epoch 1
train loss: 0.04859849993016349
valid loss scaled: 0.049774259863821814
valid loss: 8.239480139682152
valid loss clipped: 8.239480139682152
train loss: 0.048566044032315264
valid loss scaled: 0.05388942821210208
valid loss: 8.920693268651688
valid loss clipped: 8.920693268651688
train loss: 0.048645840325942186
valid loss scaled: 0.05395363235295337
valid loss: 8.931319168623325
valid loss clipped: 8.931319168623325
train loss: 0.04814379534040526
valid loss scaled: 0.0505393534476585
valid loss: 8.366129820820378
valid loss clipped: 8.366112817712773
train loss: 0.04762666338418628
valid loss scaled: 0.0559599088881008
valid loss: 9.263433639650358
valid loss clipped: 9.263433639650358
train loss: 0.047656307189366594
valid loss scaled: 0.05023614907590337
valid loss: 8.315944918367261
valid loss clipped: 8.315944918367261
train loss: 0.04783966918808628
valid loss scaled: 0.05002744041130589
valid loss: 8.281391949540064
valid loss clipped: 8.281391949540064
train loss: 0.0470857543381118
valid loss scaled: 0.05007423239511511
valid loss: 8.289137880244128
valid loss clipped: 8.289137880244128
train loss: 0.04748013890490943
valid loss scaled: 0.048425306807186
valid loss: 8.01617955284889
valid loss clipped: 8.01617955284889
train loss: 0.04756126793491012
valid loss scaled: 0.05838172127632288
valid loss: 9.664334319586205
valid loss clipped: 9.664334319586205
train loss: 0.04653691036105432
valid loss scaled: 0.04918750869152843
valid loss: 8.14234734821747
valid loss clipped: 8.14225860100907
train loss: 0.047854342213726093
valid loss scaled: 0.04825916678037553
valid loss: 7.988678617267059
valid loss clipped: 7.988678617267059
train loss: 0.04773736108071371
valid loss scaled: 0.04693122838927694
valid loss: 7.768854512974581
valid loss clipped: 7.768764906393828
train loss: 0.04762865603151333
valid loss scaled: 0.05211583033031525
valid loss: 8.627096447506986
valid loss clipped: 8.627096447506986
train loss: 0.04704306549118133
valid loss scaled: 0.048404824273360625
valid loss: 8.012792317972876
valid loss clipped: 8.012792317972876
train loss: 0.047052753271308353
valid loss scaled: 0.047940905819710194
valid loss: 7.935991222098737
valid loss clipped: 7.935991222098737
train loss: 0.047227875012458086
valid loss scaled: 0.04819175795767938
valid loss: 7.977517309701107
valid loss clipped: 7.977427346370881
train loss: 0.04710633027268819
valid loss scaled: 0.06383742791518349
valid loss: 10.567450933423165
valid loss clipped: 10.567450933423165
train loss: 0.04722847116132579
valid loss scaled: 0.050155074146505475
valid loss: 8.302519880598078
valid loss clipped: 8.302519880598078
train loss: 0.04704662696396088
valid loss scaled: 0.049731758847849555
valid loss: 8.232442667326458
valid loss clipped: 8.231381949146131
train loss: 0.04640111281869779
valid loss scaled: 0.04439513521075078
valid loss: 7.349034015457206
valid loss clipped: 7.349034015457206
train loss: 0.04608356887790085
valid loss scaled: 0.04921432936453068
valid loss: 8.146788631177284
valid loss clipped: 8.146788631177284
train loss: 0.04667254772723141
valid loss scaled: 0.04490475325651902
valid loss: 7.433397045419523
valid loss clipped: 7.433397045419523
train loss: 0.04732138091412676
valid loss scaled: 0.050358595447874285
valid loss: 8.336210587795785
valid loss clipped: 8.336210587795785
train loss: 0.046673220858461255
valid loss scaled: 0.05013948923902952
valid loss: 8.29994005436984
valid loss clipped: 8.29994005436984
train loss: 0.04553470963703993
valid loss scaled: 0.04682243262330682
valid loss: 7.750845583038412
valid loss clipped: 7.750845583038412
train loss: 0.046360710414935315
valid loss scaled: 0.045567268129456086
valid loss: 7.543068199989672
valid loss clipped: 7.543068199989672
train loss: 0.04552500833500083
valid loss scaled: 0.04640085875333069
valid loss: 7.681058866001286
valid loss clipped: 7.681058866001286
train loss: 0.0452476728574542
valid loss scaled: 0.046615049866360565
valid loss: 7.716513190971443
valid loss clipped: 7.716513190971443
train loss: 0.045669575255858894
valid loss scaled: 0.04750326795304539
valid loss: 7.8635460404556
valid loss clipped: 7.8635460404556
train loss: 0.04647987461932483
valid loss scaled: 0.04860825732568691
valid loss: 8.046460836562115
valid loss clipped: 8.046460836562115
train loss: 0.046793957499830896
valid loss scaled: 0.04575770256354883
valid loss: 7.574592460104707
valid loss clipped: 7.574592460104707
train loss: 0.04585571441358016
valid loss scaled: 0.04501184464469181
valid loss: 7.451126169626989
valid loss clipped: 7.451126169626989
train loss: 0.04610028065682122
valid loss scaled: 0.046051780212043245
valid loss: 7.623273458385178
valid loss clipped: 7.623083864668157
train loss: 0.046177845466100334
valid loss scaled: 0.06025220972281953
valid loss: 9.973968222380467
valid loss clipped: 9.973968222380467
train loss: 0.044866837962671394
valid loss scaled: 0.04543754842407298
valid loss: 7.5215926794190535
valid loss clipped: 7.5215926794190535
train loss: 0.044414652746250796
valid loss scaled: 0.044867695763598454
valid loss: 7.4272631776866636
valid loss clipped: 7.427128339884669
train loss: 0.04534780122814873
valid loss scaled: 0.045269020539499126
valid loss: 7.493695395829885
valid loss clipped: 7.493620279412681
train loss: 0.04439006437582844
valid loss scaled: 0.04507342677914177
valid loss: 7.461320119658475
valid loss clipped: 7.461320119658475
epoch 1 done, EVALUATION ON FULL DEV:
valid loss scaled: 0.04480323708981578
valid loss: 7.416593515309324
valid loss clipped: 7.416387303935181
evaluation done
epoch 2
train loss: 0.04156288899572168
valid loss scaled: 0.04604067304159161
valid loss: 7.621432637256123
valid loss clipped: 7.621432637256123
train loss: 0.04100707919178113
valid loss scaled: 0.052927161282336356
valid loss: 8.761404249024789
valid loss clipped: 8.761404249024789
train loss: 0.04102466219659839
valid loss scaled: 0.0449000373685904
valid loss: 7.432613525687441
valid loss clipped: 7.432613525687441
train loss: 0.04056007982480108
valid loss scaled: 0.04401167129402493
valid loss: 7.285561051000317
valid loss clipped: 7.285561051000317
train loss: 0.04043435892664155
valid loss scaled: 0.0482961730545656
valid loss: 7.9948042408749025
valid loss clipped: 7.9948042408749025
train loss: 0.040083140595004084
valid loss scaled: 0.05474324524206743
valid loss: 9.062036144331996
valid loss clipped: 9.061846768141493
train loss: 0.040820699932255504
valid loss scaled: 0.0489683078886907
valid loss: 8.106066491790516
valid loss clipped: 8.106061770281157
train loss: 0.04011373971574097
valid loss scaled: 0.04683979676309513
valid loss: 7.75372067780672
valid loss clipped: 7.75372067780672
train loss: 0.040472179534386435
valid loss scaled: 0.04310859235241129
valid loss: 7.13606575431109
valid loss clipped: 7.13606575431109
train loss: 0.040441015502005524
valid loss scaled: 0.05050364392711698
valid loss: 8.360220506489952
valid loss clipped: 8.360220506489952
train loss: 0.03942231891194775
valid loss scaled: 0.0488088141837757
valid loss: 8.079661712550005
valid loss clipped: 8.079636855330948
train loss: 0.04015028625372404
valid loss scaled: 0.04516411367366714
valid loss: 7.476330067311157
valid loss clipped: 7.476182879344117
train loss: 0.04084930318361188
valid loss scaled: 0.04687766826105318
valid loss: 7.7599889055078215
valid loss clipped: 7.759971606690415
train loss: 0.04066393936852439
valid loss scaled: 0.04189752550714794
valid loss: 6.935592127831488
valid loss clipped: 6.935592127831488
train loss: 0.03997029890541135
valid loss scaled: 0.044517524088250444
valid loss: 7.369297241211859
valid loss clipped: 7.369000766181205
train loss: 0.03993799675444034
valid loss scaled: 0.04544862752523738
valid loss: 7.523431609784254
valid loss clipped: 7.523394388992367
train loss: 0.040077832674748126
valid loss scaled: 0.04967513279215436
valid loss: 8.223067645108816
valid loss clipped: 8.223067645108816
train loss: 0.03990507605068726
valid loss scaled: 0.05784423713065072
valid loss: 9.575362301216071
valid loss clipped: 9.575185345956458
train loss: 0.040589384846144605
valid loss scaled: 0.0496796067080015
valid loss: 8.223812362251003
valid loss clipped: 8.223812362251003
train loss: 0.04017873370107866
valid loss scaled: 0.046199073425154856
valid loss: 7.647655631709408
valid loss clipped: 7.647218458815767
train loss: 0.04044891785946012
valid loss scaled: 0.04150905843239381
valid loss: 6.871281648789468
valid loss clipped: 6.871281648789468
train loss: 0.03947431928291259
valid loss scaled: 0.05190902248718164
valid loss: 8.592867224099864
valid loss clipped: 8.592867224099864
train loss: 0.03990533801732083
valid loss scaled: 0.04276042828750439
valid loss: 7.078432226253377
valid loss clipped: 7.078432226253377
train loss: 0.040009430050001356
valid loss scaled: 0.04785424872916424
valid loss: 7.92164959531741
valid loss clipped: 7.92164959531741
train loss: 0.03973275025215787
valid loss scaled: 0.0463738630632847
valid loss: 7.676589189399943
valid loss clipped: 7.67643024694617
train loss: 0.038768686683078965
valid loss scaled: 0.04392041103001954
valid loss: 7.270451050508478
valid loss clipped: 7.270451050508478
train loss: 0.039348882183647814
valid loss scaled: 0.04542625701933914
valid loss: 7.519721720555645
valid loss clipped: 7.519588671593333
train loss: 0.03904904227063566
valid loss scaled: 0.04735228840997471
valid loss: 7.838551573747202
valid loss clipped: 7.8383747004264395
train loss: 0.03885343865845939
valid loss scaled: 0.04507136710901525
valid loss: 7.460976738161493
valid loss clipped: 7.460948397077222
train loss: 0.039375428725538786
valid loss scaled: 0.049669896283577444
valid loss: 8.222201176726562
valid loss clipped: 8.222057108632997
train loss: 0.03995204566972607
valid loss scaled: 0.048435733190966335
valid loss: 8.017900004946629
valid loss clipped: 8.017900004946629
train loss: 0.0398491290780488
valid loss scaled: 0.04720521208137839
valid loss: 7.814206857210747
valid loss clipped: 7.814180725425085
train loss: 0.039705720615770324
valid loss scaled: 0.04476360371587965
valid loss: 7.410029695071655
valid loss clipped: 7.410029695071655
train loss: 0.03922532949685088
valid loss scaled: 0.043110160856005976
valid loss: 7.136322356669672
valid loss clipped: 7.136322356669672
train loss: 0.03944114218638488
valid loss scaled: 0.05162181950303874
valid loss: 8.54532049172949
valid loss clipped: 8.545318087656034
train loss: 0.038248045063862056
valid loss scaled: 0.045729077826710006
valid loss: 7.5698496787836405
valid loss clipped: 7.5698496787836405
train loss: 0.038367600972177264
valid loss scaled: 0.045730969331967096
valid loss: 7.570170291346379
valid loss clipped: 7.570170291346379
train loss: 0.03882607528483637
valid loss scaled: 0.04907676118577929
valid loss: 8.124020217147429
valid loss clipped: 8.124020217147429
train loss: 0.03804710767993723
valid loss scaled: 0.043570846150311024
valid loss: 7.2125900868664665
valid loss clipped: 7.2125900868664665
epoch 2 done, EVALUATION ON FULL DEV:
valid loss scaled: 0.04305710652707711
valid loss: 7.127543370821652
valid loss clipped: 7.127414565514862
evaluation done
epoch 3
train loss: 0.03567875976729889
valid loss scaled: 0.0435919204035305
valid loss: 7.216075509096017
valid loss clipped: 7.216075509096017
train loss: 0.03612129883597582
valid loss scaled: 0.05422117357473846
valid loss: 8.975609535535733
valid loss clipped: 8.975609535535733
train loss: 0.03610084946473125
valid loss scaled: 0.04305326329624392
valid loss: 7.126904973342827
valid loss clipped: 7.126904973342827
train loss: 0.0354239810019526
valid loss scaled: 0.041925094025886615
valid loss: 6.940151276296263
valid loss clipped: 6.940151276296263
train loss: 0.0352127675835109
valid loss scaled: 0.0480450296753359
valid loss: 7.9532327819071025
valid loss clipped: 7.9532327819071025
train loss: 0.03504781431121313
valid loss scaled: 0.05108940815500588
valid loss: 8.457186233906421
valid loss clipped: 8.45717667639358
train loss: 0.034990980345574955
valid loss scaled: 0.05037606369447504
valid loss: 8.339104845822392
valid loss clipped: 8.339104845822392
train loss: 0.034820230333430666
valid loss scaled: 0.04548966136491771
valid loss: 7.530221839492527
valid loss clipped: 7.530221839492527
train loss: 0.034814962549952325
valid loss scaled: 0.044887413795706835
valid loss: 7.430525821074159
valid loss clipped: 7.430525821074159
train loss: 0.035407486527384235
valid loss scaled: 0.04845567947436388
valid loss: 8.021209744958515
valid loss clipped: 8.021177381420596