reivvax commited on
Commit
0fc30b5
·
verified ·
1 Parent(s): e2dde7d

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 +68 -67
config.json CHANGED
@@ -40,7 +40,7 @@
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",
@@ -82,7 +82,7 @@
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,
 
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
  "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,
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:0eea56d5430e7e0d8877e4f9dbe33085e6256b87acb34bb596b9f7534fae0c30
3
- size 31205552
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:52dbfb9118b20b101c075fcdebb6c6ba1d42f4180ef7eab7b6507b89937800eb
3
+ size 31207604
tokenizer.json CHANGED
@@ -113,42 +113,43 @@
113
  "7": 13,
114
  "8": 14,
115
  "9": 15,
116
- "10": 16,
117
- "99": 17,
118
- "98": 18,
119
- "11": 19,
120
- "97": 20,
121
- "12": 21,
122
- "13": 22,
123
- "96": 23,
124
- "95": 24,
125
- "14": 25,
126
- "94": 26,
127
- "15": 27,
128
- "16": 28,
129
- "93": 29,
130
- "17": 30,
131
- "92": 31,
132
- "91": 32,
133
- "18": 33,
134
- "90": 34,
135
- "19": 35,
136
- "20": 36,
137
- "89": 37,
138
- "21": 38,
139
- "88": 39,
140
- "22": 40,
141
  "87": 41,
142
- "23": 42,
143
- "86": 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
  [
@@ -156,52 +157,56 @@
156
  "0"
157
  ],
158
  [
159
- "9",
160
- "9"
161
- ],
162
- [
163
- "9",
164
- "8"
165
  ],
166
  [
167
  "1",
168
- "1"
169
  ],
170
  [
171
- "9",
172
- "7"
173
  ],
174
  [
175
  "1",
176
- "2"
177
  ],
178
  [
179
  "1",
180
- "3"
181
  ],
182
  [
183
  "9",
 
 
 
 
184
  "6"
185
  ],
186
  [
187
  "9",
188
- "5"
189
  ],
190
  [
191
- "1",
192
- "4"
193
  ],
194
  [
195
  "9",
196
- "4"
197
  ],
198
  [
199
  "1",
 
 
 
 
200
  "5"
201
  ],
202
  [
203
- "1",
204
- "6"
205
  ],
206
  [
207
  "9",
@@ -209,7 +214,7 @@
209
  ],
210
  [
211
  "1",
212
- "7"
213
  ],
214
  [
215
  "9",
@@ -219,10 +224,6 @@
219
  "9",
220
  "1"
221
  ],
222
- [
223
- "1",
224
- "8"
225
- ],
226
  [
227
  "9",
228
  "0"
@@ -247,14 +248,14 @@
247
  "8",
248
  "8"
249
  ],
250
- [
251
- "2",
252
- "2"
253
- ],
254
  [
255
  "8",
256
  "7"
257
  ],
 
 
 
 
258
  [
259
  "2",
260
  "3"
@@ -263,14 +264,14 @@
263
  "8",
264
  "6"
265
  ],
266
- [
267
- "2",
268
- "4"
269
- ],
270
  [
271
  "8",
272
  "5"
273
  ],
 
 
 
 
274
  [
275
  "2",
276
  "5"
@@ -279,14 +280,14 @@
279
  "8",
280
  "4"
281
  ],
282
- [
283
- "2",
284
- "6"
285
- ],
286
  [
287
  "8",
288
  "3"
289
  ],
 
 
 
 
290
  [
291
  "2",
292
  "7"
 
113
  "7": 13,
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
+ "96": 27,
128
+ "17": 28,
129
+ "95": 29,
130
+ "94": 30,
131
+ "93": 31,
132
+ "18": 32,
133
+ "92": 33,
134
+ "91": 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
  "85": 45,
146
+ "24": 46,
147
+ "25": 47,
148
+ "84": 48,
149
  "83": 49,
150
+ "26": 50,
151
+ "27": 51,
152
+ "82": 52
153
  },
154
  "merges": [
155
  [
 
157
  "0"
158
  ],
159
  [
160
+ "1",
161
+ "1"
 
 
 
 
162
  ],
163
  [
164
  "1",
165
+ "2"
166
  ],
167
  [
168
+ "1",
169
+ "3"
170
  ],
171
  [
172
  "1",
173
+ "4"
174
  ],
175
  [
176
  "1",
177
+ "5"
178
  ],
179
  [
180
  "9",
181
+ "9"
182
+ ],
183
+ [
184
+ "1",
185
  "6"
186
  ],
187
  [
188
  "9",
189
+ "8"
190
  ],
191
  [
192
+ "9",
193
+ "7"
194
  ],
195
  [
196
  "9",
197
+ "6"
198
  ],
199
  [
200
  "1",
201
+ "7"
202
+ ],
203
+ [
204
+ "9",
205
  "5"
206
  ],
207
  [
208
+ "9",
209
+ "4"
210
  ],
211
  [
212
  "9",
 
214
  ],
215
  [
216
  "1",
217
+ "8"
218
  ],
219
  [
220
  "9",
 
224
  "9",
225
  "1"
226
  ],
 
 
 
 
227
  [
228
  "9",
229
  "0"
 
248
  "8",
249
  "8"
250
  ],
 
 
 
 
251
  [
252
  "8",
253
  "7"
254
  ],
255
+ [
256
+ "2",
257
+ "2"
258
+ ],
259
  [
260
  "2",
261
  "3"
 
264
  "8",
265
  "6"
266
  ],
 
 
 
 
267
  [
268
  "8",
269
  "5"
270
  ],
271
+ [
272
+ "2",
273
+ "4"
274
+ ],
275
  [
276
  "2",
277
  "5"
 
280
  "8",
281
  "4"
282
  ],
 
 
 
 
283
  [
284
  "8",
285
  "3"
286
  ],
287
+ [
288
+ "2",
289
+ "6"
290
+ ],
291
  [
292
  "2",
293
  "7"