FranciszekW commited on
Commit
fe651cd
·
verified ·
1 Parent(s): eb2ec25

Training in progress, step 240

Browse files
Files changed (3) hide show
  1. config.json +2 -2
  2. model.safetensors +2 -2
  3. tokenizer.json +76 -78
config.json CHANGED
@@ -40,7 +40,7 @@
40
  "tie_word_embeddings": true,
41
  "type_vocab_size": 2,
42
  "use_cache": true,
43
- "vocab_size": 54
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": 54
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:e13a7cb3d80a0a4723bcc35b941d6060d51926d4597f86235e4a12ceb5dd5402
3
- size 31209656
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e42d7c28dbd00f357ecfdf3c482dc2ce94bf44fba5c763652019387239936a0e
3
+ size 31205552
tokenizer.json CHANGED
@@ -102,65 +102,63 @@
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
- "99": 19,
120
- "98": 20,
121
- "11": 21,
122
- "97": 22,
123
- "12": 23,
124
- "96": 24,
125
- "13": 25,
126
- "14": 26,
127
- "95": 27,
128
- "94": 28,
129
- "15": 29,
130
- "93": 30,
131
- "16": 31,
132
- "92": 32,
133
- "17": 33,
134
- "18": 34,
135
- "91": 35,
136
- "90": 36,
137
- "19": 37,
138
- "20": 38,
139
- "89": 39,
140
- "21": 40,
141
- "88": 41,
142
- "87": 42,
143
- "22": 43,
144
- "23": 44,
145
- "86": 45,
146
- "85": 46,
147
- "24": 47,
148
- "25": 48,
149
- "84": 49,
150
- "83": 50,
151
- "26": 51,
152
- "27": 52,
153
- "82": 53
154
  },
155
  "merges": [
156
- [
157
- "1",
158
- "0"
159
- ],
160
  [
161
  "9",
162
  "9"
163
  ],
 
 
 
 
164
  [
165
  "9",
166
  "8"
@@ -177,14 +175,14 @@
177
  "1",
178
  "2"
179
  ],
180
- [
181
- "9",
182
- "6"
183
- ],
184
  [
185
  "1",
186
  "3"
187
  ],
 
 
 
 
188
  [
189
  "1",
190
  "4"
@@ -193,17 +191,13 @@
193
  "9",
194
  "5"
195
  ],
196
- [
197
- "9",
198
- "4"
199
- ],
200
  [
201
  "1",
202
  "5"
203
  ],
204
  [
205
  "9",
206
- "3"
207
  ],
208
  [
209
  "1",
@@ -211,36 +205,40 @@
211
  ],
212
  [
213
  "9",
214
- "2"
215
  ],
216
  [
217
  "1",
218
  "7"
219
  ],
220
  [
221
- "1",
222
- "8"
223
  ],
224
  [
225
  "9",
226
  "1"
227
  ],
228
  [
229
- "9",
230
- "0"
231
  ],
232
  [
233
  "1",
234
  "9"
235
  ],
236
  [
237
- "2",
238
  "0"
239
  ],
240
  [
241
  "8",
242
  "9"
243
  ],
 
 
 
 
244
  [
245
  "2",
246
  "1"
@@ -257,38 +255,38 @@
257
  "2",
258
  "2"
259
  ],
260
- [
261
- "2",
262
- "3"
263
- ],
264
  [
265
  "8",
266
  "6"
267
  ],
268
  [
269
- "8",
270
- "5"
271
  ],
272
  [
273
  "2",
274
  "4"
275
  ],
276
  [
277
- "2",
278
  "5"
279
  ],
280
  [
281
- "8",
282
- "4"
283
  ],
284
  [
285
  "8",
286
- "3"
287
  ],
288
  [
289
  "2",
290
  "6"
291
  ],
 
 
 
 
292
  [
293
  "2",
294
  "7"
 
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
+ "99": 16,
117
+ "10": 17,
118
+ "98": 18,
119
+ "11": 19,
120
+ "97": 20,
121
+ "12": 21,
122
+ "13": 22,
123
+ "96": 23,
124
+ "14": 24,
125
+ "95": 25,
126
+ "15": 26,
127
+ "94": 27,
128
+ "16": 28,
129
+ "93": 29,
130
+ "17": 30,
131
+ "92": 31,
132
+ "91": 32,
133
+ "18": 33,
134
+ "19": 34,
135
+ "90": 35,
136
+ "89": 36,
137
+ "20": 37,
138
+ "21": 38,
139
+ "88": 39,
140
+ "87": 40,
141
+ "22": 41,
142
+ "86": 42,
143
+ "23": 43,
144
+ "24": 44,
145
+ "85": 45,
146
+ "25": 46,
147
+ "84": 47,
148
+ "26": 48,
149
+ "83": 49,
150
+ "27": 50,
151
+ "82": 51
 
 
152
  },
153
  "merges": [
 
 
 
 
154
  [
155
  "9",
156
  "9"
157
  ],
158
+ [
159
+ "1",
160
+ "0"
161
+ ],
162
  [
163
  "9",
164
  "8"
 
175
  "1",
176
  "2"
177
  ],
 
 
 
 
178
  [
179
  "1",
180
  "3"
181
  ],
182
+ [
183
+ "9",
184
+ "6"
185
+ ],
186
  [
187
  "1",
188
  "4"
 
191
  "9",
192
  "5"
193
  ],
 
 
 
 
194
  [
195
  "1",
196
  "5"
197
  ],
198
  [
199
  "9",
200
+ "4"
201
  ],
202
  [
203
  "1",
 
205
  ],
206
  [
207
  "9",
208
+ "3"
209
  ],
210
  [
211
  "1",
212
  "7"
213
  ],
214
  [
215
+ "9",
216
+ "2"
217
  ],
218
  [
219
  "9",
220
  "1"
221
  ],
222
  [
223
+ "1",
224
+ "8"
225
  ],
226
  [
227
  "1",
228
  "9"
229
  ],
230
  [
231
+ "9",
232
  "0"
233
  ],
234
  [
235
  "8",
236
  "9"
237
  ],
238
+ [
239
+ "2",
240
+ "0"
241
+ ],
242
  [
243
  "2",
244
  "1"
 
255
  "2",
256
  "2"
257
  ],
 
 
 
 
258
  [
259
  "8",
260
  "6"
261
  ],
262
  [
263
+ "2",
264
+ "3"
265
  ],
266
  [
267
  "2",
268
  "4"
269
  ],
270
  [
271
+ "8",
272
  "5"
273
  ],
274
  [
275
+ "2",
276
+ "5"
277
  ],
278
  [
279
  "8",
280
+ "4"
281
  ],
282
  [
283
  "2",
284
  "6"
285
  ],
286
+ [
287
+ "8",
288
+ "3"
289
+ ],
290
  [
291
  "2",
292
  "7"