filip503 commited on
Commit
16d5d37
·
verified ·
1 Parent(s): 9a626e0

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 +111 -51
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": 65
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": 65
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:a86118f43b31a4694f2a2a61f6fee78fd78c74ee52d256583372d2bc09b23807
3
- size 31207604
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c5679164a1d1156d9f1ca55961eb688eb4fe1233c59f3ea1aadcd428b8afd547
3
+ size 31232228
tokenizer.json CHANGED
@@ -114,71 +114,79 @@
114
  "8": 14,
115
  "9": 15,
116
  "=": 16,
117
- "10": 17,
118
- "99": 18,
119
- "11": 19,
120
- "98": 20,
121
- "97": 21,
122
- "12": 22,
123
- "13": 23,
124
- "96": 24,
125
- "14": 25,
126
- "95": 26,
127
  "15": 27,
128
  "94": 28,
129
- "93": 29,
130
- "16": 30,
131
  "17": 31,
132
  "92": 32,
133
- "18": 33,
134
- "91": 34,
135
- "19": 35,
136
- "90": 36,
137
  "89": 37,
138
  "20": 38,
139
- "21": 39,
140
- "88": 40,
141
  "22": 41,
142
  "87": 42,
143
  "86": 43,
144
  "23": 44,
145
  "85": 45,
146
  "24": 46,
147
- "25": 47,
148
- "84": 48,
149
  "26": 49,
150
- "27": 50,
151
- "83": 51,
152
- "82": 52
 
 
 
 
 
 
 
 
 
 
 
 
153
  },
154
  "merges": [
155
- [
156
- "1",
157
- "0"
158
- ],
159
  [
160
  "9",
161
  "9"
162
  ],
163
  [
164
  "1",
165
- "1"
166
  ],
167
  [
168
  "9",
169
  "8"
170
  ],
171
  [
172
- "9",
173
- "7"
174
  ],
175
  [
176
  "1",
177
  "2"
178
  ],
179
  [
180
- "1",
181
- "3"
182
  ],
183
  [
184
  "9",
@@ -186,7 +194,7 @@
186
  ],
187
  [
188
  "1",
189
- "4"
190
  ],
191
  [
192
  "9",
@@ -194,20 +202,24 @@
194
  ],
195
  [
196
  "1",
197
- "5"
198
  ],
199
  [
200
- "9",
201
- "4"
202
  ],
203
  [
204
  "9",
205
- "3"
206
  ],
207
  [
208
  "1",
209
  "6"
210
  ],
 
 
 
 
211
  [
212
  "1",
213
  "7"
@@ -216,38 +228,38 @@
216
  "9",
217
  "2"
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
  "2",
245
- "1"
246
  ],
247
  [
248
  "8",
249
  "8"
250
  ],
 
 
 
 
251
  [
252
  "2",
253
  "2"
@@ -272,29 +284,77 @@
272
  "2",
273
  "4"
274
  ],
275
- [
276
- "2",
277
- "5"
278
- ],
279
  [
280
  "8",
281
  "4"
282
  ],
 
 
 
 
283
  [
284
  "2",
285
  "6"
286
  ],
 
 
 
 
287
  [
288
  "2",
289
  "7"
290
  ],
291
  [
292
  "8",
293
- "3"
294
  ],
295
  [
296
  "8",
297
  "2"
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
298
  ]
299
  ]
300
  }
 
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
+ "96": 23,
124
+ "13": 24,
125
+ "95": 25,
126
+ "14": 26,
127
  "15": 27,
128
  "94": 28,
129
+ "16": 29,
130
+ "93": 30,
131
  "17": 31,
132
  "92": 32,
133
+ "91": 33,
134
+ "18": 34,
135
+ "90": 35,
136
+ "19": 36,
137
  "89": 37,
138
  "20": 38,
139
+ "88": 39,
140
+ "21": 40,
141
  "22": 41,
142
  "87": 42,
143
  "86": 43,
144
  "23": 44,
145
  "85": 45,
146
  "24": 46,
147
+ "84": 47,
148
+ "25": 48,
149
  "26": 49,
150
+ "83": 50,
151
+ "27": 51,
152
+ "81": 52,
153
+ "82": 53,
154
+ "28": 54,
155
+ "80": 55,
156
+ "30": 56,
157
+ "29": 57,
158
+ "31": 58,
159
+ "79": 59,
160
+ "78": 60,
161
+ "76": 61,
162
+ "34": 62,
163
+ "77": 63,
164
+ "33": 64
165
  },
166
  "merges": [
 
 
 
 
167
  [
168
  "9",
169
  "9"
170
  ],
171
  [
172
  "1",
173
+ "0"
174
  ],
175
  [
176
  "9",
177
  "8"
178
  ],
179
  [
180
+ "1",
181
+ "1"
182
  ],
183
  [
184
  "1",
185
  "2"
186
  ],
187
  [
188
+ "9",
189
+ "7"
190
  ],
191
  [
192
  "9",
 
194
  ],
195
  [
196
  "1",
197
+ "3"
198
  ],
199
  [
200
  "9",
 
202
  ],
203
  [
204
  "1",
205
+ "4"
206
  ],
207
  [
208
+ "1",
209
+ "5"
210
  ],
211
  [
212
  "9",
213
+ "4"
214
  ],
215
  [
216
  "1",
217
  "6"
218
  ],
219
+ [
220
+ "9",
221
+ "3"
222
+ ],
223
  [
224
  "1",
225
  "7"
 
228
  "9",
229
  "2"
230
  ],
 
 
 
 
231
  [
232
  "9",
233
  "1"
234
  ],
235
  [
236
  "1",
237
+ "8"
238
  ],
239
  [
240
  "9",
241
  "0"
242
  ],
243
  [
244
+ "1",
245
  "9"
246
  ],
247
  [
248
+ "8",
249
+ "9"
250
  ],
251
  [
252
  "2",
253
+ "0"
254
  ],
255
  [
256
  "8",
257
  "8"
258
  ],
259
+ [
260
+ "2",
261
+ "1"
262
+ ],
263
  [
264
  "2",
265
  "2"
 
284
  "2",
285
  "4"
286
  ],
 
 
 
 
287
  [
288
  "8",
289
  "4"
290
  ],
291
+ [
292
+ "2",
293
+ "5"
294
+ ],
295
  [
296
  "2",
297
  "6"
298
  ],
299
+ [
300
+ "8",
301
+ "3"
302
+ ],
303
  [
304
  "2",
305
  "7"
306
  ],
307
  [
308
  "8",
309
+ "1"
310
  ],
311
  [
312
  "8",
313
  "2"
314
+ ],
315
+ [
316
+ "2",
317
+ "8"
318
+ ],
319
+ [
320
+ "8",
321
+ "0"
322
+ ],
323
+ [
324
+ "3",
325
+ "0"
326
+ ],
327
+ [
328
+ "2",
329
+ "9"
330
+ ],
331
+ [
332
+ "3",
333
+ "1"
334
+ ],
335
+ [
336
+ "7",
337
+ "9"
338
+ ],
339
+ [
340
+ "7",
341
+ "8"
342
+ ],
343
+ [
344
+ "7",
345
+ "6"
346
+ ],
347
+ [
348
+ "3",
349
+ "4"
350
+ ],
351
+ [
352
+ "7",
353
+ "7"
354
+ ],
355
+ [
356
+ "3",
357
+ "3"
358
  ]
359
  ]
360
  }