AbstractPhil commited on
Commit
2e421bd
·
verified ·
1 Parent(s): 9deff4a

Ablation H-MID-H1_svd_fp64-s2

Browse files
H/MID/H1_svd_fp64/seed2/epoch_1_checkpoint.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:89a8a75fd7a6b61f6606be73a8c83bfa5eb0b897cd528e1b3b9e48c92b677070
3
+ size 2254319
H/MID/H1_svd_fp64/seed2/final_report.json ADDED
@@ -0,0 +1,575 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "config": {
3
+ "group": "H",
4
+ "variant": "H1_svd_fp64",
5
+ "band": "MID",
6
+ "seed": 2,
7
+ "phase": 2,
8
+ "num_epochs": 1,
9
+ "batch_size": 256,
10
+ "overrides": {
11
+ "svd": "fp64"
12
+ },
13
+ "description": "H-MID-H1_svd_fp64-s2"
14
+ },
15
+ "run_config": {
16
+ "matrix_v": 64,
17
+ "D": 8,
18
+ "patch_size": 16,
19
+ "hidden": 64,
20
+ "depth": 1,
21
+ "n_cross_layers": 1,
22
+ "n_heads": 4,
23
+ "max_alpha": 0.2,
24
+ "alpha_init": -2.0,
25
+ "img_size": 64,
26
+ "batch_size": 128,
27
+ "lr": 0.0001,
28
+ "epochs": 1,
29
+ "weight_decay": 0.0,
30
+ "use_cv_ema": true,
31
+ "cv_ema_alpha": 0.01,
32
+ "cv_alignment_epochs": 0,
33
+ "cv_measure_every": 50,
34
+ "cv_sigma_scale": 0.3,
35
+ "boost": 0.5,
36
+ "cross_attn_clip": 0.5,
37
+ "allowed_types": [
38
+ 0,
39
+ 1,
40
+ 2,
41
+ 3,
42
+ 4,
43
+ 5,
44
+ 6,
45
+ 7,
46
+ 8,
47
+ 9,
48
+ 10,
49
+ 11,
50
+ 12,
51
+ 13,
52
+ 14,
53
+ 15
54
+ ],
55
+ "train_size": 1000000,
56
+ "val_size": 10000,
57
+ "num_workers": 2,
58
+ "report_every": 100,
59
+ "major_report_every": 10,
60
+ "save_every": 5,
61
+ "seed": 2,
62
+ "hf_repo": "AbstractPhil/geolip-svae-batteries",
63
+ "upload": false
64
+ },
65
+ "cv_ema_final": 0.35093768055444374,
66
+ "cv_last": 0.35780763007936683,
67
+ "predicted_band": "MID",
68
+ "expected_band": "MID",
69
+ "band_match": true,
70
+ "test_mse": 0.938351035118103,
71
+ "recon_ema": 0.9405226606858192,
72
+ "S0": 3.459611654281616,
73
+ "SD": 2.011322021484375,
74
+ "ratio": 1.720068492328098,
75
+ "erank": 7.885364532470703,
76
+ "observed_sphere_cv": 0.34849014724451893,
77
+ "uniform_sphere_cv_prediction": 0.35418994230276496,
78
+ "band_deviation": -0.005699795058246027,
79
+ "params_finite": true,
80
+ "num_epochs_run": 1,
81
+ "start_epoch": 0,
82
+ "per_epoch_metrics": [
83
+ {
84
+ "epoch": 1,
85
+ "test_mse": 0.938351035118103,
86
+ "cv_ema": 0.35093768055444374,
87
+ "observed_sphere_cv": 0.34849014724451893,
88
+ "band_deviation": -0.005699795058246027,
89
+ "erank": 7.885364532470703,
90
+ "params_finite": true,
91
+ "wallclock_seconds": 78.47087526321411,
92
+ "checkpoint_path": "/content/ablations_phase2/H/MID/H1_svd_fp64_s2/epoch_1_checkpoint.pt"
93
+ }
94
+ ],
95
+ "params_count": 183339,
96
+ "wallclock_seconds": 78.47218346595764,
97
+ "batches_completed": 3906,
98
+ "batch_limit": 3906,
99
+ "cv_trajectory": [
100
+ {
101
+ "batch": 0,
102
+ "cv": 0.3262890855598136,
103
+ "cv_ema": 0.3262890855598136,
104
+ "recon": 1.7023042440414429
105
+ },
106
+ {
107
+ "batch": 50,
108
+ "cv": 0.37712808861941644,
109
+ "cv_ema": 0.3267974755904096,
110
+ "recon": 1.5368989706039429
111
+ },
112
+ {
113
+ "batch": 100,
114
+ "cv": 0.3667197776665937,
115
+ "cv_ema": 0.3271966986111714,
116
+ "recon": 1.6507762670516968
117
+ },
118
+ {
119
+ "batch": 150,
120
+ "cv": 0.35794615370316873,
121
+ "cv_ema": 0.3275041931620914,
122
+ "recon": 1.2912814617156982
123
+ },
124
+ {
125
+ "batch": 200,
126
+ "cv": 0.35993818521877174,
127
+ "cv_ema": 0.3278285330826582,
128
+ "recon": 1.2025034427642822
129
+ },
130
+ {
131
+ "batch": 250,
132
+ "cv": 0.4102538413380495,
133
+ "cv_ema": 0.3286527861652121,
134
+ "recon": 1.209348440170288
135
+ },
136
+ {
137
+ "batch": 300,
138
+ "cv": 0.37714477255745366,
139
+ "cv_ema": 0.3291377060291345,
140
+ "recon": 1.1326255798339844
141
+ },
142
+ {
143
+ "batch": 350,
144
+ "cv": 0.35258503828538335,
145
+ "cv_ema": 0.329372179351697,
146
+ "recon": 1.0686964988708496
147
+ },
148
+ {
149
+ "batch": 400,
150
+ "cv": 0.35176822842108385,
151
+ "cv_ema": 0.3295961398423909,
152
+ "recon": 1.2096980810165405
153
+ },
154
+ {
155
+ "batch": 450,
156
+ "cv": 0.36047144171315065,
157
+ "cv_ema": 0.3299048928610985,
158
+ "recon": 1.103196382522583
159
+ },
160
+ {
161
+ "batch": 500,
162
+ "cv": 0.3705128107588807,
163
+ "cv_ema": 0.3303109720400763,
164
+ "recon": 0.9307100772857666
165
+ },
166
+ {
167
+ "batch": 550,
168
+ "cv": 0.35486413824762625,
169
+ "cv_ema": 0.3305565037021518,
170
+ "recon": 0.9779574275016785
171
+ },
172
+ {
173
+ "batch": 600,
174
+ "cv": 0.3625934681023829,
175
+ "cv_ema": 0.3308768733461541,
176
+ "recon": 1.071733832359314
177
+ },
178
+ {
179
+ "batch": 650,
180
+ "cv": 0.3979262603920208,
181
+ "cv_ema": 0.3315473672166128,
182
+ "recon": 1.0947874784469604
183
+ },
184
+ {
185
+ "batch": 700,
186
+ "cv": 0.41307888657952213,
187
+ "cv_ema": 0.3323626824102419,
188
+ "recon": 1.033246397972107
189
+ },
190
+ {
191
+ "batch": 750,
192
+ "cv": 0.3974472447606115,
193
+ "cv_ema": 0.3330135280337456,
194
+ "recon": 1.1484707593917847
195
+ },
196
+ {
197
+ "batch": 800,
198
+ "cv": 0.4000021401408734,
199
+ "cv_ema": 0.33368341415481695,
200
+ "recon": 1.0376172065734863
201
+ },
202
+ {
203
+ "batch": 850,
204
+ "cv": 0.38099994151502653,
205
+ "cv_ema": 0.33415657942841903,
206
+ "recon": 1.18581223487854
207
+ },
208
+ {
209
+ "batch": 900,
210
+ "cv": 0.40097072188961874,
211
+ "cv_ema": 0.33482472085303105,
212
+ "recon": 0.9679669141769409
213
+ },
214
+ {
215
+ "batch": 950,
216
+ "cv": 0.38105892641229755,
217
+ "cv_ema": 0.3352870629086237,
218
+ "recon": 0.9783198237419128
219
+ },
220
+ {
221
+ "batch": 1000,
222
+ "cv": 0.36286407315795816,
223
+ "cv_ema": 0.33556283301111706,
224
+ "recon": 0.9493536353111267
225
+ },
226
+ {
227
+ "batch": 1050,
228
+ "cv": 0.37076030965212775,
229
+ "cv_ema": 0.3359148077775272,
230
+ "recon": 1.010164499282837
231
+ },
232
+ {
233
+ "batch": 1100,
234
+ "cv": 0.3547761863768869,
235
+ "cv_ema": 0.3361034215635208,
236
+ "recon": 1.0217506885528564
237
+ },
238
+ {
239
+ "batch": 1150,
240
+ "cv": 0.37371827350301856,
241
+ "cv_ema": 0.3364795700829158,
242
+ "recon": 0.9661281108856201
243
+ },
244
+ {
245
+ "batch": 1200,
246
+ "cv": 0.3701923303682811,
247
+ "cv_ema": 0.33681669768576944,
248
+ "recon": 1.0261797904968262
249
+ },
250
+ {
251
+ "batch": 1250,
252
+ "cv": 0.394838569933056,
253
+ "cv_ema": 0.33739691640824226,
254
+ "recon": 0.8472411632537842
255
+ },
256
+ {
257
+ "batch": 1300,
258
+ "cv": 0.3914505178419726,
259
+ "cv_ema": 0.33793745242257955,
260
+ "recon": 1.0981405973434448
261
+ },
262
+ {
263
+ "batch": 1350,
264
+ "cv": 0.39860944690900807,
265
+ "cv_ema": 0.33854417236744383,
266
+ "recon": 0.8480774760246277
267
+ },
268
+ {
269
+ "batch": 1400,
270
+ "cv": 0.3841167612371389,
271
+ "cv_ema": 0.3389998982561408,
272
+ "recon": 1.063689112663269
273
+ },
274
+ {
275
+ "batch": 1450,
276
+ "cv": 0.3761534517283266,
277
+ "cv_ema": 0.33937143379086265,
278
+ "recon": 0.9391136169433594
279
+ },
280
+ {
281
+ "batch": 1500,
282
+ "cv": 0.38088459758946874,
283
+ "cv_ema": 0.3397865654288487,
284
+ "recon": 0.8814090490341187
285
+ },
286
+ {
287
+ "batch": 1550,
288
+ "cv": 0.35531168310042194,
289
+ "cv_ema": 0.33994181660556444,
290
+ "recon": 1.0959669351577759
291
+ },
292
+ {
293
+ "batch": 1600,
294
+ "cv": 0.3610350748972833,
295
+ "cv_ema": 0.3401527491884816,
296
+ "recon": 0.9970536231994629
297
+ },
298
+ {
299
+ "batch": 1650,
300
+ "cv": 0.3641365994966151,
301
+ "cv_ema": 0.34039258769156294,
302
+ "recon": 1.0750279426574707
303
+ },
304
+ {
305
+ "batch": 1700,
306
+ "cv": 0.38781365489484654,
307
+ "cv_ema": 0.3408667983635958,
308
+ "recon": 1.0883712768554688
309
+ },
310
+ {
311
+ "batch": 1750,
312
+ "cv": 0.38668324228243084,
313
+ "cv_ema": 0.3413249628027841,
314
+ "recon": 0.9062432050704956
315
+ },
316
+ {
317
+ "batch": 1800,
318
+ "cv": 0.38184090171104634,
319
+ "cv_ema": 0.3417301221918667,
320
+ "recon": 0.9083492755889893
321
+ },
322
+ {
323
+ "batch": 1850,
324
+ "cv": 0.38910893036072836,
325
+ "cv_ema": 0.3422039102735553,
326
+ "recon": 0.9841954112052917
327
+ },
328
+ {
329
+ "batch": 1900,
330
+ "cv": 0.4000260549442023,
331
+ "cv_ema": 0.3427821317202618,
332
+ "recon": 0.9962667226791382
333
+ },
334
+ {
335
+ "batch": 1950,
336
+ "cv": 0.3407442755441237,
337
+ "cv_ema": 0.34276175315850044,
338
+ "recon": 0.9424905776977539
339
+ },
340
+ {
341
+ "batch": 2000,
342
+ "cv": 0.39039210123826656,
343
+ "cv_ema": 0.34323805663929813,
344
+ "recon": 0.9692186117172241
345
+ },
346
+ {
347
+ "batch": 2050,
348
+ "cv": 0.3632888812035472,
349
+ "cv_ema": 0.3434385648849406,
350
+ "recon": 1.0863287448883057
351
+ },
352
+ {
353
+ "batch": 2100,
354
+ "cv": 0.3865748299470748,
355
+ "cv_ema": 0.34386992753556195,
356
+ "recon": 0.8175496459007263
357
+ },
358
+ {
359
+ "batch": 2150,
360
+ "cv": 0.3869726703991921,
361
+ "cv_ema": 0.34430095496419827,
362
+ "recon": 0.8929638862609863
363
+ },
364
+ {
365
+ "batch": 2200,
366
+ "cv": 0.3591068393945415,
367
+ "cv_ema": 0.3444490138085017,
368
+ "recon": 0.9088624715805054
369
+ },
370
+ {
371
+ "batch": 2250,
372
+ "cv": 0.3660494132772056,
373
+ "cv_ema": 0.3446650178031887,
374
+ "recon": 0.8652998805046082
375
+ },
376
+ {
377
+ "batch": 2300,
378
+ "cv": 0.3919273588939387,
379
+ "cv_ema": 0.34513764121409624,
380
+ "recon": 0.9463256597518921
381
+ },
382
+ {
383
+ "batch": 2350,
384
+ "cv": 0.38497323900523106,
385
+ "cv_ema": 0.3455359971920076,
386
+ "recon": 0.9454164505004883
387
+ },
388
+ {
389
+ "batch": 2400,
390
+ "cv": 0.3759599481100575,
391
+ "cv_ema": 0.3458402367011881,
392
+ "recon": 1.0527379512786865
393
+ },
394
+ {
395
+ "batch": 2450,
396
+ "cv": 0.3643684358468259,
397
+ "cv_ema": 0.3460255186926444,
398
+ "recon": 0.7556484937667847
399
+ },
400
+ {
401
+ "batch": 2500,
402
+ "cv": 0.3289970998923283,
403
+ "cv_ema": 0.3458552345046413,
404
+ "recon": 0.9153985977172852
405
+ },
406
+ {
407
+ "batch": 2550,
408
+ "cv": 0.33519201839728224,
409
+ "cv_ema": 0.3457486023435677,
410
+ "recon": 0.9187247157096863
411
+ },
412
+ {
413
+ "batch": 2600,
414
+ "cv": 0.4112656993565183,
415
+ "cv_ema": 0.34640377331369715,
416
+ "recon": 0.8211793303489685
417
+ },
418
+ {
419
+ "batch": 2650,
420
+ "cv": 0.36773203569261015,
421
+ "cv_ema": 0.34661705593748626,
422
+ "recon": 0.880821943283081
423
+ },
424
+ {
425
+ "batch": 2700,
426
+ "cv": 0.35978581317785147,
427
+ "cv_ema": 0.34674874350988993,
428
+ "recon": 1.0866174697875977
429
+ },
430
+ {
431
+ "batch": 2750,
432
+ "cv": 0.3952877203071792,
433
+ "cv_ema": 0.34723413327786284,
434
+ "recon": 0.962988555431366
435
+ },
436
+ {
437
+ "batch": 2800,
438
+ "cv": 0.3847646520632604,
439
+ "cv_ema": 0.34760943846571685,
440
+ "recon": 0.9300438761711121
441
+ },
442
+ {
443
+ "batch": 2850,
444
+ "cv": 0.34051894268354493,
445
+ "cv_ema": 0.3475385335078951,
446
+ "recon": 1.0651638507843018
447
+ },
448
+ {
449
+ "batch": 2900,
450
+ "cv": 0.37681693344429473,
451
+ "cv_ema": 0.3478313175072591,
452
+ "recon": 0.9571689367294312
453
+ },
454
+ {
455
+ "batch": 2950,
456
+ "cv": 0.3698431868057212,
457
+ "cv_ema": 0.34805143620024376,
458
+ "recon": 0.9835600852966309
459
+ },
460
+ {
461
+ "batch": 3000,
462
+ "cv": 0.36022727432532375,
463
+ "cv_ema": 0.3481731945814946,
464
+ "recon": 1.1070382595062256
465
+ },
466
+ {
467
+ "batch": 3050,
468
+ "cv": 0.36151602623204804,
469
+ "cv_ema": 0.34830662289800013,
470
+ "recon": 0.9987481832504272
471
+ },
472
+ {
473
+ "batch": 3100,
474
+ "cv": 0.36308597939958875,
475
+ "cv_ema": 0.34845441646301606,
476
+ "recon": 0.921568751335144
477
+ },
478
+ {
479
+ "batch": 3150,
480
+ "cv": 0.35416388797828735,
481
+ "cv_ema": 0.34851151117816875,
482
+ "recon": 0.7822057604789734
483
+ },
484
+ {
485
+ "batch": 3200,
486
+ "cv": 0.381883746626373,
487
+ "cv_ema": 0.3488452335326508,
488
+ "recon": 0.9269399046897888
489
+ },
490
+ {
491
+ "batch": 3250,
492
+ "cv": 0.365737656520453,
493
+ "cv_ema": 0.3490141577625288,
494
+ "recon": 0.8798049688339233
495
+ },
496
+ {
497
+ "batch": 3300,
498
+ "cv": 0.3931002507791081,
499
+ "cv_ema": 0.34945501869269463,
500
+ "recon": 0.8593133687973022
501
+ },
502
+ {
503
+ "batch": 3350,
504
+ "cv": 0.3437768219658751,
505
+ "cv_ema": 0.3493982367254264,
506
+ "recon": 0.961512565612793
507
+ },
508
+ {
509
+ "batch": 3400,
510
+ "cv": 0.3423002675598682,
511
+ "cv_ema": 0.3493272570337708,
512
+ "recon": 1.0110037326812744
513
+ },
514
+ {
515
+ "batch": 3450,
516
+ "cv": 0.37100556374991844,
517
+ "cv_ema": 0.3495440401009323,
518
+ "recon": 1.0252983570098877
519
+ },
520
+ {
521
+ "batch": 3500,
522
+ "cv": 0.3645964638238607,
523
+ "cv_ema": 0.3496945643381616,
524
+ "recon": 1.180680274963379
525
+ },
526
+ {
527
+ "batch": 3550,
528
+ "cv": 0.38209629882699453,
529
+ "cv_ema": 0.35001858168304995,
530
+ "recon": 0.9224287867546082
531
+ },
532
+ {
533
+ "batch": 3600,
534
+ "cv": 0.331613763543316,
535
+ "cv_ema": 0.3498345335016526,
536
+ "recon": 0.9312664866447449
537
+ },
538
+ {
539
+ "batch": 3650,
540
+ "cv": 0.3543130452172241,
541
+ "cv_ema": 0.34987931861880833,
542
+ "recon": 0.9488472938537598
543
+ },
544
+ {
545
+ "batch": 3700,
546
+ "cv": 0.3702574556207215,
547
+ "cv_ema": 0.3500830999888275,
548
+ "recon": 0.7886255979537964
549
+ },
550
+ {
551
+ "batch": 3750,
552
+ "cv": 0.3843770450510611,
553
+ "cv_ema": 0.3504260394394498,
554
+ "recon": 0.9178078174591064
555
+ },
556
+ {
557
+ "batch": 3800,
558
+ "cv": 0.3665578530429211,
559
+ "cv_ema": 0.3505873575754845,
560
+ "recon": 0.8533543348312378
561
+ },
562
+ {
563
+ "batch": 3850,
564
+ "cv": 0.3786803125169436,
565
+ "cv_ema": 0.3508682871248991,
566
+ "recon": 0.9824788570404053
567
+ },
568
+ {
569
+ "batch": 3900,
570
+ "cv": 0.35780763007936683,
571
+ "cv_ema": 0.35093768055444374,
572
+ "recon": 1.052016258239746
573
+ }
574
+ ]
575
+ }
H/MID/H1_svd_fp64/seed2/tensorboard/events.out.tfevents.1776796617.52b79b2c7612.4109.41 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:45ab286c8808cbf7f79e3ae265621da22fc6c4930e819f0c70c46aa4483659c3
3
+ size 612125