ludziej commited on
Commit
3f26068
1 Parent(s): a0dda80

Training in progress, step 500

Browse files
config.json CHANGED
@@ -78,7 +78,7 @@
78
  "typical_p": 1.0,
79
  "use_bfloat16": false,
80
  "use_cache": true,
81
- "vocab_size": 53
82
  },
83
  "decoder_start_token_id": 2,
84
  "encoder": {
@@ -157,7 +157,7 @@
157
  "typical_p": 1.0,
158
  "use_bfloat16": false,
159
  "use_cache": true,
160
- "vocab_size": 53
161
  },
162
  "eos_token_id": 0,
163
  "is_encoder_decoder": true,
 
78
  "typical_p": 1.0,
79
  "use_bfloat16": false,
80
  "use_cache": true,
81
+ "vocab_size": 105
82
  },
83
  "decoder_start_token_id": 2,
84
  "encoder": {
 
157
  "typical_p": 1.0,
158
  "use_bfloat16": false,
159
  "use_cache": true,
160
+ "vocab_size": 105
161
  },
162
  "eos_token_id": 0,
163
  "is_encoder_decoder": true,
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:88677a6bc13ac58099475a50c0deea5f6f7c48e5d529d97000cfbd253bc7678d
3
- size 31207604
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:745100c3aba7304b571ebece470b3972643f144bcb1fa9d4fc321a438645ffa9
3
+ size 31314308
runs/Feb28_17-23-45_3897ec21fae5/events.out.tfevents.1709141028.3897ec21fae5.75144.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:78c5631269409aef73cd14f48df87f32a0b026867ec72ea23309509012866399
3
+ size 11282
runs/Feb28_17-27-37_3897ec21fae5/events.out.tfevents.1709141258.3897ec21fae5.75144.1 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:71c657388662f76e4a958810b0414bca1cdcb29ca0a10c98c482ed5e841fe499
3
+ size 8343
runs/Feb28_17-27-44_3897ec21fae5/events.out.tfevents.1709141265.3897ec21fae5.75144.2 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:64e2a55b6c903519c748e5da9e880db2dd59ab83da49914d3b8a2e29f735125e
3
+ size 23252
tokenizer.json CHANGED
@@ -101,92 +101,198 @@
101
  "[CLS]": 2,
102
  "[PAD]": 3,
103
  "+": 4,
104
- "-": 5,
105
- "0": 6,
106
- "1": 7,
107
- "2": 8,
108
- "3": 9,
109
- "4": 10,
110
- "5": 11,
111
- "6": 12,
112
- "7": 13,
113
- "8": 14,
114
- "9": 15,
115
- "=": 16,
116
- "10": 17,
117
- "99": 18,
118
- "11": 19,
119
- "98": 20,
120
- "97": 21,
121
- "12": 22,
122
- "96": 23,
123
- "13": 24,
124
- "14": 25,
125
- "95": 26,
126
- "94": 27,
127
- "15": 28,
128
- "93": 29,
129
- "16": 30,
130
- "92": 31,
131
- "17": 32,
132
- "91": 33,
133
- "18": 34,
134
- "19": 35,
135
- "90": 36,
136
- "20": 37,
137
- "89": 38,
138
- "21": 39,
139
- "88": 40,
140
- "22": 41,
141
- "87": 42,
142
- "86": 43,
143
- "23": 44,
144
- "85": 45,
145
- "24": 46,
146
- "25": 47,
147
- "84": 48,
148
- "26": 49,
149
- "83": 50,
150
- "27": 51,
151
- "82": 52
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
152
  },
153
  "merges": [
154
- "1 0",
155
- "9 9",
156
- "1 1",
157
- "9 8",
158
- "9 7",
159
- "1 2",
160
- "9 6",
161
- "1 3",
162
- "1 4",
163
- "9 5",
164
- "9 4",
165
- "1 5",
166
  "9 3",
167
- "1 6",
168
- "9 2",
169
- "1 7",
170
- "9 1",
171
- "1 8",
172
  "1 9",
173
- "9 0",
174
- "2 0",
175
- "8 9",
176
  "2 1",
 
 
 
177
  "8 8",
178
- "2 2",
 
 
 
 
 
 
 
 
179
  "8 7",
180
- "8 6",
181
- "2 3",
 
 
 
 
 
 
 
 
 
 
 
182
  "8 5",
 
 
 
183
  "2 4",
184
- "2 5",
 
 
 
 
 
 
185
  "8 4",
 
 
186
  "2 6",
187
- "8 3",
188
  "2 7",
189
- "8 2"
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
190
  ]
191
  }
192
  }
 
101
  "[CLS]": 2,
102
  "[PAD]": 3,
103
  "+": 4,
104
+ "0": 5,
105
+ "1": 6,
106
+ "2": 7,
107
+ "3": 8,
108
+ "4": 9,
109
+ "5": 10,
110
+ "6": 11,
111
+ "7": 12,
112
+ "8": 13,
113
+ "9": 14,
114
+ "50": 15,
115
+ "93": 16,
116
+ "11": 17,
117
+ "60": 18,
118
+ "19": 19,
119
+ "21": 20,
120
+ "33": 21,
121
+ "36": 22,
122
+ "66": 23,
123
+ "88": 24,
124
+ "12": 25,
125
+ "17": 26,
126
+ "81": 27,
127
+ "90": 28,
128
+ "16": 29,
129
+ "39": 30,
130
+ "55": 31,
131
+ "65": 32,
132
+ "79": 33,
133
+ "87": 34,
134
+ "95": 35,
135
+ "10": 36,
136
+ "18": 37,
137
+ "25": 38,
138
+ "32": 39,
139
+ "37": 40,
140
+ "53": 41,
141
+ "54": 42,
142
+ "57": 43,
143
+ "59": 44,
144
+ "71": 45,
145
+ "72": 46,
146
+ "75": 47,
147
+ "85": 48,
148
+ "89": 49,
149
+ "91": 50,
150
+ "22": 51,
151
+ "24": 52,
152
+ "28": 53,
153
+ "35": 54,
154
+ "46": 55,
155
+ "64": 56,
156
+ "69": 57,
157
+ "78": 58,
158
+ "83": 59,
159
+ "84": 60,
160
+ "92": 61,
161
+ "94": 62,
162
+ "26": 63,
163
+ "27": 64,
164
+ "29": 65,
165
+ "30": 66,
166
+ "47": 67,
167
+ "49": 68,
168
+ "51": 69,
169
+ "58": 70,
170
+ "68": 71,
171
+ "73": 72,
172
+ "96": 73,
173
+ "13": 74,
174
+ "20": 75,
175
+ "23": 76,
176
+ "40": 77,
177
+ "61": 78,
178
+ "70": 79,
179
+ "82": 80,
180
+ "38": 81,
181
+ "74": 82,
182
+ "80": 83,
183
+ "98": 84,
184
+ "14": 85,
185
+ "41": 86,
186
+ "45": 87,
187
+ "52": 88,
188
+ "62": 89,
189
+ "63": 90,
190
+ "77": 91,
191
+ "86": 92,
192
+ "31": 93,
193
+ "34": 94,
194
+ "42": 95,
195
+ "43": 96,
196
+ "48": 97,
197
+ "76": 98,
198
+ "99": 99,
199
+ "15": 100,
200
+ "56": 101,
201
+ "97": 102,
202
+ "44": 103,
203
+ "67": 104
204
  },
205
  "merges": [
206
+ "5 0",
 
 
 
 
 
 
 
 
 
 
 
207
  "9 3",
208
+ "1 1",
209
+ "6 0",
 
 
 
210
  "1 9",
 
 
 
211
  "2 1",
212
+ "3 3",
213
+ "3 6",
214
+ "6 6",
215
  "8 8",
216
+ "1 2",
217
+ "1 7",
218
+ "8 1",
219
+ "9 0",
220
+ "1 6",
221
+ "3 9",
222
+ "5 5",
223
+ "6 5",
224
+ "7 9",
225
  "8 7",
226
+ "9 5",
227
+ "1 0",
228
+ "1 8",
229
+ "2 5",
230
+ "3 2",
231
+ "3 7",
232
+ "5 3",
233
+ "5 4",
234
+ "5 7",
235
+ "5 9",
236
+ "7 1",
237
+ "7 2",
238
+ "7 5",
239
  "8 5",
240
+ "8 9",
241
+ "9 1",
242
+ "2 2",
243
  "2 4",
244
+ "2 8",
245
+ "3 5",
246
+ "4 6",
247
+ "6 4",
248
+ "6 9",
249
+ "7 8",
250
+ "8 3",
251
  "8 4",
252
+ "9 2",
253
+ "9 4",
254
  "2 6",
 
255
  "2 7",
256
+ "2 9",
257
+ "3 0",
258
+ "4 7",
259
+ "4 9",
260
+ "5 1",
261
+ "5 8",
262
+ "6 8",
263
+ "7 3",
264
+ "9 6",
265
+ "1 3",
266
+ "2 0",
267
+ "2 3",
268
+ "4 0",
269
+ "6 1",
270
+ "7 0",
271
+ "8 2",
272
+ "3 8",
273
+ "7 4",
274
+ "8 0",
275
+ "9 8",
276
+ "1 4",
277
+ "4 1",
278
+ "4 5",
279
+ "5 2",
280
+ "6 2",
281
+ "6 3",
282
+ "7 7",
283
+ "8 6",
284
+ "3 1",
285
+ "3 4",
286
+ "4 2",
287
+ "4 3",
288
+ "4 8",
289
+ "7 6",
290
+ "9 9",
291
+ "1 5",
292
+ "5 6",
293
+ "9 7",
294
+ "4 4",
295
+ "6 7"
296
  ]
297
  }
298
  }
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:8118a074521f462d1f96fea1964baf76db0a332f1b116d058695b88f02ef809e
3
  size 4920
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1d2815be01cb68f3bfcac2870192039600e01a9bffcea46d39dfbe9c9424bf80
3
  size 4920