diff options
| author | YurenHao0426 <blackhao0426@gmail.com> | 2026-01-13 23:50:59 -0600 |
|---|---|---|
| committer | YurenHao0426 <blackhao0426@gmail.com> | 2026-01-13 23:50:59 -0600 |
| commit | 00cf667cee7ffacb144d5805fc7e0ef443f3583a (patch) | |
| tree | 77d20a3adaecf96bf3aff0612bdd3b5fa1a7dc7e /runs/slurm_logs/15261459_extreme3.out | |
| parent | c53c04aa1d6ff75cb478a9498c370baa929c74b6 (diff) | |
| parent | cd99d6b874d9d09b3bb87b8485cc787885af71f1 (diff) | |
Merge master into main
Diffstat (limited to 'runs/slurm_logs/15261459_extreme3.out')
| -rw-r--r-- | runs/slurm_logs/15261459_extreme3.out | 91 |
1 files changed, 91 insertions, 0 deletions
diff --git a/runs/slurm_logs/15261459_extreme3.out b/runs/slurm_logs/15261459_extreme3.out new file mode 100644 index 0000000..24474a9 --- /dev/null +++ b/runs/slurm_logs/15261459_extreme3.out @@ -0,0 +1,91 @@ +============================================================ +EXTREME THRESHOLD 3.0 Experiment +Job ID: 15261459 | Node: gpub057 +Start: Sat Jan 3 09:41:02 CST 2026 +============================================================ +NVIDIA A40, 46068 MiB +============================================================ +================================================================================ +DEPTH SCALING BENCHMARK +================================================================================ +Dataset: cifar100 +Depths: [4, 8, 12, 16] +Timesteps: 4 +Epochs: 150 +λ_reg: 0.3, λ_target: -0.1 +Reg type: extreme, Warmup epochs: 10 +Device: cuda +================================================================================ + +Loading cifar100... +Classes: 100, Input: (3, 32, 32) +Train: 50000, Test: 10000 + +Depth configurations: [(4, '4×1'), (8, '4×2'), (12, '4×3'), (16, '4×4')] +Regularization type: extreme +Warmup epochs: 10 +Stable init: False +Lyapunov threshold: 3.0 + +============================================================ +Depth = 4 conv layers (4 stages × 1 blocks) +============================================================ + Vanilla: depth=4, params=1,756,836 + Epoch 10: train=0.499 test=0.440 σ=9.42e-01/3.56e-08 + Epoch 20: train=0.627 test=0.513 σ=5.78e-01/2.44e-08 + Epoch 30: train=0.704 test=0.534 σ=4.77e-01/1.99e-08 + Epoch 40: train=0.754 test=0.526 σ=4.17e-01/1.73e-08 + Epoch 50: train=0.797 test=0.567 σ=3.65e-01/1.54e-08 + Epoch 60: train=0.832 test=0.568 σ=3.36e-01/1.38e-08 + Epoch 70: train=0.857 test=0.575 σ=3.25e-01/1.30e-08 + Epoch 80: train=0.884 test=0.574 σ=3.04e-01/1.20e-08 + Epoch 90: train=0.906 test=0.594 σ=2.72e-01/1.09e-08 + Epoch 100: train=0.921 test=0.592 σ=2.62e-01/1.01e-08 + Epoch 110: train=0.932 test=0.604 σ=2.63e-01/1.02e-08 + Epoch 120: train=0.943 test=0.599 σ=2.45e-01/9.00e-09 + Epoch 130: train=0.948 test=0.599 σ=2.43e-01/8.81e-09 + Epoch 140: train=0.950 test=0.602 σ=2.28e-01/8.51e-09 + Epoch 150: train=0.952 test=0.602 σ=2.39e-01/8.68e-09 + Best test acc: 0.608 + Lyapunov: depth=4, params=1,756,836 + Epoch 10: train=0.491 test=0.407 λ=2.045 σ=9.24e-01/3.51e-08 + Epoch 20: train=0.622 test=0.410 λ=1.995 σ=5.75e-01/2.43e-08 + Epoch 30: train=0.700 test=0.558 λ=1.971 σ=4.72e-01/2.00e-08 + Epoch 40: train=0.750 test=0.522 λ=1.962 σ=4.09e-01/1.70e-08 + Epoch 50: train=0.794 test=0.563 λ=1.959 σ=3.69e-01/1.53e-08 + Epoch 60: train=0.826 test=0.575 λ=1.954 σ=3.38e-01/1.41e-08 + Epoch 70: train=0.858 test=0.587 λ=1.956 σ=3.33e-01/1.30e-08 + Epoch 80: train=0.878 test=0.596 λ=1.958 σ=3.00e-01/1.19e-08 + Epoch 90: train=0.901 test=0.592 λ=1.955 σ=2.81e-01/1.11e-08 + Epoch 100: train=0.919 test=0.598 λ=1.962 σ=2.62e-01/9.87e-09 + Epoch 110: train=0.929 test=0.605 λ=1.965 σ=2.66e-01/9.85e-09 + Epoch 120: train=0.939 test=0.609 λ=1.969 σ=2.48e-01/9.16e-09 + Epoch 130: train=0.945 test=0.614 λ=1.979 σ=2.43e-01/8.93e-09 + Epoch 140: train=0.948 test=0.614 λ=1.973 σ=2.36e-01/8.59e-09 + Epoch 150: train=0.950 test=0.615 λ=1.975 σ=2.33e-01/8.44e-09 + Best test acc: 0.621 + +============================================================ +Depth = 8 conv layers (4 stages × 2 blocks) +============================================================ + Vanilla: depth=8, params=4,892,196 + Epoch 10: train=0.391 test=0.318 σ=8.18e-01/3.12e-08 + Epoch 20: train=0.541 test=0.435 σ=4.75e-01/2.17e-08 + Epoch 30: train=0.631 test=0.468 σ=3.80e-01/1.77e-08 + Epoch 40: train=0.697 test=0.492 σ=3.33e-01/1.57e-08 + Epoch 50: train=0.753 test=0.514 σ=3.09e-01/1.41e-08 + Epoch 60: train=0.798 test=0.519 σ=2.84e-01/1.29e-08 + Epoch 70: train=0.837 test=0.527 σ=2.70e-01/1.17e-08 + Epoch 80: train=0.870 test=0.537 σ=2.49e-01/1.06e-08 + Epoch 90: train=0.897 test=0.537 σ=2.32e-01/9.53e-09 + Epoch 100: train=0.918 test=0.534 σ=2.27e-01/9.08e-09 + Epoch 110: train=0.933 test=0.540 σ=2.17e-01/8.41e-09 + Epoch 120: train=0.947 test=0.541 σ=2.06e-01/7.71e-09 + Epoch 130: train=0.954 test=0.546 σ=1.95e-01/7.44e-09 + Epoch 140: train=0.956 test=0.546 σ=2.03e-01/7.55e-09 + Epoch 150: train=0.957 test=0.541 σ=1.93e-01/7.46e-09 + Best test acc: 0.554 + Lyapunov: depth=8, params=4,892,196 + Epoch 10: train=0.213 test=0.140 λ=2.795 σ=6.78e-01/2.67e-08 + Epoch 20: train=0.247 test=0.178 λ=2.574 σ=3.65e-01/1.81e-08 + Epoch 30: train=0.359 test=0.247 λ=2.635 σ=3.31e-01/1.70e-08 |
