AbstractPhil commited on
Commit
9deff4a
·
verified ·
1 Parent(s): 696f490

Ablation H-MID-H1_svd_fp64-s1

Browse files
H/MID/H1_svd_fp64/seed1/epoch_1_checkpoint.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8cbb790120c6e3168dfd15a0e80820fb6a85ebed0fd5ffc18bcf05363b64a4d3
3
+ size 2254319
H/MID/H1_svd_fp64/seed1/final_report.json ADDED
@@ -0,0 +1,575 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "config": {
3
+ "group": "H",
4
+ "variant": "H1_svd_fp64",
5
+ "band": "MID",
6
+ "seed": 1,
7
+ "phase": 2,
8
+ "num_epochs": 1,
9
+ "batch_size": 256,
10
+ "overrides": {
11
+ "svd": "fp64"
12
+ },
13
+ "description": "H-MID-H1_svd_fp64-s1"
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": 1,
62
+ "hf_repo": "AbstractPhil/geolip-svae-batteries",
63
+ "upload": false
64
+ },
65
+ "cv_ema_final": 0.36473455187352194,
66
+ "cv_last": 0.3808087981274897,
67
+ "predicted_band": "MID",
68
+ "expected_band": "MID",
69
+ "band_match": true,
70
+ "test_mse": 0.9463133215904236,
71
+ "recon_ema": 0.950522043844822,
72
+ "S0": 3.634821653366089,
73
+ "SD": 2.1206326484680176,
74
+ "ratio": 1.7140270092755943,
75
+ "erank": 7.877918720245361,
76
+ "observed_sphere_cv": 0.38411600810459884,
77
+ "uniform_sphere_cv_prediction": 0.35418994230276496,
78
+ "band_deviation": 0.02992606580183388,
79
+ "params_finite": true,
80
+ "num_epochs_run": 1,
81
+ "start_epoch": 0,
82
+ "per_epoch_metrics": [
83
+ {
84
+ "epoch": 1,
85
+ "test_mse": 0.9463133215904236,
86
+ "cv_ema": 0.36473455187352194,
87
+ "observed_sphere_cv": 0.38411600810459884,
88
+ "band_deviation": 0.02992606580183388,
89
+ "erank": 7.877918720245361,
90
+ "params_finite": true,
91
+ "wallclock_seconds": 77.72391510009766,
92
+ "checkpoint_path": "/content/ablations_phase2/H/MID/H1_svd_fp64_s1/epoch_1_checkpoint.pt"
93
+ }
94
+ ],
95
+ "params_count": 183339,
96
+ "wallclock_seconds": 77.7252471446991,
97
+ "batches_completed": 3906,
98
+ "batch_limit": 3906,
99
+ "cv_trajectory": [
100
+ {
101
+ "batch": 0,
102
+ "cv": 0.35279552109123474,
103
+ "cv_ema": 0.35279552109123474,
104
+ "recon": 1.7395108938217163
105
+ },
106
+ {
107
+ "batch": 50,
108
+ "cv": 0.35564687851963006,
109
+ "cv_ema": 0.35282403466551865,
110
+ "recon": 1.6272985935211182
111
+ },
112
+ {
113
+ "batch": 100,
114
+ "cv": 0.3395375022356627,
115
+ "cv_ema": 0.35269116934122013,
116
+ "recon": 1.3669804334640503
117
+ },
118
+ {
119
+ "batch": 150,
120
+ "cv": 0.3581503795723038,
121
+ "cv_ema": 0.352745761443531,
122
+ "recon": 1.2599258422851562
123
+ },
124
+ {
125
+ "batch": 200,
126
+ "cv": 0.3555110791922996,
127
+ "cv_ema": 0.3527734146210187,
128
+ "recon": 1.274646282196045
129
+ },
130
+ {
131
+ "batch": 250,
132
+ "cv": 0.3368160619419433,
133
+ "cv_ema": 0.3526138410942279,
134
+ "recon": 1.1936054229736328
135
+ },
136
+ {
137
+ "batch": 300,
138
+ "cv": 0.39991235379082773,
139
+ "cv_ema": 0.35308682622119386,
140
+ "recon": 1.100743293762207
141
+ },
142
+ {
143
+ "batch": 350,
144
+ "cv": 0.39472229176116214,
145
+ "cv_ema": 0.35350318087659355,
146
+ "recon": 1.1841460466384888
147
+ },
148
+ {
149
+ "batch": 400,
150
+ "cv": 0.38087881972897486,
151
+ "cv_ema": 0.35377693726511733,
152
+ "recon": 0.8235698938369751
153
+ },
154
+ {
155
+ "batch": 450,
156
+ "cv": 0.3915275719236767,
157
+ "cv_ema": 0.35415444361170295,
158
+ "recon": 1.053651213645935
159
+ },
160
+ {
161
+ "batch": 500,
162
+ "cv": 0.36030020969262516,
163
+ "cv_ema": 0.3542159012725122,
164
+ "recon": 0.9729546308517456
165
+ },
166
+ {
167
+ "batch": 550,
168
+ "cv": 0.37975428510275167,
169
+ "cv_ema": 0.35447128511081455,
170
+ "recon": 0.9853494167327881
171
+ },
172
+ {
173
+ "batch": 600,
174
+ "cv": 0.37936626264223555,
175
+ "cv_ema": 0.35472023488612875,
176
+ "recon": 0.9255595207214355
177
+ },
178
+ {
179
+ "batch": 650,
180
+ "cv": 0.382416965732802,
181
+ "cv_ema": 0.35499720219459546,
182
+ "recon": 1.0124435424804688
183
+ },
184
+ {
185
+ "batch": 700,
186
+ "cv": 0.3444827569272621,
187
+ "cv_ema": 0.3548920577419221,
188
+ "recon": 0.8815228939056396
189
+ },
190
+ {
191
+ "batch": 750,
192
+ "cv": 0.3849525164569466,
193
+ "cv_ema": 0.3551926623290724,
194
+ "recon": 0.9874274134635925
195
+ },
196
+ {
197
+ "batch": 800,
198
+ "cv": 0.42091132592479785,
199
+ "cv_ema": 0.35584984896502964,
200
+ "recon": 1.0618685483932495
201
+ },
202
+ {
203
+ "batch": 850,
204
+ "cv": 0.3809562952931882,
205
+ "cv_ema": 0.35610091342831124,
206
+ "recon": 1.08640456199646
207
+ },
208
+ {
209
+ "batch": 900,
210
+ "cv": 0.3666071844874348,
211
+ "cv_ema": 0.3562059761389025,
212
+ "recon": 0.8820624351501465
213
+ },
214
+ {
215
+ "batch": 950,
216
+ "cv": 0.3923498979243478,
217
+ "cv_ema": 0.35656741535675696,
218
+ "recon": 1.1222504377365112
219
+ },
220
+ {
221
+ "batch": 1000,
222
+ "cv": 0.3931538692927037,
223
+ "cv_ema": 0.35693327989611645,
224
+ "recon": 1.1878141164779663
225
+ },
226
+ {
227
+ "batch": 1050,
228
+ "cv": 0.3858068077454533,
229
+ "cv_ema": 0.35722201517460983,
230
+ "recon": 0.9849262237548828
231
+ },
232
+ {
233
+ "batch": 1100,
234
+ "cv": 0.3731995619094552,
235
+ "cv_ema": 0.3573817906419583,
236
+ "recon": 1.06501305103302
237
+ },
238
+ {
239
+ "batch": 1150,
240
+ "cv": 0.35678299675442837,
241
+ "cv_ema": 0.357375802703083,
242
+ "recon": 0.9671773910522461
243
+ },
244
+ {
245
+ "batch": 1200,
246
+ "cv": 0.36147295877627555,
247
+ "cv_ema": 0.3574167742638149,
248
+ "recon": 1.011407494544983
249
+ },
250
+ {
251
+ "batch": 1250,
252
+ "cv": 0.3832226370657577,
253
+ "cv_ema": 0.3576748328918343,
254
+ "recon": 1.0265220403671265
255
+ },
256
+ {
257
+ "batch": 1300,
258
+ "cv": 0.37946005911366554,
259
+ "cv_ema": 0.35789268515405265,
260
+ "recon": 0.975336492061615
261
+ },
262
+ {
263
+ "batch": 1350,
264
+ "cv": 0.36345197792454703,
265
+ "cv_ema": 0.35794827808175755,
266
+ "recon": 1.0841889381408691
267
+ },
268
+ {
269
+ "batch": 1400,
270
+ "cv": 0.35188932243433857,
271
+ "cv_ema": 0.35788768852528335,
272
+ "recon": 0.9546093940734863
273
+ },
274
+ {
275
+ "batch": 1450,
276
+ "cv": 0.35889467431512695,
277
+ "cv_ema": 0.3578977583831818,
278
+ "recon": 0.7966103553771973
279
+ },
280
+ {
281
+ "batch": 1500,
282
+ "cv": 0.41032765666463267,
283
+ "cv_ema": 0.3584220573659963,
284
+ "recon": 0.8577415347099304
285
+ },
286
+ {
287
+ "batch": 1550,
288
+ "cv": 0.3581458513479429,
289
+ "cv_ema": 0.35841929530581573,
290
+ "recon": 0.9150348901748657
291
+ },
292
+ {
293
+ "batch": 1600,
294
+ "cv": 0.3352390078066708,
295
+ "cv_ema": 0.35818749243082426,
296
+ "recon": 1.0773961544036865
297
+ },
298
+ {
299
+ "batch": 1650,
300
+ "cv": 0.391455267446833,
301
+ "cv_ema": 0.35852017018098437,
302
+ "recon": 1.0497736930847168
303
+ },
304
+ {
305
+ "batch": 1700,
306
+ "cv": 0.38724195531807387,
307
+ "cv_ema": 0.35880738803235523,
308
+ "recon": 1.054011344909668
309
+ },
310
+ {
311
+ "batch": 1750,
312
+ "cv": 0.3688061735795867,
313
+ "cv_ema": 0.35890737588782756,
314
+ "recon": 0.8008241653442383
315
+ },
316
+ {
317
+ "batch": 1800,
318
+ "cv": 0.39034626910917436,
319
+ "cv_ema": 0.35922176482004103,
320
+ "recon": 1.1303205490112305
321
+ },
322
+ {
323
+ "batch": 1850,
324
+ "cv": 0.38459637899616556,
325
+ "cv_ema": 0.3594755109618023,
326
+ "recon": 0.9880011081695557
327
+ },
328
+ {
329
+ "batch": 1900,
330
+ "cv": 0.3983339389808169,
331
+ "cv_ema": 0.3598640952419924,
332
+ "recon": 1.0282706022262573
333
+ },
334
+ {
335
+ "batch": 1950,
336
+ "cv": 0.433025168514267,
337
+ "cv_ema": 0.3605957059747152,
338
+ "recon": 0.9557077288627625
339
+ },
340
+ {
341
+ "batch": 2000,
342
+ "cv": 0.3541245034858209,
343
+ "cv_ema": 0.36053099394982624,
344
+ "recon": 0.9714743494987488
345
+ },
346
+ {
347
+ "batch": 2050,
348
+ "cv": 0.35504820759267,
349
+ "cv_ema": 0.36047616608625466,
350
+ "recon": 1.0400290489196777
351
+ },
352
+ {
353
+ "batch": 2100,
354
+ "cv": 0.39889675856321016,
355
+ "cv_ema": 0.3608603720110242,
356
+ "recon": 1.002923607826233
357
+ },
358
+ {
359
+ "batch": 2150,
360
+ "cv": 0.378768604966783,
361
+ "cv_ema": 0.3610394543405818,
362
+ "recon": 0.9223618507385254
363
+ },
364
+ {
365
+ "batch": 2200,
366
+ "cv": 0.3807664491308439,
367
+ "cv_ema": 0.3612367242884844,
368
+ "recon": 0.9840802550315857
369
+ },
370
+ {
371
+ "batch": 2250,
372
+ "cv": 0.4105364336192815,
373
+ "cv_ema": 0.3617297213817924,
374
+ "recon": 0.7422758340835571
375
+ },
376
+ {
377
+ "batch": 2300,
378
+ "cv": 0.34389730260107226,
379
+ "cv_ema": 0.36155139719398516,
380
+ "recon": 1.0067965984344482
381
+ },
382
+ {
383
+ "batch": 2350,
384
+ "cv": 0.3812218665408699,
385
+ "cv_ema": 0.361748101887454,
386
+ "recon": 0.8845947980880737
387
+ },
388
+ {
389
+ "batch": 2400,
390
+ "cv": 0.3755474399327951,
391
+ "cv_ema": 0.3618860952679074,
392
+ "recon": 0.9044098854064941
393
+ },
394
+ {
395
+ "batch": 2450,
396
+ "cv": 0.37518491537730514,
397
+ "cv_ema": 0.36201908346900136,
398
+ "recon": 0.867000937461853
399
+ },
400
+ {
401
+ "batch": 2500,
402
+ "cv": 0.3526044002511572,
403
+ "cv_ema": 0.3619249366368229,
404
+ "recon": 0.8168368339538574
405
+ },
406
+ {
407
+ "batch": 2550,
408
+ "cv": 0.386574705341651,
409
+ "cv_ema": 0.3621714343238712,
410
+ "recon": 0.9467669725418091
411
+ },
412
+ {
413
+ "batch": 2600,
414
+ "cv": 0.37276822154377703,
415
+ "cv_ema": 0.3622774021960703,
416
+ "recon": 0.8943279981613159
417
+ },
418
+ {
419
+ "batch": 2650,
420
+ "cv": 0.39750672727621716,
421
+ "cv_ema": 0.36262969544687174,
422
+ "recon": 0.9733526706695557
423
+ },
424
+ {
425
+ "batch": 2700,
426
+ "cv": 0.38700684085306586,
427
+ "cv_ema": 0.3628734669009337,
428
+ "recon": 0.8591717481613159
429
+ },
430
+ {
431
+ "batch": 2750,
432
+ "cv": 0.3415527225272767,
433
+ "cv_ema": 0.36266025945719715,
434
+ "recon": 0.9654970169067383
435
+ },
436
+ {
437
+ "batch": 2800,
438
+ "cv": 0.37070466877573716,
439
+ "cv_ema": 0.36274070355038257,
440
+ "recon": 0.9850203394889832
441
+ },
442
+ {
443
+ "batch": 2850,
444
+ "cv": 0.35509691573794855,
445
+ "cv_ema": 0.3626642656722582,
446
+ "recon": 0.8857461214065552
447
+ },
448
+ {
449
+ "batch": 2900,
450
+ "cv": 0.3961334374418692,
451
+ "cv_ema": 0.3629989573899543,
452
+ "recon": 0.9332637786865234
453
+ },
454
+ {
455
+ "batch": 2950,
456
+ "cv": 0.36680382030193315,
457
+ "cv_ema": 0.36303700601907407,
458
+ "recon": 0.8978904485702515
459
+ },
460
+ {
461
+ "batch": 3000,
462
+ "cv": 0.347666970476324,
463
+ "cv_ema": 0.36288330566364657,
464
+ "recon": 0.8949379920959473
465
+ },
466
+ {
467
+ "batch": 3050,
468
+ "cv": 0.3746081630992487,
469
+ "cv_ema": 0.3630005542380026,
470
+ "recon": 0.8828325271606445
471
+ },
472
+ {
473
+ "batch": 3100,
474
+ "cv": 0.36764194798940764,
475
+ "cv_ema": 0.3630469681755166,
476
+ "recon": 0.9268969893455505
477
+ },
478
+ {
479
+ "batch": 3150,
480
+ "cv": 0.40908514212597363,
481
+ "cv_ema": 0.3635073499150212,
482
+ "recon": 0.8494652509689331
483
+ },
484
+ {
485
+ "batch": 3200,
486
+ "cv": 0.39485664850359115,
487
+ "cv_ema": 0.3638208429009069,
488
+ "recon": 0.9985787272453308
489
+ },
490
+ {
491
+ "batch": 3250,
492
+ "cv": 0.33902808238375876,
493
+ "cv_ema": 0.3635729152957354,
494
+ "recon": 0.9445199966430664
495
+ },
496
+ {
497
+ "batch": 3300,
498
+ "cv": 0.38333636067120874,
499
+ "cv_ema": 0.36377054974949014,
500
+ "recon": 1.0312668085098267
501
+ },
502
+ {
503
+ "batch": 3350,
504
+ "cv": 0.3881649613514014,
505
+ "cv_ema": 0.3640144938655092,
506
+ "recon": 0.8617839813232422
507
+ },
508
+ {
509
+ "batch": 3400,
510
+ "cv": 0.366444556241298,
511
+ "cv_ema": 0.36403879448926707,
512
+ "recon": 0.8878062963485718
513
+ },
514
+ {
515
+ "batch": 3450,
516
+ "cv": 0.37183852224087793,
517
+ "cv_ema": 0.3641167917667832,
518
+ "recon": 0.9879652857780457
519
+ },
520
+ {
521
+ "batch": 3500,
522
+ "cv": 0.3549576311753205,
523
+ "cv_ema": 0.36402520016086853,
524
+ "recon": 1.0996479988098145
525
+ },
526
+ {
527
+ "batch": 3550,
528
+ "cv": 0.3647691243496031,
529
+ "cv_ema": 0.3640326394027559,
530
+ "recon": 0.8355409502983093
531
+ },
532
+ {
533
+ "batch": 3600,
534
+ "cv": 0.39868191102894057,
535
+ "cv_ema": 0.36437913211901773,
536
+ "recon": 0.9277450442314148
537
+ },
538
+ {
539
+ "batch": 3650,
540
+ "cv": 0.36714678351215185,
541
+ "cv_ema": 0.36440680863294905,
542
+ "recon": 0.9606571197509766
543
+ },
544
+ {
545
+ "batch": 3700,
546
+ "cv": 0.3478068319749233,
547
+ "cv_ema": 0.3642408088663688,
548
+ "recon": 1.1231796741485596
549
+ },
550
+ {
551
+ "batch": 3750,
552
+ "cv": 0.3646478089733264,
553
+ "cv_ema": 0.3642448788674384,
554
+ "recon": 0.9133890867233276
555
+ },
556
+ {
557
+ "batch": 3800,
558
+ "cv": 0.3506506321981932,
559
+ "cv_ema": 0.36410893640074593,
560
+ "recon": 0.9213093519210815
561
+ },
562
+ {
563
+ "batch": 3850,
564
+ "cv": 0.4104338713006016,
565
+ "cv_ema": 0.3645721857497445,
566
+ "recon": 1.0875505208969116
567
+ },
568
+ {
569
+ "batch": 3900,
570
+ "cv": 0.3808087981274897,
571
+ "cv_ema": 0.36473455187352194,
572
+ "recon": 0.9862040877342224
573
+ }
574
+ ]
575
+ }
H/MID/H1_svd_fp64/seed1/tensorboard/events.out.tfevents.1776796537.52b79b2c7612.4109.40 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:70b4d8e2dea480aee1144765418949471e4685e0b7d50713ead8fb3fa6ef8ff4
3
+ size 612125