Create rapid_prototype_2_output_a50_n37.txt

rapid_prototype_2_output_a50_n37.txt

@@ -0,0 +1,134 @@
+=================================================================
+RAPID PROTOTYPE v2: Differentiation-Centered Bank
+=================================================================
+  Device: cuda
+
+=================================================================
+PHASE 0: EXTRACTION
+=================================================================
+  Captions: 20,000
+
+  Extracting: bert...
+Loading weights: 100%
+ 199/199 [00:00<00:00, 4355.73it/s, Materializing param=pooler.dense.weight]
+BertModel LOAD REPORT from: google-bert/bert-base-uncased
+Key                                        | Status     |  | 
+-------------------------------------------+------------+--+-
+cls.predictions.transform.LayerNorm.weight | UNEXPECTED |  | 
+cls.predictions.transform.LayerNorm.bias   | UNEXPECTED |  | 
+cls.predictions.transform.dense.weight     | UNEXPECTED |  | 
+cls.predictions.bias                       | UNEXPECTED |  | 
+cls.seq_relationship.weight                | UNEXPECTED |  | 
+cls.seq_relationship.bias                  | UNEXPECTED |  | 
+cls.predictions.transform.dense.bias       | UNEXPECTED |  | 
+
+Notes:
+- UNEXPECTED	:can be ignored when loading from different task/architecture; not ok if you expect identical arch.
+    bert: 100%|██████████| 157/157 [00:24<00:00,  6.52it/s]
+    Shape: torch.Size([20000, 768])
+
+  Extracting: modern...
+Loading weights: 100%
+ 134/134 [00:00<00:00, 4050.54it/s, Materializing param=layers.21.mlp_norm.weight]
+ModernBertModel LOAD REPORT from: answerdotai/ModernBERT-base
+Key               | Status     |  | 
+------------------+------------+--+-
+head.norm.weight  | UNEXPECTED |  | 
+head.dense.weight | UNEXPECTED |  | 
+decoder.bias      | UNEXPECTED |  | 
+
+Notes:
+- UNEXPECTED	:can be ignored when loading from different task/architecture; not ok if you expect identical arch.
+    modern: 100%|██████████| 157/157 [00:36<00:00,  4.26it/s]
+    Shape: torch.Size([20000, 768])
+
+=================================================================
+PHASE 0b: PROCRUSTES ALIGNMENT + CONSENSUS STATISTICS
+=================================================================
+  bert      : cos 1.0000 → 1.0000
+  modern    : cos -0.0025 → 0.4849
+  cos(consensus, bert): 0.9574
+  cos(consensus, modern): 0.9584
+
+  Measuring consensus statistics...
+    CV:       0.1316
+    Mean cos: 0.0009
+    Eff dim:  223.7
+    Spectral: [0.0203, 0.0193, 0.0167, 0.0147, 0.0144...]
+
+=================================================================
+PHASE 1: TRAIN STUDENT
+=================================================================
+  Student: 11,269,632 params
+  CV target: 0.1316
+  E1: 3s  loss=3.1731  t_acc=0.322  t_cos=0.287  v_acc=0.448  v_cos=0.437  v_cv=0.227
+  E2: 2s  loss=1.6553  t_acc=0.713  t_cos=0.470  v_acc=0.649  v_cos=0.518  v_cv=0.197
+  E3: 2s  loss=1.1581  t_acc=0.858  t_cos=0.531  v_acc=0.814  v_cos=0.566  v_cv=0.169
+  E4: 2s  loss=0.8765  t_acc=0.922  t_cos=0.567  v_acc=0.867  v_cos=0.598  v_cv=0.167
+  E5: 2s  loss=0.7007  t_acc=0.954  t_cos=0.593  v_acc=0.892  v_cos=0.612  v_cv=0.169
+
+  Student saved. v_cos=0.612, v_cv=0.169
+
+=================================================================
+PHASE 2: TRAIN ALIGNMENT BANK (student frozen)
+=================================================================
+  Pre-encoding through frozen student...
+  Student embeddings: torch.Size([18000, 768])
+    Expert 0 (bert): rotation + whitener loaded, cos_after=1.0000
+    Expert 1 (modern): rotation + whitener loaded, cos_after=0.4849
+    Anchors: 512 initialized from consensus embeddings
+    Targets: CV=0.1316, mean_cos=0.0009
+  Bank: 2,918,784 params
+  Bank targets: CV=0.1316, mean_cos=0.0009
+  E 1: 1s  loss=0.4916  v_loss=0.4484  agr=0.00001  ortho=0.00005  spread=0.02713  b_cv=0.2574  e_cv=0.1571  x_var=0.00238  a_max=0.632  exp=0.894±0.020
+  E 2: 1s  loss=0.4055  v_loss=0.3992  agr=0.00001  ortho=0.00008  spread=0.02210  b_cv=0.2201  e_cv=0.1541  x_var=0.00050  a_max=0.653  exp=0.910±0.021
+  E 3: 1s  loss=0.3751  v_loss=0.3778  agr=0.00001  ortho=0.00007  spread=0.02176  b_cv=0.2053  e_cv=0.1571  x_var=0.00058  a_max=0.654  exp=0.906±0.023
+  E 4: 1s  loss=0.3637  v_loss=0.3686  agr=0.00001  ortho=0.00009  spread=0.02109  b_cv=0.1990  e_cv=0.1611  x_var=0.00060  a_max=0.655  exp=0.890±0.032
+  E 5: 1s  loss=0.3575  v_loss=0.3670  agr=0.00001  ortho=0.00010  spread=0.02014  b_cv=0.1970  e_cv=0.1608  x_var=0.00050  a_max=0.655  exp=0.873±0.033
+  E 6: 1s  loss=0.3530  v_loss=0.3679  agr=0.00001  ortho=0.00011  spread=0.01906  b_cv=0.1993  e_cv=0.1539  x_var=0.00045  a_max=0.656  exp=0.848±0.045
+  E 7: 1s  loss=0.3504  v_loss=0.3512  agr=0.00001  ortho=0.00011  spread=0.01847  b_cv=0.1941  e_cv=0.1582  x_var=0.00043  a_max=0.656  exp=0.835±0.048
+  E 8: 1s  loss=0.3472  v_loss=0.3467  agr=0.00002  ortho=0.00011  spread=0.01799  b_cv=0.1894  e_cv=0.1562  x_var=0.00041  a_max=0.656  exp=0.832±0.047
+  E 9: 1s  loss=0.3454  v_loss=0.3396  agr=0.00002  ortho=0.00012  spread=0.01804  b_cv=0.1860  e_cv=0.1597  x_var=0.00056  a_max=0.656  exp=0.817±0.051
+  E10: 1s  loss=0.3434  v_loss=0.3447  agr=0.00001  ortho=0.00012  spread=0.01758  b_cv=0.1873  e_cv=0.1536  x_var=0.00050  a_max=0.656  exp=0.795±0.055
+  E11: 1s  loss=0.3418  v_loss=0.3351  agr=0.00001  ortho=0.00012  spread=0.01724  b_cv=0.1855  e_cv=0.1547  x_var=0.00053  a_max=0.656  exp=0.823±0.055
+  E12: 1s  loss=0.3403  v_loss=0.3520  agr=0.00001  ortho=0.00012  spread=0.01733  b_cv=0.1794  e_cv=0.1578  x_var=0.00045  a_max=0.656  exp=0.824±0.054
+  E13: 1s  loss=0.3408  v_loss=0.3619  agr=0.00001  ortho=0.00013  spread=0.01707  b_cv=0.1841  e_cv=0.1535  x_var=0.00049  a_max=0.656  exp=0.786±0.057
+  E14: 1s  loss=0.3404  v_loss=0.3545  agr=0.00001  ortho=0.00015  spread=0.01681  b_cv=0.1808  e_cv=0.1594  x_var=0.00060  a_max=0.656  exp=0.787±0.055
+  E15: 1s  loss=0.3379  v_loss=0.3419  agr=0.00001  ortho=0.00012  spread=0.01671  b_cv=0.1757  e_cv=0.1575  x_var=0.00058  a_max=0.656  exp=0.828±0.050
+  E16: 1s  loss=0.3374  v_loss=0.3486  agr=0.00001  ortho=0.00013  spread=0.01662  b_cv=0.1798  e_cv=0.1546  x_var=0.00061  a_max=0.656  exp=0.812±0.050
+  E17: 1s  loss=0.3367  v_loss=0.3522  agr=0.00001  ortho=0.00012  spread=0.01684  b_cv=0.1745  e_cv=0.1557  x_var=0.00048  a_max=0.656  exp=0.835±0.047
+  E18: 1s  loss=0.3368  v_loss=0.3625  agr=0.00001  ortho=0.00010  spread=0.01685  b_cv=0.1778  e_cv=0.1545  x_var=0.00041  a_max=0.657  exp=0.815±0.051
+  E19: 1s  loss=0.3373  v_loss=0.3441  agr=0.00001  ortho=0.00012  spread=0.01689  b_cv=0.1756  e_cv=0.1602  x_var=0.00047  a_max=0.656  exp=0.793±0.057
+  E20: 1s  loss=0.3365  v_loss=0.3377  agr=0.00001  ortho=0.00012  spread=0.01701  b_cv=0.1780  e_cv=0.1556  x_var=0.00051  a_max=0.656  exp=0.794±0.052
+
+=================================================================
+PHASE 3: GEOMETRIC VERIFICATION
+=================================================================
+  Passthrough:     1.000000 (target: 1.000)
+  Emb CV:          0.1592 (consensus: 0.1316)
+  Geo context CV:  0.1790
+  Geo eff_dim:     33.7 / 128
+  Expert cos:      0.794 ± 0.052
+  Anchor max cos:  0.656
+  Cross-expert:    0.032
+
+=================================================================
+PHASE 4: CLASSIFIER STABILITY TEST
+=================================================================
+  with_bank      : train=0.782  val=0.505  gap=0.277
+  without_bank   : train=0.512  val=0.372  gap=0.140
+
+=================================================================
+SUMMARY
+=================================================================
+  Consensus CV:     0.1316
+  Consensus eff_dim:223.7
+  Student v_cos:    0.612
+  Student v_cv:     0.169
+  Bank params:      2,918,784
+  Bank geo_eff_dim: 33.7
+  Bank geo_cv:      0.1790
+
+=================================================================
+DONE
+=================================================================