File size: 24,853 Bytes
3b30195 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 | Note: Environment variable`HF_TOKEN` is set and is the current active token independently from the token you've just configured.
WARNING:huggingface_hub._login:Note: Environment variable`HF_TOKEN` is set and is the current active token independently from the token you've just configured.
======================================================================
OMEGA PROCESSOR — CIFAR-10 (Freckles features)
======================================================================
Freckles: 2,557,539 params (frozen)
Feature dim: 64
Classifier: 836,234 params
Architecture: d_model=128, heads=4, layers=4
CIFAR-10: 50K train, 10K test, 64×64
Batch: 128, lr=0.0003, epochs=30
======================================================================
100%|██████████| 170M/170M [00:01<00:00, 91.1MB/s]
Ep 1/30: 100%|████████████████████| 390/390 [00:10<00:00, 38.21it/s, loss=1.8147 acc=23.6%]
ep 1 | loss=2.0322 train=23.6% test=33.4% | best=truck=70% worst=bird=1% | 12s
class acc
----------------------
airplane 44.1% ████████
automobile 30.2% ██████
bird 1.5%
cat 22.7% ████
deer 10.3% ██
dog 34.8% ██████
frog 54.1% ██████████
horse 36.4% ███████
ship 30.6% ██████
truck 69.7% █████████████
Ep 2/30: 100%|████████████████████| 390/390 [00:09<00:00, 39.18it/s, loss=1.6721 acc=38.5%]
ep 2 | loss=1.6980 train=38.5% test=44.8% | best=automobile=74% worst=cat=16% | 11s
Ep 3/30: 100%|████████████████████| 390/390 [00:09<00:00, 39.22it/s, loss=1.4097 acc=45.5%]
ep 3 | loss=1.5104 train=45.5% test=49.5% | best=frog=70% worst=cat=25% | 11s
Ep 4/30: 100%|████████████████████| 390/390 [00:09<00:00, 39.03it/s, loss=1.4453 acc=50.1%]
ep 4 | loss=1.3840 train=50.1% test=54.4% | best=ship=81% worst=cat=25% | 11s
Ep 5/30: 100%|████████████████████| 390/390 [00:09<00:00, 39.01it/s, loss=1.0432 acc=54.3%]
ep 5 | loss=1.2717 train=54.3% test=56.5% | best=ship=80% worst=bird=37% | 11s
class acc
----------------------
airplane 67.3% █████████████
automobile 72.4% ██████████████
bird 36.7% ███████
cat 40.3% ████████
deer 57.8% ███████████
dog 39.3% ███████
frog 72.9% ██████████████
horse 59.3% ███████████
ship 80.0% ████████████████
truck 39.3% ███████
Ep 6/30: 100%|████████████████████| 390/390 [00:10<00:00, 38.27it/s, loss=1.1928 acc=57.3%]
ep 6 | loss=1.1949 train=57.3% test=59.5% | best=ship=81% worst=dog=44% | 11s
Ep 7/30: 100%|████████████████████| 390/390 [00:09<00:00, 39.01it/s, loss=1.3602 acc=59.9%]
ep 7 | loss=1.1243 train=59.9% test=59.7% | best=truck=78% worst=cat=23% | 11s
Ep 8/30: 100%|████████████████████| 390/390 [00:09<00:00, 39.22it/s, loss=1.2255 acc=61.8%]
ep 8 | loss=1.0698 train=61.8% test=62.1% | best=truck=74% worst=cat=44% | 11s
Ep 9/30: 100%|████████████████████| 390/390 [00:10<00:00, 38.57it/s, loss=1.2281 acc=63.4%]
ep 9 | loss=1.0276 train=63.4% test=64.0% | best=automobile=83% worst=cat=35% | 11s
Ep 10/30: 100%|████████████████████| 390/390 [00:09<00:00, 39.23it/s, loss=0.9413 acc=64.9%]
ep 10 | loss=0.9864 train=64.9% test=65.4% | best=ship=81% worst=bird=46% | 11s
class acc
----------------------
airplane 63.3% ████████████
automobile 76.2% ███████████████
bird 46.0% █████████
cat 51.0% ██████████
deer 60.8% ████████████
dog 51.8% ██████████
frog 73.0% ██████████████
horse 73.2% ██████████████
ship 81.5% ████████████████
truck 77.1% ███████████████
Ep 11/30: 100%|████████████████████| 390/390 [00:09<00:00, 39.06it/s, loss=0.9747 acc=66.2%]
ep 11 | loss=0.9497 train=66.2% test=66.0% | best=truck=81% worst=bird=47% | 11s
Ep 12/30: 100%|████████████████████| 390/390 [00:09<00:00, 39.04it/s, loss=0.9341 acc=67.2%]
ep 12 | loss=0.9211 train=67.2% test=65.3% | best=airplane=82% worst=cat=36% | 11s
Ep 13/30: 100%|████████████████████| 390/390 [00:09<00:00, 39.17it/s, loss=1.0120 acc=68.2%]
ep 13 | loss=0.8938 train=68.2% test=67.7% | best=frog=84% worst=cat=44% | 11s
Ep 14/30: 100%|████████████████████| 390/390 [00:09<00:00, 39.26it/s, loss=0.8321 acc=69.3%]
ep 14 | loss=0.8613 train=69.3% test=68.3% | best=automobile=83% worst=cat=50% | 11s
Ep 15/30: 100%|████████████████████| 390/390 [00:09<00:00, 39.07it/s, loss=0.8031 acc=70.3%]
ep 15 | loss=0.8360 train=70.3% test=68.7% | best=automobile=81% worst=cat=39% | 11s
class acc
----------------------
airplane 76.6% ███████████████
automobile 81.1% ████████████████
bird 52.9% ██████████
cat 39.1% ███████
deer 67.7% █████████████
dog 60.5% ████████████
frog 76.6% ███████████████
horse 75.8% ███████████████
ship 78.3% ███████████████
truck 78.7% ███████████████
Ep 16/30: 100%|████████████████████| 390/390 [00:09<00:00, 39.20it/s, loss=1.0140 acc=71.0%]
ep 16 | loss=0.8155 train=71.0% test=69.5% | best=frog=86% worst=cat=43% | 11s
Ep 17/30: 100%|████████████████████| 390/390 [00:09<00:00, 39.14it/s, loss=0.6555 acc=72.0%]
ep 17 | loss=0.7903 train=72.0% test=69.3% | best=automobile=86% worst=cat=43% | 11s
Ep 18/30: 100%|████████████████████| 390/390 [00:09<00:00, 39.01it/s, loss=0.6629 acc=72.9%]
ep 18 | loss=0.7687 train=72.9% test=70.1% | best=ship=84% worst=cat=53% | 11s
Ep 19/30: 100%|████████████████████| 390/390 [00:09<00:00, 39.17it/s, loss=0.6098 acc=73.3%]
ep 19 | loss=0.7525 train=73.3% test=70.9% | best=automobile=82% worst=cat=47% | 11s
Ep 20/30: 100%|████████████████████| 390/390 [00:10<00:00, 38.60it/s, loss=0.6530 acc=74.3%]
ep 20 | loss=0.7280 train=74.3% test=71.1% | best=airplane=85% worst=cat=52% | 11s
class acc
----------------------
airplane 84.9% ████████████████
automobile 82.0% ████████████████
bird 55.6% ███████████
cat 52.3% ██████████
deer 66.9% █████████████
dog 59.3% ███████████
frog 78.5% ███████████████
horse 73.4% ██████████████
ship 79.5% ███████████████
truck 78.3% ███████████████
Ep 21/30: 100%|████████████████████| 390/390 [00:09<00:00, 39.05it/s, loss=0.7255 acc=74.7%]
ep 21 | loss=0.7183 train=74.7% test=70.8% | best=frog=89% worst=cat=52% | 11s
Ep 22/30: 100%|████████████████████| 390/390 [00:09<00:00, 39.13it/s, loss=0.7117 acc=75.2%]
ep 22 | loss=0.7030 train=75.2% test=71.1% | best=automobile=87% worst=cat=49% | 11s
Ep 23/30: 100%|████████████████████| 390/390 [00:09<00:00, 39.06it/s, loss=0.6705 acc=75.5%]
ep 23 | loss=0.6897 train=75.5% test=71.0% | best=truck=83% worst=cat=47% | 11s
Ep 24/30: 100%|████████████████████| 390/390 [00:09<00:00, 39.19it/s, loss=0.6390 acc=76.0%]
ep 24 | loss=0.6786 train=76.0% test=72.0% | best=automobile=82% worst=cat=50% | 11s
Ep 25/30: 100%|████████████████████| 390/390 [00:09<00:00, 39.14it/s, loss=0.5925 acc=76.5%]
ep 25 | loss=0.6657 train=76.5% test=72.0% | best=automobile=86% worst=cat=52% | 11s
class acc
----------------------
airplane 81.2% ████████████████
automobile 86.2% █████████████████
bird 56.9% ███████████
cat 52.5% ██████████
deer 68.3% █████████████
dog 58.5% ███████████
frog 80.0% ████████████████
horse 76.0% ███████████████
ship 80.9% ████████████████
truck 79.1% ███████████████
Ep 26/30: 100%|████████████████████| 390/390 [00:09<00:00, 39.24it/s, loss=0.6332 acc=76.5%]
ep 26 | loss=0.6614 train=76.5% test=72.1% | best=frog=84% worst=cat=48% | 11s
Ep 27/30: 100%|████████████████████| 390/390 [00:09<00:00, 39.14it/s, loss=0.6383 acc=76.9%]
ep 27 | loss=0.6535 train=76.9% test=72.2% | best=automobile=86% worst=cat=50% | 11s
Ep 28/30: 100%|████████████████████| 390/390 [00:09<00:00, 39.15it/s, loss=0.7896 acc=77.2%]
ep 28 | loss=0.6451 train=77.2% test=72.6% | best=automobile=84% worst=cat=50% | 11s
Ep 29/30: 100%|████████████████████| 390/390 [00:09<00:00, 39.19it/s, loss=0.6121 acc=77.4%]
ep 29 | loss=0.6410 train=77.4% test=72.7% | best=automobile=84% worst=cat=51% | 11s
Ep 30/30: 100%|████████████████████| 390/390 [00:10<00:00, 38.99it/s, loss=0.7349 acc=77.5%]
ep 30 | loss=0.6390 train=77.5% test=72.7% | best=automobile=84% worst=cat=51% | 11s
class acc
----------------------
airplane 79.0% ███████████████
automobile 83.8% ████████████████
bird 59.8% ███████████
cat 50.9% ██████████
deer 68.4% █████████████
dog 63.1% ████████████
frog 82.0% ████████████████
horse 76.9% ███████████████
ship 82.2% ████████████████
truck 80.6% ████████████████
======================================================================
OMEGA PROCESSOR COMPLETE
Best test accuracy: 72.7%
Classifier params: 836,234
Random chance: 10.0%
======================================================================
======================================================================
BASELINE — CIFAR-10 (Raw patches, no Freckles)
======================================================================
Classifier: 867,050 params
Architecture: d_model=128, heads=4, layers=4
CIFAR-10: 50K train, 10K test, 64×64
Batch: 128, lr=0.0003, epochs=30
======================================================================
Ep 1/30: 100%|████████████████████| 390/390 [00:05<00:00, 65.28it/s, loss=1.6722 acc=32.2%]
ep 1 | loss=1.8088 train=32.2% test=42.3% | best=ship=62% worst=deer=18% | 6s
class acc
----------------------
airplane 47.6% █████████
automobile 59.5% ███████████
bird 32.3% ██████
cat 21.7% ████
deer 17.9% ███
dog 32.4% ██████
frog 61.7% ████████████
horse 41.3% ████████
ship 62.4% ████████████
truck 46.5% █████████
Ep 2/30: 100%|████████████████████| 390/390 [00:05<00:00, 65.53it/s, loss=1.3286 acc=46.4%]
ep 2 | loss=1.4489 train=46.4% test=51.9% | best=frog=66% worst=cat=20% | 6s
Ep 3/30: 100%|████████████████████| 390/390 [00:05<00:00, 65.29it/s, loss=1.2655 acc=53.2%]
ep 3 | loss=1.2781 train=53.2% test=56.3% | best=ship=75% worst=cat=28% | 6s
Ep 4/30: 100%|████████████████████| 390/390 [00:05<00:00, 65.42it/s, loss=1.1440 acc=57.9%]
ep 4 | loss=1.1618 train=57.9% test=60.5% | best=frog=82% worst=bird=34% | 6s
Ep 5/30: 100%|████████████████████| 390/390 [00:06<00:00, 64.95it/s, loss=1.1496 acc=61.4%]
ep 5 | loss=1.0723 train=61.4% test=63.2% | best=horse=79% worst=cat=31% | 6s
class acc
----------------------
airplane 69.9% █████████████
automobile 76.8% ███████████████
bird 38.6% ███████
cat 30.7% ██████
deer 56.0% ███████████
dog 62.7% ████████████
frog 72.4% ██████████████
horse 78.7% ███████████████
ship 74.0% ██████████████
truck 72.2% ██████████████
Ep 6/30: 100%|████████████████████| 390/390 [00:05<00:00, 65.25it/s, loss=0.8846 acc=64.0%]
ep 6 | loss=1.0052 train=64.0% test=62.7% | best=frog=89% worst=bird=32% | 6s
Ep 7/30: 100%|████████████████████| 390/390 [00:05<00:00, 65.29it/s, loss=0.9618 acc=66.5%]
ep 7 | loss=0.9418 train=66.5% test=65.2% | best=truck=87% worst=cat=32% | 6s
Ep 8/30: 100%|████████████████████| 390/390 [00:05<00:00, 65.03it/s, loss=0.9562 acc=68.3%]
ep 8 | loss=0.8913 train=68.3% test=68.0% | best=ship=87% worst=cat=45% | 6s
Ep 9/30: 100%|████████████████████| 390/390 [00:05<00:00, 65.12it/s, loss=0.8684 acc=69.9%]
ep 9 | loss=0.8519 train=69.9% test=68.6% | best=truck=85% worst=cat=40% | 6s
Ep 10/30: 100%|████████████████████| 390/390 [00:05<00:00, 65.12it/s, loss=0.6724 acc=71.3%]
ep 10 | loss=0.8107 train=71.3% test=68.6% | best=ship=84% worst=dog=46% | 6s
class acc
----------------------
airplane 76.5% ███████████████
automobile 64.8% ████████████
bird 53.8% ██████████
cat 54.7% ██████████
deer 72.6% ██████████████
dog 45.9% █████████
frog 78.4% ███████████████
horse 76.2% ███████████████
ship 84.2% ████████████████
truck 78.7% ███████████████
Ep 11/30: 100%|████████████████████| 390/390 [00:05<00:00, 65.00it/s, loss=0.7053 acc=72.5%]
ep 11 | loss=0.7744 train=72.5% test=71.5% | best=automobile=85% worst=cat=47% | 6s
Ep 12/30: 100%|████████████████████| 390/390 [00:05<00:00, 65.48it/s, loss=0.7007 acc=73.8%]
ep 12 | loss=0.7421 train=73.8% test=71.8% | best=truck=88% worst=cat=50% | 6s
Ep 13/30: 100%|████████████████████| 390/390 [00:06<00:00, 64.77it/s, loss=0.6695 acc=74.9%]
ep 13 | loss=0.7130 train=74.9% test=72.3% | best=automobile=90% worst=cat=49% | 6s
Ep 14/30: 100%|████████████████████| 390/390 [00:06<00:00, 64.57it/s, loss=0.6952 acc=75.6%]
ep 14 | loss=0.6917 train=75.6% test=72.6% | best=automobile=90% worst=dog=48% | 6s
Ep 15/30: 100%|████████████████████| 390/390 [00:05<00:00, 65.06it/s, loss=0.5546 acc=76.7%]
ep 15 | loss=0.6617 train=76.7% test=73.7% | best=automobile=91% worst=dog=60% | 6s
class acc
----------------------
airplane 80.5% ████████████████
automobile 91.3% ██████████████████
bird 68.1% █████████████
cat 60.3% ████████████
deer 66.4% █████████████
dog 60.0% ████████████
frog 79.0% ███████████████
horse 76.0% ███████████████
ship 85.4% █████████████████
truck 70.2% ██████████████
Ep 16/30: 100%|████████████████████| 390/390 [00:05<00:00, 65.05it/s, loss=0.6078 acc=77.6%]
ep 16 | loss=0.6364 train=77.6% test=74.0% | best=automobile=88% worst=cat=55% | 6s
Ep 17/30: 100%|████████████████████| 390/390 [00:05<00:00, 65.26it/s, loss=0.7195 acc=78.4%]
ep 17 | loss=0.6122 train=78.4% test=74.8% | best=automobile=89% worst=cat=60% | 6s
Ep 18/30: 100%|████████████████████| 390/390 [00:05<00:00, 65.46it/s, loss=0.5613 acc=79.2%]
ep 18 | loss=0.5941 train=79.2% test=75.1% | best=automobile=90% worst=cat=57% | 6s
Ep 19/30: 100%|████████████████████| 390/390 [00:05<00:00, 65.23it/s, loss=0.4989 acc=79.7%]
ep 19 | loss=0.5762 train=79.7% test=75.1% | best=automobile=90% worst=cat=60% | 6s
Ep 20/30: 100%|████████████████████| 390/390 [00:05<00:00, 65.22it/s, loss=0.6564 acc=80.5%]
ep 20 | loss=0.5553 train=80.5% test=75.8% | best=automobile=87% worst=bird=60% | 6s
class acc
----------------------
airplane 81.0% ████████████████
automobile 87.0% █████████████████
bird 59.8% ███████████
cat 62.9% ████████████
deer 71.4% ██████████████
dog 63.7% ████████████
frog 85.0% █████████████████
horse 79.0% ███████████████
ship 85.2% █████████████████
truck 82.6% ████████████████
Ep 21/30: 100%|████████████████████| 390/390 [00:05<00:00, 65.23it/s, loss=0.5616 acc=81.1%]
ep 21 | loss=0.5364 train=81.1% test=75.2% | best=truck=88% worst=cat=53% | 6s
Ep 22/30: 100%|████████████████████| 390/390 [00:05<00:00, 65.54it/s, loss=0.4172 acc=81.8%]
ep 22 | loss=0.5201 train=81.8% test=76.4% | best=automobile=88% worst=cat=56% | 6s
Ep 23/30: 100%|████████████████████| 390/390 [00:06<00:00, 64.75it/s, loss=0.6296 acc=82.2%]
ep 23 | loss=0.5099 train=82.2% test=76.2% | best=automobile=88% worst=cat=65% | 6s
Ep 24/30: 100%|████████████████████| 390/390 [00:06<00:00, 64.50it/s, loss=0.4983 acc=82.9%]
ep 24 | loss=0.4906 train=82.9% test=76.4% | best=ship=87% worst=cat=60% | 6s
Ep 25/30: 100%|████████████████████| 390/390 [00:05<00:00, 65.10it/s, loss=0.5062 acc=83.0%]
ep 25 | loss=0.4863 train=83.0% test=76.6% | best=automobile=89% worst=cat=56% | 6s
class acc
----------------------
airplane 80.5% ████████████████
automobile 89.4% █████████████████
bird 62.5% ████████████
cat 55.9% ███████████
deer 73.8% ██████████████
dog 71.9% ██████████████
frog 86.9% █████████████████
horse 78.0% ███████████████
ship 83.2% ████████████████
truck 84.0% ████████████████
Ep 26/30: 100%|████████████████████| 390/390 [00:05<00:00, 65.21it/s, loss=0.4450 acc=83.4%]
ep 26 | loss=0.4762 train=83.4% test=77.0% | best=truck=87% worst=cat=58% | 6s
Ep 27/30: 100%|████████████████████| 390/390 [00:05<00:00, 65.23it/s, loss=0.5349 acc=83.5%]
ep 27 | loss=0.4720 train=83.5% test=77.2% | best=automobile=87% worst=cat=59% | 6s
Ep 28/30: 100%|████████████████████| 390/390 [00:05<00:00, 65.20it/s, loss=0.4860 acc=83.8%]
ep 28 | loss=0.4634 train=83.8% test=76.9% | best=automobile=87% worst=cat=59% | 6s
Ep 29/30: 100%|████████████████████| 390/390 [00:05<00:00, 65.58it/s, loss=0.3822 acc=84.0%]
ep 29 | loss=0.4596 train=84.0% test=77.0% | best=automobile=88% worst=cat=59% | 6s
Ep 30/30: 100%|████████████████████| 390/390 [00:05<00:00, 65.01it/s, loss=0.3785 acc=84.1%]
ep 30 | loss=0.4574 train=84.1% test=77.1% | best=automobile=87% worst=cat=59% | 6s
class acc
----------------------
airplane 79.7% ███████████████
automobile 87.2% █████████████████
bird 65.3% █████████████
cat 58.8% ███████████
deer 74.4% ██████████████
dog 70.8% ██████████████
frog 84.1% ████████████████
horse 80.2% ████████████████
ship 84.5% ████████████████
truck 86.1% █████████████████
======================================================================
BASELINE COMPLETE
Best test accuracy: 77.2%
Classifier params: 867,050
Random chance: 10.0%
======================================================================
======================================================================
HEAD-TO-HEAD COMPARISON
======================================================================
Omega Processor (Freckles features): 72.7%
Baseline (raw patches): 77.2%
Delta: -4.5%
Random chance: 10.0%
====================================================================== |