
 
 
281
  "4"
282
  ],
283
  [
284
- "2",
285
- "6"
286
  ],
287
  [
288
- "8",
289
  "3"
290
  ],
291
  [
292
- "2",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
293
  "7"
294
  ],
295
  [
296
  "8",
297
  "2"
 
 
 
 
298
  ]
299
  ]
300
  }
 
101
  "[UNK]": 1,
102
  "[CLS]": 2,
103
  "[PAD]": 3,
104
+ "*": 4,
105
+ "+": 5,
106
+ "-": 6,
107
+ "0": 7,
108
+ "1": 8,
109
+ "2": 9,
110
+ "3": 10,
111
+ "4": 11,
112
+ "5": 12,
113
+ "6": 13,
114
+ "7": 14,
115
+ "8": 15,
116
+ "9": 16,
117
+ "=": 17,
118
  "10": 18,
119
+ "12": 19,
120
+ "99": 20,
121
+ "98": 21,
122
+ "11": 22,
123
+ "14": 23,
124
+ "97": 24,
125
+ "96": 25,
126
+ "16": 26,
127
+ "15": 27,
128
+ "13": 28,
129
+ "95": 29,
130
+ "18": 30,
131
+ "94": 31,
132
+ "93": 32,
133
+ "24": 33,
134
+ "17": 34,
135
+ "92": 35,
136
+ "36": 36,
137
+ "20": 37,
138
+ "91": 38,
139
+ "40": 39,
140
+ "32": 40,
141
+ "48": 41,
142
+ "28": 42,
143
+ "54": 43,
144
+ "21": 44,
145
+ "30": 45,
146
+ "45": 46,
147
+ "72": 47,
148
+ "35": 48,
149
+ "27": 49,
150
+ "42": 50,
151
+ "19": 51,
152
+ "63": 52,
153
+ "25": 53,
154
+ "56": 54,
155
+ "90": 55,
156
+ "49": 56,
157
+ "22": 57,
158
+ "23": 58,
159
+ "26": 59,
160
+ "81": 60,
161
+ "88": 61,
162
+ "64": 62,
163
+ "89": 63,
164
+ "44": 64,
165
+ "80": 65,
166
+ "86": 66,
167
+ "60": 67,
168
+ "84": 68,
169
+ "70": 69,
170
+ "29": 70,
171
+ "33": 71,
172
+ "37": 72,
173
+ "41": 73,
174
+ "52": 74,
175
+ "38": 75,
176
+ "68": 76,
177
+ "87": 77,
178
+ "34": 78,
179
+ "39": 79,
180
+ "31": 80,
181
+ "47": 81,
182
+ "51": 82,
183
+ "50": 83,
184
+ "85": 84,
185
+ "43": 85,
186
+ "46": 86,
187
+ "57": 87,
188
+ "58": 88,
189
+ "65": 89,
190
+ "55": 90,
191
+ "73": 91,
192
+ "59": 92,
193
+ "62": 93,
194
+ "83": 94,
195
+ "61": 95,
196
+ "69": 96,
197
+ "74": 97,
198
+ "78": 98,
199
+ "53": 99,
200
+ "67": 100,
201
+ "71": 101,
202
+ "76": 102,
203
+ "79": 103,
204
+ "66": 104,
205
+ "77": 105,
206
+ "82": 106,
207
+ "75": 107
208
  },
209
  "merges": [
210
  [
211
+ "1",
212
+ "0"
213
  ],
214
  [
215
  "1",
216
+ "2"
217
+ ],
218
+ [
219
+ "9",
220
+ "9"
221
  ],
222
  [
223
  "9",
 
229
  ],
230
  [
231
  "1",
232
+ "4"
233
  ],
234
  [
235
  "9",
236
  "7"
237
  ],
238
+ [
239
+ "9",
240
+ "6"
241
+ ],
242
+ [
243
+ "1",
244
+ "6"
245
+ ],
246
+ [
247
+ "1",
248
+ "5"
249
+ ],
250
  [
251
  "1",
252
  "3"
253
  ],
254
  [
255
  "9",
256
+ "5"
257
  ],
258
  [
259
  "1",
260
+ "8"
261
  ],
262
  [
263
  "9",
264
+ "4"
265
  ],
266
  [
267
  "9",
268
+ "3"
269
+ ],
270
+ [
271
+ "2",
272
  "4"
273
  ],
274
  [
275
  "1",
276
+ "7"
277
  ],
278
  [
279
+ "9",
280
+ "2"
281
+ ],
282
+ [
283
+ "3",
284
  "6"
285
  ],
286
  [
287
+ "2",
288
+ "0"
289
  ],
290
  [
291
  "9",
292
+ "1"
293
+ ],
294
+ [
295
+ "4",
296
+ "0"
297
+ ],
298
+ [
299
+ "3",
300
  "2"
301
  ],
302
  [
303
+ "4",
304
+ "8"
305
  ],
306
  [
307
+ "2",
308
  "8"
309
  ],
310
  [
311
+ "5",
312
+ "4"
313
+ ],
314
+ [
315
+ "2",
316
  "1"
317
  ],
318
+ [
319
+ "3",
320
+ "0"
321
+ ],
322
+ [
323
+ "4",
324
+ "5"
325
+ ],
326
+ [
327
+ "7",
328
+ "2"
329
+ ],
330
+ [
331
+ "3",
332
+ "5"
333
+ ],
334
+ [
335
+ "2",
336
+ "7"
337
+ ],
338
+ [
339
+ "4",
340
+ "2"
341
+ ],
342
  [
343
  "1",
344
  "9"
345
  ],
346
+ [
347
+ "6",
348
+ "3"
349
+ ],
350
+ [
351
+ "2",
352
+ "5"
353
+ ],
354
+ [
355
+ "5",
356
+ "6"
357
+ ],
358
  [
359
  "9",
360
  "0"
361
  ],
362
  [
363
+ "4",
364
  "9"
365
  ],
366
  [
367
  "2",
368
+ "2"
369
  ],
370
  [
371
+ "2",
372
+ "3"
373
  ],
374
  [
375
  "2",
376
+ "6"
377
+ ],
378
+ [
379
+ "8",
380
  "1"
381
  ],
382
  [
383
  "8",
384
+ "8"
385
  ],
386
  [
387
+ "6",
388
+ "4"
389
+ ],
390
+ [
391
+ "8",
392
+ "9"
393
+ ],
394
+ [
395
+ "4",
396
+ "4"
397
+ ],
398
+ [
399
+ "8",
400
+ "0"
401
  ],
402
  [
403
  "8",
404
  "6"
405
  ],
406
+ [
407
+ "6",
408
+ "0"
409
+ ],
410
+ [
411
+ "8",
412
+ "4"
413
+ ],
414
+ [
415
+ "7",
416
+ "0"
417
+ ],
418
  [
419
  "2",
420
+ "9"
421
+ ],
422
+ [
423
+ "3",
424
  "3"
425
  ],
426
  [
427
+ "3",
428
+ "7"
429
+ ],
430
+ [
431
+ "4",
432
+ "1"
433
+ ],
434
+ [
435
+ "5",
436
+ "2"
437
+ ],
438
+ [
439
+ "3",
440
+ "8"
441
+ ],
442
+ [
443
+ "6",
444
+ "8"
445
+ ],
446
+ [
447
+ "8",
448
+ "7"
449
+ ],
450
+ [
451
+ "3",
452
  "4"
453
  ],
454
+ [
455
+ "3",
456
+ "9"
457
+ ],
458
+ [
459
+ "3",
460
+ "1"
461
+ ],
462
+ [
463
+ "4",
464
+ "7"
465
+ ],
466
+ [
467
+ "5",
468
+ "1"
469
+ ],
470
+ [
471
+ "5",
472
+ "0"
473
+ ],
474
  [
475
  "8",
476
  "5"
477
  ],
478
  [
479
+ "4",
480
+ "3"
481
+ ],
482
+ [
483
+ "4",
484
+ "6"
485
+ ],
486
+ [
487
+ "5",
488
+ "7"
489
+ ],
490
+ [
491
+ "5",
492
+ "8"
493
+ ],
494
+ [
495
+ "6",
496
+ "5"
497
+ ],
498
+ [
499
+ "5",
500
  "5"
501
  ],
502
+ [
503
+ "7",
504
+ "3"
505
+ ],
506
+ [
507
+ "5",
508
+ "9"
509
+ ],
510
+ [
511
+ "6",
512
+ "2"
513
+ ],
514
  [
515
  "8",
516
+ "3"
517
+ ],
518
+ [
519
+ "6",
520
+ "1"
521
+ ],
522
+ [
523
+ "6",
524
+ "9"
525
+ ],
526
+ [
527
+ "7",
528
  "4"
529
  ],
530
  [
531
+ "7",
532
+ "8"
533
  ],
534
  [
535
+ "5",
536
  "3"
537
  ],
538
  [
539
+ "6",
540
+ "7"
541
+ ],
542
+ [
543
+ "7",
544
+ "1"
545
+ ],
546
+ [
547
+ "7",
548
+ "6"
549
+ ],
550
+ [
551
+ "7",
552
+ "9"
553
+ ],
554
+ [
555
+ "6",
556
+ "6"
557
+ ],
558
+ [
559
+ "7",
560
  "7"
561
  ],
562
  [
563
  "8",
564
  "2"
565
+ ],
566
+ [
567
+ "7",
568
+ "5"
569
  ]
570
  ]
571
  }