Create experiment_3_dual_teacher_output.txt
Browse files
experiment_2/experiment_3_dual_teacher_output.txt
ADDED
|
@@ -0,0 +1,66 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
=================================================================
|
| 2 |
+
DUAL-TEACHER PROCRUSTES CONSENSUS DISTILLATION
|
| 3 |
+
=================================================================
|
| 4 |
+
Device: cuda
|
| 5 |
+
|
| 6 |
+
Generating data...
|
| 7 |
+
Train: 15,000 Val: 3,000
|
| 8 |
+
|
| 9 |
+
=================================================================
|
| 10 |
+
STAGE 1A: TEACHER A — Raw Adam
|
| 11 |
+
=================================================================
|
| 12 |
+
[A] E 1: t=0.073 v=0.200 cv=1.3069
|
| 13 |
+
[A] E10: t=0.612 v=0.613 cv=1.4364
|
| 14 |
+
[A] E20: t=0.655 v=0.590 cv=1.4770
|
| 15 |
+
[A] E30: t=0.690 v=0.699 cv=1.3797
|
| 16 |
+
|
| 17 |
+
=================================================================
|
| 18 |
+
STAGE 1B: TEACHER B — Geometric (+spr+ort)
|
| 19 |
+
=================================================================
|
| 20 |
+
[B] E 1: t=0.072 v=0.184 cv=1.4589
|
| 21 |
+
[B] E10: t=0.578 v=0.606 cv=1.5603
|
| 22 |
+
[B] E20: t=0.614 v=0.667 cv=1.5950
|
| 23 |
+
[B] E30: t=0.658 v=0.649 cv=1.8004
|
| 24 |
+
|
| 25 |
+
=================================================================
|
| 26 |
+
STAGE 2: EXTRACT + PROCRUSTES ALIGN
|
| 27 |
+
=================================================================
|
| 28 |
+
Teacher A embeddings: torch.Size([15000, 768])
|
| 29 |
+
Teacher B embeddings: torch.Size([15000, 768])
|
| 30 |
+
Raw cos(A, B): 0.4360
|
| 31 |
+
GPA iter 1: delta=0.12673541
|
| 32 |
+
GPA iter 5: delta=0.01321763
|
| 33 |
+
GPA iter 10: delta=0.00224325
|
| 34 |
+
cos(consensus, a): 0.8251
|
| 35 |
+
cos(consensus, b): 0.8226
|
| 36 |
+
Consensus CV: 0.1774
|
| 37 |
+
Consensus anchors: torch.Size([30, 768])
|
| 38 |
+
Teacher A anchors cos: 0.0008
|
| 39 |
+
Teacher B anchors cos: -0.0160
|
| 40 |
+
|
| 41 |
+
=================================================================
|
| 42 |
+
STAGE 3: STUDENT — Consensus distillation + classification
|
| 43 |
+
=================================================================
|
| 44 |
+
E 1: t=0.081 v=0.203 cos=0.230 cv=1.1871 rig=4.8/34.4 [polygon=0.04 curve=0.00 star=0.36 structure=0.35]
|
| 45 |
+
E 5: t=0.610 v=0.618 cos=0.451 cv=0.6686 rig=12.9/98.8 [polygon=0.38 curve=0.83 star=0.67 structure=0.70]
|
| 46 |
+
E10: t=0.660 v=0.659 cos=0.550 cv=0.5453 rig=15.5/99.6 [polygon=0.41 curve=0.94 star=0.71 structure=0.72]
|
| 47 |
+
E15: t=0.711 v=0.702 cos=0.625 cv=0.4492 rig=18.7/97.8 [polygon=0.39 curve=0.88 star=0.93 structure=0.76]
|
| 48 |
+
E20: t=0.735 v=0.703 cos=0.671 cv=0.4598 rig=18.8/96.4 [polygon=0.45 curve=1.00 star=0.84 structure=0.70]
|
| 49 |
+
E25: t=0.745 v=0.736 cos=0.693 cv=0.4261 rig=18.3/92.9 [polygon=0.48 curve=1.00 star=0.92 structure=0.73]
|
| 50 |
+
E30: t=0.763 v=0.761 cos=0.704 cv=0.3359 rig=17.9/90.4 [polygon=0.50 curve=0.98 star=0.97 structure=0.76]
|
| 51 |
+
|
| 52 |
+
=================================================================
|
| 53 |
+
FINAL COMPARISON
|
| 54 |
+
=================================================================
|
| 55 |
+
|
| 56 |
+
Model v_acc cv poly curve star struct
|
| 57 |
+
-------------------------------------------------------
|
| 58 |
+
Teacher_A 0.699 1.4312 0.42 0.99 0.83 0.72
|
| 59 |
+
Teacher_B 0.649 1.5969 0.38 0.95 0.79 0.66
|
| 60 |
+
Student 0.761 0.3329 0.50 0.98 0.97 0.76
|
| 61 |
+
|
| 62 |
+
Student anchor drift from consensus: mean=0.4458 max=0.6453
|
| 63 |
+
|
| 64 |
+
=================================================================
|
| 65 |
+
DONE
|
| 66 |
+
=================================================================
|