Upload config with huggingface_hub
Browse files
config
ADDED
|
@@ -0,0 +1,582 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
!ModelConfig
|
| 2 |
+
config_data: !DataConfig
|
| 3 |
+
data_statistics: !DataStatistics
|
| 4 |
+
average_len_target_per_bucket:
|
| 5 |
+
- 2.0
|
| 6 |
+
- 4.08629215026444
|
| 7 |
+
- 4.545226437122044
|
| 8 |
+
- 4.607174400985307
|
| 9 |
+
- 4.87920905184137
|
| 10 |
+
- 4.982440502718691
|
| 11 |
+
- 5.27049292873817
|
| 12 |
+
- 5.803196211897
|
| 13 |
+
- 6.028941176470578
|
| 14 |
+
- 6.834719710669081
|
| 15 |
+
- 7.802691790040373
|
| 16 |
+
- 15.045793000744581
|
| 17 |
+
- 8.642659279778403
|
| 18 |
+
- 8.640957446808516
|
| 19 |
+
- 10.678657074340526
|
| 20 |
+
- 9.13441955193482
|
| 21 |
+
- 9.305970149253735
|
| 22 |
+
- 8.974063400576362
|
| 23 |
+
- 10.971887550200805
|
| 24 |
+
- 13.464285714285717
|
| 25 |
+
- 13.232323232323239
|
| 26 |
+
- 11.324468085106396
|
| 27 |
+
- 13.984732824427489
|
| 28 |
+
- 16.142276422764233
|
| 29 |
+
- 16.61032863849766
|
| 30 |
+
- 16.427509293680295
|
| 31 |
+
- 16.655589123867063
|
| 32 |
+
- 19.258675078864364
|
| 33 |
+
- 21.614285714285707
|
| 34 |
+
- 21.643258426966298
|
| 35 |
+
- 20.918478260869566
|
| 36 |
+
- 22.957871396895783
|
| 37 |
+
- 23.638766519823815
|
| 38 |
+
- 24.167002012072434
|
| 39 |
+
- 25.362318840579718
|
| 40 |
+
- 25.95375722543352
|
| 41 |
+
- 26.010489510489492
|
| 42 |
+
- 27.22000000000001
|
| 43 |
+
- 26.959999999999987
|
| 44 |
+
- 27.297697368421066
|
| 45 |
+
- 28.040998217468793
|
| 46 |
+
- 30.042830540037233
|
| 47 |
+
- 29.966735966735982
|
| 48 |
+
- 30.934859154929573
|
| 49 |
+
- 30.47868217054262
|
| 50 |
+
- 30.527777777777782
|
| 51 |
+
- 31.492779783393505
|
| 52 |
+
- 32.51171874999999
|
| 53 |
+
- 32.6358024691358
|
| 54 |
+
- 34.461538461538474
|
| 55 |
+
- 32.60769230769231
|
| 56 |
+
- 5.5
|
| 57 |
+
- 43.0
|
| 58 |
+
- null
|
| 59 |
+
- null
|
| 60 |
+
- null
|
| 61 |
+
- null
|
| 62 |
+
- null
|
| 63 |
+
- null
|
| 64 |
+
- null
|
| 65 |
+
- null
|
| 66 |
+
- null
|
| 67 |
+
- null
|
| 68 |
+
- null
|
| 69 |
+
- null
|
| 70 |
+
buckets:
|
| 71 |
+
- !!python/tuple
|
| 72 |
+
- 8
|
| 73 |
+
- 8
|
| 74 |
+
- !!python/tuple
|
| 75 |
+
- 16
|
| 76 |
+
- 16
|
| 77 |
+
- !!python/tuple
|
| 78 |
+
- 24
|
| 79 |
+
- 24
|
| 80 |
+
- !!python/tuple
|
| 81 |
+
- 32
|
| 82 |
+
- 32
|
| 83 |
+
- !!python/tuple
|
| 84 |
+
- 40
|
| 85 |
+
- 40
|
| 86 |
+
- !!python/tuple
|
| 87 |
+
- 48
|
| 88 |
+
- 48
|
| 89 |
+
- !!python/tuple
|
| 90 |
+
- 56
|
| 91 |
+
- 56
|
| 92 |
+
- !!python/tuple
|
| 93 |
+
- 64
|
| 94 |
+
- 64
|
| 95 |
+
- !!python/tuple
|
| 96 |
+
- 72
|
| 97 |
+
- 72
|
| 98 |
+
- !!python/tuple
|
| 99 |
+
- 80
|
| 100 |
+
- 80
|
| 101 |
+
- !!python/tuple
|
| 102 |
+
- 88
|
| 103 |
+
- 88
|
| 104 |
+
- !!python/tuple
|
| 105 |
+
- 96
|
| 106 |
+
- 96
|
| 107 |
+
- !!python/tuple
|
| 108 |
+
- 104
|
| 109 |
+
- 104
|
| 110 |
+
- !!python/tuple
|
| 111 |
+
- 112
|
| 112 |
+
- 112
|
| 113 |
+
- !!python/tuple
|
| 114 |
+
- 120
|
| 115 |
+
- 120
|
| 116 |
+
- !!python/tuple
|
| 117 |
+
- 128
|
| 118 |
+
- 128
|
| 119 |
+
- !!python/tuple
|
| 120 |
+
- 136
|
| 121 |
+
- 129
|
| 122 |
+
- !!python/tuple
|
| 123 |
+
- 144
|
| 124 |
+
- 129
|
| 125 |
+
- !!python/tuple
|
| 126 |
+
- 152
|
| 127 |
+
- 129
|
| 128 |
+
- !!python/tuple
|
| 129 |
+
- 160
|
| 130 |
+
- 129
|
| 131 |
+
- !!python/tuple
|
| 132 |
+
- 168
|
| 133 |
+
- 129
|
| 134 |
+
- !!python/tuple
|
| 135 |
+
- 176
|
| 136 |
+
- 129
|
| 137 |
+
- !!python/tuple
|
| 138 |
+
- 184
|
| 139 |
+
- 129
|
| 140 |
+
- !!python/tuple
|
| 141 |
+
- 192
|
| 142 |
+
- 129
|
| 143 |
+
- !!python/tuple
|
| 144 |
+
- 200
|
| 145 |
+
- 129
|
| 146 |
+
- !!python/tuple
|
| 147 |
+
- 208
|
| 148 |
+
- 129
|
| 149 |
+
- !!python/tuple
|
| 150 |
+
- 216
|
| 151 |
+
- 129
|
| 152 |
+
- !!python/tuple
|
| 153 |
+
- 224
|
| 154 |
+
- 129
|
| 155 |
+
- !!python/tuple
|
| 156 |
+
- 232
|
| 157 |
+
- 129
|
| 158 |
+
- !!python/tuple
|
| 159 |
+
- 240
|
| 160 |
+
- 129
|
| 161 |
+
- !!python/tuple
|
| 162 |
+
- 248
|
| 163 |
+
- 129
|
| 164 |
+
- !!python/tuple
|
| 165 |
+
- 256
|
| 166 |
+
- 129
|
| 167 |
+
- !!python/tuple
|
| 168 |
+
- 264
|
| 169 |
+
- 129
|
| 170 |
+
- !!python/tuple
|
| 171 |
+
- 272
|
| 172 |
+
- 129
|
| 173 |
+
- !!python/tuple
|
| 174 |
+
- 280
|
| 175 |
+
- 129
|
| 176 |
+
- !!python/tuple
|
| 177 |
+
- 288
|
| 178 |
+
- 129
|
| 179 |
+
- !!python/tuple
|
| 180 |
+
- 296
|
| 181 |
+
- 129
|
| 182 |
+
- !!python/tuple
|
| 183 |
+
- 304
|
| 184 |
+
- 129
|
| 185 |
+
- !!python/tuple
|
| 186 |
+
- 312
|
| 187 |
+
- 129
|
| 188 |
+
- !!python/tuple
|
| 189 |
+
- 320
|
| 190 |
+
- 129
|
| 191 |
+
- !!python/tuple
|
| 192 |
+
- 328
|
| 193 |
+
- 129
|
| 194 |
+
- !!python/tuple
|
| 195 |
+
- 336
|
| 196 |
+
- 129
|
| 197 |
+
- !!python/tuple
|
| 198 |
+
- 344
|
| 199 |
+
- 129
|
| 200 |
+
- !!python/tuple
|
| 201 |
+
- 352
|
| 202 |
+
- 129
|
| 203 |
+
- !!python/tuple
|
| 204 |
+
- 360
|
| 205 |
+
- 129
|
| 206 |
+
- !!python/tuple
|
| 207 |
+
- 368
|
| 208 |
+
- 129
|
| 209 |
+
- !!python/tuple
|
| 210 |
+
- 376
|
| 211 |
+
- 129
|
| 212 |
+
- !!python/tuple
|
| 213 |
+
- 384
|
| 214 |
+
- 129
|
| 215 |
+
- !!python/tuple
|
| 216 |
+
- 392
|
| 217 |
+
- 129
|
| 218 |
+
- !!python/tuple
|
| 219 |
+
- 400
|
| 220 |
+
- 129
|
| 221 |
+
- !!python/tuple
|
| 222 |
+
- 408
|
| 223 |
+
- 129
|
| 224 |
+
- !!python/tuple
|
| 225 |
+
- 416
|
| 226 |
+
- 129
|
| 227 |
+
- !!python/tuple
|
| 228 |
+
- 424
|
| 229 |
+
- 129
|
| 230 |
+
- !!python/tuple
|
| 231 |
+
- 432
|
| 232 |
+
- 129
|
| 233 |
+
- !!python/tuple
|
| 234 |
+
- 440
|
| 235 |
+
- 129
|
| 236 |
+
- !!python/tuple
|
| 237 |
+
- 448
|
| 238 |
+
- 129
|
| 239 |
+
- !!python/tuple
|
| 240 |
+
- 456
|
| 241 |
+
- 129
|
| 242 |
+
- !!python/tuple
|
| 243 |
+
- 464
|
| 244 |
+
- 129
|
| 245 |
+
- !!python/tuple
|
| 246 |
+
- 472
|
| 247 |
+
- 129
|
| 248 |
+
- !!python/tuple
|
| 249 |
+
- 480
|
| 250 |
+
- 129
|
| 251 |
+
- !!python/tuple
|
| 252 |
+
- 488
|
| 253 |
+
- 129
|
| 254 |
+
- !!python/tuple
|
| 255 |
+
- 496
|
| 256 |
+
- 129
|
| 257 |
+
- !!python/tuple
|
| 258 |
+
- 504
|
| 259 |
+
- 129
|
| 260 |
+
- !!python/tuple
|
| 261 |
+
- 512
|
| 262 |
+
- 129
|
| 263 |
+
- !!python/tuple
|
| 264 |
+
- 513
|
| 265 |
+
- 129
|
| 266 |
+
length_ratio_mean: 0.16320710693441579
|
| 267 |
+
length_ratio_stats_per_bucket:
|
| 268 |
+
- !!python/tuple
|
| 269 |
+
- 0.3333333333333333
|
| 270 |
+
- 0.0
|
| 271 |
+
- !!python/tuple
|
| 272 |
+
- 0.28246393697985434
|
| 273 |
+
- 0.17868752447804973
|
| 274 |
+
- !!python/tuple
|
| 275 |
+
- 0.21840710265332788
|
| 276 |
+
- 0.1330505772378312
|
| 277 |
+
- !!python/tuple
|
| 278 |
+
- 0.16560142798704922
|
| 279 |
+
- 0.09581195473826641
|
| 280 |
+
- !!python/tuple
|
| 281 |
+
- 0.13801367492489092
|
| 282 |
+
- 0.12429965021659338
|
| 283 |
+
- !!python/tuple
|
| 284 |
+
- 0.11863212215522084
|
| 285 |
+
- 0.1208393385452983
|
| 286 |
+
- !!python/tuple
|
| 287 |
+
- 0.10151133866588294
|
| 288 |
+
- 0.11072333780515448
|
| 289 |
+
- !!python/tuple
|
| 290 |
+
- 0.09838819717267734
|
| 291 |
+
- 0.12557601720946082
|
| 292 |
+
- !!python/tuple
|
| 293 |
+
- 0.09394884997066442
|
| 294 |
+
- 0.16159177653077658
|
| 295 |
+
- !!python/tuple
|
| 296 |
+
- 0.10402554625981722
|
| 297 |
+
- 0.26083679437294416
|
| 298 |
+
- !!python/tuple
|
| 299 |
+
- 0.1012342945734544
|
| 300 |
+
- 0.20099512839826167
|
| 301 |
+
- !!python/tuple
|
| 302 |
+
- 0.17281772320739658
|
| 303 |
+
- 0.257545103018524
|
| 304 |
+
- !!python/tuple
|
| 305 |
+
- 0.10845391475564008
|
| 306 |
+
- 0.3239959561352876
|
| 307 |
+
- !!python/tuple
|
| 308 |
+
- 0.09935073708696769
|
| 309 |
+
- 0.3059573403277105
|
| 310 |
+
- !!python/tuple
|
| 311 |
+
- 0.12635011083619693
|
| 312 |
+
- 0.3907243857496131
|
| 313 |
+
- !!python/tuple
|
| 314 |
+
- 0.08949294838769961
|
| 315 |
+
- 0.3144956594612652
|
| 316 |
+
- !!python/tuple
|
| 317 |
+
- 0.07007331783529426
|
| 318 |
+
- 0.05774973922713284
|
| 319 |
+
- !!python/tuple
|
| 320 |
+
- 0.06411727035132861
|
| 321 |
+
- 0.04726139664600602
|
| 322 |
+
- !!python/tuple
|
| 323 |
+
- 0.07382609782015778
|
| 324 |
+
- 0.050521761108029695
|
| 325 |
+
- !!python/tuple
|
| 326 |
+
- 0.08590157138956556
|
| 327 |
+
- 0.05569960282284095
|
| 328 |
+
- !!python/tuple
|
| 329 |
+
- 0.08062786352616935
|
| 330 |
+
- 0.058790770283216505
|
| 331 |
+
- !!python/tuple
|
| 332 |
+
- 0.06563222282256796
|
| 333 |
+
- 0.05450274064413921
|
| 334 |
+
- !!python/tuple
|
| 335 |
+
- 0.07765999502184046
|
| 336 |
+
- 0.05057167990395854
|
| 337 |
+
- !!python/tuple
|
| 338 |
+
- 0.08555873649959676
|
| 339 |
+
- 0.05960592960682603
|
| 340 |
+
- !!python/tuple
|
| 341 |
+
- 0.08460026909745419
|
| 342 |
+
- 0.05238594583690578
|
| 343 |
+
- !!python/tuple
|
| 344 |
+
- 0.08041439956489124
|
| 345 |
+
- 0.04792844419538253
|
| 346 |
+
- !!python/tuple
|
| 347 |
+
- 0.07859928415542815
|
| 348 |
+
- 0.05072523580179588
|
| 349 |
+
- !!python/tuple
|
| 350 |
+
- 0.08720935617277471
|
| 351 |
+
- 0.04382289790728185
|
| 352 |
+
- !!python/tuple
|
| 353 |
+
- 0.0945785744419209
|
| 354 |
+
- 0.05412799726318098
|
| 355 |
+
- !!python/tuple
|
| 356 |
+
- 0.09147639273741005
|
| 357 |
+
- 0.04623369327444139
|
| 358 |
+
- !!python/tuple
|
| 359 |
+
- 0.08566937441195915
|
| 360 |
+
- 0.03729904156233976
|
| 361 |
+
- !!python/tuple
|
| 362 |
+
- 0.09089250053653752
|
| 363 |
+
- 0.03997864088737986
|
| 364 |
+
- !!python/tuple
|
| 365 |
+
- 0.09074467762227163
|
| 366 |
+
- 0.03321456251239067
|
| 367 |
+
- !!python/tuple
|
| 368 |
+
- 0.09005889312946631
|
| 369 |
+
- 0.03317108879820214
|
| 370 |
+
- !!python/tuple
|
| 371 |
+
- 0.09171046510420815
|
| 372 |
+
- 0.035431771966381115
|
| 373 |
+
- !!python/tuple
|
| 374 |
+
- 0.09113688227781093
|
| 375 |
+
- 0.03302006652634936
|
| 376 |
+
- !!python/tuple
|
| 377 |
+
- 0.08893255045731784
|
| 378 |
+
- 0.03625826248767216
|
| 379 |
+
- !!python/tuple
|
| 380 |
+
- 0.09050986731593187
|
| 381 |
+
- 0.03584124694886162
|
| 382 |
+
- !!python/tuple
|
| 383 |
+
- 0.08733945286928807
|
| 384 |
+
- 0.02963045027026122
|
| 385 |
+
- !!python/tuple
|
| 386 |
+
- 0.08629641342788141
|
| 387 |
+
- 0.030504004395265606
|
| 388 |
+
- !!python/tuple
|
| 389 |
+
- 0.08640318897032012
|
| 390 |
+
- 0.030675309542199148
|
| 391 |
+
- !!python/tuple
|
| 392 |
+
- 0.09036990980396087
|
| 393 |
+
- 0.03191849333402471
|
| 394 |
+
- !!python/tuple
|
| 395 |
+
- 0.08807433400728253
|
| 396 |
+
- 0.03290646726223996
|
| 397 |
+
- !!python/tuple
|
| 398 |
+
- 0.08866806731559626
|
| 399 |
+
- 0.03728726453805084
|
| 400 |
+
- !!python/tuple
|
| 401 |
+
- 0.08543343575292071
|
| 402 |
+
- 0.02718762541630789
|
| 403 |
+
- !!python/tuple
|
| 404 |
+
- 0.08370713516111211
|
| 405 |
+
- 0.030215511135668078
|
| 406 |
+
- !!python/tuple
|
| 407 |
+
- 0.08456814843330018
|
| 408 |
+
- 0.02546164231510412
|
| 409 |
+
- !!python/tuple
|
| 410 |
+
- 0.08555768005742968
|
| 411 |
+
- 0.026243070552042298
|
| 412 |
+
- !!python/tuple
|
| 413 |
+
- 0.08406679695149158
|
| 414 |
+
- 0.02438826052491033
|
| 415 |
+
- !!python/tuple
|
| 416 |
+
- 0.08695308402142944
|
| 417 |
+
- 0.02926245130243095
|
| 418 |
+
- !!python/tuple
|
| 419 |
+
- 0.08093989759976632
|
| 420 |
+
- 0.032639773078865474
|
| 421 |
+
- !!python/tuple
|
| 422 |
+
- 0.013364278458885503
|
| 423 |
+
- 0.0011988283372310012
|
| 424 |
+
- !!python/tuple
|
| 425 |
+
- 0.10311750599520383
|
| 426 |
+
- 0.007770672300249229
|
| 427 |
+
- &id001 !!python/tuple
|
| 428 |
+
- null
|
| 429 |
+
- null
|
| 430 |
+
- *id001
|
| 431 |
+
- *id001
|
| 432 |
+
- *id001
|
| 433 |
+
- *id001
|
| 434 |
+
- *id001
|
| 435 |
+
- *id001
|
| 436 |
+
- *id001
|
| 437 |
+
- *id001
|
| 438 |
+
- *id001
|
| 439 |
+
- *id001
|
| 440 |
+
- *id001
|
| 441 |
+
length_ratio_std: 0.1371393774100467
|
| 442 |
+
max_observed_len_source: 417
|
| 443 |
+
max_observed_len_target: 128
|
| 444 |
+
num_discarded: 20
|
| 445 |
+
num_sents: 354707
|
| 446 |
+
num_sents_per_bucket:
|
| 447 |
+
- 1
|
| 448 |
+
- 29122
|
| 449 |
+
- 57389
|
| 450 |
+
- 123411
|
| 451 |
+
- 36766
|
| 452 |
+
- 44876
|
| 453 |
+
- 21849
|
| 454 |
+
- 6758
|
| 455 |
+
- 8500
|
| 456 |
+
- 2765
|
| 457 |
+
- 3715
|
| 458 |
+
- 2686
|
| 459 |
+
- 722
|
| 460 |
+
- 752
|
| 461 |
+
- 417
|
| 462 |
+
- 491
|
| 463 |
+
- 402
|
| 464 |
+
- 347
|
| 465 |
+
- 249
|
| 466 |
+
- 196
|
| 467 |
+
- 297
|
| 468 |
+
- 376
|
| 469 |
+
- 262
|
| 470 |
+
- 246
|
| 471 |
+
- 213
|
| 472 |
+
- 269
|
| 473 |
+
- 331
|
| 474 |
+
- 317
|
| 475 |
+
- 350
|
| 476 |
+
- 356
|
| 477 |
+
- 368
|
| 478 |
+
- 451
|
| 479 |
+
- 454
|
| 480 |
+
- 497
|
| 481 |
+
- 483
|
| 482 |
+
- 519
|
| 483 |
+
- 572
|
| 484 |
+
- 500
|
| 485 |
+
- 550
|
| 486 |
+
- 608
|
| 487 |
+
- 561
|
| 488 |
+
- 537
|
| 489 |
+
- 481
|
| 490 |
+
- 568
|
| 491 |
+
- 516
|
| 492 |
+
- 468
|
| 493 |
+
- 554
|
| 494 |
+
- 512
|
| 495 |
+
- 486
|
| 496 |
+
- 455
|
| 497 |
+
- 130
|
| 498 |
+
- 2
|
| 499 |
+
- 4
|
| 500 |
+
- 0
|
| 501 |
+
- 0
|
| 502 |
+
- 0
|
| 503 |
+
- 0
|
| 504 |
+
- 0
|
| 505 |
+
- 0
|
| 506 |
+
- 0
|
| 507 |
+
- 0
|
| 508 |
+
- 0
|
| 509 |
+
- 0
|
| 510 |
+
- 0
|
| 511 |
+
- 0
|
| 512 |
+
num_tokens_source: 15626084
|
| 513 |
+
num_tokens_target: 2020240
|
| 514 |
+
num_unks_source: 1
|
| 515 |
+
num_unks_target: 116
|
| 516 |
+
size_vocab_source: 1232
|
| 517 |
+
size_vocab_target: 5976
|
| 518 |
+
eop_id: -1
|
| 519 |
+
max_seq_len_source: 513
|
| 520 |
+
max_seq_len_target: 129
|
| 521 |
+
num_source_factors: 1
|
| 522 |
+
num_target_factors: 1
|
| 523 |
+
config_decoder: !TransformerConfig
|
| 524 |
+
act_type: relu
|
| 525 |
+
attention_heads: 8
|
| 526 |
+
block_prepended_cross_attention: false
|
| 527 |
+
decoder_type: transformer
|
| 528 |
+
depth_key_value: 512
|
| 529 |
+
dropout_act: 0.1
|
| 530 |
+
dropout_attention: 0.1
|
| 531 |
+
dropout_prepost: 0.1
|
| 532 |
+
feed_forward_num_hidden: 2048
|
| 533 |
+
max_seq_len_source: 513
|
| 534 |
+
max_seq_len_target: 129
|
| 535 |
+
model_size: 512
|
| 536 |
+
num_layers: 6
|
| 537 |
+
positional_embedding_type: fixed
|
| 538 |
+
postprocess_sequence: dr
|
| 539 |
+
preprocess_sequence: n
|
| 540 |
+
use_glu: false
|
| 541 |
+
use_lhuc: false
|
| 542 |
+
config_embed_source: !EmbeddingConfig
|
| 543 |
+
allow_sparse_grad: false
|
| 544 |
+
dropout: 0.0
|
| 545 |
+
factor_configs: null
|
| 546 |
+
num_embed: 512
|
| 547 |
+
num_factors: 1
|
| 548 |
+
vocab_size: 1232
|
| 549 |
+
config_embed_target: !EmbeddingConfig
|
| 550 |
+
allow_sparse_grad: false
|
| 551 |
+
dropout: 0.0
|
| 552 |
+
factor_configs: null
|
| 553 |
+
num_embed: 512
|
| 554 |
+
num_factors: 1
|
| 555 |
+
vocab_size: 5976
|
| 556 |
+
config_encoder: !TransformerConfig
|
| 557 |
+
act_type: relu
|
| 558 |
+
attention_heads: 8
|
| 559 |
+
block_prepended_cross_attention: false
|
| 560 |
+
decoder_type: transformer
|
| 561 |
+
depth_key_value: 512
|
| 562 |
+
dropout_act: 0.1
|
| 563 |
+
dropout_attention: 0.1
|
| 564 |
+
dropout_prepost: 0.1
|
| 565 |
+
feed_forward_num_hidden: 2048
|
| 566 |
+
max_seq_len_source: 513
|
| 567 |
+
max_seq_len_target: 129
|
| 568 |
+
model_size: 512
|
| 569 |
+
num_layers: 6
|
| 570 |
+
positional_embedding_type: fixed
|
| 571 |
+
postprocess_sequence: dr
|
| 572 |
+
preprocess_sequence: n
|
| 573 |
+
use_glu: false
|
| 574 |
+
use_lhuc: false
|
| 575 |
+
config_length_task: null
|
| 576 |
+
dtype: float32
|
| 577 |
+
lhuc: false
|
| 578 |
+
neural_vocab_selection: null
|
| 579 |
+
neural_vocab_selection_block_loss: false
|
| 580 |
+
vocab_source_size: 1232
|
| 581 |
+
vocab_target_size: 5976
|
| 582 |
+
weight_tying_type: none
|