summaryrefslogtreecommitdiff
path: root/results/h2_no_residual_full_s42/snapshot_evolution_s42.json
diff options
context:
space:
mode:
Diffstat (limited to 'results/h2_no_residual_full_s42/snapshot_evolution_s42.json')
-rw-r--r--results/h2_no_residual_full_s42/snapshot_evolution_s42.json5475
1 files changed, 5475 insertions, 0 deletions
diff --git a/results/h2_no_residual_full_s42/snapshot_evolution_s42.json b/results/h2_no_residual_full_s42/snapshot_evolution_s42.json
new file mode 100644
index 0000000..5a96cce
--- /dev/null
+++ b/results/h2_no_residual_full_s42/snapshot_evolution_s42.json
@@ -0,0 +1,5475 @@
+{
+ "config": {
+ "output_dir": "results/h2_no_residual_full_s42",
+ "epochs": 100,
+ "lr": 0.001,
+ "wd": 0.01,
+ "seed": 42,
+ "depth": 4,
+ "d_hidden": 256,
+ "log_every": 1,
+ "no_residual_add": true,
+ "w2_std": 0.5
+ },
+ "depth": 4,
+ "d_hidden": 256,
+ "num_classes": 10,
+ "bp_log": [
+ {
+ "hidden_norms": [
+ 8.718070030212402,
+ 42.802223205566406,
+ 44.22697067260742,
+ 46.05121994018555,
+ 46.896766662597656
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 0.0010791289387270808,
+ 0.0002035950165009126,
+ 0.00014324457151815295,
+ 0.00011542177526280284,
+ 9.846547618508339e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0380445197224617,
+ 0.00666962843388319,
+ 0.0046500470489263535,
+ 0.003719294909387827,
+ 0.003155440790578723
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.080078125,
+ "loss_eval": 2.446239709854126,
+ "epoch": 0
+ },
+ {
+ "hidden_norms": [
+ 86.00836944580078,
+ 130.30899047851562,
+ 107.6886215209961,
+ 127.31951141357422,
+ 155.31224060058594
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 5.636421337840147e-05,
+ 4.504464232013561e-05,
+ 4.254864325048402e-05,
+ 4.0549883124185726e-05,
+ 3.8371628761524335e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0019479888724163175,
+ 0.0014867106219753623,
+ 0.0013919975608587265,
+ 0.001322419848293066,
+ 0.0012390650808811188
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.1435546875,
+ "loss_eval": 2.293787717819214,
+ "epoch": 1
+ },
+ {
+ "hidden_norms": [
+ 114.72047424316406,
+ 154.09420776367188,
+ 128.2491455078125,
+ 160.83230590820312,
+ 193.06747436523438
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.478770642890595e-05,
+ 3.7948859244352207e-05,
+ 3.570393528207205e-05,
+ 3.381549322512001e-05,
+ 3.158334220643155e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0015459235291928053,
+ 0.0012464966857805848,
+ 0.0011667043436318636,
+ 0.0010993058094754815,
+ 0.0010170178720727563
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.1728515625,
+ "loss_eval": 2.2832698822021484,
+ "epoch": 2
+ },
+ {
+ "hidden_norms": [
+ 129.28399658203125,
+ 169.31143188476562,
+ 136.48431396484375,
+ 176.9506378173828,
+ 216.20181274414062
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.090799848199822e-05,
+ 3.4604578104335815e-05,
+ 3.2474679755978286e-05,
+ 3.0787930882070214e-05,
+ 2.844005575752817e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0014615722466260195,
+ 0.0011723904171958566,
+ 0.0010821294272318482,
+ 0.001009885803796351,
+ 0.000921491184271872
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.177734375,
+ "loss_eval": 2.248687267303467,
+ "epoch": 3
+ },
+ {
+ "hidden_norms": [
+ 146.55203247070312,
+ 179.31271362304688,
+ 144.7793426513672,
+ 195.2471923828125,
+ 244.1785888671875
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.7983172660460696e-05,
+ 3.2681429729564115e-05,
+ 3.063849362661131e-05,
+ 2.877583210647572e-05,
+ 2.62654539255891e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0013270865892991424,
+ 0.0011031923349946737,
+ 0.0010214283829554915,
+ 0.0009468899224884808,
+ 0.0008542603463865817
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.201171875,
+ "loss_eval": 2.1866140365600586,
+ "epoch": 4
+ },
+ {
+ "hidden_norms": [
+ 163.96316528320312,
+ 186.10421752929688,
+ 151.66432189941406,
+ 204.73275756835938,
+ 266.4998474121094
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.63540711987298e-05,
+ 3.1559447961626574e-05,
+ 2.965202475024853e-05,
+ 2.767328805930447e-05,
+ 2.4988346922327764e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0012765713036060333,
+ 0.0010719435522332788,
+ 0.0009894924005493522,
+ 0.0009105658973567188,
+ 0.0008120943675749004
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.205078125,
+ "loss_eval": 2.213620185852051,
+ "epoch": 5
+ },
+ {
+ "hidden_norms": [
+ 173.03375244140625,
+ 192.33702087402344,
+ 151.7676544189453,
+ 211.76028442382812,
+ 273.14984130859375
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.612042564782314e-05,
+ 3.138732790830545e-05,
+ 2.92498261842411e-05,
+ 2.741635762504302e-05,
+ 2.4648286853334866e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0012719081714749336,
+ 0.0010634849313646555,
+ 0.0009788337629288435,
+ 0.0008956211968325078,
+ 0.0007930789142847061
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.201171875,
+ "loss_eval": 2.20696759223938,
+ "epoch": 6
+ },
+ {
+ "hidden_norms": [
+ 184.99200439453125,
+ 200.6100311279297,
+ 154.7901611328125,
+ 216.4141387939453,
+ 294.1932067871094
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.5208213375881314e-05,
+ 3.101788752246648e-05,
+ 2.887230584747158e-05,
+ 2.6716055799624883e-05,
+ 2.381021658948157e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0012596799060702324,
+ 0.0010643631685525179,
+ 0.0009746645228005946,
+ 0.0008811713196337223,
+ 0.0007715557585470378
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.189453125,
+ "loss_eval": 2.26588773727417,
+ "epoch": 7
+ },
+ {
+ "hidden_norms": [
+ 199.71249389648438,
+ 206.28636169433594,
+ 159.15716552734375,
+ 225.3829803466797,
+ 316.6280517578125
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.397777982172556e-05,
+ 3.0055914976401255e-05,
+ 2.8245711291674525e-05,
+ 2.577536906755995e-05,
+ 2.2829581212135963e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0012129175011068583,
+ 0.0010332722449675202,
+ 0.0009465343900956213,
+ 0.0008514428627677262,
+ 0.0007388851954601705
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.203125,
+ "loss_eval": 2.232454776763916,
+ "epoch": 8
+ },
+ {
+ "hidden_norms": [
+ 209.29306030273438,
+ 211.71087646484375,
+ 155.6585235595703,
+ 229.6345672607422,
+ 335.6332092285156
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.381286660442129e-05,
+ 3.0025108571862802e-05,
+ 2.795294494717382e-05,
+ 2.5483084755251184e-05,
+ 2.2450465621659532e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.001203069812618196,
+ 0.0010322949383407831,
+ 0.0009428044431842864,
+ 0.0008445970015600324,
+ 0.0007283088052645326
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.189453125,
+ "loss_eval": 2.2567524909973145,
+ "epoch": 9
+ },
+ {
+ "hidden_norms": [
+ 219.5225067138672,
+ 215.77891540527344,
+ 158.00711059570312,
+ 239.82870483398438,
+ 364.106689453125
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.357209061505273e-05,
+ 3.0128332582535222e-05,
+ 2.8080758056603372e-05,
+ 2.5434510462218896e-05,
+ 2.2072978026699275e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.001222757389768958,
+ 0.0010488828411325812,
+ 0.0009540626197122037,
+ 0.000844798400066793,
+ 0.0007152045727707446
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.1884765625,
+ "loss_eval": 2.298661708831787,
+ "epoch": 10
+ },
+ {
+ "hidden_norms": [
+ 228.8721160888672,
+ 219.33489990234375,
+ 159.22947692871094,
+ 242.5733642578125,
+ 373.5672607421875
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.361105336807668e-05,
+ 3.0123681426630355e-05,
+ 2.82819164567627e-05,
+ 2.5511151761747897e-05,
+ 2.200892413384281e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0012033756356686354,
+ 0.0010383055778220296,
+ 0.0009499595616944134,
+ 0.0008407059940509498,
+ 0.000707409402821213
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.18359375,
+ "loss_eval": 2.3219099044799805,
+ "epoch": 11
+ },
+ {
+ "hidden_norms": [
+ 238.0825958251953,
+ 220.4098663330078,
+ 157.96127319335938,
+ 246.49105834960938,
+ 399.103515625
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.3689280826365575e-05,
+ 3.044224104087334e-05,
+ 2.8301144993747585e-05,
+ 2.544306880736258e-05,
+ 2.16337175515946e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0012109626550227404,
+ 0.0010498397750779986,
+ 0.0009549973183311522,
+ 0.0008336003520525992,
+ 0.0006956221768632531
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.197265625,
+ "loss_eval": 2.292271614074707,
+ "epoch": 12
+ },
+ {
+ "hidden_norms": [
+ 245.9247589111328,
+ 225.1177520751953,
+ 161.000244140625,
+ 247.5177001953125,
+ 405.64068603515625
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.318578455946408e-05,
+ 2.9796459784847684e-05,
+ 2.778910129563883e-05,
+ 2.4826545995892957e-05,
+ 2.1260259018163197e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0012085892958566546,
+ 0.0010467838728800416,
+ 0.0009537216974422336,
+ 0.0008304294897243381,
+ 0.0006893921527080238
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.1865234375,
+ "loss_eval": 2.309349536895752,
+ "epoch": 13
+ },
+ {
+ "hidden_norms": [
+ 254.78964233398438,
+ 226.26568603515625,
+ 158.11354064941406,
+ 246.54937744140625,
+ 402.22564697265625
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.351089617353864e-05,
+ 3.0367529689101502e-05,
+ 2.8210330128786154e-05,
+ 2.5112614821409807e-05,
+ 2.127070547430776e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0012080274755135179,
+ 0.0010523217497393489,
+ 0.0009614708251319826,
+ 0.0008391732117161155,
+ 0.0006922443280927837
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.177734375,
+ "loss_eval": 2.3668038845062256,
+ "epoch": 14
+ },
+ {
+ "hidden_norms": [
+ 262.1316833496094,
+ 226.88943481445312,
+ 157.7928009033203,
+ 248.3592529296875,
+ 416.15966796875
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.413746162550524e-05,
+ 3.107329393969849e-05,
+ 2.889345705625601e-05,
+ 2.570425385783892e-05,
+ 2.1452229702845216e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0012374932412058115,
+ 0.001085828524082899,
+ 0.0009924235055223107,
+ 0.0008545225136913359,
+ 0.0006974710267968476
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.1650390625,
+ "loss_eval": 2.3762776851654053,
+ "epoch": 15
+ },
+ {
+ "hidden_norms": [
+ 270.8565979003906,
+ 227.8741455078125,
+ 160.908935546875,
+ 249.23391723632812,
+ 426.4018859863281
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.346315133967437e-05,
+ 3.0396649890462868e-05,
+ 2.8140580980107188e-05,
+ 2.509099431335926e-05,
+ 2.0963307179044932e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0012071001110598445,
+ 0.0010572251630946994,
+ 0.0009650305728428066,
+ 0.0008328591356985271,
+ 0.0006829687627032399
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.1748046875,
+ "loss_eval": 2.3659489154815674,
+ "epoch": 16
+ },
+ {
+ "hidden_norms": [
+ 277.88897705078125,
+ 229.93283081054688,
+ 160.6474151611328,
+ 249.1276092529297,
+ 439.88336181640625
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.291336179245263e-05,
+ 3.02154276141664e-05,
+ 2.813426181091927e-05,
+ 2.503528776287567e-05,
+ 2.0815306925214827e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0011904252460226417,
+ 0.0010531789157539606,
+ 0.000964420149102807,
+ 0.0008325970848090947,
+ 0.0006798306130804121
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.17578125,
+ "loss_eval": 2.3654253482818604,
+ "epoch": 17
+ },
+ {
+ "hidden_norms": [
+ 286.05938720703125,
+ 230.5788116455078,
+ 160.4178009033203,
+ 247.1944122314453,
+ 438.8507995605469
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.318934977869503e-05,
+ 3.0551069357898086e-05,
+ 2.8402588213793933e-05,
+ 2.525982381484937e-05,
+ 2.0943261915817857e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0012075696140527725,
+ 0.001069555408321321,
+ 0.0009757071966305375,
+ 0.0008412246825173497,
+ 0.0006814986118115485
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.1748046875,
+ "loss_eval": 2.3814916610717773,
+ "epoch": 18
+ },
+ {
+ "hidden_norms": [
+ 291.21563720703125,
+ 232.70465087890625,
+ 157.34620666503906,
+ 246.0981903076172,
+ 446.42889404296875
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.315764479339123e-05,
+ 3.062548057641834e-05,
+ 2.8670148822129704e-05,
+ 2.5318840926047415e-05,
+ 2.096824573527556e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0012071789242327213,
+ 0.0010707362089306116,
+ 0.0009821956045925617,
+ 0.0008455720380879939,
+ 0.0006821187562309206
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.1728515625,
+ "loss_eval": 2.4200758934020996,
+ "epoch": 19
+ },
+ {
+ "hidden_norms": [
+ 296.9744873046875,
+ 232.92347717285156,
+ 155.9208221435547,
+ 242.89227294921875,
+ 464.36663818359375
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.340119292261079e-05,
+ 3.0905790481483564e-05,
+ 2.8828215363319032e-05,
+ 2.5400462618563324e-05,
+ 2.0934525309712626e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.001213042763993144,
+ 0.001074956264346838,
+ 0.000986237544566393,
+ 0.0008489739266224205,
+ 0.0006839185953140259
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.1669921875,
+ "loss_eval": 2.4464807510375977,
+ "epoch": 20
+ },
+ {
+ "hidden_norms": [
+ 302.8837890625,
+ 232.27020263671875,
+ 156.21949768066406,
+ 240.65646362304688,
+ 465.4597473144531
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.269123408244923e-05,
+ 3.057413414353505e-05,
+ 2.8494312573457137e-05,
+ 2.536452484491747e-05,
+ 2.0975066945538856e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0011949212057515979,
+ 0.0010672758799046278,
+ 0.0009809982730075717,
+ 0.0008468242012895644,
+ 0.0006826177705079317
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.171875,
+ "loss_eval": 2.4423742294311523,
+ "epoch": 21
+ },
+ {
+ "hidden_norms": [
+ 309.28985595703125,
+ 234.24151611328125,
+ 157.79649353027344,
+ 237.83340454101562,
+ 481.6331787109375
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.304004712845199e-05,
+ 3.070462844334543e-05,
+ 2.8894632123410702e-05,
+ 2.556288927735295e-05,
+ 2.0964585928595625e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.00119681132491678,
+ 0.001071512931957841,
+ 0.0009897719137370586,
+ 0.0008524276199750602,
+ 0.0006829827907495201
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.1591796875,
+ "loss_eval": 2.4570655822753906,
+ "epoch": 22
+ },
+ {
+ "hidden_norms": [
+ 313.0410461425781,
+ 231.53106689453125,
+ 157.91920471191406,
+ 237.67922973632812,
+ 458.6028137207031
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.3489413908682764e-05,
+ 3.1275969377020374e-05,
+ 2.9323471608222462e-05,
+ 2.5867013391689397e-05,
+ 2.119375130860135e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.001231541857123375,
+ 0.0011033288901671767,
+ 0.001017258851788938,
+ 0.0008692976552993059,
+ 0.0006914011319167912
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.1435546875,
+ "loss_eval": 2.4768424034118652,
+ "epoch": 23
+ },
+ {
+ "hidden_norms": [
+ 317.760986328125,
+ 230.58580017089844,
+ 159.59475708007812,
+ 231.33895874023438,
+ 457.3432922363281
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.3722881198627874e-05,
+ 3.157257742714137e-05,
+ 2.952909198938869e-05,
+ 2.6216532205580734e-05,
+ 2.1260124412947334e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0012356883380562067,
+ 0.001110993092879653,
+ 0.001023975433781743,
+ 0.0008810298750177026,
+ 0.0006986238295212388
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.1611328125,
+ "loss_eval": 2.49592661857605,
+ "epoch": 24
+ },
+ {
+ "hidden_norms": [
+ 322.651123046875,
+ 232.43505859375,
+ 152.87049865722656,
+ 231.94496154785156,
+ 473.4416198730469
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.427052797633223e-05,
+ 3.222951272618957e-05,
+ 3.0394427085411735e-05,
+ 2.6714942578109913e-05,
+ 2.156150912924204e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0012650383869186044,
+ 0.001135875703766942,
+ 0.0010507238330319524,
+ 0.0008947087335400283,
+ 0.0007038050098344684
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.1455078125,
+ "loss_eval": 2.5300588607788086,
+ "epoch": 25
+ },
+ {
+ "hidden_norms": [
+ 330.46783447265625,
+ 230.2061004638672,
+ 152.9967498779297,
+ 227.8012237548828,
+ 478.8476867675781
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.374157677171752e-05,
+ 3.1537194445263594e-05,
+ 2.9856541004846804e-05,
+ 2.6372779757366516e-05,
+ 2.1366411601775326e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0012416786048561335,
+ 0.001122219255194068,
+ 0.0010412678821012378,
+ 0.0008921810076572001,
+ 0.0006990613183006644
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.140625,
+ "loss_eval": 2.537388801574707,
+ "epoch": 26
+ },
+ {
+ "hidden_norms": [
+ 334.9617919921875,
+ 229.8689727783203,
+ 156.80023193359375,
+ 226.36166381835938,
+ 473.52545166015625
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.429998105275445e-05,
+ 3.2245541660813615e-05,
+ 3.0571496608899906e-05,
+ 2.689965367608238e-05,
+ 2.1567262592725456e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.001256743329577148,
+ 0.0011394708417356014,
+ 0.0010550402803346515,
+ 0.0009035509428940713,
+ 0.0007058348273858428
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.138671875,
+ "loss_eval": 2.5099682807922363,
+ "epoch": 27
+ },
+ {
+ "hidden_norms": [
+ 338.32244873046875,
+ 230.9785614013672,
+ 152.46780395507812,
+ 224.96499633789062,
+ 486.111083984375
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.443951572990045e-05,
+ 3.224790998501703e-05,
+ 3.0727285775355995e-05,
+ 2.7226298698224127e-05,
+ 2.1652953364537098e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0012660843785852194,
+ 0.0011493951315060258,
+ 0.0010706180473789573,
+ 0.0009175407467409968,
+ 0.0007115012849681079
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.1298828125,
+ "loss_eval": 2.5662076473236084,
+ "epoch": 28
+ },
+ {
+ "hidden_norms": [
+ 342.5646057128906,
+ 229.08815002441406,
+ 153.96673583984375,
+ 216.54530334472656,
+ 468.6747131347656
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.46860324498266e-05,
+ 3.2538478990318254e-05,
+ 3.093785562668927e-05,
+ 2.7406946173869073e-05,
+ 2.1759695300715975e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0012609382392838597,
+ 0.001147899660281837,
+ 0.0010664825094863772,
+ 0.0009160453919321299,
+ 0.0007114250329323113
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.123046875,
+ "loss_eval": 2.586763381958008,
+ "epoch": 29
+ },
+ {
+ "hidden_norms": [
+ 348.8407287597656,
+ 229.21829223632812,
+ 152.20892333984375,
+ 214.9100799560547,
+ 478.5781555175781
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.3881755371112376e-05,
+ 3.1849580409470946e-05,
+ 3.0435230655712076e-05,
+ 2.698182470339816e-05,
+ 2.157314338546712e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0012431713985279202,
+ 0.0011320828925818205,
+ 0.0010550167644396424,
+ 0.0009089091327041388,
+ 0.000706658698618412
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.1318359375,
+ "loss_eval": 2.5531229972839355,
+ "epoch": 30
+ },
+ {
+ "hidden_norms": [
+ 351.8199768066406,
+ 225.8287811279297,
+ 153.97621154785156,
+ 207.98619079589844,
+ 449.3790588378906
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.457794082351029e-05,
+ 3.269680746598169e-05,
+ 3.108347300440073e-05,
+ 2.75895836239215e-05,
+ 2.188902180932928e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0012726614950224757,
+ 0.0011605133768171072,
+ 0.0010800294112414122,
+ 0.0009289697627536952,
+ 0.0007184858550317585
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.130859375,
+ "loss_eval": 2.5874619483947754,
+ "epoch": 31
+ },
+ {
+ "hidden_norms": [
+ 353.6206359863281,
+ 225.6538543701172,
+ 151.9993896484375,
+ 207.81591796875,
+ 458.920166015625
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.461301821516827e-05,
+ 3.2766867661848664e-05,
+ 3.128241587546654e-05,
+ 2.7638012397801504e-05,
+ 2.198204310843721e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.001283104415051639,
+ 0.0011686854995787144,
+ 0.0010903337970376015,
+ 0.0009369406034238636,
+ 0.000721081392839551
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.1376953125,
+ "loss_eval": 2.583052635192871,
+ "epoch": 32
+ },
+ {
+ "hidden_norms": [
+ 355.174560546875,
+ 226.53065490722656,
+ 150.45091247558594,
+ 200.73980712890625,
+ 457.2731018066406
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.5071778256678954e-05,
+ 3.3138821891043335e-05,
+ 3.159284096909687e-05,
+ 2.8334745366009884e-05,
+ 2.23859660763992e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0012956986902281642,
+ 0.0011821272782981396,
+ 0.001104547642171383,
+ 0.0009524936904199421,
+ 0.0007312360685318708
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.12890625,
+ "loss_eval": 2.605353832244873,
+ "epoch": 33
+ },
+ {
+ "hidden_norms": [
+ 360.4849853515625,
+ 223.40980529785156,
+ 148.3797607421875,
+ 195.64169311523438,
+ 461.1542663574219
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.4972203138750046e-05,
+ 3.31850751535967e-05,
+ 3.171996650053188e-05,
+ 2.820813824655488e-05,
+ 2.2246067601372488e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0012813522480428219,
+ 0.0011739785550162196,
+ 0.00110049347858876,
+ 0.0009508885559625924,
+ 0.0007318558637052774
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.1279296875,
+ "loss_eval": 2.609264373779297,
+ "epoch": 34
+ },
+ {
+ "hidden_norms": [
+ 365.4316711425781,
+ 222.55001831054688,
+ 148.93753051757812,
+ 196.41476440429688,
+ 464.26507568359375
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.474177356110886e-05,
+ 3.299827585578896e-05,
+ 3.161515633109957e-05,
+ 2.8241847758181393e-05,
+ 2.233032682852354e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.001280988217331469,
+ 0.0011732708662748337,
+ 0.0010997118661180139,
+ 0.0009513412951491773,
+ 0.0007320611621253192
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.123046875,
+ "loss_eval": 2.6414568424224854,
+ "epoch": 35
+ },
+ {
+ "hidden_norms": [
+ 365.99407958984375,
+ 220.5921173095703,
+ 147.05055236816406,
+ 194.5092315673828,
+ 456.3348083496094
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.5159438994014636e-05,
+ 3.344163633300923e-05,
+ 3.1946114177117124e-05,
+ 2.86933845927706e-05,
+ 2.2498443286167458e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0013043885119259357,
+ 0.0011966620804741979,
+ 0.0011218151776120067,
+ 0.0009711913880892098,
+ 0.0007424909272231162
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.1337890625,
+ "loss_eval": 2.6415724754333496,
+ "epoch": 36
+ },
+ {
+ "hidden_norms": [
+ 370.3197326660156,
+ 220.85414123535156,
+ 148.2594451904297,
+ 188.3152313232422,
+ 446.8343505859375
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.53947798430454e-05,
+ 3.362595452927053e-05,
+ 3.218940764782019e-05,
+ 2.885511094063986e-05,
+ 2.269916149089113e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.001304175122641027,
+ 0.001198453363031149,
+ 0.0011262644547969103,
+ 0.0009740483365021646,
+ 0.0007447644020430744
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.1279296875,
+ "loss_eval": 2.6847548484802246,
+ "epoch": 37
+ },
+ {
+ "hidden_norms": [
+ 372.6519775390625,
+ 219.43032836914062,
+ 146.2524871826172,
+ 186.9598388671875,
+ 419.9220275878906
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.5350865800864995e-05,
+ 3.369166006450541e-05,
+ 3.231317532481626e-05,
+ 2.9215334507171065e-05,
+ 2.2943044314160943e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0013154788175597787,
+ 0.0012110902462154627,
+ 0.00113767315633595,
+ 0.000987510895356536,
+ 0.0007552221650257707
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.134765625,
+ "loss_eval": 2.663112163543701,
+ "epoch": 38
+ },
+ {
+ "hidden_norms": [
+ 377.5515441894531,
+ 217.7505340576172,
+ 144.7539520263672,
+ 182.63914489746094,
+ 424.8920593261719
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.550875771907158e-05,
+ 3.3909564081113786e-05,
+ 3.234989708289504e-05,
+ 2.9385402740444988e-05,
+ 2.3012338715489022e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0013136330526322126,
+ 0.0012126904912292957,
+ 0.001143725821748376,
+ 0.0009956557769328356,
+ 0.000757638190407306
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.1123046875,
+ "loss_eval": 2.6713695526123047,
+ "epoch": 39
+ },
+ {
+ "hidden_norms": [
+ 375.5976867675781,
+ 217.32119750976562,
+ 143.29833984375,
+ 178.9922332763672,
+ 416.0475769042969
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.6261131754145026e-05,
+ 3.459263825789094e-05,
+ 3.311317050247453e-05,
+ 3.0077193514443934e-05,
+ 2.332564690732397e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0013422674965113401,
+ 0.0012380937114357948,
+ 0.0011671145912259817,
+ 0.0010137255303561687,
+ 0.0007670389022678137
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.1171875,
+ "loss_eval": 2.6898365020751953,
+ "epoch": 40
+ },
+ {
+ "hidden_norms": [
+ 375.6454772949219,
+ 215.73880004882812,
+ 143.0796356201172,
+ 173.5657958984375,
+ 413.10693359375
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.5994391510030255e-05,
+ 3.4521792258601636e-05,
+ 3.3360858651576564e-05,
+ 3.016969822056126e-05,
+ 2.3851269361330196e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0013558906503021717,
+ 0.0012535321293398738,
+ 0.0011830577859655023,
+ 0.0010295362444594502,
+ 0.0007781152380630374
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.119140625,
+ "loss_eval": 2.6808576583862305,
+ "epoch": 41
+ },
+ {
+ "hidden_norms": [
+ 381.1890869140625,
+ 214.87374877929688,
+ 142.17715454101562,
+ 172.16009521484375,
+ 413.11871337890625
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.590781852835789e-05,
+ 3.450449730735272e-05,
+ 3.32859308400657e-05,
+ 3.011901571881026e-05,
+ 2.3740196411381476e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0013439194299280643,
+ 0.0012433268129825592,
+ 0.0011747550452128053,
+ 0.0010278060799464583,
+ 0.0007790841045789421
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.1083984375,
+ "loss_eval": 2.7061455249786377,
+ "epoch": 42
+ },
+ {
+ "hidden_norms": [
+ 382.1147766113281,
+ 212.32655334472656,
+ 142.33558654785156,
+ 169.52694702148438,
+ 398.5701904296875
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.659713547676802e-05,
+ 3.507183282636106e-05,
+ 3.372716673766263e-05,
+ 3.064091288251802e-05,
+ 2.416563620499801e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0013747483026236296,
+ 0.0012707116547971964,
+ 0.0012008449994027615,
+ 0.0010481667704880238,
+ 0.0007909793639555573
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.107421875,
+ "loss_eval": 2.7121200561523438,
+ "epoch": 43
+ },
+ {
+ "hidden_norms": [
+ 388.24554443359375,
+ 211.70144653320312,
+ 141.58709716796875,
+ 166.8690643310547,
+ 390.73895263671875
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.605844176490791e-05,
+ 3.457901402725838e-05,
+ 3.344014839967713e-05,
+ 3.0472821890725754e-05,
+ 2.386603955528699e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0013416680740192533,
+ 0.0012422216823324561,
+ 0.0011764775263145566,
+ 0.0010343515314161777,
+ 0.0007849647663533688
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.1240234375,
+ "loss_eval": 2.6901586055755615,
+ "epoch": 44
+ },
+ {
+ "hidden_norms": [
+ 387.182861328125,
+ 211.15060424804688,
+ 141.0326385498047,
+ 165.57235717773438,
+ 375.9649963378906
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.6760768125532195e-05,
+ 3.520414247759618e-05,
+ 3.39620492013637e-05,
+ 3.100339381489903e-05,
+ 2.4274080715258606e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0013679155381396413,
+ 0.0012680977815762162,
+ 0.001201036386191845,
+ 0.0010586031712591648,
+ 0.0007975373882800341
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.115234375,
+ "loss_eval": 2.7196202278137207,
+ "epoch": 45
+ },
+ {
+ "hidden_norms": [
+ 389.78302001953125,
+ 209.2168426513672,
+ 140.19239807128906,
+ 162.1501007080078,
+ 373.6092529296875
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.6972556699765846e-05,
+ 3.547927553881891e-05,
+ 3.4321546991122887e-05,
+ 3.136819213978015e-05,
+ 2.4795872377580963e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.001378108048811555,
+ 0.0012792085763067007,
+ 0.0012131200637668371,
+ 0.0010711504146456718,
+ 0.0008094239747151732
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.1123046875,
+ "loss_eval": 2.732020854949951,
+ "epoch": 46
+ },
+ {
+ "hidden_norms": [
+ 390.8168640136719,
+ 207.98391723632812,
+ 140.21983337402344,
+ 161.1061248779297,
+ 358.555419921875
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.740100146387704e-05,
+ 3.60443300451152e-05,
+ 3.478667713352479e-05,
+ 3.1701372790848836e-05,
+ 2.49275763053447e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0013912947615608573,
+ 0.0012947055511176586,
+ 0.0012257345952093601,
+ 0.00108279287815094,
+ 0.000817548600025475
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.11328125,
+ "loss_eval": 2.734851598739624,
+ "epoch": 47
+ },
+ {
+ "hidden_norms": [
+ 392.3091735839844,
+ 206.94676208496094,
+ 137.84091186523438,
+ 156.7150115966797,
+ 363.1738586425781
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.7181769585004076e-05,
+ 3.575249866116792e-05,
+ 3.458252103882842e-05,
+ 3.156786260660738e-05,
+ 2.489473990863189e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0013799670850858092,
+ 0.0012849466875195503,
+ 0.0012195904273539782,
+ 0.0010820929892361164,
+ 0.0008166414918377995
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.1083984375,
+ "loss_eval": 2.730489492416382,
+ "epoch": 48
+ },
+ {
+ "hidden_norms": [
+ 392.93621826171875,
+ 206.63357543945312,
+ 137.95150756835938,
+ 156.19265747070312,
+ 348.4825744628906
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.750698306248523e-05,
+ 3.625384488259442e-05,
+ 3.4952467103721574e-05,
+ 3.2122828997671604e-05,
+ 2.5343802917632274e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0014074875507503748,
+ 0.0013119739014655352,
+ 0.0012457441771402955,
+ 0.0011056199437007308,
+ 0.0008317606407217681
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.10546875,
+ "loss_eval": 2.7619447708129883,
+ "epoch": 49
+ },
+ {
+ "hidden_norms": [
+ 394.5487976074219,
+ 204.12411499023438,
+ 137.66342163085938,
+ 151.4534454345703,
+ 343.6612854003906
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.779921098612249e-05,
+ 3.6266086681280285e-05,
+ 3.5253593523520976e-05,
+ 3.246645428589545e-05,
+ 2.558354208304081e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0014083774294704199,
+ 0.0013136735651642084,
+ 0.0012480755103752017,
+ 0.0011111509520560503,
+ 0.0008380677900277078
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.095703125,
+ "loss_eval": 2.7554924488067627,
+ "epoch": 50
+ },
+ {
+ "hidden_norms": [
+ 397.6114196777344,
+ 202.55885314941406,
+ 136.6888427734375,
+ 150.49545288085938,
+ 329.96112060546875
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.777034726226702e-05,
+ 3.6449018807616085e-05,
+ 3.515862044878304e-05,
+ 3.240678779548034e-05,
+ 2.560876600909978e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.001405128394253552,
+ 0.001311970641836524,
+ 0.001248166779987514,
+ 0.0011124340817332268,
+ 0.0008421412203460932
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.10546875,
+ "loss_eval": 2.7597320079803467,
+ "epoch": 51
+ },
+ {
+ "hidden_norms": [
+ 397.78509521484375,
+ 202.12451171875,
+ 135.5238800048828,
+ 148.92816162109375,
+ 322.4108581542969
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.840518184006214e-05,
+ 3.6834608181379735e-05,
+ 3.5840494092553854e-05,
+ 3.275332346674986e-05,
+ 2.6062174583785236e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.001429337658919394,
+ 0.00133553147315979,
+ 0.0012706448324024677,
+ 0.001131666824221611,
+ 0.0008531482890248299
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.099609375,
+ "loss_eval": 2.7737417221069336,
+ "epoch": 52
+ },
+ {
+ "hidden_norms": [
+ 397.3096618652344,
+ 199.6620635986328,
+ 134.52984619140625,
+ 145.349853515625,
+ 298.2867431640625
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.877158087561838e-05,
+ 3.721338725881651e-05,
+ 3.603506775107235e-05,
+ 3.3124437322840095e-05,
+ 2.6345343940192834e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0014398579951375723,
+ 0.001346803386695683,
+ 0.0012816074304282665,
+ 0.001146289287135005,
+ 0.0008667311049066484
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.1025390625,
+ "loss_eval": 2.799215316772461,
+ "epoch": 53
+ },
+ {
+ "hidden_norms": [
+ 400.7160949707031,
+ 198.93077087402344,
+ 134.9965057373047,
+ 147.8721923828125,
+ 299.2564392089844
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.8850699638715014e-05,
+ 3.7408095522550866e-05,
+ 3.61659076588694e-05,
+ 3.348653990542516e-05,
+ 2.6619676646078005e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0014573318185284734,
+ 0.0013647449668496847,
+ 0.001298584509640932,
+ 0.001162077416665852,
+ 0.0008761510835029185
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.08984375,
+ "loss_eval": 2.8113207817077637,
+ "epoch": 54
+ },
+ {
+ "hidden_norms": [
+ 399.78253173828125,
+ 198.36306762695312,
+ 132.11000061035156,
+ 142.8116912841797,
+ 299.17218017578125
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.933380503440276e-05,
+ 3.791782364714891e-05,
+ 3.674212348414585e-05,
+ 3.40558253810741e-05,
+ 2.7044263333664276e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.001484063221141696,
+ 0.0013903625076636672,
+ 0.0013244440779089928,
+ 0.0011864654952660203,
+ 0.0008941980777308345
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.0888671875,
+ "loss_eval": 2.847215175628662,
+ "epoch": 55
+ },
+ {
+ "hidden_norms": [
+ 401.17156982421875,
+ 196.801025390625,
+ 134.57815551757812,
+ 139.7723388671875,
+ 288.41851806640625
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.922020914615132e-05,
+ 3.7812937080161646e-05,
+ 3.68127366527915e-05,
+ 3.415128594497219e-05,
+ 2.7148824301548302e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0014704873319715261,
+ 0.001381139038130641,
+ 0.0013163810363039374,
+ 0.001184670371003449,
+ 0.0008946889429353178
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.0908203125,
+ "loss_eval": 2.816185712814331,
+ "epoch": 56
+ },
+ {
+ "hidden_norms": [
+ 400.9927978515625,
+ 196.3038330078125,
+ 132.07135009765625,
+ 139.7966766357422,
+ 279.3277587890625
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.982979615102522e-05,
+ 3.841472425847314e-05,
+ 3.7321347917895764e-05,
+ 3.467328497208655e-05,
+ 2.7518768547452055e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0014958448009565473,
+ 0.001403261674568057,
+ 0.0013373151887208223,
+ 0.0012032209197059274,
+ 0.0009093029075302184
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.09765625,
+ "loss_eval": 2.8296005725860596,
+ "epoch": 57
+ },
+ {
+ "hidden_norms": [
+ 401.6940612792969,
+ 195.37176513671875,
+ 131.6783447265625,
+ 139.06954956054688,
+ 268.53350830078125
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.964872666983865e-05,
+ 3.833413211395964e-05,
+ 3.708717122208327e-05,
+ 3.448158531682566e-05,
+ 2.7542468160390854e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0014878648798912764,
+ 0.0013977688504382968,
+ 0.00133328081574291,
+ 0.0012016715481877327,
+ 0.0009097974398173392
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.09375,
+ "loss_eval": 2.80366849899292,
+ "epoch": 58
+ },
+ {
+ "hidden_norms": [
+ 402.2669372558594,
+ 195.07772827148438,
+ 130.2777099609375,
+ 136.1566925048828,
+ 263.0674133300781
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.008352698292583e-05,
+ 3.8822814531158656e-05,
+ 3.763332642847672e-05,
+ 3.510116948746145e-05,
+ 2.7965274057351053e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0015098530566319823,
+ 0.0014185438631102443,
+ 0.001352492137812078,
+ 0.0012215198948979378,
+ 0.0009253643220290542
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.09375,
+ "loss_eval": 2.833949089050293,
+ "epoch": 59
+ },
+ {
+ "hidden_norms": [
+ 404.5005798339844,
+ 192.42762756347656,
+ 130.96656799316406,
+ 135.46536254882812,
+ 255.70611572265625
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.0377384721068665e-05,
+ 3.883212775690481e-05,
+ 3.7497047742363065e-05,
+ 3.505245331325568e-05,
+ 2.8039796234224923e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0015065876068547368,
+ 0.0014162716688588262,
+ 0.0013495896710082889,
+ 0.0012195424642413855,
+ 0.0009257213678210974
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.103515625,
+ "loss_eval": 2.8347878456115723,
+ "epoch": 60
+ },
+ {
+ "hidden_norms": [
+ 405.3465270996094,
+ 191.92613220214844,
+ 129.0675048828125,
+ 135.35177612304688,
+ 252.48280334472656
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.047132460982539e-05,
+ 3.890526568284258e-05,
+ 3.791163908317685e-05,
+ 3.532003393047489e-05,
+ 2.8360002033878118e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.001506851869635284,
+ 0.0014191134832799435,
+ 0.0013552922755479813,
+ 0.0012275888584554195,
+ 0.0009371190681122243
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.0947265625,
+ "loss_eval": 2.8487884998321533,
+ "epoch": 61
+ },
+ {
+ "hidden_norms": [
+ 405.92303466796875,
+ 191.02706909179688,
+ 130.1326904296875,
+ 131.96170043945312,
+ 243.21319580078125
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.0983595681609586e-05,
+ 3.9633086998946965e-05,
+ 3.8324356864904985e-05,
+ 3.588865001802333e-05,
+ 2.8768741685780697e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.001533527858555317,
+ 0.0014434503391385078,
+ 0.0013784898910671473,
+ 0.001250342931598425,
+ 0.0009516954887658358
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.099609375,
+ "loss_eval": 2.853466749191284,
+ "epoch": 62
+ },
+ {
+ "hidden_norms": [
+ 404.5056457519531,
+ 189.7686004638672,
+ 128.53379821777344,
+ 131.6268768310547,
+ 239.69944763183594
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.129076478420757e-05,
+ 3.997457315563224e-05,
+ 3.881670272676274e-05,
+ 3.6345707485452294e-05,
+ 2.9240014555398375e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0015445395838469267,
+ 0.001454798853956163,
+ 0.0013908768305554986,
+ 0.0012628707336261868,
+ 0.000962753314524889
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.09375,
+ "loss_eval": 2.864095687866211,
+ "epoch": 63
+ },
+ {
+ "hidden_norms": [
+ 406.03680419921875,
+ 189.29116821289062,
+ 127.50506591796875,
+ 131.68568420410156,
+ 233.78379821777344
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.1155919461743906e-05,
+ 4.004100264864974e-05,
+ 3.874011599691585e-05,
+ 3.628453123383224e-05,
+ 2.934946860477794e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0015430431813001633,
+ 0.0014545356389135122,
+ 0.0013910855632275343,
+ 0.0012647509574890137,
+ 0.0009666763944551349
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.095703125,
+ "loss_eval": 2.8635716438293457,
+ "epoch": 64
+ },
+ {
+ "hidden_norms": [
+ 406.0451354980469,
+ 188.50218200683594,
+ 126.54427337646484,
+ 131.3389129638672,
+ 226.51025390625
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.183232886134647e-05,
+ 4.0587648982182145e-05,
+ 3.94381677324418e-05,
+ 3.678469511214644e-05,
+ 2.9675620680791326e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.001558265183120966,
+ 0.0014702302869409323,
+ 0.0014062377158552408,
+ 0.0012798203388229012,
+ 0.0009780466789379716
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.095703125,
+ "loss_eval": 2.8676958084106445,
+ "epoch": 65
+ },
+ {
+ "hidden_norms": [
+ 408.6777038574219,
+ 187.3246307373047,
+ 127.4869155883789,
+ 129.45443725585938,
+ 217.5286102294922
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.219069887767546e-05,
+ 4.0918450395110995e-05,
+ 3.9726855902699754e-05,
+ 3.7175828765612096e-05,
+ 2.9963695851620287e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0015777976950630546,
+ 0.001490283408202231,
+ 0.0014262630138546228,
+ 0.001298666698858142,
+ 0.0009915514383465052
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.0869140625,
+ "loss_eval": 2.8961129188537598,
+ "epoch": 66
+ },
+ {
+ "hidden_norms": [
+ 407.8295593261719,
+ 186.51223754882812,
+ 126.99847412109375,
+ 128.33078002929688,
+ 215.7896270751953
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.240999624016695e-05,
+ 4.097919736523181e-05,
+ 3.9745518733980134e-05,
+ 3.734237179742195e-05,
+ 3.010852560692001e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0015762250404804945,
+ 0.00148971623275429,
+ 0.0014267999213188887,
+ 0.0013026067754253745,
+ 0.0009967804653570056
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.0927734375,
+ "loss_eval": 2.8922109603881836,
+ "epoch": 67
+ },
+ {
+ "hidden_norms": [
+ 408.5052490234375,
+ 185.35150146484375,
+ 126.04090881347656,
+ 127.63972473144531,
+ 211.70339965820312
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.2915176891256124e-05,
+ 4.142584293731488e-05,
+ 4.0363862353842705e-05,
+ 3.7879472074564546e-05,
+ 3.063846452278085e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0015976275317370892,
+ 0.001511436770670116,
+ 0.0014480556128546596,
+ 0.001321145216934383,
+ 0.001012138556689024
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.08984375,
+ "loss_eval": 2.9307312965393066,
+ "epoch": 68
+ },
+ {
+ "hidden_norms": [
+ 408.47100830078125,
+ 185.3424835205078,
+ 126.64398956298828,
+ 127.55502319335938,
+ 211.88119506835938
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.3488638766575605e-05,
+ 4.180728137725964e-05,
+ 4.041391730424948e-05,
+ 3.8047244743211195e-05,
+ 3.0868024623487145e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.00161185662727803,
+ 0.0015253758756443858,
+ 0.0014606004115194082,
+ 0.0013330455403774977,
+ 0.0010200685355812311
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.087890625,
+ "loss_eval": 2.932239532470703,
+ "epoch": 69
+ },
+ {
+ "hidden_norms": [
+ 408.52923583984375,
+ 184.1455535888672,
+ 125.52372741699219,
+ 125.75703430175781,
+ 202.25552368164062
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.348695438238792e-05,
+ 4.2108520574402064e-05,
+ 4.075234028277919e-05,
+ 3.836580435745418e-05,
+ 3.1048515666043386e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0016183900879696012,
+ 0.0015315954806283116,
+ 0.0014678208390250802,
+ 0.001341602997854352,
+ 0.0010267652105540037
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.0869140625,
+ "loss_eval": 2.9302639961242676,
+ "epoch": 70
+ },
+ {
+ "hidden_norms": [
+ 408.8894958496094,
+ 183.35772705078125,
+ 125.07991027832031,
+ 125.64179229736328,
+ 197.8079376220703
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.373578849481419e-05,
+ 4.239315967424773e-05,
+ 4.112703027203679e-05,
+ 3.845592436846346e-05,
+ 3.1269701139535755e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0016269702464342117,
+ 0.0015422124415636063,
+ 0.0014783325605094433,
+ 0.001352085149846971,
+ 0.0010349294170737267
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.091796875,
+ "loss_eval": 2.9367189407348633,
+ "epoch": 71
+ },
+ {
+ "hidden_norms": [
+ 408.4325256347656,
+ 182.9261932373047,
+ 124.68492126464844,
+ 124.7309341430664,
+ 192.49610900878906
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.388380693853833e-05,
+ 4.263465234544128e-05,
+ 4.1247698391089216e-05,
+ 3.877044946420938e-05,
+ 3.1494288123212755e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0016377014108002186,
+ 0.0015517075080424547,
+ 0.0014868414727970958,
+ 0.001360182068310678,
+ 0.0010441095801070333
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.0927734375,
+ "loss_eval": 2.930757999420166,
+ "epoch": 72
+ },
+ {
+ "hidden_norms": [
+ 408.14178466796875,
+ 182.40310668945312,
+ 124.50841522216797,
+ 123.8620834350586,
+ 190.0107879638672
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.45740224677138e-05,
+ 4.323892426327802e-05,
+ 4.188392267678864e-05,
+ 3.935809581889771e-05,
+ 3.189874405506998e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0016554963076487184,
+ 0.00156887574121356,
+ 0.0015039717545732856,
+ 0.0013761234004050493,
+ 0.0010567718418315053
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.0869140625,
+ "loss_eval": 2.9669477939605713,
+ "epoch": 73
+ },
+ {
+ "hidden_norms": [
+ 408.57916259765625,
+ 181.7321014404297,
+ 124.03689575195312,
+ 122.9704818725586,
+ 187.95928955078125
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.501524745137431e-05,
+ 4.3535896111279726e-05,
+ 4.2307488911319524e-05,
+ 3.976419611717574e-05,
+ 3.218180063413456e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0016716192476451397,
+ 0.001584019628353417,
+ 0.0015181623166427016,
+ 0.0013901714701205492,
+ 0.0010662488639354706
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.0908203125,
+ "loss_eval": 2.9710330963134766,
+ "epoch": 74
+ },
+ {
+ "hidden_norms": [
+ 408.1520690917969,
+ 180.9037322998047,
+ 124.13348388671875,
+ 123.0340576171875,
+ 186.64889526367188
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.489531056606211e-05,
+ 4.363941479823552e-05,
+ 4.237061511958018e-05,
+ 3.981809277320281e-05,
+ 3.235776239307597e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0016766563057899475,
+ 0.0015889391070231795,
+ 0.0015231993747875094,
+ 0.0013958020135760307,
+ 0.0010720851132646203
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.0927734375,
+ "loss_eval": 2.959228515625,
+ "epoch": 75
+ },
+ {
+ "hidden_norms": [
+ 409.3914794921875,
+ 180.49095153808594,
+ 123.59921264648438,
+ 122.37818145751953,
+ 181.8359375
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.522506060311571e-05,
+ 4.3832860683323815e-05,
+ 4.2538729758234695e-05,
+ 3.9995622501010075e-05,
+ 3.253304384998046e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0016819218872115016,
+ 0.001593844499439001,
+ 0.0015282074455171824,
+ 0.0014012405881658196,
+ 0.0010784030891954899
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.0927734375,
+ "loss_eval": 2.961671829223633,
+ "epoch": 76
+ },
+ {
+ "hidden_norms": [
+ 407.241455078125,
+ 180.02792358398438,
+ 123.85694885253906,
+ 122.00341033935547,
+ 179.4743194580078
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.5559816499007866e-05,
+ 4.4125670683570206e-05,
+ 4.277457628631964e-05,
+ 4.0345701563637704e-05,
+ 3.2811487471917644e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0016946949763223529,
+ 0.0016066118841990829,
+ 0.0015405946178361773,
+ 0.0014128233306109905,
+ 0.0010864713694900274
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.09765625,
+ "loss_eval": 2.971811294555664,
+ "epoch": 77
+ },
+ {
+ "hidden_norms": [
+ 408.3240661621094,
+ 179.1845245361328,
+ 123.02991485595703,
+ 121.22331237792969,
+ 176.96409606933594
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.576908759190701e-05,
+ 4.430625631357543e-05,
+ 4.3082043703179806e-05,
+ 4.0411658119410276e-05,
+ 3.2916912459768355e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0016961582005023956,
+ 0.0016096236649900675,
+ 0.0015443011652678251,
+ 0.0014174508396536112,
+ 0.0010930649004876614
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.0947265625,
+ "loss_eval": 2.974250316619873,
+ "epoch": 78
+ },
+ {
+ "hidden_norms": [
+ 409.5758972167969,
+ 178.83154296875,
+ 122.89900207519531,
+ 120.96637725830078,
+ 176.27029418945312
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.575210914481431e-05,
+ 4.45544756075833e-05,
+ 4.321682718000375e-05,
+ 4.0704584534978494e-05,
+ 3.307598308310844e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0017041281098499894,
+ 0.001617106725461781,
+ 0.0015517985448241234,
+ 0.0014248047955334187,
+ 0.0010987903224304318
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.0888671875,
+ "loss_eval": 2.9865829944610596,
+ "epoch": 79
+ },
+ {
+ "hidden_norms": [
+ 408.8537902832031,
+ 178.98573303222656,
+ 122.46696472167969,
+ 121.31299591064453,
+ 173.41905212402344
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.6213288442231715e-05,
+ 4.482917211134918e-05,
+ 4.371384784462862e-05,
+ 4.093564712093212e-05,
+ 3.334296343382448e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.00171999994199723,
+ 0.001632340019568801,
+ 0.001565982005558908,
+ 0.0014375611208379269,
+ 0.0011084020370617509
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.0927734375,
+ "loss_eval": 2.993680953979492,
+ "epoch": 80
+ },
+ {
+ "hidden_norms": [
+ 408.5456848144531,
+ 178.0683135986328,
+ 122.94597625732422,
+ 119.80611419677734,
+ 171.1206817626953
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.6296096115838736e-05,
+ 4.503710806602612e-05,
+ 4.384905696497299e-05,
+ 4.111307498533279e-05,
+ 3.349714825162664e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0017213152023032308,
+ 0.0016335739055648446,
+ 0.0015675420872867107,
+ 0.001440141350030899,
+ 0.001112470985390246
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.0927734375,
+ "loss_eval": 2.994321346282959,
+ "epoch": 81
+ },
+ {
+ "hidden_norms": [
+ 409.3093566894531,
+ 177.68714904785156,
+ 122.89625549316406,
+ 119.95466613769531,
+ 168.89804077148438
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.6671568270539865e-05,
+ 4.528012505033985e-05,
+ 4.418051685206592e-05,
+ 4.1504132241243497e-05,
+ 3.381213900865987e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.001732644159346819,
+ 0.0016448728274554014,
+ 0.0015783295966684818,
+ 0.001450547482818365,
+ 0.0011199831496924162
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.0947265625,
+ "loss_eval": 3.004246950149536,
+ "epoch": 82
+ },
+ {
+ "hidden_norms": [
+ 408.72955322265625,
+ 177.53179931640625,
+ 121.72450256347656,
+ 119.71387481689453,
+ 168.49583435058594
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.686604006565176e-05,
+ 4.567100768326782e-05,
+ 4.437788811628707e-05,
+ 4.169986641500145e-05,
+ 3.4023832995444536e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0017455043271183968,
+ 0.0016571393935009837,
+ 0.0015903809107840061,
+ 0.0014616175321862102,
+ 0.001127574942074716
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.09375,
+ "loss_eval": 3.015833854675293,
+ "epoch": 83
+ },
+ {
+ "hidden_norms": [
+ 409.93121337890625,
+ 177.41183471679688,
+ 121.63835144042969,
+ 119.78266906738281,
+ 166.5583953857422
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.7002995415823534e-05,
+ 4.566587813314982e-05,
+ 4.4412678107619286e-05,
+ 4.1799670725595206e-05,
+ 3.404032759135589e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0017475533531978726,
+ 0.0016598458169028163,
+ 0.0015934593975543976,
+ 0.0014648212818428874,
+ 0.0011309005785733461
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.095703125,
+ "loss_eval": 3.014139413833618,
+ "epoch": 84
+ },
+ {
+ "hidden_norms": [
+ 409.67083740234375,
+ 177.14193725585938,
+ 121.51908111572266,
+ 119.6811294555664,
+ 166.031982421875
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.713247108156793e-05,
+ 4.586786963045597e-05,
+ 4.4539196096593514e-05,
+ 4.203089338261634e-05,
+ 3.428043783060275e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0017551736673340201,
+ 0.001667042262852192,
+ 0.0016005010111257434,
+ 0.0014717906014993787,
+ 0.0011371106375008821
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.087890625,
+ "loss_eval": 3.0291876792907715,
+ "epoch": 85
+ },
+ {
+ "hidden_norms": [
+ 410.37042236328125,
+ 176.73268127441406,
+ 121.4700698852539,
+ 119.30713653564453,
+ 163.45822143554688
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.7194233047775924e-05,
+ 4.609613824868575e-05,
+ 4.4591557525563985e-05,
+ 4.207295205560513e-05,
+ 3.430316428421065e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.001759424339979887,
+ 0.0016711216885596514,
+ 0.0016040647169575095,
+ 0.0014759227633476257,
+ 0.001140325446613133
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.09375,
+ "loss_eval": 3.0314245223999023,
+ "epoch": 86
+ },
+ {
+ "hidden_norms": [
+ 409.8002624511719,
+ 176.60916137695312,
+ 121.219970703125,
+ 118.81930541992188,
+ 164.01478576660156
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.7294168325606734e-05,
+ 4.6120272600091994e-05,
+ 4.4749522203346714e-05,
+ 4.225381417199969e-05,
+ 3.438722342252731e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0017650157678872347,
+ 0.0016766699263826013,
+ 0.0016095110913738608,
+ 0.0014808655250817537,
+ 0.0011436669155955315
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.0927734375,
+ "loss_eval": 3.030081272125244,
+ "epoch": 87
+ },
+ {
+ "hidden_norms": [
+ 409.7901306152344,
+ 176.61012268066406,
+ 121.13548278808594,
+ 118.5867919921875,
+ 162.41827392578125
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.7548022848786786e-05,
+ 4.6351164201041684e-05,
+ 4.494565655477345e-05,
+ 4.2309966374887154e-05,
+ 3.455657861195505e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0017697742441669106,
+ 0.001681560999713838,
+ 0.00161420414224267,
+ 0.0014853639295324683,
+ 0.0011477968655526638
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.09375,
+ "loss_eval": 3.0332422256469727,
+ "epoch": 88
+ },
+ {
+ "hidden_norms": [
+ 409.9271240234375,
+ 176.3477783203125,
+ 120.93573760986328,
+ 118.68280029296875,
+ 161.79257202148438
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.757043279823847e-05,
+ 4.634608922060579e-05,
+ 4.5027496526017785e-05,
+ 4.236193490214646e-05,
+ 3.461442975094542e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0017714434070512652,
+ 0.0016832581022754312,
+ 0.0016157239442691207,
+ 0.0014867670834064484,
+ 0.0011494142236188054
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.091796875,
+ "loss_eval": 3.0295891761779785,
+ "epoch": 89
+ },
+ {
+ "hidden_norms": [
+ 409.3666076660156,
+ 176.4572296142578,
+ 121.02136993408203,
+ 118.42733001708984,
+ 161.92198181152344
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.7602890845155343e-05,
+ 4.65112243546173e-05,
+ 4.5134373067412525e-05,
+ 4.2457497329451144e-05,
+ 3.4702385164564475e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0017726266523823142,
+ 0.0016847110819071531,
+ 0.0016171842580661178,
+ 0.0014885812997817993,
+ 0.001151244156062603
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.09375,
+ "loss_eval": 3.0332508087158203,
+ "epoch": 90
+ },
+ {
+ "hidden_norms": [
+ 408.93438720703125,
+ 176.05850219726562,
+ 121.09625244140625,
+ 118.71532440185547,
+ 160.93397521972656
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.767462087329477e-05,
+ 4.656092278310098e-05,
+ 4.517777415458113e-05,
+ 4.260002606315538e-05,
+ 3.481149542494677e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0017778765177354217,
+ 0.0016896515153348446,
+ 0.0016218661330640316,
+ 0.0014927292941138148,
+ 0.0011542538413777947
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.0927734375,
+ "loss_eval": 3.0385818481445312,
+ "epoch": 91
+ },
+ {
+ "hidden_norms": [
+ 408.2444763183594,
+ 176.17698669433594,
+ 120.81340789794922,
+ 118.79690551757812,
+ 160.8688507080078
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.765141420648433e-05,
+ 4.667821849579923e-05,
+ 4.5252400013851e-05,
+ 4.269158671377227e-05,
+ 3.480418672552332e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0017795232124626637,
+ 0.0016910110134631395,
+ 0.0016234440263360739,
+ 0.0014941159170120955,
+ 0.0011556288227438927
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.0908203125,
+ "loss_eval": 3.0389466285705566,
+ "epoch": 92
+ },
+ {
+ "hidden_norms": [
+ 408.58831787109375,
+ 176.04327392578125,
+ 120.86355590820312,
+ 118.54495239257812,
+ 160.07017517089844
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.780395101988688e-05,
+ 4.6777106035733595e-05,
+ 4.532063030637801e-05,
+ 4.271055877325125e-05,
+ 3.487269350443967e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.001782167237251997,
+ 0.0016936222091317177,
+ 0.0016259904950857162,
+ 0.0014966638991609216,
+ 0.0011574920499697328
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.091796875,
+ "loss_eval": 3.0411715507507324,
+ "epoch": 93
+ },
+ {
+ "hidden_norms": [
+ 408.69635009765625,
+ 175.86227416992188,
+ 120.84530639648438,
+ 118.5611572265625,
+ 159.65382385253906
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.7862777137197554e-05,
+ 4.672469003708102e-05,
+ 4.5333617890719324e-05,
+ 4.2781779484357685e-05,
+ 3.488900983938947e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.001782999956049025,
+ 0.0016942995134741068,
+ 0.0016265579033643007,
+ 0.001497312681749463,
+ 0.001158081111498177
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.0927734375,
+ "loss_eval": 3.0412378311157227,
+ "epoch": 94
+ },
+ {
+ "hidden_norms": [
+ 408.5793151855469,
+ 175.8729705810547,
+ 120.83885192871094,
+ 118.5599365234375,
+ 159.589111328125
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.793615153175779e-05,
+ 4.6734887291677296e-05,
+ 4.529079524218105e-05,
+ 4.2848438170040026e-05,
+ 3.4947173844557256e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0017846085829660296,
+ 0.0016958917258307338,
+ 0.001628118334338069,
+ 0.0014987689210101962,
+ 0.0011592835653573275
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.0908203125,
+ "loss_eval": 3.043121576309204,
+ "epoch": 95
+ },
+ {
+ "hidden_norms": [
+ 408.6483459472656,
+ 175.8555908203125,
+ 120.81853485107422,
+ 118.52471160888672,
+ 159.30239868164062
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.796134089701809e-05,
+ 4.68125035695266e-05,
+ 4.539822111837566e-05,
+ 4.2882278648903593e-05,
+ 3.495233249850571e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0017852310556918383,
+ 0.0016965895192697644,
+ 0.001628802390769124,
+ 0.0014994175871834159,
+ 0.0011597864795476198
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.0908203125,
+ "loss_eval": 3.0433754920959473,
+ "epoch": 96
+ },
+ {
+ "hidden_norms": [
+ 408.78167724609375,
+ 175.81280517578125,
+ 120.72789001464844,
+ 118.362060546875,
+ 159.2758026123047
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.7979057853808627e-05,
+ 4.6818935516057536e-05,
+ 4.545195770333521e-05,
+ 4.2866002331720665e-05,
+ 3.4974782465724275e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0017855397891253233,
+ 0.0016968693817034364,
+ 0.0016290779458358884,
+ 0.001499759964644909,
+ 0.001160072861239314
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.091796875,
+ "loss_eval": 3.0432236194610596,
+ "epoch": 97
+ },
+ {
+ "hidden_norms": [
+ 408.695068359375,
+ 175.82485961914062,
+ 120.68985748291016,
+ 118.31425476074219,
+ 159.31414794921875
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.796995563083328e-05,
+ 4.685561725636944e-05,
+ 4.54587752756197e-05,
+ 4.2861738620558754e-05,
+ 3.4979620977537706e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0017854906618595123,
+ 0.0016968429554253817,
+ 0.001629071426577866,
+ 0.0014998035039752722,
+ 0.0011602065060287714
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.091796875,
+ "loss_eval": 3.0435032844543457,
+ "epoch": 98
+ },
+ {
+ "hidden_norms": [
+ 408.69329833984375,
+ 175.80587768554688,
+ 120.6858901977539,
+ 118.34341430664062,
+ 159.32943725585938
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.796809298568405e-05,
+ 4.686374086304568e-05,
+ 4.545971387415193e-05,
+ 4.28554521931801e-05,
+ 3.4983324439963326e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0017855679616332054,
+ 0.0016969236312434077,
+ 0.001629148842766881,
+ 0.0014998704427853227,
+ 0.0011602530721575022
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.091796875,
+ "loss_eval": 3.043445348739624,
+ "epoch": 99
+ },
+ {
+ "hidden_norms": [
+ 408.6950378417969,
+ 175.79644775390625,
+ 120.68733978271484,
+ 118.34892272949219,
+ 159.32240295410156
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.7968002036213875e-05,
+ 4.6864570322213694e-05,
+ 4.546308991848491e-05,
+ 4.2857409425778314e-05,
+ 3.498443402349949e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0017855766927823424,
+ 0.0016969360876828432,
+ 0.0016291660722345114,
+ 0.0014998912811279297,
+ 0.0011602782178670168
+ ],
+ "gamma_dfa": NaN,
+ "acc_eval": 0.091796875,
+ "loss_eval": 3.0435147285461426,
+ "epoch": 100
+ }
+ ],
+ "dfa_log": [
+ {
+ "hidden_norms": [
+ 8.718070030212402,
+ 42.802223205566406,
+ 44.22697067260742,
+ 46.05121994018555,
+ 46.896766662597656
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 0.0010791289387270808,
+ 0.0002035950165009126,
+ 0.00014324457151815295,
+ 0.00011542177526280284,
+ 9.846547618508339e-05
+ ],
+ "bp_grad_norms_F": [
+ 0.0380445197224617,
+ 0.00666962843388319,
+ 0.0046500470489263535,
+ 0.003719294909387827,
+ 0.003155440790578723
+ ],
+ "gamma_dfa": 0.004864378133788705,
+ "acc_eval": 0.080078125,
+ "loss_eval": 2.446239709854126,
+ "epoch": 0
+ },
+ {
+ "hidden_norms": [
+ 1180.2318115234375,
+ 5768.2900390625,
+ 5472.90234375,
+ 6369.3427734375,
+ 5295.2470703125
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 1.2146981589467032e-06,
+ 8.634061714474228e-07,
+ 8.587696811446222e-07,
+ 8.589160529481887e-07,
+ 8.53334995554178e-07
+ ],
+ "bp_grad_norms_F": [
+ 6.592504360014573e-05,
+ 4.607137088896707e-05,
+ 4.4360855099512264e-05,
+ 4.341698149801232e-05,
+ 4.290258584660478e-05
+ ],
+ "gamma_dfa": 0.0467059058137238,
+ "acc_eval": 0.15625,
+ "loss_eval": 2.2713985443115234,
+ "epoch": 1
+ },
+ {
+ "hidden_norms": [
+ 2712.02197265625,
+ 35799.15625,
+ 30196.22265625,
+ 31429.328125,
+ 21571.224609375
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.2316361853190756e-07,
+ 1.6252558054929978e-07,
+ 1.6230175958753534e-07,
+ 1.6301386551731412e-07,
+ 1.6222111298702657e-07
+ ],
+ "bp_grad_norms_F": [
+ 2.881550244637765e-05,
+ 1.2372604032862e-05,
+ 1.1472440746729262e-05,
+ 1.1236465070396662e-05,
+ 1.1124561751785222e-05
+ ],
+ "gamma_dfa": 0.035421818611212075,
+ "acc_eval": 0.146484375,
+ "loss_eval": 2.335322856903076,
+ "epoch": 2
+ },
+ {
+ "hidden_norms": [
+ 4310.83203125,
+ 104622.7734375,
+ 77630.2421875,
+ 78666.21875,
+ 48679.88671875
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 1.8400233159354684e-07,
+ 6.495802296058173e-08,
+ 6.498943605492968e-08,
+ 6.491647042139448e-08,
+ 6.464611601586512e-08
+ ],
+ "bp_grad_norms_F": [
+ 1.769972368492745e-05,
+ 5.158871317689773e-06,
+ 4.676893240684876e-06,
+ 4.603746219800087e-06,
+ 4.549533514364157e-06
+ ],
+ "gamma_dfa": 0.03363112888473552,
+ "acc_eval": 0.150390625,
+ "loss_eval": 2.4106125831604004,
+ "epoch": 3
+ },
+ {
+ "hidden_norms": [
+ 5880.8115234375,
+ 227753.375,
+ 152491.703125,
+ 155418.390625,
+ 88801.296875
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 1.2607929988917022e-07,
+ 3.377261137416099e-08,
+ 3.3712492353288326e-08,
+ 3.3661436305010284e-08,
+ 3.362825395925029e-08
+ ],
+ "bp_grad_norms_F": [
+ 1.548601358081214e-05,
+ 2.78187621916004e-06,
+ 2.4831683731463272e-06,
+ 2.456902620906476e-06,
+ 2.4304001726704882e-06
+ ],
+ "gamma_dfa": 0.027950844378210604,
+ "acc_eval": 0.1494140625,
+ "loss_eval": 2.473954677581787,
+ "epoch": 4
+ },
+ {
+ "hidden_norms": [
+ 7446.83203125,
+ 415853.34375,
+ 244149.921875,
+ 253104.71875,
+ 141041.109375
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 1.0277101125666377e-07,
+ 2.1167430475088622e-08,
+ 2.113832486827505e-08,
+ 2.1176427722480184e-08,
+ 2.102850160667913e-08
+ ],
+ "bp_grad_norms_F": [
+ 1.4487833141174633e-05,
+ 2.0795134787476854e-06,
+ 1.5597894389429712e-06,
+ 1.522601451142691e-06,
+ 1.5034931948321173e-06
+ ],
+ "gamma_dfa": 0.02746052150905598,
+ "acc_eval": 0.1435546875,
+ "loss_eval": 2.5455238819122314,
+ "epoch": 5
+ },
+ {
+ "hidden_norms": [
+ 8986.072265625,
+ 683627.0,
+ 382622.5625,
+ 394341.96875,
+ 225869.703125
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 8.362774650549909e-08,
+ 1.4089099664715832e-08,
+ 1.4079772014952141e-08,
+ 1.4096787737116756e-08,
+ 1.4042676355074946e-08
+ ],
+ "bp_grad_norms_F": [
+ 1.094696290238062e-05,
+ 1.43898739679571e-06,
+ 1.0249204933643341e-06,
+ 1.0037305173682398e-06,
+ 9.89082195701485e-07
+ ],
+ "gamma_dfa": 0.022385360905900598,
+ "acc_eval": 0.1416015625,
+ "loss_eval": 2.617079973220825,
+ "epoch": 6
+ },
+ {
+ "hidden_norms": [
+ 10599.609375,
+ 1064012.5,
+ 545495.875,
+ 573164.9375,
+ 330028.71875
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 7.213265718064577e-08,
+ 1.0183534016050544e-08,
+ 1.0178969667151705e-08,
+ 1.0194342259239875e-08,
+ 1.0160833951999848e-08
+ ],
+ "bp_grad_norms_F": [
+ 1.0159021258004941e-05,
+ 8.805947686596483e-07,
+ 7.233961696329061e-07,
+ 7.142662639125774e-07,
+ 7.038501053102664e-07
+ ],
+ "gamma_dfa": 0.014009941252879798,
+ "acc_eval": 0.1337890625,
+ "loss_eval": 2.7312068939208984,
+ "epoch": 7
+ },
+ {
+ "hidden_norms": [
+ 12143.3994140625,
+ 1557775.25,
+ 759262.3125,
+ 792826.0,
+ 460775.59375
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 6.697734988847515e-08,
+ 7.604014307105444e-09,
+ 7.596973716772482e-09,
+ 7.606048235686558e-09,
+ 7.604833207608408e-09
+ ],
+ "bp_grad_norms_F": [
+ 9.400837370776571e-06,
+ 6.404222290257167e-07,
+ 5.235326625552261e-07,
+ 5.169007977201545e-07,
+ 5.090913077765435e-07
+ ],
+ "gamma_dfa": 0.011664194520562887,
+ "acc_eval": 0.1328125,
+ "loss_eval": 2.825345516204834,
+ "epoch": 8
+ },
+ {
+ "hidden_norms": [
+ 13694.29296875,
+ 2172737.5,
+ 997486.0625,
+ 1070732.625,
+ 629918.4375
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 6.409224795334012e-08,
+ 6.097121918458015e-09,
+ 6.086279480399526e-09,
+ 6.08825212466968e-09,
+ 6.0656004663428575e-09
+ ],
+ "bp_grad_norms_F": [
+ 9.614421287551522e-06,
+ 5.113795964462042e-07,
+ 4.071621901857725e-07,
+ 4.02141125732669e-07,
+ 3.9655543560002116e-07
+ ],
+ "gamma_dfa": 0.008997007156722248,
+ "acc_eval": 0.1298828125,
+ "loss_eval": 2.988190174102783,
+ "epoch": 9
+ },
+ {
+ "hidden_norms": [
+ 15223.474609375,
+ 2927110.5,
+ 1275342.625,
+ 1378191.375,
+ 828821.8125
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 6.029451782296746e-08,
+ 5.071707498416345e-09,
+ 5.069247244193775e-09,
+ 5.075359243988942e-09,
+ 5.0706532306321606e-09
+ ],
+ "bp_grad_norms_F": [
+ 9.640240023145452e-06,
+ 4.242011186761374e-07,
+ 3.2627585255795566e-07,
+ 3.2204775379796047e-07,
+ 3.1740532335788885e-07
+ ],
+ "gamma_dfa": 0.0051322957733646035,
+ "acc_eval": 0.125,
+ "loss_eval": 3.152592658996582,
+ "epoch": 10
+ },
+ {
+ "hidden_norms": [
+ 16733.46484375,
+ 3849004.0,
+ 1605019.5,
+ 1758574.0,
+ 1046866.9375
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 5.804990976798763e-08,
+ 4.271067943761864e-09,
+ 4.260845898329535e-09,
+ 4.262723063419571e-09,
+ 4.251643481723022e-09
+ ],
+ "bp_grad_norms_F": [
+ 9.824444532569032e-06,
+ 3.8355952369784063e-07,
+ 2.739863020906341e-07,
+ 2.699064225453185e-07,
+ 2.6583589374240546e-07
+ ],
+ "gamma_dfa": 0.002556616673246026,
+ "acc_eval": 0.125,
+ "loss_eval": 3.3216772079467773,
+ "epoch": 11
+ },
+ {
+ "hidden_norms": [
+ 18226.0703125,
+ 4895676.0,
+ 1964835.75,
+ 2174918.0,
+ 1297900.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 5.8149943527041614e-08,
+ 3.753563682096228e-09,
+ 3.750809440816738e-09,
+ 3.756827737788626e-09,
+ 3.745073584582315e-09
+ ],
+ "bp_grad_norms_F": [
+ 1.0199929420195986e-05,
+ 3.2010626682676957e-07,
+ 2.3428944473380398e-07,
+ 2.3066610310706892e-07,
+ 2.272569474826014e-07
+ ],
+ "gamma_dfa": -0.001056009205058217,
+ "acc_eval": 0.1162109375,
+ "loss_eval": 3.5638060569763184,
+ "epoch": 12
+ },
+ {
+ "hidden_norms": [
+ 19695.107421875,
+ 6176501.5,
+ 2397138.0,
+ 2692062.25,
+ 1632551.125
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 5.818218085096305e-08,
+ 3.2458047360961473e-09,
+ 3.240681722971317e-09,
+ 3.251705793516635e-09,
+ 3.2418265849543104e-09
+ ],
+ "bp_grad_norms_F": [
+ 1.0194554306508508e-05,
+ 2.884790433199669e-07,
+ 2.0161056113465747e-07,
+ 1.9845649035232782e-07,
+ 1.9566277842386626e-07
+ ],
+ "gamma_dfa": -0.003356708213686943,
+ "acc_eval": 0.11328125,
+ "loss_eval": 3.7695584297180176,
+ "epoch": 13
+ },
+ {
+ "hidden_norms": [
+ 21182.841796875,
+ 7649758.0,
+ 2859185.75,
+ 3238327.25,
+ 2017791.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 5.6009717752658617e-08,
+ 2.8617739289416022e-09,
+ 2.86194334897516e-09,
+ 2.864514181410982e-09,
+ 2.8607307633876644e-09
+ ],
+ "bp_grad_norms_F": [
+ 9.991773367801215e-06,
+ 2.805488747981144e-07,
+ 1.7382579642344353e-07,
+ 1.7105583083321108e-07,
+ 1.6860903429005702e-07
+ ],
+ "gamma_dfa": -0.004406693275086582,
+ "acc_eval": 0.1103515625,
+ "loss_eval": 3.9298348426818848,
+ "epoch": 14
+ },
+ {
+ "hidden_norms": [
+ 22575.759765625,
+ 9260891.0,
+ 3426286.0,
+ 3904244.75,
+ 2469897.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 5.523341073399024e-08,
+ 2.458340642164103e-09,
+ 2.4571535917061738e-09,
+ 2.4605368853514165e-09,
+ 2.455863956640769e-09
+ ],
+ "bp_grad_norms_F": [
+ 9.669157407188322e-06,
+ 2.321831686913356e-07,
+ 1.4871888254219812e-07,
+ 1.466667072236305e-07,
+ 1.44689224157446e-07
+ ],
+ "gamma_dfa": -0.005493569537065923,
+ "acc_eval": 0.11328125,
+ "loss_eval": 4.065067291259766,
+ "epoch": 15
+ },
+ {
+ "hidden_norms": [
+ 23971.93359375,
+ 11148851.0,
+ 3986279.5,
+ 4682323.5,
+ 2980997.5
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 5.2560817920266345e-08,
+ 2.1839356989517e-09,
+ 2.1809569705766307e-09,
+ 2.1860839805043497e-09,
+ 2.177982905138265e-09
+ ],
+ "bp_grad_norms_F": [
+ 9.612061148800422e-06,
+ 2.2171894897837774e-07,
+ 1.2985647401819733e-07,
+ 1.2793493908702658e-07,
+ 1.2610371413757093e-07
+ ],
+ "gamma_dfa": -0.0057282960042357445,
+ "acc_eval": 0.1103515625,
+ "loss_eval": 4.23506498336792,
+ "epoch": 16
+ },
+ {
+ "hidden_norms": [
+ 25387.59765625,
+ 13199074.0,
+ 4584581.0,
+ 5421912.0,
+ 3520717.25
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 5.063454011633439e-08,
+ 1.9527632844074105e-09,
+ 1.9520551841623046e-09,
+ 1.956988127105319e-09,
+ 1.9527153227727467e-09
+ ],
+ "bp_grad_norms_F": [
+ 9.531939213047735e-06,
+ 1.855236746450828e-07,
+ 1.1443902536711903e-07,
+ 1.1294694246544168e-07,
+ 1.1146746459189671e-07
+ ],
+ "gamma_dfa": -0.00637711986200884,
+ "acc_eval": 0.1123046875,
+ "loss_eval": 4.372166633605957,
+ "epoch": 17
+ },
+ {
+ "hidden_norms": [
+ 26803.72265625,
+ 15607512.0,
+ 5231978.5,
+ 6327852.0,
+ 4171921.5
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 5.073666287103151e-08,
+ 1.7431257548494727e-09,
+ 1.7403234409130164e-09,
+ 1.7429053755790846e-09,
+ 1.737441190918787e-09
+ ],
+ "bp_grad_norms_F": [
+ 9.275222510041203e-06,
+ 1.6788693812941347e-07,
+ 1.0197032906944514e-07,
+ 1.0054727539454689e-07,
+ 9.927033062240298e-08
+ ],
+ "gamma_dfa": -0.006377041107043624,
+ "acc_eval": 0.1142578125,
+ "loss_eval": 4.511043071746826,
+ "epoch": 18
+ },
+ {
+ "hidden_norms": [
+ 28166.5859375,
+ 18256806.0,
+ 6062269.0,
+ 7361214.0,
+ 4908964.5
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.8937458529962896e-08,
+ 1.5624438409744812e-09,
+ 1.5615452264583496e-09,
+ 1.5637011685498692e-09,
+ 1.559517959215384e-09
+ ],
+ "bp_grad_norms_F": [
+ 9.249049981008284e-06,
+ 1.5261937846844376e-07,
+ 9.117609778286351e-08,
+ 8.977448828773049e-08,
+ 8.863072764597746e-08
+ ],
+ "gamma_dfa": -0.007660938150365837,
+ "acc_eval": 0.1171875,
+ "loss_eval": 4.64661979675293,
+ "epoch": 19
+ },
+ {
+ "hidden_norms": [
+ 29548.751953125,
+ 21145650.0,
+ 6936910.0,
+ 8491332.0,
+ 5632708.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.774140904828528e-08,
+ 1.397353122101208e-09,
+ 1.3845558033409588e-09,
+ 1.3843480806130515e-09,
+ 1.3776775276141962e-09
+ ],
+ "bp_grad_norms_F": [
+ 9.035198672791012e-06,
+ 1.426667637360879e-07,
+ 8.229534387282911e-08,
+ 8.094935566305139e-08,
+ 7.995085837819715e-08
+ ],
+ "gamma_dfa": -0.00677196380274836,
+ "acc_eval": 0.111328125,
+ "loss_eval": 4.76383113861084,
+ "epoch": 20
+ },
+ {
+ "hidden_norms": [
+ 30869.275390625,
+ 24201464.0,
+ 7880455.0,
+ 9702563.0,
+ 6424945.5
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.703911571368735e-08,
+ 1.2679121086378586e-09,
+ 1.2570423590929636e-09,
+ 1.2592289433399628e-09,
+ 1.2545631200566731e-09
+ ],
+ "bp_grad_norms_F": [
+ 8.886428986443207e-06,
+ 1.178441522142748e-07,
+ 7.369149557234778e-08,
+ 7.258120149344904e-08,
+ 7.175417238158843e-08
+ ],
+ "gamma_dfa": -0.008610141419922002,
+ "acc_eval": 0.1162109375,
+ "loss_eval": 4.926914691925049,
+ "epoch": 21
+ },
+ {
+ "hidden_norms": [
+ 32181.666015625,
+ 27506424.0,
+ 8773789.0,
+ 11014517.0,
+ 7322481.5
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.4904634677322974e-08,
+ 1.1420530077188573e-09,
+ 1.121353787603141e-09,
+ 1.124021098419803e-09,
+ 1.120450510150306e-09
+ ],
+ "bp_grad_norms_F": [
+ 8.90475439518923e-06,
+ 1.1447492909155699e-07,
+ 6.648813410947696e-08,
+ 6.533218765980564e-08,
+ 6.45067217419637e-08
+ ],
+ "gamma_dfa": -0.007374132976110559,
+ "acc_eval": 0.111328125,
+ "loss_eval": 5.004192352294922,
+ "epoch": 22
+ },
+ {
+ "hidden_norms": [
+ 33471.7265625,
+ 31266642.0,
+ 9769161.0,
+ 12538190.0,
+ 8328704.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.517446683394155e-08,
+ 1.0276371975592724e-09,
+ 1.005762251260478e-09,
+ 1.0072189748910887e-09,
+ 1.0049648890841922e-09
+ ],
+ "bp_grad_norms_F": [
+ 8.657435500936117e-06,
+ 1.0238768766157591e-07,
+ 6.021300436032107e-08,
+ 5.9172805322305067e-08,
+ 5.8442100936417773e-08
+ ],
+ "gamma_dfa": -0.006727739999405458,
+ "acc_eval": 0.1123046875,
+ "loss_eval": 5.116036415100098,
+ "epoch": 23
+ },
+ {
+ "hidden_norms": [
+ 34792.7890625,
+ 35134068.0,
+ 11038143.0,
+ 14059113.0,
+ 9444097.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.4753964090205045e-08,
+ 9.320837435211615e-10,
+ 9.107828935484008e-10,
+ 9.128516276213361e-10,
+ 9.105118325969386e-10
+ ],
+ "bp_grad_norms_F": [
+ 8.577152584621217e-06,
+ 9.559200009334745e-08,
+ 5.512932332862874e-08,
+ 5.411101611230151e-08,
+ 5.343352782460897e-08
+ ],
+ "gamma_dfa": -0.00556181388310506,
+ "acc_eval": 0.1142578125,
+ "loss_eval": 5.2026848793029785,
+ "epoch": 24
+ },
+ {
+ "hidden_norms": [
+ 36056.9921875,
+ 39258876.0,
+ 12277943.0,
+ 15686179.0,
+ 10635754.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 4.2437520164639864e-08,
+ 8.431049747237296e-10,
+ 8.279052998716452e-10,
+ 8.256172967513464e-10,
+ 8.249605998322806e-10
+ ],
+ "bp_grad_norms_F": [
+ 8.401701961702202e-06,
+ 9.383651189409647e-08,
+ 5.072890729707069e-08,
+ 4.9743736241225633e-08,
+ 4.90946696629635e-08
+ ],
+ "gamma_dfa": -0.004373190506157698,
+ "acc_eval": 0.115234375,
+ "loss_eval": 5.283636569976807,
+ "epoch": 25
+ },
+ {
+ "hidden_norms": [
+ 37332.3203125,
+ 43838448.0,
+ 13573116.0,
+ 17373072.0,
+ 11853122.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.96461210527832e-08,
+ 7.721222550216567e-10,
+ 7.523985323665272e-10,
+ 7.522342748700339e-10,
+ 7.507660604311184e-10
+ ],
+ "bp_grad_norms_F": [
+ 8.241736395575572e-06,
+ 8.553880093131738e-08,
+ 4.659301566789509e-08,
+ 4.568136446891913e-08,
+ 4.507716511170656e-08
+ ],
+ "gamma_dfa": -0.003043898326723138,
+ "acc_eval": 0.1123046875,
+ "loss_eval": 5.340336322784424,
+ "epoch": 26
+ },
+ {
+ "hidden_norms": [
+ 38604.234375,
+ 48637388.0,
+ 14943648.0,
+ 19237954.0,
+ 13168041.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.845103080379886e-08,
+ 7.063521434425013e-10,
+ 6.949120723298563e-10,
+ 6.971247468179342e-10,
+ 6.954581355245182e-10
+ ],
+ "bp_grad_norms_F": [
+ 8.12488178780768e-06,
+ 8.518332350604396e-08,
+ 4.348786220020884e-08,
+ 4.2559751278758995e-08,
+ 4.1998429622935873e-08
+ ],
+ "gamma_dfa": -0.0021704566534026526,
+ "acc_eval": 0.1103515625,
+ "loss_eval": 5.403829574584961,
+ "epoch": 27
+ },
+ {
+ "hidden_norms": [
+ 39856.5078125,
+ 53569880.0,
+ 16230254.0,
+ 21087022.0,
+ 14556169.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.739526022172868e-08,
+ 6.497330451438188e-10,
+ 6.402843810704439e-10,
+ 6.384167083872683e-10,
+ 6.366632776533265e-10
+ ],
+ "bp_grad_norms_F": [
+ 8.042790796025656e-06,
+ 7.930053413929272e-08,
+ 4.0553047142566356e-08,
+ 3.9663138551304655e-08,
+ 3.909939749746627e-08
+ ],
+ "gamma_dfa": 0.00027063591551268473,
+ "acc_eval": 0.11328125,
+ "loss_eval": 5.486921310424805,
+ "epoch": 28
+ },
+ {
+ "hidden_norms": [
+ 41114.66015625,
+ 59315272.0,
+ 17822804.0,
+ 22986032.0,
+ 16084622.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.655359392951141e-08,
+ 6.013742281929524e-10,
+ 5.905386180060646e-10,
+ 5.911486300469448e-10,
+ 5.910953393417628e-10
+ ],
+ "bp_grad_norms_F": [
+ 7.960659786476754e-06,
+ 8.031720000190035e-08,
+ 3.805743276075191e-08,
+ 3.7164745947393385e-08,
+ 3.6596219388229656e-08
+ ],
+ "gamma_dfa": 0.0034554791564005427,
+ "acc_eval": 0.1142578125,
+ "loss_eval": 5.604689598083496,
+ "epoch": 29
+ },
+ {
+ "hidden_norms": [
+ 42325.58984375,
+ 64740044.0,
+ 19249378.0,
+ 25010146.0,
+ 17657480.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.541291704323157e-08,
+ 5.539400049769938e-10,
+ 5.454215967759524e-10,
+ 5.459841467825299e-10,
+ 5.440591865912836e-10
+ ],
+ "bp_grad_norms_F": [
+ 7.836213626433164e-06,
+ 7.777267541086985e-08,
+ 3.578357521405451e-08,
+ 3.4922344127608085e-08,
+ 3.437838813624694e-08
+ ],
+ "gamma_dfa": 0.003963275059504667,
+ "acc_eval": 0.115234375,
+ "loss_eval": 5.663756370544434,
+ "epoch": 30
+ },
+ {
+ "hidden_norms": [
+ 43534.375,
+ 70424152.0,
+ 20966526.0,
+ 27032410.0,
+ 19126838.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.4974387830288833e-08,
+ 5.135932235056373e-10,
+ 5.051635221242634e-10,
+ 5.056056129326691e-10,
+ 5.03480979130444e-10
+ ],
+ "bp_grad_norms_F": [
+ 7.886262210377026e-06,
+ 8.217509162022907e-08,
+ 3.401517645329477e-08,
+ 3.309881080326704e-08,
+ 3.255568614690674e-08
+ ],
+ "gamma_dfa": 0.006612789164137212,
+ "acc_eval": 0.1162109375,
+ "loss_eval": 5.706640243530273,
+ "epoch": 31
+ },
+ {
+ "hidden_norms": [
+ 44720.9453125,
+ 76344112.0,
+ 22771320.0,
+ 29203482.0,
+ 20864830.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.4229628909088206e-08,
+ 4.800634334500842e-10,
+ 4.701866673784139e-10,
+ 4.693791466614528e-10,
+ 4.668225250803459e-10
+ ],
+ "bp_grad_norms_F": [
+ 7.781368367432151e-06,
+ 7.414258362814508e-08,
+ 3.207212984079888e-08,
+ 3.1193785332561674e-08,
+ 3.0666903683140845e-08
+ ],
+ "gamma_dfa": 0.008515000299667008,
+ "acc_eval": 0.1162109375,
+ "loss_eval": 5.720961093902588,
+ "epoch": 32
+ },
+ {
+ "hidden_norms": [
+ 45875.8046875,
+ 82383720.0,
+ 24638848.0,
+ 31558806.0,
+ 22520176.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.390868030805905e-08,
+ 4.458567681719927e-10,
+ 4.385545260277013e-10,
+ 4.3886239087242984e-10,
+ 4.3707018559935307e-10
+ ],
+ "bp_grad_norms_F": [
+ 7.905513484729454e-06,
+ 7.275451707755565e-08,
+ 3.047262708832932e-08,
+ 2.960254974482268e-08,
+ 2.9090815090171418e-08
+ ],
+ "gamma_dfa": 0.01025733859205502,
+ "acc_eval": 0.1142578125,
+ "loss_eval": 5.7539262771606445,
+ "epoch": 33
+ },
+ {
+ "hidden_norms": [
+ 47003.1328125,
+ 89200944.0,
+ 26495704.0,
+ 33934388.0,
+ 24310298.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.3027241386207606e-08,
+ 4.1779490977944533e-10,
+ 4.123023311652929e-10,
+ 4.1292305685836084e-10,
+ 4.1066355871421933e-10
+ ],
+ "bp_grad_norms_F": [
+ 8.257205990958028e-06,
+ 7.720848316239426e-08,
+ 2.8986395506080953e-08,
+ 2.8067159263400754e-08,
+ 2.7556911419424068e-08
+ ],
+ "gamma_dfa": 0.012584943462570664,
+ "acc_eval": 0.11328125,
+ "loss_eval": 5.796531677246094,
+ "epoch": 34
+ },
+ {
+ "hidden_norms": [
+ 48090.08984375,
+ 95793720.0,
+ 28464040.0,
+ 36264388.0,
+ 26293806.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.217191490989535e-08,
+ 3.894292666117849e-10,
+ 3.8594555329396485e-10,
+ 3.8626962739485293e-10,
+ 3.8498901289152343e-10
+ ],
+ "bp_grad_norms_F": [
+ 8.125247404677793e-06,
+ 7.330743301281473e-08,
+ 2.7641750222073824e-08,
+ 2.6769551908500944e-08,
+ 2.6268399011541987e-08
+ ],
+ "gamma_dfa": 0.015375214185041841,
+ "acc_eval": 0.1181640625,
+ "loss_eval": 5.804059982299805,
+ "epoch": 35
+ },
+ {
+ "hidden_norms": [
+ 49138.04296875,
+ 102879768.0,
+ 30344818.0,
+ 38660032.0,
+ 28092164.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.081565580487222e-08,
+ 3.644603230323895e-10,
+ 3.576897944501667e-10,
+ 3.568374207230107e-10,
+ 3.5416741761551407e-10
+ ],
+ "bp_grad_norms_F": [
+ 8.033912308746949e-06,
+ 7.398465839969504e-08,
+ 2.61401691403762e-08,
+ 2.530256892896432e-08,
+ 2.481599992165684e-08
+ ],
+ "gamma_dfa": 0.01655786213814281,
+ "acc_eval": 0.1162109375,
+ "loss_eval": 5.779043197631836,
+ "epoch": 36
+ },
+ {
+ "hidden_norms": [
+ 50205.23828125,
+ 109785504.0,
+ 32500932.0,
+ 40993492.0,
+ 30124936.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.087046351879508e-08,
+ 3.450236207847013e-10,
+ 3.3626229578587186e-10,
+ 3.3652211572920976e-10,
+ 3.357169819917516e-10
+ ],
+ "bp_grad_norms_F": [
+ 8.317977517435793e-06,
+ 6.930265783466893e-08,
+ 2.4881831706125013e-08,
+ 2.413608513052168e-08,
+ 2.36848762824593e-08
+ ],
+ "gamma_dfa": 0.01839016478334088,
+ "acc_eval": 0.1162109375,
+ "loss_eval": 5.815016746520996,
+ "epoch": 37
+ },
+ {
+ "hidden_norms": [
+ 51234.28125,
+ 116882832.0,
+ 34622344.0,
+ 43672052.0,
+ 32010828.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.011625793192252e-08,
+ 3.242739410325157e-10,
+ 3.1924299315200244e-10,
+ 3.1952127055312474e-10,
+ 3.181244712102682e-10
+ ],
+ "bp_grad_norms_F": [
+ 8.41012479213532e-06,
+ 7.246134714478103e-08,
+ 2.3759820777513596e-08,
+ 2.3049024022725462e-08,
+ 2.2618014128283903e-08
+ ],
+ "gamma_dfa": 0.019554060958398622,
+ "acc_eval": 0.1142578125,
+ "loss_eval": 5.829615116119385,
+ "epoch": 38
+ },
+ {
+ "hidden_norms": [
+ 52247.84375,
+ 124228312.0,
+ 36552372.0,
+ 46094684.0,
+ 33968656.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.024800321327348e-08,
+ 3.052329222708039e-10,
+ 3.0248267779420246e-10,
+ 3.022799510699059e-10,
+ 3.012357030485191e-10
+ ],
+ "bp_grad_norms_F": [
+ 7.947497579152696e-06,
+ 6.409469932577849e-08,
+ 2.2781945219207955e-08,
+ 2.209960925370069e-08,
+ 2.1678969730487552e-08
+ ],
+ "gamma_dfa": 0.022201491628948133,
+ "acc_eval": 0.1142578125,
+ "loss_eval": 5.778614044189453,
+ "epoch": 39
+ },
+ {
+ "hidden_norms": [
+ 53254.6640625,
+ 131932448.0,
+ 38786720.0,
+ 48743312.0,
+ 36026440.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.9781913823967443e-08,
+ 2.8849589384094543e-10,
+ 2.856256064998064e-10,
+ 2.8570401600092055e-10,
+ 2.844837143634038e-10
+ ],
+ "bp_grad_norms_F": [
+ 7.602731329825474e-06,
+ 6.632863147615353e-08,
+ 2.174161650714268e-08,
+ 2.107603691570148e-08,
+ 2.067151072537854e-08
+ ],
+ "gamma_dfa": 0.02325869471678743,
+ "acc_eval": 0.119140625,
+ "loss_eval": 5.798803329467773,
+ "epoch": 40
+ },
+ {
+ "hidden_norms": [
+ 54228.43359375,
+ 139390368.0,
+ 41071128.0,
+ 51348924.0,
+ 38183484.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.9682835744893055e-08,
+ 2.752674477246586e-10,
+ 2.7137878055860654e-10,
+ 2.716726288376492e-10,
+ 2.71538042051489e-10
+ ],
+ "bp_grad_norms_F": [
+ 7.424123396049254e-06,
+ 6.37739248077196e-08,
+ 2.099002749389456e-08,
+ 2.0336601735948534e-08,
+ 1.9947020035715468e-08
+ ],
+ "gamma_dfa": 0.024572641068516532,
+ "acc_eval": 0.119140625,
+ "loss_eval": 5.8440093994140625,
+ "epoch": 41
+ },
+ {
+ "hidden_norms": [
+ 55136.65625,
+ 147294320.0,
+ 43301224.0,
+ 54132172.0,
+ 40002916.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.0416718033166035e-08,
+ 2.623444794735974e-10,
+ 2.586598435438958e-10,
+ 2.587878522586351e-10,
+ 2.5755084176459775e-10
+ ],
+ "bp_grad_norms_F": [
+ 7.195616944954963e-06,
+ 5.543640568816954e-08,
+ 2.0126753597082825e-08,
+ 1.9496084746606357e-08,
+ 1.9124053451946565e-08
+ ],
+ "gamma_dfa": 0.026229552553559188,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.841940879821777,
+ "epoch": 42
+ },
+ {
+ "hidden_norms": [
+ 56057.48828125,
+ 155058496.0,
+ 45581136.0,
+ 56721952.0,
+ 42172944.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.9346542973485157e-08,
+ 2.4918875296542353e-10,
+ 2.4687848987348104e-10,
+ 2.470146309718757e-10,
+ 2.4589502656269246e-10
+ ],
+ "bp_grad_norms_F": [
+ 7.236937562993262e-06,
+ 4.957384902581907e-08,
+ 1.9364815528888357e-08,
+ 1.8756255215635065e-08,
+ 1.840485097659439e-08
+ ],
+ "gamma_dfa": 0.027025546743971063,
+ "acc_eval": 0.12109375,
+ "loss_eval": 5.79724645614624,
+ "epoch": 43
+ },
+ {
+ "hidden_norms": [
+ 56951.0546875,
+ 163286592.0,
+ 47715684.0,
+ 59381272.0,
+ 44453996.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.9479124918907473e-08,
+ 2.39016806347081e-10,
+ 2.3503707313743405e-10,
+ 2.33648544956111e-10,
+ 2.335798499064623e-10
+ ],
+ "bp_grad_norms_F": [
+ 7.3719347710721195e-06,
+ 5.521000900898798e-08,
+ 1.8681236113593513e-08,
+ 1.8048309513574168e-08,
+ 1.7682033615074033e-08
+ ],
+ "gamma_dfa": 0.02818910058704205,
+ "acc_eval": 0.1181640625,
+ "loss_eval": 5.779097080230713,
+ "epoch": 44
+ },
+ {
+ "hidden_norms": [
+ 57889.91015625,
+ 171572336.0,
+ 49962976.0,
+ 62099764.0,
+ 46403104.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 3.00151405951965e-08,
+ 2.294373580014053e-10,
+ 2.2455036441382248e-10,
+ 2.2508651886798958e-10,
+ 2.2402991961545382e-10
+ ],
+ "bp_grad_norms_F": [
+ 7.484923116862774e-06,
+ 5.0286196540128e-08,
+ 1.8117480848900414e-08,
+ 1.7483015923858147e-08,
+ 1.7114837547183015e-08
+ ],
+ "gamma_dfa": 0.029963871067593573,
+ "acc_eval": 0.1220703125,
+ "loss_eval": 5.8359575271606445,
+ "epoch": 45
+ },
+ {
+ "hidden_norms": [
+ 58731.421875,
+ 180056896.0,
+ 52137220.0,
+ 64717056.0,
+ 48647116.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.929727926925807e-08,
+ 2.1976008512947232e-10,
+ 2.159663420320257e-10,
+ 2.163312862180078e-10,
+ 2.1587925891353166e-10
+ ],
+ "bp_grad_norms_F": [
+ 7.681007446080912e-06,
+ 5.031828109736125e-08,
+ 1.7532725493651924e-08,
+ 1.6936246183263393e-08,
+ 1.659478598980968e-08
+ ],
+ "gamma_dfa": 0.03088831547938753,
+ "acc_eval": 0.1220703125,
+ "loss_eval": 5.786240100860596,
+ "epoch": 46
+ },
+ {
+ "hidden_norms": [
+ 59567.96875,
+ 187949040.0,
+ 54233264.0,
+ 67384816.0,
+ 50735356.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.870891080419824e-08,
+ 2.1069677946794485e-10,
+ 2.057157916235397e-10,
+ 2.0573617809382938e-10,
+ 2.051469827346608e-10
+ ],
+ "bp_grad_norms_F": [
+ 7.31024283595616e-06,
+ 4.820083177037304e-08,
+ 1.691453910268592e-08,
+ 1.6327849294839325e-08,
+ 1.5993451896179067e-08
+ ],
+ "gamma_dfa": 0.03192764806590276,
+ "acc_eval": 0.119140625,
+ "loss_eval": 5.707631587982178,
+ "epoch": 47
+ },
+ {
+ "hidden_norms": [
+ 60402.91796875,
+ 196285488.0,
+ 56440304.0,
+ 70085304.0,
+ 52941472.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8424729237030988e-08,
+ 2.033633955678127e-10,
+ 1.9889993818633656e-10,
+ 1.9922237470826332e-10,
+ 1.9868894030050654e-10
+ ],
+ "bp_grad_norms_F": [
+ 7.644263860129286e-06,
+ 4.852574519986774e-08,
+ 1.6383141954179337e-08,
+ 1.581754283108694e-08,
+ 1.548989914113008e-08
+ ],
+ "gamma_dfa": 0.033413520381145645,
+ "acc_eval": 0.1220703125,
+ "loss_eval": 5.704823017120361,
+ "epoch": 48
+ },
+ {
+ "hidden_norms": [
+ 61196.53515625,
+ 204414672.0,
+ 58842936.0,
+ 72912304.0,
+ 55068312.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.859230185947581e-08,
+ 1.9821513874695995e-10,
+ 1.9266777062654228e-10,
+ 1.9288849684162557e-10,
+ 1.9265547490654455e-10
+ ],
+ "bp_grad_norms_F": [
+ 7.879517397668678e-06,
+ 4.184777324667266e-08,
+ 1.6105406785982268e-08,
+ 1.550557016116727e-08,
+ 1.5175970702330233e-08
+ ],
+ "gamma_dfa": 0.034091180186806014,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.725862979888916,
+ "epoch": 49
+ },
+ {
+ "hidden_norms": [
+ 61914.14453125,
+ 212112608.0,
+ 61176224.0,
+ 75542504.0,
+ 57474876.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8669807861092522e-08,
+ 1.8937640344773854e-10,
+ 1.8534798695846177e-10,
+ 1.855716413867725e-10,
+ 1.852704378801917e-10
+ ],
+ "bp_grad_norms_F": [
+ 8.087788955890574e-06,
+ 4.2019124180114886e-08,
+ 1.5675068354426003e-08,
+ 1.5084628657291432e-08,
+ 1.4754542476680399e-08
+ ],
+ "gamma_dfa": 0.034466555964172585,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.820840358734131,
+ "epoch": 50
+ },
+ {
+ "hidden_norms": [
+ 62603.82421875,
+ 219453520.0,
+ 63197896.0,
+ 78229512.0,
+ 59644488.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.9210220020559063e-08,
+ 1.8399998191753753e-10,
+ 1.7923225403837506e-10,
+ 1.792543474765651e-10,
+ 1.785920855645884e-10
+ ],
+ "bp_grad_norms_F": [
+ 8.000329216883983e-06,
+ 4.107307916001446e-08,
+ 1.5165383615567407e-08,
+ 1.4602372644390016e-08,
+ 1.4272954373950597e-08
+ ],
+ "gamma_dfa": 0.035090556057184585,
+ "acc_eval": 0.12109375,
+ "loss_eval": 5.784890174865723,
+ "epoch": 51
+ },
+ {
+ "hidden_norms": [
+ 63365.48828125,
+ 227326704.0,
+ 65434160.0,
+ 80771200.0,
+ 61832948.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.9006677948473225e-08,
+ 1.7990454959093682e-10,
+ 1.7418685938075384e-10,
+ 1.7404670760168273e-10,
+ 1.7386522777051994e-10
+ ],
+ "bp_grad_norms_F": [
+ 8.04007868282497e-06,
+ 3.7175201583750095e-08,
+ 1.4899267597456856e-08,
+ 1.4325389763314433e-08,
+ 1.4007627946455159e-08
+ ],
+ "gamma_dfa": 0.035336821872988367,
+ "acc_eval": 0.12109375,
+ "loss_eval": 5.837845802307129,
+ "epoch": 52
+ },
+ {
+ "hidden_norms": [
+ 64122.77734375,
+ 235505280.0,
+ 67571832.0,
+ 83154408.0,
+ 63815212.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8882961800036355e-08,
+ 1.752420292211454e-10,
+ 1.6966168747689636e-10,
+ 1.6997936391760504e-10,
+ 1.6931231416883463e-10
+ ],
+ "bp_grad_norms_F": [
+ 7.732651283731684e-06,
+ 3.6642365586203596e-08,
+ 1.4491905453439813e-08,
+ 1.39411024946412e-08,
+ 1.3628048911584756e-08
+ ],
+ "gamma_dfa": 0.035559086416469654,
+ "acc_eval": 0.12109375,
+ "loss_eval": 5.813376426696777,
+ "epoch": 53
+ },
+ {
+ "hidden_norms": [
+ 64808.83203125,
+ 243721104.0,
+ 69766352.0,
+ 85766896.0,
+ 65980356.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8568885923618836e-08,
+ 1.6757319143412275e-10,
+ 1.6364704036320177e-10,
+ 1.6374304689925623e-10,
+ 1.6309503747535814e-10
+ ],
+ "bp_grad_norms_F": [
+ 7.502960215788335e-06,
+ 3.409801507814336e-08,
+ 1.4090593580817767e-08,
+ 1.3566921808205734e-08,
+ 1.3258036446472943e-08
+ ],
+ "gamma_dfa": 0.03633021883433685,
+ "acc_eval": 0.12109375,
+ "loss_eval": 5.748073101043701,
+ "epoch": 54
+ },
+ {
+ "hidden_norms": [
+ 65405.05078125,
+ 251434864.0,
+ 71765544.0,
+ 88222840.0,
+ 68040816.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8620791070466112e-08,
+ 1.6591342189009595e-10,
+ 1.611620559227589e-10,
+ 1.6055762275257734e-10,
+ 1.6039196359951546e-10
+ ],
+ "bp_grad_norms_F": [
+ 7.535967142757727e-06,
+ 3.341937571121889e-08,
+ 1.3891366279494832e-08,
+ 1.3369996665346662e-08,
+ 1.3064806125839823e-08
+ ],
+ "gamma_dfa": 0.037189101487456355,
+ "acc_eval": 0.1220703125,
+ "loss_eval": 5.769567966461182,
+ "epoch": 55
+ },
+ {
+ "hidden_norms": [
+ 65975.4296875,
+ 259364544.0,
+ 73868528.0,
+ 90804584.0,
+ 70148048.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8815582808761064e-08,
+ 1.6121720625150715e-10,
+ 1.5669744668489471e-10,
+ 1.568625784820199e-10,
+ 1.56342438994983e-10
+ ],
+ "bp_grad_norms_F": [
+ 7.530327820859384e-06,
+ 2.989135694519973e-08,
+ 1.3588412173248798e-08,
+ 1.308291253110383e-08,
+ 1.2791012693469384e-08
+ ],
+ "gamma_dfa": 0.037349180958699435,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.802239418029785,
+ "epoch": 56
+ },
+ {
+ "hidden_norms": [
+ 66535.546875,
+ 267235600.0,
+ 75921592.0,
+ 93206176.0,
+ 72113992.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8679746577608967e-08,
+ 1.5684299692342307e-10,
+ 1.530628263024525e-10,
+ 1.5322397517447683e-10,
+ 1.5271124642612932e-10
+ ],
+ "bp_grad_norms_F": [
+ 7.578343684144784e-06,
+ 2.8776035776445497e-08,
+ 1.3372356555407805e-08,
+ 1.2855631226216246e-08,
+ 1.2555028128247159e-08
+ ],
+ "gamma_dfa": 0.03780493976228172,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.850732803344727,
+ "epoch": 57
+ },
+ {
+ "hidden_norms": [
+ 67107.3125,
+ 274158304.0,
+ 77594136.0,
+ 95315520.0,
+ 73880288.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.9038343285492374e-08,
+ 1.5442255807407435e-10,
+ 1.5004263109741345e-10,
+ 1.5028490951696227e-10,
+ 1.4968959405337046e-10
+ ],
+ "bp_grad_norms_F": [
+ 7.555737283837516e-06,
+ 3.160908335075874e-08,
+ 1.3153115929753767e-08,
+ 1.2644002289619038e-08,
+ 1.2345022781801163e-08
+ ],
+ "gamma_dfa": 0.03859367490440491,
+ "acc_eval": 0.12109375,
+ "loss_eval": 5.83157205581665,
+ "epoch": 58
+ },
+ {
+ "hidden_norms": [
+ 67633.5703125,
+ 281346784.0,
+ 79395712.0,
+ 97521760.0,
+ 75716480.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8924493022941533e-08,
+ 1.5015137744267548e-10,
+ 1.4681762749990668e-10,
+ 1.469668414744163e-10,
+ 1.4640397227871915e-10
+ ],
+ "bp_grad_norms_F": [
+ 7.076869223965332e-06,
+ 3.03394500633658e-08,
+ 1.2924256331814377e-08,
+ 1.2421836004250508e-08,
+ 1.2128483106721433e-08
+ ],
+ "gamma_dfa": 0.03882967439312779,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.850924491882324,
+ "epoch": 59
+ },
+ {
+ "hidden_norms": [
+ 68127.2578125,
+ 288068512.0,
+ 81019736.0,
+ 99671120.0,
+ 77532304.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.900766915558961e-08,
+ 1.4631622302641034e-10,
+ 1.434221075236053e-10,
+ 1.4353812582967862e-10,
+ 1.428595158836643e-10
+ ],
+ "bp_grad_norms_F": [
+ 7.026245839369949e-06,
+ 3.037761331370348e-08,
+ 1.2673269544904997e-08,
+ 1.216829925709817e-08,
+ 1.1867063776094255e-08
+ ],
+ "gamma_dfa": 0.039077571367670316,
+ "acc_eval": 0.119140625,
+ "loss_eval": 5.843035697937012,
+ "epoch": 60
+ },
+ {
+ "hidden_norms": [
+ 68607.2890625,
+ 295220896.0,
+ 83038584.0,
+ 101714504.0,
+ 79438864.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.889952988027744e-08,
+ 1.4436063455747217e-10,
+ 1.410637162635453e-10,
+ 1.4117192137508283e-10,
+ 1.405108668306454e-10
+ ],
+ "bp_grad_norms_F": [
+ 6.8223289417801425e-06,
+ 2.9571756599011678e-08,
+ 1.237319757763089e-08,
+ 1.1902468344260342e-08,
+ 1.1619527562345411e-08
+ ],
+ "gamma_dfa": 0.0394883609842509,
+ "acc_eval": 0.119140625,
+ "loss_eval": 5.8329925537109375,
+ "epoch": 61
+ },
+ {
+ "hidden_norms": [
+ 69038.0703125,
+ 302174528.0,
+ 84795640.0,
+ 103621440.0,
+ 81134560.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8758867287592693e-08,
+ 1.4096314393530207e-10,
+ 1.380290187702471e-10,
+ 1.3813312993438132e-10,
+ 1.3748156779680443e-10
+ ],
+ "bp_grad_norms_F": [
+ 6.821945589763345e-06,
+ 2.9822015079616904e-08,
+ 1.2211526900784975e-08,
+ 1.1739270888710962e-08,
+ 1.1454956982959175e-08
+ ],
+ "gamma_dfa": 0.03950453496872797,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.862898826599121,
+ "epoch": 62
+ },
+ {
+ "hidden_norms": [
+ 69486.8671875,
+ 308758272.0,
+ 86559352.0,
+ 105690288.0,
+ 82830320.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.880081950706881e-08,
+ 1.381648684350978e-10,
+ 1.3562810596834396e-10,
+ 1.3572509782733277e-10,
+ 1.3508719143295878e-10
+ ],
+ "bp_grad_norms_F": [
+ 6.841747108410345e-06,
+ 3.094388745239485e-08,
+ 1.2084786504829026e-08,
+ 1.1623659368353856e-08,
+ 1.1344519990075241e-08
+ ],
+ "gamma_dfa": 0.03953796969653922,
+ "acc_eval": 0.1181640625,
+ "loss_eval": 5.886589527130127,
+ "epoch": 63
+ },
+ {
+ "hidden_norms": [
+ 69910.8515625,
+ 314931680.0,
+ 87987448.0,
+ 107443096.0,
+ 84227784.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8549854036441502e-08,
+ 1.3562842515746354e-10,
+ 1.3309958690754797e-10,
+ 1.3318925129457426e-10,
+ 1.32547681164219e-10
+ ],
+ "bp_grad_norms_F": [
+ 6.486627626145491e-06,
+ 3.26463052147119e-08,
+ 1.1863860116534397e-08,
+ 1.1407089495207856e-08,
+ 1.112555558790973e-08
+ ],
+ "gamma_dfa": 0.040154138219804736,
+ "acc_eval": 0.12109375,
+ "loss_eval": 5.841930389404297,
+ "epoch": 64
+ },
+ {
+ "hidden_norms": [
+ 70309.921875,
+ 320869664.0,
+ 89479256.0,
+ 109445496.0,
+ 85720672.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8119913508817262e-08,
+ 1.333930049751686e-10,
+ 1.311449004948173e-10,
+ 1.3122899988893266e-10,
+ 1.306234981290899e-10
+ ],
+ "bp_grad_norms_F": [
+ 6.1845657910453156e-06,
+ 3.1657755528158305e-08,
+ 1.1685694190077811e-08,
+ 1.1245842479468138e-08,
+ 1.0970555130995763e-08
+ ],
+ "gamma_dfa": 0.04057217295485316,
+ "acc_eval": 0.1181640625,
+ "loss_eval": 5.787144184112549,
+ "epoch": 65
+ },
+ {
+ "hidden_norms": [
+ 70717.453125,
+ 326733312.0,
+ 90990512.0,
+ 111385000.0,
+ 87145456.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.868856441295975e-08,
+ 1.3173599711091555e-10,
+ 1.290058199154842e-10,
+ 1.2908775437470155e-10,
+ 1.284874151519233e-10
+ ],
+ "bp_grad_norms_F": [
+ 6.056342499505263e-06,
+ 3.0463933597957293e-08,
+ 1.1565422397552538e-08,
+ 1.1122933685214775e-08,
+ 1.0847451825668486e-08
+ ],
+ "gamma_dfa": 0.040889364270697115,
+ "acc_eval": 0.119140625,
+ "loss_eval": 5.847908020019531,
+ "epoch": 66
+ },
+ {
+ "hidden_norms": [
+ 71084.71875,
+ 332338880.0,
+ 92455056.0,
+ 113026960.0,
+ 88541256.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.7999890406249506e-08,
+ 1.2975863439290691e-10,
+ 1.2677238425684578e-10,
+ 1.268507521245965e-10,
+ 1.2626712175833887e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.962062004982727e-06,
+ 2.9298774961716845e-08,
+ 1.14416129903816e-08,
+ 1.1010567568803253e-08,
+ 1.0736493472052189e-08
+ ],
+ "gamma_dfa": 0.040890415915782796,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.816187381744385,
+ "epoch": 67
+ },
+ {
+ "hidden_norms": [
+ 71436.84375,
+ 337832864.0,
+ 93879264.0,
+ 114665400.0,
+ 89926616.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.7683153547286565e-08,
+ 1.2797991832957933e-10,
+ 1.2511817970572991e-10,
+ 1.2518648617731998e-10,
+ 1.246210634686662e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.7771489991864655e-06,
+ 2.8285153774731953e-08,
+ 1.1245044895247247e-08,
+ 1.0827025498372223e-08,
+ 1.0562406060898866e-08
+ ],
+ "gamma_dfa": 0.04072478835223592,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.783463478088379,
+ "epoch": 68
+ },
+ {
+ "hidden_norms": [
+ 71769.5546875,
+ 342918304.0,
+ 95263584.0,
+ 116254568.0,
+ 91446520.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.775840357571724e-08,
+ 1.2646382552272684e-10,
+ 1.2334086529897093e-10,
+ 1.2337944554907665e-10,
+ 1.2288142725580542e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.667419827659614e-06,
+ 2.6888709925287912e-08,
+ 1.1133559851828068e-08,
+ 1.0716635578944533e-08,
+ 1.0457153365450722e-08
+ ],
+ "gamma_dfa": 0.040744281255683745,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.804702281951904,
+ "epoch": 69
+ },
+ {
+ "hidden_norms": [
+ 72080.078125,
+ 347772576.0,
+ 96482160.0,
+ 117685240.0,
+ 92638296.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8613341029881667e-08,
+ 1.2489977108121053e-10,
+ 1.2201274712797527e-10,
+ 1.220573642157774e-10,
+ 1.2148008987633574e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.642392352456227e-06,
+ 2.7134319680044428e-08,
+ 1.1023663759601732e-08,
+ 1.0605543110386861e-08,
+ 1.034114660569685e-08
+ ],
+ "gamma_dfa": 0.04128229359048419,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.805394172668457,
+ "epoch": 70
+ },
+ {
+ "hidden_norms": [
+ 72369.9921875,
+ 352377056.0,
+ 97691240.0,
+ 119151816.0,
+ 93761800.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8514245187238885e-08,
+ 1.2353804090814435e-10,
+ 1.2041069530344117e-10,
+ 1.2046977304613904e-10,
+ 1.1990240744719216e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.6565527302154806e-06,
+ 2.8678897479039733e-08,
+ 1.0947891482260275e-08,
+ 1.0533200978102286e-08,
+ 1.0270048811378274e-08
+ ],
+ "gamma_dfa": 0.04154024633317022,
+ "acc_eval": 0.119140625,
+ "loss_eval": 5.7999067306518555,
+ "epoch": 71
+ },
+ {
+ "hidden_norms": [
+ 72651.671875,
+ 356846048.0,
+ 98782024.0,
+ 120475080.0,
+ 94848032.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8226885717685946e-08,
+ 1.221410056428951e-10,
+ 1.1935313848354667e-10,
+ 1.193777160457543e-10,
+ 1.1868718508001308e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.597235940513201e-06,
+ 2.7164126947809564e-08,
+ 1.078571720825039e-08,
+ 1.038548358422986e-08,
+ 1.0128220928606879e-08
+ ],
+ "gamma_dfa": 0.041652156607597135,
+ "acc_eval": 0.119140625,
+ "loss_eval": 5.782708168029785,
+ "epoch": 72
+ },
+ {
+ "hidden_norms": [
+ 72915.9765625,
+ 361214752.0,
+ 99899064.0,
+ 121794264.0,
+ 95956256.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8156543763202535e-08,
+ 1.208296795951469e-10,
+ 1.1789087761560069e-10,
+ 1.1794409893184366e-10,
+ 1.1745009131924888e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.609212166746147e-06,
+ 2.767807316672588e-08,
+ 1.074356958952194e-08,
+ 1.0335670985739398e-08,
+ 1.007653693818611e-08
+ ],
+ "gamma_dfa": 0.041388919182281825,
+ "acc_eval": 0.1171875,
+ "loss_eval": 5.816926956176758,
+ "epoch": 73
+ },
+ {
+ "hidden_norms": [
+ 73160.8984375,
+ 365176576.0,
+ 100896352.0,
+ 123055840.0,
+ 96952304.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8052935974187676e-08,
+ 1.2002213112261018e-10,
+ 1.1708690961231838e-10,
+ 1.1711043246265262e-10,
+ 1.1664859356219637e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.528850579139544e-06,
+ 2.6534554109503006e-08,
+ 1.062906651583262e-08,
+ 1.0228863978056779e-08,
+ 9.976772297193293e-09
+ ],
+ "gamma_dfa": 0.04158145491965115,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.83203125,
+ "epoch": 74
+ },
+ {
+ "hidden_norms": [
+ 73388.9921875,
+ 369039872.0,
+ 101870976.0,
+ 124229664.0,
+ 97872176.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.811547972214612e-08,
+ 1.1867723470615488e-10,
+ 1.160038107239636e-10,
+ 1.1605346544873996e-10,
+ 1.1549688982981365e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.516625151358312e-06,
+ 2.6522718243882082e-08,
+ 1.0559543461852172e-08,
+ 1.0160210450749219e-08,
+ 9.905477327265544e-09
+ ],
+ "gamma_dfa": 0.04162494709453313,
+ "acc_eval": 0.1181640625,
+ "loss_eval": 5.825821876525879,
+ "epoch": 75
+ },
+ {
+ "hidden_norms": [
+ 73607.875,
+ 372436096.0,
+ 102747608.0,
+ 125356400.0,
+ 98735744.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8296041065800637e-08,
+ 1.1797214594100325e-10,
+ 1.1553259737784316e-10,
+ 1.1555564144449804e-10,
+ 1.148080935875484e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.4743991313443985e-06,
+ 2.5615005228019072e-08,
+ 1.0501097769122225e-08,
+ 1.0109610926178902e-08,
+ 9.859975946824306e-09
+ ],
+ "gamma_dfa": 0.04183118613582337,
+ "acc_eval": 0.1171875,
+ "loss_eval": 5.838399887084961,
+ "epoch": 76
+ },
+ {
+ "hidden_norms": [
+ 73806.21875,
+ 375504928.0,
+ 103559744.0,
+ 126298048.0,
+ 99592752.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8072012270285995e-08,
+ 1.16896228807839e-10,
+ 1.1455134063531602e-10,
+ 1.1459744958530749e-10,
+ 1.1405828365118609e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.35501703780028e-06,
+ 2.4375134799470288e-08,
+ 1.0415713624922773e-08,
+ 1.0033582853452572e-08,
+ 9.78571357279634e-09
+ ],
+ "gamma_dfa": 0.04175542373559438,
+ "acc_eval": 0.1181640625,
+ "loss_eval": 5.806183815002441,
+ "epoch": 77
+ },
+ {
+ "hidden_norms": [
+ 73982.1328125,
+ 378528768.0,
+ 104296416.0,
+ 127198352.0,
+ 100364600.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.802979714999765e-08,
+ 1.1618311174244056e-10,
+ 1.1359845702996196e-10,
+ 1.1364404556291063e-10,
+ 1.1310929276531212e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.359067017707275e-06,
+ 2.4563517442288685e-08,
+ 1.0355730495348325e-08,
+ 9.969440384338668e-09,
+ 9.722052496385913e-09
+ ],
+ "gamma_dfa": 0.041852476922940696,
+ "acc_eval": 0.1171875,
+ "loss_eval": 5.812763690948486,
+ "epoch": 78
+ },
+ {
+ "hidden_norms": [
+ 74143.3515625,
+ 381338400.0,
+ 104995776.0,
+ 128103664.0,
+ 101143248.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.803018084307496e-08,
+ 1.1504715929921971e-10,
+ 1.1309976566398205e-10,
+ 1.1314182923882754e-10,
+ 1.1259981835820554e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.305423201207304e-06,
+ 2.479905347740896e-08,
+ 1.0304642472647174e-08,
+ 9.922130672634921e-09,
+ 9.675199308389892e-09
+ ],
+ "gamma_dfa": 0.04194222000660375,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.790335655212402,
+ "epoch": 79
+ },
+ {
+ "hidden_norms": [
+ 74296.140625,
+ 383779872.0,
+ 105624680.0,
+ 128832568.0,
+ 101739872.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8086292402917934e-08,
+ 1.1405020677868194e-10,
+ 1.1259575216637785e-10,
+ 1.126363308179279e-10,
+ 1.121035694828798e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.285005499899853e-06,
+ 2.497140805246545e-08,
+ 1.0247953596831394e-08,
+ 9.865908978667903e-09,
+ 9.618767116137406e-09
+ ],
+ "gamma_dfa": 0.042092169023817405,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.78256893157959,
+ "epoch": 80
+ },
+ {
+ "hidden_norms": [
+ 74432.171875,
+ 386197824.0,
+ 106270032.0,
+ 129540264.0,
+ 102336408.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8129701235002358e-08,
+ 1.1369888364143321e-10,
+ 1.1195557675369727e-10,
+ 1.1199620397750465e-10,
+ 1.1146324141453334e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.2314603635750245e-06,
+ 2.418116018532146e-08,
+ 1.0211192780218425e-08,
+ 9.833079239740528e-09,
+ 9.588974059226985e-09
+ ],
+ "gamma_dfa": 0.0422149248624919,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.794626235961914,
+ "epoch": 81
+ },
+ {
+ "hidden_norms": [
+ 74557.1328125,
+ 388309664.0,
+ 106804240.0,
+ 130160096.0,
+ 102871888.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8006962082827158e-08,
+ 1.1333335658836319e-10,
+ 1.1151064099879093e-10,
+ 1.1155010942731636e-10,
+ 1.1102004732199688e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.231111572356895e-06,
+ 2.3960881279094792e-08,
+ 1.0185694065967255e-08,
+ 9.808036161018663e-09,
+ 9.561717639883227e-09
+ ],
+ "gamma_dfa": 0.042091405965038575,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.809223175048828,
+ "epoch": 82
+ },
+ {
+ "hidden_norms": [
+ 74670.6484375,
+ 390046144.0,
+ 107241528.0,
+ 130709072.0,
+ 103311792.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.80954566278524e-08,
+ 1.1238011909942003e-10,
+ 1.1107993691528151e-10,
+ 1.1111822573184327e-10,
+ 1.1058326476742764e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.195881840336369e-06,
+ 2.3944403793052516e-08,
+ 1.0142765738407888e-08,
+ 9.765725117460988e-09,
+ 9.519601995577887e-09
+ ],
+ "gamma_dfa": 0.042151874920818955,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.789272785186768,
+ "epoch": 83
+ },
+ {
+ "hidden_norms": [
+ 74771.78125,
+ 391740192.0,
+ 107714400.0,
+ 131203208.0,
+ 103730008.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8205375812717648e-08,
+ 1.1187289983283222e-10,
+ 1.1065893340544974e-10,
+ 1.106964936381516e-10,
+ 1.1018032319842774e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.16868840350071e-06,
+ 2.363896278723132e-08,
+ 1.0106472991822102e-08,
+ 9.731052408312735e-09,
+ 9.487330032698083e-09
+ ],
+ "gamma_dfa": 0.042111018407013034,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.778679847717285,
+ "epoch": 84
+ },
+ {
+ "hidden_norms": [
+ 74860.828125,
+ 393210464.0,
+ 108154200.0,
+ 131674336.0,
+ 104148832.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8289822040505896e-08,
+ 1.1139363043088935e-10,
+ 1.1020819673523974e-10,
+ 1.1024554186223057e-10,
+ 1.0972771996575759e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.15709007231635e-06,
+ 2.3636799184600932e-08,
+ 1.008097960664145e-08,
+ 9.703811088002112e-09,
+ 9.460902283819905e-09
+ ],
+ "gamma_dfa": 0.04205825741155422,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.787901878356934,
+ "epoch": 85
+ },
+ {
+ "hidden_norms": [
+ 74942.4453125,
+ 394569024.0,
+ 108544648.0,
+ 132082208.0,
+ 104483976.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8258760664812144e-08,
+ 1.1120066673031559e-10,
+ 1.0991466070642275e-10,
+ 1.0995130500512929e-10,
+ 1.0943435047039429e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.150290235178545e-06,
+ 2.3486826705720887e-08,
+ 1.0058711197302728e-08,
+ 9.684677060306512e-09,
+ 9.442134185633222e-09
+ ],
+ "gamma_dfa": 0.042173042878857814,
+ "acc_eval": 0.119140625,
+ "loss_eval": 5.789056301116943,
+ "epoch": 86
+ },
+ {
+ "hidden_norms": [
+ 75009.5703125,
+ 395748704.0,
+ 108812392.0,
+ 132420432.0,
+ 104755352.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.825358436098213e-08,
+ 1.1081059680950744e-10,
+ 1.0968410901757153e-10,
+ 1.0972003861020596e-10,
+ 1.0920536697156535e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.127661552251084e-06,
+ 2.3339271848499266e-08,
+ 1.002943950112467e-08,
+ 9.659677502327213e-09,
+ 9.417544966083824e-09
+ ],
+ "gamma_dfa": 0.04217808429893921,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.776318550109863,
+ "epoch": 87
+ },
+ {
+ "hidden_norms": [
+ 75066.4921875,
+ 396774016.0,
+ 109070288.0,
+ 132708432.0,
+ 105002960.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.824552858271545e-08,
+ 1.1050411280466577e-10,
+ 1.0939484040850544e-10,
+ 1.0943065897883741e-10,
+ 1.0891805513058017e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.1081005949527025e-06,
+ 2.2895029871961015e-08,
+ 1.0016128371148625e-08,
+ 9.644456788748812e-09,
+ 9.40405175953174e-09
+ ],
+ "gamma_dfa": 0.04221886159211863,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.782362937927246,
+ "epoch": 88
+ },
+ {
+ "hidden_norms": [
+ 75117.3984375,
+ 397629600.0,
+ 109292216.0,
+ 132974672.0,
+ 105237816.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8004475183251998e-08,
+ 1.1032175173397718e-10,
+ 1.092313739459172e-10,
+ 1.092670676161589e-10,
+ 1.0875097350426799e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.101501756143989e-06,
+ 2.2761090789913396e-08,
+ 9.993788907536327e-09,
+ 9.627679986579096e-09,
+ 9.387123967030675e-09
+ ],
+ "gamma_dfa": 0.04218835258870968,
+ "acc_eval": 0.119140625,
+ "loss_eval": 5.786335468292236,
+ "epoch": 89
+ },
+ {
+ "hidden_norms": [
+ 75159.3515625,
+ 398356864.0,
+ 109470496.0,
+ 133184088.0,
+ 105416456.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.806424248547046e-08,
+ 1.1018109341565108e-10,
+ 1.0910816694575942e-10,
+ 1.0914344428236689e-10,
+ 1.0862655219767703e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.100384441902861e-06,
+ 2.2749302885927136e-08,
+ 9.977647152936697e-09,
+ 9.612988627338837e-09,
+ 9.37321864569185e-09
+ ],
+ "gamma_dfa": 0.04216382784579764,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.787578582763672,
+ "epoch": 90
+ },
+ {
+ "hidden_norms": [
+ 75196.53125,
+ 398958272.0,
+ 109628448.0,
+ 133366784.0,
+ 105577720.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8093891657476888e-08,
+ 1.0998777583148822e-10,
+ 1.0895360308404989e-10,
+ 1.0898840857587189e-10,
+ 1.084766096393075e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.099497684568632e-06,
+ 2.276015820257271e-08,
+ 9.975322790012342e-09,
+ 9.609090412254773e-09,
+ 9.369199638342707e-09
+ ],
+ "gamma_dfa": 0.0421357955710846,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.788777828216553,
+ "epoch": 91
+ },
+ {
+ "hidden_norms": [
+ 75225.75,
+ 399436800.0,
+ 109753008.0,
+ 133503392.0,
+ 105709880.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.7958691362073296e-08,
+ 1.098953844591577e-10,
+ 1.0885654877501594e-10,
+ 1.0889117385559643e-10,
+ 1.0837702263399862e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.099244390294189e-06,
+ 2.2776946551061883e-08,
+ 9.965032354841696e-09,
+ 9.599201433729831e-09,
+ 9.359114372387012e-09
+ ],
+ "gamma_dfa": 0.04213721772248391,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.787423133850098,
+ "epoch": 92
+ },
+ {
+ "hidden_norms": [
+ 75247.3671875,
+ 399815808.0,
+ 109848648.0,
+ 133619136.0,
+ 105800656.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.813173693994031e-08,
+ 1.0978142006567992e-10,
+ 1.0878217771015386e-10,
+ 1.0881660156281114e-10,
+ 1.0830317198617934e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.097780103824334e-06,
+ 2.278613919770578e-08,
+ 9.958157853873217e-09,
+ 9.592699079519207e-09,
+ 9.35304633742362e-09
+ ],
+ "gamma_dfa": 0.042139860528550344,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.78769588470459,
+ "epoch": 93
+ },
+ {
+ "hidden_norms": [
+ 75264.9921875,
+ 400121888.0,
+ 109927240.0,
+ 133712496.0,
+ 105877656.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8044253141956688e-08,
+ 1.0970312158686824e-10,
+ 1.0871267774881233e-10,
+ 1.0874703221253057e-10,
+ 1.0823512919255762e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.0932972044392955e-06,
+ 2.2714601755069452e-08,
+ 9.954023383329513e-09,
+ 9.589306237955952e-09,
+ 9.350141105812781e-09
+ ],
+ "gamma_dfa": 0.04215262952493504,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.787428379058838,
+ "epoch": 94
+ },
+ {
+ "hidden_norms": [
+ 75277.7421875,
+ 400340192.0,
+ 109981456.0,
+ 133783368.0,
+ 105935424.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8024039977481152e-08,
+ 1.0965856001021734e-10,
+ 1.0866757493843693e-10,
+ 1.0870188776879175e-10,
+ 1.0818954065960895e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.093556865176652e-06,
+ 2.269677956689975e-08,
+ 9.948016632677081e-09,
+ 9.583557947223653e-09,
+ 9.344440776715146e-09
+ ],
+ "gamma_dfa": 0.04216338234982686,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.787187099456787,
+ "epoch": 95
+ },
+ {
+ "hidden_norms": [
+ 75285.703125,
+ 400477120.0,
+ 110015360.0,
+ 133823992.0,
+ 105968568.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.7959323745108122e-08,
+ 1.096286394997037e-10,
+ 1.0863549643191917e-10,
+ 1.0866971211775933e-10,
+ 1.0815728868074359e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.094433618069161e-06,
+ 2.2707608238192734e-08,
+ 9.945019030510593e-09,
+ 9.580433335543148e-09,
+ 9.341129647566504e-09
+ ],
+ "gamma_dfa": 0.04217854590388015,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.785982608795166,
+ "epoch": 96
+ },
+ {
+ "hidden_norms": [
+ 75289.953125,
+ 400549248.0,
+ 110035048.0,
+ 133846800.0,
+ 105987544.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.7980583183762064e-08,
+ 1.0960886365207756e-10,
+ 1.0861883614765588e-10,
+ 1.0865301020013263e-10,
+ 1.0814063533537421e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.094651896797586e-06,
+ 2.2699452983943047e-08,
+ 9.943881273954958e-09,
+ 9.579290249916994e-09,
+ 9.340033635396594e-09
+ ],
+ "gamma_dfa": 0.04217962139227893,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.78618860244751,
+ "epoch": 97
+ },
+ {
+ "hidden_norms": [
+ 75293.8203125,
+ 400615232.0,
+ 110052384.0,
+ 133866488.0,
+ 106003008.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.800710419137431e-08,
+ 1.0960048146824164e-10,
+ 1.0860957966318807e-10,
+ 1.0864373983787701e-10,
+ 1.0813145517873934e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.094465905131074e-06,
+ 2.2699449431229368e-08,
+ 9.942947798435853e-09,
+ 9.578524640119213e-09,
+ 9.339320428125575e-09
+ ],
+ "gamma_dfa": 0.04218013053832692,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.786159515380859,
+ "epoch": 98
+ },
+ {
+ "hidden_norms": [
+ 75295.7109375,
+ 400646048.0,
+ 110060248.0,
+ 133875320.0,
+ 106009248.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8009079500179723e-08,
+ 1.0959386176345731e-10,
+ 1.0860166238524371e-10,
+ 1.0863580868214484e-10,
+ 1.0812370443424868e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.094112111692084e-06,
+ 2.26939054215336e-08,
+ 9.942577428034838e-09,
+ 9.578169368751333e-09,
+ 9.33901134203552e-09
+ ],
+ "gamma_dfa": 0.042181356100627454,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.786299705505371,
+ "epoch": 99
+ },
+ {
+ "hidden_norms": [
+ 75295.7421875,
+ 400646208.0,
+ 110060504.0,
+ 133875616.0,
+ 106009536.0
+ ],
+ "bp_grad_norms_per_sample_med": [
+ 2.8005374019812734e-08,
+ 1.0959412544142566e-10,
+ 1.0860218280228651e-10,
+ 1.0863632909918763e-10,
+ 1.0812418321792805e-10
+ ],
+ "bp_grad_norms_F": [
+ 5.0940698201884516e-06,
+ 2.269248788877576e-08,
+ 9.942581868926936e-09,
+ 9.57817469782185e-09,
+ 9.339016671106037e-09
+ ],
+ "gamma_dfa": 0.042181452532531694,
+ "acc_eval": 0.1201171875,
+ "loss_eval": 5.786318302154541,
+ "epoch": 100
+ }
+ ]
+} \ No newline at end of file