pancake5000 commited on
Commit
ae58701
·
verified ·
1 Parent(s): 8ca7c54

Training in progress, step 400

Browse files
Files changed (3) hide show
  1. config.json +2 -2
  2. model.safetensors +2 -2
  3. tokenizer.json +74 -79
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": 52
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": 52
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 sha256:015709ee7668471e29d13a4ba7bac028e761e0f0a0eda9f7bfa5d4ea58465836
3
- size 31207604
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:332ee77d14c5237bf36502e873bac4394e709ceb4d862be0c5208e7bc1ace2a8
3
+ size 31205552
tokenizer.json CHANGED
@@ -114,56 +114,47 @@
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
- "15": 27,
128
- "94": 28,
129
- "93": 29,
130
- "16": 30,
131
- "17": 31,
132
- "92": 32,
133
- "91": 33,
134
- "18": 34,
135
- "90": 35,
136
- "19": 36,
137
- "20": 37,
138
- "89": 38,
139
- "21": 39,
140
- "88": 40,
141
- "87": 41,
142
- "22": 42,
143
- "23": 43,
144
- "86": 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",
165
- "8"
166
- ],
167
  [
168
  "1",
169
  "1"
@@ -172,105 +163,101 @@
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
  "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"
214
  ],
215
  [
216
  "9",
217
- "2"
218
  ],
219
  [
220
  "9",
221
- "1"
222
- ],
223
- [
224
- "1",
225
- "8"
226
  ],
227
  [
228
  "9",
229
- "0"
230
  ],
231
  [
232
  "1",
233
  "9"
234
  ],
 
 
 
 
235
  [
236
  "2",
237
  "0"
238
  ],
239
  [
240
- "8",
241
- "9"
242
  ],
243
  [
244
  "2",
245
  "1"
246
  ],
247
  [
248
- "8",
249
- "8"
250
  ],
251
  [
252
- "8",
253
- "7"
254
  ],
255
  [
256
  "2",
257
- "2"
258
  ],
259
  [
260
  "2",
261
- "3"
262
  ],
263
  [
264
  "8",
265
- "6"
266
- ],
267
- [
268
- "2",
269
- "4"
270
  ],
271
  [
272
  "8",
273
- "5"
274
  ],
275
  [
276
  "2",
@@ -278,7 +265,7 @@
278
  ],
279
  [
280
  "8",
281
- "4"
282
  ],
283
  [
284
  "2",
@@ -288,13 +275,21 @@
288
  "8",
289
  "3"
290
  ],
 
 
 
 
291
  [
292
  "2",
293
  "7"
294
  ],
295
  [
296
  "8",
297
- "2"
 
 
 
 
298
  ]
299
  ]
300
  }
 
114
  "8": 14,
115
  "9": 15,
116
  "=": 16,
117
+ "10": 17,
118
+ "11": 18,
119
+ "12": 19,
120
+ "13": 20,
121
+ "14": 21,
122
+ "15": 22,
123
+ "99": 23,
124
+ "16": 24,
125
+ "98": 25,
126
+ "97": 26,
127
+ "17": 27,
128
+ "96": 28,
129
+ "18": 29,
130
+ "95": 30,
131
+ "94": 31,
132
+ "93": 32,
133
+ "19": 33,
134
+ "92": 34,
135
+ "20": 35,
136
+ "91": 36,
137
+ "21": 37,
138
+ "22": 38,
139
+ "90": 39,
140
+ "23": 40,
141
+ "24": 41,
142
+ "88": 42,
143
+ "89": 43,
144
+ "25": 44,
145
+ "86": 45,
146
+ "26": 46,
147
+ "83": 47,
148
+ "87": 48,
149
+ "27": 49,
150
+ "84": 50,
151
+ "85": 51
 
152
  },
153
  "merges": [
 
 
 
 
154
  [
155
  "1",
156
  "0"
157
  ],
 
 
 
 
158
  [
159
  "1",
160
  "1"
 
163
  "1",
164
  "2"
165
  ],
 
 
 
 
166
  [
167
  "1",
168
  "3"
169
  ],
170
  [
171
+ "1",
172
+ "4"
173
  ],
174
  [
175
  "1",
176
+ "5"
177
  ],
178
  [
179
  "9",
180
+ "9"
181
  ],
182
  [
183
  "1",
184
+ "6"
185
  ],
186
  [
187
  "9",
188
+ "8"
189
  ],
190
  [
191
  "9",
192
+ "7"
193
  ],
194
  [
195
  "1",
196
+ "7"
197
+ ],
198
+ [
199
+ "9",
200
  "6"
201
  ],
202
  [
203
  "1",
204
+ "8"
205
  ],
206
  [
207
  "9",
208
+ "5"
209
  ],
210
  [
211
  "9",
212
+ "4"
 
 
 
 
213
  ],
214
  [
215
  "9",
216
+ "3"
217
  ],
218
  [
219
  "1",
220
  "9"
221
  ],
222
+ [
223
+ "9",
224
+ "2"
225
+ ],
226
  [
227
  "2",
228
  "0"
229
  ],
230
  [
231
+ "9",
232
+ "1"
233
  ],
234
  [
235
  "2",
236
  "1"
237
  ],
238
  [
239
+ "2",
240
+ "2"
241
  ],
242
  [
243
+ "9",
244
+ "0"
245
  ],
246
  [
247
  "2",
248
+ "3"
249
  ],
250
  [
251
  "2",
252
+ "4"
253
  ],
254
  [
255
  "8",
256
+ "8"
 
 
 
 
257
  ],
258
  [
259
  "8",
260
+ "9"
261
  ],
262
  [
263
  "2",
 
265
  ],
266
  [
267
  "8",
268
+ "6"
269
  ],
270
  [
271
  "2",
 
275
  "8",
276
  "3"
277
  ],
278
+ [
279
+ "8",
280
+ "7"
281
+ ],
282
  [
283
  "2",
284
  "7"
285
  ],
286
  [
287
  "8",
288
+ "4"
289
+ ],
290
+ [
291
+ "8",
292
+ "5"
293
  ]
294
  ]
295
  }