Source identity
| Field | Recorded value |
|---|---|
| Source type | archive member |
| Original record | prepare_observed_tables.py |
| Date | Not specified in source |
| Path or archive member | archives/libsrc-6c9313f7cdd5/Declaration_Study_v1/code/prepare_observed_tables.py |
| Original SHA-256 | 0212e2b7afd90d26250aaf094e86788518ec069303642e4a68f6691ce98e7f7b |
| Reading-copy SHA-256 | 0212e2b7afd90d26250aaf094e86788518ec069303642e4a68f6691ce98e7f7b |
| Archive source | {"sourceId": "libsrc-6c9313f7cdd5", "member": "Declaration_Study_v1/code/prepare_observed_tables.py"} |
Recovered source transcript
Download complete source text ↓59 lines · 4.3 KB
"""Record selected values returned by the source-pinned remote tensor calculation.
These are tabulated secondary-analysis outputs, not newly simulated data.
The complete 167-row, 16-column result is regenerated by checkpoint_analysis.py.
"""
from pathlib import Path
import csv,json,hashlib
ROOT=Path(__file__).resolve().parents[1]
rows=[
(1,0.0,None),
(2,0.0046036369444359855,0.3985287186299244),
(150,2.8504898167593167,0.6941041629381179),
(160,3.0040141585823124,0.7112780843117966),
(165,3.073528341079451,0.718911649558106),
(170,3.1384999275095744,0.7254468696989013),
(175,3.200147005459472,0.7311468532536727),
(180,3.2626957625018402,0.736751441068763),
(185,3.3271927363139024,0.7432716395944564),
(190,3.3905990592799675,0.751358065847903),
(200,3.506372033377101,0.7698784377188664),
(300,4.430487121740427,0.9127275074077696),
(400,4.890018376298764,0.9735345812636843),
(600,5.233161470448734,0.9989614239570225),
(790,5.324971464809088,0.9999999999429887),
(792,5.324990583460912,1.0)]
with (ROOT/'data/selected_checkpoint_metrics.csv').open('w',newline='') as f:
w=csv.writer(f);w.writerow(['step','effective_update_frobenius_norm','cosine_with_final_effective_update']);w.writerows(rows)
# Authors' factor-amplitude eligibility rule: retain if either A or B movement exceeds threshold.
values=[
(5,0,156,725,-.788852194988483,-.8127091218762812),
(5,.0001,152,725,-.788852194988483,-.8127091218762812),
(5,.0005,124,550,-.8681123933885858,-.8783802293795547),
(5,.001,86,430,-.9291340350042407,-.9338299212834139),
(5,.002,66,190,-.9345827374326346,-.9401716153531431),
(5,.005,30,10,-.9872944014719269,-.9870983111016234),
(10,0,154,725,-.7271026474442789,-.759444763918197),
(10,.0001,154,725,-.7271026474442789,-.759444763918197),
(10,.0005,139,635,-.768928181070317,-.7797311679079444),
(10,.001,121,630,-.7964121690283363,-.8132492072811176),
(10,.002,86,190,-.8062624523173085,-.8222161976197946),
(10,.005,59,190,-.8062624523173085,-.8222161976197946),
(15,0,152,190,-.6914293279094361,-.7155692011406185),
(15,.0001,152,190,-.6914293279094361,-.7155692011406185),
(15,.0005,146,190,-.6914293279094361,-.7155692011406185),
(15,.001,133,190,-.6914293279094361,-.7155692011406185),
(15,.002,108,190,-.6914293279094361,-.7155692011406185),
(15,.005,72,190,-.6914293279094361,-.7155692011406185)]
with (ROOT/'data/corner_sensitivity.csv').open('w',newline='') as f:
w=csv.writer(f);w.writerow(['lag','amplitude_threshold','eligible_centers','peak_step_B_and_effective','B_peak_cosine','effective_peak_cosine']);w.writerows(values)
comparisons={
'150_to_200':{'A_cosine':.9975956592106836,'B_cosine':.9646791461958234,'effective_cosine':.9623597287760218,'effective_distance':1.087474736567148},
'175_to_190':{'A_cosine':.9997845947432347,'B_cosine':.9964775753994548,'effective_cosine':.9962629288914651,'effective_distance':.3425923712149229},
'180_to_200':{'A_cosine':.9996580076932067,'B_cosine':.9951625812829412,'effective_cosine':.9948222433361337,'effective_distance':.42171972832701626},
'180_to_792':{'A_cosine':.9738402370012251,'B_cosine':.7565424112454661,'effective_cosine':.736751441068763,'effective_distance':3.6606458859649993},
'300_to_792':{'A_cosine':.9938960687382065,'B_cosine':.9183329486014732,'effective_cosine':.9127275074077696,'effective_distance':2.2176687042008556}}
(ROOT/'data/checkpoint_comparisons.json').write_text(json.dumps(comparisons,indent=2))
validation={'n_checkpoints':167,'n_training_steps':792,'n_eval_observations':79,'parameter_layer_index':21,'rank':1,'scale':64,'A_shape':[1,13824],'B_shape':[5120,1],'configurations_identical':True,'zero_effective_update_steps':[1],'effective_norm_strictly_increasing':True,'B_norm_strictly_increasing':True,'pca_variance_ratios':[.8109369033497655,.13896290956024188,.03310634943994352,.010324782029499699,.003771529911710381],'max_stable_vs_gram_corner_error':4.490124128064821e-7,'stable_corner_checks':462,'remote_derived_geometry_sha256':'6760abfbdf676bb5ad3bc85ea777bfe9313fcbcbe1a0ea692a06d57eece6f0f1','derived_table_scope':'The distributed checkpoint and sensitivity CSV files tabulate selected remote outputs. Recompute the full geometry using checkpoint_analysis.py and the public pinned sources.'}
(ROOT/'data/tensor_validation.json').write_text(json.dumps(validation,indent=2))
print('Selected checkpoint rows:',len(rows),'Sensitivity rows:',len(values))