File size: 13,060 Bytes
================================================================================
UNIVERSAL SVAE DIAGNOSTIC BATTERY
================================================================================
v16_johanna_omega/checkpoints/best.pt: 100%
 203M/203M [00:07<00:00, 62.8MB/s]
  Loading: /root/.cache/huggingface/hub/models--AbstractPhil--geolip-SVAE/snapshots/1befe15db2d45e53868547b00e98659aea6a20b1/v16_johanna_omega/checkpoints/best.pt
  Epoch: 40, MSE: 0.019033
  Config: {'V': 256, 'D': 16, 'patch_size': 16, 'hidden': 768, 'depth': 4, 'n_cross_layers': 2, 'target_cv': 0.125, 'dataset': 'omega_noise_16types_128', 'pretrained_from': 'v14_noise/epoch_0200.pt', 'img_size': 128, 'lr': 0.0001}
  Params: 16,942,419
  Resolution: 128×128, samples_per_test: 25

================================================================================
IMAGE DATASET BATTERY (128×128, n=25)
================================================================================
  dataset                       MSE        std        min        max |     S0     SD  ratio  erank
----------------------------------------------------------------------------------------------------
  CIFAR-10→128             0.001462   0.001278   0.000670   0.006966 |  4.745  3.270   1.45  15.92
  MNIST→128                0.001713   0.000479   0.000779   0.002767 |  4.743  3.274   1.45  15.92
  TinyImageNet→128         0.005592   0.007961   0.000530   0.034349 |  4.750  3.275   1.45  15.92
  ImageNet-128→128         0.002344   0.002470   0.000976   0.013297 |  4.746  3.265   1.45  15.92
Resolving data files: 100%
 40/40 [00:00<00:00, 12172.40it/s]
Resolving data files: 100%
 40/40 [00:00<00:00, 13518.02it/s]
  ImageNet-256→128         0.002214   0.002376   0.000841   0.012717 |  4.746  3.266   1.45  15.92

================================================================================
NOISE TYPE BATTERY (128×128, n=25)
================================================================================
  type                      MSE        std |     S0     SD  ratio  erank | byte_acc   ±1_acc
----------------------------------------------------------------------------------------------------
  gaussian             0.029415   0.001089 |  4.736  3.277   1.45  15.92 |    9.48%   27.57%
  uniform              0.010246   0.000487 |  4.735  3.279   1.44  15.92 |   15.65%   43.36%
  uniform_scaled       0.041399   0.001460 |  4.736  3.275   1.45  15.92 |    8.02%   23.50%
  poisson              0.005948   0.006064 |  4.734  3.280   1.44  15.92 |   23.32%   58.94%
  pink                 0.000261   0.000349 |  4.775  3.233   1.48  15.91 |   71.06%   99.15%
  brown                0.000377   0.000728 |  4.782  3.235   1.48  15.91 |   70.16%   96.65%
  salt_pepper          0.023009   0.000825 |  4.735  3.261   1.45  15.92 |    9.70%   28.32%
  sparse               0.014341   0.000402 |  4.736  3.278   1.44  15.92 |   14.08%   38.88%
  block                0.002015   0.000442 |  4.735  3.278   1.44  15.92 |   31.44%   75.25%
  gradient             0.009110   0.000343 |  4.752  3.264   1.46  15.92 |   15.93%   44.27%
  checkerboard         0.003647   0.000122 |  4.734  3.269   1.45  15.92 |   23.19%   60.60%
  mixed                0.010457   0.003563 |  4.735  3.279   1.44  15.92 |   15.98%   43.97%
  structural           0.018725   0.001017 |  4.732  3.230   1.46  15.91 |   28.97%   48.61%
  cauchy               0.061531   0.001877 |  4.732  3.269   1.45  15.92 |    6.73%   19.93%
  exponential          0.023947   0.000893 |  4.736  3.277   1.45  15.92 |   10.74%   30.75%
  laplace              0.052132   0.001897 |  4.737  3.274   1.45  15.92 |    7.62%   21.51%

================================================================================
TEXT BYTE RECONSTRUCTION (128×128)
================================================================================

  In:  'Hello, world! This is a test of the geometric encoder.'
  Out: 'D]be^'jjgg^Meae_bKh^njc_!e_YU\^``jjeT[kZd[]a)'
  MSE: 0.001004  Byte: 0.0%

  In:  'The quick brown fox jumps over the lazy dog. 0123456789'
  Out: 'NbZ_kWZbYland!Yek XkZgheoZe m_X_TocZg`+*,1+,0/)1*'
  MSE: 0.001007  Byte: 3.6%

  In:  'import torch; model = PatchSVAE(); output = model(x)'
  Out: 'befgbhfkf^b0clY]`-?Vl[`OI<>(#1dfi^lj5 ^f]a`%k''
  MSE: 0.001005  Byte: 3.8%

  In:  'E = mc² — Albert Einstein, theoretical physicist, 1905'
  Out: '@5Z\��(͂�:caZhh5_\gm\ch#je[eg^c`PXaedjia_^ij(+))0'
  MSE: 0.001006  Byte: 0.0%

  In:  'To be, or not to be, that is the question. — Shakespeare'
  Out: 'PfXW&djdkcikV]&hTZibjfb_cl]bkWeg&"�y�!N]Wb]bgTVlZ'
  MSE: 0.001006  Byte: 0.0%

  In:  '∫₀^∞ e^(-x²) dx = √π/2 — Gaussian integral'
  Out: 'Ӂ��stTΉ�!^Q$(u��&Vk-'Ά��y(2�}�<Pjlic\__lj[_hRa'
  MSE: 0.001008  Byte: 1.9%

  In:  '01101000 01100101 01101100 01101100 01101111 — binary hello'
  Out: '.-,*$)#'(-.&+,0)*+#*#(,(-'++1*'*"+#),*/)�|�WbcShc^^dg'
  MSE: 0.001002  Byte: 3.3%

  In:  'SELECT * FROM models WHERE cv BETWEEN 0.20 AND 0.23;'
  Out: 'N@E>5J <LL?cl[YegL9;M<]f:EKM>>>"&.(<C=/')/4'
  MSE: 0.001000  Byte: 1.9%

================================================================================
PIECEMEAL 512→128 TILED RECONSTRUCTION
================================================================================
  gaussian          : 16 tiles, MSE=0.028989
  uniform           : 16 tiles, MSE=0.010235
  pink              : 16 tiles, MSE=0.000095
  salt_pepper       : 16 tiles, MSE=0.022889
  cauchy            : 16 tiles, MSE=0.062530

================================================================================
SIGNAL ENERGY SURVIVAL
================================================================================
  source                   survival     SNR_dB     orig_E    recon_E
----------------------------------------------------------------------
  CIFAR-10→128               101.3%      28.5dB     1.0330     1.0463
  MNIST→128                   97.5%      26.5dB     0.7737     0.7547
  TinyImageNet→128           100.0%      21.4dB     0.7748     0.7751
  ImageNet-128→128           100.5%      27.7dB     1.3908     1.3980
Resolving data files: 100%
 40/40 [00:00<00:00, 13618.98it/s]
Resolving data files: 100%
 40/40 [00:00<00:00, 15745.86it/s]
  ImageNet-256→128           100.6%      27.9dB     1.3549     1.3630
  noise/gaussian               96.8%      15.3dB     1.0017     0.9701
  noise/pink                  100.4%      39.1dB     1.5183     1.5243
  noise/salt_pepper           101.7%      22.4dB     4.0096     4.0786
  noise/cauchy                 98.0%      16.8dB     2.9569     2.8966

================================================================================
ALPHA PROFILE
================================================================================

  Layer 0: mean=0.03755  max=0.04310  min=0.03551  std=0.001793
    α[ 0]: 0.03865  ████████████████████████████████████████████
    α[ 1]: 0.03915  █████████████████████████████████████████████
    α[ 2]: 0.03683  ██████████████████████████████████████████
    α[ 3]: 0.03551  █████████████████████████████████████████
    α[ 4]: 0.04310  █████████████████████████████████████████████████
    α[ 5]: 0.03675  ██████████████████████████████████████████
    α[ 6]: 0.03792  ███████████████████████████████████████████
    α[ 7]: 0.03684  ██████████████████████████████████████████
    α[ 8]: 0.03733  ███████████████████████████████████████████
    α[ 9]: 0.03574  █████████████████████████████████████████
    α[10]: 0.03686  ██████████████████████████████████████████
    α[11]: 0.03657  ██████████████████████████████████████████
    α[12]: 0.03738  ███████████████████████████████████████████
    α[13]: 0.03635  ██████████████████████████████████████████
    α[14]: 0.03868  ████████████████████████████████████████████
    α[15]: 0.03714  ███████████████████████████████████████████

  Layer 1: mean=0.04468  max=0.04837  min=0.04178  std=0.001530
    α[ 0]: 0.04363  █████████████████████████████████████████████
    α[ 1]: 0.04572  ███████████████████████████████████████████████
    α[ 2]: 0.04407  █████████████████████████████████████████████
    α[ 3]: 0.04506  ██████████████████████████████████████████████
    α[ 4]: 0.04837  █████████████████████████████████████████████████
    α[ 5]: 0.04680  ████████████████████████████████████████████████
    α[ 6]: 0.04377  █████████████████████████████████████████████
    α[ 7]: 0.04445  █████████████████████████████████████████████
    α[ 8]: 0.04391  █████████████████████████████████████████████
    α[ 9]: 0.04413  █████████████████████████████████████████████
    α[10]: 0.04483  ██████████████████████████████████████████████
    α[11]: 0.04178  ███████████████████████████████████████████
    α[12]: 0.04583  ███████████████████████████████████████████████
    α[13]: 0.04309  ████████████████████████████████████████████
    α[14]: 0.04433  █████████████████████████████████████████████
    α[15]: 0.04504  ██████████████████████████████████████████████

================================================================================
COMPRESSION METRICS
================================================================================
  Input:   128×128×3 = 49,152 values
  Latent:  16×64 = 1,024 omega tokens
  Ratio:   48.0:1
  8-bit: input=48.0KB  latent=1.0KB  ratio=48.0:1
  16-bit: input=96.0KB  latent=2.0KB  ratio=48.0:1
  32-bit: input=192.0KB  latent=4.0KB  ratio=48.0:1

================================================================================
RECONSTRUCTION GRID
================================================================================
Resolving data files: 100%
 40/40 [00:00<00:00, 13917.23it/s]
Resolving data files: 100%
 40/40 [00:00<00:00, 14231.25it/s]
  Saved: universal_diagnostic_grid.png

  Results: diagnostic_v16_johanna_omega.json

================================================================================
DIAGNOSTIC COMPLETE
================================================================================