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 — Noise Classification Prototype ====================================================================== Freckles: 2,557,539 params (frozen) Feature dim: 64 per patch Patches: 16×16 = 256 Classifier: 837,008 params (trainable) Architecture: d_model=128, heads=4, layers=4 Batch: 128, lr=0.0003, epochs=20 ====================================================================== Ep 1/20: 100%|████████████████████| 1250/1250 [00:32<00:00, 38.88it/s, loss=0.2265 acc=68.8%] ep 1 | loss=0.8137 train=68.8% val=83.2% | best=uniform=100% worst=brown=0% | 34s type acc ---------------------- gaussian 35.1% ███████ uniform 100.0% ████████████████████ uniform_sc 99.9% ███████████████████ poisson 76.6% ███████████████ pink 100.0% ████████████████████ brown 0.0% salt_pepper 100.0% ████████████████████ sparse 100.0% ████████████████████ block 94.7% ██████████████████ gradient 100.0% ████████████████████ checker 100.0% ████████████████████ mixed 31.4% ██████ structural 100.0% ████████████████████ cauchy 100.0% ████████████████████ exponential 94.1% ██████████████████ laplace 99.8% ███████████████████ Ep 2/20: 100%|████████████████████| 1250/1250 [00:31<00:00, 39.39it/s, loss=0.3427 acc=84.1%] ep 2 | loss=0.3447 train=84.1% val=90.3% | best=salt_pepper=100% worst=pink=1% | 34s Ep 3/20: 100%|████████████████████| 1250/1250 [00:31<00:00, 39.45it/s, loss=0.7114 acc=86.0%] ep 3 | loss=0.3002 train=86.0% val=80.8% | best=uniform=100% worst=pink=0% | 34s Ep 4/20: 100%|████████████████████| 1250/1250 [00:31<00:00, 39.44it/s, loss=0.1888 acc=86.7%] ep 4 | loss=0.2876 train=86.7% val=89.9% | best=pink=100% worst=brown=0% | 34s Ep 5/20: 100%|████████████████████| 1250/1250 [00:31<00:00, 39.54it/s, loss=0.4770 acc=88.7%] ep 5 | loss=0.2326 train=88.7% val=90.0% | best=uniform_sc=100% worst=pink=1% | 34s type acc ---------------------- gaussian 89.9% █████████████████ uniform 76.2% ███████████████ uniform_sc 100.0% ████████████████████ poisson 82.9% ████████████████ pink 0.7% brown 99.3% ███████████████████ salt_pepper 99.6% ███████████████████ sparse 100.0% ████████████████████ block 100.0% ████████████████████ gradient 100.0% ████████████████████ checker 100.0% ████████████████████ mixed 92.0% ██████████████████ structural 100.0% ████████████████████ cauchy 100.0% ████████████████████ exponential 99.9% ███████████████████ laplace 100.0% ████████████████████ Ep 6/20: 100%|████████████████████| 1250/1250 [00:31<00:00, 39.40it/s, loss=0.2170 acc=88.8%] ep 6 | loss=0.2268 train=88.8% val=88.3% | best=uniform_sc=100% worst=brown=0% | 34s Ep 7/20: 100%|████████████████████| 1250/1250 [00:31<00:00, 39.44it/s, loss=0.2996 acc=89.5%] ep 7 | loss=0.2065 train=89.5% val=87.7% | best=sparse=100% worst=pink=0% | 34s Ep 8/20: 100%|████████████████████| 1250/1250 [00:31<00:00, 39.41it/s, loss=0.5350 acc=90.0%] ep 8 | loss=0.1977 train=90.0% val=81.6% | best=brown=100% worst=pink=0% | 34s Ep 9/20: 100%|████████████████████| 1250/1250 [00:31<00:00, 39.45it/s, loss=0.2076 acc=90.6%] ep 9 | loss=0.1760 train=90.6% val=91.8% | best=salt_pepper=100% worst=brown=1% | 34s Ep 10/20: 100%|████████████████████| 1250/1250 [00:31<00:00, 39.44it/s, loss=0.2252 acc=91.0%] ep 10 | loss=0.1658 train=91.0% val=91.6% | best=uniform_sc=100% worst=pink=12% | 34s type acc ---------------------- gaussian 98.9% ███████████████████ uniform 99.8% ███████████████████ uniform_sc 100.0% ████████████████████ poisson 85.7% █████████████████ pink 12.3% ██ brown 87.3% █████████████████ salt_pepper 98.8% ███████████████████ sparse 100.0% ████████████████████ block 100.0% ████████████████████ gradient 100.0% ████████████████████ checker 100.0% ████████████████████ mixed 88.5% █████████████████ structural 100.0% ████████████████████ cauchy 100.0% ████████████████████ exponential 100.0% ████████████████████ laplace 94.9% ██████████████████ Ep 11/20: 100%|████████████████████| 1250/1250 [00:31<00:00, 39.50it/s, loss=0.1168 acc=91.4%] ep 11 | loss=0.1526 train=91.4% val=92.2% | best=gaussian=100% worst=pink=0% | 34s Ep 12/20: 100%|████████████████████| 1250/1250 [00:31<00:00, 39.47it/s, loss=0.1941 acc=91.6%] ep 12 | loss=0.1458 train=91.6% val=90.8% | best=uniform_sc=100% worst=brown=0% | 34s Ep 13/20: 100%|████████████████████| 1250/1250 [00:31<00:00, 39.46it/s, loss=0.1693 acc=91.8%] ep 13 | loss=0.1386 train=91.8% val=92.4% | best=uniform_sc=100% worst=pink=0% | 34s Ep 14/20: 100%|████████████████████| 1250/1250 [00:31<00:00, 39.42it/s, loss=0.1358 acc=92.1%] ep 14 | loss=0.1326 train=92.1% val=92.6% | best=gaussian=100% worst=brown=0% | 34s Ep 15/20: 100%|████████████████████| 1250/1250 [00:31<00:00, 39.33it/s, loss=0.1334 acc=92.3%] ep 15 | loss=0.1296 train=92.3% val=92.5% | best=salt_pepper=100% worst=brown=1% | 34s type acc ---------------------- gaussian 99.8% ███████████████████ uniform 99.9% ███████████████████ uniform_sc 99.9% ███████████████████ poisson 91.0% ██████████████████ pink 99.1% ███████████████████ brown 0.9% salt_pepper 100.0% ████████████████████ sparse 100.0% ████████████████████ block 100.0% ████████████████████ gradient 100.0% ████████████████████ checker 100.0% ████████████████████ mixed 89.6% █████████████████ structural 100.0% ████████████████████ cauchy 100.0% ████████████████████ exponential 99.6% ███████████████████ laplace 100.0% ████████████████████ Ep 16/20: 100%|████████████████████| 1250/1250 [00:31<00:00, 39.37it/s, loss=0.1978 acc=92.5%] ep 16 | loss=0.1211 train=92.5% val=92.5% | best=uniform_sc=100% worst=pink=17% | 34s Ep 17/20: 100%|████████████████████| 1250/1250 [00:31<00:00, 39.48it/s, loss=0.1129 acc=92.6%] ep 17 | loss=0.1194 train=92.6% val=92.7% | best=brown=100% worst=pink=0% | 34s Ep 18/20: 100%|████████████████████| 1250/1250 [00:31<00:00, 39.27it/s, loss=0.0775 acc=92.7%] ep 18 | loss=0.1165 train=92.7% val=92.7% | best=uniform_sc=100% worst=pink=20% | 34s Ep 19/20: 100%|████████████████████| 1250/1250 [00:31<00:00, 39.42it/s, loss=0.0913 acc=92.7%] ep 19 | loss=0.1147 train=92.7% val=92.8% | best=gaussian=100% worst=pink=24% | 34s Ep 20/20: 100%|████████████████████| 1250/1250 [00:31<00:00, 39.42it/s, loss=0.1398 acc=92.9%] ep 20 | loss=0.1128 train=92.9% val=92.8% | best=gaussian=100% worst=pink=33% | 34s type acc ---------------------- gaussian 100.0% ████████████████████ uniform 98.6% ███████████████████ uniform_sc 99.9% ███████████████████ poisson 94.3% ██████████████████ pink 33.4% ██████ brown 64.3% ████████████ salt_pepper 100.0% ████████████████████ sparse 100.0% ████████████████████ block 100.0% ████████████████████ gradient 100.0% ████████████████████ checker 100.0% ████████████████████ mixed 94.5% ██████████████████ structural 100.0% ████████████████████ cauchy 100.0% ████████████████████ exponential 100.0% ████████████████████ laplace 100.0% ████████████████████ ====================================================================== TRAINING COMPLETE Best val accuracy: 92.8% Classifier params: 837,008 Random chance: 6.2% ======================================================================