diff --git a/configs/db_new.yaml b/configs/db_new.yaml
new file mode 100644
index 0000000000000000000000000000000000000000..8b4c9a08b6c988f47fd0a49addd9e99579c4151d
--- /dev/null
+++ b/configs/db_new.yaml
@@ -0,0 +1,70 @@
+main_config:
+ max_missing : 120
+ db_params:
+ user: c##mspils
+ password: cobalt_deviancy
+ dsn: localhost/XE
+ # Auch als Liste möglich
+ zrxp_folder : ../data/dwd_ens_zrpx/
+ sensor_folder : ../data/db_in/
+ zrxp_out_folder : ../data/zrxp_out/
+ #if True tries to put all files in folder in the external forecast table
+ load_sensor : False
+ load_zrxp : False
+ ensemble : True
+ single : False
+ dummy : True
+ #start: 2019-12-01 02:00
+ #start: 2023-04-19 03:00
+ #start: 2023-11-19 03:00
+ start: 2024-09-13 09:00
+ #end: !!str 2021-02-06
+ # if range is true will try all values from start to end for predictions
+ #range: !!bool True
+ range: !!bool False
+ zrxp: !!bool True
+gauge_configs:
+ - gauge: hollingstedt
+ model_folder: ../models_torch/hollingstedt_version_14
+ columns:
+ - 4466,SHum,vwsl
+ - 4466,SHum,bfwls
+ - 4466,AT,h.Cmd-2
+ - 114069,S,15m.Cmd
+ - 111111,S,5m.Cmd
+ - 9520081,S_Tide,1m.Cmd
+ - 9530010,S_Tide,1m.Cmd
+ - 112211,Precip,h.Cmd
+ - 112211,S,5m.Cmd
+ external_fcst:
+ - 9530010,S_Tide,1m.Cmd
+ - 112211,Precip,h.Cmd
+ - gauge: 114547,S,60m.Cmd
+ model_folder: ../models_torch/tarp_version_49/
+ columns:
+ - 4466,SHum,vwsl
+ - 4466,SHum,bfwls
+ - 4466,AT,h.Cmd
+ - 114435,S,60m.Cmd
+ - 114050,S,60m.Cmd
+ - 114547,S,60m.Cmd
+ - 114547,Precip,h.Cmd
+ external_fcst:
+ - 114547,Precip,h.Cmd
+ #114547,Precip,h.Cmd : Tarp
+ - gauge: 114069,Precip,h.Cmd
+ model_folder: ../models_torch/version_74_treia/
+ columns:
+ - 4466,SHum,vwsl
+ - 4466,SHum,bfwls
+ - 4466,AT,h.Cmd
+ - 114435,S,60m.Cmd
+ - 114050,S,60m.Cmd
+ - 114547,S,60m.Cmd
+ - 114224,S,60m.Cmd
+ - 114061,S,60m.Cmd
+ - 114069,Precip,h.Cmd
+ - 114069,S,15m.Cmd
+ external_fcst:
+ - 114069,Precip,h.Cmd
+ #114069,Precip,h.Cmd : Treia
\ No newline at end of file
diff --git a/configs/icon_exp_ensemble.yaml b/configs/icon_exp_ensemble.yaml
new file mode 100644
index 0000000000000000000000000000000000000000..27338164b0b4cc002aa2f22c9c15bec9c040bb38
--- /dev/null
+++ b/configs/icon_exp_ensemble.yaml
@@ -0,0 +1,9 @@
+shape_folder: ../data/meta_other/Gebiete/
+index_folder: ../data/meta_other/ensemble_index/
+source_folder: ../data/ICON_ENSEMBLE/
+target_folder: ../data/ICON_ENSEMBLE_EXPORT/
+mode: ensemble
+force_index_calculation: False
+log_file: ../data/ICON_ENSEMBLE_EXPORT/icon_export.log
+tidy_up: True
+debug: False
\ No newline at end of file
diff --git a/configs/icon_exp_single.yaml b/configs/icon_exp_single.yaml
new file mode 100644
index 0000000000000000000000000000000000000000..5441644f63829b0105a9bc68f0ef12f64af10655
--- /dev/null
+++ b/configs/icon_exp_single.yaml
@@ -0,0 +1,9 @@
+shape_folder: ../data/meta_other/Gebiete/
+index_folder: ../data/meta_other/single_index/
+source_folder: ../data/ICON_D2/
+target_folder: ../data/ICON_D2_EXPORT/
+mode: single
+force_index_calculation: False
+log_file: ../data/ICON_ENSEMBLE_EXPORT/icon_export.log
+tidy_up: True
+debug: False
\ No newline at end of file
diff --git a/notebooks/data/dk.json b/notebooks/data/dk.json
new file mode 100644
index 0000000000000000000000000000000000000000..3d80c640b643d3847a108707ec8d9c8eb905bc5b
--- /dev/null
+++ b/notebooks/data/dk.json
@@ -0,0 +1,10071 @@
+{
+ "type": "FeatureCollection",
+ "features": [
+ {
+ "geometry": {
+ "type": "MultiPolygon",
+ "coordinates": [
+ [
+ [
+ [
+ 8.17560707357738,
+ 55.80147645034364
+ ],
+ [
+ 8.175811394924775,
+ 55.80147796784839
+ ],
+ [
+ 8.368306106083875,
+ 55.80253734910015
+ ],
+ [
+ 8.39920861463416,
+ 55.7901091909284
+ ],
+ [
+ 8.42447838826987,
+ 55.78375297076611
+ ],
+ [
+ 8.459721712590886,
+ 55.77106637546227
+ ],
+ [
+ 8.471969027931252,
+ 55.76941274952962
+ ],
+ [
+ 8.490159131276158,
+ 55.77147978437919
+ ],
+ [
+ 8.503026561730286,
+ 55.7792829105895
+ ],
+ [
+ 8.553307732933744,
+ 55.811296498605216
+ ],
+ [
+ 8.555478140129413,
+ 55.81349270440006
+ ],
+ [
+ 8.555943231512725,
+ 55.81607657225882
+ ],
+ [
+ 8.554237906052201,
+ 55.81912543998764
+ ],
+ [
+ 8.54912194349245,
+ 55.82545586408289
+ ],
+ [
+ 8.548501822773193,
+ 55.82868558605206
+ ],
+ [
+ 8.55361779392601,
+ 55.831527857785694
+ ],
+ [
+ 8.560749141086577,
+ 55.83411167230695
+ ],
+ [
+ 8.569585809065101,
+ 55.83540355393319
+ ],
+ [
+ 8.578732535573415,
+ 55.83524849887616
+ ],
+ [
+ 8.588499385666147,
+ 55.831476173896185
+ ],
+ [
+ 8.598886348943562,
+ 55.829202417393134
+ ],
+ [
+ 8.610875277338248,
+ 55.827936303139445
+ ],
+ [
+ 8.62761844153198,
+ 55.82979665549971
+ ],
+ [
+ 8.64405154841968,
+ 55.833904933787856
+ ],
+ [
+ 8.654593538502695,
+ 55.83478341441331
+ ],
+ [
+ 8.687201370536842,
+ 55.82995170572608
+ ],
+ [
+ 8.757481315770384,
+ 55.80178801830286
+ ],
+ [
+ 8.774586213717082,
+ 55.7988166042023
+ ],
+ [
+ 8.80698733725518,
+ 55.80455269904055
+ ],
+ [
+ 8.823110387699552,
+ 55.81034049952672
+ ],
+ [
+ 8.82295535353463,
+ 55.81680000457975
+ ],
+ [
+ 8.817684358139942,
+ 55.833749880058846
+ ],
+ [
+ 8.828278026543742,
+ 55.84284492031668
+ ],
+ [
+ 8.843470898212058,
+ 55.87230050114686
+ ],
+ [
+ 8.880057814373021,
+ 55.882584143794276
+ ],
+ [
+ 8.929098749185627,
+ 55.88000033168568
+ ],
+ [
+ 8.959381140724814,
+ 55.86400649390912
+ ],
+ [
+ 8.999275340451677,
+ 55.834731729562215
+ ],
+ [
+ 9.013072949743426,
+ 55.8287889521982
+ ],
+ [
+ 9.047334428007561,
+ 55.830907732000924
+ ],
+ [
+ 9.058083123524849,
+ 55.836617956922545
+ ],
+ [
+ 9.070950553214702,
+ 55.835713584575274
+ ],
+ [
+ 9.079787223967422,
+ 55.8482709664429
+ ],
+ [
+ 9.085058218700397,
+ 55.853541949899
+ ],
+ [
+ 9.093894893636694,
+ 55.859329773001974
+ ],
+ [
+ 9.10288658789021,
+ 55.86320544626469
+ ],
+ [
+ 9.12107669118206,
+ 55.86604763999335
+ ],
+ [
+ 9.137974890102475,
+ 55.873902523732745
+ ],
+ [
+ 9.184380325599165,
+ 55.907104616236445
+ ],
+ [
+ 9.211665482605554,
+ 55.93834303653782
+ ],
+ [
+ 9.249389277637226,
+ 55.94299390050005
+ ],
+ [
+ 9.343130320788484,
+ 55.92547553445804
+ ],
+ [
+ 9.374756301344647,
+ 55.91436514245439
+ ],
+ [
+ 9.399922725675944,
+ 55.90165276715717
+ ],
+ [
+ 9.450668982651276,
+ 55.865789329469074
+ ],
+ [
+ 9.458265416105375,
+ 55.85845124894569
+ ],
+ [
+ 9.459660681670151,
+ 55.853309463079334
+ ],
+ [
+ 9.462916296440518,
+ 55.84610060767762
+ ],
+ [
+ 9.468807407800863,
+ 55.83685050143775
+ ],
+ [
+ 9.491131627334669,
+ 55.82170931107715
+ ],
+ [
+ 9.516453079968407,
+ 55.797576426697766
+ ],
+ [
+ 9.530405715603825,
+ 55.78755118193151
+ ],
+ [
+ 9.546218706457406,
+ 55.78450230435875
+ ],
+ [
+ 9.581617059133832,
+ 55.78682771060045
+ ],
+ [
+ 9.600685661506827,
+ 55.78452809944401
+ ],
+ [
+ 9.605491573785457,
+ 55.77886956315117
+ ],
+ [
+ 9.602546012833916,
+ 55.77266835042036
+ ],
+ [
+ 9.598360225006857,
+ 55.76690646748059
+ ],
+ [
+ 9.596344843368806,
+ 55.76083447556116
+ ],
+ [
+ 9.597895136632635,
+ 55.7533930762072
+ ],
+ [
+ 9.608747184175755,
+ 55.74946561617856
+ ],
+ [
+ 9.641665073962033,
+ 55.74693350975045
+ ],
+ [
+ 9.663989289291997,
+ 55.7429543765599
+ ],
+ [
+ 9.675513138155502,
+ 55.73706332971859
+ ],
+ [
+ 9.680629104828782,
+ 55.72848503775071
+ ],
+ [
+ 9.679698927323647,
+ 55.71964835017899
+ ],
+ [
+ 9.675564814682227,
+ 55.70871880301719
+ ],
+ [
+ 9.675520593530397,
+ 55.70859818519244
+ ],
+ [
+ 9.58301842565813,
+ 55.708889073058614
+ ],
+ [
+ 9.559580922573973,
+ 55.715399451221316
+ ],
+ [
+ 9.556407097715837,
+ 55.711574619789154
+ ],
+ [
+ 9.554535349228082,
+ 55.70880764995218
+ ],
+ [
+ 9.553477410079664,
+ 55.70579661797767
+ ],
+ [
+ 9.552744987393648,
+ 55.70111724754837
+ ],
+ [
+ 9.573578319770531,
+ 55.695461314922625
+ ],
+ [
+ 9.645518422359826,
+ 55.69489163686437
+ ],
+ [
+ 9.658539258899422,
+ 55.69037506577659
+ ],
+ [
+ 9.710297071045407,
+ 55.66014232054865
+ ],
+ [
+ 9.720957880166207,
+ 55.64826082310041
+ ],
+ [
+ 9.728200717344835,
+ 55.63751862616716
+ ],
+ [
+ 9.73829186429071,
+ 55.62970613758564
+ ],
+ [
+ 9.75814863414666,
+ 55.62665436576682
+ ],
+ [
+ 9.85377037914811,
+ 55.62665436576682
+ ],
+ [
+ 9.837412956073978,
+ 55.61522043392491
+ ],
+ [
+ 9.785492384007366,
+ 55.59194570508788
+ ],
+ [
+ 9.757660351929335,
+ 55.571600652146245
+ ],
+ [
+ 9.744151237149831,
+ 55.56635161972159
+ ],
+ [
+ 9.720876498698033,
+ 55.56460196345832
+ ],
+ [
+ 9.7124129562298,
+ 55.56028879465636
+ ],
+ [
+ 9.710622593448813,
+ 55.55093009758133
+ ],
+ [
+ 9.71029707258739,
+ 55.541571375268184
+ ],
+ [
+ 9.70687910073204,
+ 55.53729887960059
+ ],
+ [
+ 9.643890821389626,
+ 55.52830638676298
+ ],
+ [
+ 9.621104363115515,
+ 55.518947659415055
+ ],
+ [
+ 9.607269726567162,
+ 55.51679107768524
+ ],
+ [
+ 9.595225456492976,
+ 55.51837799076683
+ ],
+ [
+ 9.572764518424071,
+ 55.52391184390417
+ ],
+ [
+ 9.559580923492513,
+ 55.52362700140266
+ ],
+ [
+ 9.562185092801702,
+ 55.51951732988752
+ ],
+ [
+ 9.566416862569017,
+ 55.51679107768524
+ ],
+ [
+ 9.551931185564605,
+ 55.50511301762154
+ ],
+ [
+ 9.514903192305688,
+ 55.49819572119229
+ ],
+ [
+ 9.49805748833703,
+ 55.489447337187244
+ ],
+ [
+ 9.515635614402576,
+ 55.48314038439945
+ ],
+ [
+ 9.534434439495858,
+ 55.483832079375595
+ ],
+ [
+ 9.5836694664389,
+ 55.49115631008924
+ ],
+ [
+ 9.592784051191279,
+ 55.4942894696549
+ ],
+ [
+ 9.602386913984125,
+ 55.49579497651419
+ ],
+ [
+ 9.660166863421669,
+ 55.4779727281527
+ ],
+ [
+ 9.664235874172954,
+ 55.466701579333595
+ ],
+ [
+ 9.659922722676862,
+ 55.45302969827726
+ ],
+ [
+ 9.648936394034388,
+ 55.4410667994202
+ ],
+ [
+ 9.635264520273964,
+ 55.43573641062711
+ ],
+ [
+ 9.601410350913916,
+ 55.43183013162742
+ ],
+ [
+ 9.586761915082118,
+ 55.42743561501192
+ ],
+ [
+ 9.592051628105969,
+ 55.41632720497663
+ ],
+ [
+ 9.604340041430191,
+ 55.40119051461637
+ ],
+ [
+ 9.607920768390393,
+ 55.393255919491075
+ ],
+ [
+ 9.60515384131264,
+ 55.38202544952899
+ ],
+ [
+ 9.599294468407335,
+ 55.37335848335683
+ ],
+ [
+ 9.600840689547587,
+ 55.36591215908493
+ ],
+ [
+ 9.621104364564076,
+ 55.358547287253366
+ ],
+ [
+ 9.63347415416924,
+ 55.351996150783876
+ ],
+ [
+ 9.645274285828965,
+ 55.34271882718258
+ ],
+ [
+ 9.64828535143677,
+ 55.3329124950804
+ ],
+ [
+ 9.634613476524187,
+ 55.32501861614474
+ ],
+ [
+ 9.634613476355934,
+ 55.31757232507297
+ ],
+ [
+ 9.644867384196797,
+ 55.310939848418954
+ ],
+ [
+ 9.681895378775444,
+ 55.27464427024593
+ ],
+ [
+ 9.694183789662452,
+ 55.27163320086574
+ ],
+ [
+ 9.706065300480738,
+ 55.265082103879735
+ ],
+ [
+ 9.71029707169057,
+ 55.248683994446814
+ ],
+ [
+ 9.70655358250278,
+ 55.23700593315368
+ ],
+ [
+ 9.699473504757053,
+ 55.22980378727489
+ ],
+ [
+ 9.692881707753767,
+ 55.22101472524301
+ ],
+ [
+ 9.689789258701413,
+ 55.204901430334786
+ ],
+ [
+ 9.684418165801725,
+ 55.1975772258461
+ ],
+ [
+ 9.671560091360043,
+ 55.19245025412797
+ ],
+ [
+ 9.655772332527341,
+ 55.18943920149645
+ ],
+ [
+ 9.642100457188864,
+ 55.18842194432937
+ ],
+ [
+ 9.583343946606071,
+ 55.1946475355221
+ ],
+ [
+ 9.569834831845437,
+ 55.192531641077665
+ ],
+ [
+ 9.561534049384047,
+ 55.187201231334384
+ ],
+ [
+ 9.554942253758345,
+ 55.18056874898944
+ ],
+ [
+ 9.545909050111222,
+ 55.17413971738564
+ ],
+ [
+ 9.521820509343181,
+ 55.165432037289875
+ ],
+ [
+ 9.507172071189414,
+ 55.16193268637121
+ ],
+ [
+ 9.494395378724967,
+ 55.16054921755007
+ ],
+ [
+ 9.486338738265216,
+ 55.156236073328046
+ ],
+ [
+ 9.486338738048264,
+ 55.14740631760494
+ ],
+ [
+ 9.494639519095053,
+ 55.140082099191496
+ ],
+ [
+ 9.511892123095185,
+ 55.140082099191496
+ ],
+ [
+ 9.51189212281142,
+ 55.13320546863723
+ ],
+ [
+ 9.482432487933002,
+ 55.13377513115708
+ ],
+ [
+ 9.46876061325089,
+ 55.13178132015613
+ ],
+ [
+ 9.456553582031837,
+ 55.12571849240018
+ ],
+ [
+ 9.466644726946955,
+ 55.12140534033129
+ ],
+ [
+ 9.479177279933268,
+ 55.11969635516042
+ ],
+ [
+ 9.508067254074023,
+ 55.119533596927006
+ ],
+ [
+ 9.521332227373646,
+ 55.115871490677044
+ ],
+ [
+ 9.535817904870262,
+ 55.1068789713773
+ ],
+ [
+ 9.55958092479999,
+ 55.08539459550287
+ ],
+ [
+ 9.533702018955712,
+ 55.065822657444464
+ ],
+ [
+ 9.516612174926012,
+ 55.0565453140707
+ ],
+ [
+ 9.498057487958585,
+ 55.05068593947829
+ ],
+ [
+ 9.452403191174048,
+ 55.045355538197924
+ ],
+ [
+ 9.43433678501543,
+ 55.037543036194315
+ ],
+ [
+ 9.442881707086386,
+ 55.02338288109222
+ ],
+ [
+ 9.455251498001637,
+ 55.023179429020075
+ ],
+ [
+ 9.511892122979443,
+ 55.0363630227412
+ ],
+ [
+ 9.537364128999789,
+ 55.03538645999635
+ ],
+ [
+ 9.545909049979404,
+ 55.0363630227412
+ ],
+ [
+ 9.553233268909093,
+ 55.03961822986234
+ ],
+ [
+ 9.5602319668574,
+ 55.04441966221794
+ ],
+ [
+ 9.568532748091803,
+ 55.04877350614511
+ ],
+ [
+ 9.579925976958215,
+ 55.05068593947829
+ ],
+ [
+ 9.601573112806816,
+ 55.045396223596086
+ ],
+ [
+ 9.641856315981128,
+ 55.019720769765065
+ ],
+ [
+ 9.662608269005544,
+ 55.00971100506752
+ ],
+ [
+ 9.717539910000141,
+ 55.00116608300078
+ ],
+ [
+ 9.730967634208934,
+ 54.995428657673976
+ ],
+ [
+ 9.738617384347835,
+ 54.98086172527219
+ ],
+ [
+ 9.760752793745374,
+ 54.904974600182385
+ ],
+ [
+ 9.763194214662695,
+ 54.900458169096815
+ ],
+ [
+ 9.761566598287489,
+ 54.89740627139821
+ ],
+ [
+ 9.751963736498542,
+ 54.89301176753617
+ ],
+ [
+ 9.737071160338674,
+ 54.892767649169954
+ ],
+ [
+ 9.722992378965934,
+ 54.89765039592069
+ ],
+ [
+ 9.71143638424735,
+ 54.89858620959221
+ ],
+ [
+ 9.703461141444187,
+ 54.886175940011945
+ ],
+ [
+ 9.715993693787642,
+ 54.88377522813949
+ ],
+ [
+ 9.727875205552087,
+ 54.879462098791386
+ ],
+ [
+ 9.737803575272267,
+ 54.873521143106785
+ ],
+ [
+ 9.744476750329023,
+ 54.86627828021659
+ ],
+ [
+ 9.748301616468707,
+ 54.857855381702095
+ ],
+ [
+ 9.750254752226777,
+ 54.84625883416297
+ ],
+ [
+ 9.745941600347193,
+ 54.83600492963127
+ ],
+ [
+ 9.731211778232606,
+ 54.83152895546649
+ ],
+ [
+ 9.719574418590241,
+ 54.83584223436277
+ ],
+ [
+ 9.70533286685405,
+ 54.85455948677131
+ ],
+ [
+ 9.69662518627868,
+ 54.85887266866664
+ ],
+ [
+ 9.65503991521983,
+ 54.85708248889256
+ ],
+ [
+ 9.642100447333359,
+ 54.85887266866664
+ ],
+ [
+ 9.623301639062129,
+ 54.865301950477324
+ ],
+ [
+ 9.624685079444063,
+ 54.86928929846198
+ ],
+ [
+ 9.632823112793242,
+ 54.87311432242379
+ ],
+ [
+ 9.634613488047739,
+ 54.879340048554255
+ ],
+ [
+ 9.616953961993252,
+ 54.89423236422661
+ ],
+ [
+ 9.610362173122594,
+ 54.90322497857955
+ ],
+ [
+ 9.617523624347543,
+ 54.90729389169413
+ ],
+ [
+ 9.644216333912572,
+ 54.9132753574955
+ ],
+ [
+ 9.638845235807219,
+ 54.9247499154399
+ ],
+ [
+ 9.615000849914543,
+ 54.932684673309545
+ ],
+ [
+ 9.586761924293263,
+ 54.92772056113519
+ ],
+ [
+ 9.581228056658931,
+ 54.92177963571003
+ ],
+ [
+ 9.568044470857693,
+ 54.90143468533682
+ ],
+ [
+ 9.556813992140066,
+ 54.890326165465375
+ ],
+ [
+ 9.550547716384926,
+ 54.88271711947688
+ ],
+ [
+ 9.545909060946043,
+ 54.879340048554255
+ ],
+ [
+ 9.538340695791549,
+ 54.87791581726283
+ ],
+ [
+ 9.522146041443072,
+ 54.87929936515986
+ ],
+ [
+ 9.515147339700981,
+ 54.87592203909655
+ ],
+ [
+ 9.498057491648527,
+ 54.86432530809857
+ ],
+ [
+ 9.456390811646466,
+ 54.843817331229346
+ ],
+ [
+ 9.442881700433123,
+ 54.83152895546649
+ ],
+ [
+ 9.43628990133417,
+ 54.81045145462693
+ ],
+ [
+ 9.437503025229184,
+ 54.81041110414442
+ ],
+ [
+ 9.436922237825144,
+ 54.81014390156533
+ ],
+ [
+ 9.422142786558886,
+ 54.8072501319129
+ ],
+ [
+ 9.405192909137645,
+ 54.80838691919905
+ ],
+ [
+ 9.385039094237555,
+ 54.819393910814675
+ ],
+ [
+ 9.366952354960588,
+ 54.81701694497835
+ ],
+ [
+ 9.35599694771287,
+ 54.811332495311206
+ ],
+ [
+ 9.341734265085929,
+ 54.809058891495006
+ ],
+ [
+ 9.332019086123955,
+ 54.803219327357674
+ ],
+ [
+ 9.317032927368201,
+ 54.80161745087352
+ ],
+ [
+ 9.24406580607896,
+ 54.80177249106897
+ ],
+ [
+ 9.22639245957918,
+ 54.805906520628426
+ ],
+ [
+ 9.219054415111097,
+ 54.81779221185391
+ ],
+ [
+ 9.216057170714539,
+ 54.831744712344474
+ ],
+ [
+ 9.211509651938867,
+ 54.84182169749087
+ ],
+ [
+ 9.194766481215852,
+ 54.850399897874475
+ ],
+ [
+ 8.982686407978816,
+ 54.87933877719802
+ ],
+ [
+ 8.904138224106939,
+ 54.89794220143914
+ ],
+ [
+ 8.8242464522875,
+ 54.905900278812105
+ ],
+ [
+ 8.800888718498516,
+ 54.90383340251752
+ ],
+ [
+ 8.73277916991628,
+ 54.88905378418294
+ ],
+ [
+ 8.69588221081347,
+ 54.88998415054627
+ ],
+ [
+ 8.660775938016705,
+ 54.896311211608904
+ ],
+ [
+ 8.668793158135317,
+ 54.913519503092054
+ ],
+ [
+ 8.661387555528917,
+ 54.92031470113451
+ ],
+ [
+ 8.67481530286424,
+ 54.94794346653396
+ ],
+ [
+ 8.66635176611271,
+ 54.97333415020506
+ ],
+ [
+ 8.650401248308818,
+ 54.997992397502145
+ ],
+ [
+ 8.640961134079038,
+ 55.02338288109222
+ ],
+ [
+ 8.64478600409789,
+ 55.055731513350935
+ ],
+ [
+ 8.656504753869235,
+ 55.08250559919663
+ ],
+ [
+ 8.670909050192723,
+ 55.10712311664887
+ ],
+ [
+ 8.681895378827914,
+ 55.13320546863723
+ ],
+ [
+ 8.575531446003332,
+ 55.144110419045376
+ ],
+ [
+ 8.563649935800209,
+ 55.143377994220394
+ ],
+ [
+ 8.5599064458335,
+ 55.13507721368162
+ ],
+ [
+ 8.55738365988096,
+ 55.096380925339744
+ ],
+ [
+ 8.551117383700921,
+ 55.09137603982726
+ ],
+ [
+ 8.541351759227348,
+ 55.0902774141742
+ ],
+ [
+ 8.528575066124937,
+ 55.081976631746606
+ ],
+ [
+ 8.514821810967796,
+ 55.07078685054798
+ ],
+ [
+ 8.499359570988963,
+ 55.06557851784406
+ ],
+ [
+ 8.48389733210726,
+ 55.06761302450786
+ ],
+ [
+ 8.470225456956383,
+ 55.078599350384444
+ ],
+ [
+ 8.458832226774858,
+ 55.10346100182525
+ ],
+ [
+ 8.46021569098076,
+ 55.12738678572781
+ ],
+ [
+ 8.46648196699231,
+ 55.15159739789119
+ ],
+ [
+ 8.470225456362806,
+ 55.17755767504625
+ ],
+ [
+ 8.488942905351012,
+ 55.197170319917745
+ ],
+ [
+ 8.530528190421357,
+ 55.200751035931305
+ ],
+ [
+ 8.572601758763724,
+ 55.19367096272045
+ ],
+ [
+ 8.593760613317961,
+ 55.18097565740169
+ ],
+ [
+ 8.580332878739016,
+ 55.1789411443792
+ ],
+ [
+ 8.56950931161871,
+ 55.17377351359249
+ ],
+ [
+ 8.552093945752262,
+ 55.16054921755007
+ ],
+ [
+ 8.55518639366981,
+ 55.149359437402225
+ ],
+ [
+ 8.614756706687587,
+ 55.14435455467931
+ ],
+ [
+ 8.669281445794743,
+ 55.13690826717837
+ ],
+ [
+ 8.689463737645074,
+ 55.14158762713324
+ ],
+ [
+ 8.687185092204567,
+ 55.160589913803065
+ ],
+ [
+ 8.668793165548237,
+ 55.1946475355221
+ ],
+ [
+ 8.65593509262145,
+ 55.23749421652995
+ ],
+ [
+ 8.648448112966872,
+ 55.28025950654425
+ ],
+ [
+ 8.650401238383944,
+ 55.291734122265
+ ],
+ [
+ 8.659353061526872,
+ 55.30581290021579
+ ],
+ [
+ 8.661387565708639,
+ 55.31443919101
+ ],
+ [
+ 8.659678581507512,
+ 55.32721588025758
+ ],
+ [
+ 8.648448113482958,
+ 55.352280998613985
+ ],
+ [
+ 8.63941490970425,
+ 55.39740631294858
+ ],
+ [
+ 8.632090692441972,
+ 55.4187686417526
+ ],
+ [
+ 8.617686393771129,
+ 55.43793365785988
+ ],
+ [
+ 8.592539910508489,
+ 55.44928620699118
+ ],
+ [
+ 8.556651239028431,
+ 55.45429109719838
+ ],
+ [
+ 8.490000847751606,
+ 55.455308345457375
+ ],
+ [
+ 8.441905144419705,
+ 55.46393464287095
+ ],
+ [
+ 8.401133658984431,
+ 55.48566313872918
+ ],
+ [
+ 8.31039472791438,
+ 55.562689532961016
+ ],
+ [
+ 8.311208529085114,
+ 55.56976958503396
+ ],
+ [
+ 8.33236738526004,
+ 55.57200757180966
+ ],
+ [
+ 8.332367385110652,
+ 55.57827384569545
+ ],
+ [
+ 8.31316165560184,
+ 55.582709060523605
+ ],
+ [
+ 8.290212436488272,
+ 55.58380769093401
+ ],
+ [
+ 8.267832878595701,
+ 55.580511780240855
+ ],
+ [
+ 8.250498895247686,
+ 55.57200757180966
+ ],
+ [
+ 8.239024284135542,
+ 55.55792876563688
+ ],
+ [
+ 8.24203535163392,
+ 55.54852935876459
+ ],
+ [
+ 8.271739128343198,
+ 55.53046294863682
+ ],
+ [
+ 8.26246178524464,
+ 55.52920156848994
+ ],
+ [
+ 8.25635826872602,
+ 55.52680084408582
+ ],
+ [
+ 8.243662956628619,
+ 55.51679107768524
+ ],
+ [
+ 8.2667749356007,
+ 55.51439036130193
+ ],
+ [
+ 8.293955924990433,
+ 55.504339910862626
+ ],
+ [
+ 8.316742383518553,
+ 55.48940663616638
+ ],
+ [
+ 8.326182486754506,
+ 55.472113330334665
+ ],
+ [
+ 8.313324415145892,
+ 55.46930573307157
+ ],
+ [
+ 8.193614128162096,
+ 55.52594634394013
+ ],
+ [
+ 8.16920006603527,
+ 55.53408437750778
+ ],
+ [
+ 8.11060631515597,
+ 55.5402285707321
+ ],
+ [
+ 8.095225456630736,
+ 55.549058329623165
+ ],
+ [
+ 8.094004754717746,
+ 55.563910233445114
+ ],
+ [
+ 8.150238476726635,
+ 55.64923736805667
+ ],
+ [
+ 8.16911868754386,
+ 55.687933683190366
+ ],
+ [
+ 8.18165123761594,
+ 55.72907135049285
+ ],
+ [
+ 8.183360224078854,
+ 55.76959872976427
+ ],
+ [
+ 8.17560707357738,
+ 55.80147645034364
+ ]
+ ]
+ ],
+ [
+ [
+ [
+ 10.450205936170002,
+ 54.89980716115108
+ ],
+ [
+ 10.466644718324003,
+ 54.89057028059205
+ ],
+ [
+ 10.485687705388898,
+ 54.88369385939922
+ ],
+ [
+ 10.499196813618319,
+ 54.874579198919406
+ ],
+ [
+ 10.498708519474526,
+ 54.85887266866664
+ ],
+ [
+ 10.508148639101282,
+ 54.85492590627374
+ ],
+ [
+ 10.518565299877851,
+ 54.851996159604916
+ ],
+ [
+ 10.440114787779645,
+ 54.841498203481926
+ ],
+ [
+ 10.418630399413951,
+ 54.827826174591685
+ ],
+ [
+ 10.406260620158449,
+ 54.82469318364205
+ ],
+ [
+ 10.383474148793761,
+ 54.830755855204266
+ ],
+ [
+ 10.211436397379803,
+ 54.93935785463954
+ ],
+ [
+ 10.195160349701462,
+ 54.96185941501375
+ ],
+ [
+ 10.18921959670927,
+ 54.97748443858516
+ ],
+ [
+ 10.203868030044573,
+ 54.96808496988032
+ ],
+ [
+ 10.288747587810368,
+ 54.940741229231854
+ ],
+ [
+ 10.313731325997992,
+ 54.92772056113519
+ ],
+ [
+ 10.337250187560384,
+ 54.90591045113062
+ ],
+ [
+ 10.34961998817365,
+ 54.89671458701082
+ ],
+ [
+ 10.364268423404269,
+ 54.89301176753617
+ ],
+ [
+ 10.379405142092576,
+ 54.89305245396206
+ ],
+ [
+ 10.391368043222768,
+ 54.8954532868832
+ ],
+ [
+ 10.395843943210103,
+ 54.903469110826904
+ ],
+ [
+ 10.38884523544047,
+ 54.92031470113451
+ ],
+ [
+ 10.403086773602622,
+ 54.90839251367642
+ ],
+ [
+ 10.41521243272885,
+ 54.87848538132226
+ ],
+ [
+ 10.426036014434958,
+ 54.872504093084274
+ ],
+ [
+ 10.446787969577972,
+ 54.87238204745899
+ ],
+ [
+ 10.46094811694495,
+ 54.87449783424342
+ ],
+ [
+ 10.46363365701385,
+ 54.882473014767264
+ ],
+ [
+ 10.450205936170002,
+ 54.89980716115108
+ ]
+ ]
+ ],
+ [
+ [
+ [
+ 10.011403835368746,
+ 54.975531237746765
+ ],
+ [
+ 10.025889521517511,
+ 54.959784277055974
+ ],
+ [
+ 10.065114769596327,
+ 54.890814396042586
+ ],
+ [
+ 10.064463727131873,
+ 54.88100807546556
+ ],
+ [
+ 10.050466333670308,
+ 54.87885153057662
+ ],
+ [
+ 10.021983280491938,
+ 54.879340048554255
+ ],
+ [
+ 10.00709068967912,
+ 54.87482329316303
+ ],
+ [
+ 9.98755943924883,
+ 54.86591215595805
+ ],
+ [
+ 9.968028192634538,
+ 54.85952384968041
+ ],
+ [
+ 9.953623907000386,
+ 54.86261643474964
+ ],
+ [
+ 9.937185102720004,
+ 54.870917188439385
+ ],
+ [
+ 9.891123905886525,
+ 54.87929936515986
+ ],
+ [
+ 9.874278198575254,
+ 54.886175940011945
+ ],
+ [
+ 9.874278189479712,
+ 54.89301176753617
+ ],
+ [
+ 9.900889525744638,
+ 54.90070229767064
+ ],
+ [
+ 9.932302280044915,
+ 54.892808335541815
+ ],
+ [
+ 9.963877803202015,
+ 54.87995030054906
+ ],
+ [
+ 9.991547081000085,
+ 54.872504093084274
+ ],
+ [
+ 9.991547082457021,
+ 54.879340048554255
+ ],
+ [
+ 9.943695505917669,
+ 54.90363186583856
+ ],
+ [
+ 9.898285343276806,
+ 54.91299052137614
+ ],
+ [
+ 9.887950058169896,
+ 54.913519503092054
+ ],
+ [
+ 9.877940289401584,
+ 54.91132220440013
+ ],
+ [
+ 9.863536002513195,
+ 54.90306222392806
+ ],
+ [
+ 9.85377038953248,
+ 54.89980716115108
+ ],
+ [
+ 9.814707884250783,
+ 54.9010684911391
+ ],
+ [
+ 9.780935092277268,
+ 54.91681550039663
+ ],
+ [
+ 9.761241085303888,
+ 54.94306065453227
+ ],
+ [
+ 9.764414903532382,
+ 54.975531237746765
+ ],
+ [
+ 9.75814862753651,
+ 54.975531237746765
+ ],
+ [
+ 9.758148636960843,
+ 54.98175693929783
+ ],
+ [
+ 9.78028404428746,
+ 54.972845700107776
+ ],
+ [
+ 9.804047068829648,
+ 54.957180053498845
+ ],
+ [
+ 9.827403179286081,
+ 54.94546118726518
+ ],
+ [
+ 9.847504109503829,
+ 54.948187658239775
+ ],
+ [
+ 9.843272324980585,
+ 54.95172752061992
+ ],
+ [
+ 9.836680536522799,
+ 54.95856356253031
+ ],
+ [
+ 9.832774283830954,
+ 54.96515532676951
+ ],
+ [
+ 9.836680530222894,
+ 54.96808496988032
+ ],
+ [
+ 9.84213301051801,
+ 54.9702823124562
+ ],
+ [
+ 9.835215699849782,
+ 54.975165213661896
+ ],
+ [
+ 9.793711782987808,
+ 54.9921735454269
+ ],
+ [
+ 9.786794467000117,
+ 55.00014883000086
+ ],
+ [
+ 9.799082879005613,
+ 55.00971100506752
+ ],
+ [
+ 9.799082879020935,
+ 55.01654694225437
+ ],
+ [
+ 9.720550976998437,
+ 55.012274480979904
+ ],
+ [
+ 9.706879101955433,
+ 55.013128972450346
+ ],
+ [
+ 9.696950716992406,
+ 55.01752350490582
+ ],
+ [
+ 9.684825066013111,
+ 55.0185407571608
+ ],
+ [
+ 9.6741642589505,
+ 55.02106354338773
+ ],
+ [
+ 9.66879316492514,
+ 55.03021881607416
+ ],
+ [
+ 9.672373894091065,
+ 55.034654040123684
+ ],
+ [
+ 9.691254101941746,
+ 55.042792058281215
+ ],
+ [
+ 9.696625195957703,
+ 55.05068593947829
+ ],
+ [
+ 9.678233268870711,
+ 55.05019765640373
+ ],
+ [
+ 9.661631706892349,
+ 55.04653554766841
+ ],
+ [
+ 9.645518424958007,
+ 55.044867254478994
+ ],
+ [
+ 9.628428581958014,
+ 55.05068593947829
+ ],
+ [
+ 9.628428581901334,
+ 55.058050846775586
+ ],
+ [
+ 9.716644726875185,
+ 55.08222076846654
+ ],
+ [
+ 9.751963737795956,
+ 55.08539459550287
+ ],
+ [
+ 9.793467644098452,
+ 55.080755928198464
+ ],
+ [
+ 9.853037956996921,
+ 55.04010651196177
+ ],
+ [
+ 9.884776237923482,
+ 55.03021881607416
+ ],
+ [
+ 9.929047071091386,
+ 55.02619049809902
+ ],
+ [
+ 9.960948112973046,
+ 55.01504140767561
+ ],
+ [
+ 9.986582881120201,
+ 54.99774813938632
+ ],
+ [
+ 10.011403835368746,
+ 54.975531237746765
+ ]
+ ]
+ ],
+ [
+ [
+ [
+ 10.854991081902677,
+ 55.0438500009283
+ ],
+ [
+ 10.861989780056026,
+ 55.07208893461515
+ ],
+ [
+ 10.895518425158832,
+ 55.12152741273652
+ ],
+ [
+ 10.934255405235449,
+ 55.159654041563364
+ ],
+ [
+ 10.956716341899812,
+ 55.1537132819106
+ ],
+ [
+ 10.908946159915544,
+ 55.03021881607416
+ ],
+ [
+ 10.889903177323907,
+ 54.99738175282415
+ ],
+ [
+ 10.837738479544045,
+ 54.9372419571147
+ ],
+ [
+ 10.741058781045634,
+ 54.75238668564687
+ ],
+ [
+ 10.731293165941244,
+ 54.74005769456112
+ ],
+ [
+ 10.717539896946489,
+ 54.73578507631932
+ ],
+ [
+ 10.693858268610699,
+ 54.73529693761481
+ ],
+ [
+ 10.681895392013429,
+ 54.74481857368004
+ ],
+ [
+ 10.656423360525931,
+ 54.80365617628867
+ ],
+ [
+ 10.645518418570633,
+ 54.814601882421414
+ ],
+ [
+ 10.632090693905937,
+ 54.82343174083612
+ ],
+ [
+ 10.601084833774252,
+ 54.8383243060986
+ ],
+ [
+ 10.625336130705339,
+ 54.83930081076209
+ ],
+ [
+ 10.64087976293952,
+ 54.84320719887792
+ ],
+ [
+ 10.653493696091447,
+ 54.85163006371646
+ ],
+ [
+ 10.669281436506099,
+ 54.86627828021659
+ ],
+ [
+ 10.678558779577445,
+ 54.88304259292659
+ ],
+ [
+ 10.68441815483303,
+ 54.88784402887487
+ ],
+ [
+ 10.69385826219115,
+ 54.8827578036266
+ ],
+ [
+ 10.698903828140368,
+ 54.88117081060754
+ ],
+ [
+ 10.724457235227463,
+ 54.886175940011945
+ ],
+ [
+ 10.722178586320611,
+ 54.89606359231655
+ ],
+ [
+ 10.717784039243305,
+ 54.90729389169413
+ ],
+ [
+ 10.708994986718787,
+ 54.90298084665635
+ ],
+ [
+ 10.700368689769318,
+ 54.901556750231435
+ ],
+ [
+ 10.691905142720836,
+ 54.90298084665635
+ ],
+ [
+ 10.682953310278243,
+ 54.90729389169413
+ ],
+ [
+ 10.697032085607441,
+ 54.93134173641795
+ ],
+ [
+ 10.718272330241486,
+ 54.9491641103546
+ ],
+ [
+ 10.774099162127976,
+ 54.98655841413487
+ ],
+ [
+ 10.798187704865208,
+ 54.9965681991756
+ ],
+ [
+ 10.854991081902677,
+ 55.0438500009283
+ ]
+ ]
+ ],
+ [
+ [
+ [
+ 8.476410352019505,
+ 55.34544505427165
+ ],
+ [
+ 8.451670769020973,
+ 55.33974844029284
+ ],
+ [
+ 8.425303582488972,
+ 55.35553620687327
+ ],
+ [
+ 8.402354362199478,
+ 55.38003163172106
+ ],
+ [
+ 8.370941602071719,
+ 55.42743561501192
+ ],
+ [
+ 8.364268424786678,
+ 55.44904205598535
+ ],
+ [
+ 8.374359571322282,
+ 55.46328360054461
+ ],
+ [
+ 8.408213736896347,
+ 55.46841053349749
+ ],
+ [
+ 8.411957225773904,
+ 55.43870675577208
+ ],
+ [
+ 8.428558788624667,
+ 55.43353911071081
+ ],
+ [
+ 8.44890384071442,
+ 55.43553292701415
+ ],
+ [
+ 8.463389519072525,
+ 55.42743561501192
+ ],
+ [
+ 8.463063997781898,
+ 55.41571686095161
+ ],
+ [
+ 8.450205924979855,
+ 55.39337799671668
+ ],
+ [
+ 8.45289147117176,
+ 55.38336821940135
+ ],
+ [
+ 8.46045983159376,
+ 55.37238189731395
+ ],
+ [
+ 8.46745853142103,
+ 55.35370516487879
+ ],
+ [
+ 8.476410352019505,
+ 55.34544505427165
+ ]
+ ]
+ ],
+ [
+ [
+ [
+ 10.66627037913914,
+ 55.064276435556344
+ ],
+ [
+ 10.621429884138543,
+ 55.064276435556344
+ ],
+ [
+ 10.604746940990548,
+ 55.06256744989268
+ ],
+ [
+ 10.586110872889401,
+ 55.05752187575211
+ ],
+ [
+ 10.568858268830168,
+ 55.04987213508023
+ ],
+ [
+ 10.556407096996702,
+ 55.04010651196177
+ ],
+ [
+ 10.538259310962179,
+ 55.029730535570685
+ ],
+ [
+ 10.516856316010402,
+ 55.031317450117264
+ ],
+ [
+ 10.477549674906069,
+ 55.0438500009283
+ ],
+ [
+ 10.410655143950128,
+ 55.04694244942693
+ ],
+ [
+ 10.388845247954656,
+ 55.05068593947829
+ ],
+ [
+ 10.37916100379613,
+ 55.05451080665386
+ ],
+ [
+ 10.371348504047862,
+ 55.05902741155003
+ ],
+ [
+ 10.36255944116261,
+ 55.06273021187225
+ ],
+ [
+ 10.350922071135017,
+ 55.064276435556344
+ ],
+ [
+ 10.325938346829322,
+ 55.06289296902591
+ ],
+ [
+ 10.31275475394861,
+ 55.06366608240427
+ ],
+ [
+ 10.234222852047766,
+ 55.09223053655587
+ ],
+ [
+ 10.20525149839707,
+ 55.09369538464015
+ ],
+ [
+ 10.197601758649801,
+ 55.08608632590246
+ ],
+ [
+ 10.197032097133034,
+ 55.064276435556344
+ ],
+ [
+ 10.181895379145494,
+ 55.070868232704555
+ ],
+ [
+ 10.158376497826332,
+ 55.08665598995992
+ ],
+ [
+ 10.14177493604731,
+ 55.09223053655587
+ ],
+ [
+ 10.121267122585314,
+ 55.09247467211241
+ ],
+ [
+ 10.083506706610983,
+ 55.086981507396715
+ ],
+ [
+ 10.06592858204699,
+ 55.09223053655587
+ ],
+ [
+ 10.14258873776995,
+ 55.125799869296046
+ ],
+ [
+ 10.156016472101667,
+ 55.140082099191496
+ ],
+ [
+ 10.1386824883511,
+ 55.154282949122766
+ ],
+ [
+ 10.126149936039596,
+ 55.17064036746492
+ ],
+ [
+ 10.112966342065567,
+ 55.18378327077089
+ ],
+ [
+ 10.094004754197709,
+ 55.18842194432937
+ ],
+ [
+ 10.078949415716782,
+ 55.18593985446082
+ ],
+ [
+ 10.067556185548955,
+ 55.18227773066841
+ ],
+ [
+ 10.055837436406074,
+ 55.17975495480498
+ ],
+ [
+ 10.039317254371431,
+ 55.18097565740169
+ ],
+ [
+ 10.016612174423653,
+ 55.19428131015439
+ ],
+ [
+ 10.003591342269122,
+ 55.197333078199385
+ ],
+ [
+ 9.991547071195688,
+ 55.18842194432937
+ ],
+ [
+ 9.994476758325675,
+ 55.17279693396882
+ ],
+ [
+ 10.029144726584947,
+ 55.14240142906952
+ ],
+ [
+ 10.025645379033746,
+ 55.12571849240018
+ ],
+ [
+ 10.020681185789915,
+ 55.12445709550095
+ ],
+ [
+ 9.985524935862239,
+ 55.12775299535552
+ ],
+ [
+ 9.982920768663558,
+ 55.130601299983184
+ ],
+ [
+ 9.984385613312474,
+ 55.13467031872778
+ ],
+ [
+ 9.984141472099946,
+ 55.140082099191496
+ ],
+ [
+ 9.986989780308171,
+ 55.139349680675494
+ ],
+ [
+ 9.989756706766967,
+ 55.143377994220394
+ ],
+ [
+ 9.988780144385869,
+ 55.15184154359939
+ ],
+ [
+ 9.974294467366468,
+ 55.174017649372324
+ ],
+ [
+ 9.97201582060627,
+ 55.18471913830127
+ ],
+ [
+ 9.974294467524063,
+ 55.19525788624932
+ ],
+ [
+ 9.980723503692467,
+ 55.204901430334786
+ ],
+ [
+ 9.979502799841496,
+ 55.21283600311051
+ ],
+ [
+ 9.96412194133203,
+ 55.21816641196038
+ ],
+ [
+ 9.924001498158022,
+ 55.22394440489207
+ ],
+ [
+ 9.911387565608264,
+ 55.228501690301705
+ ],
+ [
+ 9.900157096554675,
+ 55.23411692365382
+ ],
+ [
+ 9.89535566472633,
+ 55.23899973871322
+ ],
+ [
+ 9.893239780248472,
+ 55.24445221898316
+ ],
+ [
+ 9.883962435326753,
+ 55.253485410909555
+ ],
+ [
+ 9.88168378977043,
+ 55.25950755124045
+ ],
+ [
+ 9.883962436303584,
+ 55.26561107364629
+ ],
+ [
+ 9.893239779770504,
+ 55.27464427024593
+ ],
+ [
+ 9.895355664962064,
+ 55.28025950654425
+ ],
+ [
+ 9.89323977969637,
+ 55.29124583135807
+ ],
+ [
+ 9.88396243549658,
+ 55.308783263959086
+ ],
+ [
+ 9.881683789667559,
+ 55.31443919101
+ ],
+ [
+ 9.884776237233998,
+ 55.348863013818445
+ ],
+ [
+ 9.87452233206857,
+ 55.353338934821686
+ ],
+ [
+ 9.858897331987555,
+ 55.35407135584959
+ ],
+ [
+ 9.829844597548924,
+ 55.352280998613985
+ ],
+ [
+ 9.816579623363888,
+ 55.35521068339016
+ ],
+ [
+ 9.80241946702246,
+ 55.36237213727084
+ ],
+ [
+ 9.778086783780262,
+ 55.380275768233055
+ ],
+ [
+ 9.7949324871601,
+ 55.38385650184981
+ ],
+ [
+ 9.813649937008709,
+ 55.3851586081649
+ ],
+ [
+ 9.830088736929644,
+ 55.38890207211294
+ ],
+ [
+ 9.840098503195733,
+ 55.40009185432919
+ ],
+ [
+ 9.78549238492329,
+ 55.41376375015809
+ ],
+ [
+ 9.793711785483225,
+ 55.41278718283455
+ ],
+ [
+ 9.819590691926528,
+ 55.41376375015809
+ ],
+ [
+ 9.809255405561906,
+ 55.425482495772144
+ ],
+ [
+ 9.791026236892481,
+ 55.43183013162742
+ ],
+ [
+ 9.75131269645672,
+ 55.43488190853541
+ ],
+ [
+ 9.73731529931474,
+ 55.438177793681234
+ ],
+ [
+ 9.703461133117822,
+ 55.462144262260786
+ ],
+ [
+ 9.7034611327263,
+ 55.46841053349749
+ ],
+ [
+ 9.73023522162668,
+ 55.46149322646713
+ ],
+ [
+ 9.758474156547944,
+ 55.447495853743064
+ ],
+ [
+ 9.78630618710932,
+ 55.43720125239728
+ ],
+ [
+ 9.812185092034932,
+ 55.4410667994202
+ ],
+ [
+ 9.713389518897008,
+ 55.49103424574749
+ ],
+ [
+ 9.676280143179644,
+ 55.49567290899384
+ ],
+ [
+ 9.683929884991167,
+ 55.50604890707477
+ ],
+ [
+ 9.69353274851258,
+ 55.51019929223739
+ ],
+ [
+ 9.70476321603026,
+ 55.51239654021229
+ ],
+ [
+ 9.717784049562201,
+ 55.51679107768524
+ ],
+ [
+ 9.755137566103793,
+ 55.54413483425302
+ ],
+ [
+ 9.809092645416975,
+ 55.55027905602217
+ ],
+ [
+ 9.827403191077227,
+ 55.548407294925084
+ ],
+ [
+ 9.830821161406048,
+ 55.54311759685677
+ ],
+ [
+ 9.83130943938249,
+ 55.534735399602695
+ ],
+ [
+ 9.840098502448265,
+ 55.52362700140266
+ ],
+ [
+ 9.88819420627223,
+ 55.508693743316286
+ ],
+ [
+ 9.940277540457153,
+ 55.51923249442325
+ ],
+ [
+ 10.039317254046738,
+ 55.55776601854763
+ ],
+ [
+ 10.168630406190562,
+ 55.58226147378536
+ ],
+ [
+ 10.212657096884394,
+ 55.60065338566679
+ ],
+ [
+ 10.237315300582194,
+ 55.60553620669113
+ ],
+ [
+ 10.264170767786522,
+ 55.6066348190875
+ ],
+ [
+ 10.272146030584128,
+ 55.60553620669113
+ ],
+ [
+ 10.272797071676962,
+ 55.60252513975437
+ ],
+ [
+ 10.279063347533423,
+ 55.58832429210833
+ ],
+ [
+ 10.279551630326392,
+ 55.585109783190866
+ ],
+ [
+ 10.293630405405752,
+ 55.588080149640064
+ ],
+ [
+ 10.297129755584754,
+ 55.59564852311406
+ ],
+ [
+ 10.293793163996485,
+ 55.61611561553079
+ ],
+ [
+ 10.307627800466742,
+ 55.61823151732566
+ ],
+ [
+ 10.338389519013925,
+ 55.610093492157546
+ ],
+ [
+ 10.407399936317791,
+ 55.58295319559485
+ ],
+ [
+ 10.423024937576159,
+ 55.57200757180966
+ ],
+ [
+ 10.435394727884422,
+ 55.55971915298427
+ ],
+ [
+ 10.44418379090594,
+ 55.55316804622046
+ ],
+ [
+ 10.453623895510075,
+ 55.55027905602217
+ ],
+ [
+ 10.471934441874337,
+ 55.54901765483858
+ ],
+ [
+ 10.488536002447987,
+ 55.54547758855931
+ ],
+ [
+ 10.518809439476504,
+ 55.533433317923496
+ ],
+ [
+ 10.502452018864341,
+ 55.537176823476415
+ ],
+ [
+ 10.486827018864547,
+ 55.537176823476415
+ ],
+ [
+ 10.475271029151596,
+ 55.5319684824506
+ ],
+ [
+ 10.470713737518661,
+ 55.52020904657752
+ ],
+ [
+ 10.474375845616782,
+ 55.515814504422785
+ ],
+ [
+ 10.489105663715645,
+ 55.509100638554116
+ ],
+ [
+ 10.49122155113084,
+ 55.50311922071705
+ ],
+ [
+ 10.486013217039657,
+ 55.49477773644582
+ ],
+ [
+ 10.477793817074064,
+ 55.492499103097465
+ ],
+ [
+ 10.46778404890904,
+ 55.49213286769851
+ ],
+ [
+ 10.457041863371053,
+ 55.489447337187244
+ ],
+ [
+ 10.419606966568162,
+ 55.45880767910378
+ ],
+ [
+ 10.436045769024572,
+ 55.442124742277294
+ ],
+ [
+ 10.4715275405342,
+ 55.443060620027985
+ ],
+ [
+ 10.491221551701535,
+ 55.46527743015443
+ ],
+ [
+ 10.502452018936758,
+ 55.46356842628779
+ ],
+ [
+ 10.566905142637065,
+ 55.4826520697724
+ ],
+ [
+ 10.604746941376291,
+ 55.489447337187244
+ ],
+ [
+ 10.609141471613277,
+ 55.492580466696914
+ ],
+ [
+ 10.60718834784971,
+ 55.499416417451805
+ ],
+ [
+ 10.602305535499237,
+ 55.50625235155454
+ ],
+ [
+ 10.597666862828552,
+ 55.50934479209061
+ ],
+ [
+ 10.588715039622256,
+ 55.51068756279093
+ ],
+ [
+ 10.579112174824804,
+ 55.514105534045584
+ ],
+ [
+ 10.571055536306217,
+ 55.51870353257262
+ ],
+ [
+ 10.566905142333656,
+ 55.52362700140266
+ ],
+ [
+ 10.569834833695124,
+ 55.53241610190702
+ ],
+ [
+ 10.58228600437224,
+ 55.52912018814671
+ ],
+ [
+ 10.595550977079057,
+ 55.521389065879134
+ ],
+ [
+ 10.601084831522392,
+ 55.51679107768524
+ ],
+ [
+ 10.613780144582275,
+ 55.527899487467856
+ ],
+ [
+ 10.6105249359641,
+ 55.54364655160074
+ ],
+ [
+ 10.60084069266351,
+ 55.561224695486196
+ ],
+ [
+ 10.594248895412154,
+ 55.57827384569545
+ ],
+ [
+ 10.599131708121279,
+ 55.57367586045816
+ ],
+ [
+ 10.614593946762222,
+ 55.56460196345832
+ ],
+ [
+ 10.615000847911244,
+ 55.58144766810706
+ ],
+ [
+ 10.608409050495453,
+ 55.606268627494764
+ ],
+ [
+ 10.608571811966439,
+ 55.61298249971747
+ ],
+ [
+ 10.627614780144961,
+ 55.613592842604334
+ ],
+ [
+ 10.65691165392684,
+ 55.59487538114644
+ ],
+ [
+ 10.704112176546262,
+ 55.55027905602217
+ ],
+ [
+ 10.706065298257984,
+ 55.543890672821085
+ ],
+ [
+ 10.707041862042445,
+ 55.534979548455794
+ ],
+ [
+ 10.709239128368214,
+ 55.52704498204273
+ ],
+ [
+ 10.71436608031039,
+ 55.52362700140266
+ ],
+ [
+ 10.716807489745976,
+ 55.521429774212734
+ ],
+ [
+ 10.738129101826269,
+ 55.50934479209061
+ ],
+ [
+ 10.744476758413478,
+ 55.49555084155718
+ ],
+ [
+ 10.742360873809709,
+ 55.48749421689665
+ ],
+ [
+ 10.73308353070581,
+ 55.48187898076292
+ ],
+ [
+ 10.703949416511954,
+ 55.47113680267958
+ ],
+ [
+ 10.657481316302636,
+ 55.45970287035339
+ ],
+ [
+ 10.588389520315193,
+ 55.46336499066999
+ ],
+ [
+ 10.571299675682166,
+ 55.461249098621295
+ ],
+ [
+ 10.559418165934762,
+ 55.455308345457375
+ ],
+ [
+ 10.554047071211933,
+ 55.44257233537236
+ ],
+ [
+ 10.563324414964208,
+ 55.43622467659879
+ ],
+ [
+ 10.579437696984174,
+ 55.43585847699492
+ ],
+ [
+ 10.59424889403775,
+ 55.4410667994202
+ ],
+ [
+ 10.58961022030816,
+ 55.446234423120316
+ ],
+ [
+ 10.588552281130914,
+ 55.4494082826193
+ ],
+ [
+ 10.595957878921256,
+ 55.45966217612125
+ ],
+ [
+ 10.61573327048357,
+ 55.456854575508444
+ ],
+ [
+ 10.682953320357562,
+ 55.44790272889342
+ ],
+ [
+ 10.697927280406107,
+ 55.44025300148607
+ ],
+ [
+ 10.796641473755175,
+ 55.358547287253366
+ ],
+ [
+ 10.803070509728627,
+ 55.34975821298953
+ ],
+ [
+ 10.828379754095794,
+ 55.30658600604835
+ ],
+ [
+ 10.834320509038998,
+ 55.290269273426425
+ ],
+ [
+ 10.826914910038992,
+ 55.290269273426425
+ ],
+ [
+ 10.81511477917325,
+ 55.30084869093182
+ ],
+ [
+ 10.79688561251295,
+ 55.30768463164912
+ ],
+ [
+ 10.78028404917265,
+ 55.30491770089634
+ ],
+ [
+ 10.772959831725215,
+ 55.28709544197238
+ ],
+ [
+ 10.778656446705163,
+ 55.274847730764975
+ ],
+ [
+ 10.805349155799307,
+ 55.25043366978602
+ ],
+ [
+ 10.81332441511044,
+ 55.23558177421286
+ ],
+ [
+ 10.815196160178393,
+ 55.21405670996024
+ ],
+ [
+ 10.812673372377803,
+ 55.19428131015439
+ ],
+ [
+ 10.806000195150439,
+ 55.1773949146429
+ ],
+ [
+ 10.790782097261758,
+ 55.15692780488745
+ ],
+ [
+ 10.78785240970201,
+ 55.14948150870935
+ ],
+ [
+ 10.786631706615385,
+ 55.129461976750285
+ ],
+ [
+ 10.783376497889787,
+ 55.12421295678142
+ ],
+ [
+ 10.758799674638372,
+ 55.10529205499124
+ ],
+ [
+ 10.749359570941726,
+ 55.093085027352714
+ ],
+ [
+ 10.742686393808503,
+ 55.08169179687228
+ ],
+ [
+ 10.733653190783334,
+ 55.07184479159011
+ ],
+ [
+ 10.717784050139826,
+ 55.064276435556344
+ ],
+ [
+ 10.705577018990457,
+ 55.06256744989268
+ ],
+ [
+ 10.66627037913914,
+ 55.064276435556344
+ ]
+ ]
+ ]
+ ]
+ },
+ "type": "Feature",
+ "properties": {
+ "source": "https://simplemaps.com",
+ "id": "DK83",
+ "name": "Syddanmark"
+ },
+ "id": 1
+ },
+ {
+ "geometry": {
+ "type": "MultiPolygon",
+ "coordinates": [
+ [
+ [
+ [
+ 9.675520593530397,
+ 55.70859818519244
+ ],
+ [
+ 9.675564814682227,
+ 55.70871880301719
+ ],
+ [
+ 9.679698927323647,
+ 55.71964835017899
+ ],
+ [
+ 9.680629104828782,
+ 55.72848503775071
+ ],
+ [
+ 9.675513138155502,
+ 55.73706332971859
+ ],
+ [
+ 9.663989289291997,
+ 55.7429543765599
+ ],
+ [
+ 9.641665073962033,
+ 55.74693350975045
+ ],
+ [
+ 9.608747184175755,
+ 55.74946561617856
+ ],
+ [
+ 9.597895136632635,
+ 55.7533930762072
+ ],
+ [
+ 9.596344843368806,
+ 55.76083447556116
+ ],
+ [
+ 9.598360225006857,
+ 55.76690646748059
+ ],
+ [
+ 9.602546012833916,
+ 55.77266835042036
+ ],
+ [
+ 9.605491573785457,
+ 55.77886956315117
+ ],
+ [
+ 9.600685661506827,
+ 55.78452809944401
+ ],
+ [
+ 9.581617059133832,
+ 55.78682771060045
+ ],
+ [
+ 9.546218706457406,
+ 55.78450230435875
+ ],
+ [
+ 9.530405715603825,
+ 55.78755118193151
+ ],
+ [
+ 9.516453079968407,
+ 55.797576426697766
+ ],
+ [
+ 9.491131627334669,
+ 55.82170931107715
+ ],
+ [
+ 9.468807407800863,
+ 55.83685050143775
+ ],
+ [
+ 9.462916296440518,
+ 55.84610060767762
+ ],
+ [
+ 9.459660681670151,
+ 55.853309463079334
+ ],
+ [
+ 9.458265416105375,
+ 55.85845124894569
+ ],
+ [
+ 9.450668982651276,
+ 55.865789329469074
+ ],
+ [
+ 9.399922725675944,
+ 55.90165276715717
+ ],
+ [
+ 9.374756301344647,
+ 55.91436514245439
+ ],
+ [
+ 9.343130320788484,
+ 55.92547553445804
+ ],
+ [
+ 9.249389277637226,
+ 55.94299390050005
+ ],
+ [
+ 9.211665482605554,
+ 55.93834303653782
+ ],
+ [
+ 9.184380325599165,
+ 55.907104616236445
+ ],
+ [
+ 9.137974890102475,
+ 55.873902523732745
+ ],
+ [
+ 9.12107669118206,
+ 55.86604763999335
+ ],
+ [
+ 9.10288658789021,
+ 55.86320544626469
+ ],
+ [
+ 9.093894893636694,
+ 55.859329773001974
+ ],
+ [
+ 9.085058218700397,
+ 55.853541949899
+ ],
+ [
+ 9.079787223967422,
+ 55.8482709664429
+ ],
+ [
+ 9.070950553214702,
+ 55.835713584575274
+ ],
+ [
+ 9.058083123524849,
+ 55.836617956922545
+ ],
+ [
+ 9.047334428007561,
+ 55.830907732000924
+ ],
+ [
+ 9.013072949743426,
+ 55.8287889521982
+ ],
+ [
+ 8.999275340451677,
+ 55.834731729562215
+ ],
+ [
+ 8.959381140724814,
+ 55.86400649390912
+ ],
+ [
+ 8.929098749185627,
+ 55.88000033168568
+ ],
+ [
+ 8.880057814373021,
+ 55.882584143794276
+ ],
+ [
+ 8.843470898212058,
+ 55.87230050114686
+ ],
+ [
+ 8.828278026543742,
+ 55.84284492031668
+ ],
+ [
+ 8.817684358139942,
+ 55.833749880058846
+ ],
+ [
+ 8.82295535353463,
+ 55.81680000457975
+ ],
+ [
+ 8.823110387699552,
+ 55.81034049952672
+ ],
+ [
+ 8.80698733725518,
+ 55.80455269904055
+ ],
+ [
+ 8.774586213717082,
+ 55.7988166042023
+ ],
+ [
+ 8.757481315770384,
+ 55.80178801830286
+ ],
+ [
+ 8.687201370536842,
+ 55.82995170572608
+ ],
+ [
+ 8.654593538502695,
+ 55.83478341441331
+ ],
+ [
+ 8.64405154841968,
+ 55.833904933787856
+ ],
+ [
+ 8.62761844153198,
+ 55.82979665549971
+ ],
+ [
+ 8.610875277338248,
+ 55.827936303139445
+ ],
+ [
+ 8.598886348943562,
+ 55.829202417393134
+ ],
+ [
+ 8.588499385666147,
+ 55.831476173896185
+ ],
+ [
+ 8.578732535573415,
+ 55.83524849887616
+ ],
+ [
+ 8.569585809065101,
+ 55.83540355393319
+ ],
+ [
+ 8.560749141086577,
+ 55.83411167230695
+ ],
+ [
+ 8.55361779392601,
+ 55.831527857785694
+ ],
+ [
+ 8.548501822773193,
+ 55.82868558605206
+ ],
+ [
+ 8.54912194349245,
+ 55.82545586408289
+ ],
+ [
+ 8.554237906052201,
+ 55.81912543998764
+ ],
+ [
+ 8.555943231512725,
+ 55.81607657225882
+ ],
+ [
+ 8.555478140129413,
+ 55.81349270440006
+ ],
+ [
+ 8.553307732933744,
+ 55.811296498605216
+ ],
+ [
+ 8.503026561730286,
+ 55.7792829105895
+ ],
+ [
+ 8.490159131276158,
+ 55.77147978437919
+ ],
+ [
+ 8.471969027931252,
+ 55.76941274952962
+ ],
+ [
+ 8.459721712590886,
+ 55.77106637546227
+ ],
+ [
+ 8.42447838826987,
+ 55.78375297076611
+ ],
+ [
+ 8.39920861463416,
+ 55.7901091909284
+ ],
+ [
+ 8.368306106083875,
+ 55.80253734910015
+ ],
+ [
+ 8.175811394924775,
+ 55.80147796784839
+ ],
+ [
+ 8.17560707357738,
+ 55.80147645034364
+ ],
+ [
+ 8.135590041469221,
+ 55.96600985405177
+ ],
+ [
+ 8.13070722865471,
+ 55.97614169271765
+ ],
+ [
+ 8.12777753920905,
+ 55.97821685265813
+ ],
+ [
+ 8.126800976441542,
+ 55.98297759799166
+ ],
+ [
+ 8.127777541827136,
+ 55.98761630419096
+ ],
+ [
+ 8.130707225214998,
+ 55.989732138421765
+ ],
+ [
+ 8.136729364638711,
+ 55.98859286446319
+ ],
+ [
+ 8.139414911500449,
+ 55.98602938547778
+ ],
+ [
+ 8.141123892420685,
+ 55.98346586439417
+ ],
+ [
+ 8.14405358106405,
+ 55.982326551607095
+ ],
+ [
+ 8.150564000043884,
+ 55.977036880224276
+ ],
+ [
+ 8.175466343170756,
+ 55.90033602172184
+ ],
+ [
+ 8.193369988373481,
+ 55.87384675232547
+ ],
+ [
+ 8.195974154880979,
+ 55.862779037301955
+ ],
+ [
+ 8.193614129273355,
+ 55.853216867899455
+ ],
+ [
+ 8.183848503707283,
+ 55.83950429481554
+ ],
+ [
+ 8.181651237176492,
+ 55.828355198223186
+ ],
+ [
+ 8.18506920778882,
+ 55.814642656277755
+ ],
+ [
+ 8.193614127332868,
+ 55.81183500418458
+ ],
+ [
+ 8.238536001515204,
+ 55.82607652971108
+ ],
+ [
+ 8.286306187323277,
+ 55.84747957767322
+ ],
+ [
+ 8.308848506627879,
+ 55.85199619797794
+ ],
+ [
+ 8.347829621677539,
+ 55.86933012648509
+ ],
+ [
+ 8.38770592589628,
+ 55.89288972847932
+ ],
+ [
+ 8.39332116068166,
+ 55.909165766480484
+ ],
+ [
+ 8.38689212581569,
+ 55.926499781176844
+ ],
+ [
+ 8.372894726673353,
+ 55.943426820448494
+ ],
+ [
+ 8.356944207345775,
+ 55.9586856188232
+ ],
+ [
+ 8.319509312533606,
+ 55.98602938547778
+ ],
+ [
+ 8.319509310435356,
+ 56.01703521509781
+ ],
+ [
+ 8.308116080454482,
+ 56.05292387336136
+ ],
+ [
+ 8.28239993859334,
+ 56.07587315039778
+ ],
+ [
+ 8.247731967723352,
+ 56.08982983019081
+ ],
+ [
+ 8.17058353044788,
+ 56.10956452636605
+ ],
+ [
+ 8.149668817704294,
+ 56.1111514582867
+ ],
+ [
+ 8.14063561281474,
+ 56.10272858035521
+ ],
+ [
+ 8.141612174763925,
+ 56.06000396962665
+ ],
+ [
+ 8.139903189371493,
+ 56.039740278720366
+ ],
+ [
+ 8.133799674448005,
+ 56.01703521509781
+ ],
+ [
+ 8.137461782535219,
+ 56.00039293572009
+ ],
+ [
+ 8.134776240110654,
+ 55.993963964223944
+ ],
+ [
+ 8.120127799871216,
+ 55.996568099151304
+ ],
+ [
+ 8.115977409379386,
+ 56.00185780609272
+ ],
+ [
+ 8.11166425995508,
+ 56.01227449470862
+ ],
+ [
+ 8.107920767400937,
+ 56.02431872904197
+ ],
+ [
+ 8.103363476194632,
+ 56.09686919645187
+ ],
+ [
+ 8.106455927004376,
+ 56.119452244709215
+ ],
+ [
+ 8.130625844773292,
+ 56.181952184249475
+ ],
+ [
+ 8.133799676059224,
+ 56.20547111108664
+ ],
+ [
+ 8.122406450734253,
+ 56.551662568936344
+ ],
+ [
+ 8.12663822089232,
+ 56.568426879980585
+ ],
+ [
+ 8.133799679348174,
+ 56.58563880033836
+ ],
+ [
+ 8.17872155340174,
+ 56.6720645676193
+ ],
+ [
+ 8.198985226836095,
+ 56.6988793664857
+ ],
+ [
+ 8.213877804427952,
+ 56.711818813657985
+ ],
+ [
+ 8.219981318678856,
+ 56.70917393227664
+ ],
+ [
+ 8.223968949118222,
+ 56.70042556637944
+ ],
+ [
+ 8.233164912618985,
+ 56.695461366397815
+ ],
+ [
+ 8.23243248375862,
+ 56.68838119303595
+ ],
+ [
+ 8.209971544562185,
+ 56.65595116313434
+ ],
+ [
+ 8.202159047595972,
+ 56.64704007242005
+ ],
+ [
+ 8.2021590500083,
+ 56.640204169114746
+ ],
+ [
+ 8.208343948979776,
+ 56.63678624159911
+ ],
+ [
+ 8.21509849872238,
+ 56.63495507137631
+ ],
+ [
+ 8.233164915441412,
+ 56.63397865874161
+ ],
+ [
+ 8.236582880130427,
+ 56.62970613772931
+ ],
+ [
+ 8.240489133298649,
+ 56.62006262680262
+ ],
+ [
+ 8.245778843101515,
+ 56.610419027317825
+ ],
+ [
+ 8.253916862920942,
+ 56.60610585990039
+ ],
+ [
+ 8.29493248620512,
+ 56.60199613618065
+ ],
+ [
+ 8.305918821033613,
+ 56.59780514951501
+ ],
+ [
+ 8.291514519671043,
+ 56.593003657283404
+ ],
+ [
+ 8.295176630571296,
+ 56.58380770573496
+ ],
+ [
+ 8.292491083107722,
+ 56.56769441833343
+ ],
+ [
+ 8.298350455167522,
+ 56.55829496364317
+ ],
+ [
+ 8.30738366576653,
+ 56.55434812871056
+ ],
+ [
+ 8.315196163519667,
+ 56.55687088360297
+ ],
+ [
+ 8.318044469189315,
+ 56.56362542321678
+ ],
+ [
+ 8.31202232798222,
+ 56.57196680851626
+ ],
+ [
+ 8.312022333747704,
+ 56.578802826129255
+ ],
+ [
+ 8.343516469891366,
+ 56.584540076999275
+ ],
+ [
+ 8.3921004514014,
+ 56.58527245144984
+ ],
+ [
+ 8.436534048243395,
+ 56.58022693110437
+ ],
+ [
+ 8.455902534734788,
+ 56.568548823521304
+ ],
+ [
+ 8.465179884770581,
+ 56.56688061144914
+ ],
+ [
+ 8.51465905367859,
+ 56.54466385362288
+ ],
+ [
+ 8.528493687631835,
+ 56.54816317497519
+ ],
+ [
+ 8.541351758003476,
+ 56.55683013160224
+ ],
+ [
+ 8.552093949745739,
+ 56.56801997427812
+ ],
+ [
+ 8.559580926799763,
+ 56.578802826129255
+ ],
+ [
+ 8.565765822801042,
+ 56.578802826129255
+ ],
+ [
+ 8.596039261955584,
+ 56.53180577050534
+ ],
+ [
+ 8.596853059489398,
+ 56.524155971805776
+ ],
+ [
+ 8.593028193341276,
+ 56.51992425231421
+ ],
+ [
+ 8.594574417555165,
+ 56.51048415117733
+ ],
+ [
+ 8.601084833490955,
+ 56.50112541293175
+ ],
+ [
+ 8.617442250926125,
+ 56.49380113297416
+ ],
+ [
+ 8.636973500229224,
+ 56.47972234776898
+ ],
+ [
+ 8.648448107247043,
+ 56.475734691458456
+ ],
+ [
+ 8.665537951948185,
+ 56.47589746091975
+ ],
+ [
+ 8.726573114561422,
+ 56.48322176758303
+ ],
+ [
+ 8.73316490479291,
+ 56.48505273340931
+ ],
+ [
+ 8.73853600676175,
+ 56.49005772035075
+ ],
+ [
+ 8.742035355241322,
+ 56.497748156637826
+ ],
+ [
+ 8.743418814501615,
+ 56.5070661282958
+ ],
+ [
+ 8.743662962575444,
+ 56.52167396828435
+ ],
+ [
+ 8.744965041888491,
+ 56.53156161581221
+ ],
+ [
+ 8.750824417449953,
+ 56.552069435150976
+ ],
+ [
+ 8.755137568315167,
+ 56.55780673836247
+ ],
+ [
+ 8.760427284411445,
+ 56.56012610181757
+ ],
+ [
+ 8.76254316274512,
+ 56.563381222455426
+ ],
+ [
+ 8.757578967766854,
+ 56.57196680851626
+ ],
+ [
+ 8.726573117665046,
+ 56.58563880033836
+ ],
+ [
+ 8.716563346613768,
+ 56.58616770491466
+ ],
+ [
+ 8.70899499330235,
+ 56.588568496853576
+ ],
+ [
+ 8.704437690509964,
+ 56.59369530764334
+ ],
+ [
+ 8.702972851165326,
+ 56.60236236099911
+ ],
+ [
+ 8.698496946254393,
+ 56.6112328842656
+ ],
+ [
+ 8.68148847584155,
+ 56.61627837169969
+ ],
+ [
+ 8.681895374489326,
+ 56.62653223444577
+ ],
+ [
+ 8.695160357432876,
+ 56.633246232606275
+ ],
+ [
+ 8.740896024384549,
+ 56.63495507137631
+ ],
+ [
+ 8.80543053780102,
+ 56.695461366397815
+ ],
+ [
+ 8.82048586887807,
+ 56.70038477953351
+ ],
+ [
+ 8.861582884263688,
+ 56.7074649789459
+ ],
+ [
+ 8.916758656161551,
+ 56.7088076347582
+ ],
+ [
+ 8.918711785010679,
+ 56.71100495013594
+ ],
+ [
+ 8.894786002955312,
+ 56.71596912956676
+ ],
+ [
+ 8.851084837591616,
+ 56.71678301724762
+ ],
+ [
+ 8.84083092457538,
+ 56.72235748350683
+ ],
+ [
+ 8.850108273028225,
+ 56.74017980702219
+ ],
+ [
+ 8.864756707674243,
+ 56.7558454536888
+ ],
+ [
+ 8.876638219351777,
+ 56.76430901626274
+ ],
+ [
+ 8.908457873771225,
+ 56.777411132976084
+ ],
+ [
+ 8.955739786007904,
+ 56.80296466155356
+ ],
+ [
+ 8.974294470882054,
+ 56.80536535136014
+ ],
+ [
+ 8.985199421170517,
+ 56.804144676365254
+ ],
+ [
+ 9.0036727217887,
+ 56.79905833232055
+ ],
+ [
+ 9.014903196652808,
+ 56.797837703564795
+ ],
+ [
+ 9.02662193485363,
+ 56.79962791928897
+ ],
+ [
+ 9.032725452567554,
+ 56.80377833100231
+ ],
+ [
+ 9.037608267899401,
+ 56.808539116103795
+ ],
+ [
+ 9.045583532032277,
+ 56.81220126463207
+ ],
+ [
+ 9.063161656417934,
+ 56.812567467848716
+ ],
+ [
+ 9.086110868116132,
+ 56.80841699766937
+ ],
+ [
+ 9.1059676383431,
+ 56.80011620710157
+ ],
+ [
+ 9.114512561906691,
+ 56.787990575730376
+ ],
+ [
+ 9.120453322073976,
+ 56.76459383044985
+ ],
+ [
+ 9.134450717557533,
+ 56.75071849897378
+ ],
+ [
+ 9.150564001374276,
+ 56.74091221914551
+ ],
+ [
+ 9.162852406713741,
+ 56.72964097699567
+ ],
+ [
+ 9.1760360039933,
+ 56.7085635439155
+ ],
+ [
+ 9.148203978170283,
+ 56.70473875115853
+ ],
+ [
+ 9.118825722949117,
+ 56.678778461681915
+ ],
+ [
+ 9.093435088713615,
+ 56.6749941666245
+ ],
+ [
+ 9.093435086383348,
+ 56.66754784827368
+ ],
+ [
+ 9.097504105725617,
+ 56.658433381902476
+ ],
+ [
+ 9.086680532204243,
+ 56.649603547752214
+ ],
+ [
+ 9.05933679098743,
+ 56.63397865874161
+ ],
+ [
+ 9.049978067127956,
+ 56.61981850462772
+ ],
+ [
+ 9.048838737638105,
+ 56.60716379941246
+ ],
+ [
+ 9.052989131355975,
+ 56.5753848628679
+ ],
+ [
+ 9.060557486289383,
+ 56.56574125632476
+ ],
+ [
+ 9.076996288499137,
+ 56.567368863017855
+ ],
+ [
+ 9.093435087195209,
+ 56.57807031935969
+ ],
+ [
+ 9.106781449770896,
+ 56.61078527048138
+ ],
+ [
+ 9.1215926449409,
+ 56.613023190826624
+ ],
+ [
+ 9.13965905123838,
+ 56.611029379536134
+ ],
+ [
+ 9.15544681268776,
+ 56.61294182013723
+ ],
+ [
+ 9.149180539434546,
+ 56.62250402856251
+ ],
+ [
+ 9.14112389078222,
+ 56.62604397164926
+ ],
+ [
+ 9.131521030945358,
+ 56.62482332086677
+ ],
+ [
+ 9.121348500674664,
+ 56.620347356204164
+ ],
+ [
+ 9.128916862668182,
+ 56.63125234183311
+ ],
+ [
+ 9.13672935922895,
+ 56.63906479870432
+ ],
+ [
+ 9.143728062853125,
+ 56.64769117121754
+ ],
+ [
+ 9.148610869280976,
+ 56.661322286443976
+ ],
+ [
+ 9.15544680727819,
+ 56.661322286443976
+ ],
+ [
+ 9.149424677058171,
+ 56.63886142077581
+ ],
+ [
+ 9.175547716642097,
+ 56.63544336008088
+ ],
+ [
+ 9.211192256072056,
+ 56.63886142077581
+ ],
+ [
+ 9.233653188641572,
+ 56.637152381729756
+ ],
+ [
+ 9.247243691767405,
+ 56.63202552947973
+ ],
+ [
+ 9.262461791008482,
+ 56.63373451634153
+ ],
+ [
+ 9.276215039257542,
+ 56.632961320826276
+ ],
+ [
+ 9.28532961961485,
+ 56.620347356204164
+ ],
+ [
+ 9.282399942377218,
+ 56.610541161781065
+ ],
+ [
+ 9.25945071226655,
+ 56.5919456463518
+ ],
+ [
+ 9.250336135944972,
+ 56.578802826129255
+ ],
+ [
+ 9.261241083752829,
+ 56.576361416721014
+ ],
+ [
+ 9.26335696252233,
+ 56.57196680851626
+ ],
+ [
+ 9.264008014935795,
+ 56.56590411255629
+ ],
+ [
+ 9.270843943952746,
+ 56.55829496364317
+ ],
+ [
+ 9.280609571315143,
+ 56.5528018288988
+ ],
+ [
+ 9.305674679020324,
+ 56.54466385362288
+ ],
+ [
+ 9.309743687854509,
+ 56.54051343688258
+ ],
+ [
+ 9.312185087346867,
+ 56.535101572590754
+ ],
+ [
+ 9.316416861771847,
+ 56.530829153851734
+ ],
+ [
+ 9.326182484805742,
+ 56.530340847643835
+ ],
+ [
+ 9.340586786264598,
+ 56.547064535549005
+ ],
+ [
+ 9.346690302616796,
+ 56.552069435150976
+ ],
+ [
+ 9.360850457597001,
+ 56.54474519132456
+ ],
+ [
+ 9.373057491453492,
+ 56.552720487311284
+ ],
+ [
+ 9.372731970605795,
+ 56.56537511116512
+ ],
+ [
+ 9.350352405480313,
+ 56.57196680851626
+ ],
+ [
+ 9.319346545381228,
+ 56.56171290109638
+ ],
+ [
+ 9.31739342937094,
+ 56.55866125386369
+ ],
+ [
+ 9.312836131415686,
+ 56.55776598613549
+ ],
+ [
+ 9.30225670494582,
+ 56.55829496364317
+ ],
+ [
+ 9.293711784646893,
+ 56.56102122163668
+ ],
+ [
+ 9.293142126405458,
+ 56.56679926306699
+ ],
+ [
+ 9.296153190743057,
+ 56.571600649826415
+ ],
+ [
+ 9.298838733505216,
+ 56.57196680851626
+ ],
+ [
+ 9.30062910457032,
+ 56.60911696069075
+ ],
+ [
+ 9.317067908508559,
+ 56.641099394148235
+ ],
+ [
+ 9.326019727647243,
+ 56.66941967194571
+ ],
+ [
+ 9.397700635418184,
+ 56.64982411537245
+ ],
+ [
+ 9.435424432982122,
+ 56.63850699479308
+ ],
+ [
+ 9.459557321721157,
+ 56.661192890710765
+ ],
+ [
+ 9.450720657778541,
+ 56.67522311645477
+ ],
+ [
+ 9.458782181578247,
+ 56.68511915375927
+ ],
+ [
+ 9.487669308539491,
+ 56.68382726504193
+ ],
+ [
+ 9.502500439101365,
+ 56.6747321430792
+ ],
+ [
+ 9.49629926598918,
+ 56.65088345400259
+ ],
+ [
+ 9.524049518039673,
+ 56.63370112937423
+ ],
+ [
+ 9.637737657719962,
+ 56.633003375379715
+ ],
+ [
+ 9.657323037027822,
+ 56.62246149512716
+ ],
+ [
+ 9.631278109720947,
+ 56.60579579757894
+ ],
+ [
+ 9.6240950904214,
+ 56.583962556098186
+ ],
+ [
+ 9.643267044123556,
+ 56.58044853094458
+ ],
+ [
+ 9.675926551527683,
+ 56.59042210945879
+ ],
+ [
+ 9.725432571655704,
+ 56.55908033064203
+ ],
+ [
+ 9.769305861512892,
+ 56.565643221762734
+ ],
+ [
+ 9.796125929180937,
+ 56.55740082095066
+ ],
+ [
+ 9.813024118136479,
+ 56.55525621344174
+ ],
+ [
+ 9.854623654962744,
+ 56.558925342707504
+ ],
+ [
+ 9.887903272277963,
+ 56.580655207959175
+ ],
+ [
+ 9.927694124798018,
+ 56.59352263550065
+ ],
+ [
+ 9.952033720682651,
+ 56.597682542861435
+ ],
+ [
+ 10.011616651631176,
+ 56.60153249863015
+ ],
+ [
+ 10.080139604291302,
+ 56.62514866636879
+ ],
+ [
+ 10.11347090760862,
+ 56.65129698155062
+ ],
+ [
+ 10.159979690639924,
+ 56.672897668630384
+ ],
+ [
+ 10.193207225917762,
+ 56.69358957284981
+ ],
+ [
+ 10.24195396672329,
+ 56.68838119303595
+ ],
+ [
+ 10.283539252222727,
+ 56.689764640577515
+ ],
+ [
+ 10.310313353081414,
+ 56.6988793664857
+ ],
+ [
+ 10.334239130379848,
+ 56.70384350727417
+ ],
+ [
+ 10.352793821325553,
+ 56.681830203880615
+ ],
+ [
+ 10.357676625439932,
+ 56.648749051623284
+ ],
+ [
+ 10.340342640230258,
+ 56.620347356204164
+ ],
+ [
+ 10.272959831999422,
+ 56.59170156499268
+ ],
+ [
+ 10.223317905785061,
+ 56.560614333816446
+ ],
+ [
+ 10.214366079744398,
+ 56.55829496364317
+ ],
+ [
+ 10.207286009551435,
+ 56.52578365849927
+ ],
+ [
+ 10.20386803460589,
+ 56.516669007559564
+ ],
+ [
+ 10.211436387802069,
+ 56.50531640321249
+ ],
+ [
+ 10.20679771512625,
+ 56.490952789539634
+ ],
+ [
+ 10.195485876750483,
+ 56.47768793532434
+ ],
+ [
+ 10.183360223469702,
+ 56.469549888608064
+ ],
+ [
+ 10.214121937067972,
+ 56.470445009698246
+ ],
+ [
+ 10.225596543197911,
+ 56.475734691458456
+ ],
+ [
+ 10.224375843213199,
+ 56.5136172124239
+ ],
+ [
+ 10.227549677979344,
+ 56.54108310846218
+ ],
+ [
+ 10.237640819277013,
+ 56.554388719672744
+ ],
+ [
+ 10.279551624031155,
+ 56.57196680851626
+ ],
+ [
+ 10.312673379960419,
+ 56.596625144422596
+ ],
+ [
+ 10.33033288604101,
+ 56.60321694217178
+ ],
+ [
+ 10.350922077818426,
+ 56.59589272056415
+ ],
+ [
+ 10.354502797988411,
+ 56.589544966709816
+ ],
+ [
+ 10.359873895737316,
+ 56.568101323249664
+ ],
+ [
+ 10.360850451865026,
+ 56.56171290109638
+ ],
+ [
+ 10.365244988116974,
+ 56.55841706029491
+ ],
+ [
+ 10.394867388490852,
+ 56.54466385362288
+ ],
+ [
+ 10.469899938085563,
+ 56.52090080589351
+ ],
+ [
+ 10.551524288002149,
+ 56.51557041270423
+ ],
+ [
+ 10.7890731072545,
+ 56.53636296181861
+ ],
+ [
+ 10.822276241715347,
+ 56.53424717644245
+ ],
+ [
+ 10.854991080092631,
+ 56.524155971805776
+ ],
+ [
+ 10.869883666428247,
+ 56.514715954978314
+ ],
+ [
+ 10.902110223950665,
+ 56.48322176758303
+ ],
+ [
+ 10.964121942817739,
+ 56.448431727088696
+ ],
+ [
+ 10.964121936444675,
+ 56.44220607415228
+ ],
+ [
+ 10.94092858163071,
+ 56.4295108051109
+ ],
+ [
+ 10.928477417391218,
+ 56.40648028795688
+ ],
+ [
+ 10.923513211340763,
+ 56.37986882224197
+ ],
+ [
+ 10.922618033382731,
+ 56.35626860491031
+ ],
+ [
+ 10.911631704165607,
+ 56.3340517950024
+ ],
+ [
+ 10.819590692574854,
+ 56.260931724839246
+ ],
+ [
+ 10.772959830036868,
+ 56.24298735090657
+ ],
+ [
+ 10.755381709754587,
+ 56.22931552665423
+ ],
+ [
+ 10.750010614168248,
+ 56.216050535586675
+ ],
+ [
+ 10.752614781991603,
+ 56.1849633044705
+ ],
+ [
+ 10.738617382364483,
+ 56.15452706733051
+ ],
+ [
+ 10.707041861396508,
+ 56.15086495862358
+ ],
+ [
+ 10.672211132208702,
+ 56.165187873532496
+ ],
+ [
+ 10.648936396170745,
+ 56.188381275627066
+ ],
+ [
+ 10.674652538819684,
+ 56.190822645183665
+ ],
+ [
+ 10.68523196885155,
+ 56.19623446208058
+ ],
+ [
+ 10.69044030232542,
+ 56.208889090199065
+ ],
+ [
+ 10.687673374160049,
+ 56.222235431569636
+ ],
+ [
+ 10.677582226989701,
+ 56.22630442888756
+ ],
+ [
+ 10.664561395217719,
+ 56.22760652522093
+ ],
+ [
+ 10.65259850718061,
+ 56.232733500568024
+ ],
+ [
+ 10.628591342609603,
+ 56.23753490463916
+ ],
+ [
+ 10.603770379290923,
+ 56.221014719176665
+ ],
+ [
+ 10.581228063378923,
+ 56.199204846051884
+ ],
+ [
+ 10.54916425926488,
+ 56.17820872891046
+ ],
+ [
+ 10.556162953399571,
+ 56.15493394643114
+ ],
+ [
+ 10.569346551712718,
+ 56.12946199981034
+ ],
+ [
+ 10.57374108462783,
+ 56.11261630686188
+ ],
+ [
+ 10.563975457235633,
+ 56.10561758259007
+ ],
+ [
+ 10.548024935703118,
+ 56.101263735729816
+ ],
+ [
+ 10.53101647058773,
+ 56.10016509041951
+ ],
+ [
+ 10.518565299760608,
+ 56.10272858035521
+ ],
+ [
+ 10.508474154770525,
+ 56.11102936146334
+ ],
+ [
+ 10.498545770566647,
+ 56.13019442032174
+ ],
+ [
+ 10.491221547884221,
+ 56.13996000459415
+ ],
+ [
+ 10.49919681404505,
+ 56.144720803654444
+ ],
+ [
+ 10.516123893287398,
+ 56.15159739013722
+ ],
+ [
+ 10.525401239604937,
+ 56.1536319206892
+ ],
+ [
+ 10.51823977802126,
+ 56.16478097018065
+ ],
+ [
+ 10.509450714583782,
+ 56.17011136833807
+ ],
+ [
+ 10.46810957178726,
+ 56.17706941171897
+ ],
+ [
+ 10.458994986631973,
+ 56.17609282583332
+ ],
+ [
+ 10.436696814314514,
+ 56.167914166830116
+ ],
+ [
+ 10.408702018531724,
+ 56.1648623667812
+ ],
+ [
+ 10.39144941382995,
+ 56.172674858942784
+ ],
+ [
+ 10.376231317212914,
+ 56.18667228154888
+ ],
+ [
+ 10.354665560373139,
+ 56.20205311998487
+ ],
+ [
+ 10.376963738445669,
+ 56.20331452497532
+ ],
+ [
+ 10.410411001249493,
+ 56.21849189839832
+ ],
+ [
+ 10.429209830565123,
+ 56.22247954306578
+ ],
+ [
+ 10.443369986057245,
+ 56.21662016245215
+ ],
+ [
+ 10.460703973550354,
+ 56.20620353517001
+ ],
+ [
+ 10.477793815274604,
+ 56.20233794997725
+ ],
+ [
+ 10.491221547670495,
+ 56.21629463828026
+ ],
+ [
+ 10.474294465611797,
+ 56.23224516865464
+ ],
+ [
+ 10.475433789057949,
+ 56.2446963455906
+ ],
+ [
+ 10.50473066378589,
+ 56.27777739763327
+ ],
+ [
+ 10.491221550133046,
+ 56.27765534246566
+ ],
+ [
+ 10.476817256555302,
+ 56.280910577202874
+ ],
+ [
+ 10.463633658741013,
+ 56.286851290087824
+ ],
+ [
+ 10.45362389553505,
+ 56.29450106097364
+ ],
+ [
+ 10.442637562441844,
+ 56.29865139461579
+ ],
+ [
+ 10.416758662202446,
+ 56.29189690144159
+ ],
+ [
+ 10.403005401553143,
+ 56.29824449470244
+ ],
+ [
+ 10.399261913045468,
+ 56.28485746726773
+ ],
+ [
+ 10.39380943982506,
+ 56.27651600492583
+ ],
+ [
+ 10.385020376505732,
+ 56.272162149219426
+ ],
+ [
+ 10.371104363750087,
+ 56.270941481364545
+ ],
+ [
+ 10.363780144896786,
+ 56.265570390009024
+ ],
+ [
+ 10.33692467808637,
+ 56.232733500568024
+ ],
+ [
+ 10.312510610110426,
+ 56.215155308543174
+ ],
+ [
+ 10.258474157378325,
+ 56.190253023870966
+ ],
+ [
+ 10.234222853864008,
+ 56.171332119119796
+ ],
+ [
+ 10.220225458008352,
+ 56.14789460944701
+ ],
+ [
+ 10.22689863443663,
+ 56.133246165955015
+ ],
+ [
+ 10.243418814366441,
+ 56.11957426748165
+ ],
+ [
+ 10.259043818326077,
+ 56.09902580934033
+ ],
+ [
+ 10.2656356101426,
+ 56.07371650664059
+ ],
+ [
+ 10.261485221238654,
+ 56.05442942036943
+ ],
+ [
+ 10.249196810122218,
+ 56.036932674033764
+ ],
+ [
+ 10.2311304043056,
+ 56.01703521509781
+ ],
+ [
+ 10.255137562685965,
+ 56.02362698238252
+ ],
+ [
+ 10.272715692250298,
+ 56.02484773716607
+ ],
+ [
+ 10.277517122476887,
+ 56.01776763907378
+ ],
+ [
+ 10.252452016499527,
+ 55.9862327865821
+ ],
+ [
+ 10.250010616155409,
+ 55.973456157879525
+ ],
+ [
+ 10.251638217451957,
+ 55.948187572141286
+ ],
+ [
+ 10.250743034523483,
+ 55.93073150657497
+ ],
+ [
+ 10.249278192880714,
+ 55.92304110441422
+ ],
+ [
+ 10.245371942682018,
+ 55.914618250163215
+ ],
+ [
+ 10.221364782105091,
+ 55.89691806005135
+ ],
+ [
+ 10.217539911543009,
+ 55.88983797563868
+ ],
+ [
+ 10.21265709711756,
+ 55.88987864934358
+ ],
+ [
+ 10.20183352784075,
+ 55.88589099327467
+ ],
+ [
+ 10.190196160125884,
+ 55.879706123442425
+ ],
+ [
+ 10.183360220740383,
+ 55.873724662544774
+ ],
+ [
+ 10.188324414683688,
+ 55.87116119637112
+ ],
+ [
+ 10.195567256022578,
+ 55.86131421818802
+ ],
+ [
+ 10.199392125101404,
+ 55.85073479208969
+ ],
+ [
+ 10.19060306115059,
+ 55.843491930727666
+ ],
+ [
+ 10.187347850030916,
+ 55.8386090943921
+ ],
+ [
+ 10.182139522466624,
+ 55.83368565382492
+ ],
+ [
+ 10.17269941221861,
+ 55.83148831601413
+ ],
+ [
+ 10.160411002783954,
+ 55.83331938899874
+ ],
+ [
+ 10.151052279933396,
+ 55.838120834232086
+ ],
+ [
+ 10.135508663205592,
+ 55.85199619797794
+ ],
+ [
+ 10.131032746498333,
+ 55.85846586967185
+ ],
+ [
+ 10.128103059151174,
+ 55.86587146466397
+ ],
+ [
+ 10.123789907315587,
+ 55.873277054013585
+ ],
+ [
+ 10.11451256658686,
+ 55.87986888877809
+ ],
+ [
+ 10.101817252446311,
+ 55.88239165103053
+ ],
+ [
+ 10.004649285580486,
+ 55.87986888877809
+ ],
+ [
+ 9.991709830895351,
+ 55.87734608308122
+ ],
+ [
+ 9.983653193685472,
+ 55.87140537243335
+ ],
+ [
+ 9.977386912242672,
+ 55.86456937069787
+ ],
+ [
+ 9.970469594876024,
+ 55.85944242509661
+ ],
+ [
+ 9.944509312659145,
+ 55.85358309450582
+ ],
+ [
+ 9.891774933764882,
+ 55.85537341358272
+ ],
+ [
+ 9.867442257120807,
+ 55.85199619797794
+ ],
+ [
+ 9.892588741269153,
+ 55.836249129653034
+ ],
+ [
+ 10.035655145253743,
+ 55.81842683961854
+ ],
+ [
+ 10.045746292487912,
+ 55.813950942984214
+ ],
+ [
+ 10.042165559380752,
+ 55.80341215812287
+ ],
+ [
+ 10.025645376136858,
+ 55.7830264005488
+ ],
+ [
+ 10.017344596426073,
+ 55.76557037328519
+ ],
+ [
+ 10.019786005212314,
+ 55.75787996418022
+ ],
+ [
+ 10.03370201774753,
+ 55.7560488803678
+ ],
+ [
+ 10.059825069219633,
+ 55.75641514351241
+ ],
+ [
+ 10.059825066405011,
+ 55.749497793718824
+ ],
+ [
+ 10.018809438913099,
+ 55.73627348057128
+ ],
+ [
+ 10.006846550727818,
+ 55.72504304453098
+ ],
+ [
+ 10.018239779702247,
+ 55.708563540511534
+ ],
+ [
+ 10.01449628875622,
+ 55.70685454442307
+ ],
+ [
+ 10.00521894536491,
+ 55.70111724754837
+ ],
+ [
+ 9.981700063351516,
+ 55.70941800486073
+ ],
+ [
+ 9.871836781868152,
+ 55.69122952175045
+ ],
+ [
+ 9.821787958045313,
+ 55.67597078249979
+ ],
+ [
+ 9.792246939769703,
+ 55.67446522124108
+ ],
+ [
+ 9.730967646163029,
+ 55.688137137099474
+ ],
+ [
+ 9.695648634080827,
+ 55.70579661797767
+ ],
+ [
+ 9.686534049712154,
+ 55.708563540511534
+ ],
+ [
+ 9.675520593530397,
+ 55.70859818519244
+ ]
+ ]
+ ],
+ [
+ [
+ [
+ 10.559418164909685,
+ 55.86627838599967
+ ],
+ [
+ 10.573090039809683,
+ 55.87885162789616
+ ],
+ [
+ 10.58106530288328,
+ 55.894517351824156
+ ],
+ [
+ 10.580251495725154,
+ 55.91152575789926
+ ],
+ [
+ 10.566905146115145,
+ 55.9282901293565
+ ],
+ [
+ 10.548594598982612,
+ 55.93573641692674
+ ],
+ [
+ 10.528330921993893,
+ 55.94061917469174
+ ],
+ [
+ 10.515391470489508,
+ 55.951239308247004
+ ],
+ [
+ 10.51856530214067,
+ 55.97614169271765
+ ],
+ [
+ 10.528819203878422,
+ 55.9892031574574
+ ],
+ [
+ 10.543955923481946,
+ 55.99799223822899
+ ],
+ [
+ 10.556813997752535,
+ 55.99579498626887
+ ],
+ [
+ 10.564219599734619,
+ 55.953843522185075
+ ],
+ [
+ 10.580251497587632,
+ 55.936509502452424
+ ],
+ [
+ 10.602793813101835,
+ 55.92523831610057
+ ],
+ [
+ 10.628265823142186,
+ 55.92084384052331
+ ],
+ [
+ 10.619395376667798,
+ 55.910956096360565
+ ],
+ [
+ 10.615407749520154,
+ 55.90033602172184
+ ],
+ [
+ 10.616384314085106,
+ 55.8897973019432
+ ],
+ [
+ 10.621429884616246,
+ 55.87986888877809
+ ],
+ [
+ 10.63868248857638,
+ 55.86611563333127
+ ],
+ [
+ 10.650645377196785,
+ 55.8701032372131
+ ],
+ [
+ 10.659190299624619,
+ 55.880764060884594
+ ],
+ [
+ 10.66627038176776,
+ 55.88670485031125
+ ],
+ [
+ 10.669281447028816,
+ 55.88003165426549
+ ],
+ [
+ 10.655039910296452,
+ 55.86538320825103
+ ],
+ [
+ 10.628265822036479,
+ 55.84577058640335
+ ],
+ [
+ 10.62330162907934,
+ 55.82770416987296
+ ],
+ [
+ 10.624359568315947,
+ 55.79433825341711
+ ],
+ [
+ 10.614593948937907,
+ 55.77684157142379
+ ],
+ [
+ 10.58220462440506,
+ 55.75975171056095
+ ],
+ [
+ 10.547129754171236,
+ 55.76520417090167
+ ],
+ [
+ 10.520274288252132,
+ 55.78856032520235
+ ],
+ [
+ 10.511566604981441,
+ 55.82526279518501
+ ],
+ [
+ 10.517263218093392,
+ 55.84552644615681
+ ],
+ [
+ 10.527679882257631,
+ 55.85390857664946
+ ],
+ [
+ 10.54216556208591,
+ 55.85814037904115
+ ],
+ [
+ 10.559418164909685,
+ 55.86627838599967
+ ]
+ ]
+ ],
+ [
+ [
+ [
+ 11.512217647662029,
+ 56.695461366397815
+ ],
+ [
+ 11.509532097396018,
+ 56.710150461934816
+ ],
+ [
+ 11.520843948609365,
+ 56.71539950690267
+ ],
+ [
+ 11.606130397217369,
+ 56.72040428731851
+ ],
+ [
+ 11.649424670821913,
+ 56.72964097699567
+ ],
+ [
+ 11.649424671478307,
+ 56.722154004430884
+ ],
+ [
+ 11.645192909043846,
+ 56.72199131770778
+ ],
+ [
+ 11.64193770119654,
+ 56.720648556042256
+ ],
+ [
+ 11.635752798633375,
+ 56.71596912956676
+ ],
+ [
+ 11.618011914182572,
+ 56.711330462951814
+ ],
+ [
+ 11.604258662022403,
+ 56.70115796493851
+ ],
+ [
+ 11.592539905192151,
+ 56.68984604853613
+ ],
+ [
+ 11.581065296019627,
+ 56.68121975465745
+ ],
+ [
+ 11.564219592051636,
+ 56.6774355650139
+ ],
+ [
+ 11.543711781021791,
+ 56.67865627754905
+ ],
+ [
+ 11.524587440795575,
+ 56.684719190626815
+ ],
+ [
+ 11.512217647662029,
+ 56.695461366397815
+ ]
+ ]
+ ]
+ ]
+ },
+ "type": "Feature",
+ "properties": {
+ "source": "https://simplemaps.com",
+ "id": "DK82",
+ "name": "Midtjylland"
+ },
+ "id": 2
+ },
+ {
+ "geometry": {
+ "type": "MultiPolygon",
+ "coordinates": [
+ [
+ [
+ [
+ 10.193207225917762,
+ 56.69358957284981
+ ],
+ [
+ 10.159979690639924,
+ 56.672897668630384
+ ],
+ [
+ 10.11347090760862,
+ 56.65129698155062
+ ],
+ [
+ 10.080139604291302,
+ 56.62514866636879
+ ],
+ [
+ 10.011616651631176,
+ 56.60153249863015
+ ],
+ [
+ 9.952033720682651,
+ 56.597682542861435
+ ],
+ [
+ 9.927694124798018,
+ 56.59352263550065
+ ],
+ [
+ 9.887903272277963,
+ 56.580655207959175
+ ],
+ [
+ 9.854623654962744,
+ 56.558925342707504
+ ],
+ [
+ 9.813024118136479,
+ 56.55525621344174
+ ],
+ [
+ 9.796125929180937,
+ 56.55740082095066
+ ],
+ [
+ 9.769305861512892,
+ 56.565643221762734
+ ],
+ [
+ 9.725432571655704,
+ 56.55908033064203
+ ],
+ [
+ 9.675926551527683,
+ 56.59042210945879
+ ],
+ [
+ 9.643267044123556,
+ 56.58044853094458
+ ],
+ [
+ 9.6240950904214,
+ 56.583962556098186
+ ],
+ [
+ 9.631278109720947,
+ 56.60579579757894
+ ],
+ [
+ 9.657323037027822,
+ 56.62246149512716
+ ],
+ [
+ 9.637737657719962,
+ 56.633003375379715
+ ],
+ [
+ 9.524049518039673,
+ 56.63370112937423
+ ],
+ [
+ 9.49629926598918,
+ 56.65088345400259
+ ],
+ [
+ 9.502500439101365,
+ 56.6747321430792
+ ],
+ [
+ 9.487669308539491,
+ 56.68382726504193
+ ],
+ [
+ 9.458782181578247,
+ 56.68511915375927
+ ],
+ [
+ 9.450720657778541,
+ 56.67522311645477
+ ],
+ [
+ 9.459557321721157,
+ 56.661192890710765
+ ],
+ [
+ 9.435424432982122,
+ 56.63850699479308
+ ],
+ [
+ 9.397700635418184,
+ 56.64982411537245
+ ],
+ [
+ 9.326019727647243,
+ 56.66941967194571
+ ],
+ [
+ 9.325938344795059,
+ 56.669623086927594
+ ],
+ [
+ 9.30567467796015,
+ 56.695461366397815
+ ],
+ [
+ 9.295909052133293,
+ 56.701564872254046
+ ],
+ [
+ 9.290700718240487,
+ 56.70384350727417
+ ],
+ [
+ 9.270843941505182,
+ 56.70233790253155
+ ],
+ [
+ 9.230642123673102,
+ 56.692084060344044
+ ],
+ [
+ 9.220713737140725,
+ 56.68427154133272
+ ],
+ [
+ 9.19906659923478,
+ 56.676581087586726
+ ],
+ [
+ 9.177907742865939,
+ 56.67584863595898
+ ],
+ [
+ 9.169118682428925,
+ 56.68866603842685
+ ],
+ [
+ 9.180349156172545,
+ 56.70075105861235
+ ],
+ [
+ 9.226410355946934,
+ 56.716254021923106
+ ],
+ [
+ 9.236827022822274,
+ 56.733059033304265
+ ],
+ [
+ 9.242523636752741,
+ 56.74705641603474
+ ],
+ [
+ 9.240489123041355,
+ 56.75397362632163
+ ],
+ [
+ 9.215505405379606,
+ 56.75885651670541
+ ],
+ [
+ 9.205251495397771,
+ 56.76382067270825
+ ],
+ [
+ 9.196543822394405,
+ 56.77045327140939
+ ],
+ [
+ 9.189707873606142,
+ 56.777411132976084
+ ],
+ [
+ 9.184418163881741,
+ 56.784002983098276
+ ],
+ [
+ 9.17872155543238,
+ 56.79413490055297
+ ],
+ [
+ 9.174978064499332,
+ 56.806057076519906
+ ],
+ [
+ 9.176036004316222,
+ 56.81842682893014
+ ],
+ [
+ 9.181162951086751,
+ 56.821722649535694
+ ],
+ [
+ 9.204112170121496,
+ 56.84162012357209
+ ],
+ [
+ 9.206716339517602,
+ 56.84634020526394
+ ],
+ [
+ 9.21143639362597,
+ 56.851507875370906
+ ],
+ [
+ 9.210622587097715,
+ 56.863104498353664
+ ],
+ [
+ 9.205902536166544,
+ 56.874579121565155
+ ],
+ [
+ 9.199554885233635,
+ 56.87982820727238
+ ],
+ [
+ 9.187836134100765,
+ 56.88117096724837
+ ],
+ [
+ 9.178965689998103,
+ 56.884588921568565
+ ],
+ [
+ 9.16285241159137,
+ 56.89411043877302
+ ],
+ [
+ 9.181488472052713,
+ 56.91909408168849
+ ],
+ [
+ 9.20020593085657,
+ 56.93891029539453
+ ],
+ [
+ 9.220469592648225,
+ 56.95506414134463
+ ],
+ [
+ 9.251475460080915,
+ 56.97256097784063
+ ],
+ [
+ 9.258474158844786,
+ 56.97483963218505
+ ],
+ [
+ 9.265961134330604,
+ 56.97622305305322
+ ],
+ [
+ 9.274668819351598,
+ 56.976629991058175
+ ],
+ [
+ 9.279958533450221,
+ 56.979803820239475
+ ],
+ [
+ 9.285492381911617,
+ 56.99396390845612
+ ],
+ [
+ 9.28874759097369,
+ 56.99713774945202
+ ],
+ [
+ 9.437022326862003,
+ 57.02440007021194
+ ],
+ [
+ 9.470062699583565,
+ 57.02383051394371
+ ],
+ [
+ 9.586761913940776,
+ 56.99713774945202
+ ],
+ [
+ 9.586761914307125,
+ 56.98969146479006
+ ],
+ [
+ 9.573985227151065,
+ 56.985012172117095
+ ],
+ [
+ 9.573090034601545,
+ 56.97882715192989
+ ],
+ [
+ 9.58082116554944,
+ 56.97296790681679
+ ],
+ [
+ 9.59424889449745,
+ 56.96918366689119
+ ],
+ [
+ 9.607920772347711,
+ 56.9698754325948
+ ],
+ [
+ 9.617686399362848,
+ 56.97504309935661
+ ],
+ [
+ 9.654144722178824,
+ 57.006781260300976
+ ],
+ [
+ 9.670664905656531,
+ 57.02484767202911
+ ],
+ [
+ 9.688975452572684,
+ 57.03912989816185
+ ],
+ [
+ 9.728200718637952,
+ 57.04706453909633
+ ],
+ [
+ 9.761241079439936,
+ 57.058579790260886
+ ],
+ [
+ 9.809092639078795,
+ 57.051743825102506
+ ],
+ [
+ 9.915782094399411,
+ 57.058579790260886
+ ],
+ [
+ 9.856618690547672,
+ 57.08869054927573
+ ],
+ [
+ 9.826670765685359,
+ 57.094875354786474
+ ],
+ [
+ 9.80640709108457,
+ 57.10423404867845
+ ],
+ [
+ 9.795664914140982,
+ 57.1063907398754
+ ],
+ [
+ 9.787608269593742,
+ 57.10394929286969
+ ],
+ [
+ 9.752452021552083,
+ 57.07778557865701
+ ],
+ [
+ 9.733246293020736,
+ 57.066799256182435
+ ],
+ [
+ 9.711924672158709,
+ 57.06126533382857
+ ],
+ [
+ 9.693207222978161,
+ 57.06915919996528
+ ],
+ [
+ 9.67514082511577,
+ 57.07916906020953
+ ],
+ [
+ 9.656097851594938,
+ 57.07636139024477
+ ],
+ [
+ 9.621104360476695,
+ 57.058579790260886
+ ],
+ [
+ 9.580821165982902,
+ 57.04930254682269
+ ],
+ [
+ 9.49398847731529,
+ 57.04857005776567
+ ],
+ [
+ 9.265147333138623,
+ 57.00340405295286
+ ],
+ [
+ 9.245371946715599,
+ 57.00234616619543
+ ],
+ [
+ 9.22641035283153,
+ 57.00702546823144
+ ],
+ [
+ 9.114512563609562,
+ 57.058579790260886
+ ],
+ [
+ 9.12086021916716,
+ 57.045396190770155
+ ],
+ [
+ 9.112803585231902,
+ 57.0386417092328
+ ],
+ [
+ 9.07984459343319,
+ 57.031236025426516
+ ],
+ [
+ 9.05543053610782,
+ 57.0211449378836
+ ],
+ [
+ 9.042653837056765,
+ 57.01772689295164
+ ],
+ [
+ 9.025726762414081,
+ 57.01756423793354
+ ],
+ [
+ 9.031993038417813,
+ 57.02383051394371
+ ],
+ [
+ 9.014008008506512,
+ 57.030178120787184
+ ],
+ [
+ 8.996755399371759,
+ 57.02806217101133
+ ],
+ [
+ 8.979665556300953,
+ 57.020493822607065
+ ],
+ [
+ 8.963145375288958,
+ 57.01019923899576
+ ],
+ [
+ 8.94695071096266,
+ 57.011175771395216
+ ],
+ [
+ 8.926442906796579,
+ 57.00014885285384
+ ],
+ [
+ 8.909027544838914,
+ 56.98818600169403
+ ],
+ [
+ 8.901621935020133,
+ 56.986273428690076
+ ],
+ [
+ 8.893077024450848,
+ 56.999660618600494
+ ],
+ [
+ 8.873220252796639,
+ 57.004828253376544
+ ],
+ [
+ 8.850433792719473,
+ 57.00348545388797
+ ],
+ [
+ 8.833343945024032,
+ 56.99713774945202
+ ],
+ [
+ 8.847992379948721,
+ 56.98847071799051
+ ],
+ [
+ 8.81169681407025,
+ 56.9698754325948
+ ],
+ [
+ 8.762054878697858,
+ 56.95547102720816
+ ],
+ [
+ 8.726410350806923,
+ 56.96124907545893
+ ],
+ [
+ 8.702972852934105,
+ 56.958441485199685
+ ],
+ [
+ 8.67953535683694,
+ 56.952541471355794
+ ],
+ [
+ 8.668793162554593,
+ 56.945298537922476
+ ],
+ [
+ 8.663340686287057,
+ 56.9259300181098
+ ],
+ [
+ 8.649912956851706,
+ 56.914129948048014
+ ],
+ [
+ 8.63412520108447,
+ 56.90517819402782
+ ],
+ [
+ 8.62110436449729,
+ 56.89411043877302
+ ],
+ [
+ 8.615407753112041,
+ 56.86562741360491
+ ],
+ [
+ 8.614268421594348,
+ 56.863348654951395
+ ],
+ [
+ 8.612559441713975,
+ 56.85468171744676
+ ],
+ [
+ 8.607920764081065,
+ 56.84967675986452
+ ],
+ [
+ 8.601328976094548,
+ 56.84552648826429
+ ],
+ [
+ 8.569590691295334,
+ 56.81842682893014
+ ],
+ [
+ 8.547048372326387,
+ 56.817328183008854
+ ],
+ [
+ 8.536306189280742,
+ 56.814520618844725
+ ],
+ [
+ 8.531748888209458,
+ 56.80878319255899
+ ],
+ [
+ 8.524261913685656,
+ 56.804754932404556
+ ],
+ [
+ 8.507823108976027,
+ 56.801092787022505
+ ],
+ [
+ 8.49138430689679,
+ 56.79425684784418
+ ],
+ [
+ 8.48389733216455,
+ 56.78115469221353
+ ],
+ [
+ 8.49244224893334,
+ 56.74725976482086
+ ],
+ [
+ 8.491465686409867,
+ 56.73139063020126
+ ],
+ [
+ 8.476410349437536,
+ 56.722154004430884
+ ],
+ [
+ 8.476410351004446,
+ 56.71596912956676
+ ],
+ [
+ 8.516368040334106,
+ 56.718207169626496
+ ],
+ [
+ 8.525563996998654,
+ 56.71596912956676
+ ],
+ [
+ 8.533050982831165,
+ 56.703802880176575
+ ],
+ [
+ 8.527028840141444,
+ 56.695298545059046
+ ],
+ [
+ 8.514008003893059,
+ 56.69025292835464
+ ],
+ [
+ 8.50098717168919,
+ 56.68866603842685
+ ],
+ [
+ 8.489268420516233,
+ 56.69074121755575
+ ],
+ [
+ 8.465830925664426,
+ 56.70014070097824
+ ],
+ [
+ 8.45289146790178,
+ 56.70233790253155
+ ],
+ [
+ 8.440928578055477,
+ 56.70038477953351
+ ],
+ [
+ 8.420909055260468,
+ 56.69135170148675
+ ],
+ [
+ 8.408213734725315,
+ 56.68866603842685
+ ],
+ [
+ 8.408213735110149,
+ 56.68121975465745
+ ],
+ [
+ 8.455902536706724,
+ 56.68866603842685
+ ],
+ [
+ 8.470225457141865,
+ 56.68610260791589
+ ],
+ [
+ 8.49122155539534,
+ 56.67698809573879
+ ],
+ [
+ 8.50098717192777,
+ 56.6749941666245
+ ],
+ [
+ 8.571543819451136,
+ 56.69196202707372
+ ],
+ [
+ 8.586273630655944,
+ 56.68866603842685
+ ],
+ [
+ 8.589040566752278,
+ 56.681341940660026
+ ],
+ [
+ 8.586273631241475,
+ 56.66229885820681
+ ],
+ [
+ 8.586273637724014,
+ 56.65387609367957
+ ],
+ [
+ 8.601328973425783,
+ 56.64061108899264
+ ],
+ [
+ 8.608083535885937,
+ 56.63104897336726
+ ],
+ [
+ 8.593272330201621,
+ 56.61635974400443
+ ],
+ [
+ 8.58301841920802,
+ 56.60932030261431
+ ],
+ [
+ 8.56959069091792,
+ 56.60610585990039
+ ],
+ [
+ 8.555430537642662,
+ 56.60488519342441
+ ],
+ [
+ 8.546641468078725,
+ 56.60171127737501
+ ],
+ [
+ 8.547048371283905,
+ 56.597479535967835
+ ],
+ [
+ 8.559580925692737,
+ 56.593003657283404
+ ],
+ [
+ 8.539317259016187,
+ 56.592759574389575
+ ],
+ [
+ 8.511485223465519,
+ 56.60700106374487
+ ],
+ [
+ 8.486989779194154,
+ 56.62531158011541
+ ],
+ [
+ 8.476410349834978,
+ 56.637152381729756
+ ],
+ [
+ 8.467295774103585,
+ 56.65204505804329
+ ],
+ [
+ 8.444834834974284,
+ 56.664048610331655
+ ],
+ [
+ 8.416758663500735,
+ 56.6720645676193
+ ],
+ [
+ 8.391123890967474,
+ 56.6749941666245
+ ],
+ [
+ 8.373383009160847,
+ 56.68024323319805
+ ],
+ [
+ 8.352224160318617,
+ 56.692938616869085
+ ],
+ [
+ 8.332530138669888,
+ 56.70840070982771
+ ],
+ [
+ 8.319509308484943,
+ 56.722154004430884
+ ],
+ [
+ 8.308360221717031,
+ 56.75820546710781
+ ],
+ [
+ 8.302012570676782,
+ 56.763739389347506
+ ],
+ [
+ 8.288422073665112,
+ 56.766343528726495
+ ],
+ [
+ 8.26026451489687,
+ 56.77667872127639
+ ],
+ [
+ 8.24366295216143,
+ 56.777411132976084
+ ],
+ [
+ 8.258067257372966,
+ 56.740546012682344
+ ],
+ [
+ 8.263194204503616,
+ 56.71893948545044
+ ],
+ [
+ 8.260752802674645,
+ 56.705389753037885
+ ],
+ [
+ 8.245860216546964,
+ 56.704006176344954
+ ],
+ [
+ 8.238291863654226,
+ 56.72504304507891
+ ],
+ [
+ 8.236175980011796,
+ 56.77118565576224
+ ],
+ [
+ 8.24789472457522,
+ 56.81244534645829
+ ],
+ [
+ 8.270355669042594,
+ 56.84324784472327
+ ],
+ [
+ 8.44971764396072,
+ 57.003973699471075
+ ],
+ [
+ 8.503591341254559,
+ 57.03465402905361
+ ],
+ [
+ 8.586273637629715,
+ 57.1063907398754
+ ],
+ [
+ 8.618011910901679,
+ 57.122870130163264
+ ],
+ [
+ 8.65381921235167,
+ 57.12250404300879
+ ],
+ [
+ 8.737152543693501,
+ 57.1063907398754
+ ],
+ [
+ 8.783946155715057,
+ 57.10480372275003
+ ],
+ [
+ 8.874359575166626,
+ 57.11611568016025
+ ],
+ [
+ 8.914561392146,
+ 57.12742745245742
+ ],
+ [
+ 8.962575712992873,
+ 57.15656154042745
+ ],
+ [
+ 8.980479364164603,
+ 57.16103751666595
+ ],
+ [
+ 9.230642124980795,
+ 57.14109937528141
+ ],
+ [
+ 9.321543818177306,
+ 57.14704013242578
+ ],
+ [
+ 9.41211998301237,
+ 57.165431973072266
+ ],
+ [
+ 9.49781334921667,
+ 57.19700758037148
+ ],
+ [
+ 9.573252805895546,
+ 57.24290606150845
+ ],
+ [
+ 9.788422068857306,
+ 57.45929594142355
+ ],
+ [
+ 9.826426622084604,
+ 57.489935535048744
+ ],
+ [
+ 9.89535566188599,
+ 57.531561555731564
+ ],
+ [
+ 9.929860868015943,
+ 57.567450205798245
+ ],
+ [
+ 9.943125844089257,
+ 57.572495802224694
+ ],
+ [
+ 9.96697024625269,
+ 57.59137602438129
+ ],
+ [
+ 10.015391466633195,
+ 57.59662500704154
+ ],
+ [
+ 10.107676630245914,
+ 57.5929629717925
+ ],
+ [
+ 10.197601752086225,
+ 57.60122297314783
+ ],
+ [
+ 10.27955163042313,
+ 57.620917074480516
+ ],
+ [
+ 10.350596543789212,
+ 57.648993163945725
+ ],
+ [
+ 10.472504102440322,
+ 57.716986399689986
+ ],
+ [
+ 10.539073123548665,
+ 57.74384197559005
+ ],
+ [
+ 10.566091335304085,
+ 57.74945702735636
+ ],
+ [
+ 10.596039246679975,
+ 57.751165953393574
+ ],
+ [
+ 10.625010623569203,
+ 57.747504083893105
+ ],
+ [
+ 10.648936405789948,
+ 57.73700605045504
+ ],
+ [
+ 10.594411659178583,
+ 57.725897571991815
+ ],
+ [
+ 10.54249108070871,
+ 57.698919975328494
+ ],
+ [
+ 10.460459829101179,
+ 57.630560583017406
+ ],
+ [
+ 10.430430540016054,
+ 57.57054276384266
+ ],
+ [
+ 10.458343943742024,
+ 57.52545805079961
+ ],
+ [
+ 10.5089624423464,
+ 57.484035616757964
+ ],
+ [
+ 10.546560092322496,
+ 57.43528880743455
+ ],
+ [
+ 10.517425981356647,
+ 57.39130284858253
+ ],
+ [
+ 10.51791425630189,
+ 57.33917874911419
+ ],
+ [
+ 10.539073116234418,
+ 57.236721130644064
+ ],
+ [
+ 10.521983274402015,
+ 57.221747194977425
+ ],
+ [
+ 10.470876505300469,
+ 57.20115802377594
+ ],
+ [
+ 10.446787954934406,
+ 57.184881880649705
+ ],
+ [
+ 10.41146894625541,
+ 57.14443594277496
+ ],
+ [
+ 10.395518425172973,
+ 57.117254951883034
+ ],
+ [
+ 10.384043811583865,
+ 57.07420477179594
+ ],
+ [
+ 10.35922284794932,
+ 57.02610907262601
+ ],
+ [
+ 10.347178577715914,
+ 57.01019923899576
+ ],
+ [
+ 10.338715037560112,
+ 57.00324125120324
+ ],
+ [
+ 10.330577018513987,
+ 56.99933502265667
+ ],
+ [
+ 10.319672077093282,
+ 56.99750404005188
+ ],
+ [
+ 10.272471548787905,
+ 56.99787017059957
+ ],
+ [
+ 10.25798587590796,
+ 56.99583571119346
+ ],
+ [
+ 10.239756713812723,
+ 56.992824686562656
+ ],
+ [
+ 10.15601647584315,
+ 57.02383051394371
+ ],
+ [
+ 10.094004752469694,
+ 57.06049223780963
+ ],
+ [
+ 10.052744988464926,
+ 57.069728913241676
+ ],
+ [
+ 10.019379104210397,
+ 57.08844638099549
+ ],
+ [
+ 9.994639518823382,
+ 57.09271881499704
+ ],
+ [
+ 9.974457225044386,
+ 57.08649320878353
+ ],
+ [
+ 9.946136921951027,
+ 57.06118406523756
+ ],
+ [
+ 9.926931183396446,
+ 57.058579790260886
+ ],
+ [
+ 9.94548587611335,
+ 57.05731846249587
+ ],
+ [
+ 9.966563351090022,
+ 57.0722110515149
+ ],
+ [
+ 9.984141468531915,
+ 57.07904690263256
+ ],
+ [
+ 10.001963733615304,
+ 57.083889015322804
+ ],
+ [
+ 10.022634310099294,
+ 57.07908762181481
+ ],
+ [
+ 10.12126712738605,
+ 57.02212153518102
+ ],
+ [
+ 10.153981974072328,
+ 57.013251123064755
+ ],
+ [
+ 10.192393424078409,
+ 56.994045304730015
+ ],
+ [
+ 10.214366081261776,
+ 56.98969146479006
+ ],
+ [
+ 10.295095245040725,
+ 56.988592808349594
+ ],
+ [
+ 10.31373130941015,
+ 56.982855463413735
+ ],
+ [
+ 10.285655144008023,
+ 56.96369049708775
+ ],
+ [
+ 10.27955162265674,
+ 56.95551171584963
+ ],
+ [
+ 10.2747501931051,
+ 56.942287469959346
+ ],
+ [
+ 10.269297727983616,
+ 56.91502519830449
+ ],
+ [
+ 10.265879756860823,
+ 56.90770094571675
+ ],
+ [
+ 10.273122595887296,
+ 56.891343535084744
+ ],
+ [
+ 10.284027532911585,
+ 56.8403180224169
+ ],
+ [
+ 10.288340684608563,
+ 56.80011620710157
+ ],
+ [
+ 10.295420761969932,
+ 56.778875966029474
+ ],
+ [
+ 10.319997595966095,
+ 56.73647699892911
+ ],
+ [
+ 10.328461138857866,
+ 56.72980383177139
+ ],
+ [
+ 10.338226756182637,
+ 56.72394437183867
+ ],
+ [
+ 10.340586788492509,
+ 56.71735264462469
+ ],
+ [
+ 10.326833529992411,
+ 56.7085635439155
+ ],
+ [
+ 10.318695510216426,
+ 56.70823803648404
+ ],
+ [
+ 10.295176633413805,
+ 56.71430089703291
+ ],
+ [
+ 10.282562694792297,
+ 56.71596912956676
+ ],
+ [
+ 10.237559439797598,
+ 56.71596912956676
+ ],
+ [
+ 10.221853057010126,
+ 56.712103538354846
+ ],
+ [
+ 10.203868030052833,
+ 56.70233790253155
+ ],
+ [
+ 10.187022326060728,
+ 56.712225586314645
+ ],
+ [
+ 10.168142120581265,
+ 56.72003812579755
+ ],
+ [
+ 10.146494987512652,
+ 56.72235748350683
+ ],
+ [
+ 10.121267121811226,
+ 56.71596912956676
+ ],
+ [
+ 10.09896894920698,
+ 56.706122173325866
+ ],
+ [
+ 10.089610223946876,
+ 56.70404696407359
+ ],
+ [
+ 10.073496936116044,
+ 56.70233790253155
+ ],
+ [
+ 10.058360219570032,
+ 56.69822822935727
+ ],
+ [
+ 10.027517122065499,
+ 56.68427154133272
+ ],
+ [
+ 9.990896028394332,
+ 56.67694731460128
+ ],
+ [
+ 9.945567247257936,
+ 56.65814842735746
+ ],
+ [
+ 9.805918813125894,
+ 56.64704007242005
+ ],
+ [
+ 9.8059188160099,
+ 56.640204169114746
+ ],
+ [
+ 9.851328973639628,
+ 56.635728276072385
+ ],
+ [
+ 10.001475452934299,
+ 56.661322286443976
+ ],
+ [
+ 10.0349227211484,
+ 56.672837622282955
+ ],
+ [
+ 10.049571156368094,
+ 56.6749941666245
+ ],
+ [
+ 10.052907744111653,
+ 56.67715073872065
+ ],
+ [
+ 10.053477415171583,
+ 56.681830203880615
+ ],
+ [
+ 10.055430532508321,
+ 56.68650947863871
+ ],
+ [
+ 10.062836130080862,
+ 56.68866603842685
+ ],
+ [
+ 10.07097415206388,
+ 56.687486036632905
+ ],
+ [
+ 10.084320506657336,
+ 56.68235909040863
+ ],
+ [
+ 10.090586781531904,
+ 56.68121975465745
+ ],
+ [
+ 10.101328978273054,
+ 56.68268470808155
+ ],
+ [
+ 10.113617380144698,
+ 56.68671291437081
+ ],
+ [
+ 10.12378990426307,
+ 56.692328127151626
+ ],
+ [
+ 10.128103066973903,
+ 56.6988793664857
+ ],
+ [
+ 10.134776238897338,
+ 56.71238842412603
+ ],
+ [
+ 10.150564000449258,
+ 56.71442294659631
+ ],
+ [
+ 10.183360221992565,
+ 56.7085635439155
+ ],
+ [
+ 10.180837434851979,
+ 56.704901421099656
+ ],
+ [
+ 10.17579186623692,
+ 56.695461366397815
+ ],
+ [
+ 10.193207225917762,
+ 56.69358957284981
+ ]
+ ]
+ ],
+ [
+ [
+ [
+ 8.825856969761675,
+ 56.74384189976032
+ ],
+ [
+ 8.812673378665838,
+ 56.7345645934932
+ ],
+ [
+ 8.771250845969831,
+ 56.71596912956676
+ ],
+ [
+ 8.766123894216294,
+ 56.69871654182198
+ ],
+ [
+ 8.686371289849399,
+ 56.68659088501487
+ ],
+ [
+ 8.658376502289403,
+ 56.6781274027111
+ ],
+ [
+ 8.648448118645883,
+ 56.6810571077253
+ ],
+ [
+ 8.62175540132782,
+ 56.713934588286094
+ ],
+ [
+ 8.610850454396903,
+ 56.722154004430884
+ ],
+ [
+ 8.586273637299831,
+ 56.73647699892911
+ ],
+ [
+ 8.56226647450189,
+ 56.743231545393705
+ ],
+ [
+ 8.555918818677247,
+ 56.74384189976032
+ ],
+ [
+ 8.524912962136634,
+ 56.73729082587258
+ ],
+ [
+ 8.51465904501759,
+ 56.738226563114196
+ ],
+ [
+ 8.518077021320037,
+ 56.750067481121675
+ ],
+ [
+ 8.522797071260046,
+ 56.75507233648518
+ ],
+ [
+ 8.530284053862667,
+ 56.75934484572938
+ ],
+ [
+ 8.539235878677434,
+ 56.7625186979483
+ ],
+ [
+ 8.548675981815718,
+ 56.763739389347506
+ ],
+ [
+ 8.553233271403515,
+ 56.76813388809437
+ ],
+ [
+ 8.551605665700956,
+ 56.77777742035894
+ ],
+ [
+ 8.548024941480847,
+ 56.78725827819759
+ ],
+ [
+ 8.545258005929366,
+ 56.791083033980705
+ ],
+ [
+ 8.573008658771343,
+ 56.80255766764401
+ ],
+ [
+ 8.643809437635438,
+ 56.808050803585424
+ ],
+ [
+ 8.668793168758866,
+ 56.82583247345247
+ ],
+ [
+ 8.656260611032152,
+ 56.82514066607107
+ ],
+ [
+ 8.643809442234117,
+ 56.825873130282424
+ ],
+ [
+ 8.632009311119157,
+ 56.82807038157371
+ ],
+ [
+ 8.621104357469092,
+ 56.832017246840564
+ ],
+ [
+ 8.6267195968653,
+ 56.841538802218736
+ ],
+ [
+ 8.6467391341718,
+ 56.85635007016282
+ ],
+ [
+ 8.655121290684125,
+ 56.86684805800479
+ ],
+ [
+ 8.651866081506084,
+ 56.86603424049468
+ ],
+ [
+ 8.649750193195297,
+ 56.87425361106286
+ ],
+ [
+ 8.64844811291625,
+ 56.89069244889631
+ ],
+ [
+ 8.651621940004175,
+ 56.89419178589387
+ ],
+ [
+ 8.659353063425995,
+ 56.894476661896704
+ ],
+ [
+ 8.668223503385242,
+ 56.89378488992415
+ ],
+ [
+ 8.675059442506658,
+ 56.89411043877302
+ ],
+ [
+ 8.70378665196756,
+ 56.90005113433761
+ ],
+ [
+ 8.762461783754013,
+ 56.905422251813754
+ ],
+ [
+ 8.791758663322138,
+ 56.91453689399943
+ ],
+ [
+ 8.813649935876292,
+ 56.91034576840197
+ ],
+ [
+ 8.833750849708155,
+ 56.92475010187667
+ ],
+ [
+ 8.867442253361073,
+ 56.96295806680949
+ ],
+ [
+ 8.89039146858317,
+ 56.95482001030748
+ ],
+ [
+ 8.905039913766345,
+ 56.96600988106698
+ ],
+ [
+ 8.916270376054399,
+ 56.97907131846535
+ ],
+ [
+ 8.928884314226604,
+ 56.976629991058175
+ ],
+ [
+ 8.923106314335104,
+ 56.968573287795095
+ ],
+ [
+ 8.911387571526602,
+ 56.95612220749948
+ ],
+ [
+ 8.908457878292484,
+ 56.94871652996888
+ ],
+ [
+ 8.910411008579791,
+ 56.94440344544499
+ ],
+ [
+ 8.920420767962002,
+ 56.930487358762576
+ ],
+ [
+ 8.922699416490154,
+ 56.922023849000574
+ ],
+ [
+ 8.920420773233658,
+ 56.91990799900012
+ ],
+ [
+ 8.908457882794385,
+ 56.88727452950182
+ ],
+ [
+ 8.90040123690991,
+ 56.87482329504766
+ ],
+ [
+ 8.895274284874622,
+ 56.87030670639326
+ ],
+ [
+ 8.887950066702551,
+ 56.86684805800479
+ ],
+ [
+ 8.887705931117841,
+ 56.87872963740132
+ ],
+ [
+ 8.882660349616154,
+ 56.88637926843774
+ ],
+ [
+ 8.875010611930213,
+ 56.89109933727384
+ ],
+ [
+ 8.867442255540064,
+ 56.89411043877302
+ ],
+ [
+ 8.843760615967915,
+ 56.88344964921835
+ ],
+ [
+ 8.839610223185392,
+ 56.87982820727238
+ ],
+ [
+ 8.83643640016492,
+ 56.86928953448189
+ ],
+ [
+ 8.839040555035833,
+ 56.86286034211504
+ ],
+ [
+ 8.843597853262972,
+ 56.85822176327372
+ ],
+ [
+ 8.847015826836552,
+ 56.8530948691891
+ ],
+ [
+ 8.855153840189603,
+ 56.81976957467872
+ ],
+ [
+ 8.86394290915155,
+ 56.809515745438546
+ ],
+ [
+ 8.881114132841779,
+ 56.80536535136014
+ ],
+ [
+ 8.874766473001845,
+ 56.80036043188197
+ ],
+ [
+ 8.868418815326491,
+ 56.7980817643313
+ ],
+ [
+ 8.86133874037303,
+ 56.79755283256295
+ ],
+ [
+ 8.853282102551468,
+ 56.797837703564795
+ ],
+ [
+ 8.860687690766337,
+ 56.79490787758155
+ ],
+ [
+ 8.867442250813605,
+ 56.791083033980705
+ ],
+ [
+ 8.85417728371998,
+ 56.77993403397182
+ ],
+ [
+ 8.839121942538183,
+ 56.75726961788023
+ ],
+ [
+ 8.825856969761675,
+ 56.74384189976032
+ ]
+ ]
+ ],
+ [
+ [
+ [
+ 11.142344592013188,
+ 57.332342790595156
+ ],
+ [
+ 11.16993248552196,
+ 57.33270902701456
+ ],
+ [
+ 11.181488479074925,
+ 57.33030835589918
+ ],
+ [
+ 11.190114785573531,
+ 57.32550695910287
+ ],
+ [
+ 11.197276232006054,
+ 57.31338119168454
+ ],
+ [
+ 11.196501701216176,
+ 57.30282124550524
+ ],
+ [
+ 11.18911338963492,
+ 57.29428256779825
+ ],
+ [
+ 11.166875951785547,
+ 57.296775050740784
+ ],
+ [
+ 11.143231241170417,
+ 57.30135125618067
+ ],
+ [
+ 11.066579629887661,
+ 57.29132728616404
+ ],
+ [
+ 11.08269291039726,
+ 57.283880981911565
+ ],
+ [
+ 11.094164610441222,
+ 57.26899600315104
+ ],
+ [
+ 11.09110072083318,
+ 57.252784136695496
+ ],
+ [
+ 11.06716630688235,
+ 57.238361948600286
+ ],
+ [
+ 11.053517090142824,
+ 57.22620208381221
+ ],
+ [
+ 11.045137616154513,
+ 57.21271488076637
+ ],
+ [
+ 11.028835403866951,
+ 57.201914227924966
+ ],
+ [
+ 10.997904584524852,
+ 57.19791000976349
+ ],
+ [
+ 10.977961917198284,
+ 57.20673291716783
+ ],
+ [
+ 10.951814885380303,
+ 57.23372968091941
+ ],
+ [
+ 10.916536637386958,
+ 57.23736859527434
+ ],
+ [
+ 10.888799322871035,
+ 57.245344107837155
+ ],
+ [
+ 10.871104363000443,
+ 57.25462474200373
+ ],
+ [
+ 10.854991083793413,
+ 57.26463452563711
+ ],
+ [
+ 10.854991080001577,
+ 57.27090076223399
+ ],
+ [
+ 10.888424553471385,
+ 57.274864878955384
+ ],
+ [
+ 10.921263282411266,
+ 57.29228003550066
+ ],
+ [
+ 10.946601230285697,
+ 57.3046432384645
+ ],
+ [
+ 10.986869593747608,
+ 57.30599270295381
+ ],
+ [
+ 11.010083618981037,
+ 57.30987338244718
+ ],
+ [
+ 11.027723185323206,
+ 57.320690553163026
+ ],
+ [
+ 11.101927417912405,
+ 57.32041686322719
+ ],
+ [
+ 11.142344592013188,
+ 57.332342790595156
+ ]
+ ]
+ ]
+ ]
+ },
+ "type": "Feature",
+ "properties": {
+ "source": "https://simplemaps.com",
+ "id": "DK81",
+ "name": "Nordjylland"
+ },
+ "id": 3
+ },
+ {
+ "geometry": {
+ "type": "MultiPolygon",
+ "coordinates": [
+ [
+ [
+ [
+ 11.845225460623086,
+ 54.731919700839164
+ ],
+ [
+ 11.833832223326482,
+ 54.72455471760293
+ ],
+ [
+ 11.82357830960239,
+ 54.72211323186349
+ ],
+ [
+ 11.814707868141676,
+ 54.71869526926492
+ ],
+ [
+ 11.806976772532268,
+ 54.70799402013389
+ ],
+ [
+ 11.846446167708578,
+ 54.69782143444661
+ ],
+ [
+ 11.84262129388001,
+ 54.67621495052116
+ ],
+ [
+ 11.810720241309175,
+ 54.65509667862272
+ ],
+ [
+ 11.766123906274881,
+ 54.6465519509475
+ ],
+ [
+ 11.610524932693899,
+ 54.66640856164108
+ ],
+ [
+ 11.5600692192957,
+ 54.659613476363404
+ ],
+ [
+ 11.512217656009998,
+ 54.6465519509475
+ ],
+ [
+ 11.474375856144098,
+ 54.62612548926378
+ ],
+ [
+ 11.45183351687243,
+ 54.628363209535415
+ ],
+ [
+ 11.358734583832662,
+ 54.66054935734039
+ ],
+ [
+ 11.297699420342623,
+ 54.69155513701995
+ ],
+ [
+ 11.187754757942226,
+ 54.73224522643956
+ ],
+ [
+ 11.138194215322757,
+ 54.73859293098146
+ ],
+ [
+ 11.084646021939323,
+ 54.753119120464696
+ ],
+ [
+ 11.030528194394009,
+ 54.76064691402952
+ ],
+ [
+ 11.009287963593168,
+ 54.77163327705111
+ ],
+ [
+ 10.999196797588041,
+ 54.786769777356184
+ ],
+ [
+ 11.00570721411709,
+ 54.80365617628867
+ ],
+ [
+ 11.030772325703206,
+ 54.81216048039607
+ ],
+ [
+ 11.06128989811818,
+ 54.80931211289472
+ ],
+ [
+ 11.08969161309564,
+ 54.81061444524824
+ ],
+ [
+ 11.108164902275016,
+ 54.83152895546649
+ ],
+ [
+ 11.09197025311721,
+ 54.835028442409026
+ ],
+ [
+ 11.078379745481078,
+ 54.841131810657885
+ ],
+ [
+ 11.025238487825996,
+ 54.879462098791386
+ ],
+ [
+ 11.018809449453293,
+ 54.886175940011945
+ ],
+ [
+ 11.018809438369066,
+ 54.903387733375105
+ ],
+ [
+ 11.03239994658802,
+ 54.91925701800389
+ ],
+ [
+ 11.073985230438417,
+ 54.948187658239775
+ ],
+ [
+ 11.088145372108036,
+ 54.944159172566664
+ ],
+ [
+ 11.231700077055248,
+ 54.96124920881458
+ ],
+ [
+ 11.25603274357512,
+ 54.954575866937105
+ ],
+ [
+ 11.303477410494047,
+ 54.9324405022349
+ ],
+ [
+ 11.321055545974465,
+ 54.92772056113519
+ ],
+ [
+ 11.338877810252818,
+ 54.91913494135289
+ ],
+ [
+ 11.36890708354355,
+ 54.88108944301853
+ ],
+ [
+ 11.392751509402464,
+ 54.872504093084274
+ ],
+ [
+ 11.416758674506582,
+ 54.86933029746676
+ ],
+ [
+ 11.594737161378692,
+ 54.811102464810304
+ ],
+ [
+ 11.585134305907584,
+ 54.82851791819645
+ ],
+ [
+ 11.573578315827504,
+ 54.83934148532813
+ ],
+ [
+ 11.572764522148955,
+ 54.84796787366821
+ ],
+ [
+ 11.594737163375743,
+ 54.85887266866664
+ ],
+ [
+ 11.622080913348329,
+ 54.85887266866664
+ ],
+ [
+ 11.646332237296038,
+ 54.86538331109131
+ ],
+ [
+ 11.656260602627857,
+ 54.86627828021659
+ ],
+ [
+ 11.647227398102201,
+ 54.877142522450825
+ ],
+ [
+ 11.644786004067926,
+ 54.88963450769874
+ ],
+ [
+ 11.646332228097899,
+ 54.90228914130091
+ ],
+ [
+ 11.649424665775003,
+ 54.913519503092054
+ ],
+ [
+ 11.725759305988953,
+ 54.87335841472575
+ ],
+ [
+ 11.731944195238176,
+ 54.85887266866664
+ ],
+ [
+ 11.755625837960297,
+ 54.841131810657885
+ ],
+ [
+ 11.770681189424478,
+ 54.83429600094265
+ ],
+ [
+ 11.7872827398027,
+ 54.83152895546649
+ ],
+ [
+ 11.777191614027863,
+ 54.82599530669044
+ ],
+ [
+ 11.772959840098588,
+ 54.82469318364205
+ ],
+ [
+ 11.78760826368005,
+ 54.806870780752284
+ ],
+ [
+ 11.837738478075808,
+ 54.78278230492996
+ ],
+ [
+ 11.848643412699744,
+ 54.77317930410614
+ ],
+ [
+ 11.85132897075778,
+ 54.756821994383195
+ ],
+ [
+ 11.85547935091771,
+ 54.749456975310345
+ ],
+ [
+ 11.85547936009498,
+ 54.74359769349557
+ ],
+ [
+ 11.845225460623086,
+ 54.731919700839164
+ ]
+ ]
+ ],
+ [
+ [
+ [
+ 12.417165560903817,
+ 55.03021881607416
+ ],
+ [
+ 12.504730664936336,
+ 55.01976145939095
+ ],
+ [
+ 12.53760826900682,
+ 55.002386786064235
+ ],
+ [
+ 12.55372154390337,
+ 54.96808496988032
+ ],
+ [
+ 12.551768422221377,
+ 54.957180053498845
+ ],
+ [
+ 12.542653852731997,
+ 54.95115804744394
+ ],
+ [
+ 12.526215027097674,
+ 54.948675724667865
+ ],
+ [
+ 12.50277754952713,
+ 54.948187658239775
+ ],
+ [
+ 12.384613477671735,
+ 54.96308015149127
+ ],
+ [
+ 12.348399295061654,
+ 54.9600284845324
+ ],
+ [
+ 12.313975460963622,
+ 54.9508731525377
+ ],
+ [
+ 12.280039906145827,
+ 54.9345563444049
+ ],
+ [
+ 12.226817255540526,
+ 54.89264559010829
+ ],
+ [
+ 12.20777429436542,
+ 54.886175940011945
+ ],
+ [
+ 12.178721557317544,
+ 54.88898353805816
+ ],
+ [
+ 12.115000834841005,
+ 54.90729389169413
+ ],
+ [
+ 12.115000837406315,
+ 54.913519503092054
+ ],
+ [
+ 12.129160994395093,
+ 54.913519503092054
+ ],
+ [
+ 12.124685100600102,
+ 54.92373289998612
+ ],
+ [
+ 12.128916853482357,
+ 54.931626600998236
+ ],
+ [
+ 12.138194209849308,
+ 54.9372419571147
+ ],
+ [
+ 12.149099151473372,
+ 54.940822623235334
+ ],
+ [
+ 12.144053567516108,
+ 54.94379289317412
+ ],
+ [
+ 12.13453208548344,
+ 54.95160542236959
+ ],
+ [
+ 12.129160995262616,
+ 54.95441306675911
+ ],
+ [
+ 12.151377788317765,
+ 54.96889871494803
+ ],
+ [
+ 12.164398619729067,
+ 54.97264217961834
+ ],
+ [
+ 12.177012560086311,
+ 54.96808496988032
+ ],
+ [
+ 12.184336776929486,
+ 54.975531237746765
+ ],
+ [
+ 12.177012559396879,
+ 54.98920312714131
+ ],
+ [
+ 12.193125835201323,
+ 54.98834858425947
+ ],
+ [
+ 12.217295781128865,
+ 54.97866457377333
+ ],
+ [
+ 12.232188338897794,
+ 54.975531237746765
+ ],
+ [
+ 12.242198113041608,
+ 54.97662995040588
+ ],
+ [
+ 12.27475021217615,
+ 54.984442607641455
+ ],
+ [
+ 12.286875840337306,
+ 54.98920312714131
+ ],
+ [
+ 12.281504758105072,
+ 54.99909092696695
+ ],
+ [
+ 12.28126061298283,
+ 55.007147527832096
+ ],
+ [
+ 12.285655143943591,
+ 55.013128972450346
+ ],
+ [
+ 12.294200066026246,
+ 55.01654694225437
+ ],
+ [
+ 12.294200066112452,
+ 55.02338288109222
+ ],
+ [
+ 12.261566602115444,
+ 55.03681061512405
+ ],
+ [
+ 12.252696159890135,
+ 55.0438500009283
+ ],
+ [
+ 12.247569207110306,
+ 55.05385976907461
+ ],
+ [
+ 12.250987174882193,
+ 55.0607363938512
+ ],
+ [
+ 12.261973504159947,
+ 55.064276435556344
+ ],
+ [
+ 12.280039910160186,
+ 55.064276435556344
+ ],
+ [
+ 12.30176842515586,
+ 55.042059638512434
+ ],
+ [
+ 12.335297070956194,
+ 55.033270574575106
+ ],
+ [
+ 12.417165560903817,
+ 55.03021881607416
+ ]
+ ]
+ ],
+ [
+ [
+ [
+ 12.359205489553343,
+ 55.60459309122142
+ ],
+ [
+ 12.331309441686834,
+ 55.59540437055523
+ ],
+ [
+ 12.31202233376881,
+ 55.58397046191442
+ ],
+ [
+ 12.248383008653981,
+ 55.54625071466998
+ ],
+ [
+ 12.248220248456937,
+ 55.546128652396355
+ ],
+ [
+ 12.19809004091944,
+ 55.48749421689665
+ ],
+ [
+ 12.22486412957257,
+ 55.43488190853541
+ ],
+ [
+ 12.273448112901432,
+ 55.41022369904938
+ ],
+ [
+ 12.290537957949557,
+ 55.40692781113812
+ ],
+ [
+ 12.352061392945194,
+ 55.404445694897305
+ ],
+ [
+ 12.368825715988994,
+ 55.40009185432919
+ ],
+ [
+ 12.412119985986495,
+ 55.37913644079626
+ ],
+ [
+ 12.434743684535524,
+ 55.36367420705257
+ ],
+ [
+ 12.444590690035609,
+ 55.348863013818445
+ ],
+ [
+ 12.446299675048309,
+ 55.331691799459215
+ ],
+ [
+ 12.450856967719215,
+ 55.31720612884864
+ ],
+ [
+ 12.46501712304485,
+ 55.290269273426425
+ ],
+ [
+ 12.450205924410247,
+ 55.27968984037714
+ ],
+ [
+ 12.444590691068491,
+ 55.2765567086525
+ ],
+ [
+ 12.392425976521851,
+ 55.25531646841701
+ ],
+ [
+ 12.182790560150604,
+ 55.227199602713156
+ ],
+ [
+ 12.167002800192483,
+ 55.21735260787991
+ ],
+ [
+ 12.141368034333517,
+ 55.207993875471686
+ ],
+ [
+ 12.121104363201272,
+ 55.191066800974646
+ ],
+ [
+ 12.109060091757371,
+ 55.17072174461459
+ ],
+ [
+ 12.115000846889258,
+ 55.1537132819106
+ ],
+ [
+ 12.107432488248671,
+ 55.14508698344561
+ ],
+ [
+ 12.098968945712947,
+ 55.14142486917319
+ ],
+ [
+ 12.09009850363263,
+ 55.142238670379705
+ ],
+ [
+ 12.080739779925821,
+ 55.14687734526829
+ ],
+ [
+ 12.08814537916541,
+ 55.16836172262812
+ ],
+ [
+ 12.076670769550468,
+ 55.18195222142304
+ ],
+ [
+ 12.056407096758033,
+ 55.185695702611234
+ ],
+ [
+ 12.036794466094433,
+ 55.17755767504625
+ ],
+ [
+ 12.02784264348068,
+ 55.17267486686016
+ ],
+ [
+ 12.020030144927066,
+ 55.170965896179474
+ ],
+ [
+ 12.014659050893867,
+ 55.167466547855454
+ ],
+ [
+ 12.012461785034107,
+ 55.15713125233733
+ ],
+ [
+ 12.017914259067021,
+ 55.1530215526635
+ ],
+ [
+ 12.046641472120578,
+ 55.140082099191496
+ ],
+ [
+ 12.130544466824041,
+ 55.135891016271344
+ ],
+ [
+ 12.168630404778035,
+ 55.12848541682601
+ ],
+ [
+ 12.177012565590688,
+ 55.10529205499124
+ ],
+ [
+ 12.171397332175514,
+ 55.0954450557191
+ ],
+ [
+ 12.165049675448072,
+ 55.0912946683926
+ ],
+ [
+ 12.157725456721337,
+ 55.089178775265076
+ ],
+ [
+ 12.14909915474574,
+ 55.08539459550287
+ ],
+ [
+ 12.128428582052921,
+ 55.078843492519844
+ ],
+ [
+ 12.118500196177802,
+ 55.07355377975029
+ ],
+ [
+ 12.125743035246046,
+ 55.0711123744209
+ ],
+ [
+ 12.163340691111259,
+ 55.02338288109222
+ ],
+ [
+ 12.162282748004234,
+ 55.00438060104112
+ ],
+ [
+ 12.148936384941022,
+ 54.994126605825144
+ ],
+ [
+ 12.128428577349311,
+ 54.98997625813666
+ ],
+ [
+ 12.104991075435928,
+ 54.98920312714131
+ ],
+ [
+ 12.089528844388004,
+ 54.98566317662728
+ ],
+ [
+ 12.055918802169598,
+ 54.97077043772056
+ ],
+ [
+ 12.032969591156272,
+ 54.96808496988032
+ ],
+ [
+ 11.998057483543878,
+ 54.976507835561854
+ ],
+ [
+ 11.943858269030692,
+ 55.00608958530637
+ ],
+ [
+ 11.91016686300681,
+ 55.00971100506752
+ ],
+ [
+ 11.90951582099313,
+ 55.00495026193042
+ ],
+ [
+ 11.911143425006369,
+ 55.00381094006302
+ ],
+ [
+ 11.913828971987568,
+ 55.003851629874376
+ ],
+ [
+ 11.916840039996298,
+ 55.00287506696223
+ ],
+ [
+ 11.9113875660031,
+ 55.00092194203031
+ ],
+ [
+ 11.908376509967857,
+ 54.99860272617312
+ ],
+ [
+ 11.904633018773572,
+ 54.9965681991756
+ ],
+ [
+ 11.897146018029318,
+ 54.995428657673976
+ ],
+ [
+ 11.901540552226553,
+ 54.97638572079925
+ ],
+ [
+ 11.894379095758357,
+ 54.96499251719451
+ ],
+ [
+ 11.885020388302207,
+ 54.95514566869857
+ ],
+ [
+ 11.882823109550706,
+ 54.940822623235334
+ ],
+ [
+ 11.890147331617431,
+ 54.92963287614468
+ ],
+ [
+ 11.903330928024944,
+ 54.92300042344825
+ ],
+ [
+ 11.91846763971696,
+ 54.921942406939635
+ ],
+ [
+ 11.931162968565891,
+ 54.92772056113519
+ ],
+ [
+ 11.928558777785014,
+ 54.93138243133094
+ ],
+ [
+ 11.923675973538836,
+ 54.940822623235334
+ ],
+ [
+ 11.932383667628544,
+ 54.94147377656314
+ ],
+ [
+ 11.957855674111904,
+ 54.948187658239775
+ ],
+ [
+ 11.965993681953197,
+ 54.94261298249913
+ ],
+ [
+ 11.980235234662349,
+ 54.92426191263591
+ ],
+ [
+ 11.982188332514237,
+ 54.92031470113451
+ ],
+ [
+ 11.991465689476021,
+ 54.91632720076877
+ ],
+ [
+ 12.023936385407016,
+ 54.89801658331203
+ ],
+ [
+ 12.040375194146172,
+ 54.89301176753617
+ ],
+ [
+ 12.058929870458066,
+ 54.8943544243411
+ ],
+ [
+ 12.078298365472063,
+ 54.897853833270304
+ ],
+ [
+ 12.096446150858279,
+ 54.898097958386884
+ ],
+ [
+ 12.111827019507949,
+ 54.88959382202874
+ ],
+ [
+ 12.139903180514333,
+ 54.86493551034474
+ ],
+ [
+ 12.170176622766075,
+ 54.84454981049402
+ ],
+ [
+ 12.170176631036847,
+ 54.8383243060986
+ ],
+ [
+ 12.081716342933536,
+ 54.79645417828962
+ ],
+ [
+ 12.067149299453042,
+ 54.7868107580399
+ ],
+ [
+ 12.02800540569684,
+ 54.74355703997523
+ ],
+ [
+ 12.022715690562311,
+ 54.73529693761481
+ ],
+ [
+ 12.012705911989062,
+ 54.73110573054387
+ ],
+ [
+ 11.985850450053752,
+ 54.712306963159826
+ ],
+ [
+ 11.975271043725133,
+ 54.70799402013389
+ ],
+ [
+ 11.960703967756812,
+ 54.69407782422585
+ ],
+ [
+ 11.960215686975578,
+ 54.66233954439386
+ ],
+ [
+ 11.971446149741864,
+ 54.60561747643132
+ ],
+ [
+ 11.969004742733691,
+ 54.56976948411546
+ ],
+ [
+ 11.949066616400767,
+ 54.56858973081493
+ ],
+ [
+ 11.928233282145229,
+ 54.58397057827691
+ ],
+ [
+ 11.923675971145741,
+ 54.59813054164454
+ ],
+ [
+ 11.909353049365398,
+ 54.60952371193509
+ ],
+ [
+ 11.876638203817894,
+ 54.64695859200255
+ ],
+ [
+ 11.86915125062445,
+ 54.659613476363404
+ ],
+ [
+ 11.870616075192352,
+ 54.68036523282794
+ ],
+ [
+ 11.882660359861722,
+ 54.69643797276442
+ ],
+ [
+ 11.910166865058796,
+ 54.72166577782497
+ ],
+ [
+ 11.889903192205171,
+ 54.72418855621003
+ ],
+ [
+ 11.87582440914097,
+ 54.72972233857388
+ ],
+ [
+ 11.872325078063387,
+ 54.73993573635425
+ ],
+ [
+ 11.882823113686953,
+ 54.75641511295742
+ ],
+ [
+ 11.871429882281001,
+ 54.760321338572595
+ ],
+ [
+ 11.86524499154493,
+ 54.765448344950556
+ ],
+ [
+ 11.855479363311744,
+ 54.77692292216238
+ ],
+ [
+ 11.809092653018478,
+ 54.804917801797004
+ ],
+ [
+ 11.805430542711507,
+ 54.80951577050808
+ ],
+ [
+ 11.80095460978061,
+ 54.817857029395824
+ ],
+ [
+ 11.798838724259534,
+ 54.825872974095205
+ ],
+ [
+ 11.800059446399633,
+ 54.83120357296399
+ ],
+ [
+ 11.799815306972512,
+ 54.83657468196597
+ ],
+ [
+ 11.793467637959045,
+ 54.84454981049402
+ ],
+ [
+ 11.776052281603485,
+ 54.85614656834425
+ ],
+ [
+ 11.760590040131047,
+ 54.863226631336644
+ ],
+ [
+ 11.749685091747752,
+ 54.87311432242379
+ ],
+ [
+ 11.745616080191573,
+ 54.89301176753617
+ ],
+ [
+ 11.745371928397926,
+ 54.91132220440013
+ ],
+ [
+ 11.742035342821978,
+ 54.92544170535037
+ ],
+ [
+ 11.731944213175325,
+ 54.93528886704727
+ ],
+ [
+ 11.711436390600463,
+ 54.940822623235334
+ ],
+ [
+ 11.711436402924152,
+ 54.948187658239775
+ ],
+ [
+ 11.749766456432706,
+ 54.965724843432746
+ ],
+ [
+ 11.759287951289178,
+ 54.96808496988032
+ ],
+ [
+ 11.773610868039578,
+ 54.96503321957469
+ ],
+ [
+ 11.79297937676343,
+ 54.951320844731924
+ ],
+ [
+ 11.803965699994658,
+ 54.948187658239775
+ ],
+ [
+ 11.814789258887432,
+ 54.952093815858696
+ ],
+ [
+ 11.837575713663643,
+ 54.966131869268324
+ ],
+ [
+ 11.841807503579913,
+ 54.96466721646516
+ ],
+ [
+ 11.848317890520025,
+ 54.95913305870945
+ ],
+ [
+ 11.862478047797072,
+ 54.96369036378688
+ ],
+ [
+ 11.882823105129203,
+ 54.975531237746765
+ ],
+ [
+ 11.882823116605525,
+ 54.98175693929783
+ ],
+ [
+ 11.872325075363035,
+ 54.98956949732511
+ ],
+ [
+ 11.853526237945724,
+ 55.02008698045201
+ ],
+ [
+ 11.841807487908317,
+ 55.03021881607416
+ ],
+ [
+ 11.774668816157412,
+ 55.04857005559562
+ ],
+ [
+ 11.755625846948723,
+ 55.05068593947829
+ ],
+ [
+ 11.75025475380844,
+ 55.052801823052626
+ ],
+ [
+ 11.736989779824334,
+ 55.062201237212925
+ ],
+ [
+ 11.728770379152994,
+ 55.064276435556344
+ ],
+ [
+ 11.665537956783496,
+ 55.06639231778394
+ ],
+ [
+ 11.646657747804463,
+ 55.07233307299589
+ ],
+ [
+ 11.630381707136898,
+ 55.08112213840362
+ ],
+ [
+ 11.615896030054175,
+ 55.09223053655587
+ ],
+ [
+ 11.669606967271775,
+ 55.085638741777906
+ ],
+ [
+ 11.775238476761748,
+ 55.057440494583716
+ ],
+ [
+ 11.827484570948393,
+ 55.05068593947829
+ ],
+ [
+ 11.80534915496288,
+ 55.07387929862434
+ ],
+ [
+ 11.743418816226786,
+ 55.10053131930205
+ ],
+ [
+ 11.717621290039412,
+ 55.12571849240018
+ ],
+ [
+ 11.803965690567331,
+ 55.13483307063208
+ ],
+ [
+ 11.809580924837514,
+ 55.13886139336001
+ ],
+ [
+ 11.806976758927522,
+ 55.14687734526829
+ ],
+ [
+ 11.785411004135748,
+ 55.15721263337069
+ ],
+ [
+ 11.728526238216995,
+ 55.15477122820228
+ ],
+ [
+ 11.711436394461423,
+ 55.167914134689134
+ ],
+ [
+ 11.726247591141602,
+ 55.17694732428767
+ ],
+ [
+ 11.73471113369988,
+ 55.18866607995658
+ ],
+ [
+ 11.733083529464961,
+ 55.20014068657598
+ ],
+ [
+ 11.717621290853144,
+ 55.20831941165712
+ ],
+ [
+ 11.695078972729206,
+ 55.20807526941348
+ ],
+ [
+ 11.67823326890401,
+ 55.199367579021576
+ ],
+ [
+ 11.663828971479502,
+ 55.18829986668994
+ ],
+ [
+ 11.649424675430996,
+ 55.18097565740169
+ ],
+ [
+ 11.636241081791297,
+ 55.181626692864945
+ ],
+ [
+ 11.599375847016528,
+ 55.19131094016861
+ ],
+ [
+ 11.52418053410674,
+ 55.20209380778066
+ ],
+ [
+ 11.506032748837727,
+ 55.20831941165712
+ ],
+ [
+ 11.492360873621697,
+ 55.20416901843293
+ ],
+ [
+ 11.423187696785993,
+ 55.2219099061782
+ ],
+ [
+ 11.410329622600706,
+ 55.220119528839625
+ ],
+ [
+ 11.382985872915475,
+ 55.21116770911658
+ ],
+ [
+ 11.36890709782778,
+ 55.20831941165712
+ ],
+ [
+ 11.299978061822719,
+ 55.20831941165712
+ ],
+ [
+ 11.299978061740706,
+ 55.20148347379554
+ ],
+ [
+ 11.320485873715903,
+ 55.1946475355221
+ ],
+ [
+ 11.285899284875034,
+ 55.19537994868205
+ ],
+ [
+ 11.26099694020241,
+ 55.20058827457524
+ ],
+ [
+ 11.245371940253325,
+ 55.21478912210494
+ ],
+ [
+ 11.238536003830053,
+ 55.242417708202645
+ ],
+ [
+ 11.253103061527058,
+ 55.236273510564864
+ ],
+ [
+ 11.267832879605496,
+ 55.235907300384405
+ ],
+ [
+ 11.2800399108165,
+ 55.24233633859871
+ ],
+ [
+ 11.28630618519703,
+ 55.25608957655029
+ ],
+ [
+ 11.276052279163718,
+ 55.25161366519073
+ ],
+ [
+ 11.265391471627082,
+ 55.249457094067104
+ ],
+ [
+ 11.241953972799505,
+ 55.248683994446814
+ ],
+ [
+ 11.238780144823496,
+ 55.25413646870143
+ ],
+ [
+ 11.241547071700618,
+ 55.27850984039653
+ ],
+ [
+ 11.235118034819514,
+ 55.28400299509278
+ ],
+ [
+ 11.21713300851041,
+ 55.29022857782053
+ ],
+ [
+ 11.162119987692769,
+ 55.319159243735825
+ ],
+ [
+ 11.152842643463767,
+ 55.328436585299954
+ ],
+ [
+ 11.161957227315348,
+ 55.33087800034785
+ ],
+ [
+ 11.17644290506829,
+ 55.331854559723794
+ ],
+ [
+ 11.188243036078866,
+ 55.335272539427166
+ ],
+ [
+ 11.190114780025699,
+ 55.34544505427165
+ ],
+ [
+ 11.182302279411676,
+ 55.35040924076694
+ ],
+ [
+ 11.167979363291282,
+ 55.35276927608878
+ ],
+ [
+ 11.152598502213515,
+ 55.35321684502423
+ ],
+ [
+ 11.142344597622206,
+ 55.352280998613985
+ ],
+ [
+ 11.13542727982451,
+ 55.34686920613249
+ ],
+ [
+ 11.130869986958954,
+ 55.338446344915894
+ ],
+ [
+ 11.123545769068011,
+ 55.331854559723794
+ ],
+ [
+ 11.108164910067913,
+ 55.331854559723794
+ ],
+ [
+ 11.101898632508913,
+ 55.33633045508252
+ ],
+ [
+ 11.087738478802478,
+ 55.358547287253366
+ ],
+ [
+ 11.087738477452657,
+ 55.365952871833585
+ ],
+ [
+ 11.114593944821046,
+ 55.364406630438026
+ ],
+ [
+ 11.128916864352218,
+ 55.36517976938758
+ ],
+ [
+ 11.13892662793693,
+ 55.369370823698866
+ ],
+ [
+ 11.147715692449738,
+ 55.37592195739491
+ ],
+ [
+ 11.159190299796421,
+ 55.38129303383966
+ ],
+ [
+ 11.171722851791724,
+ 55.38507721379276
+ ],
+ [
+ 11.1832788436571,
+ 55.386420006536554
+ ],
+ [
+ 11.204356314555604,
+ 55.39183175774402
+ ],
+ [
+ 11.214203321572922,
+ 55.40379467004302
+ ],
+ [
+ 11.213145378163368,
+ 55.41575754517551
+ ],
+ [
+ 11.20093834739421,
+ 55.42121003216065
+ ],
+ [
+ 11.20484459829513,
+ 55.431341877662916
+ ],
+ [
+ 11.196462434363138,
+ 55.45384347472771
+ ],
+ [
+ 11.176442905396529,
+ 55.489447337187244
+ ],
+ [
+ 11.163910351439373,
+ 55.50189849907388
+ ],
+ [
+ 11.151133661296738,
+ 55.510321369717325
+ ],
+ [
+ 11.135590041493469,
+ 55.51512280481872
+ ],
+ [
+ 11.115000846499239,
+ 55.51679107768524
+ ],
+ [
+ 11.097178580186652,
+ 55.50950753472209
+ ],
+ [
+ 11.08578535243469,
+ 55.5075544336253
+ ],
+ [
+ 11.080902538693184,
+ 55.513088269088
+ ],
+ [
+ 11.082774284855729,
+ 55.52655670646411
+ ],
+ [
+ 11.088552279424825,
+ 55.53384022988425
+ ],
+ [
+ 11.098155145501051,
+ 55.53668855994345
+ ],
+ [
+ 11.111582877548551,
+ 55.53729887960059
+ ],
+ [
+ 11.137950064129111,
+ 55.54291411020055
+ ],
+ [
+ 11.148448111862042,
+ 55.55731841497261
+ ],
+ [
+ 11.14568118661247,
+ 55.57611726147084
+ ],
+ [
+ 11.13209068915189,
+ 55.59528227445812
+ ],
+ [
+ 11.087901239809568,
+ 55.626939214192994
+ ],
+ [
+ 11.03923587259579,
+ 55.64345937269504
+ ],
+ [
+ 10.929942254051111,
+ 55.66014232054865
+ ],
+ [
+ 10.929942254229687,
+ 55.66762930146802
+ ],
+ [
+ 11.087738477051841,
+ 55.66014232054865
+ ],
+ [
+ 11.071055537640907,
+ 55.6769473610169
+ ],
+ [
+ 11.008636916723791,
+ 55.69228752038356
+ ],
+ [
+ 10.984629753302741,
+ 55.70111724754837
+ ],
+ [
+ 10.97429446804269,
+ 55.710842201149234
+ ],
+ [
+ 10.96827233373578,
+ 55.718410567566785
+ ],
+ [
+ 10.960215689223787,
+ 55.72435128498792
+ ],
+ [
+ 10.943614128486258,
+ 55.72907135049285
+ ],
+ [
+ 10.891449413292984,
+ 55.73126860361674
+ ],
+ [
+ 10.875498892465103,
+ 55.73651762844807
+ ],
+ [
+ 10.87549889061176,
+ 55.74274319446913
+ ],
+ [
+ 11.033050976482048,
+ 55.72907135049285
+ ],
+ [
+ 11.065684439619737,
+ 55.73224516936971
+ ],
+ [
+ 11.128184443165532,
+ 55.746323982814346
+ ],
+ [
+ 11.16277103044937,
+ 55.749497793718824
+ ],
+ [
+ 11.162771026522215,
+ 55.74274319446913
+ ],
+ [
+ 11.151703318215032,
+ 55.743638384717784
+ ],
+ [
+ 11.141123893670155,
+ 55.74323150852776
+ ],
+ [
+ 11.13086998976822,
+ 55.74103426562653
+ ],
+ [
+ 11.121836783430332,
+ 55.73651762844807
+ ],
+ [
+ 11.137543164890147,
+ 55.71906158984218
+ ],
+ [
+ 11.163422074181126,
+ 55.70538974944428
+ ],
+ [
+ 11.192637567887518,
+ 55.70034416279445
+ ],
+ [
+ 11.218028190666626,
+ 55.708563540511534
+ ],
+ [
+ 11.176442904475289,
+ 55.72907135049285
+ ],
+ [
+ 11.190114778420654,
+ 55.73651762844807
+ ],
+ [
+ 11.199473507381192,
+ 55.73110589640942
+ ],
+ [
+ 11.216156448434358,
+ 55.72874586045725
+ ],
+ [
+ 11.252207878471767,
+ 55.72907135049285
+ ],
+ [
+ 11.268402538873948,
+ 55.73273345427919
+ ],
+ [
+ 11.302419468288994,
+ 55.74652742437044
+ ],
+ [
+ 11.336680532415349,
+ 55.7509219152723
+ ],
+ [
+ 11.352224157925917,
+ 55.75580478521013
+ ],
+ [
+ 11.364024286219097,
+ 55.76512280157037
+ ],
+ [
+ 11.368907094072032,
+ 55.779974646830766
+ ],
+ [
+ 11.369965039831863,
+ 55.787787175267454
+ ],
+ [
+ 11.374685095517313,
+ 55.80158115820151
+ ],
+ [
+ 11.375661653215527,
+ 55.807928759637534
+ ],
+ [
+ 11.372080928609416,
+ 55.8158633851437
+ ],
+ [
+ 11.354665563245073,
+ 55.83022697813047
+ ],
+ [
+ 11.348399286594058,
+ 55.83832430242831
+ ],
+ [
+ 11.37387129011109,
+ 55.82876211214201
+ ],
+ [
+ 11.398448116452467,
+ 55.82640212043672
+ ],
+ [
+ 11.421722852737085,
+ 55.83197663754851
+ ],
+ [
+ 11.443369989115974,
+ 55.84577058640335
+ ],
+ [
+ 11.477549678900688,
+ 55.84153884374579
+ ],
+ [
+ 11.505218944043222,
+ 55.868719774122404
+ ],
+ [
+ 11.510508663296642,
+ 55.908270608436816
+ ],
+ [
+ 11.477386911672125,
+ 55.9413515961765
+ ],
+ [
+ 11.448415561504708,
+ 55.948187572141286
+ ],
+ [
+ 11.384287957422004,
+ 55.95083242832225
+ ],
+ [
+ 11.30453535337036,
+ 55.96784090113025
+ ],
+ [
+ 11.28191165370975,
+ 55.978949272671926
+ ],
+ [
+ 11.272634310821141,
+ 55.996568099151304
+ ],
+ [
+ 11.400563999022541,
+ 55.961981522456576
+ ],
+ [
+ 11.426931186467694,
+ 55.96552155677095
+ ],
+ [
+ 11.581065303262973,
+ 55.95184970985653
+ ],
+ [
+ 11.599375843892053,
+ 55.94208402074429
+ ],
+ [
+ 11.640879753545843,
+ 55.943426820448494
+ ],
+ [
+ 11.711436391587137,
+ 55.95563383197502
+ ],
+ [
+ 11.747813350211619,
+ 55.96824782086688
+ ],
+ [
+ 11.768321159506486,
+ 55.970526429111075
+ ],
+ [
+ 11.779795769892973,
+ 55.9618594508372
+ ],
+ [
+ 11.775401241317716,
+ 55.95184970985653
+ ],
+ [
+ 11.747325063598433,
+ 55.93280668414642
+ ],
+ [
+ 11.738780146366025,
+ 55.92084384052331
+ ],
+ [
+ 11.750498891653283,
+ 55.91038643668394
+ ],
+ [
+ 11.743988476746129,
+ 55.9037539704752
+ ],
+ [
+ 11.727875195707872,
+ 55.902289127090796
+ ],
+ [
+ 11.711436390421193,
+ 55.907171906321324
+ ],
+ [
+ 11.713552280855303,
+ 55.91083406052398
+ ],
+ [
+ 11.717621292361741,
+ 55.92084384052331
+ ],
+ [
+ 11.685801629605821,
+ 55.919867261050065
+ ],
+ [
+ 11.67286217531134,
+ 55.91592031811254
+ ],
+ [
+ 11.66309654643597,
+ 55.907171906321324
+ ],
+ [
+ 11.7126571005798,
+ 55.85809971272858
+ ],
+ [
+ 11.731618684020786,
+ 55.83071521626328
+ ],
+ [
+ 11.717621288470731,
+ 55.81098052072379
+ ],
+ [
+ 11.702403190897638,
+ 55.809475000975326
+ ],
+ [
+ 11.671234570729771,
+ 55.814276431289564
+ ],
+ [
+ 11.656260611478762,
+ 55.81098052072379
+ ],
+ [
+ 11.637950065575446,
+ 55.79608795372773
+ ],
+ [
+ 11.626231312786677,
+ 55.790594761632434
+ ],
+ [
+ 11.607758009686085,
+ 55.79047272992114
+ ],
+ [
+ 11.607758005685001,
+ 55.7830264005488
+ ],
+ [
+ 11.62427819178041,
+ 55.78001537486348
+ ],
+ [
+ 11.673594593666225,
+ 55.7830264005488
+ ],
+ [
+ 11.687022329049313,
+ 55.786444373429354
+ ],
+ [
+ 11.702810090195634,
+ 55.79340239294499
+ ],
+ [
+ 11.720388217438149,
+ 55.7986514383765
+ ],
+ [
+ 11.738780145430493,
+ 55.7973493652674
+ ],
+ [
+ 11.745453324799955,
+ 55.78937412888774
+ ],
+ [
+ 11.75586998988046,
+ 55.77293530774021
+ ],
+ [
+ 11.763682484324447,
+ 55.75682197037696
+ ],
+ [
+ 11.762705925473561,
+ 55.749497793718824
+ ],
+ [
+ 11.682302283299082,
+ 55.74127842011318
+ ],
+ [
+ 11.664561394260977,
+ 55.7371279996229
+ ],
+ [
+ 11.65723717848307,
+ 55.73086175604403
+ ],
+ [
+ 11.669932489921235,
+ 55.72284578931272
+ ],
+ [
+ 11.68433678513369,
+ 55.7216250683415
+ ],
+ [
+ 11.729258658606266,
+ 55.72357817805957
+ ],
+ [
+ 11.738780147206834,
+ 55.7259789400447
+ ],
+ [
+ 11.746348502199954,
+ 55.731024463025584
+ ],
+ [
+ 11.763031447416793,
+ 55.721380941129695
+ ],
+ [
+ 11.787282747251759,
+ 55.70111724754837
+ ],
+ [
+ 11.794444209229011,
+ 55.687933683190366
+ ],
+ [
+ 11.793793164618812,
+ 55.6830101542037
+ ],
+ [
+ 11.78793379281189,
+ 55.6769473610169
+ ],
+ [
+ 11.779795769055085,
+ 55.66014232054865
+ ],
+ [
+ 11.7976180390053,
+ 55.66962315388267
+ ],
+ [
+ 11.81055748698889,
+ 55.67910389794443
+ ],
+ [
+ 11.852305533166701,
+ 55.73090239285704
+ ],
+ [
+ 11.854404201249086,
+ 55.73215351630363
+ ],
+ [
+ 11.855638462155492,
+ 55.73205068609124
+ ],
+ [
+ 11.893517283879806,
+ 55.733652624758726
+ ],
+ [
+ 11.905764598997523,
+ 55.736882413791534
+ ],
+ [
+ 11.907521598852634,
+ 55.739440403447716
+ ],
+ [
+ 11.909071894654756,
+ 55.74062899228263
+ ],
+ [
+ 11.92488488146605,
+ 55.73794180453806
+ ],
+ [
+ 11.97266686693785,
+ 55.72142165558555
+ ],
+ [
+ 11.965342646706738,
+ 55.70490146054979
+ ],
+ [
+ 11.958181189718621,
+ 55.69342686150855
+ ],
+ [
+ 11.923024939680078,
+ 55.67413975225489
+ ],
+ [
+ 11.910166863055695,
+ 55.66014232054865
+ ],
+ [
+ 11.930430533460264,
+ 55.65761951783294
+ ],
+ [
+ 11.959157747947033,
+ 55.6651065114792
+ ],
+ [
+ 11.982188346638525,
+ 55.678900454455984
+ ],
+ [
+ 11.985199411719364,
+ 55.69489163686437
+ ],
+ [
+ 12.001149938843394,
+ 55.69867587581191
+ ],
+ [
+ 12.017588739538805,
+ 55.69570548603143
+ ],
+ [
+ 12.031911651328809,
+ 55.68720120317136
+ ],
+ [
+ 12.040375194487208,
+ 55.67446522124108
+ ],
+ [
+ 12.009287958200504,
+ 55.68048738373947
+ ],
+ [
+ 12.005137563542096,
+ 55.67816801195262
+ ],
+ [
+ 12.011403843544446,
+ 55.66901277710518
+ ],
+ [
+ 12.026621942196272,
+ 55.66205476668677
+ ],
+ [
+ 12.060313347306188,
+ 55.65399811698287
+ ],
+ [
+ 12.069102408731863,
+ 55.674953505657285
+ ],
+ [
+ 12.09205162920757,
+ 55.706732491013874
+ ],
+ [
+ 12.095062695432182,
+ 55.72907135049285
+ ],
+ [
+ 12.076088193567447,
+ 55.77892115166933
+ ],
+ [
+ 12.076296830281649,
+ 55.77892123152771
+ ],
+ [
+ 12.112935421647217,
+ 55.77881789478984
+ ],
+ [
+ 12.151589390997403,
+ 55.75757885450273
+ ],
+ [
+ 12.16786746484549,
+ 55.748561293969836
+ ],
+ [
+ 12.240472855667889,
+ 55.741249118363235
+ ],
+ [
+ 12.22667524467127,
+ 55.73380766988351
+ ],
+ [
+ 12.228380566775812,
+ 55.72838165083446
+ ],
+ [
+ 12.230550979544354,
+ 55.72584952862235
+ ],
+ [
+ 12.232876418085002,
+ 55.72347239903002
+ ],
+ [
+ 12.236442089109945,
+ 55.72119860434869
+ ],
+ [
+ 12.255717397844352,
+ 55.7145323448421
+ ],
+ [
+ 12.258042840268807,
+ 55.712206937012766
+ ],
+ [
+ 12.254942249755779,
+ 55.70776272342719
+ ],
+ [
+ 12.215668167435675,
+ 55.691484701978915
+ ],
+ [
+ 12.192672150345633,
+ 55.686342855716745
+ ],
+ [
+ 12.157945590272014,
+ 55.67055572237396
+ ],
+ [
+ 12.165542027708844,
+ 55.66430289898831
+ ],
+ [
+ 12.166317174396083,
+ 55.662442541528854
+ ],
+ [
+ 12.167092319578037,
+ 55.660168775752645
+ ],
+ [
+ 12.166627230340989,
+ 55.65771412477824
+ ],
+ [
+ 12.165076940026612,
+ 55.65469107687858
+ ],
+ [
+ 12.163836704744895,
+ 55.65071199109456
+ ],
+ [
+ 12.162596471078688,
+ 55.64373567000555
+ ],
+ [
+ 12.164456822637751,
+ 55.638413002986375
+ ],
+ [
+ 12.16910769943336,
+ 55.63404633094207
+ ],
+ [
+ 12.17019290475286,
+ 55.62221242534733
+ ],
+ [
+ 12.201095412256285,
+ 55.61428009919232
+ ],
+ [
+ 12.220164015831859,
+ 55.61490022010679
+ ],
+ [
+ 12.23985273542413,
+ 55.62156646767534
+ ],
+ [
+ 12.251996697700063,
+ 55.616347150761534
+ ],
+ [
+ 12.281297236281251,
+ 55.61898266056878
+ ],
+ [
+ 12.316695592103654,
+ 55.61655387320317
+ ],
+ [
+ 12.359121940002405,
+ 55.604616586275746
+ ],
+ [
+ 12.359205489553343,
+ 55.60459309122142
+ ]
+ ]
+ ]
+ ]
+ },
+ "type": "Feature",
+ "properties": {
+ "source": "https://simplemaps.com",
+ "id": "DK85",
+ "name": "Sjaælland"
+ },
+ "id": 4
+ },
+ {
+ "geometry": {
+ "type": "MultiPolygon",
+ "coordinates": [
+ [
+ [
+ [
+ 11.97266686693785,
+ 55.72142165558555
+ ],
+ [
+ 11.92488488146605,
+ 55.73794180453806
+ ],
+ [
+ 11.909071894654756,
+ 55.74062899228263
+ ],
+ [
+ 11.907521598852634,
+ 55.739440403447716
+ ],
+ [
+ 11.905764598997523,
+ 55.736882413791534
+ ],
+ [
+ 11.893517283879806,
+ 55.733652624758726
+ ],
+ [
+ 11.855638462155492,
+ 55.73205068609124
+ ],
+ [
+ 11.854404201249086,
+ 55.73215351630363
+ ],
+ [
+ 11.85547935922123,
+ 55.73281480861671
+ ],
+ [
+ 11.853282095926788,
+ 55.73997629338215
+ ],
+ [
+ 11.844004750473028,
+ 55.75267158209198
+ ],
+ [
+ 11.84180749178994,
+ 55.75641514351241
+ ],
+ [
+ 11.844981314322053,
+ 55.767889699400115
+ ],
+ [
+ 11.852712435146506,
+ 55.78229400356512
+ ],
+ [
+ 11.869151235990493,
+ 55.80475493017968
+ ],
+ [
+ 11.884043818869664,
+ 55.813666111260304
+ ],
+ [
+ 11.919444209463188,
+ 55.81940341118172
+ ],
+ [
+ 11.937347855385827,
+ 55.82526279518501
+ ],
+ [
+ 11.953623892887538,
+ 55.853501684118406
+ ],
+ [
+ 11.933116085262888,
+ 55.88613518994075
+ ],
+ [
+ 11.909515824890143,
+ 55.915472761127745
+ ],
+ [
+ 11.91684004376976,
+ 55.93451572890702
+ ],
+ [
+ 11.934743686453116,
+ 55.933661204429285
+ ],
+ [
+ 11.969981316259638,
+ 55.91258372853133
+ ],
+ [
+ 11.988617380336466,
+ 55.907171906321324
+ ],
+ [
+ 11.998220246209474,
+ 55.90281808357406
+ ],
+ [
+ 11.997894724984262,
+ 55.89276762537649
+ ],
+ [
+ 11.993825713441622,
+ 55.881537142337585
+ ],
+ [
+ 11.992035350516625,
+ 55.873724662544774
+ ],
+ [
+ 11.996348504229038,
+ 55.86107005623102
+ ],
+ [
+ 12.000173374354311,
+ 55.85651277519341
+ ],
+ [
+ 12.004893426425802,
+ 55.85346101888511
+ ],
+ [
+ 12.012461786171466,
+ 55.84577058640335
+ ],
+ [
+ 12.031260612244127,
+ 55.816595762646806
+ ],
+ [
+ 12.049001500090608,
+ 55.77533602532673
+ ],
+ [
+ 12.051931184614052,
+ 55.73663970251359
+ ],
+ [
+ 12.026133656947984,
+ 55.715399451221316
+ ],
+ [
+ 11.99000084644076,
+ 55.722805074527386
+ ],
+ [
+ 11.97266686693785,
+ 55.72142165558555
+ ]
+ ]
+ ],
+ [
+ [
+ [
+ 12.359205489553343,
+ 55.60459309122142
+ ],
+ [
+ 12.359121940002405,
+ 55.604616586275746
+ ],
+ [
+ 12.316695592103654,
+ 55.61655387320317
+ ],
+ [
+ 12.281297236281251,
+ 55.61898266056878
+ ],
+ [
+ 12.251996697700063,
+ 55.616347150761534
+ ],
+ [
+ 12.23985273542413,
+ 55.62156646767534
+ ],
+ [
+ 12.220164015831859,
+ 55.61490022010679
+ ],
+ [
+ 12.201095412256285,
+ 55.61428009919232
+ ],
+ [
+ 12.17019290475286,
+ 55.62221242534733
+ ],
+ [
+ 12.16910769943336,
+ 55.63404633094207
+ ],
+ [
+ 12.164456822637751,
+ 55.638413002986375
+ ],
+ [
+ 12.162596471078688,
+ 55.64373567000555
+ ],
+ [
+ 12.163836704744895,
+ 55.65071199109456
+ ],
+ [
+ 12.165076940026612,
+ 55.65469107687858
+ ],
+ [
+ 12.166627230340989,
+ 55.65771412477824
+ ],
+ [
+ 12.167092319578037,
+ 55.660168775752645
+ ],
+ [
+ 12.166317174396083,
+ 55.662442541528854
+ ],
+ [
+ 12.165542027708844,
+ 55.66430289898831
+ ],
+ [
+ 12.157945590272014,
+ 55.67055572237396
+ ],
+ [
+ 12.192672150345633,
+ 55.686342855716745
+ ],
+ [
+ 12.215668167435675,
+ 55.691484701978915
+ ],
+ [
+ 12.254942249755779,
+ 55.70776272342719
+ ],
+ [
+ 12.258042840268807,
+ 55.712206937012766
+ ],
+ [
+ 12.255717397844352,
+ 55.7145323448421
+ ],
+ [
+ 12.236442089109945,
+ 55.72119860434869
+ ],
+ [
+ 12.232876418085002,
+ 55.72347239903002
+ ],
+ [
+ 12.230550979544354,
+ 55.72584952862235
+ ],
+ [
+ 12.228380566775812,
+ 55.72838165083446
+ ],
+ [
+ 12.22667524467127,
+ 55.73380766988351
+ ],
+ [
+ 12.240472855667889,
+ 55.741249118363235
+ ],
+ [
+ 12.16786746484549,
+ 55.748561293969836
+ ],
+ [
+ 12.151589390997403,
+ 55.75757885450273
+ ],
+ [
+ 12.112935421647217,
+ 55.77881789478984
+ ],
+ [
+ 12.076296830281649,
+ 55.77892123152771
+ ],
+ [
+ 12.076088193567447,
+ 55.77892115166933
+ ],
+ [
+ 12.053477411693088,
+ 55.83832430242831
+ ],
+ [
+ 12.053477408509035,
+ 55.873724662544774
+ ],
+ [
+ 12.04981529655972,
+ 55.8880068696489
+ ],
+ [
+ 12.026133656513036,
+ 55.9413515961765
+ ],
+ [
+ 12.01059003853187,
+ 55.95526763074563
+ ],
+ [
+ 11.992442254012786,
+ 55.96059804912358
+ ],
+ [
+ 11.970713737951074,
+ 55.95986562652268
+ ],
+ [
+ 11.894297724884726,
+ 55.942287530289725
+ ],
+ [
+ 11.865489132264011,
+ 55.9393578470896
+ ],
+ [
+ 11.848643425522397,
+ 55.948187572141286
+ ],
+ [
+ 11.865570507074159,
+ 55.972723681081824
+ ],
+ [
+ 11.917328319325891,
+ 55.99762602763652
+ ],
+ [
+ 12.012461783142959,
+ 56.03070708000749
+ ],
+ [
+ 12.16537519656009,
+ 56.103705150347466
+ ],
+ [
+ 12.245127802469614,
+ 56.12872958241207
+ ],
+ [
+ 12.328461134799328,
+ 56.12689851252647
+ ],
+ [
+ 12.41488691781484,
+ 56.09902580934033
+ ],
+ [
+ 12.486094599831008,
+ 56.09902580934033
+ ],
+ [
+ 12.505381705184858,
+ 56.0931663911338
+ ],
+ [
+ 12.540375199281172,
+ 56.075384863415465
+ ],
+ [
+ 12.570078971643118,
+ 56.06826406169815
+ ],
+ [
+ 12.621918168814245,
+ 56.04376865715721
+ ],
+ [
+ 12.608409053024259,
+ 56.02802157991647
+ ],
+ [
+ 12.567393424800635,
+ 55.996568099151304
+ ],
+ [
+ 12.552744986683214,
+ 55.979681695774694
+ ],
+ [
+ 12.525563994809978,
+ 55.9398867562898
+ ],
+ [
+ 12.512705927504612,
+ 55.9282901293565
+ ],
+ [
+ 12.512705927522013,
+ 55.92084384052331
+ ],
+ [
+ 12.540293818392625,
+ 55.8982608312804
+ ],
+ [
+ 12.554453972538004,
+ 55.88287995500568
+ ],
+ [
+ 12.560557486278357,
+ 55.869940480980674
+ ],
+ [
+ 12.565114780986613,
+ 55.8511416781805
+ ],
+ [
+ 12.576019728613252,
+ 55.829169027006905
+ ],
+ [
+ 12.589691601889854,
+ 55.809475000975326
+ ],
+ [
+ 12.602224156535662,
+ 55.7973493652674
+ ],
+ [
+ 12.59799238827741,
+ 55.787583765764076
+ ],
+ [
+ 12.595876496640711,
+ 55.77423735935655
+ ],
+ [
+ 12.594737172788962,
+ 55.75299711642062
+ ],
+ [
+ 12.589040562598806,
+ 55.73574454789345
+ ],
+ [
+ 12.587901237409062,
+ 55.72907135049285
+ ],
+ [
+ 12.58920332213372,
+ 55.7242699345637
+ ],
+ [
+ 12.592784046443004,
+ 55.71914293785914
+ ],
+ [
+ 12.602224154625437,
+ 55.708563540511534
+ ],
+ [
+ 12.58122806481676,
+ 55.693060649616676
+ ],
+ [
+ 12.541026238752604,
+ 55.64887116805303
+ ],
+ [
+ 12.503265821872343,
+ 55.63401928120328
+ ],
+ [
+ 12.496755405911227,
+ 55.621771560576434
+ ],
+ [
+ 12.48715254091024,
+ 55.609849359576096
+ ],
+ [
+ 12.465017123708822,
+ 55.60553620669113
+ ],
+ [
+ 12.448578321394276,
+ 55.60887279272605
+ ],
+ [
+ 12.436778190301723,
+ 55.61440663639387
+ ],
+ [
+ 12.423675978511822,
+ 55.61884187831393
+ ],
+ [
+ 12.403575063999211,
+ 55.61920805602437
+ ],
+ [
+ 12.359205489553343,
+ 55.60459309122142
+ ]
+ ]
+ ],
+ [
+ [
+ [
+ 15.151377799699626,
+ 55.13320546863723
+ ],
+ [
+ 15.146657748589611,
+ 55.12824128263833
+ ],
+ [
+ 15.144541863021463,
+ 55.12482330916832
+ ],
+ [
+ 15.142344597514276,
+ 55.12213776604713
+ ],
+ [
+ 15.137705925117812,
+ 55.119533596927006
+ ],
+ [
+ 15.145030143887588,
+ 55.10488515711437
+ ],
+ [
+ 15.14771569065473,
+ 55.09560781228082
+ ],
+ [
+ 15.144704623448108,
+ 55.08710358652932
+ ],
+ [
+ 15.134287956713033,
+ 55.074855858736655
+ ],
+ [
+ 15.112803581866446,
+ 55.05463287894425
+ ],
+ [
+ 15.10661868591341,
+ 55.0446638033151
+ ],
+ [
+ 15.110362174967346,
+ 55.0363630227412
+ ],
+ [
+ 15.110362174882985,
+ 55.03021881607416
+ ],
+ [
+ 15.113242231028034,
+ 55.01829487422116
+ ],
+ [
+ 15.097353256984684,
+ 55.007822962878265
+ ],
+ [
+ 15.08668593561237,
+ 54.99921337492748
+ ],
+ [
+ 15.072666222229683,
+ 54.991637976082735
+ ],
+ [
+ 15.023663845453589,
+ 54.999339321486666
+ ],
+ [
+ 14.98286981900447,
+ 55.001013408035035
+ ],
+ [
+ 14.927265174002398,
+ 55.013691882018705
+ ],
+ [
+ 14.858606616102898,
+ 55.038226353826566
+ ],
+ [
+ 14.780832726085134,
+ 55.051128327687174
+ ],
+ [
+ 14.699554883663037,
+ 55.089178775265076
+ ],
+ [
+ 14.684172686869642,
+ 55.10134680694097
+ ],
+ [
+ 14.697044746375662,
+ 55.122653717045786
+ ],
+ [
+ 14.701426629604299,
+ 55.16469961589287
+ ],
+ [
+ 14.705739778939227,
+ 55.22565337842627
+ ],
+ [
+ 14.70923912897663,
+ 55.235825913810565
+ ],
+ [
+ 14.717539910126169,
+ 55.246771553017695
+ ],
+ [
+ 14.73707116072436,
+ 55.2666690178351
+ ],
+ [
+ 14.745127800978143,
+ 55.281561598872194
+ ],
+ [
+ 14.752126497751574,
+ 55.298773503001954
+ ],
+ [
+ 14.760915559844317,
+ 55.30927154271457
+ ],
+ [
+ 14.774587434970488,
+ 55.303900449734755
+ ],
+ [
+ 14.814952019097953,
+ 55.27041250278426
+ ],
+ [
+ 14.828949414029816,
+ 55.26170481223073
+ ],
+ [
+ 14.85142220523577,
+ 55.24819986088343
+ ],
+ [
+ 14.889008008366588,
+ 55.233221741944156
+ ],
+ [
+ 14.926799098885601,
+ 55.21579872612391
+ ],
+ [
+ 14.954354432958475,
+ 55.219254408720374
+ ],
+ [
+ 14.973399284455681,
+ 55.21613189867759
+ ],
+ [
+ 14.989756707522488,
+ 55.196030996145524
+ ],
+ [
+ 15.000498894293779,
+ 55.18842194432937
+ ],
+ [
+ 15.012868686289377,
+ 55.18382396229323
+ ],
+ [
+ 15.08454287843215,
+ 55.15457949151666
+ ],
+ [
+ 15.131602409617736,
+ 55.14545318899037
+ ],
+ [
+ 15.151377799699626,
+ 55.13320546863723
+ ]
+ ]
+ ],
+ [
+ [
+ [
+ 12.615733271804235,
+ 55.688137137099474
+ ],
+ [
+ 12.673838738720733,
+ 55.60553620669113
+ ],
+ [
+ 12.674571158594317,
+ 55.60126372594711
+ ],
+ [
+ 12.639414911684735,
+ 55.57827384569545
+ ],
+ [
+ 12.624196810743124,
+ 55.5756289706033
+ ],
+ [
+ 12.605723503426296,
+ 55.56928131163867
+ ],
+ [
+ 12.588877797997286,
+ 55.561346728257355
+ ],
+ [
+ 12.577972850907424,
+ 55.55402251776397
+ ],
+ [
+ 12.556325715120503,
+ 55.55048247136016
+ ],
+ [
+ 12.534841340304386,
+ 55.56932199106652
+ ],
+ [
+ 12.520518426804275,
+ 55.596747153960536
+ ],
+ [
+ 12.520192902980352,
+ 55.61920805602437
+ ],
+ [
+ 12.531016472324815,
+ 55.63190339004734
+ ],
+ [
+ 12.543955926921893,
+ 55.64036694701188
+ ],
+ [
+ 12.556814000024898,
+ 55.64695874495491
+ ],
+ [
+ 12.567393425319057,
+ 55.65399811698287
+ ],
+ [
+ 12.578135615085353,
+ 55.66644930547757
+ ],
+ [
+ 12.595062697527752,
+ 55.69135164424099
+ ],
+ [
+ 12.608409049199672,
+ 55.70111724754837
+ ],
+ [
+ 12.612559440804974,
+ 55.69513580718399
+ ],
+ [
+ 12.614105666025866,
+ 55.69245027154784
+ ],
+ [
+ 12.615733271804235,
+ 55.688137137099474
+ ]
+ ]
+ ]
+ ]
+ },
+ "type": "Feature",
+ "properties": {
+ "source": "https://simplemaps.com",
+ "id": "DK84",
+ "name": "Hovedstaden"
+ },
+ "id": 5
+ }
+ ]
+}
\ No newline at end of file
diff --git a/src/ensemble_rl_example.py b/src/ensemble_rl_example.py
deleted file mode 100644
index 8f13326f67dc2f763a40417905885db61b79c95f..0000000000000000000000000000000000000000
--- a/src/ensemble_rl_example.py
+++ /dev/null
@@ -1,346 +0,0 @@
-from collections import OrderedDict, deque, namedtuple
-from typing import Iterator, List, Tuple
-import gymnasium as gym
-import lightning.pytorch as pl
-import numpy as np
-import torch
-from lightning.pytorch.loggers import CSVLogger
-from torch import Tensor, nn
-from torch.optim import Adam, Optimizer
-from torch.utils.data import DataLoader
-from torch.utils.data.dataset import IterableDataset
-
-use_cuda = torch.cuda.is_available()
-
-if use_cuda:
- accelerator = 'gpu'
- torch.set_float32_matmul_precision('high')
-else:
- accelerator = 'cpu'
-
-
-class DQN(nn.Module):
- """Simple MLP network."""
-
- def __init__(self, obs_size: int, n_actions: int, hidden_size: int = 128):
- """
- Args:
- obs_size: observation/state size of the environment
- n_actions: number of discrete actions available in the environment
- hidden_size: size of hidden layers
- """
- super().__init__()
- self.net = nn.Sequential(
- nn.Linear(obs_size, hidden_size),
- nn.ReLU(),
- nn.Linear(hidden_size, n_actions),
- )
-
- def forward(self, x):
- return self.net(x.float())
-
-# Named tuple for storing experience steps gathered in training
-Experience = namedtuple(
- "Experience",
- field_names=["state", "action", "reward", "done", "new_state"],
-)
-
-class ReplayBuffer:
- """Replay Buffer for storing past experiences allowing the agent to learn from them.
-
- Args:
- capacity: size of the buffer
- """
-
- def __init__(self, capacity: int) -> None:
- self.buffer = deque(maxlen=capacity)
-
- def __len__(self) -> int:
- return len(self.buffer)
-
- def append(self, experience: Experience) -> None:
- """Add experience to the buffer.
-
- Args:
- experience: tuple (state, action, reward, done, new_state)
- """
- self.buffer.append(experience)
-
- def sample(self, batch_size: int) -> Tuple:
- indices = np.random.choice(len(self.buffer), batch_size, replace=False)
- states, actions, rewards, dones, next_states = zip(*(self.buffer[idx] for idx in indices))
-
- return (
- np.array(states),
- np.array(actions),
- np.array(rewards, dtype=np.float32),
- np.array(dones, dtype=bool),
- np.array(next_states),
- )
-
-class RLDataset(IterableDataset):
- """Iterable Dataset containing the ExperienceBuffer which will be updated with new experiences during training.
-
- Args:
- buffer: replay buffer
- sample_size: number of experiences to sample at a time
- """
-
- def __init__(self, buffer: ReplayBuffer, sample_size: int = 200) -> None:
- self.buffer = buffer
- self.sample_size = sample_size
-
- def __iter__(self) -> Iterator[Tuple]:
- states, actions, rewards, dones, new_states = self.buffer.sample(self.sample_size)
- for i in range(len(dones)):
- yield states[i], actions[i], rewards[i], dones[i], new_states[i]
-
-class Agent:
- """Base Agent class handeling the interaction with the environment."""
-
- def __init__(self, env: gym.Env, replay_buffer: ReplayBuffer) -> None:
- """
- Args:
- env: training environment
- replay_buffer: replay buffer storing experiences
- """
- self.env = env
- self.replay_buffer = replay_buffer
- self.reset()
- self.state = self.env.reset()
-
- def reset(self) -> None:
- """Resents the environment and updates the state."""
- self.state, _ = self.env.reset()
-
- def get_action(self, net: nn.Module, epsilon: float, device: str) -> int:
- """Using the given network, decide what action to carry out using an epsilon-greedy policy.
-
- Args:
- net: DQN network
- epsilon: value to determine likelihood of taking a random action
- device: current device
-
- Returns:
- action
- """
- if np.random.random() < epsilon:
- action = self.env.action_space.sample()
- else:
- state = torch.tensor([self.state])
-
- if device not in ["cpu"]:
- state = state.cuda(device)
-
- q_values = net(state)
- _, action = torch.max(q_values, dim=1)
- action = int(action.item())
-
- return action
-
- @torch.no_grad()
- def play_step(
- self,
- net: nn.Module,
- epsilon: float = 0.0,
- device: str = "cpu",
- ) -> Tuple[float, bool]:
- """Carries out a single interaction step between the agent and the environment.
-
- Args:
- net: DQN network
- epsilon: value to determine likelihood of taking a random action
- device: current device
-
- Returns:
- reward, done
- """
-
- action = self.get_action(net, epsilon, device)
-
- # do step in the environment
- #new_state, reward, done, _ = self.env.step(action)
- new_state, reward, done, truncated, info = self.env.step(action)
-
- exp = Experience(self.state, action, reward, done, new_state)
-
- self.replay_buffer.append(exp)
-
- self.state = new_state
- if done:
- self.reset()
- return reward, done
-
-class DQNLightning(pl.LightningModule):
- """Basic DQN Model."""
-
- def __init__(
- self,
- batch_size: int = 16,
- lr: float = 1e-2,
- env: str = "CartPole-v1",
- gamma: float = 0.99,
- sync_rate: int = 10,
- replay_size: int = 1000,
- warm_start_size: int = 1000,
- eps_last_frame: int = 1000,
- eps_start: float = 1.0,
- eps_end: float = 0.01,
- episode_length: int = 200,
- warm_start_steps: int = 1000,
- ) -> None:
- """
- Args:
- batch_size: size of the batches")
- lr: learning rate
- env: gym environment tag
- gamma: discount factor
- sync_rate: how many frames do we update the target network
- replay_size: capacity of the replay buffer
- warm_start_size: how many samples do we use to fill our buffer at the start of training
- eps_last_frame: what frame should epsilon stop decaying
- eps_start: starting value of epsilon
- eps_end: final value of epsilon
- episode_length: max length of an episode
- warm_start_steps: max episode reward in the environment
- """
- super().__init__()
- self.save_hyperparameters()
-
- self.env = gym.make(self.hparams.env)
- obs_size = self.env.observation_space.shape[0]
- n_actions = self.env.action_space.n
-
- self.net = DQN(obs_size, n_actions)
- self.target_net = DQN(obs_size, n_actions)
-
- self.buffer = ReplayBuffer(self.hparams.replay_size)
- self.agent = Agent(self.env, self.buffer)
- self.total_reward = 0
- self.episode_reward = 0
- self.populate(self.hparams.warm_start_steps)
-
- def populate(self, steps: int = 1000) -> None:
- """Carries out several random steps through the environment to initially fill up the replay buffer with
- experiences.
-
- Args:
- steps: number of random steps to populate the buffer with
- """
- for _ in range(steps):
- self.agent.play_step(self.net, epsilon=1.0)
-
- def forward(self, x: Tensor) -> Tensor:
- """Passes in a state x through the network and gets the q_values of each action as an output.
-
- Args:
- x: environment state
-
- Returns:
- q values
- """
- output = self.net(x)
- return output
-
- def dqn_mse_loss(self, batch: Tuple[Tensor, Tensor]) -> Tensor:
- """Calculates the mse loss using a mini batch from the replay buffer.
-
- Args:
- batch: current mini batch of replay data
-
- Returns:
- loss
- """
- states, actions, rewards, dones, next_states = batch
-
- state_action_values = self.net(states).gather(1, actions.long().unsqueeze(-1)).squeeze(-1)
-
- with torch.no_grad():
- next_state_values = self.target_net(next_states).max(1)[0]
- next_state_values[dones] = 0.0
- next_state_values = next_state_values.detach()
-
- expected_state_action_values = next_state_values * self.hparams.gamma + rewards
-
- return nn.MSELoss()(state_action_values, expected_state_action_values)
-
- def get_epsilon(self, start: int, end: int, frames: int) -> float:
- if self.global_step > frames:
- return end
- return start - (self.global_step / frames) * (start - end)
-
- def training_step(self, batch: Tuple[Tensor, Tensor], nb_batch) -> OrderedDict:
- """Carries out a single step through the environment to update the replay buffer. Then calculates loss
- based on the minibatch recieved.
-
- Args:
- batch: current mini batch of replay data
- nb_batch: batch number
-
- Returns:
- Training loss and log metrics
- """
- device = self.get_device(batch)
- epsilon = self.get_epsilon(self.hparams.eps_start, self.hparams.eps_end, self.hparams.eps_last_frame)
- self.log("epsilon", epsilon)
-
- # step through environment with agent
- reward, done = self.agent.play_step(self.net, epsilon, device)
- self.episode_reward += reward
- self.log("episode reward", self.episode_reward)
-
- # calculates training loss
- loss = self.dqn_mse_loss(batch)
-
- if done:
- self.total_reward = self.episode_reward
- self.episode_reward = 0
-
- # Soft update of target network
- if self.global_step % self.hparams.sync_rate == 0:
- self.target_net.load_state_dict(self.net.state_dict())
-
- self.log_dict(
- {
- "reward": reward,
- "train_loss": loss,
- }
- )
- self.log("total_reward", self.total_reward, prog_bar=True)
- self.log("steps", self.global_step, logger=False, prog_bar=True)
-
- return loss
-
- def configure_optimizers(self) -> List[Optimizer]:
- """Initialize Adam optimizer."""
- optimizer = Adam(self.net.parameters(), lr=self.hparams.lr)
- return optimizer
-
- def __dataloader(self) -> DataLoader:
- """Initialize the Replay Buffer dataset used for retrieving experiences."""
- dataset = RLDataset(self.buffer, self.hparams.episode_length)
- dataloader = DataLoader(
- dataset=dataset,
- batch_size=self.hparams.batch_size,
- )
- return dataloader
-
- def train_dataloader(self) -> DataLoader:
- """Get train loader."""
- return self.__dataloader()
-
- def get_device(self, batch) -> str:
- """Retrieve device currently being used by minibatch."""
- return batch[0].device.index if self.on_gpu else "cpu"
-
-model = DQNLightning()
-
-trainer = pl.Trainer(
- accelerator="auto",
- devices=1 if torch.cuda.is_available() else None, # limiting got iPython runs
- max_epochs=150,
- val_check_interval=50,
- logger=CSVLogger(save_dir="logs/"),
-)
-
-trainer.fit(model)
\ No newline at end of file
diff --git a/src/icaart_ensemble_hyper.py b/src/icaart_ensemble_hyper.py
deleted file mode 100644
index 1d99c9a2559b6408c82dd5f6fd1721c805d9bbd4..0000000000000000000000000000000000000000
--- a/src/icaart_ensemble_hyper.py
+++ /dev/null
@@ -1,236 +0,0 @@
-import argparse
-import logging
-import pickle
-import sys
-from pathlib import Path
-
-import lightning.pytorch as pl
-import optuna
-import torch
-import yaml
-from lightning.pytorch.callbacks import EarlyStopping, ModelCheckpoint
-from lightning.pytorch.loggers import TensorBoardLogger
-
-import utils.helpers as hp
-import utils.utility as ut
-from callbacks import OptunaPruningCallback, WaVoCallback
-from data_tools.data_module import WaVoDataModule
-from ensemble_models import WaVoLightningEnsemble, WaVoLightningAttentionEnsemble
-
-use_cuda = torch.cuda.is_available()
-if use_cuda:
- accelerator = 'cuda'
- torch.set_float32_matmul_precision('high')
-
- max_free = 0
- best_device = None
- for j in range(torch.cuda.device_count()):
- free , _ = torch.cuda.mem_get_info(j)
- if free > max_free:
- max_free = free
- best_device = j
- devices = [best_device]
-
-else:
- accelerator = 'cpu'
- devices = 'auto'
-
-storage_base = 'sqlite:///../../../../data-project/KIWaVo/models/optuna/'
-
-if ut.debugger_is_active():
- max_epochs = 2
- default_storage_name = 'sqlite:///../../../data-project/KIWaVo/models/optuna/icaart_ensemble_debug_01.db'
-
-else:
- max_epochs = 2000
- default_storage_name = 'sqlite:///../../../../data-project/KIWaVo/models/optuna/icaart_ensemble_01.db'
-
-
-class Objective:
- """
- This class defines the objective function for hyperparameter tuning using Optuna library.
-
- Args:
- filename (str|Path): Path to .csv file, first column should be a timeindex
- model_dir (str|Path): Path to the directory containing the base models
- log_dir (str|Path): Path to the logging directory
- select_strat (str): How to select the base models (random or first n)
- model_count (int): How many base models to use
- monitor (str, optional): metric to monitor. Defaults to 'hp/val_loss'.
- """
- def __init__(self, filename,modeldir,logdir,select_strat,model_count,monitor= 'hp/val_loss', **kwargs):
- # Hold these implementation specific arguments as the fields of the class.
- self.filename = filename
- self.model_dir = modeldir
- self.log_dir = logdir
- self.select_strat = select_strat
- self.model_count = model_count
- self.monitor = monitor
-
- def _get_callbacks(self, trial):
- checkpoint_callback = ModelCheckpoint(save_top_k=1, monitor=self.monitor,save_weights_only=True)
- pruning_callback = OptunaPruningCallback(trial, monitor=self.monitor)
- early_stop_callback = EarlyStopping(monitor=self.monitor, mode="min", patience=3)
- my_callback = WaVoCallback(ensemble=True)
-
- return my_callback, [checkpoint_callback,pruning_callback, early_stop_callback, my_callback]
-
- def _get_model_params(self,trial):
- model_params = dict(
- hidden_size = trial.suggest_int("hidden_size", 32, 512),
- num_layers = trial.suggest_int("n_layers", 2, 4),
- dropout = 0.25,
- learning_rate = trial.suggest_float("lr", 0.00001, 0.01),
- norm_func = trial.suggest_categorical("norm_func", ['softmax','minmax']),
- )
-
- return model_params
-
- def __call__(self, trial):
-
- model_params = self._get_model_params(trial)
- #log_dir = '../../../data-project/KIWaVo/models/ensemble_debug/'
- #model_dir = Path('../../../data-project/KIWaVo/models/icaarts_hollingstedt/lightning_logs/')
- model_list = []
- model_path_list = []
- yaml_data = None
-
- if self.select_strat == 'random':
- all_models = [x.name.split('_')[1] for x in self.model_dir.iterdir()]
-
- elif self.select_strat == 'first':
- model_choice = list(range(self.model_count))
-
- #for s in [0,1,2]:
- for s in model_choice:
- temp_dir = self.model_dir / f'version_{s}/'
-
- model_list.append(hp.load_model_cuda(temp_dir,use_cuda=use_cuda,devices=devices))
- model_path_list.append(str(temp_dir.resolve()))
-
- if yaml_data is None:
- yaml_data = hp.load_settings_model(temp_dir)
- #with open(temp_dir / 'hparams.yaml', 'r') as file:
- # yaml_data = yaml.load(file, Loader=yaml.FullLoader)
- # yaml_data['scaler'] = pickle.loads(yaml_data['scaler'])
-
- config = {
- 'scaler' : yaml_data['scaler'], #TODO test how this works without giving scaler etc. outside of jupyterlab
- #'filename' : yaml_data['filename'],
- 'filename' : str(self.filename),
- 'level_name_org' : yaml_data['level_name_org'],
- 'out_size' : yaml_data['out_size'],
- 'threshold' : yaml_data['threshold'],
- 'feature_count' : yaml_data['feature_count'],
- 'differencing' : yaml_data['differencing'],
- 'model_architecture' : 'ensemble'}
-
- print("ACHTUNG GGF. FALSCHES MODELL")
- #ensemble_model = WaVoLightningEnsemble(model_list,model_path_list,**model_params)#TODO 'ÄNDERN!
- ensemble_model = WaVoLightningAttentionEnsemble(model_list,model_path_list,**model_params)
-
- config['in_size'] = ensemble_model.max_in_size
-
- data_module = WaVoDataModule(**config)
-
-
-
- logging.info("Params: %s", trial.params)
-
-
- my_callback, callbacks = self._get_callbacks(trial)
- logger = TensorBoardLogger(self.log_dir, default_hp_metric=False)
-
-
- trainer = pl.Trainer(default_root_dir=self.log_dir,
- gradient_clip_val=0.5,
- logger=logger,
- accelerator=accelerator,
- devices=devices,
- callbacks=callbacks,
- max_epochs=max_epochs,
- log_every_n_steps=10,
- )
- trainer.fit(ensemble_model, data_module)
-
- #save metrics to optuna
- model_path = str(Path(trainer.log_dir).resolve())
- logging.info("model_path: %s", model_path)
- trial.set_user_attr("model_path", model_path)
- for metric in ['hp/val_nse', 'hp/val_mae', 'hp/val_mae_flood']:
- for i in [23, 47]:
- trial.set_user_attr(f'{metric}_{i}', my_callback.metrics[metric][i].item())
-
- return my_callback.metrics[self.monitor].item()
-
-
-
-def parse_args() -> argparse.Namespace:
- """ Parse all the arguments and provides some help in the command line
- """
-
- parser: argparse.ArgumentParser = argparse.ArgumentParser(description='Execute experiments for exp_icaart.')
- parser.add_argument('filename', metavar='datafile', type=Path,help='The path to your input data.')
- parser.add_argument('modeldir', metavar='modeldir', type=Path,help='The path to your base models.')
- parser.add_argument('logdir', type=Path,help='set a directory for logs and model checkpoints.')
- parser.add_argument('trials', metavar='trials',type=int, default=100,help='How many trials to run.')
- parser.add_argument('select_strat', choices=['random','first'])
- parser.add_argument('model_count', metavar='mc',type=int, default=5,help='How many base models to use.')
-
- parser.add_argument('--expname', metavar='experiment_name',type=str, default='nameless',help='The name of the experiment.')
- parser.add_argument('--storagename', metavar='storage_name',type=str, default=None,help='The database for the experiment.')
-
- return parser.parse_args()
-
-
-
-
-def main():
- parsed_args = parse_args()
- if not parsed_args.logdir.exists():
- parsed_args.logdir.mkdir(parents=True)
-
- if False:
- pruner = optuna.pruners.HyperbandPruner(
- min_resource=1, max_resource='auto', reduction_factor=3, bootstrap_count=0)
- else:
- pruner = optuna.pruners.NopPruner()
-
-
-
- study_name = f"{parsed_args.filename.stem} {parsed_args.expname}" # Unique identifier of the study.
- storage_name = default_storage_name if parsed_args.storagename is None else f"{storage_base}{parsed_args.storagename}.db"
-
- # Logging, add stream handler of stdout to show the messages
- logging.basicConfig(level=logging.INFO)
- logFormatter = logging.Formatter('%(asctime)s;%(levelname)s;%(message)s',datefmt='%Y-%m-%d %H:%M:%S')
- fileHandler = logging.FileHandler(parsed_args.logdir / 'ensemble_hyper.log',)
- consoleHandler = logging.StreamHandler(sys.stdout)
- fileHandler.setFormatter(logFormatter)
- consoleHandler.setFormatter(logFormatter)
- logging.getLogger().addHandler(fileHandler)
- #logging.getLogger().addHandler(consoleHandler)
- optuna.logging.get_logger("optuna").addHandler(fileHandler)
- optuna.logging.get_logger("optuna").addHandler(consoleHandler)
-
- logging.info('Start of this execution======================================================================')
- logging.info("Executing %s with device %s and parameters %s ",sys.argv[0],devices,sys.argv[1:])
-
-
- study = optuna.create_study(
- study_name=study_name,
- storage=storage_name,
- direction="minimize",
- pruner=pruner,
- load_if_exists=True)
-
- study.set_metric_names(["hp/val_loss"])
- objective = Objective(**vars(parsed_args),gc_after_trial=True)
-
-
- study.optimize(objective,n_trials=parsed_args.trials,timeout=None,callbacks=[lambda study, trial: torch.cuda.empty_cache()])
-
-
-if __name__ == "__main__":
- #TODO command line arguments
- main()
diff --git a/src/icaart_random_base.py b/src/icaart_random_base.py
deleted file mode 100644
index 8a6b3f5dea646ae16bd1edab3e937af8418b4e48..0000000000000000000000000000000000000000
--- a/src/icaart_random_base.py
+++ /dev/null
@@ -1,233 +0,0 @@
-import argparse
-import logging
-import pickle
-import sys
-from pathlib import Path
-import random
-import lightning.pytorch as pl
-import optuna
-import torch
-import yaml
-from lightning.pytorch.callbacks import EarlyStopping, ModelCheckpoint
-from lightning.pytorch.loggers import TensorBoardLogger
-
-import utils.helpers as hp
-import utils.utility as ut
-from callbacks import OptunaPruningCallback, WaVoCallback
-from data_tools.data_module import WaVoDataModule
-from ensemble_models import WaVoLightningEnsemble
-
-use_cuda = torch.cuda.is_available()
-if use_cuda:
- accelerator = 'cuda'
- torch.set_float32_matmul_precision('high')
-
- max_free = 0
- best_device = None
- for j in range(torch.cuda.device_count()):
- free , _ = torch.cuda.mem_get_info(j)
- if free > max_free:
- max_free = free
- best_device = j
- devices = [best_device]
-
-else:
- accelerator = 'cpu'
- devices = 'auto'
-
-storage_base = 'sqlite:///../../../../data-project/KIWaVo/models/optuna/'
-
-if ut.debugger_is_active():
- max_epochs = 2
- default_storage_name = 'sqlite:///../../../data-project/KIWaVo/models/optuna/icaart_ensemble_debug_01.db'
-
-else:
- max_epochs = 100
- default_storage_name = 'sqlite:///../../../../data-project/KIWaVo/models/optuna/icaart_ensemble_01.db'
-
-
-class Objective:
- """
- This class defines the objective function for hyperparameter tuning using Optuna library.
-
- Args:
- filename (str|Path): Path to .csv file, first column should be a timeindex
- model_dir (str|Path): Path to the directory containing the base models
- log_dir (str|Path): Path to the logging directory
- model_count (int): How many base models to use
- monitor (str, optional): metric to monitor. Defaults to 'hp/val_loss'.
- """
- def __init__(self, filename,modeldir,logdir,model_count,monitor= 'hp/val_loss', **kwargs):
- # Hold these implementation specific arguments as the fields of the class.
- self.filename = filename
- self.model_dir = modeldir
- self.log_dir = logdir
- self.model_count = model_count
- self.monitor = monitor
- self.all_models = [int(x.name.split('_')[1]) for x in self.model_dir.iterdir()]
-
-
- def _get_callbacks(self, trial):
- checkpoint_callback = ModelCheckpoint(save_top_k=1, monitor=self.monitor,save_weights_only=True)
- pruning_callback = OptunaPruningCallback(trial, monitor=self.monitor)
- early_stop_callback = EarlyStopping(monitor=self.monitor, mode="min", patience=3)
- my_callback = WaVoCallback(ensemble=True)
-
- return my_callback, [checkpoint_callback,pruning_callback, early_stop_callback, my_callback]
-
- def _get_model_params(self,trial):
- #This is just so that optuna will actually execute the number of trials we want, it stops when it testes all parameter combinations
- place_holder = trial.suggest_int("place_holder", 0, 10000)
-
- model_params = dict(
- hidden_size = trial.suggest_int("hidden_size", 512, 512),
- num_layers = trial.suggest_int("n_layers", 2, 2),
- dropout = 0.25,
- learning_rate = trial.suggest_float("lr", 0.002, 0.002),
- norm_func = trial.suggest_categorical("norm_func", ['minmax']),
- )
-
- return model_params
-
- def __call__(self, trial):
-
- model_params = self._get_model_params(trial)
- #log_dir = '../../../data-project/KIWaVo/models/ensemble_debug/'
- #model_dir = Path('../../../data-project/KIWaVo/models/icaarts_hollingstedt/lightning_logs/')
- model_list = []
- model_path_list = []
- yaml_data = None
-
-
-
- model_choice = random.sample(self.all_models, self.model_count)
-
- #for s in [0,1,2]:
- for s in model_choice:
- temp_dir = self.model_dir / f'version_{s}/'
-
- model_list.append(hp.load_model_cuda(temp_dir,use_cuda=use_cuda,devices=devices))
- model_path_list.append(str(temp_dir.resolve()))
-
- if yaml_data is None:
- yaml_data = hp.load_settings_model(temp_dir)
- #with open(temp_dir / 'hparams.yaml', 'r') as file:
- # yaml_data = yaml.load(file, Loader=yaml.FullLoader)
- # yaml_data['scaler'] = pickle.loads(yaml_data['scaler'])
-
- config = {
- 'scaler' : yaml_data['scaler'], #TODO test how this works without giving scaler etc. outside of jupyterlab
- #'filename' : yaml_data['filename'],
- 'filename' : str(self.filename),
- 'level_name_org' : yaml_data['level_name_org'],
- 'out_size' : yaml_data['out_size'],
- 'threshold' : yaml_data['threshold'],
- 'feature_count' : yaml_data['feature_count'],
- 'differencing' : yaml_data['differencing'],
- 'model_architecture' : 'ensemble'}
-
- ensemble_model = WaVoLightningEnsemble(model_list,model_path_list,**model_params)
- config['in_size'] = ensemble_model.max_in_size
-
- data_module = WaVoDataModule(**config)
-
-
-
- logging.info("Params: %s", trial.params)
-
-
- my_callback, callbacks = self._get_callbacks(trial)
- logger = TensorBoardLogger(self.log_dir, default_hp_metric=False)
-
-
- trainer = pl.Trainer(default_root_dir=self.log_dir,
- gradient_clip_val=0.5,
- logger=logger,
- accelerator=accelerator,
- devices=devices,
- callbacks=callbacks,
- max_epochs=max_epochs,
- log_every_n_steps=10,
- )
- trainer.fit(ensemble_model, data_module)
-
- #save metrics to optuna
- model_path = str(Path(trainer.log_dir).resolve())
- logging.info("model_path: %s", model_path)
- trial.set_user_attr("model_path", model_path)
- for metric in ['hp/val_nse', 'hp/val_mae', 'hp/val_mae_flood']:
- for i in [23, 47]:
- trial.set_user_attr(f'{metric}_{i}', my_callback.metrics[metric][i].item())
-
- return my_callback.metrics[self.monitor].item()
-
-
-
-def parse_args() -> argparse.Namespace:
- """ Parse all the arguments and provides some help in the command line
- """
-
- parser: argparse.ArgumentParser = argparse.ArgumentParser(description='Execute experiments for exp_icaart.')
- parser.add_argument('filename', metavar='datafile', type=Path,help='The path to your input data.')
- parser.add_argument('modeldir', metavar='modeldir', type=Path,help='The path to your base models.')
- parser.add_argument('logdir', type=Path,help='set a directory for logs and model checkpoints.')
- parser.add_argument('trials', metavar='trials',type=int, default=100,help='How many trials to run.')
- parser.add_argument('model_count', metavar='mc',type=int, default=5,help='How many base models to use.')
-
- parser.add_argument('--expname', metavar='experiment_name',type=str, default='nameless',help='The name of the experiment.')
- parser.add_argument('--storagename', metavar='storage_name',type=str, default=None,help='The database for the experiment.')
-
- return parser.parse_args()
-
-
-
-
-def main():
- parsed_args = parse_args()
- if not parsed_args.logdir.exists():
- parsed_args.logdir.mkdir(parents=True)
-
- if False:
- pruner = optuna.pruners.HyperbandPruner(
- min_resource=1, max_resource='auto', reduction_factor=3, bootstrap_count=0)
- else:
- pruner = optuna.pruners.NopPruner()
-
-
-
- study_name = f"{parsed_args.filename.stem} {parsed_args.expname}" # Unique identifier of the study.
- storage_name = default_storage_name if parsed_args.storagename is None else f"{storage_base}{parsed_args.storagename}.db"
-
- # Logging, add stream handler of stdout to show the messages
- logging.basicConfig(level=logging.INFO)
- logFormatter = logging.Formatter('%(asctime)s;%(levelname)s;%(message)s',datefmt='%Y-%m-%d %H:%M:%S')
- fileHandler = logging.FileHandler(parsed_args.logdir / 'ensemble_hyper.log',)
- consoleHandler = logging.StreamHandler(sys.stdout)
- fileHandler.setFormatter(logFormatter)
- consoleHandler.setFormatter(logFormatter)
- logging.getLogger().addHandler(fileHandler)
- #logging.getLogger().addHandler(consoleHandler)
- optuna.logging.get_logger("optuna").addHandler(fileHandler)
- optuna.logging.get_logger("optuna").addHandler(consoleHandler)
-
- logging.info('Start of this execution======================================================================')
- logging.info("Executing %s with device %s and parameters %s ",sys.argv[0],devices,sys.argv[1:])
-
-
- study = optuna.create_study(
- study_name=study_name,
- storage=storage_name,
- direction="minimize",
- pruner=pruner,
- load_if_exists=True)
-
- study.set_metric_names(["hp/val_loss"])
- objective = Objective(**vars(parsed_args),gc_after_trial=True)
-
-
- study.optimize(objective,n_trials=parsed_args.trials,timeout=None,callbacks=[lambda study, trial: torch.cuda.empty_cache()])
-
-
-if __name__ == "__main__":
- #TODO command line arguments
- main()
diff --git a/src/icaart_sizes.py b/src/icaart_sizes.py
deleted file mode 100644
index 532255d01dacbdb16abea421c4eca05a482fe6e0..0000000000000000000000000000000000000000
--- a/src/icaart_sizes.py
+++ /dev/null
@@ -1,237 +0,0 @@
-import argparse
-import logging
-import pickle
-import sys
-from pathlib import Path
-import random
-import lightning.pytorch as pl
-import optuna
-import torch
-import yaml
-from lightning.pytorch.callbacks import EarlyStopping, ModelCheckpoint
-from lightning.pytorch.loggers import TensorBoardLogger
-
-import utils.helpers as hp
-import utils.utility as ut
-from callbacks import OptunaPruningCallback, WaVoCallback
-from data_tools.data_module import WaVoDataModule
-from ensemble_models import WaVoLightningEnsemble
-
-use_cuda = torch.cuda.is_available()
-if use_cuda:
- accelerator = 'cuda'
- torch.set_float32_matmul_precision('high')
-
- max_free = 0
- best_device = None
- for j in range(torch.cuda.device_count()):
- free , _ = torch.cuda.mem_get_info(j)
- if free > max_free:
- max_free = free
- best_device = j
- devices = [best_device]
-
-else:
- accelerator = 'cpu'
- devices = 'auto'
-
-storage_base = 'sqlite:///../../../../data-project/KIWaVo/models/optuna/'
-
-if ut.debugger_is_active():
- max_epochs = 2
- default_storage_name = 'sqlite:///../../../data-project/KIWaVo/models/optuna/icaart_ensemble_debug_01.db'
-
-else:
- max_epochs = 100
- default_storage_name = 'sqlite:///../../../../data-project/KIWaVo/models/optuna/icaart_ensemble_01.db'
-
-
-class Objective:
- """
- This class defines the objective function for hyperparameter tuning using Optuna library.
-
- Args:
- filename (str|Path): Path to .csv file, first column should be a timeindex
- model_dir (str|Path): Path to the directory containing the base models
- log_dir (str|Path): Path to the logging directory
- monitor (str, optional): metric to monitor. Defaults to 'hp/val_loss'.
- """
- def __init__(self, filename,modeldir,logdir,monitor= 'hp/val_loss', **kwargs):
- # Hold these implementation specific arguments as the fields of the class.
- self.filename = filename
- self.model_dir = modeldir
- self.log_dir = logdir
- self.monitor = monitor
- self.all_models = [int(x.name.split('_')[1]) for x in self.model_dir.iterdir()]
-
-
- def _get_callbacks(self, trial):
- checkpoint_callback = ModelCheckpoint(save_top_k=1, monitor=self.monitor,save_weights_only=True)
- pruning_callback = OptunaPruningCallback(trial, monitor=self.monitor)
- early_stop_callback = EarlyStopping(monitor=self.monitor, mode="min", patience=3)
- my_callback = WaVoCallback(ensemble=True)
-
- return my_callback, [checkpoint_callback,pruning_callback, early_stop_callback, my_callback]
-
- def _get_model_params(self,trial):
- #This is just so that optuna will actually execute the number of trials we want, it stops when it testes all parameter combinations
- ensemble_size = trial.suggest_categorical("ensemble_size",[5,10,15,20])
- #temperature = trial.suggest_categorical("temperature",[1/10,1,10])
-
- model_params = dict(
- hidden_size = trial.suggest_int("hidden_size", 512, 512),
- num_layers = trial.suggest_int("n_layers", 2, 2),
- dropout = 0.25,
- learning_rate = trial.suggest_float("lr", 0.002, 0.002),
- norm_func = trial.suggest_categorical("norm_func", ['softmax']),
- )
-
- return model_params, ensemble_size
-
- def __call__(self, trial):
-
- model_params,ensemble_size = self._get_model_params(trial)
- #log_dir = '../../../data-project/KIWaVo/models/ensemble_debug/'
- #model_dir = Path('../../../data-project/KIWaVo/models/icaarts_hollingstedt/lightning_logs/')
- model_list = []
- model_path_list = []
- yaml_data = None
-
-
-
- model_choice = random.sample(self.all_models, ensemble_size)
-
- #for s in [0,1,2]:
- for s in model_choice:
- temp_dir = self.model_dir / f'version_{s}/'
-
- model_list.append(hp.load_model_cuda(temp_dir,use_cuda=use_cuda,devices=devices))
- model_path_list.append(str(temp_dir.resolve()))
-
- if yaml_data is None:
- yaml_data = hp.load_settings_model(temp_dir)
- #with open(temp_dir / 'hparams.yaml', 'r') as file:
- # yaml_data = yaml.load(file, Loader=yaml.FullLoader)
- # yaml_data['scaler'] = pickle.loads(yaml_data['scaler'])
-
- config = {
- 'scaler' : yaml_data['scaler'], #TODO test how this works without giving scaler etc. outside of jupyterlab
- #'filename' : yaml_data['filename'],
- 'filename' : str(self.filename),
- 'level_name_org' : yaml_data['level_name_org'],
- 'out_size' : yaml_data['out_size'],
- 'threshold' : yaml_data['threshold'],
- 'feature_count' : yaml_data['feature_count'],
- 'differencing' : yaml_data['differencing'],
- 'model_architecture' : 'ensemble'}
-
- ensemble_model = WaVoLightningEnsemble(model_list,model_path_list,**model_params)
- config['in_size'] = ensemble_model.max_in_size
-
- data_module = WaVoDataModule(**config)
-
-
-
- logging.info("Params: %s", trial.params)
-
-
- my_callback, callbacks = self._get_callbacks(trial)
- logger = TensorBoardLogger(self.log_dir, default_hp_metric=False)
-
-
- trainer = pl.Trainer(default_root_dir=self.log_dir,
- gradient_clip_val=0.5,
- logger=logger,
- accelerator=accelerator,
- devices=devices,
- callbacks=callbacks,
- max_epochs=max_epochs,
- log_every_n_steps=10,
- )
- trainer.fit(ensemble_model, data_module)
-
- #save metrics to optuna
- model_path = str(Path(trainer.log_dir).resolve())
- logging.info("model_path: %s", model_path)
- trial.set_user_attr("model_path", model_path)
- for metric in ['hp/val_nse', 'hp/val_mae', 'hp/val_mae_flood']:
- for i in [23, 47]:
- trial.set_user_attr(f'{metric}_{i}', my_callback.metrics[metric][i].item())
-
- return my_callback.metrics[self.monitor].item()
-
-
-
-def parse_args() -> argparse.Namespace:
- """ Parse all the arguments and provides some help in the command line
- """
-
- parser: argparse.ArgumentParser = argparse.ArgumentParser(description='Execute experiments for exp_icaart.')
- parser.add_argument('filename', metavar='datafile', type=Path,help='The path to your input data.')
- parser.add_argument('modeldir', metavar='modeldir', type=Path,help='The path to your base models.')
- parser.add_argument('logdir', type=Path,help='set a directory for logs and model checkpoints.')
- parser.add_argument('trials', metavar='trials',type=int, default=100,help='How many trials to run.')
- #parser.add_argument('model_count', metavar='mc',type=int, default=5,help='How many base models to use.')
-
- parser.add_argument('--expname', metavar='experiment_name',type=str, default='nameless',help='The name of the experiment.')
- parser.add_argument('--storagename', metavar='storage_name',type=str, default=None,help='The database for the experiment.')
-
- return parser.parse_args()
-
-
-
-
-def main():
- parsed_args = parse_args()
- if not parsed_args.logdir.exists():
- parsed_args.logdir.mkdir(parents=True)
-
- if False:
- pruner = optuna.pruners.HyperbandPruner(
- min_resource=1, max_resource='auto', reduction_factor=3, bootstrap_count=0)
- else:
- pruner = optuna.pruners.NopPruner()
-
-
-
- study_name = f"{parsed_args.filename.stem} {parsed_args.expname}" # Unique identifier of the study.
- storage_name = default_storage_name if parsed_args.storagename is None else f"{storage_base}{parsed_args.storagename}.db"
-
- # Logging, add stream handler of stdout to show the messages
- logging.basicConfig(level=logging.INFO)
- logFormatter = logging.Formatter('%(asctime)s;%(levelname)s;%(message)s',datefmt='%Y-%m-%d %H:%M:%S')
- fileHandler = logging.FileHandler(parsed_args.logdir / 'ensemble_hyper.log',)
- consoleHandler = logging.StreamHandler(sys.stdout)
- fileHandler.setFormatter(logFormatter)
- consoleHandler.setFormatter(logFormatter)
- logging.getLogger().addHandler(fileHandler)
- #logging.getLogger().addHandler(consoleHandler)
- optuna.logging.get_logger("optuna").addHandler(fileHandler)
- optuna.logging.get_logger("optuna").addHandler(consoleHandler)
-
- logging.info('Start of this execution======================================================================')
- logging.info("Executing %s with device %s and parameters %s ",sys.argv[0],devices,sys.argv[1:])
-
-
- ens_sizes =[5,10,15,20]
- search_space = {"ensemble_size":ens_sizes*parsed_args.trials }
- sampler=optuna.samplers.GridSampler(search_space)
-
- study = optuna.create_study(
- sampler=sampler,
- study_name=study_name,
- storage=storage_name,
- direction="minimize",
- pruner=pruner,
- load_if_exists=True)
-
- study.set_metric_names(["hp/val_loss"])
- objective = Objective(**vars(parsed_args),gc_after_trial=True)
-
-
- study.optimize(objective,n_trials=len(ens_sizes)*parsed_args.trials,timeout=None,callbacks=[lambda study, trial: torch.cuda.empty_cache()])
-
-
-if __name__ == "__main__":
- #TODO command line arguments
- main()
diff --git a/src/main.py b/src/main.py
deleted file mode 100644
index 9ba378440bbfbca27b0387bbb557cd9dd65de492..0000000000000000000000000000000000000000
--- a/src/main.py
+++ /dev/null
@@ -1,90 +0,0 @@
-"""
-Main Module to start training
-"""
-import lightning.pytorch as pl
-import torch
-from lightning.pytorch.callbacks.early_stopping import EarlyStopping
-
-from lightning.pytorch.loggers import TensorBoardLogger
-
-import metrics as mt
-from callbacks import WaVoCallback
-from data_tools import WaVoDataModule
-from models import WaVoLightningModule
-
-# print('set float matmul precision to high')
-torch.set_float32_matmul_precision('high')
-
-
-def main():
- """Main Method
- """
-
- config = {
- # 'filename': '../../data/input/Poetrau20.csv',
- # 'log_dir': '../../models_torch/pötrau',
- # 'level_name_org': 'WPoetrau_pegel_cm',
- 'filename': '../../data/input/Treia.csv',
- 'log_dir': '../../models_torch/treia',
- 'level_name_org': 'Treia_pegel_cm',
- 'train': 0.7,
- 'val': 0.15,
- 'test': 0.15,
- 'batch_size': 2048,
- 'in_size': 144,
- 'out_size': 48,
- # 'scale_target':False,
- # 'differencing':1,
- 'differencing': 0,
- 'percentile': 0.95
- }
-
- found_lr = None
- tune_lr = False
-
- for _ in range(10):
-
- data_module = WaVoDataModule(**config)
- data_module.prepare_data()
- data_module.setup(stage='fit')
-
- # TODO kwargs for neural network?
- model = WaVoLightningModule(
- data_module.mean,
- data_module.scale,
- feature_count=data_module.feature_count,
- in_size=data_module.hparams.in_size,
- out_size=data_module.hparams.out_size,
- hidden_size_lstm=128,
- hidden_size=64,
- learning_rate=found_lr or 0.001) # TODO tidy up scaler mess
-
- # Callbacks & Logging
- early_stop_callback = EarlyStopping(
- monitor="hp/val_loss", mode="min", patience=3)
- my_callback = WaVoCallback()
- callbacks = [early_stop_callback, my_callback]
- logger = TensorBoardLogger(config['log_dir'], default_hp_metric=False)
-
- trainer = pl.Trainer(default_root_dir=config['log_dir'],
- logger=logger,
- # accelerator="auto",
- accelerator="gpu",
- devices=1,
- callbacks=callbacks,
- max_epochs=50,
- log_every_n_steps=1,
- )
-
- if found_lr is None and tune_lr:
- tuner = pl.tuner.Tuner(trainer)
- # tuner.scale_batch_size(model, datamodule=data_module,init_val=512, max_trials=3)
- # TODO calling data_module here messes with the hyperparameter logging
- lr_finder = tuner.lr_find(model, datamodule=data_module)
- found_lr = lr_finder.suggestion()
-
- trainer.fit(model, data_module)
-
-
-if __name__ == "__main__":
- main()
diff --git a/src/main_cli.py b/src/main_cli.py
deleted file mode 100644
index a2dd91eb1b047906dc3a766a7c8e3f56390431ce..0000000000000000000000000000000000000000
--- a/src/main_cli.py
+++ /dev/null
@@ -1,66 +0,0 @@
-"""
-Main Module to start training
-"""
-import lightning.pytorch as pl
-import torch
-from lightning.pytorch.callbacks.early_stopping import EarlyStopping
-from lightning.pytorch.cli import LightningCLI
-
-from lightning.pytorch.loggers import TensorBoardLogger
-
-import metrics as mt
-from callbacks import WaVoCallback
-from data_tools import WaVoDataModule
-from models import WaVoLightningModule
-
-# print('set float matmul precision to high')
-torch.set_float32_matmul_precision('high')
-
-
-class MyLightningCLI(LightningCLI):
- # def __init__():
- # super().__init__()
-
- def add_arguments_to_parser(self, parser):
- parser.link_arguments("data.mean", "model.mean",
- apply_on="instantiate")
- parser.link_arguments("data.scale", "model.scale",
- apply_on="instantiate")
- parser.link_arguments(
- "data.threshold", "model.threshold", apply_on="instantiate")
- parser.link_arguments("data.feature_count",
- "model.feature_count", apply_on="instantiate")
- parser.link_arguments("data.in_size", "model.in_size")
- parser.link_arguments("data.out_size", "model.out_size")
-
- parser.add_lightning_class_args(WaVoCallback, "wavo_callback")
- # parser.set_defaults({"wavo_callback.chosen_metrics": ['nse', 'p10']})
-
- parser.add_lightning_class_args(EarlyStopping, "early_stopping")
- parser.set_defaults({"early_stopping.monitor": "hp/val_loss",
- 'early_stopping.mode': 'min', "early_stopping.patience": 2})
- # monitor="hp/val_loss", mode="min"
-
-
-if __name__ == "__main__":
-
- cli = MyLightningCLI(model_class=WaVoLightningModule,
- datamodule_class=WaVoDataModuleCLI,
- )
- # run=False
- # hidden_size_lstm: 128
- # hidden_size: 64
- # num_layers: 2
- # dropout: 0.2
- # learning_rate: 0.001
- # lightning_model = WaVoLightningModule(
- # cli.data_module.mean,
- # cli.data_module.scale,
- # cli.data_module.threshold,
- # feature_count=cli.data_module.feature_count,
- # in_size=cli.data_module.hparams.in_size,
- # out_size=cli.data_module.hparams.out_size,
- # hidden_size_lstm=cli.model.hidden_size_lstm,
- # hidden_size=cli.model.hidden_size,
- # learning_rate=cli.model.learning_rate)
- # cli.trainer.fit(lightning_model, datamodule=cli.datamodule)
diff --git a/src/main_hyper.py b/src/main_hyper.py
index 50dd34890235179cfef1d5620183aa3fdb6bf47d..03a514d7553324e37fc2ee7ed634a32ee230388e 100644
--- a/src/main_hyper.py
+++ b/src/main_hyper.py
@@ -1,6 +1,7 @@
"""
Main Module to start training
"""
+
import argparse
import gc
import logging
@@ -15,46 +16,49 @@ from lightning.pytorch.loggers import TensorBoardLogger
from lightning.pytorch.tuner import Tuner
import utils.utility as ut
-from callbacks import OptunaPruningCallback, WaVoCallback
+from utils.callbacks import OptunaPruningCallback, WaVoCallback
from data_tools.data_module import WaVoDataModule
from models.lightning_module import WaVoLightningModule
use_cuda = torch.cuda.is_available()
if use_cuda:
- accelerator = 'cuda'
- #accelerator = 'gpu'
- torch.set_float32_matmul_precision('high')
+ accelerator = "cuda"
+ # accelerator = 'gpu'
+ torch.set_float32_matmul_precision("high")
else:
- accelerator = 'cpu'
+ accelerator = "cpu"
# CONSTANTS
-#storage_base = 'sqlite:///../../models_torch/optuna/'
-storage_base = 'sqlite:///../../../../data-project/KIWaVo/models/optuna/'
+# storage_base = 'sqlite:///../../models_torch/optuna/'
+storage_base = "sqlite:///../../../../data-project/KIWaVo/models/optuna/"
if ut.debugger_is_active():
- #default_storage_name = 'sqlite:///../models_torch/optuna/debug_01.db'
- default_storage_name = 'sqlite:///../../../data-project/KIWaVo/models/optuna/debug_03.db'
+ # default_storage_name = 'sqlite:///../models_torch/optuna/debug_01.db'
+ default_storage_name = (
+ "sqlite:///../../../data-project/KIWaVo/models/optuna/debug_02.db"
+ )
n_trials = 2
max_epochs = 2
default_max_trials = 1
else:
- #default_storage_name = 'sqlite:///../../models_torch/optuna/optimization_01.db'
- default_storage_name = 'sqlite:///../../../../data-project/KIWaVo/models/optuna/optimization_01.db'
+ # default_storage_name = 'sqlite:///../../models_torch/optuna/optimization_01.db'
+ default_storage_name = (
+ "sqlite:///../../../../data-project/KIWaVo/models/optuna/optimization_01.db"
+ )
n_trials = 100
max_epochs = 2000
default_max_trials = None
if use_cuda:
- #default_batch_size = 1024
+ # default_batch_size = 1024
default_batch_size = 256
-
- #This is for multi GPU training, doesn't work with optuna
- #devices = []
- #for j in range(torch.cuda.device_count()):
+ # This is for multi GPU training, doesn't work with optuna
+ # devices = []
+ # for j in range(torch.cuda.device_count()):
# free , max_gpu = torch.cuda.mem_get_info(j)
# if free > (max_gpu /2):
# devices.append(j)
@@ -62,15 +66,15 @@ if use_cuda:
max_free = 0
best_device = None
for j in range(torch.cuda.device_count()):
- free , _ = torch.cuda.mem_get_info(j)
+ free, _ = torch.cuda.mem_get_info(j)
if free > max_free:
max_free = free
best_device = j
devices = [best_device]
- #torch.cuda.set_per_process_memory_fraction(0.7, best_device)
+ # torch.cuda.set_per_process_memory_fraction(0.7, best_device)
else:
- devices = 'auto'
+ devices = "auto"
default_batch_size = 2048
@@ -89,7 +93,21 @@ class Objective:
train (float, optional): share of training data (first train %). Defaults to 0.7.
val (float, optional): share of validation data (directly after train). Defaults to .15.
"""
- def __init__(self, filename, gauge, logdir, in_size=144, out_size=48, batch_size=2048, percentile=0.95, train=0.7, val=.15,monitor= 'hp/val_loss', **kwargs):
+
+ def __init__(
+ self,
+ filename,
+ gauge,
+ logdir,
+ in_size=144,
+ out_size=48,
+ batch_size=2048,
+ percentile=0.95,
+ train=0.7,
+ val=0.15,
+ monitor="hp/val_loss",
+ **kwargs,
+ ):
# Hold these implementation specific arguments as the fields of the class.
self.filename = filename
self.level_name_org = gauge
@@ -103,19 +121,28 @@ class Objective:
self.monitor = monitor
def _get_callbacks(self, trial):
- checkpoint_callback = ModelCheckpoint(save_top_k=1, monitor=self.monitor,save_weights_only=True)
+ checkpoint_callback = ModelCheckpoint(
+ save_top_k=1, monitor=self.monitor, save_weights_only=True
+ )
pruning_callback = OptunaPruningCallback(trial, monitor=self.monitor)
- early_stop_callback = EarlyStopping(monitor=self.monitor, mode="min", patience=3)
+ early_stop_callback = EarlyStopping(
+ monitor=self.monitor, mode="min", patience=3
+ )
# TODO: update when optuna support the new namespace # pruning_callback = PyTorchLightningPruningCallback(trial, monitor="hp/val_loss")
my_callback = WaVoCallback()
- return my_callback, [checkpoint_callback,pruning_callback, early_stop_callback, my_callback]
+ return my_callback, [
+ checkpoint_callback,
+ pruning_callback,
+ early_stop_callback,
+ my_callback,
+ ]
def _get_model_params(self,trial : optuna.Trial):
#differencing = 0
differencing = trial.suggest_int("differencing", 0, 1)
learning_rate = trial.suggest_float("lr", 0.00001, 0.01)
- #optimizer = trial.suggest_categorical("optimizer", ["adam","adamw"])
+ # optimizer = trial.suggest_categorical("optimizer", ["adam","adamw"])
optimizer = trial.suggest_categorical("optimizer", ["adam"])
embed_time = trial.suggest_categorical("embed_time", [True,False])
#model_architecture = trial.suggest_categorical("model_architecture", ["classic_lstm"])
@@ -131,73 +158,79 @@ class Objective:
if model_architecture in ["chained_dense"]:
model_params = dict(
- activation=trial.suggest_categorical("activation", ["relu","gelu"]),
+ activation=trial.suggest_categorical("activation", ["relu", "gelu"]),
hidden_size=trial.suggest_int("hidden_size", 32, 512),
- dropout = trial.suggest_float("dropout", 0, 0.5),
+ dropout=trial.suggest_float("dropout", 0, 0.5),
)
if model_architecture in ["tsmixer"]:
model_params = dict(
- norm_type=trial.suggest_categorical("norm_type", ["L","B"]),
- activation=trial.suggest_categorical("activation", ["relu","gelu"]),
+ norm_type=trial.suggest_categorical("norm_type", ["L", "B"]),
+ activation=trial.suggest_categorical("activation", ["relu", "gelu"]),
n_block=trial.suggest_int("n_block", 1, 8),
- dropout = trial.suggest_float("dropout", 0, 0.5),
+ dropout=trial.suggest_float("dropout", 0, 0.5),
ff_dim=trial.suggest_int("ff_dim", 32, 1024),
)
- if model_architecture in ["classic_lstm","last_lstm"]:
+ if model_architecture in ["classic_lstm", "last_lstm"]:
model_params = dict(
- hidden_size_lstm = trial.suggest_int("hidden_size_lstm", 32, 512),
- num_layers_lstm = trial.suggest_int("n_layers_lstm", 1, 3),
- #num_layers_lstm = trial.suggest_int("n_layers_lstm", 1, 1),
- hidden_size = trial.suggest_int("hidden_size", 32, 512),
- num_layers = trial.suggest_int("n_layers", 2, 4),
- #num_layers = trial.suggest_int("n_layers", 2, 2),
- dropout = trial.suggest_float("dropout", 0, 0.5),
- )
+ hidden_size_lstm=trial.suggest_int("hidden_size_lstm", 32, 512),
+ num_layers_lstm=trial.suggest_int("n_layers_lstm", 1, 3),
+ # num_layers_lstm = trial.suggest_int("n_layers_lstm", 1, 1),
+ hidden_size=trial.suggest_int("hidden_size", 32, 512),
+ num_layers=trial.suggest_int("n_layers", 2, 4),
+ # num_layers = trial.suggest_int("n_layers", 2, 2),
+ dropout=trial.suggest_float("dropout", 0, 0.5),
+ )
- if model_architecture in ["transformer","autoformer","informer"]:
- n_heads_temp = 8 #trial.suggest_int("n_heads", 8, 32)
- d_model_temp = trial.suggest_int("d_model", 2*n_heads_temp, 256,2) #even number only, must be larger than n_heads (factor 2 is just guessing)
+ if model_architecture in ["transformer", "autoformer", "informer"]:
+ n_heads_temp = 8 # trial.suggest_int("n_heads", 8, 32)
+ d_model_temp = trial.suggest_int(
+ "d_model", 2 * n_heads_temp, 256, 2
+ ) # even number only, must be larger than n_heads (factor 2 is just guessing)
model_params = dict(
- d_model = d_model_temp,
- n_heads = n_heads_temp,
- #output_attention=trial.suggest_categorical("output_attention", [True,False]),
+ d_model=d_model_temp,
+ n_heads=n_heads_temp,
+ # output_attention=trial.suggest_categorical("output_attention", [True,False]),
e_layers=trial.suggest_int("e_layers", 1, 2),
d_layers=trial.suggest_int("d_layers", 1, 1),
d_ff=trial.suggest_int("d_ff", 32, 1024),
- dropout = trial.suggest_float("dropout", 0, 0.2),
- factor = trial.suggest_int("attn_factor", 1, 8),
+ dropout=trial.suggest_float("dropout", 0, 0.2),
+ factor=trial.suggest_int("attn_factor", 1, 8),
output_attention=False,
- #e_layers=2,
- #d_layers=1,
- activation=trial.suggest_categorical("activation", ["relu","gelu"]),
- #embed_type=trial.suggest_categorical("embed", ["fixed","timeF","neural"]),
- embed_type="timeF", #TODO remve
- )
+ # e_layers=2,
+ # d_layers=1,
+ activation=trial.suggest_categorical("activation", ["relu", "gelu"]),
+ # embed_type=trial.suggest_categorical("embed", ["fixed","timeF","neural"]),
+ embed_type="timeF", # TODO remve
+ )
if model_architecture == "autoformer":
- model_params["moving_avg"] = trial.suggest_int("moving_avg",23,71,2) #uneven numbers only
+ model_params["moving_avg"] = trial.suggest_int(
+ "moving_avg", 23, 71, 2
+ ) # uneven numbers only
if model_architecture == "informer":
- model_params["distil"] = trial.suggest_categorical("distil", [True,False])
+ model_params["distil"] = trial.suggest_categorical(
+ "distil", [True, False]
+ )
if model_architecture == "dlinear":
model_params = dict(
- individual=trial.suggest_categorical("individual", [True,False]),
- moving_avg = trial.suggest_int("moving_avg",23,71,2), #uneven numbers only
+ individual=trial.suggest_categorical("individual", [True, False]),
+ moving_avg=trial.suggest_int(
+ "moving_avg", 23, 71, 2
+ ), # uneven numbers only
)
return model_architecture, differencing, learning_rate,optimizer,embed_time, model_params
def __call__(self, trial):
- #def __call__(self, single_trial): Failed attempt at distributed training
- #trial = optuna.integration.TorchDistributedTrial(single_trial)
+ # def __call__(self, single_trial): Failed attempt at distributed training
+ # trial = optuna.integration.TorchDistributedTrial(single_trial)
model_architecture, differencing, learning_rate, optimizer,embed_time, model_params = self._get_model_params(trial)
#TODO gradient clipping?
#TODO to seed or not to seed?
#pl.seed_everything(42, workers=True)
- #prepare data
-
data_module = WaVoDataModule(
str(self.filename),
self.level_name_org,
@@ -213,7 +246,7 @@ class Objective:
embed_time=embed_time,
)
- #prepare model
+ # prepare model
model = WaVoLightningModule(
model_architecture=model_architecture,
feature_count=data_module.feature_count,
@@ -228,47 +261,54 @@ class Objective:
embed_time=data_module.hparams.embed_time,
**model_params)
- #find optimal batch_size
- #if default_max_trials is None:
+ # find optimal batch_size
+ # if default_max_trials is None:
# max_trials = 2 if model_architecture in ["transformer","autoformer"] else 5
- #else:
+ # else:
# max_trials = default_max_trials
- #logging.info("Trial %d Optimizing batch size with %d trials", trial.number, max_trials)
+ # logging.info("Trial %d Optimizing batch size with %d trials", trial.number, max_trials)
logging.info("Params: %s", trial.params)
- #logging.info("Before batch_size %.3f GB / %.3f GB", torch.cuda.mem_get_info(devices[0])[0]/1073741824, torch.cuda.mem_get_info(devices[0])[1]/1073741824)
- #trainer_bs = pl.Trainer(logger=False,accelerator=accelerator,devices=devices,max_epochs=1,)
- #tuner = Tuner(trainer_bs)
- #tuner.scale_batch_size(model, datamodule=data_module,init_val=256,max_trials=max_trials)
- #clear memory from optimizing the batch size
- #tuner = None
- #trainer_bs = None
- #gc.collect()
- #torch.cuda.empty_cache()
-
- logging.info("After batch_size %.3f GB / %.3f GB", torch.cuda.mem_get_info(devices[0])[0]/1073741824, torch.cuda.mem_get_info(devices[0])[1]/1073741824)
-
+ # logging.info("Before batch_size %.3f GB / %.3f GB", torch.cuda.mem_get_info(devices[0])[0]/1073741824, torch.cuda.mem_get_info(devices[0])[1]/1073741824)
+ # trainer_bs = pl.Trainer(logger=False,accelerator=accelerator,devices=devices,max_epochs=1,)
+ # tuner = Tuner(trainer_bs)
+ # tuner.scale_batch_size(model, datamodule=data_module,init_val=256,max_trials=max_trials)
+ # clear memory from optimizing the batch size
+ # tuner = None
+ # trainer_bs = None
+ # gc.collect()
+ # torch.cuda.empty_cache()
+
+ logging.info(
+ "After batch_size %.3f GB / %.3f GB",
+ torch.cuda.mem_get_info(devices[0])[0] / 1073741824,
+ torch.cuda.mem_get_info(devices[0])[1] / 1073741824,
+ )
my_callback, callbacks = self._get_callbacks(trial)
logger = TensorBoardLogger(self.log_dir, default_hp_metric=False)
-
- trainer = pl.Trainer(default_root_dir=self.log_dir,
- logger=logger,
- accelerator=accelerator,
- devices=devices,
- callbacks=callbacks,
- max_epochs=max_epochs,
- #log_every_n_steps=int((len(data_module.train_set) / data_module.hparams.batch_size)/4),#TODO remove
- log_every_n_steps=10,
- )
+ trainer = pl.Trainer(
+ default_root_dir=self.log_dir,
+ logger=logger,
+ accelerator=accelerator,
+ devices=devices,
+ callbacks=callbacks,
+ max_epochs=max_epochs,
+ # log_every_n_steps=int((len(data_module.train_set) / data_module.hparams.batch_size)/4),#TODO remove
+ log_every_n_steps=10,
+ )
trainer.fit(model, data_module)
- logging.info("After trial %.3f GB / %.3f GB", torch.cuda.mem_get_info(devices[0])[0]/1073741824, torch.cuda.mem_get_info(devices[0])[1]/1073741824)
+ logging.info(
+ "After trial %.3f GB / %.3f GB",
+ torch.cuda.mem_get_info(devices[0])[0] / 1073741824,
+ torch.cuda.mem_get_info(devices[0])[1] / 1073741824,
+ )
- #save metrics to optuna
+ # save metrics to optuna
model_path = str(Path(trainer.log_dir).resolve())
logging.info("model_path: %s", model_path)
trial.set_user_attr("model_path", model_path)
- for metric in ['hp/val_nse', 'hp/val_mae', 'hp/val_mae_flood']:
+ for metric in ["hp/val_nse", "hp/val_mae", "hp/val_mae_flood"]:
for i in [23, 47]:
trial.set_user_attr(f'{metric}_{i+1}', my_callback.metrics[metric][i].item())
@@ -277,23 +317,49 @@ class Objective:
def parse_args() -> argparse.Namespace:
- """ Parse all the arguments and provides some help in the command line
- """
+ """Parse all the arguments and provides some help in the command line"""
parser: argparse.ArgumentParser = argparse.ArgumentParser(
- description='Execute Hyperparameter optimization with optuna and torchlightning.')
- parser.add_argument('filename', metavar='datafile', type=Path,
- help='The path to your input data.')
- parser.add_argument('gauge', metavar='gaugename',
- type=str, help='The name of the gauge column.')
- parser.add_argument('logdir', type=Path,
- help='set a directory for logs and model checkpoints.')
- parser.add_argument('trials', metavar='trials',type=int, default=n_trials,help='How many trials to run.')
-
- parser.add_argument('--expname', metavar='experiment_name',type=str, default='nameless',help='The name of the experiment.')
- parser.add_argument('--storagename', metavar='storage_name',type=str, default=None,help='The database for the experiment.')
- parser.add_argument('--architecture', metavar='architecture',type=str, default=None,help='Hyperparameter to decide an architecture.')
+ description="Execute Hyperparameter optimization with optuna and torchlightning."
+ )
+ parser.add_argument(
+ "filename", metavar="datafile", type=Path, help="The path to your input data."
+ )
+ parser.add_argument(
+ "gauge", metavar="gaugename", type=str, help="The name of the gauge column."
+ )
+ parser.add_argument(
+ "logdir", type=Path, help="set a directory for logs and model checkpoints."
+ )
+ parser.add_argument(
+ "trials",
+ metavar="trials",
+ type=int,
+ default=n_trials,
+ help="How many trials to run.",
+ )
+ parser.add_argument(
+ "--expname",
+ metavar="experiment_name",
+ type=str,
+ default="nameless",
+ help="The name of the experiment.",
+ )
+ parser.add_argument(
+ "--storagename",
+ metavar="storage_name",
+ type=str,
+ default=None,
+ help="The database for the experiment.",
+ )
+ parser.add_argument(
+ "--architecture",
+ metavar="architecture",
+ type=str,
+ default=None,
+ help="Hyperparameter to decide an architecture.",
+ )
parser.add_argument(
"--pruning",
@@ -306,57 +372,80 @@ def parse_args() -> argparse.Namespace:
return parser.parse_args()
+
def main():
- """Start a hyperparameter optimization with optuna and torchlightning.
- """
+ """Start a hyperparameter optimization with optuna and torchlightning."""
parsed_args = parse_args()
if not parsed_args.logdir.exists():
parsed_args.logdir.mkdir(parents=True)
-
if parsed_args.pruning:
- pruner = optuna.pruners.HyperbandPruner(min_resource=1, max_resource='auto', reduction_factor=3, bootstrap_count=0)
+ pruner = optuna.pruners.HyperbandPruner(
+ min_resource=1, max_resource="auto", reduction_factor=3, bootstrap_count=0
+ )
else:
pruner = optuna.pruners.NopPruner()
-
-
- study_name = f"{parsed_args.filename.stem} {parsed_args.expname}" # Unique identifier of the study.
- storage_name = default_storage_name if parsed_args.storagename is None else f"{storage_base}{parsed_args.storagename}.db"
+ study_name = f"{parsed_args.filename.stem} {parsed_args.expname}" # Unique identifier of the study.
+ storage_name = (
+ default_storage_name
+ if parsed_args.storagename is None
+ else f"{storage_base}{parsed_args.storagename}.db"
+ )
# Logging, add stream handler of stdout to show the messages
logging.basicConfig(level=logging.INFO)
- logFormatter = logging.Formatter('%(asctime)s;%(levelname)s;%(message)s',datefmt='%Y-%m-%d %H:%M:%S')
- fileHandler = logging.FileHandler(parsed_args.logdir / 'optuna.log',)
+ logFormatter = logging.Formatter(
+ "%(asctime)s;%(levelname)s;%(message)s", datefmt="%Y-%m-%d %H:%M:%S"
+ )
+ fileHandler = logging.FileHandler(
+ parsed_args.logdir / "optuna.log",
+ )
consoleHandler = logging.StreamHandler(sys.stdout)
fileHandler.setFormatter(logFormatter)
consoleHandler.setFormatter(logFormatter)
logging.getLogger().addHandler(fileHandler)
- #logging.getLogger().addHandler(consoleHandler)
+ # logging.getLogger().addHandler(consoleHandler)
optuna.logging.get_logger("optuna").addHandler(fileHandler)
optuna.logging.get_logger("optuna").addHandler(consoleHandler)
- logging.info('Start of this execution======================================================================')
- logging.info("Executing %s with device %s and parameters %s ",sys.argv[0],devices,sys.argv[1:])
-
+ logging.info(
+ "Start of this execution======================================================================"
+ )
+ logging.info(
+ "Executing %s with device %s and parameters %s ",
+ sys.argv[0],
+ devices,
+ sys.argv[1:],
+ )
study = optuna.create_study(
study_name=study_name,
storage=storage_name,
direction="minimize",
pruner=pruner,
- load_if_exists=True)
+ load_if_exists=True,
+ )
study.set_metric_names(["hp/val_loss"])
- objective = Objective(**vars(parsed_args),batch_size=1024)#TODO REMOVE batch_size
+ objective = Objective(
+ **vars(parsed_args), batch_size=1024
+ ) # TODO REMOVE batch_size
if parsed_args.architecture is not None:
for _ in range(parsed_args.trials):
study.enqueue_trial({"model_architecture": parsed_args.architecture})
- study.optimize(objective, n_trials=parsed_args.trials, gc_after_trial=True,timeout=None,callbacks=[lambda study, trial: torch.cuda.empty_cache()])
+ study.optimize(
+ objective,
+ n_trials=parsed_args.trials,
+ gc_after_trial=True,
+ timeout=None,
+ callbacks=[lambda study, trial: torch.cuda.empty_cache()],
+ )
+
if __name__ == "__main__":
main()
- #python main_hyper.py ../../../../data-project/KIWaVo/data/input/mergedPreetzAll.csv Preetz_pegel_cm ../../../../data-project/KIWaVo/models/lfu/preetz/ 100 --expname lstms_1 --storagename lfu
+ # python main_hyper.py ../../../../data-project/KIWaVo/data/input/mergedPreetzAll.csv Preetz_pegel_cm ../../../../data-project/KIWaVo/models/lfu/preetz/ 100 --expname lstms_1 --storagename lfu
diff --git a/src/ensemble_models.py b/src/models/ensemble_models.py
similarity index 100%
rename from src/ensemble_models.py
rename to src/models/ensemble_models.py
diff --git a/src/models/lightning_module.py b/src/models/lightning_module.py
index 0b575b282a33d656d7d4c0c593ed5d9506165b85..2d987a75f06cff996c759972bab514941d36fb53 100644
--- a/src/models/lightning_module.py
+++ b/src/models/lightning_module.py
@@ -150,7 +150,8 @@ class WaVoLightningModule(pl.LightningModule):
x_org = self.scaler.inverse_transform(x.reshape(-1, self.feature_count).cpu()).reshape(-1,self.in_size,self.feature_count)
x_base = x_org[:,-1,self.gauge_idx].round(2) #round to get rid of floating point errors, the measure value is always an integer or has at most 1 decimal
x_base = torch.from_numpy(x_base).unsqueeze(1)
- pred = x_base.cuda() + pred.cumsum(dim=1)
+ x_base = x_base.to(pred.device)
+ pred = x_base + pred.cumsum(dim=1)
return pred
diff --git a/src/example_sarah.py b/src/old/example_sarah.py
similarity index 55%
rename from src/example_sarah.py
rename to src/old/example_sarah.py
index 696ab285c6c0ed41c961ebebb47e270421c3a8b0..b27c9e5acafa56139b1977edbb9c06244c812dc8 100644
--- a/src/example_sarah.py
+++ b/src/old/example_sarah.py
@@ -1,6 +1,7 @@
"""
Main Module to start training
"""
+
import argparse
import logging
import sys
@@ -14,9 +15,10 @@ from lightning.pytorch.loggers import TensorBoardLogger
from models.lstms import LSTM_LAST
import utils.utility as ut
-from callbacks import OptunaPruningCallback, WaVoCallback
-#from data_tools.data_module import WaVoDataModule
-#from models.lightning_module import WaVoLightningModule
+from utils.callbacks import OptunaPruningCallback, WaVoCallback
+
+# from data_tools.data_module import WaVoDataModule
+# from models.lightning_module import WaVoLightningModule
from datetime import datetime
import pickle
@@ -28,25 +30,24 @@ from sklearn.preprocessing import StandardScaler
from torch.utils.data import DataLoader
import torch.nn as nn
import torch.optim as optim
-from data_tools.datasets import TimeSeriesDataSet # pylint: disable=import-error
-import utils.utility as ut # pylint: disable=import-error
+from data_tools.datasets import TimeSeriesDataSet # pylint: disable=import-error
+import utils.utility as ut # pylint: disable=import-error
use_cuda = torch.cuda.is_available()
if use_cuda:
- accelerator = 'cuda'
- torch.set_float32_matmul_precision('high')
+ accelerator = "cuda"
+ torch.set_float32_matmul_precision("high")
devices = [0]
else:
- accelerator = 'cpu'
+ accelerator = "cpu"
devices = None
# CONSTANTS
-storage_base = 'sqlite:///../../../../data-project/KIWaVo/sarah/optuna/'
+storage_base = "sqlite:///../../../../data-project/KIWaVo/sarah/optuna/"
max_epochs = 200
n_trials = 3
-
class WaVoLightningModule(pl.LightningModule):
"""LightningModule that defines the training and other steps, logging and the neural network.
@@ -82,16 +83,24 @@ class WaVoLightningModule(pl.LightningModule):
"""
# pylint: disable-next=unused-argument
- def __init__(self,feature_count, in_size, out_size,learning_rate,optimizer,scaler,target_idx,**kwargs):
+ def __init__(
+ self,
+ feature_count,
+ in_size,
+ out_size,
+ learning_rate,
+ optimizer,
+ scaler,
+ target_idx,
+ **kwargs,
+ ):
super().__init__()
+ self.save_hyperparameters(ignore=["scaler"])
+ # self.save_hyperparameters({"scaler": pickle.dumps(self.scaler),})
-
- self.save_hyperparameters(ignore=['scaler'])
- #self.save_hyperparameters({"scaler": pickle.dumps(self.scaler),})
-
- if 'timestamp' in kwargs: #TODO test if this works
- self.timestamp = kwargs['timestamp']
+ if "timestamp" in kwargs: # TODO test if this works
+ self.timestamp = kwargs["timestamp"]
else:
self.timestamp = datetime.now().strftime("%Y%m%d-%H%M%S")
self.save_hyperparameters({"timestamp": self.timestamp})
@@ -103,7 +112,6 @@ class WaVoLightningModule(pl.LightningModule):
self.target_idx = target_idx
self.model = self.create_model(**kwargs)
-
# pylint: disable-next=arguments-differ
def forward(self, x):
return self.model(x)
@@ -118,7 +126,9 @@ class WaVoLightningModule(pl.LightningModule):
# pylint: disable-next=unused-argument, arguments-differ
def validation_step(self, batch, batch_idx):
val_loss = self._common_step(batch)
- self.log("hp/val_loss", val_loss, sync_dist=True)# syncdist needed for optuna hyperparameter optimization to work
+ self.log(
+ "hp/val_loss", val_loss, sync_dist=True
+ ) # syncdist needed for optuna hyperparameter optimization to work
# pylint: disable-next=unused-argument, arguments-differ
def test_step(self, batch, batch_idx):
@@ -127,21 +137,25 @@ class WaVoLightningModule(pl.LightningModule):
def predict_step(self, batch, batch_idx, dataloader_idx=0):
x = batch[0]
- #This is not great, but necessary if i want to use a scaler.
+ # This is not great, but necessary if i want to use a scaler.
pred = self.model(x)
- pred = ut.inv_standard(pred, self.scaler.mean_[self.target_idx], self.scaler.scale_[self.target_idx])
+ pred = ut.inv_standard(
+ pred,
+ self.scaler.mean_[self.target_idx],
+ self.scaler.scale_[self.target_idx],
+ )
return pred
- def _common_step(self,batch):
+ def _common_step(self, batch):
x, y = batch
y_hat = self.model(x)
- loss = nn.functional.mse_loss(y_hat, y) #TODO different loss functions
+ loss = nn.functional.mse_loss(y_hat, y) # TODO different loss functions
return loss
def configure_optimizers(self):
- if 'optimizer' not in self.hparams or self.hparams.optimizer == 'adam':
+ if "optimizer" not in self.hparams or self.hparams.optimizer == "adam":
optimizer = optim.Adam(self.parameters(), lr=self.hparams.learning_rate)
- elif self.hparams.optimizer =='adamw':
+ elif self.hparams.optimizer == "adamw":
optimizer = optim.AdamW(self.parameters(), lr=self.hparams.learning_rate)
return optimizer
@@ -153,7 +167,7 @@ class WaVoLightningModule(pl.LightningModule):
"""
return self.model
- def create_model(self,**kwargs) -> nn.Module:
+ def create_model(self, **kwargs) -> nn.Module:
"""Creates the neural network model and passes the kwargs to the model.
Returns:
@@ -163,7 +177,9 @@ class WaVoLightningModule(pl.LightningModule):
feature_count=self.feature_count,
in_size=self.in_size,
out_size=self.out_size,
- **kwargs)
+ **kwargs,
+ )
+
class WaVoDataModule(pl.LightningDataModule):
"""
@@ -181,15 +197,19 @@ class WaVoDataModule(pl.LightningDataModule):
scaler (StandardScaler, optional): Scaler to use for input data. Only given when the DataModule is created for prediction. Defaults to None.
"""
- def __init__(self, filename, level_name_org,
- batch_size=128,
- in_size=144,
- out_size=48,
- percentile=0.95,
- train=0.7,
- val=.15,
- scaler=None,
- **kwargs) -> None:
+ def __init__(
+ self,
+ filename,
+ level_name_org,
+ batch_size=128,
+ in_size=144,
+ out_size=48,
+ percentile=0.95,
+ train=0.7,
+ val=0.15,
+ scaler=None,
+ **kwargs,
+ ) -> None:
super().__init__()
self.filename = str(filename)
self.target_column = level_name_org
@@ -200,57 +220,75 @@ class WaVoDataModule(pl.LightningDataModule):
self.train = train
self.val = val
self.scaler = scaler
- #maybe tidy up the next 4 lines
- self.threshold = kwargs['threshold'] if 'threshold' in kwargs else None
- self.feature_count = kwargs['feature_count'] if 'feature_count' in kwargs else None
+ # maybe tidy up the next 4 lines
+ self.threshold = kwargs["threshold"] if "threshold" in kwargs else None
+ self.feature_count = (
+ kwargs["feature_count"] if "feature_count" in kwargs else None
+ )
self.df = self._load_df()
- self.target_idx =self.df.columns.get_loc(self.target_column)
+ self.target_idx = self.df.columns.get_loc(self.target_column)
# if we instantiate the object from scratch we do not have these values. If we load it with the config/hparams from a model we do.
- if all(attr is None for attr in [self.scaler,self.threshold,self.feature_count]):
+ if all(
+ attr is None for attr in [self.scaler, self.threshold, self.feature_count]
+ ):
self._set_threshold()
df = self.df.copy()
# drop the first value, it's nan if differencing, but for comparing metrics we need to always drop it
- val_idx = int(self.train*len(df))
- test_idx = int(val_idx + self.val*len(df))
+ val_idx = int(self.train * len(df))
+ test_idx = int(val_idx + self.val * len(df))
- df_train= df[:val_idx]
+ df_train = df[:val_idx]
self.feature_count = df_train.shape[-1]
self.scaler = StandardScaler()
self.scaler.fit(df_train)
- self.save_hyperparameters(ignore=['scaler','threshold','feature_count'])
- self.save_hyperparameters({
- "scaler": pickle.dumps(self.scaler),
- "threshold": self.threshold,
- #These aren't really hyperparameters, but maybe useful to have them logged
- "train_start":str(df.index[0]),
- "val_start":str(df.index[val_idx]),
- "test_start":str(df.index[test_idx]),
- })
-
- elif any(attr is None for attr in [self.scaler, self.threshold,self.feature_count]):
- raise ValueError("If you provide a scaler, you also need to provide threshold and feature_count.")
+ self.save_hyperparameters(ignore=["scaler", "threshold", "feature_count"])
+ self.save_hyperparameters(
+ {
+ "scaler": pickle.dumps(self.scaler),
+ "threshold": self.threshold,
+ # These aren't really hyperparameters, but maybe useful to have them logged
+ "train_start": str(df.index[0]),
+ "val_start": str(df.index[val_idx]),
+ "test_start": str(df.index[test_idx]),
+ }
+ )
+
+ elif any(
+ attr is None for attr in [self.scaler, self.threshold, self.feature_count]
+ ):
+ raise ValueError(
+ "If you provide a scaler, you also need to provide threshold and feature_count."
+ )
else:
- self.save_hyperparameters(ignore=['scaler'])
+ self.save_hyperparameters(ignore=["scaler"])
self.save_hyperparameters({"scaler": pickle.dumps(self.scaler)})
- self.df_scaled = pd.DataFrame(self.scaler.transform(self.df),index=self.df.index,columns=self.df.columns)
+ self.df_scaled = pd.DataFrame(
+ self.scaler.transform(self.df), index=self.df.index, columns=self.df.columns
+ )
def _load_df(self):
df = pd.read_csv(self.filename, index_col=0, parse_dates=True)
- if not isinstance(df.index,pd.core.indexes.datetimes.DatetimeIndex):
- df = pd.read_csv(self.filename,index_col=0,parse_dates=True,date_format='%d.%m.%Y %H:00')
+ if not isinstance(df.index, pd.core.indexes.datetimes.DatetimeIndex):
+ df = pd.read_csv(
+ self.filename,
+ index_col=0,
+ parse_dates=True,
+ date_format="%d.%m.%Y %H:00",
+ )
df = ut.fill_missing_values(df)
return df
def _set_threshold(self):
if 0 < self.percentile < 1:
- self.threshold = self.df[self.target_column].quantile(
- self.percentile).item()
+ self.threshold = (
+ self.df[self.target_column].quantile(self.percentile).item()
+ )
else:
self.threshold = self.percentile
@@ -260,29 +298,54 @@ class WaVoDataModule(pl.LightningDataModule):
# I don't really do setup i guess?
# That train_set objects etc don't have a state, so no need to reload them or anything.
# Additionally, i need all of them in my custom callback so i'll just load them once in the beginning and then ignore the setup method
- if not hasattr(self,'train_set'):
+ if not hasattr(self, "train_set"):
df_len = len(self.df)
- val_idx = int(self.train*df_len)
- test_idx = int(val_idx + self.val*df_len)
- self.train_set = self.make_dataset(1,val_idx+1)
- self.val_set = self.make_dataset(val_idx+1,test_idx+1)
- self.test_set = self.make_dataset(test_idx+1)
-
+ val_idx = int(self.train * df_len)
+ test_idx = int(val_idx + self.val * df_len)
+ self.train_set = self.make_dataset(1, val_idx + 1)
+ self.val_set = self.make_dataset(val_idx + 1, test_idx + 1)
+ self.test_set = self.make_dataset(test_idx + 1)
def train_dataloader(self):
- return DataLoader(self.train_set, batch_size=self.hparams.batch_size, shuffle=True, num_workers=8,persistent_workers=True)
+ return DataLoader(
+ self.train_set,
+ batch_size=self.hparams.batch_size,
+ shuffle=True,
+ num_workers=8,
+ persistent_workers=True,
+ )
def val_dataloader(self):
- return DataLoader(self.val_set, batch_size=self.hparams.batch_size, shuffle=False, num_workers=8,persistent_workers=True)
+ return DataLoader(
+ self.val_set,
+ batch_size=self.hparams.batch_size,
+ shuffle=False,
+ num_workers=8,
+ persistent_workers=True,
+ )
def test_dataloader(self):
- return DataLoader(self.test_set, batch_size=self.hparams.batch_size, shuffle=False, num_workers=8,persistent_workers=True)
+ return DataLoader(
+ self.test_set,
+ batch_size=self.hparams.batch_size,
+ shuffle=False,
+ num_workers=8,
+ persistent_workers=True,
+ )
def predict_dataloader(self):
# return the trainset, but sorted
- return DataLoader(self.train_set, batch_size=self.hparams.batch_size, shuffle=False, num_workers=8,persistent_workers=True)
+ return DataLoader(
+ self.train_set,
+ batch_size=self.hparams.batch_size,
+ shuffle=False,
+ num_workers=8,
+ persistent_workers=True,
+ )
- def make_dataset(self,start:int,end:int=None,exog=False) -> TimeSeriesDataSet:
+ def make_dataset(
+ self, start: int, end: int = None, exog=False
+ ) -> TimeSeriesDataSet:
"""Makes a TimeseriesDataset from the given range of the available data.
Uses values from self.df_scaled.
@@ -296,11 +359,9 @@ class WaVoDataModule(pl.LightningDataModule):
x = torch.Tensor(self.df_scaled[start:end].values)
y = torch.Tensor(self.df_scaled[self.target_column][start:end].values)
- #y = torch.Tensor(self.df_y[start:end].values[:,0])
-
- return TimeSeriesDataSet(x,y,self.in_size,self.out_size)
-
+ # y = torch.Tensor(self.df_y[start:end].values[:,0])
+ return TimeSeriesDataSet(x, y, self.in_size, self.out_size)
class Objective:
@@ -319,7 +380,21 @@ class Objective:
val (float, optional): share of validation data (directly after train). Defaults to .15.
monitor (str, optional): metric to monitor. Defaults to 'hp/val_loss'.
"""
- def __init__(self, filename, gauge, logdir, in_size=144, out_size=48, batch_size=2048, percentile=0.95, train=0.7, val=.15,monitor= 'hp/val_loss', **kwargs):
+
+ def __init__(
+ self,
+ filename,
+ gauge,
+ logdir,
+ in_size=144,
+ out_size=48,
+ batch_size=2048,
+ percentile=0.95,
+ train=0.7,
+ val=0.15,
+ monitor="hp/val_loss",
+ **kwargs,
+ ):
# Hold these implementation specific arguments as the fields of the class.
self.filename = filename
self.level_name_org = gauge
@@ -333,36 +408,45 @@ class Objective:
self.monitor = monitor
def _get_callbacks(self, trial):
- checkpoint_callback = ModelCheckpoint(save_top_k=1, monitor=self.monitor,save_weights_only=True)
+ checkpoint_callback = ModelCheckpoint(
+ save_top_k=1, monitor=self.monitor, save_weights_only=True
+ )
pruning_callback = OptunaPruningCallback(trial, monitor=self.monitor)
- early_stop_callback = EarlyStopping(monitor=self.monitor, mode="min", patience=3)
+ early_stop_callback = EarlyStopping(
+ monitor=self.monitor, mode="min", patience=3
+ )
my_callback = WaVoCallback()
- return my_callback, [checkpoint_callback,pruning_callback, early_stop_callback, my_callback]
-
- def _get_model_params(self,trial):
+ return my_callback, [
+ checkpoint_callback,
+ pruning_callback,
+ early_stop_callback,
+ my_callback,
+ ]
+
+ def _get_model_params(self, trial):
learning_rate = trial.suggest_float("lr", 0.00001, 0.01)
optimizer = trial.suggest_categorical("optimizer", ["adam"])
model_params = dict(
- hidden_size_lstm = trial.suggest_int("hidden_size_lstm", 32, 512),
- num_layers_lstm = trial.suggest_int("n_layers_lstm", 1, 3),
- hidden_size = trial.suggest_int("hidden_size", 32, 512),
- num_layers = trial.suggest_int("n_layers", 2, 4),
- dropout = trial.suggest_float("dropout", 0, 0.5),
- )
-
- return learning_rate,optimizer, model_params
+ hidden_size_lstm=trial.suggest_int("hidden_size_lstm", 32, 512),
+ num_layers_lstm=trial.suggest_int("n_layers_lstm", 1, 3),
+ hidden_size=trial.suggest_int("hidden_size", 32, 512),
+ num_layers=trial.suggest_int("n_layers", 2, 4),
+ dropout=trial.suggest_float("dropout", 0, 0.5),
+ )
+
+ return learning_rate, optimizer, model_params
def __call__(self, trial):
- #def __call__(self, single_trial): Failed attempt at distributed training
- #trial = optuna.integration.TorchDistributedTrial(single_trial)
+ # def __call__(self, single_trial): Failed attempt at distributed training
+ # trial = optuna.integration.TorchDistributedTrial(single_trial)
learning_rate, optimizer, model_params = self._get_model_params(trial)
- #TODO gradient clipping?
- #TODO to seed or not to seed?
- #pl.seed_everything(42, workers=True)
+ # TODO gradient clipping?
+ # TODO to seed or not to seed?
+ # pl.seed_everything(42, workers=True)
- #prepare data (This could maybe be done in init?)
+ # prepare data (This could maybe be done in init?)
data_module = WaVoDataModule(
str(self.filename),
self.level_name_org,
@@ -374,7 +458,7 @@ class Objective:
val=self.val,
)
- #prepare model
+ # prepare model
model = WaVoLightningModule(
feature_count=data_module.feature_count,
in_size=data_module.hparams.in_size,
@@ -383,57 +467,85 @@ class Objective:
target_idx=data_module.target_idx,
learning_rate=learning_rate,
optimizer=optimizer,
- **model_params)
+ **model_params,
+ )
logging.info("Params: %s", trial.params)
- #logging.info("After batch_size %.3f GB / %.3f GB", torch.cuda.mem_get_info(devices[0])[0]/1073741824, torch.cuda.mem_get_info(devices[0])[1]/1073741824)
-
+ # logging.info("After batch_size %.3f GB / %.3f GB", torch.cuda.mem_get_info(devices[0])[0]/1073741824, torch.cuda.mem_get_info(devices[0])[1]/1073741824)
my_callback, callbacks = self._get_callbacks(trial)
logger = TensorBoardLogger(self.log_dir, default_hp_metric=False)
-
- trainer = pl.Trainer(default_root_dir=self.log_dir,
- logger=logger,
- accelerator=accelerator,
- devices=devices,
- callbacks=callbacks,
- max_epochs=max_epochs,
- #log_every_n_steps=int((len(data_module.train_set) / data_module.hparams.batch_size)/4),#TODO remove
- log_every_n_steps=10,
- )
+ trainer = pl.Trainer(
+ default_root_dir=self.log_dir,
+ logger=logger,
+ accelerator=accelerator,
+ devices=devices,
+ callbacks=callbacks,
+ max_epochs=max_epochs,
+ # log_every_n_steps=int((len(data_module.train_set) / data_module.hparams.batch_size)/4),#TODO remove
+ log_every_n_steps=10,
+ )
trainer.fit(model, data_module)
- #logging.info("After trial %.3f GB / %.3f GB", torch.cuda.mem_get_info(devices[0])[0]/1073741824, torch.cuda.mem_get_info(devices[0])[1]/1073741824)
+ # logging.info("After trial %.3f GB / %.3f GB", torch.cuda.mem_get_info(devices[0])[0]/1073741824, torch.cuda.mem_get_info(devices[0])[1]/1073741824)
- #save metrics to optuna
+ # save metrics to optuna
model_path = str(Path(trainer.log_dir).resolve())
logging.info("model_path: %s", model_path)
trial.set_user_attr("model_path", model_path)
- for metric in ['hp/val_nse', 'hp/val_mae', 'hp/val_mae_flood']:
+ for metric in ["hp/val_nse", "hp/val_mae", "hp/val_mae_flood"]:
for i in [23, 47]:
- trial.set_user_attr(f'{metric}_{i}', my_callback.metrics[metric][i].item())
+ trial.set_user_attr(
+ f"{metric}_{i}", my_callback.metrics[metric][i].item()
+ )
return my_callback.metrics[self.monitor].item()
def parse_args() -> argparse.Namespace:
- """ Parse all the arguments and provides some help in the command line
- """
+ """Parse all the arguments and provides some help in the command line"""
parser: argparse.ArgumentParser = argparse.ArgumentParser(
- description='Execute Hyperparameter optimization with optuna and torchlightning.')
- parser.add_argument('filename', metavar='datafile', type=Path,
- help='The path to your input data.')
- parser.add_argument('gauge', metavar='gaugename',
- type=str, help='The name of the gauge column.')
- parser.add_argument('logdir', type=Path,
- help='set a directory for logs and model checkpoints.')
- parser.add_argument('trials', metavar='trials',type=int, default=n_trials,help='How many trials to run.')
-
- parser.add_argument('--expname', metavar='experiment_name',type=str, default='nameless',help='The name of the experiment.')
- parser.add_argument('--storagename', metavar='storage_name',type=str, default=None,help='The database for the experiment.')
- parser.add_argument('--architecture', metavar='architecture',type=str, default=None,help='Hyperparameter to decide an architecture.')
+ description="Execute Hyperparameter optimization with optuna and torchlightning."
+ )
+ parser.add_argument(
+ "filename", metavar="datafile", type=Path, help="The path to your input data."
+ )
+ parser.add_argument(
+ "gauge", metavar="gaugename", type=str, help="The name of the gauge column."
+ )
+ parser.add_argument(
+ "logdir", type=Path, help="set a directory for logs and model checkpoints."
+ )
+ parser.add_argument(
+ "trials",
+ metavar="trials",
+ type=int,
+ default=n_trials,
+ help="How many trials to run.",
+ )
+ parser.add_argument(
+ "--expname",
+ metavar="experiment_name",
+ type=str,
+ default="nameless",
+ help="The name of the experiment.",
+ )
+ parser.add_argument(
+ "--storagename",
+ metavar="storage_name",
+ type=str,
+ default=None,
+ help="The database for the experiment.",
+ )
+ parser.add_argument(
+ "--architecture",
+ metavar="architecture",
+ type=str,
+ default=None,
+ help="Hyperparameter to decide an architecture.",
+ )
parser.add_argument(
"--pruning",
@@ -446,57 +558,79 @@ def parse_args() -> argparse.Namespace:
return parser.parse_args()
+
def main():
- """Start a hyperparameter optimization with optuna and torchlightning.
- """
+ """Start a hyperparameter optimization with optuna and torchlightning."""
parsed_args = parse_args()
if not parsed_args.logdir.exists():
parsed_args.logdir.mkdir(parents=True)
-
if parsed_args.pruning:
- pruner = optuna.pruners.HyperbandPruner(min_resource=1, max_resource='auto', reduction_factor=3, bootstrap_count=0)
+ pruner = optuna.pruners.HyperbandPruner(
+ min_resource=1, max_resource="auto", reduction_factor=3, bootstrap_count=0
+ )
else:
pruner = optuna.pruners.NopPruner()
-
-
- study_name = f"{parsed_args.filename.stem} {parsed_args.expname}" # Unique identifier of the study.
- storage_name = f"{storage_base}default.db" if parsed_args.storagename is None else f"{storage_base}{parsed_args.storagename}.db"
+ study_name = f"{parsed_args.filename.stem} {parsed_args.expname}" # Unique identifier of the study.
+ storage_name = (
+ f"{storage_base}default.db"
+ if parsed_args.storagename is None
+ else f"{storage_base}{parsed_args.storagename}.db"
+ )
# Logging, add stream handler of stdout to show the messages
logging.basicConfig(level=logging.INFO)
- logFormatter = logging.Formatter('%(asctime)s;%(levelname)s;%(message)s',datefmt='%Y-%m-%d %H:%M:%S')
- fileHandler = logging.FileHandler(parsed_args.logdir / 'optuna.log',)
+ logFormatter = logging.Formatter(
+ "%(asctime)s;%(levelname)s;%(message)s", datefmt="%Y-%m-%d %H:%M:%S"
+ )
+ fileHandler = logging.FileHandler(
+ parsed_args.logdir / "optuna.log",
+ )
consoleHandler = logging.StreamHandler(sys.stdout)
fileHandler.setFormatter(logFormatter)
consoleHandler.setFormatter(logFormatter)
logging.getLogger().addHandler(fileHandler)
- #logging.getLogger().addHandler(consoleHandler)
+ # logging.getLogger().addHandler(consoleHandler)
optuna.logging.get_logger("optuna").addHandler(fileHandler)
optuna.logging.get_logger("optuna").addHandler(consoleHandler)
- logging.info('Start of this execution======================================================================')
- logging.info("Executing %s with device %s and parameters %s ",sys.argv[0],devices,sys.argv[1:])
-
+ logging.info(
+ "Start of this execution======================================================================"
+ )
+ logging.info(
+ "Executing %s with device %s and parameters %s ",
+ sys.argv[0],
+ devices,
+ sys.argv[1:],
+ )
study = optuna.create_study(
study_name=study_name,
storage=storage_name,
direction="minimize",
pruner=pruner,
- load_if_exists=True)
+ load_if_exists=True,
+ )
study.set_metric_names(["hp/val_loss"])
- objective = Objective(**vars(parsed_args),batch_size=1024, gc_after_trial=True)#TODO REMOVE batch_size
+ objective = Objective(
+ **vars(parsed_args), batch_size=1024, gc_after_trial=True
+ ) # TODO REMOVE batch_size
if parsed_args.architecture is not None:
for _ in range(parsed_args.trials):
study.enqueue_trial({"model_architecture": parsed_args.architecture})
- study.optimize(objective, n_trials=parsed_args.trials, timeout=None,callbacks=[lambda study, trial: torch.cuda.empty_cache()])
+ study.optimize(
+ objective,
+ n_trials=parsed_args.trials,
+ timeout=None,
+ callbacks=[lambda study, trial: torch.cuda.empty_cache()],
+ )
+
if __name__ == "__main__":
main()
- #python main_hyper.py ../../../../data-project/KIWaVo/data/input/mergedPreetzAll.csv Preetz_pegel_cm ../../../../data-project/KIWaVo/models/lfu/preetz/ 100 --expname lstms_1 --storagename lfu
+ # python main_hyper.py ../../../../data-project/KIWaVo/data/input/mergedPreetzAll.csv Preetz_pegel_cm ../../../../data-project/KIWaVo/models/lfu/preetz/ 100 --expname lstms_1 --storagename lfu
diff --git a/src/exp_icaart.py b/src/old/exp_icaart.py
similarity index 50%
rename from src/exp_icaart.py
rename to src/old/exp_icaart.py
index a608f93b5ea09ac3a1b31e0449177ddad0bf670a..7de5748c69618c2f99d73d12913a5e3c3d4a3828 100644
--- a/src/exp_icaart.py
+++ b/src/old/exp_icaart.py
@@ -1,6 +1,7 @@
"""
Main Module to start training
"""
+
import argparse
import gc
import logging
@@ -14,33 +15,37 @@ from lightning.pytorch.callbacks import EarlyStopping, ModelCheckpoint
from lightning.pytorch.loggers import TensorBoardLogger
import utils.utility as ut
-from callbacks import OptunaPruningCallback, WaVoCallback
+from utils.callbacks import OptunaPruningCallback, WaVoCallback
from data_tools.data_module import WaVoDataModule
from models.lightning_module import WaVoLightningModule
use_cuda = torch.cuda.is_available()
if use_cuda:
- accelerator = 'cuda'
- #accelerator = 'gpu'
- torch.set_float32_matmul_precision('high')
+ accelerator = "cuda"
+ # accelerator = 'gpu'
+ torch.set_float32_matmul_precision("high")
else:
- accelerator = 'cpu'
+ accelerator = "cpu"
# CONSTANTS
-#storage_base = 'sqlite:///../../models_torch/optuna/'
-storage_base = 'sqlite:///../../../../data-project/KIWaVo/models/optuna/'
+# storage_base = 'sqlite:///../../models_torch/optuna/'
+storage_base = "sqlite:///../../../../data-project/KIWaVo/models/optuna/"
if ut.debugger_is_active():
- #default_storage_name = 'sqlite:///../models_torch/optuna/debug_01.db'
- default_storage_name = 'sqlite:///../../../data-project/KIWaVo/models/optuna/debug_icaart.db'
+ # default_storage_name = 'sqlite:///../models_torch/optuna/debug_01.db'
+ default_storage_name = (
+ "sqlite:///../../../data-project/KIWaVo/models/optuna/debug_icaart.db"
+ )
n_trials = 2
max_epochs = 2
default_max_trials = 1
else:
- #default_storage_name = 'sqlite:///../../models_torch/optuna/optimization_01.db'
- default_storage_name = 'sqlite:///../../../../data-project/KIWaVo/models/optuna/icaart_01.db'
+ # default_storage_name = 'sqlite:///../../models_torch/optuna/optimization_01.db'
+ default_storage_name = (
+ "sqlite:///../../../../data-project/KIWaVo/models/optuna/icaart_01.db"
+ )
n_trials = 100
max_epochs = 2000
default_max_trials = None
@@ -50,14 +55,14 @@ if use_cuda:
max_free = 0
best_device = None
for j in range(torch.cuda.device_count()):
- free , _ = torch.cuda.mem_get_info(j)
+ free, _ = torch.cuda.mem_get_info(j)
if free > max_free:
max_free = free
best_device = j
devices = [best_device]
else:
- devices = 'auto'
+ devices = "auto"
class Objective:
@@ -76,7 +81,21 @@ class Objective:
val (float, optional): share of validation data (directly after train). Defaults to .15.
monitor (str, optional): metric to monitor. Defaults to 'hp/val_loss'.
"""
- def __init__(self, filename, gauge, logdir, in_size=144, out_size=48, batch_size=2048, percentile=0.95, train=0.7, val=.15,monitor= 'hp/val_loss', **kwargs):
+
+ def __init__(
+ self,
+ filename,
+ gauge,
+ logdir,
+ in_size=144,
+ out_size=48,
+ batch_size=2048,
+ percentile=0.95,
+ train=0.7,
+ val=0.15,
+ monitor="hp/val_loss",
+ **kwargs,
+ ):
# Hold these implementation specific arguments as the fields of the class.
self.filename = filename
self.level_name_org = gauge
@@ -90,31 +109,44 @@ class Objective:
self.monitor = monitor
def _get_callbacks(self, trial):
- checkpoint_callback = ModelCheckpoint(save_top_k=1, monitor=self.monitor,save_weights_only=True)
+ checkpoint_callback = ModelCheckpoint(
+ save_top_k=1, monitor=self.monitor, save_weights_only=True
+ )
pruning_callback = OptunaPruningCallback(trial, monitor=self.monitor)
- early_stop_callback = EarlyStopping(monitor=self.monitor, mode="min", patience=3)
+ early_stop_callback = EarlyStopping(
+ monitor=self.monitor, mode="min", patience=3
+ )
my_callback = WaVoCallback()
- return my_callback, [checkpoint_callback,pruning_callback, early_stop_callback, my_callback]
+ return my_callback, [
+ checkpoint_callback,
+ pruning_callback,
+ early_stop_callback,
+ my_callback,
+ ]
- def _get_model_params(self,trial):
+ def _get_model_params(self, trial):
differencing = 0
learning_rate = 0.001
- model_architecture = 'last_lstm'
+ model_architecture = "last_lstm"
model_params = dict(
- hidden_size_lstm = trial.suggest_categorical("hidden_size_lstm", [64, 128, 256]),
- num_layers_lstm = trial.suggest_categorical("n_layers_lstm", [1,2,3]),
- hidden_size = trial.suggest_categorical("hidden_size", [128, 256]),
- num_layers = trial.suggest_categorical("n_layers", [2,3]),
- dropout = 0.25
- )
+ hidden_size_lstm=trial.suggest_categorical(
+ "hidden_size_lstm", [64, 128, 256]
+ ),
+ num_layers_lstm=trial.suggest_categorical("n_layers_lstm", [1, 2, 3]),
+ hidden_size=trial.suggest_categorical("hidden_size", [128, 256]),
+ num_layers=trial.suggest_categorical("n_layers", [2, 3]),
+ dropout=0.25,
+ )
return model_architecture, differencing, learning_rate, model_params
def __call__(self, trial):
- model_architecture, differencing, learning_rate, model_params = self._get_model_params(trial)
- #prepare data
+ model_architecture, differencing, learning_rate, model_params = (
+ self._get_model_params(trial)
+ )
+ # prepare data
data_module = WaVoDataModule(
str(self.filename),
self.level_name_org,
@@ -126,10 +158,14 @@ class Objective:
train=self.train,
val=self.val,
model_architecture=model_architecture,
- time_enc=model_params['embed_type'] if model_architecture in ["transformer","autoformer"] else 'fixed',
+ time_enc=(
+ model_params["embed_type"]
+ if model_architecture in ["transformer", "autoformer"]
+ else "fixed"
+ ),
)
- #prepare model
+ # prepare model
model = WaVoLightningModule(
model_architecture=model_architecture,
feature_count=data_module.feature_count,
@@ -138,49 +174,81 @@ class Objective:
scaler=data_module.scaler,
target_idx=data_module.target_idx,
learning_rate=learning_rate,
- **model_params)
+ **model_params,
+ )
logging.info("Params: %s", trial.params)
-
my_callback, callbacks = self._get_callbacks(trial)
logger = TensorBoardLogger(self.log_dir, default_hp_metric=False)
-
- trainer = pl.Trainer(default_root_dir=self.log_dir,
- logger=logger,
- accelerator=accelerator,
- devices=devices,
- callbacks=callbacks,
- max_epochs=max_epochs,
- log_every_n_steps=3,
- )
+ trainer = pl.Trainer(
+ default_root_dir=self.log_dir,
+ logger=logger,
+ accelerator=accelerator,
+ devices=devices,
+ callbacks=callbacks,
+ max_epochs=max_epochs,
+ log_every_n_steps=3,
+ )
trainer.fit(model, data_module)
- #save metrics to optuna
+ # save metrics to optuna
model_path = str(Path(trainer.log_dir).resolve())
logging.info("model_path: %s", model_path)
trial.set_user_attr("model_path", model_path)
- for metric in ['hp/val_nse', 'hp/val_mae', 'hp/val_mae_flood']:
+ for metric in ["hp/val_nse", "hp/val_mae", "hp/val_mae_flood"]:
for i in [23, 47]:
- trial.set_user_attr(f'{metric}_{i}', my_callback.metrics[metric][i].item())
+ trial.set_user_attr(
+ f"{metric}_{i}", my_callback.metrics[metric][i].item()
+ )
return my_callback.metrics[self.monitor].item()
def parse_args() -> argparse.Namespace:
- """ Parse all the arguments and provides some help in the command line
- """
-
- parser: argparse.ArgumentParser = argparse.ArgumentParser(description='Execute experiments for exp_icaart.')
- parser.add_argument('filename', metavar='datafile', type=Path,help='The path to your input data.')
- parser.add_argument('gauge', metavar='gaugename',type=str, help='The name of the gauge column.')
- parser.add_argument('logdir', type=Path,help='set a directory for logs and model checkpoints.')
- parser.add_argument('trials', metavar='trials',type=int, default=n_trials,help='How many trials to run.')
+ """Parse all the arguments and provides some help in the command line"""
- parser.add_argument('--expname', metavar='experiment_name',type=str, default='nameless',help='The name of the experiment.')
- parser.add_argument('--storagename', metavar='storage_name',type=str, default=None,help='The database for the experiment.')
- parser.add_argument('--architecture', metavar='architecture',type=str, default=None,help='Hyperparameter to decide an architecture.')
+ parser: argparse.ArgumentParser = argparse.ArgumentParser(
+ description="Execute experiments for exp_icaart."
+ )
+ parser.add_argument(
+ "filename", metavar="datafile", type=Path, help="The path to your input data."
+ )
+ parser.add_argument(
+ "gauge", metavar="gaugename", type=str, help="The name of the gauge column."
+ )
+ parser.add_argument(
+ "logdir", type=Path, help="set a directory for logs and model checkpoints."
+ )
+ parser.add_argument(
+ "trials",
+ metavar="trials",
+ type=int,
+ default=n_trials,
+ help="How many trials to run.",
+ )
+ parser.add_argument(
+ "--expname",
+ metavar="experiment_name",
+ type=str,
+ default="nameless",
+ help="The name of the experiment.",
+ )
+ parser.add_argument(
+ "--storagename",
+ metavar="storage_name",
+ type=str,
+ default=None,
+ help="The database for the experiment.",
+ )
+ parser.add_argument(
+ "--architecture",
+ metavar="architecture",
+ type=str,
+ default=None,
+ help="Hyperparameter to decide an architecture.",
+ )
parser.add_argument(
"--pruning",
@@ -192,49 +260,62 @@ def parse_args() -> argparse.Namespace:
return parser.parse_args()
+
def main():
- """Start a hyperparameter optimization with optuna and torchlightning.
- """
+ """Start a hyperparameter optimization with optuna and torchlightning."""
parsed_args = parse_args()
if not parsed_args.logdir.exists():
parsed_args.logdir.mkdir(parents=True)
-
if parsed_args.pruning:
pruner = optuna.pruners.HyperbandPruner(
- min_resource=1, max_resource='auto', reduction_factor=3, bootstrap_count=0)
+ min_resource=1, max_resource="auto", reduction_factor=3, bootstrap_count=0
+ )
else:
pruner = optuna.pruners.NopPruner()
-
-
- study_name = f"{parsed_args.filename.stem} {parsed_args.expname}" # Unique identifier of the study.
- storage_name = default_storage_name if parsed_args.storagename is None else f"{storage_base}{parsed_args.storagename}.db"
+ study_name = f"{parsed_args.filename.stem} {parsed_args.expname}" # Unique identifier of the study.
+ storage_name = (
+ default_storage_name
+ if parsed_args.storagename is None
+ else f"{storage_base}{parsed_args.storagename}.db"
+ )
# Logging, add stream handler of stdout to show the messages
logging.basicConfig(level=logging.INFO)
- logFormatter = logging.Formatter('%(asctime)s;%(levelname)s;%(message)s',datefmt='%Y-%m-%d %H:%M:%S')
- fileHandler = logging.FileHandler(parsed_args.logdir / 'optuna.log',)
+ logFormatter = logging.Formatter(
+ "%(asctime)s;%(levelname)s;%(message)s", datefmt="%Y-%m-%d %H:%M:%S"
+ )
+ fileHandler = logging.FileHandler(
+ parsed_args.logdir / "optuna.log",
+ )
consoleHandler = logging.StreamHandler(sys.stdout)
fileHandler.setFormatter(logFormatter)
consoleHandler.setFormatter(logFormatter)
logging.getLogger().addHandler(fileHandler)
- #logging.getLogger().addHandler(consoleHandler)
+ # logging.getLogger().addHandler(consoleHandler)
optuna.logging.get_logger("optuna").addHandler(fileHandler)
optuna.logging.get_logger("optuna").addHandler(consoleHandler)
- logging.info('Start of this execution======================================================================')
- logging.info("Executing %s with device %s and parameters %s ",sys.argv[0],devices,sys.argv[1:])
+ logging.info(
+ "Start of this execution======================================================================"
+ )
+ logging.info(
+ "Executing %s with device %s and parameters %s ",
+ sys.argv[0],
+ devices,
+ sys.argv[1:],
+ )
search_space = {
- "hidden_size_lstm":[64, 128, 256],
- "n_layers_lstm": [1,2],
- "hidden_size": [128, 256],
- "n_layers": [2,3],
+ "hidden_size_lstm": [64, 128, 256],
+ "n_layers_lstm": [1, 2],
+ "hidden_size": [128, 256],
+ "n_layers": [2, 3],
}
trials = parsed_args.trials
- space_sizes = [len(v) for k,v in search_space.items()]
+ space_sizes = [len(v) for k, v in search_space.items()]
for el in space_sizes:
trials *= el
@@ -244,16 +325,24 @@ def main():
storage=storage_name,
direction="minimize",
pruner=pruner,
- load_if_exists=True)
+ load_if_exists=True,
+ )
study.set_metric_names(["hp/val_loss"])
- objective = Objective(**vars(parsed_args),batch_size=4096, gc_after_trial=True)#TODO REMOVE batch_size
+ objective = Objective(
+ **vars(parsed_args), batch_size=4096, gc_after_trial=True
+ ) # TODO REMOVE batch_size
- #if parsed_args.architecture is not None:
+ # if parsed_args.architecture is not None:
# for _ in range(parsed_args.trials):
# study.enqueue_trial({"model_architecture": parsed_args.architecture})
- study.optimize(objective, timeout=None,callbacks=[lambda study, trial: torch.cuda.empty_cache()])
+ study.optimize(
+ objective,
+ timeout=None,
+ callbacks=[lambda study, trial: torch.cuda.empty_cache()],
+ )
+
if __name__ == "__main__":
main()
diff --git a/src/old/icaart_ensemble_hyper.py b/src/old/icaart_ensemble_hyper.py
new file mode 100644
index 0000000000000000000000000000000000000000..ad140aa5a7f9523b369aef1db7758d79dc0e6874
--- /dev/null
+++ b/src/old/icaart_ensemble_hyper.py
@@ -0,0 +1,309 @@
+import argparse
+import logging
+import pickle
+import sys
+from pathlib import Path
+
+import lightning.pytorch as pl
+import optuna
+import torch
+import yaml
+from lightning.pytorch.callbacks import EarlyStopping, ModelCheckpoint
+from lightning.pytorch.loggers import TensorBoardLogger
+
+import utils.helpers as hp
+import utils.utility as ut
+from utils.callbacks import OptunaPruningCallback, WaVoCallback
+from data_tools.data_module import WaVoDataModule
+from models.ensemble_models import WaVoLightningEnsemble, WaVoLightningAttentionEnsemble
+
+use_cuda = torch.cuda.is_available()
+if use_cuda:
+ accelerator = "cuda"
+ torch.set_float32_matmul_precision("high")
+
+ max_free = 0
+ best_device = None
+ for j in range(torch.cuda.device_count()):
+ free, _ = torch.cuda.mem_get_info(j)
+ if free > max_free:
+ max_free = free
+ best_device = j
+ devices = [best_device]
+
+else:
+ accelerator = "cpu"
+ devices = "auto"
+
+storage_base = "sqlite:///../../../../data-project/KIWaVo/models/optuna/"
+
+if ut.debugger_is_active():
+ max_epochs = 2
+ default_storage_name = "sqlite:///../../../data-project/KIWaVo/models/optuna/icaart_ensemble_debug_01.db"
+
+else:
+ max_epochs = 2000
+ default_storage_name = (
+ "sqlite:///../../../../data-project/KIWaVo/models/optuna/icaart_ensemble_01.db"
+ )
+
+
+class Objective:
+ """
+ This class defines the objective function for hyperparameter tuning using Optuna library.
+
+ Args:
+ filename (str|Path): Path to .csv file, first column should be a timeindex
+ model_dir (str|Path): Path to the directory containing the base models
+ log_dir (str|Path): Path to the logging directory
+ select_strat (str): How to select the base models (random or first n)
+ model_count (int): How many base models to use
+ monitor (str, optional): metric to monitor. Defaults to 'hp/val_loss'.
+ """
+
+ def __init__(
+ self,
+ filename,
+ modeldir,
+ logdir,
+ select_strat,
+ model_count,
+ monitor="hp/val_loss",
+ **kwargs,
+ ):
+ # Hold these implementation specific arguments as the fields of the class.
+ self.filename = filename
+ self.model_dir = modeldir
+ self.log_dir = logdir
+ self.select_strat = select_strat
+ self.model_count = model_count
+ self.monitor = monitor
+
+ def _get_callbacks(self, trial):
+ checkpoint_callback = ModelCheckpoint(
+ save_top_k=1, monitor=self.monitor, save_weights_only=True
+ )
+ pruning_callback = OptunaPruningCallback(trial, monitor=self.monitor)
+ early_stop_callback = EarlyStopping(
+ monitor=self.monitor, mode="min", patience=3
+ )
+ my_callback = WaVoCallback(ensemble=True)
+
+ return my_callback, [
+ checkpoint_callback,
+ pruning_callback,
+ early_stop_callback,
+ my_callback,
+ ]
+
+ def _get_model_params(self, trial):
+ model_params = dict(
+ hidden_size=trial.suggest_int("hidden_size", 32, 512),
+ num_layers=trial.suggest_int("n_layers", 2, 4),
+ dropout=0.25,
+ learning_rate=trial.suggest_float("lr", 0.00001, 0.01),
+ norm_func=trial.suggest_categorical("norm_func", ["softmax", "minmax"]),
+ )
+
+ return model_params
+
+ def __call__(self, trial):
+
+ model_params = self._get_model_params(trial)
+ # log_dir = '../../../data-project/KIWaVo/models/ensemble_debug/'
+ # model_dir = Path('../../../data-project/KIWaVo/models/icaarts_hollingstedt/lightning_logs/')
+ model_list = []
+ model_path_list = []
+ yaml_data = None
+
+ if self.select_strat == "random":
+ all_models = [x.name.split("_")[1] for x in self.model_dir.iterdir()]
+
+ elif self.select_strat == "first":
+ model_choice = list(range(self.model_count))
+
+ # for s in [0,1,2]:
+ for s in model_choice:
+ temp_dir = self.model_dir / f"version_{s}/"
+
+ model_list.append(
+ hp.load_model_cuda(temp_dir, use_cuda=use_cuda, devices=devices)
+ )
+ model_path_list.append(str(temp_dir.resolve()))
+
+ if yaml_data is None:
+ yaml_data = hp.load_settings_model(temp_dir)
+ # with open(temp_dir / 'hparams.yaml', 'r') as file:
+ # yaml_data = yaml.load(file, Loader=yaml.FullLoader)
+ # yaml_data['scaler'] = pickle.loads(yaml_data['scaler'])
+
+ config = {
+ "scaler": yaml_data[
+ "scaler"
+ ], # TODO test how this works without giving scaler etc. outside of jupyterlab
+ #'filename' : yaml_data['filename'],
+ "filename": str(self.filename),
+ "level_name_org": yaml_data["level_name_org"],
+ "out_size": yaml_data["out_size"],
+ "threshold": yaml_data["threshold"],
+ "feature_count": yaml_data["feature_count"],
+ "differencing": yaml_data["differencing"],
+ "model_architecture": "ensemble",
+ }
+
+ print("ACHTUNG GGF. FALSCHES MODELL")
+ # ensemble_model = WaVoLightningEnsemble(model_list,model_path_list,**model_params)#TODO 'ÄNDERN!
+ ensemble_model = WaVoLightningAttentionEnsemble(
+ model_list, model_path_list, **model_params
+ )
+
+ config["in_size"] = ensemble_model.max_in_size
+
+ data_module = WaVoDataModule(**config)
+
+ logging.info("Params: %s", trial.params)
+
+ my_callback, callbacks = self._get_callbacks(trial)
+ logger = TensorBoardLogger(self.log_dir, default_hp_metric=False)
+
+ trainer = pl.Trainer(
+ default_root_dir=self.log_dir,
+ gradient_clip_val=0.5,
+ logger=logger,
+ accelerator=accelerator,
+ devices=devices,
+ callbacks=callbacks,
+ max_epochs=max_epochs,
+ log_every_n_steps=10,
+ )
+ trainer.fit(ensemble_model, data_module)
+
+ # save metrics to optuna
+ model_path = str(Path(trainer.log_dir).resolve())
+ logging.info("model_path: %s", model_path)
+ trial.set_user_attr("model_path", model_path)
+ for metric in ["hp/val_nse", "hp/val_mae", "hp/val_mae_flood"]:
+ for i in [23, 47]:
+ trial.set_user_attr(
+ f"{metric}_{i}", my_callback.metrics[metric][i].item()
+ )
+
+ return my_callback.metrics[self.monitor].item()
+
+
+def parse_args() -> argparse.Namespace:
+ """Parse all the arguments and provides some help in the command line"""
+
+ parser: argparse.ArgumentParser = argparse.ArgumentParser(
+ description="Execute experiments for exp_icaart."
+ )
+ parser.add_argument(
+ "filename", metavar="datafile", type=Path, help="The path to your input data."
+ )
+ parser.add_argument(
+ "modeldir", metavar="modeldir", type=Path, help="The path to your base models."
+ )
+ parser.add_argument(
+ "logdir", type=Path, help="set a directory for logs and model checkpoints."
+ )
+ parser.add_argument(
+ "trials",
+ metavar="trials",
+ type=int,
+ default=100,
+ help="How many trials to run.",
+ )
+ parser.add_argument("select_strat", choices=["random", "first"])
+ parser.add_argument(
+ "model_count",
+ metavar="mc",
+ type=int,
+ default=5,
+ help="How many base models to use.",
+ )
+
+ parser.add_argument(
+ "--expname",
+ metavar="experiment_name",
+ type=str,
+ default="nameless",
+ help="The name of the experiment.",
+ )
+ parser.add_argument(
+ "--storagename",
+ metavar="storage_name",
+ type=str,
+ default=None,
+ help="The database for the experiment.",
+ )
+
+ return parser.parse_args()
+
+
+def main():
+ parsed_args = parse_args()
+ if not parsed_args.logdir.exists():
+ parsed_args.logdir.mkdir(parents=True)
+
+ if False:
+ pruner = optuna.pruners.HyperbandPruner(
+ min_resource=1, max_resource="auto", reduction_factor=3, bootstrap_count=0
+ )
+ else:
+ pruner = optuna.pruners.NopPruner()
+
+ study_name = f"{parsed_args.filename.stem} {parsed_args.expname}" # Unique identifier of the study.
+ storage_name = (
+ default_storage_name
+ if parsed_args.storagename is None
+ else f"{storage_base}{parsed_args.storagename}.db"
+ )
+
+ # Logging, add stream handler of stdout to show the messages
+ logging.basicConfig(level=logging.INFO)
+ logFormatter = logging.Formatter(
+ "%(asctime)s;%(levelname)s;%(message)s", datefmt="%Y-%m-%d %H:%M:%S"
+ )
+ fileHandler = logging.FileHandler(
+ parsed_args.logdir / "ensemble_hyper.log",
+ )
+ consoleHandler = logging.StreamHandler(sys.stdout)
+ fileHandler.setFormatter(logFormatter)
+ consoleHandler.setFormatter(logFormatter)
+ logging.getLogger().addHandler(fileHandler)
+ # logging.getLogger().addHandler(consoleHandler)
+ optuna.logging.get_logger("optuna").addHandler(fileHandler)
+ optuna.logging.get_logger("optuna").addHandler(consoleHandler)
+
+ logging.info(
+ "Start of this execution======================================================================"
+ )
+ logging.info(
+ "Executing %s with device %s and parameters %s ",
+ sys.argv[0],
+ devices,
+ sys.argv[1:],
+ )
+
+ study = optuna.create_study(
+ study_name=study_name,
+ storage=storage_name,
+ direction="minimize",
+ pruner=pruner,
+ load_if_exists=True,
+ )
+
+ study.set_metric_names(["hp/val_loss"])
+ objective = Objective(**vars(parsed_args), gc_after_trial=True)
+
+ study.optimize(
+ objective,
+ n_trials=parsed_args.trials,
+ timeout=None,
+ callbacks=[lambda study, trial: torch.cuda.empty_cache()],
+ )
+
+
+if __name__ == "__main__":
+ # TODO command line arguments
+ main()
diff --git a/src/old/icaart_random_base.py b/src/old/icaart_random_base.py
new file mode 100644
index 0000000000000000000000000000000000000000..be973b7023e841ee917f9e5b209861fb3dab189f
--- /dev/null
+++ b/src/old/icaart_random_base.py
@@ -0,0 +1,296 @@
+import argparse
+import logging
+import pickle
+import sys
+from pathlib import Path
+import random
+import lightning.pytorch as pl
+import optuna
+import torch
+import yaml
+from lightning.pytorch.callbacks import EarlyStopping, ModelCheckpoint
+from lightning.pytorch.loggers import TensorBoardLogger
+
+import utils.helpers as hp
+import utils.utility as ut
+from utils.callbacks import OptunaPruningCallback, WaVoCallback
+from data_tools.data_module import WaVoDataModule
+from models.ensemble_models import WaVoLightningEnsemble
+
+use_cuda = torch.cuda.is_available()
+if use_cuda:
+ accelerator = "cuda"
+ torch.set_float32_matmul_precision("high")
+
+ max_free = 0
+ best_device = None
+ for j in range(torch.cuda.device_count()):
+ free, _ = torch.cuda.mem_get_info(j)
+ if free > max_free:
+ max_free = free
+ best_device = j
+ devices = [best_device]
+
+else:
+ accelerator = "cpu"
+ devices = "auto"
+
+storage_base = "sqlite:///../../../../data-project/KIWaVo/models/optuna/"
+
+if ut.debugger_is_active():
+ max_epochs = 2
+ default_storage_name = "sqlite:///../../../data-project/KIWaVo/models/optuna/icaart_ensemble_debug_01.db"
+
+else:
+ max_epochs = 100
+ default_storage_name = (
+ "sqlite:///../../../../data-project/KIWaVo/models/optuna/icaart_ensemble_01.db"
+ )
+
+
+class Objective:
+ """
+ This class defines the objective function for hyperparameter tuning using Optuna library.
+
+ Args:
+ filename (str|Path): Path to .csv file, first column should be a timeindex
+ model_dir (str|Path): Path to the directory containing the base models
+ log_dir (str|Path): Path to the logging directory
+ model_count (int): How many base models to use
+ monitor (str, optional): metric to monitor. Defaults to 'hp/val_loss'.
+ """
+
+ def __init__(
+ self, filename, modeldir, logdir, model_count, monitor="hp/val_loss", **kwargs
+ ):
+ # Hold these implementation specific arguments as the fields of the class.
+ self.filename = filename
+ self.model_dir = modeldir
+ self.log_dir = logdir
+ self.model_count = model_count
+ self.monitor = monitor
+ self.all_models = [int(x.name.split("_")[1]) for x in self.model_dir.iterdir()]
+
+ def _get_callbacks(self, trial):
+ checkpoint_callback = ModelCheckpoint(
+ save_top_k=1, monitor=self.monitor, save_weights_only=True
+ )
+ pruning_callback = OptunaPruningCallback(trial, monitor=self.monitor)
+ early_stop_callback = EarlyStopping(
+ monitor=self.monitor, mode="min", patience=3
+ )
+ my_callback = WaVoCallback(ensemble=True)
+
+ return my_callback, [
+ checkpoint_callback,
+ pruning_callback,
+ early_stop_callback,
+ my_callback,
+ ]
+
+ def _get_model_params(self, trial):
+ # This is just so that optuna will actually execute the number of trials we want, it stops when it testes all parameter combinations
+ place_holder = trial.suggest_int("place_holder", 0, 10000)
+
+ model_params = dict(
+ hidden_size=trial.suggest_int("hidden_size", 512, 512),
+ num_layers=trial.suggest_int("n_layers", 2, 2),
+ dropout=0.25,
+ learning_rate=trial.suggest_float("lr", 0.002, 0.002),
+ norm_func=trial.suggest_categorical("norm_func", ["minmax"]),
+ )
+
+ return model_params
+
+ def __call__(self, trial):
+
+ model_params = self._get_model_params(trial)
+ # log_dir = '../../../data-project/KIWaVo/models/ensemble_debug/'
+ # model_dir = Path('../../../data-project/KIWaVo/models/icaarts_hollingstedt/lightning_logs/')
+ model_list = []
+ model_path_list = []
+ yaml_data = None
+
+ model_choice = random.sample(self.all_models, self.model_count)
+
+ # for s in [0,1,2]:
+ for s in model_choice:
+ temp_dir = self.model_dir / f"version_{s}/"
+
+ model_list.append(
+ hp.load_model_cuda(temp_dir, use_cuda=use_cuda, devices=devices)
+ )
+ model_path_list.append(str(temp_dir.resolve()))
+
+ if yaml_data is None:
+ yaml_data = hp.load_settings_model(temp_dir)
+ # with open(temp_dir / 'hparams.yaml', 'r') as file:
+ # yaml_data = yaml.load(file, Loader=yaml.FullLoader)
+ # yaml_data['scaler'] = pickle.loads(yaml_data['scaler'])
+
+ config = {
+ "scaler": yaml_data[
+ "scaler"
+ ], # TODO test how this works without giving scaler etc. outside of jupyterlab
+ #'filename' : yaml_data['filename'],
+ "filename": str(self.filename),
+ "level_name_org": yaml_data["level_name_org"],
+ "out_size": yaml_data["out_size"],
+ "threshold": yaml_data["threshold"],
+ "feature_count": yaml_data["feature_count"],
+ "differencing": yaml_data["differencing"],
+ "model_architecture": "ensemble",
+ }
+
+ ensemble_model = WaVoLightningEnsemble(
+ model_list, model_path_list, **model_params
+ )
+ config["in_size"] = ensemble_model.max_in_size
+
+ data_module = WaVoDataModule(**config)
+
+ logging.info("Params: %s", trial.params)
+
+ my_callback, callbacks = self._get_callbacks(trial)
+ logger = TensorBoardLogger(self.log_dir, default_hp_metric=False)
+
+ trainer = pl.Trainer(
+ default_root_dir=self.log_dir,
+ gradient_clip_val=0.5,
+ logger=logger,
+ accelerator=accelerator,
+ devices=devices,
+ callbacks=callbacks,
+ max_epochs=max_epochs,
+ log_every_n_steps=10,
+ )
+ trainer.fit(ensemble_model, data_module)
+
+ # save metrics to optuna
+ model_path = str(Path(trainer.log_dir).resolve())
+ logging.info("model_path: %s", model_path)
+ trial.set_user_attr("model_path", model_path)
+ for metric in ["hp/val_nse", "hp/val_mae", "hp/val_mae_flood"]:
+ for i in [23, 47]:
+ trial.set_user_attr(
+ f"{metric}_{i}", my_callback.metrics[metric][i].item()
+ )
+
+ return my_callback.metrics[self.monitor].item()
+
+
+def parse_args() -> argparse.Namespace:
+ """Parse all the arguments and provides some help in the command line"""
+
+ parser: argparse.ArgumentParser = argparse.ArgumentParser(
+ description="Execute experiments for exp_icaart."
+ )
+ parser.add_argument(
+ "filename", metavar="datafile", type=Path, help="The path to your input data."
+ )
+ parser.add_argument(
+ "modeldir", metavar="modeldir", type=Path, help="The path to your base models."
+ )
+ parser.add_argument(
+ "logdir", type=Path, help="set a directory for logs and model checkpoints."
+ )
+ parser.add_argument(
+ "trials",
+ metavar="trials",
+ type=int,
+ default=100,
+ help="How many trials to run.",
+ )
+ parser.add_argument(
+ "model_count",
+ metavar="mc",
+ type=int,
+ default=5,
+ help="How many base models to use.",
+ )
+
+ parser.add_argument(
+ "--expname",
+ metavar="experiment_name",
+ type=str,
+ default="nameless",
+ help="The name of the experiment.",
+ )
+ parser.add_argument(
+ "--storagename",
+ metavar="storage_name",
+ type=str,
+ default=None,
+ help="The database for the experiment.",
+ )
+
+ return parser.parse_args()
+
+
+def main():
+ parsed_args = parse_args()
+ if not parsed_args.logdir.exists():
+ parsed_args.logdir.mkdir(parents=True)
+
+ if False:
+ pruner = optuna.pruners.HyperbandPruner(
+ min_resource=1, max_resource="auto", reduction_factor=3, bootstrap_count=0
+ )
+ else:
+ pruner = optuna.pruners.NopPruner()
+
+ study_name = f"{parsed_args.filename.stem} {parsed_args.expname}" # Unique identifier of the study.
+ storage_name = (
+ default_storage_name
+ if parsed_args.storagename is None
+ else f"{storage_base}{parsed_args.storagename}.db"
+ )
+
+ # Logging, add stream handler of stdout to show the messages
+ logging.basicConfig(level=logging.INFO)
+ logFormatter = logging.Formatter(
+ "%(asctime)s;%(levelname)s;%(message)s", datefmt="%Y-%m-%d %H:%M:%S"
+ )
+ fileHandler = logging.FileHandler(
+ parsed_args.logdir / "ensemble_hyper.log",
+ )
+ consoleHandler = logging.StreamHandler(sys.stdout)
+ fileHandler.setFormatter(logFormatter)
+ consoleHandler.setFormatter(logFormatter)
+ logging.getLogger().addHandler(fileHandler)
+ # logging.getLogger().addHandler(consoleHandler)
+ optuna.logging.get_logger("optuna").addHandler(fileHandler)
+ optuna.logging.get_logger("optuna").addHandler(consoleHandler)
+
+ logging.info(
+ "Start of this execution======================================================================"
+ )
+ logging.info(
+ "Executing %s with device %s and parameters %s ",
+ sys.argv[0],
+ devices,
+ sys.argv[1:],
+ )
+
+ study = optuna.create_study(
+ study_name=study_name,
+ storage=storage_name,
+ direction="minimize",
+ pruner=pruner,
+ load_if_exists=True,
+ )
+
+ study.set_metric_names(["hp/val_loss"])
+ objective = Objective(**vars(parsed_args), gc_after_trial=True)
+
+ study.optimize(
+ objective,
+ n_trials=parsed_args.trials,
+ timeout=None,
+ callbacks=[lambda study, trial: torch.cuda.empty_cache()],
+ )
+
+
+if __name__ == "__main__":
+ # TODO command line arguments
+ main()
diff --git a/src/old/icaart_sizes.py b/src/old/icaart_sizes.py
new file mode 100644
index 0000000000000000000000000000000000000000..57a4e5342511f3cc16fd09e68e6ff2c553ff6df8
--- /dev/null
+++ b/src/old/icaart_sizes.py
@@ -0,0 +1,292 @@
+import argparse
+import logging
+import pickle
+import sys
+from pathlib import Path
+import random
+import lightning.pytorch as pl
+import optuna
+import torch
+import yaml
+from lightning.pytorch.callbacks import EarlyStopping, ModelCheckpoint
+from lightning.pytorch.loggers import TensorBoardLogger
+
+import utils.helpers as hp
+import utils.utility as ut
+from utils.callbacks import OptunaPruningCallback, WaVoCallback
+from data_tools.data_module import WaVoDataModule
+from models.ensemble_models import WaVoLightningEnsemble
+
+use_cuda = torch.cuda.is_available()
+if use_cuda:
+ accelerator = "cuda"
+ torch.set_float32_matmul_precision("high")
+
+ max_free = 0
+ best_device = None
+ for j in range(torch.cuda.device_count()):
+ free, _ = torch.cuda.mem_get_info(j)
+ if free > max_free:
+ max_free = free
+ best_device = j
+ devices = [best_device]
+
+else:
+ accelerator = "cpu"
+ devices = "auto"
+
+storage_base = "sqlite:///../../../../data-project/KIWaVo/models/optuna/"
+
+if ut.debugger_is_active():
+ max_epochs = 2
+ default_storage_name = "sqlite:///../../../data-project/KIWaVo/models/optuna/icaart_ensemble_debug_01.db"
+
+else:
+ max_epochs = 100
+ default_storage_name = (
+ "sqlite:///../../../../data-project/KIWaVo/models/optuna/icaart_ensemble_01.db"
+ )
+
+
+class Objective:
+ """
+ This class defines the objective function for hyperparameter tuning using Optuna library.
+
+ Args:
+ filename (str|Path): Path to .csv file, first column should be a timeindex
+ model_dir (str|Path): Path to the directory containing the base models
+ log_dir (str|Path): Path to the logging directory
+ monitor (str, optional): metric to monitor. Defaults to 'hp/val_loss'.
+ """
+
+ def __init__(self, filename, modeldir, logdir, monitor="hp/val_loss", **kwargs):
+ # Hold these implementation specific arguments as the fields of the class.
+ self.filename = filename
+ self.model_dir = modeldir
+ self.log_dir = logdir
+ self.monitor = monitor
+ self.all_models = [int(x.name.split("_")[1]) for x in self.model_dir.iterdir()]
+
+ def _get_callbacks(self, trial):
+ checkpoint_callback = ModelCheckpoint(
+ save_top_k=1, monitor=self.monitor, save_weights_only=True
+ )
+ pruning_callback = OptunaPruningCallback(trial, monitor=self.monitor)
+ early_stop_callback = EarlyStopping(
+ monitor=self.monitor, mode="min", patience=3
+ )
+ my_callback = WaVoCallback(ensemble=True)
+
+ return my_callback, [
+ checkpoint_callback,
+ pruning_callback,
+ early_stop_callback,
+ my_callback,
+ ]
+
+ def _get_model_params(self, trial):
+ # This is just so that optuna will actually execute the number of trials we want, it stops when it testes all parameter combinations
+ ensemble_size = trial.suggest_categorical("ensemble_size", [5, 10, 15, 20])
+ # temperature = trial.suggest_categorical("temperature",[1/10,1,10])
+
+ model_params = dict(
+ hidden_size=trial.suggest_int("hidden_size", 512, 512),
+ num_layers=trial.suggest_int("n_layers", 2, 2),
+ dropout=0.25,
+ learning_rate=trial.suggest_float("lr", 0.002, 0.002),
+ norm_func=trial.suggest_categorical("norm_func", ["softmax"]),
+ )
+
+ return model_params, ensemble_size
+
+ def __call__(self, trial):
+
+ model_params, ensemble_size = self._get_model_params(trial)
+ # log_dir = '../../../data-project/KIWaVo/models/ensemble_debug/'
+ # model_dir = Path('../../../data-project/KIWaVo/models/icaarts_hollingstedt/lightning_logs/')
+ model_list = []
+ model_path_list = []
+ yaml_data = None
+
+ model_choice = random.sample(self.all_models, ensemble_size)
+
+ # for s in [0,1,2]:
+ for s in model_choice:
+ temp_dir = self.model_dir / f"version_{s}/"
+
+ model_list.append(
+ hp.load_model_cuda(temp_dir, use_cuda=use_cuda, devices=devices)
+ )
+ model_path_list.append(str(temp_dir.resolve()))
+
+ if yaml_data is None:
+ yaml_data = hp.load_settings_model(temp_dir)
+ # with open(temp_dir / 'hparams.yaml', 'r') as file:
+ # yaml_data = yaml.load(file, Loader=yaml.FullLoader)
+ # yaml_data['scaler'] = pickle.loads(yaml_data['scaler'])
+
+ config = {
+ "scaler": yaml_data[
+ "scaler"
+ ], # TODO test how this works without giving scaler etc. outside of jupyterlab
+ #'filename' : yaml_data['filename'],
+ "filename": str(self.filename),
+ "level_name_org": yaml_data["level_name_org"],
+ "out_size": yaml_data["out_size"],
+ "threshold": yaml_data["threshold"],
+ "feature_count": yaml_data["feature_count"],
+ "differencing": yaml_data["differencing"],
+ "model_architecture": "ensemble",
+ }
+
+ ensemble_model = WaVoLightningEnsemble(
+ model_list, model_path_list, **model_params
+ )
+ config["in_size"] = ensemble_model.max_in_size
+
+ data_module = WaVoDataModule(**config)
+
+ logging.info("Params: %s", trial.params)
+
+ my_callback, callbacks = self._get_callbacks(trial)
+ logger = TensorBoardLogger(self.log_dir, default_hp_metric=False)
+
+ trainer = pl.Trainer(
+ default_root_dir=self.log_dir,
+ gradient_clip_val=0.5,
+ logger=logger,
+ accelerator=accelerator,
+ devices=devices,
+ callbacks=callbacks,
+ max_epochs=max_epochs,
+ log_every_n_steps=10,
+ )
+ trainer.fit(ensemble_model, data_module)
+
+ # save metrics to optuna
+ model_path = str(Path(trainer.log_dir).resolve())
+ logging.info("model_path: %s", model_path)
+ trial.set_user_attr("model_path", model_path)
+ for metric in ["hp/val_nse", "hp/val_mae", "hp/val_mae_flood"]:
+ for i in [23, 47]:
+ trial.set_user_attr(
+ f"{metric}_{i}", my_callback.metrics[metric][i].item()
+ )
+
+ return my_callback.metrics[self.monitor].item()
+
+
+def parse_args() -> argparse.Namespace:
+ """Parse all the arguments and provides some help in the command line"""
+
+ parser: argparse.ArgumentParser = argparse.ArgumentParser(
+ description="Execute experiments for exp_icaart."
+ )
+ parser.add_argument(
+ "filename", metavar="datafile", type=Path, help="The path to your input data."
+ )
+ parser.add_argument(
+ "modeldir", metavar="modeldir", type=Path, help="The path to your base models."
+ )
+ parser.add_argument(
+ "logdir", type=Path, help="set a directory for logs and model checkpoints."
+ )
+ parser.add_argument(
+ "trials",
+ metavar="trials",
+ type=int,
+ default=100,
+ help="How many trials to run.",
+ )
+ # parser.add_argument('model_count', metavar='mc',type=int, default=5,help='How many base models to use.')
+
+ parser.add_argument(
+ "--expname",
+ metavar="experiment_name",
+ type=str,
+ default="nameless",
+ help="The name of the experiment.",
+ )
+ parser.add_argument(
+ "--storagename",
+ metavar="storage_name",
+ type=str,
+ default=None,
+ help="The database for the experiment.",
+ )
+
+ return parser.parse_args()
+
+
+def main():
+ parsed_args = parse_args()
+ if not parsed_args.logdir.exists():
+ parsed_args.logdir.mkdir(parents=True)
+
+ if False:
+ pruner = optuna.pruners.HyperbandPruner(
+ min_resource=1, max_resource="auto", reduction_factor=3, bootstrap_count=0
+ )
+ else:
+ pruner = optuna.pruners.NopPruner()
+
+ study_name = f"{parsed_args.filename.stem} {parsed_args.expname}" # Unique identifier of the study.
+ storage_name = (
+ default_storage_name
+ if parsed_args.storagename is None
+ else f"{storage_base}{parsed_args.storagename}.db"
+ )
+
+ # Logging, add stream handler of stdout to show the messages
+ logging.basicConfig(level=logging.INFO)
+ logFormatter = logging.Formatter(
+ "%(asctime)s;%(levelname)s;%(message)s", datefmt="%Y-%m-%d %H:%M:%S"
+ )
+ fileHandler = logging.FileHandler(
+ parsed_args.logdir / "ensemble_hyper.log",
+ )
+ consoleHandler = logging.StreamHandler(sys.stdout)
+ fileHandler.setFormatter(logFormatter)
+ consoleHandler.setFormatter(logFormatter)
+ logging.getLogger().addHandler(fileHandler)
+ # logging.getLogger().addHandler(consoleHandler)
+ optuna.logging.get_logger("optuna").addHandler(fileHandler)
+ optuna.logging.get_logger("optuna").addHandler(consoleHandler)
+
+ logging.info(
+ "Start of this execution======================================================================"
+ )
+ logging.info(
+ "Executing %s with device %s and parameters %s ",
+ sys.argv[0],
+ devices,
+ sys.argv[1:],
+ )
+
+ ens_sizes = [5, 10, 15, 20]
+ search_space = {"ensemble_size": ens_sizes * parsed_args.trials}
+ sampler = optuna.samplers.GridSampler(search_space)
+
+ study = optuna.create_study(
+ sampler=sampler,
+ study_name=study_name,
+ storage=storage_name,
+ direction="minimize",
+ pruner=pruner,
+ load_if_exists=True,
+ )
+
+ study.set_metric_names(["hp/val_loss"])
+ objective = Objective(**vars(parsed_args), gc_after_trial=True)
+
+ study.optimize(
+ objective,
+ n_trials=len(ens_sizes) * parsed_args.trials,
+ timeout=None,
+ callbacks=[lambda study, trial: torch.cuda.empty_cache()],
+ )
+
+
+if __name__ == "__main__":
+ # TODO command line arguments
+ main()
diff --git a/src/predict_database.py b/src/predict_database.py
index 0c29c2d20cc88bc4d61bb7d0edd672d1293108aa..ab30b028f4508aad912fbcf66d00f12d25ff2bdb 100644
--- a/src/predict_database.py
+++ b/src/predict_database.py
@@ -1,6 +1,7 @@
"""
Skript to forecast water levels using a trained torch model and writing the results to an Oracle Database.
"""
+
import argparse
import logging
import sys
@@ -27,41 +28,52 @@ def get_configs(passed_args: argparse.Namespace) -> Tuple[dict, List[dict]]:
yaml_path = passed_args.yaml_path
gauge_configs = []
with open(yaml_path, "r", encoding="utf-8") as file:
- docs = yaml.safe_load_all(file)
- for doc in docs:
- if doc["document"] == "main":
- main_config = doc
- if isinstance(main_config["zrxp_folder"],str):
- main_config["zrxp_folder"] = [main_config["zrxp_folder"]]
- main_config["zrxp_folder"] = list(map(Path,main_config["zrxp_folder"]))
- main_config["sensor_folder"] = Path(main_config["sensor_folder"])
- main_config["start"] = pd.to_datetime(main_config["start"])
- if "end" in main_config:
- main_config["end"] = pd.to_datetime(main_config["end"])
- else:
- doc["model_folder"] = list(map(Path, doc["model_folder"]))
- gauge_configs.append(doc)
-
+ configs = yaml.safe_load(file)
+
+ main_config = configs["main_config"]
+ if isinstance(main_config["zrxp_folder"], str):
+ main_config["zrxp_folder"] = [main_config["zrxp_folder"]]
+ main_config["zrxp_folder"] = list(map(Path, main_config["zrxp_folder"]))
+ main_config["sensor_folder"] = Path(main_config["sensor_folder"])
+ main_config["start"] = pd.to_datetime(main_config["start"])
+ if "end" in main_config:
+ main_config["end"] = pd.to_datetime(main_config["end"])
if passed_args.start is not None:
main_config["start"] = passed_args.start
- main_config.pop("end",None)
+ main_config.pop("end", None)
main_config["range"] = False
+ gauge_configs = configs["gauge_configs"]
+ for gauge_config in gauge_configs:
+ gauge_config["model_folder"] = Path(gauge_config["model_folder"])
+ assert all([isinstance(c,str) for c in gauge_config["columns"]])
+ assert isinstance(gauge_config["gauge"],str)
+ #assert all([isinstance(k,str) and isinstance(v,str) for k,v in gauge_config["external_fcst"].items()])
+ assert all([isinstance(c,str) for c in gauge_config["external_fcst"]])
+
return main_config, gauge_configs
def _prepare_logging_1() -> None:
logging.basicConfig(level=logging.INFO)
- logFormatter = logging.Formatter(
- "%(asctime)s;%(levelname)s;%(message)s", datefmt="%Y-%m-%d %H:%M:%S"
- )
- consoleHandler = logging.StreamHandler(sys.stdout)
- consoleHandler.setFormatter(logFormatter)
- logging.getLogger().addHandler(consoleHandler)
logging.info("Executing %s with parameters %s ", sys.argv[0], sys.argv[1:])
+
def _prepare_logging_2(con) -> None:
+
+ old_factory = logging.getLogRecordFactory()
+ def record_factory(*args, **kwargs):
+ record = old_factory(*args, **kwargs)
+ record.gauge_id = kwargs.pop("gauge_id", None)
+ return record
+ logging.setLogRecordFactory(record_factory)
+
+ logFormatter = logging.Formatter(
+ "%(asctime)s;%(levelname)s;%(message)s", datefmt="%Y-%m-%d %H:%M:%S"
+ )
dbHandler = OracleDBHandler(con)
+ dbHandler.setFormatter(logFormatter)
+
dbHandler.setLevel(logging.INFO)
logging.getLogger().addHandler(dbHandler)
logging.info("Executing %s with parameters %s ", sys.argv[0], sys.argv[1:])
@@ -85,11 +97,9 @@ def parse_args() -> argparse.Namespace:
type=pd.to_datetime,
help="The start date for the prediction. Format: YYYY-MM-DD HH:MM, overwrites start/end/range in the config file.",
)
- parser.add_argument("--zrxp", action="store_true", help="Save predictions as ZRXP files")
-
- # parser.add_argument('-load_sensor',action='store_true',help="Force the")
- # load_sensor
- # load_zrxp
+ parser.add_argument(
+ "--zrxp", action="store_true", help="Save predictions as ZRXP files (not yet implemented)"
+ ) #TODO: Implement ZRXP
return parser.parse_args()
@@ -100,23 +110,28 @@ def main(passed_args) -> None:
_prepare_logging_1()
if "config_dir" in main_config["db_params"]:
- logging.info("Initiating Thick mode with executable %s", main_config["db_params"]["config_dir"])
+ logging.info(
+ "Initiating Thick mode with executable %s",
+ main_config["db_params"]["config_dir"],
+ )
oracledb.init_oracle_client(lib_dir=main_config["db_params"]["config_dir"])
else:
logging.info("Initiating Thin mode")
con = oracledb.connect(**main_config["db_params"])
_prepare_logging_2(con)
- #This is logged twice into Stout, because if thinks break it would not be logged at all.
+ # This is logged twice into Stout, because if thinks break it would not be logged at all.
if "config_dir" in main_config["db_params"]:
- logging.info("Initiating Thick mode with executable %s", main_config["db_params"]["config_dir"])
+ logging.info(
+ "Initiating Thick mode with executable %s",
+ main_config["db_params"]["config_dir"],
+ )
else:
logging.info("Initiating Thin mode")
connector = OracleWaVoConnection(con, main_config)
- #connector.maybe_create_tables() # Create tables if they don't exist
+ # connector.maybe_create_tables() # Create tables if they don't exist
connector.maybe_update_tables() # Potentially load new data
-
for gauge_config in gauge_configs:
connector.handle_gauge(gauge_config)
diff --git a/src/utils/bsh_extractor_v2.py b/src/scripts/bsh_extractor_v2.py
similarity index 100%
rename from src/utils/bsh_extractor_v2.py
rename to src/scripts/bsh_extractor_v2.py
diff --git a/src/scripts/icon_export_v3.py b/src/scripts/icon_export_v3.py
new file mode 100755
index 0000000000000000000000000000000000000000..f7d765420c58bba913f4dfd76d475236cdfb6c3a
--- /dev/null
+++ b/src/scripts/icon_export_v3.py
@@ -0,0 +1,548 @@
+"""
+This module provides tools to extract zrxp files from grib2 ensemble predictions.
+"""
+
+import argparse
+import bz2
+import logging
+import logging.handlers
+import re
+import subprocess as sp
+import time
+from pathlib import Path
+from typing import List, Tuple
+
+import numpy as np
+import pandas as pd
+import pygrib
+from shapely.geometry import Point
+from shapely.geometry.polygon import Polygon
+from sympy import elliptic_f
+import yaml
+from io import BytesIO
+
+logger = logging.getLogger("icon_export")
+logger.setLevel(logging.DEBUG)
+
+
+def setup_logging(log_file: Path, debug: bool, max_bytes: int = 50000) -> None:
+ """Set up rotating log files
+
+ Args:
+ log_file (Path): Where to log to
+ debug (bool): Wheather to log debug level messages
+ max_bytes (int, optional): maximum filesize for the logfile. Defaults to 50000.
+ """
+ log_file.parent.mkdir(parents=True, exist_ok=True)
+ formatter = logging.Formatter(
+ fmt="%(asctime)s;%(levelname)s;%(message)s", datefmt="%Y-%m-%d %H:%M:%S"
+ )
+
+ logHandlerFile = logging.handlers.RotatingFileHandler(
+ log_file, maxBytes=max_bytes, backupCount=1
+ )
+ logHandlerFile.setLevel(logging.DEBUG if debug else logging.INFO)
+ logHandlerFile.setFormatter(formatter)
+ logger.addHandler(logHandlerFile)
+
+ logHandlerStream = logging.StreamHandler()
+ logHandlerStream.setLevel(logging.DEBUG if debug else logging.INFO)
+ logHandlerStream.setFormatter(formatter)
+ logger.addHandler(logHandlerStream)
+
+
+def parse_args() -> argparse.Namespace:
+ """Parse all the arguments and provides some help in the command line"""
+ parser: argparse.ArgumentParser = argparse.ArgumentParser(
+ description="Extracts predictions from ICON Ensembles into zrx files"
+ )
+
+ parser.add_argument(
+ "yaml_path",
+ metavar="yaml_path",
+ type=Path,
+ help="""The path to your config file.
+ Each files is expected to contain:
+ shape_folder: The path to the directory with your areafiles or to a single area file.
+ index_folder: The path to the directory where you want to store your index files.
+ source_folder: The path to the directory with your .grib or .bz2 files. All files should have the same format. e.g. the same coordinates.
+ target_folder: The path to the directory for your zrx files.
+ mode: 'ensemble', 'hourly' or 'single' ICON forecast
+ Hourly means a folder with one .grib2 file per hour with 15 minute messages, not currently in use
+ force_index_calculation: Force generation of new index files, Defaults fo False
+ log_file, optional: Where to log information.
+ tidy_up: Weather to delete the input files after execution
+ debug: Weather to increase the log level to debug.
+
+ Example:
+ shape_folder: ../data/meta_other/Gebiete/
+ index_folder: ../data/meta_other/ensemble_index/
+ source_folder: ../data/ICON_ENSEMBLE/
+ target_folder: ../data/ICON_ENSEMBLE_EXPORT/
+ mode: ensemble
+ force_index_calculation: False
+ log_file: ../data/ICON_ENSEMBLE_EXPORT/icon_export.log
+ tidy_up: True
+ debug: False
+ """,
+ )
+ return parser.parse_args()
+
+
+def get_config(passed_args: argparse.Namespace) -> dict:
+ """
+ Retrieves the configurations from a YAML file.
+
+ Args:
+ passed_args (argparse.Namespace): The command-line arguments.
+
+ Returns: dict: The configuration
+
+ """
+
+ with open(passed_args.yaml_path, "r", encoding="utf-8") as file:
+ config = yaml.safe_load(file)
+ config["shape_folder"] = Path(config["shape_folder"])
+ config["index_folder"] = Path(config["index_folder"])
+ config["source_folder"] = Path(config["source_folder"])
+ config["target_folder"] = Path(config["target_folder"])
+ if "log_file" in config:
+ config["log_file"] = Path(config["log_file"])
+
+
+ assert config["shape_folder"].exists(), f"shape_folder {config['shape_folder']} does not exist"
+ assert config["source_folder"].exists(), f"source_folder {config['source_folder']} does not exist"
+ assert config["source_folder"].is_dir(), f"source_folder {config['source_folder']} is not a directory"
+
+ return config
+
+
+def get_area_name(area_path: Path) -> str:
+ """Generate an area name from the given filenname.
+ Returns the part of the filename after the first _ and the next .
+ For example: '../ICON/areas/ezg_willenscharen.txt' -> willenscharen
+ This is used for the name of the zrx file
+ Args:
+ area_path (Path): Path to a shape file (Tab seperated values).
+
+ Returns:
+ str: name
+ """
+ return area_path.stem.rsplit("_", maxsplit=1)[-1]
+
+
+
+
+def get_mask_from_grid(lats: np.array, lons: np.array, area_file: Path) -> np.array:
+ """Returns a mask that is False everywhere, but at the indexes in the grid that
+ correspond to points in the polygon made from the area_file.
+
+ Args:
+ lats (np.array): Latitudes of a grib file.
+ lons (np.array): Longitudes of a grib file.
+ area_file (Path): Path to one shape file (Tab seperated values).
+
+ Returns:
+ np.array: _description_
+ """
+ logger.info("Generating mask from %s", area_file)
+ try:
+ df_area = pd.read_csv(area_file, index_col=0, sep=",", engine="python")
+ assert all(
+ pd.api.types.is_numeric_dtype(df_area[col]) for col in ["Lat", "Long"]
+ )
+ except (KeyError, pd.errors.ParserError):
+ try:
+ df_area = pd.read_csv(
+ area_file, index_col=0, sep=None, engine="python", decimal=","
+ )
+ assert all(
+ pd.api.types.is_numeric_dtype(df_area[col]) for col in ["Lat", "Long"]
+ )
+ except (KeyError, pd.errors.ParserError):
+ df_area = pd.read_csv(area_file, index_col=0, sep=",")
+ assert all(
+ pd.api.types.is_numeric_dtype(df_area[col]) for col in ["Lat", "Long"]
+ )
+
+ polygon = Polygon(df_area.values)
+
+ min_index_lat = np.argmax(lats > df_area["Lat"].min(), axis=0)[0]
+ max_index_lat = np.argmax(lats > df_area["Lat"].max(), axis=0)[0]
+ min_index_lon = np.argmax(lons > df_area["Long"].min(), axis=1)[0]
+ max_index_lon = np.argmax(lons > df_area["Long"].max(), axis=1)[0]
+
+ mask = np.full(lats.shape, False)
+ for i in range(min_index_lat, max_index_lat + 1):
+ for j in range(min_index_lon, max_index_lon + 1):
+ mask[i, j] = polygon.contains(Point(lats[i][j], lons[i][j]))
+
+ return mask
+
+def dict_to_zrx(
+ forecast_dict: dict,
+ target_folder: Path,
+ source_file: Path,
+ name: str,
+ member: int = 0,
+) -> None:
+ """Takes a dictionary containing a forecast and saves it into an zrx file.
+
+ Args:
+ forecast_dict (dict): forecast
+ target_folder (Path): Where to save the zrx file
+ source_file (Path): Input grib2 file
+ name (str): Name of the current area
+ member (int, optional): Ensemble number. Defaults to 0.
+ """
+ df = pd.DataFrame(forecast_dict).astype({"forecast": "datetime64[ns]"})
+ # df['value'] = df['value']-df['value'].shift(1).fillna(df['value'][0])
+ df["value"] = df["value"] - df["value"].shift(1).fillna(0)
+ df["value"] = df["value"].apply(lambda x: max(0, x))
+ df = df.round({"value": 3})
+
+ target_file = target_folder / f"{re.findall('20[2-9][0-9][0-1][0-9][0-3][0-9][0-2][0-9]',source_file.name)[0]}_{name}_{member}.zrx"
+ if target_file.exists():
+ logger.warning("File %s already exists, file will not be saved.", target_file)
+ else:
+ with open(target_file, "w", encoding="utf-8") as file:
+ file.write("#REXCHANGEWISKI." + name + ".N.KNN|*|\n")
+ file.write("#RINVAL-777|*|\n")
+ # if mode in ['hourly','single']:
+ if member == 0:
+ file.write("#LAYOUT(timestamp,value)|*|\n")
+ df = df[["forecast", "value"]]
+ else:
+ file.write("#LAYOUT(timestamp,forecast, member,value)|*|\n")
+
+ # df[['timestamp','value']]
+ df.to_csv(
+ path_or_buf=target_file,
+ header=False,
+ index=False,
+ mode="a",
+ sep=" ",
+ date_format="%Y%m%d%H%M",
+ )
+
+
+class ICON_Exporter:
+ """
+ Class to extract zrx files from ICON (ensemble) predictions.
+ Args:
+ source_folder (Path): directory containing grib2 and/or bz2 files.
+ target_folder (Path): where to save the zrx files
+ shape_folder (Path): directory or file with shape files (Tab seperated values).
+ index_folder (Path): where to save the index files
+ gen_indexes (bool): generate (new) index files.
+ mode (str): 'ensemble', 'hourly' or 'single' ICON forecast
+ Hourly means a folder with one .grib2 file per hour with 15 minute messages, not currently in use
+
+
+ """
+ def __init__(
+ self,
+ source_folder,
+ target_folder,
+ shape_folder,
+ index_folder,
+ gen_indexes,
+ mode,
+ tidy_up,
+ ):
+ self.source_folder = source_folder
+ self.target_folder = target_folder
+ self.shape_folder = shape_folder
+ self.index_folder = index_folder
+ self.gen_indexes = gen_indexes
+ self.mode = mode
+ self.tidy_up=tidy_up
+ self.indexes = None
+ self.names = None
+
+
+ if self.mode not in ["hourly", "single", "ensemble"]:
+ raise ValueError(
+ f"The value {mode} is not a valid mode. Must be hourly, single or ensemble."
+ )
+
+ def setup(self):
+ self.index_folder.mkdir(parents=True, exist_ok=True)
+ self.target_folder.mkdir(parents=True, exist_ok=True)
+
+ def tear_down(self):
+ """Deletes old .grib2 and .bz2 files
+ """
+ if self.tidy_up:
+ for data_file in filter(lambda x: x.suffix in [".grib2", ".bz2"], self.source_folder.iterdir()):
+ logger.info("Deleting file %s", data_file)
+ data_file.unlink()
+
+
+ def unpack_bz(self) -> None:
+ """Unpacks all files ending with .bz2 in a given folder or subfolder and saves them with the same name, just without the .bz2
+ This should not really be necessary, but for some reason the pygrib library cannot handle the unzipped and loaded data
+ """
+ #TODO Once pygrib 2.1.5 is released this wont be necessary anymore.
+ for source_file in filter(
+ lambda x: x.suffix == ".bz2"
+ and not x.with_suffix(".grib2").exists()
+ and not (x.parent / x.stem).exists(),
+ self.source_folder.iterdir(),
+ ):
+ if source_file.suffixes == [".grib2", ".bz2"]:
+ unpacked_source_file = source_file.parent / source_file.stem
+ else:
+ unpacked_source_file = source_file.with_suffix(".grib2")
+ logger.info("Unpacking file %s", source_file)
+ bz2_file = bz2.BZ2File(source_file)
+ bz2_data = bz2_file.read()
+ with open(unpacked_source_file, "wb") as file:
+ file.write(bz2_data)
+
+ def set_indexes(self):
+ if self.names is None:
+ if (self.gen_indexes
+ or len(list(filter(lambda x: x.suffix == ".index", self.index_folder.iterdir()))) == 0
+ ):
+
+ #Select the smallest file
+ try:
+ grib_file = sorted(
+ filter(lambda x: x.suffix == ".grib2", self.source_folder.iterdir()),
+ key=lambda x: x.stat().st_size,
+ )[0]
+ except IndexError:
+ raise FileNotFoundError("No .grib2 files found in source folder")
+ lats, lons = pygrib.open(str(grib_file))[1].latlons() # pylint: disable=no-member
+ logger.info(
+ "Generating index files from %s, grid shape is %s",
+ grib_file.name,
+ lats.shape,
+ )
+
+ # because not all ICON files use the same grid, but we have no way of knowing which one until we load a message.
+ shape_files = list(filter(lambda x: x.suffix in [".txt", ".csv"], self.shape_folder.iterdir()))
+ index_files = [self.index_folder / shape_file.with_suffix(".index").name for shape_file in shape_files]
+ self.indexes = list(map(lambda x: np.where(get_mask_from_grid(lats, lons, x).flatten())[0]+ 1,shape_files))
+ self.names = list(map(get_area_name, shape_files))
+
+ for name, index, index_file in zip(self.names, self.indexes, index_files):
+ if index_file.exists():
+ logger.info("File %s already exists, replacing it.", index_file)
+ pd.DataFrame(index, columns=[name]).to_csv(index_file)
+
+ else:
+ index_files = list(filter(lambda x: x.suffix == ".index", self.index_folder.iterdir()))
+ logger.info("Loading index files %s", index_files)
+ self.indexes = [pd.read_csv(index_file, index_col=0).values.flatten() for index_file in index_files]
+ self.names = list(map(get_area_name, index_files))
+
+
+ def process(self) -> None:
+ if self.mode == "ensemble":
+ self.process_ensembles()
+ elif self.mode == "single":
+ self.process_singles()
+ elif self.mode == "hourly":
+ self.process_hourly()
+
+ def process_singles(self) -> None:
+ self.set_indexes()
+
+ for source_file in filter(lambda x: x.suffix == ".grib2", self.source_folder.iterdir()):
+ logger.info("Processing single forecast for file %s", source_file)
+
+ gribfile = pygrib.open(str(source_file)) # pylint: disable=no-member
+ dict_dict = {}
+ first = True
+
+ for message in gribfile.select(shortName=["tp"]):
+ start_time = pd.to_datetime(str(message.dataDate) + str(message.dataTime), format="%Y%m%d%H")
+ for index, name in zip(self.indexes, self.names):
+ if first:
+ dict_dict[name] = {
+ "timestamp": 48 * [start_time],
+ "forecast": pd.period_range(
+ start=start_time + pd.offsets.Hour(1), periods=48, freq="h"
+ ),
+ "member": 48 * [0],
+ "value": [],
+ }
+ else:
+ if isinstance(message.values, np.ma.core.MaskedArray):
+ dict_dict[name]["value"].append(
+ message.values.data.flatten()[index - 1].mean()
+ )
+ else:
+ dict_dict[name]["value"].append(
+ message.values.flatten()[index - 1].mean()
+ )
+ first = False
+
+ for name, cur_dict in dict_dict.items():
+ dict_to_zrx(cur_dict, self.target_folder, source_file, name, member=0)
+
+
+
+ def process_ensembles(self) -> None:
+ """Helper function that splits the input ensemble file based on ensemble number, gets names and indexes,
+ and then calls extract_ensemble on all .grib2 files
+ """
+
+ # Iterate over all potential input files
+ tmp_folder = self.source_folder / "tmp"
+ tmp_folder.mkdir(parents=True, exist_ok=True)
+ for source_file in filter(lambda x: x.suffix == ".grib2", self.source_folder.iterdir()):
+ # split file
+ output_name = tmp_folder / source_file.name.replace(
+ ".grib2", "_[perturbationNumber].grib2"
+ ) # The pertubationNumber is not a Python thing, it's for the grib_copy command
+ logger.info("grip_copy %s into %s", source_file, output_name)
+ #You may need to install the eccodes library: https://confluence.ecmwf.int/display/ECC/ecCodes+installation#ecCodesinstallation-Quickinstallationguide
+ #Or just sudo apt install libeccodes-tools
+ sp.run(["grib_copy", source_file, output_name], capture_output=True, check=True)
+ self.set_indexes()
+ source_base = str(tmp_folder / (source_file.stem + "_"))
+ self.extract_ensemble(source_base)
+ for temp_file in tmp_folder.iterdir():
+ temp_file.unlink()
+ tmp_folder.rmdir()
+
+ def extract_ensemble(self,source_base: str) -> None:
+ """Extracts the predicted mean rainfall per hour for each ensemble member and area/catchment and exports them into zrx files.
+
+ Args:
+ source_base (str): the name base of the grib2 and zrx files
+ """
+ logger.info("Processing ensemble forecast for files %s", source_base + "*.grib2")
+ for i in range(1, 21):
+ logger.info("Ensemble member %s/20", i)
+ source_file = Path(source_base + str(i) + ".grib2")
+ gribfile_small = pygrib.open(str(source_file)) # pylint: disable=no-member
+ # doesn't work with 00:00: start_time = pd.to_datetime(str(gribfile_small[1].dataDate) +str(gribfile_small[1].dataTime),format="%Y%m%d%H%M")
+ start_time = pd.to_datetime(
+ f"{gribfile_small[1].dataDate} {gribfile_small[1].dataTime:04}",
+ format="%Y%m%d %H%M",
+ )
+ dict_dict = {}
+ for name in self.names:
+ dict_dict[name] = {
+ "timestamp": 48 * [start_time],
+ "forecast": pd.period_range(
+ start=start_time + pd.offsets.Hour(1), periods=48, freq="h"),
+ "member": 48 * [i],
+ "value": [],
+ }
+
+ for message in gribfile_small:
+ for name, index in zip(self.names, self.indexes):
+ dict_dict[name]["value"].append(message.values.flatten()[index - 1].mean())
+
+ for name, cur_dict in dict_dict.items():
+ dict_to_zrx(cur_dict, self.target_folder, source_file, name, member=i)
+
+ #Remove the grib2 file created by grib_copy
+ source_file.unlink()
+
+
+ def process_hourly(self) -> None:
+ """Extracts the predicted mean rainfall per hour for each area/catchment and exports them into zrx files.
+ """
+ print("WARNING: UNTESTED CODE")
+ self.set_indexes()
+
+ logger.info("Processing hourly forecast for folder %s", self.source_folder)
+ # get all current startimes from the file names
+ date_set = {
+ re.findall("20[2-9][0-9][0-1][0-9][0-3][0-9][0-2][0-9]", x.name)[0]
+ for x in filter(lambda x: x.suffix == ".grib2", self.source_folder.iterdir())
+ }
+
+ for icon_run_date in date_set:
+ dict_dict = {}
+ for i, source_file in enumerate(
+ sorted(
+ x
+ for x in self.source_folder.iterdir()
+ if icon_run_date in x.name and x.suffix == ".grib2"
+ )
+ ):
+ assert source_file.name.endswith(
+ f"{icon_run_date}_{i:03d}_2d_tot_prec.grib2"
+ )
+ if i == 0:
+ pass # The first file has messages vor 0, 0-15,0-30, 0-45 minutes, which we don't care about.
+ else:
+ gribfile = pygrib.open(str(source_file)) # pylint: disable=no-member
+ if i == 1:
+ start_time = pd.to_datetime(
+ str(gribfile[1].dataDate) + str(gribfile[1].dataTime),
+ format="%Y%m%d%H%M",
+ )
+ for name, index in zip(self.names, self.indexes):
+ if i == 1:
+ dict_dict[name] = {
+ "timestamp": 48 * [start_time],
+ "forecast": pd.period_range(
+ start=start_time + pd.offsets.Hour(1),
+ periods=48,
+ freq="H",
+ ),
+ "member": 48 * [0],
+ "value": [],
+ }
+
+ dict_dict[name]["value"].append(
+ gribfile[1].values.flatten()[index - 1].mean()
+ )
+
+ logger.debug("Message: %s", gribfile[1])
+ logger.debug(
+ "Grid shape is %s ,does this match the index files?",
+ gribfile[1].values.shape,
+ )
+
+ source_file = [
+ x
+ for x in self.source_folder.iterdir()
+ if icon_run_date in x.name and x.suffix == ".grib2"
+ ][0]
+ for name, cur_dict in dict_dict.items():
+ dict_to_zrx(cur_dict, self.target_folder, source_file, name, member=0)
+
+def main() -> None:
+ """_summary_"""
+
+ start = time.time()
+ parsed_args = parse_args() # configs/icon_exp_ensemble.yaml
+ config = get_config(parsed_args)
+
+ setup_logging(config["log_file"], config["debug"])
+ logger.info(
+ "Start of this execution======================================================================"
+ )
+
+ exporter = ICON_Exporter(
+ config["source_folder"],
+ config["target_folder"],
+ config["shape_folder"],
+ config["index_folder"],
+ config["force_index_calculation"],
+ config["mode"],
+ config["tidy_up"],
+ )
+ exporter.setup()
+ exporter.unpack_bz()
+ exporter.process()
+ exporter.tear_down()
+
+ end = time.time()
+ logger.info("Excecution time %s s", end - start)
+ logger.info(
+ "End of this execution======================================================================"
+ )
+
+
+if __name__ == "__main__":
+ main()
diff --git a/src/predict.py b/src/scripts/predict.py
similarity index 100%
rename from src/predict.py
rename to src/scripts/predict.py
diff --git a/src/rl_ensemble.py b/src/scripts/rl_ensemble.py
similarity index 100%
rename from src/rl_ensemble.py
rename to src/scripts/rl_ensemble.py
diff --git a/src/rl_hyper.py b/src/scripts/rl_hyper.py
similarity index 100%
rename from src/rl_hyper.py
rename to src/scripts/rl_hyper.py
diff --git a/src/callbacks.py b/src/utils/callbacks.py
similarity index 100%
rename from src/callbacks.py
rename to src/utils/callbacks.py
diff --git a/src/utils/db_tools.py b/src/utils/db_tools.py
index 5fe390ae27afc1996822021bc84166ae4c44d3d2..65af2e2533853a74b1b2acd76919407d62b38eea 100644
--- a/src/utils/db_tools.py
+++ b/src/utils/db_tools.py
@@ -17,6 +17,7 @@ from sqlalchemy import create_engine
from sqlalchemy import select, bindparam, between, update, insert
from sqlalchemy.orm import Session
from sqlalchemy.dialects.oracle import FLOAT, TIMESTAMP, VARCHAR2, NUMBER
+from sqlalchemy.exc import IntegrityError
import torch
from torch.utils.data import Dataset, DataLoader
import lightning.pytorch as pl
@@ -31,6 +32,7 @@ from utils.orm_classes import (
PegelForecasts,
Log,
Sensor,
+ InputForecastsMeta
)
# pylint: disable=unsupported-assignment-operation
@@ -47,6 +49,7 @@ class OracleWaVoConnection:
self.members = self.get_member_list()
self.times = self.get_times()
+ self.not_found = []
def handle_gauge(self, gauge_config: dict) -> None:
"""
@@ -60,69 +63,53 @@ class OracleWaVoConnection:
Returns:
None
"""
- # self.cur_config = gauge_config
- for model_folder in gauge_config["model_folder"]:
# self.cur_model_folder = model_folder
- self.handle_model(gauge_config, model_folder)
-
- def handle_model(self, gauge_config, model_folder) -> None:
- """
- Handles the model by loading it, setting it to evaluation mode, and making predictions for each time step and ensemble member.
-
- Args:
- gauge_config (dict): The configuration for the gauge.
- model_folder (Path): The folder containing the model files.
-
- Returns:
- None
- """
- model = hp.load_model(model_folder)
+
+ model = hp.load_model(gauge_config["model_folder"])
model.eval()
- model_name = model_folder.name
created = datetime.now()
for end_time in tqdm(self.times):
-
# load sensor sensor input data
df_base_input = self.load_input_db(
in_size=model.in_size,
end_time=end_time,
- columns=gauge_config["columns"],
- external_fcst=gauge_config["external_fcst"],
+ gauge_config=gauge_config,
+ created=created,
)
if df_base_input is None:
- y = torch.Tensor(48).fill_(np.nan)
- for member in self.members:
- self.insert_forecast(
- y,
- gauge_config["document"],
- model_name,
- end_time,
- member,
- created,
- )
continue
# predict
+ # Get all external forecasts for this gauge
+ #stmt = select(InputForecastsMeta).where(InputForecastsMeta.sensor_name.in_(bindparam("external_fcst")))
+ #params = {"external_fcst" : gauge_config["external_fcst"]}
+ #with Session(self.engine) as session:
+ # x = session.scalars(statement=stmt, params=params).fetchall()
+ # sensor_names = [el.sensor_name for el in x]
+ #params2 = {"ext_forecasts" : sensor_names, "tstamp" : end_time}
+ stmst2 = select(InputForecasts).where(InputForecasts.sensor_name.in_(bindparam("ext_forecasts")),InputForecasts.tstamp == (bindparam("tstamp")))
+ params2 = {"ext_forecasts" : gauge_config["external_fcst"], "tstamp" : end_time}
+ df_temp = pd.read_sql(sql=stmst2,con=self.engine,index_col="tstamp",params=params2)
+
for member in self.members:
self.handle_member(
gauge_config,
model,
- model_name,
df_base_input,
member,
end_time,
created,
)
+
def handle_member(
self,
gauge_config,
model,
- model_name,
df_base_input: pd.DataFrame,
- member,
+ member: int,
end_time,
created: pd.Timestamp = None,
) -> None:
@@ -133,10 +120,10 @@ class OracleWaVoConnection:
Args:
gauge_config (dict): The configuration for the gauge.
model (LightningModule): The model to use for prediction.
- model_name (str): The name of the model (basically name of folder containing the model files).
df_base_input (DataFrame): The base input DataFrame with measure values, still need the actual precipitation forecast.
member (int): The member identifier.
end_time (pd.Timestamp): The timestamp of the forecast.
+ created (pd.Timestamp): The timestamp of the start of the forecast creation.
Returns:
None
@@ -144,31 +131,34 @@ class OracleWaVoConnection:
if member == -1:
df_input = df_base_input.fillna(0)
else:
+
+
# replace fake forecast with external forecast
- df_input = self.merge_synth_fcst(
- df_base_input, member, end_time, gauge_config["external_fcst"]
- )
+ df_input = self.merge_synth_fcst(df_base_input, member, end_time, gauge_config,created)
if df_input is None:
return
y = pred_single_db(model, df_input)
self.insert_forecast(
- y, gauge_config["document"], model_name, end_time, member, created
+ y, gauge_config["gauge"], gauge_config["model_folder"].name, end_time, member, created
)
- def load_input_db(self, in_size, end_time, columns, external_fcst) -> pd.DataFrame:
+ def load_input_db(self, in_size, end_time, gauge_config,created) -> pd.DataFrame:
"""
Loads input data from the database based on the current configuration.
Args:
in_size (int): The input size in hours.
end_time (pd.Timestamp): The end time of the input data.
- columns (List[str]): The columns to load.
- external_fcst (Dict[str,str]): The external forecasts to load/shift.
+ gauge_config (dict): The configuration for the gauge, important here are the columns and external forecasts.
+ created (pd.Timestamp): The timestamp of the start of the forecast creation.
Returns:
pd.DataFrame: The loaded input data as a pandas DataFrame.
"""
+
+ columns=gauge_config["columns"]
+ external_fcst=gauge_config["external_fcst"]
start_time = end_time - pd.Timedelta(in_size - 1, "hours")
stmt = select(SensorData).where(
@@ -177,7 +167,6 @@ class OracleWaVoConnection:
)
df_input = None
try:
-
df_main = pd.read_sql(
sql=stmt,
con=self.engine,
@@ -196,23 +185,18 @@ class OracleWaVoConnection:
df_rest.index = df_rest.index - pd.Timedelta("48h")
if len(df_input) != in_size:
- raise LookupError(
+ raise MissingTimestampsError(
f"Some timestamps are completely missing, found {len(df_input)}/{in_size} hours from {start_time} to {end_time}"
)
for col in df_input.columns:
if df_input[col].isna().sum() > 0:
- logging.warning(
- "Missing %s values in %s, attempting interpolation",
- df_input[col].isna().sum(),
- col,
- )
+ logging.warning("Missing %s values in %s, attempting interpolation",df_input[col].isna().sum(),col,extra={"gauge":gauge_config["gauge"]})
# Interpolation
# Precipitation is filled with 0, other values are interpolated
precip_cols = df_input.columns[df_input.columns.str.contains("Precip")]
df_input.loc[:, precip_cols].fillna(0, inplace=True)
-
df_input = df_input.interpolate(
limit=self.main_config["max_missing"], limit_direction="both"
)
@@ -220,17 +204,12 @@ class OracleWaVoConnection:
if df_input.isna().sum().sum() > 0:
for col in df_input.columns:
if df_input[col].isna().sum() > 0:
- logging.warning(
- "Missing %s values in %s after interpolation",
- df_input[col].isna().sum(),
- col,
- )
- raise LookupError(
+ logging.warning("Missing %s values in %s after interpolation",df_input[col].isna().sum(),col,extra={"gauge":gauge_config["gauge"]})
+ raise MissingValuesError(
f"Missing {df_input.isna().sum().sum()} values after interpolation"
)
# Shift columns that contain external forecasts by 48 hours
-
for col in external_fcst:
df_input[col] = df_input[col].shift(-48)
# Replace the values beyond that cutoff that we actually do know.
@@ -238,14 +217,46 @@ class OracleWaVoConnection:
# This is a little messy because we don't want to interpolate those values and need to be careful about missing timestamps
df_input.loc[df_rest.index, col] = df_rest[col]
- except LookupError as e:
- if df_input is not None:
- logging.error(
- "No data for the chosen timeperiod up to %s and columns in the database.",
- end_time,
- )
- df_input = None
+ except MissingTimestampsError as e:
+ logging.error(e.args[0],extra={"gauge":gauge_config["gauge"]})
+ df_input = None
+ except MissingValuesError as e:
+ logging.error(e.args[0],extra={"gauge":gauge_config["gauge"]})
+ df_input = None
+ except KeyError as e:
+ logging.error(
+ "Some columns are missing in the database, found %s",
+ df_main.columns,extra={"gauge":gauge_config["gauge"]})
logging.error(e.args[0])
+ df_input = None
+ #except LookupError as e:
+ # if df_input is not None:
+ # logging.error(
+ # "No data for the chosen timeperiod up to %s and columns in the database.",
+ # end_time,
+ # extra={"gauge":gauge_config["gauge"]}
+ # )
+ # df_input = None
+ # logging.error(e.args[0])
+
+
+ if df_input is None:
+ logging.error("Input sensordata could not be loaded, inserting empty forecasts",extra={"gauge":gauge_config["gauge"]})
+ y = torch.Tensor(48).fill_(np.nan)
+ for member in self.members:
+ try:
+ self.insert_forecast(
+ y,
+ gauge_config["gauge"],
+ gauge_config["model_folder"].name,
+ end_time,
+ member,
+ created,
+ )
+ except IntegrityError as e:
+ logging.error( "Sensorname %s is not a valid value in PEGEL_FORECASTS. %s",gauge_config["gauge"],e.args[0])
+ break # if one member fails, don't try to insert the rest
+
return df_input
@@ -254,7 +265,8 @@ class OracleWaVoConnection:
df_base_input: pd.DataFrame,
member: int,
end_time: pd.Timestamp,
- external_fcst: dict,
+ gauge_config: dict,
+ created: pd.Timestamp = None,
) -> pd.DataFrame:
"""
Merges external forecasts into the base input dataframe by replacing the old wrong values.
@@ -263,24 +275,29 @@ class OracleWaVoConnection:
df_base_input (pd.DataFrame): The base input dataframe.
member (int): The member identifier.
end_time (pd.Timestamp): The timestamp of the forecast.
- external_fcst (dict): The external forecasts to merge.
+ gauge_config (dict): Settings for the gauge, especially the external forecasts and the gauge name.
+ created (pd.Timestamp): The timestamp of the start of the forecast creation.
Returns:
pd.DataFrame: The merged dataframe with external forecasts.
"""
+ #TODO: besprechen ob wir einfach alle spalten durch vorhersagen ergänzen falls vorhanden?
df_input = df_base_input.copy()
- for col, fcst_name in external_fcst.items():
+
+ for col in gauge_config["external_fcst"]:
try:
- ext_fcst = self.get_ext_forecast(fcst_name, member, end_time)
+ ext_fcst = self.get_ext_forecast(col, member, end_time)
except AttributeError:
logging.error(
"External forecast %s time %s ensemble %s is missing",
- fcst_name,
+ col,
end_time,
member,
+ extra={"gauge":gauge_config["gauge"]}
)
# if some external forecasts are missing, insert NaNs and skip prediction
- # self.insert_forecast(torch.Tensor(48).fill_(np.nan), self.cur_member)
+ y = torch.Tensor(48).fill_(np.nan)
+ self.insert_forecast(y, gauge_config["gauge"], gauge_config["model_folder"].name, end_time, member, created)
return None
else:
# Replace missing values with forecasts.
@@ -292,9 +309,6 @@ class OracleWaVoConnection:
df_input.iloc[nan_indices2, df_input.columns.get_loc(col)] = np.array(
ext_fcst
)[nan_indices]
- # df_input.iloc[nan_indices, df_input.columns.get_loc(col)] = np.array(
- # ext_fcst
- # )[nan_indices]
return df_input
@@ -311,6 +325,9 @@ class OracleWaVoConnection:
continue
vhs_gebiet = zrxp_file.name.split("_")[1] # [6:]
+ if vhs_gebiet in self.not_found:
+ continue
+
df_zrxp = pd.read_csv(
zrxp_file, skiprows=3, header=None, sep=" ", parse_dates=[0]
)
@@ -352,12 +369,12 @@ class OracleWaVoConnection:
member,
)
- def get_ext_forecast(self, vhs_gebiet: str, member: int, end_time) -> List[int]:
+ def get_ext_forecast(self, sensor_name: str, member: int, end_time) -> List[int]:
"""
Retrieves the external forecast data from the database.
Args:
- vhs_gebiet (str): The name of the external forecast.
+ sensor_name (str): The name of the external forecast.
member (int): The member identifier.
end_time (pd.Timestamp): The timestamp of the forecast.
@@ -365,26 +382,34 @@ class OracleWaVoConnection:
List(int): The external forecast data.
"""
# TODO check new column in MODELL_SENSOR
- if vhs_gebiet.startswith("Det_"):
- member = 0
-
- sensor = self.get_sensor_name(vhs_gebiet) # maybe check for none?
- if sensor is None:
- raise AttributeError(f"Sensor {vhs_gebiet} not found in database")
- stmt = select(InputForecasts).where(
+ #TODO this would be more efficient outside the member loop
+ #TODO error handling
+ stmt = select(InputForecastsMeta).where(InputForecastsMeta.sensor_name == (bindparam("sensor_name")))
+ params = {"sensor_name" : sensor_name}
+ with Session(self.engine) as session:
+ input_meta = session.scalar(statement=stmt, params=params)
+
+ if input_meta is None:
+ raise AttributeError(f"Forecast {sensor_name} not found in Table InputForecastsMeta")
+
+ #sensor = self.get_sensor_name(vhs_gebiet) # maybe check for none?
+ #if sensor is None:
+ # raise AttributeError(f"Sensor {vhs_gebiet} not found in database")
+
+ stmt2 = select(InputForecasts).where(
InputForecasts.tstamp == bindparam("tstamp"),
InputForecasts.sensor_name == bindparam("sensor_name"),
InputForecasts.member == bindparam("member"),
)
params = {
"tstamp": end_time,
- "sensor_name": sensor.sensor_name,
+ "sensor_name": input_meta.sensor_name,
"member": member,
}
with Session(self.engine) as session:
- input_forecast = session.scalar(statement=stmt, params=params)
+ input_forecast = session.scalar(statement=stmt2, params=params)
input_forecast = [input_forecast.__dict__[f"h{i}"] for i in range(1, 49)]
return input_forecast
@@ -506,8 +531,10 @@ class OracleWaVoConnection:
**fcst_values,
)
session.add(pegel_forecast)
+ #try:
session.commit()
+
def maybe_update_tables(self) -> None:
"""
Updates the database tables if necessary based on the configuration settings.
@@ -652,6 +679,8 @@ class OracleWaVoConnection:
Returns:
Sensor: A Sensor object (a row from the table Sensor).
"""
+
+
with Session(bind=self.engine) as session:
# get all sensor names for the vhs_gebiet from the model_sensor table and check if there is only one.
stmt = select(ModellSensor.sensor_name).where(
@@ -662,6 +691,8 @@ class OracleWaVoConnection:
)
if len(sensor_names) == 0:
logging.warning("No sensor_name found for %s", vhs_gebiet)
+ self.not_found.append(vhs_gebiet)
+
return
elif len(sensor_names) > 1:
logging.warning("Multiple sensor_names found for %s", vhs_gebiet)
@@ -859,13 +890,20 @@ class OracleDBHandler(logging.Handler):
record.exc_text = ""
log = Log(
- created=pd.to_datetime(record.asctime),
+ created=pd.to_datetime(record.created,unit="s"),
loglevelname=record.levelname,
message=record.message,
module=record.module,
funcname=record.funcName,
lineno=record.lineno,
exception=record.exc_text,
+ gauge=record.gauge_id,
)
self.session.add(log)
self.session.commit()
+
+class MissingTimestampsError(LookupError):
+ pass
+
+class MissingValuesError(LookupError):
+ pass
diff --git a/src/utils/orm_classes.py b/src/utils/orm_classes.py
index 71e0288a7387f655e2091dc1dc63b21ffa1ebd5a..a80a75f680486c6657a02544e5bf27309cef5722 100644
--- a/src/utils/orm_classes.py
+++ b/src/utils/orm_classes.py
@@ -2,27 +2,16 @@
The alternative is to use table reflection from sqlalchemy.
"""
from typing import List, Optional
+import datetime
from sqlalchemy import Double, ForeignKeyConstraint, Index, Integer, PrimaryKeyConstraint, TIMESTAMP, VARCHAR,Identity
from sqlalchemy.dialects.oracle import NUMBER
from sqlalchemy.orm import DeclarativeBase, Mapped, mapped_column, relationship
-import datetime
class Base(DeclarativeBase):
pass
-
-class ExampleTable(Base):
- __tablename__ = 'example_table'
- __table_args__ = (
- PrimaryKeyConstraint('tstamp', 'sensor_name', name='sys_c008698'),
- )
-
- tstamp: Mapped[datetime.datetime] = mapped_column(TIMESTAMP, primary_key=True)
- sensor_name: Mapped[str] = mapped_column(VARCHAR(256), primary_key=True)
- sensor_value: Mapped[Optional[float]] = mapped_column(Double)
-
class Log(Base):
__tablename__ = 'log'
__table_args__ = (
@@ -37,6 +26,19 @@ class Log(Base):
funcname: Mapped[Optional[str]] = mapped_column(VARCHAR(64))
lineno: Mapped[Optional[int]] = mapped_column(Integer)
exception: Mapped[Optional[str]] = mapped_column(VARCHAR(256))
+ gauge: Mapped[Optional[str]] = mapped_column(VARCHAR(256))
+
+
+class InputForecastsMeta(Base):
+ __tablename__ = 'input_forecasts_meta'
+ __table_args__ = (
+ PrimaryKeyConstraint('sensor_name', 'vhs_gebiet', name='UNIQUE_EXT_FORECASTS'),
+ #Index('sensordata_name', 'sensor_name'),
+ #Index('sensordata_ts', 'tstamp')
+ )
+ sensor_name : Mapped[str] = mapped_column(VARCHAR(256), primary_key=True)
+ vhs_gebiet : Mapped[str] = mapped_column(VARCHAR(256), primary_key=True)
+ ensemble_members : Mapped[int] = mapped_column(NUMBER(5, 0, False))
class Modell(Base):