Upload folder using huggingface_hub

args2_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:164e7915f7738f41fd701a5f91668e597fd5a7193a43f4099005a12ff1e1b319
+size 1753692399

checkpoint-7437/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f9ae14f80aa60c742ca67909cdba483196abfc97216a9e85091d68182885ab4f
+size 1753573036

checkpoint-7437/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e963a74430c4cf8a33ec4baa1e644a7d5eee01a73eb810b28b343a04f5d9b74
+size 3507383815

checkpoint-7437/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c507348b7fa630117c89f6b2c0c9132e866e2ed4864941c34905a0061eeb91c1
+size 14645

checkpoint-7437/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:04fece73e52b8f4a1ad1567fd1eb41d6049641d916f1515ed7201c4774fa41a0
+size 1465

checkpoint-7437/trainer_state.json

@@ -0,0 +1,1094 @@
+{
+  "best_global_step": 7437,
+  "best_metric": 0.34891217947006226,
+  "best_model_checkpoint": "/content/outputs/args2/checkpoint-7437",
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 7437,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.02016942315449778,
+      "grad_norm": 13.883692741394043,
+      "learning_rate": 5.487122060470325e-07,
+      "loss": 7.911922607421875,
+      "step": 50
+    },
+    {
+      "epoch": 0.04033884630899556,
+      "grad_norm": 15.190408706665039,
+      "learning_rate": 1.1086226203807392e-06,
+      "loss": 7.467951049804688,
+      "step": 100
+    },
+    {
+      "epoch": 0.060508269463493344,
+      "grad_norm": 17.43606948852539,
+      "learning_rate": 1.6685330347144458e-06,
+      "loss": 5.6657861328125,
+      "step": 150
+    },
+    {
+      "epoch": 0.08067769261799113,
+      "grad_norm": 2.4850821495056152,
+      "learning_rate": 2.2284434490481522e-06,
+      "loss": 3.194233093261719,
+      "step": 200
+    },
+    {
+      "epoch": 0.10084711577248891,
+      "grad_norm": 3.9329416751861572,
+      "learning_rate": 2.7883538633818593e-06,
+      "loss": 2.713049011230469,
+      "step": 250
+    },
+    {
+      "epoch": 0.12101653892698669,
+      "grad_norm": 4.511388301849365,
+      "learning_rate": 3.348264277715566e-06,
+      "loss": 2.4405267333984373,
+      "step": 300
+    },
+    {
+      "epoch": 0.14118596208148446,
+      "grad_norm": 6.108118534088135,
+      "learning_rate": 3.908174692049272e-06,
+      "loss": 2.321563720703125,
+      "step": 350
+    },
+    {
+      "epoch": 0.16135538523598225,
+      "grad_norm": 6.757564544677734,
+      "learning_rate": 4.468085106382979e-06,
+      "loss": 2.0823135375976562,
+      "step": 400
+    },
+    {
+      "epoch": 0.18152480839048002,
+      "grad_norm": 7.603394985198975,
+      "learning_rate": 5.027995520716685e-06,
+      "loss": 1.9735365295410157,
+      "step": 450
+    },
+    {
+      "epoch": 0.20169423154497781,
+      "grad_norm": 8.737318992614746,
+      "learning_rate": 5.587905935050393e-06,
+      "loss": 1.8455184936523437,
+      "step": 500
+    },
+    {
+      "epoch": 0.22186365469947558,
+      "grad_norm": 15.384092330932617,
+      "learning_rate": 6.1478163493840995e-06,
+      "loss": 1.810538330078125,
+      "step": 550
+    },
+    {
+      "epoch": 0.24203307785397338,
+      "grad_norm": 6.883424282073975,
+      "learning_rate": 6.707726763717805e-06,
+      "loss": 1.6601730346679688,
+      "step": 600
+    },
+    {
+      "epoch": 0.26220250100847114,
+      "grad_norm": 5.532959938049316,
+      "learning_rate": 7.267637178051512e-06,
+      "loss": 1.6028485107421875,
+      "step": 650
+    },
+    {
+      "epoch": 0.2823719241629689,
+      "grad_norm": 4.067587852478027,
+      "learning_rate": 7.82754759238522e-06,
+      "loss": 1.5451387023925782,
+      "step": 700
+    },
+    {
+      "epoch": 0.30254134731746674,
+      "grad_norm": 5.565714359283447,
+      "learning_rate": 8.387458006718925e-06,
+      "loss": 1.5030950927734374,
+      "step": 750
+    },
+    {
+      "epoch": 0.3227107704719645,
+      "grad_norm": 5.0564093589782715,
+      "learning_rate": 8.947368421052632e-06,
+      "loss": 1.468430633544922,
+      "step": 800
+    },
+    {
+      "epoch": 0.34288019362646227,
+      "grad_norm": 3.7779645919799805,
+      "learning_rate": 9.507278835386339e-06,
+      "loss": 1.4264836120605469,
+      "step": 850
+    },
+    {
+      "epoch": 0.36304961678096004,
+      "grad_norm": 6.055937767028809,
+      "learning_rate": 9.995708461483441e-06,
+      "loss": 1.369202880859375,
+      "step": 900
+    },
+    {
+      "epoch": 0.38321903993545786,
+      "grad_norm": 7.9874420166015625,
+      "learning_rate": 9.959945640512124e-06,
+      "loss": 1.2812403869628906,
+      "step": 950
+    },
+    {
+      "epoch": 0.40338846308995563,
+      "grad_norm": 3.193695545196533,
+      "learning_rate": 9.924182819540807e-06,
+      "loss": 1.2542407989501954,
+      "step": 1000
+    },
+    {
+      "epoch": 0.4235578862444534,
+      "grad_norm": 4.517177104949951,
+      "learning_rate": 9.888419998569488e-06,
+      "loss": 1.2273830413818358,
+      "step": 1050
+    },
+    {
+      "epoch": 0.44372730939895116,
+      "grad_norm": 6.356571674346924,
+      "learning_rate": 9.85265717759817e-06,
+      "loss": 1.2184021759033203,
+      "step": 1100
+    },
+    {
+      "epoch": 0.463896732553449,
+      "grad_norm": 4.842334270477295,
+      "learning_rate": 9.816894356626852e-06,
+      "loss": 1.1650125122070312,
+      "step": 1150
+    },
+    {
+      "epoch": 0.48406615570794675,
+      "grad_norm": 5.895249366760254,
+      "learning_rate": 9.781131535655533e-06,
+      "loss": 1.1422249603271484,
+      "step": 1200
+    },
+    {
+      "epoch": 0.5042355788624445,
+      "grad_norm": 5.94123649597168,
+      "learning_rate": 9.745368714684214e-06,
+      "loss": 1.1161700439453126,
+      "step": 1250
+    },
+    {
+      "epoch": 0.5244050020169423,
+      "grad_norm": 4.841538429260254,
+      "learning_rate": 9.709605893712897e-06,
+      "loss": 1.0551397705078125,
+      "step": 1300
+    },
+    {
+      "epoch": 0.5445744251714401,
+      "grad_norm": 4.6943230628967285,
+      "learning_rate": 9.673843072741578e-06,
+      "loss": 1.0741574096679687,
+      "step": 1350
+    },
+    {
+      "epoch": 0.5647438483259378,
+      "grad_norm": 5.972919940948486,
+      "learning_rate": 9.638080251770261e-06,
+      "loss": 0.998023910522461,
+      "step": 1400
+    },
+    {
+      "epoch": 0.5849132714804357,
+      "grad_norm": 7.286667346954346,
+      "learning_rate": 9.602317430798942e-06,
+      "loss": 0.9545994567871093,
+      "step": 1450
+    },
+    {
+      "epoch": 0.6050826946349335,
+      "grad_norm": 3.531245231628418,
+      "learning_rate": 9.566554609827625e-06,
+      "loss": 0.9311307525634765,
+      "step": 1500
+    },
+    {
+      "epoch": 0.6252521177894312,
+      "grad_norm": 2.719672679901123,
+      "learning_rate": 9.530791788856306e-06,
+      "loss": 0.9172502136230469,
+      "step": 1550
+    },
+    {
+      "epoch": 0.645421540943929,
+      "grad_norm": 5.1147565841674805,
+      "learning_rate": 9.495028967884987e-06,
+      "loss": 0.9066413879394531,
+      "step": 1600
+    },
+    {
+      "epoch": 0.6655909640984268,
+      "grad_norm": 4.618788719177246,
+      "learning_rate": 9.45926614691367e-06,
+      "loss": 0.8759637451171876,
+      "step": 1650
+    },
+    {
+      "epoch": 0.6857603872529245,
+      "grad_norm": 3.814316987991333,
+      "learning_rate": 9.423503325942351e-06,
+      "loss": 0.8547855377197265,
+      "step": 1700
+    },
+    {
+      "epoch": 0.7059298104074223,
+      "grad_norm": 6.46733283996582,
+      "learning_rate": 9.387740504971032e-06,
+      "loss": 0.8559357452392579,
+      "step": 1750
+    },
+    {
+      "epoch": 0.7260992335619201,
+      "grad_norm": 4.51616096496582,
+      "learning_rate": 9.351977683999715e-06,
+      "loss": 0.85986083984375,
+      "step": 1800
+    },
+    {
+      "epoch": 0.746268656716418,
+      "grad_norm": 5.665010929107666,
+      "learning_rate": 9.316214863028396e-06,
+      "loss": 0.8125870513916016,
+      "step": 1850
+    },
+    {
+      "epoch": 0.7664380798709157,
+      "grad_norm": 4.560214519500732,
+      "learning_rate": 9.280452042057079e-06,
+      "loss": 0.704938735961914,
+      "step": 1900
+    },
+    {
+      "epoch": 0.7866075030254135,
+      "grad_norm": 3.8082242012023926,
+      "learning_rate": 9.24468922108576e-06,
+      "loss": 0.7475352478027344,
+      "step": 1950
+    },
+    {
+      "epoch": 0.8067769261799113,
+      "grad_norm": 9.0048189163208,
+      "learning_rate": 9.208926400114441e-06,
+      "loss": 0.7419461822509765,
+      "step": 2000
+    },
+    {
+      "epoch": 0.826946349334409,
+      "grad_norm": 4.623185157775879,
+      "learning_rate": 9.173163579143124e-06,
+      "loss": 0.7527564239501953,
+      "step": 2050
+    },
+    {
+      "epoch": 0.8471157724889068,
+      "grad_norm": 3.1853349208831787,
+      "learning_rate": 9.137400758171805e-06,
+      "loss": 0.712258071899414,
+      "step": 2100
+    },
+    {
+      "epoch": 0.8672851956434046,
+      "grad_norm": 5.3499579429626465,
+      "learning_rate": 9.101637937200488e-06,
+      "loss": 0.6958197021484375,
+      "step": 2150
+    },
+    {
+      "epoch": 0.8874546187979023,
+      "grad_norm": 4.2968244552612305,
+      "learning_rate": 9.065875116229169e-06,
+      "loss": 0.6866611480712891,
+      "step": 2200
+    },
+    {
+      "epoch": 0.9076240419524002,
+      "grad_norm": 3.536168098449707,
+      "learning_rate": 9.03011229525785e-06,
+      "loss": 0.6310316467285156,
+      "step": 2250
+    },
+    {
+      "epoch": 0.927793465106898,
+      "grad_norm": 4.907815456390381,
+      "learning_rate": 8.994349474286533e-06,
+      "loss": 0.7333563995361329,
+      "step": 2300
+    },
+    {
+      "epoch": 0.9479628882613957,
+      "grad_norm": 7.752031326293945,
+      "learning_rate": 8.958586653315214e-06,
+      "loss": 0.6219098281860351,
+      "step": 2350
+    },
+    {
+      "epoch": 0.9681323114158935,
+      "grad_norm": 5.084356784820557,
+      "learning_rate": 8.922823832343897e-06,
+      "loss": 0.6567278289794922,
+      "step": 2400
+    },
+    {
+      "epoch": 0.9883017345703913,
+      "grad_norm": 5.4790873527526855,
+      "learning_rate": 8.887061011372578e-06,
+      "loss": 0.6141578674316406,
+      "step": 2450
+    },
+    {
+      "epoch": 1.0,
+      "eval_loss": 0.5905303955078125,
+      "eval_runtime": 10.4325,
+      "eval_samples_per_second": 231.872,
+      "eval_steps_per_second": 14.57,
+      "step": 2479
+    },
+    {
+      "epoch": 1.008471157724889,
+      "grad_norm": 3.408093214035034,
+      "learning_rate": 8.851298190401259e-06,
+      "loss": 0.5763191986083984,
+      "step": 2500
+    },
+    {
+      "epoch": 1.0286405808793868,
+      "grad_norm": 3.654597282409668,
+      "learning_rate": 8.815535369429942e-06,
+      "loss": 0.5724074172973633,
+      "step": 2550
+    },
+    {
+      "epoch": 1.0488100040338846,
+      "grad_norm": 3.761699676513672,
+      "learning_rate": 8.779772548458623e-06,
+      "loss": 0.5620191955566406,
+      "step": 2600
+    },
+    {
+      "epoch": 1.0689794271883823,
+      "grad_norm": 2.8509440422058105,
+      "learning_rate": 8.744009727487306e-06,
+      "loss": 0.5438192749023437,
+      "step": 2650
+    },
+    {
+      "epoch": 1.0891488503428801,
+      "grad_norm": 5.627965450286865,
+      "learning_rate": 8.708246906515987e-06,
+      "loss": 0.5647544860839844,
+      "step": 2700
+    },
+    {
+      "epoch": 1.1093182734973779,
+      "grad_norm": 4.898985862731934,
+      "learning_rate": 8.672484085544668e-06,
+      "loss": 0.5223981857299804,
+      "step": 2750
+    },
+    {
+      "epoch": 1.1294876966518759,
+      "grad_norm": 6.7661004066467285,
+      "learning_rate": 8.63672126457335e-06,
+      "loss": 0.5443586730957031,
+      "step": 2800
+    },
+    {
+      "epoch": 1.1496571198063736,
+      "grad_norm": 3.1649415493011475,
+      "learning_rate": 8.600958443602032e-06,
+      "loss": 0.48222358703613283,
+      "step": 2850
+    },
+    {
+      "epoch": 1.1698265429608714,
+      "grad_norm": 8.380264282226562,
+      "learning_rate": 8.565195622630714e-06,
+      "loss": 0.5042714309692383,
+      "step": 2900
+    },
+    {
+      "epoch": 1.1899959661153692,
+      "grad_norm": 6.042585372924805,
+      "learning_rate": 8.529432801659396e-06,
+      "loss": 0.4808406448364258,
+      "step": 2950
+    },
+    {
+      "epoch": 1.210165389269867,
+      "grad_norm": 4.3166303634643555,
+      "learning_rate": 8.493669980688077e-06,
+      "loss": 0.5053442764282227,
+      "step": 3000
+    },
+    {
+      "epoch": 1.2303348124243647,
+      "grad_norm": 3.1164700984954834,
+      "learning_rate": 8.45790715971676e-06,
+      "loss": 0.5191451644897461,
+      "step": 3050
+    },
+    {
+      "epoch": 1.2505042355788625,
+      "grad_norm": 10.162965774536133,
+      "learning_rate": 8.42214433874544e-06,
+      "loss": 0.46246818542480467,
+      "step": 3100
+    },
+    {
+      "epoch": 1.2706736587333602,
+      "grad_norm": 4.169692516326904,
+      "learning_rate": 8.386381517774123e-06,
+      "loss": 0.4679804229736328,
+      "step": 3150
+    },
+    {
+      "epoch": 1.290843081887858,
+      "grad_norm": 3.540759325027466,
+      "learning_rate": 8.350618696802804e-06,
+      "loss": 0.46444507598876955,
+      "step": 3200
+    },
+    {
+      "epoch": 1.3110125050423558,
+      "grad_norm": 11.483667373657227,
+      "learning_rate": 8.314855875831487e-06,
+      "loss": 0.4642531585693359,
+      "step": 3250
+    },
+    {
+      "epoch": 1.3311819281968535,
+      "grad_norm": 3.5641000270843506,
+      "learning_rate": 8.279093054860168e-06,
+      "loss": 0.4847731018066406,
+      "step": 3300
+    },
+    {
+      "epoch": 1.3513513513513513,
+      "grad_norm": 3.0268476009368896,
+      "learning_rate": 8.24333023388885e-06,
+      "loss": 0.47580326080322266,
+      "step": 3350
+    },
+    {
+      "epoch": 1.371520774505849,
+      "grad_norm": 7.949971675872803,
+      "learning_rate": 8.207567412917532e-06,
+      "loss": 0.4675758743286133,
+      "step": 3400
+    },
+    {
+      "epoch": 1.3916901976603469,
+      "grad_norm": 2.5996310710906982,
+      "learning_rate": 8.171804591946213e-06,
+      "loss": 0.4324866485595703,
+      "step": 3450
+    },
+    {
+      "epoch": 1.4118596208148446,
+      "grad_norm": 3.861765146255493,
+      "learning_rate": 8.136041770974894e-06,
+      "loss": 0.43339492797851564,
+      "step": 3500
+    },
+    {
+      "epoch": 1.4320290439693424,
+      "grad_norm": 3.4504647254943848,
+      "learning_rate": 8.100278950003577e-06,
+      "loss": 0.39920875549316404,
+      "step": 3550
+    },
+    {
+      "epoch": 1.4521984671238402,
+      "grad_norm": 3.7376797199249268,
+      "learning_rate": 8.064516129032258e-06,
+      "loss": 0.44303287506103517,
+      "step": 3600
+    },
+    {
+      "epoch": 1.472367890278338,
+      "grad_norm": 2.2165110111236572,
+      "learning_rate": 8.02875330806094e-06,
+      "loss": 0.4385309600830078,
+      "step": 3650
+    },
+    {
+      "epoch": 1.4925373134328357,
+      "grad_norm": 3.3696818351745605,
+      "learning_rate": 7.992990487089622e-06,
+      "loss": 0.4112298583984375,
+      "step": 3700
+    },
+    {
+      "epoch": 1.5127067365873335,
+      "grad_norm": 8.021625518798828,
+      "learning_rate": 7.957227666118305e-06,
+      "loss": 0.44179962158203123,
+      "step": 3750
+    },
+    {
+      "epoch": 1.5328761597418312,
+      "grad_norm": 3.0229880809783936,
+      "learning_rate": 7.921464845146986e-06,
+      "loss": 0.48030174255371094,
+      "step": 3800
+    },
+    {
+      "epoch": 1.5530455828963292,
+      "grad_norm": 11.356090545654297,
+      "learning_rate": 7.885702024175667e-06,
+      "loss": 0.40241451263427735,
+      "step": 3850
+    },
+    {
+      "epoch": 1.573215006050827,
+      "grad_norm": 5.1527323722839355,
+      "learning_rate": 7.84993920320435e-06,
+      "loss": 0.4361794662475586,
+      "step": 3900
+    },
+    {
+      "epoch": 1.5933844292053247,
+      "grad_norm": 3.763461112976074,
+      "learning_rate": 7.814176382233031e-06,
+      "loss": 0.4061758041381836,
+      "step": 3950
+    },
+    {
+      "epoch": 1.6135538523598225,
+      "grad_norm": 2.5149497985839844,
+      "learning_rate": 7.778413561261712e-06,
+      "loss": 0.39749195098876955,
+      "step": 4000
+    },
+    {
+      "epoch": 1.6337232755143203,
+      "grad_norm": 5.366840839385986,
+      "learning_rate": 7.742650740290395e-06,
+      "loss": 0.3932053756713867,
+      "step": 4050
+    },
+    {
+      "epoch": 1.653892698668818,
+      "grad_norm": 4.50701904296875,
+      "learning_rate": 7.706887919319076e-06,
+      "loss": 0.36299785614013674,
+      "step": 4100
+    },
+    {
+      "epoch": 1.6740621218233158,
+      "grad_norm": 5.8101887702941895,
+      "learning_rate": 7.671125098347757e-06,
+      "loss": 0.3728195571899414,
+      "step": 4150
+    },
+    {
+      "epoch": 1.6942315449778136,
+      "grad_norm": 2.5029637813568115,
+      "learning_rate": 7.63536227737644e-06,
+      "loss": 0.35891620635986327,
+      "step": 4200
+    },
+    {
+      "epoch": 1.7144009681323116,
+      "grad_norm": 4.704710483551025,
+      "learning_rate": 7.599599456405122e-06,
+      "loss": 0.37625602722167967,
+      "step": 4250
+    },
+    {
+      "epoch": 1.7345703912868093,
+      "grad_norm": 3.2719874382019043,
+      "learning_rate": 7.563836635433803e-06,
+      "loss": 0.37053287506103516,
+      "step": 4300
+    },
+    {
+      "epoch": 1.754739814441307,
+      "grad_norm": 3.344325065612793,
+      "learning_rate": 7.528073814462485e-06,
+      "loss": 0.3574095916748047,
+      "step": 4350
+    },
+    {
+      "epoch": 1.7749092375958049,
+      "grad_norm": 3.606675863265991,
+      "learning_rate": 7.492310993491168e-06,
+      "loss": 0.35118736267089845,
+      "step": 4400
+    },
+    {
+      "epoch": 1.7950786607503026,
+      "grad_norm": 3.3224735260009766,
+      "learning_rate": 7.456548172519848e-06,
+      "loss": 0.3643968963623047,
+      "step": 4450
+    },
+    {
+      "epoch": 1.8152480839048004,
+      "grad_norm": 5.382266521453857,
+      "learning_rate": 7.420785351548531e-06,
+      "loss": 0.3956292724609375,
+      "step": 4500
+    },
+    {
+      "epoch": 1.8354175070592982,
+      "grad_norm": 5.572700023651123,
+      "learning_rate": 7.385022530577213e-06,
+      "loss": 0.3590609741210937,
+      "step": 4550
+    },
+    {
+      "epoch": 1.855586930213796,
+      "grad_norm": 4.492390155792236,
+      "learning_rate": 7.349259709605895e-06,
+      "loss": 0.3465123748779297,
+      "step": 4600
+    },
+    {
+      "epoch": 1.8757563533682937,
+      "grad_norm": 4.636186122894287,
+      "learning_rate": 7.313496888634576e-06,
+      "loss": 0.32605178833007814,
+      "step": 4650
+    },
+    {
+      "epoch": 1.8959257765227915,
+      "grad_norm": 2.553628921508789,
+      "learning_rate": 7.277734067663258e-06,
+      "loss": 0.33962291717529297,
+      "step": 4700
+    },
+    {
+      "epoch": 1.9160951996772893,
+      "grad_norm": 4.120754241943359,
+      "learning_rate": 7.24197124669194e-06,
+      "loss": 0.31869050979614255,
+      "step": 4750
+    },
+    {
+      "epoch": 1.936264622831787,
+      "grad_norm": 4.462713241577148,
+      "learning_rate": 7.206208425720621e-06,
+      "loss": 0.36314701080322265,
+      "step": 4800
+    },
+    {
+      "epoch": 1.9564340459862848,
+      "grad_norm": 5.850078582763672,
+      "learning_rate": 7.170445604749303e-06,
+      "loss": 0.3359333419799805,
+      "step": 4850
+    },
+    {
+      "epoch": 1.9766034691407826,
+      "grad_norm": 7.381692409515381,
+      "learning_rate": 7.134682783777985e-06,
+      "loss": 0.3309978103637695,
+      "step": 4900
+    },
+    {
+      "epoch": 1.9967728922952803,
+      "grad_norm": 5.099913597106934,
+      "learning_rate": 7.098919962806666e-06,
+      "loss": 0.32351688385009764,
+      "step": 4950
+    },
+    {
+      "epoch": 2.0,
+      "eval_loss": 0.36766618490219116,
+      "eval_runtime": 10.2855,
+      "eval_samples_per_second": 235.186,
+      "eval_steps_per_second": 14.778,
+      "step": 4958
+    },
+    {
+      "epoch": 2.016942315449778,
+      "grad_norm": 4.931946754455566,
+      "learning_rate": 7.063157141835349e-06,
+      "loss": 0.28860326766967775,
+      "step": 5000
+    },
+    {
+      "epoch": 2.037111738604276,
+      "grad_norm": 3.009385108947754,
+      "learning_rate": 7.027394320864031e-06,
+      "loss": 0.2937843894958496,
+      "step": 5050
+    },
+    {
+      "epoch": 2.0572811617587736,
+      "grad_norm": 4.8531084060668945,
+      "learning_rate": 6.991631499892713e-06,
+      "loss": 0.27569530487060545,
+      "step": 5100
+    },
+    {
+      "epoch": 2.0774505849132714,
+      "grad_norm": 3.875870943069458,
+      "learning_rate": 6.955868678921394e-06,
+      "loss": 0.2876934051513672,
+      "step": 5150
+    },
+    {
+      "epoch": 2.097620008067769,
+      "grad_norm": 2.2998695373535156,
+      "learning_rate": 6.920105857950076e-06,
+      "loss": 0.2687033462524414,
+      "step": 5200
+    },
+    {
+      "epoch": 2.117789431222267,
+      "grad_norm": 2.7476847171783447,
+      "learning_rate": 6.884343036978758e-06,
+      "loss": 0.26827131271362303,
+      "step": 5250
+    },
+    {
+      "epoch": 2.1379588543767647,
+      "grad_norm": 4.094130516052246,
+      "learning_rate": 6.848580216007439e-06,
+      "loss": 0.2869255828857422,
+      "step": 5300
+    },
+    {
+      "epoch": 2.1581282775312625,
+      "grad_norm": 4.750766277313232,
+      "learning_rate": 6.812817395036121e-06,
+      "loss": 0.2684427070617676,
+      "step": 5350
+    },
+    {
+      "epoch": 2.1782977006857602,
+      "grad_norm": 2.9605631828308105,
+      "learning_rate": 6.777054574064803e-06,
+      "loss": 0.27501359939575193,
+      "step": 5400
+    },
+    {
+      "epoch": 2.198467123840258,
+      "grad_norm": 2.3290975093841553,
+      "learning_rate": 6.741291753093484e-06,
+      "loss": 0.2675877380371094,
+      "step": 5450
+    },
+    {
+      "epoch": 2.2186365469947558,
+      "grad_norm": 2.5787155628204346,
+      "learning_rate": 6.705528932122166e-06,
+      "loss": 0.303548698425293,
+      "step": 5500
+    },
+    {
+      "epoch": 2.2388059701492535,
+      "grad_norm": 2.6872715950012207,
+      "learning_rate": 6.669766111150849e-06,
+      "loss": 0.2784063720703125,
+      "step": 5550
+    },
+    {
+      "epoch": 2.2589753933037517,
+      "grad_norm": 1.8256844282150269,
+      "learning_rate": 6.6340032901795305e-06,
+      "loss": 0.25218706130981444,
+      "step": 5600
+    },
+    {
+      "epoch": 2.279144816458249,
+      "grad_norm": 2.204469680786133,
+      "learning_rate": 6.598240469208212e-06,
+      "loss": 0.24446300506591798,
+      "step": 5650
+    },
+    {
+      "epoch": 2.2993142396127473,
+      "grad_norm": 2.390136241912842,
+      "learning_rate": 6.562477648236894e-06,
+      "loss": 0.25136438369750974,
+      "step": 5700
+    },
+    {
+      "epoch": 2.319483662767245,
+      "grad_norm": 3.441288709640503,
+      "learning_rate": 6.5267148272655755e-06,
+      "loss": 0.28397703170776367,
+      "step": 5750
+    },
+    {
+      "epoch": 2.339653085921743,
+      "grad_norm": 3.4469010829925537,
+      "learning_rate": 6.490952006294257e-06,
+      "loss": 0.22534845352172853,
+      "step": 5800
+    },
+    {
+      "epoch": 2.3598225090762406,
+      "grad_norm": 4.857476234436035,
+      "learning_rate": 6.455189185322939e-06,
+      "loss": 0.24779975891113282,
+      "step": 5850
+    },
+    {
+      "epoch": 2.3799919322307383,
+      "grad_norm": 3.7842509746551514,
+      "learning_rate": 6.4194263643516206e-06,
+      "loss": 0.2669987106323242,
+      "step": 5900
+    },
+    {
+      "epoch": 2.400161355385236,
+      "grad_norm": 2.723820924758911,
+      "learning_rate": 6.383663543380302e-06,
+      "loss": 0.2623712348937988,
+      "step": 5950
+    },
+    {
+      "epoch": 2.420330778539734,
+      "grad_norm": 2.6213226318359375,
+      "learning_rate": 6.347900722408984e-06,
+      "loss": 0.2388707733154297,
+      "step": 6000
+    },
+    {
+      "epoch": 2.4405002016942317,
+      "grad_norm": 4.395678997039795,
+      "learning_rate": 6.312137901437666e-06,
+      "loss": 0.28599718093872073,
+      "step": 6050
+    },
+    {
+      "epoch": 2.4606696248487294,
+      "grad_norm": 5.71527099609375,
+      "learning_rate": 6.276375080466348e-06,
+      "loss": 0.25312797546386717,
+      "step": 6100
+    },
+    {
+      "epoch": 2.480839048003227,
+      "grad_norm": 2.939988851547241,
+      "learning_rate": 6.2406122594950295e-06,
+      "loss": 0.24944252014160156,
+      "step": 6150
+    },
+    {
+      "epoch": 2.501008471157725,
+      "grad_norm": 4.063046932220459,
+      "learning_rate": 6.204849438523711e-06,
+      "loss": 0.21988363265991212,
+      "step": 6200
+    },
+    {
+      "epoch": 2.5211778943122227,
+      "grad_norm": 3.4909474849700928,
+      "learning_rate": 6.169086617552393e-06,
+      "loss": 0.24219284057617188,
+      "step": 6250
+    },
+    {
+      "epoch": 2.5413473174667205,
+      "grad_norm": 2.6816914081573486,
+      "learning_rate": 6.1333237965810745e-06,
+      "loss": 0.23203304290771484,
+      "step": 6300
+    },
+    {
+      "epoch": 2.5615167406212183,
+      "grad_norm": 3.423490047454834,
+      "learning_rate": 6.0975609756097564e-06,
+      "loss": 0.27278310775756837,
+      "step": 6350
+    },
+    {
+      "epoch": 2.581686163775716,
+      "grad_norm": 4.844998359680176,
+      "learning_rate": 6.061798154638438e-06,
+      "loss": 0.23865175247192383,
+      "step": 6400
+    },
+    {
+      "epoch": 2.601855586930214,
+      "grad_norm": 3.49391508102417,
+      "learning_rate": 6.02603533366712e-06,
+      "loss": 0.2672038459777832,
+      "step": 6450
+    },
+    {
+      "epoch": 2.6220250100847116,
+      "grad_norm": 3.5145552158355713,
+      "learning_rate": 5.9902725126958014e-06,
+      "loss": 0.23499755859375,
+      "step": 6500
+    },
+    {
+      "epoch": 2.6421944332392093,
+      "grad_norm": 3.1010382175445557,
+      "learning_rate": 5.954509691724483e-06,
+      "loss": 0.2238442802429199,
+      "step": 6550
+    },
+    {
+      "epoch": 2.662363856393707,
+      "grad_norm": 4.308987617492676,
+      "learning_rate": 5.918746870753166e-06,
+      "loss": 0.22217355728149413,
+      "step": 6600
+    },
+    {
+      "epoch": 2.682533279548205,
+      "grad_norm": 5.416255950927734,
+      "learning_rate": 5.882984049781847e-06,
+      "loss": 0.22912097930908204,
+      "step": 6650
+    },
+    {
+      "epoch": 2.7027027027027026,
+      "grad_norm": 2.1434426307678223,
+      "learning_rate": 5.847221228810529e-06,
+      "loss": 0.22893272399902342,
+      "step": 6700
+    },
+    {
+      "epoch": 2.7228721258572004,
+      "grad_norm": 10.477407455444336,
+      "learning_rate": 5.811458407839211e-06,
+      "loss": 0.249277286529541,
+      "step": 6750
+    },
+    {
+      "epoch": 2.743041549011698,
+      "grad_norm": 5.964328289031982,
+      "learning_rate": 5.775695586867892e-06,
+      "loss": 0.23258882522583008,
+      "step": 6800
+    },
+    {
+      "epoch": 2.763210972166196,
+      "grad_norm": 2.736809492111206,
+      "learning_rate": 5.739932765896574e-06,
+      "loss": 0.25174449920654296,
+      "step": 6850
+    },
+    {
+      "epoch": 2.7833803953206937,
+      "grad_norm": 6.482688903808594,
+      "learning_rate": 5.704169944925256e-06,
+      "loss": 0.2076584815979004,
+      "step": 6900
+    },
+    {
+      "epoch": 2.8035498184751915,
+      "grad_norm": 3.5733821392059326,
+      "learning_rate": 5.668407123953938e-06,
+      "loss": 0.23874975204467774,
+      "step": 6950
+    },
+    {
+      "epoch": 2.8237192416296892,
+      "grad_norm": 3.554004669189453,
+      "learning_rate": 5.632644302982619e-06,
+      "loss": 0.21515037536621093,
+      "step": 7000
+    },
+    {
+      "epoch": 2.8438886647841874,
+      "grad_norm": 3.9487385749816895,
+      "learning_rate": 5.596881482011301e-06,
+      "loss": 0.22660268783569337,
+      "step": 7050
+    },
+    {
+      "epoch": 2.8640580879386848,
+      "grad_norm": 3.470832347869873,
+      "learning_rate": 5.561118661039984e-06,
+      "loss": 0.23147939682006835,
+      "step": 7100
+    },
+    {
+      "epoch": 2.884227511093183,
+      "grad_norm": 8.422386169433594,
+      "learning_rate": 5.525355840068664e-06,
+      "loss": 0.2193218994140625,
+      "step": 7150
+    },
+    {
+      "epoch": 2.9043969342476803,
+      "grad_norm": 2.3972811698913574,
+      "learning_rate": 5.489593019097347e-06,
+      "loss": 0.2347504997253418,
+      "step": 7200
+    },
+    {
+      "epoch": 2.9245663574021785,
+      "grad_norm": 3.2007927894592285,
+      "learning_rate": 5.453830198126029e-06,
+      "loss": 0.22932228088378906,
+      "step": 7250
+    },
+    {
+      "epoch": 2.944735780556676,
+      "grad_norm": 2.436324119567871,
+      "learning_rate": 5.41806737715471e-06,
+      "loss": 0.20108043670654296,
+      "step": 7300
+    },
+    {
+      "epoch": 2.964905203711174,
+      "grad_norm": 4.076231956481934,
+      "learning_rate": 5.382304556183392e-06,
+      "loss": 0.22045228958129884,
+      "step": 7350
+    },
+    {
+      "epoch": 2.9850746268656714,
+      "grad_norm": 4.013544082641602,
+      "learning_rate": 5.346541735212074e-06,
+      "loss": 0.2202323532104492,
+      "step": 7400
+    },
+    {
+      "epoch": 3.0,
+      "eval_loss": 0.34891217947006226,
+      "eval_runtime": 10.4596,
+      "eval_samples_per_second": 231.27,
+      "eval_steps_per_second": 14.532,
+      "step": 7437
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 14874,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 6,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 0.0,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-7437/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d58909b889894bf332483adc7572ec6f0871c70082613ece0a5e87c6c34d05d8
+size 5265

role2id.json

@@ -0,0 +1 @@
+{"role2id": {"<NULL>": 0, "Abundant_entities": 1, "Abuser": 2, "Accessibility": 3, "Accessory": 4, "Accoutrement": 5, "Accuracy": 6, "Accused": 7, "Act": 8, "Action": 9, "Activists": 10, "Activity": 11, "Actor": 12, "Added_set": 13, "Addict": 14, "Addictant": 15, "Address": 16, "Address_term": 17, "Addressee": 18, "Adequacy": 19, "Adversity": 20, "Affected": 21, "Affected_party": 22, "Affliction": 23, "Age": 24, "Agency": 25, "Agent": 26, "Aggregate": 27, "Aggregate_property": 28, "Aggressor": 29, "Aggressors": 30, "Agreement": 31, "Agriculturist": 32, "Ailment": 33, "Air": 34, "Alignment_match": 35, "Alignment_mismatch": 36, "Alliance": 37, "Allowed_category": 38, "Alter": 39, "Alterant": 40, "Amends": 41, "Ammunition": 42, "Amount": 43, "Amount_of_discussion": 44, "Amount_of_information": 45, "Amount_of_progress": 46, "Amount_of_reassurance": 47, "Anchor": 48, "Angle": 49, "Animal": 50, "Anti_consequence": 51, "Apparatus": 52, "Apparent_conclusion": 53, "Appraisal": 54, "Area": 55, "Arguer": 56, "Arguer1": 57, "Arguer2": 58, "Arguers": 59, "Armor": 60, "Arraign_authority": 61, "Artifact": 62, "Artist": 63, "Artwork": 64, "Assailant": 65, "Assessor": 66, "Asset": 67, "Assets": 68, "Astronomical_entity": 69, "Attachment": 70, "Attack": 71, "Attacked": 72, "Attended_person": 73, "Attendee": 74, "Attendees": 75, "Attitude": 76, "Attraction": 77, "Attribute": 78, "Audience": 79, "Authenticity": 80, "Author": 81, "Authorities": 82, "Authority": 83, "Avenger": 84, "Averted_situation": 85, "Back": 86, "Bad_action": 87, "Bad_entity": 88, "Bad_outcome": 89, "Baggage": 90, "Barrier": 91, "Basis": 92, "Basis_of_order": 93, "Beam": 94, "Bearer": 95, "Bed": 96, "Behavior": 97, "Behavior_product": 98, "Believer": 99, "Benefactor": 100, "Beneficial_situation": 101, "Beneficiary": 102, "Benefit": 103, "Benefited_party": 104, "Benefiting_action": 105, "Billing_authority": 106, "Binding_principle": 107, "Body": 108, "Body_location": 109, "Body_mark": 110, "Body_part": 111, "Body_system": 112, "Bodypart_of_agent": 113, "Bomb": 114, "Bomber": 115, "Booker": 116, "Borrower": 117, "Boundary": 118, "Bounded_area": 119, "Branch": 120, "Brand": 121, "Building": 122, "Building_part": 123, "Business": 124, "Business_name": 125, "Buyer": 126, "Cache": 127, "Calculator": 128, "Capitulator": 129, "Captive": 130, "Captor": 131, "Capture": 132, "Cargo": 133, "Carrier": 134, "Case": 135, "Category": 136, "Cause": 137, "Cause_of_shine": 138, "Change_agent": 139, "Character_of_end_state": 140, "Characteristic": 141, "Characterization": 142, "Charge": 143, "Charges": 144, "Chemical_entity": 145, "Child": 146, "Chosen": 147, "Circumstance": 148, "Circumstances": 149, "Claimant": 150, "Class": 151, "Client": 152, "Clothing": 153, "Co-participant": 154, "Co-timed_event": 155, "Co-variable": 156, "Co_resident": 157, "Code": 158, "Cognate_event": 159, "Cognizer": 160, "Cognizer_1": 161, "Cognizer_2": 162, "Cognizer_agent": 163, "Cognizers": 164, "Collection": 165, "Collection_name": 166, "Colonists": 167, "Color": 168, "Color_qualifier": 169, "Combatant": 170, "Combatants": 171, "Commitment": 172, "Commonality": 173, "Communication": 174, "Communicative_act": 175, "Communicative_force": 176, "Communicator": 177, "Company": 178, "Comparand": 179, "Comparison_activity": 180, "Comparison_event": 181, "Comparison_set": 182, "Compeller": 183, "Compensation": 184, "Competing_consideration": 185, "Competition": 186, "Competitor": 187, "Competitors": 188, "Complainer": 189, "Complaint": 190, "Completeness": 191, "Complex": 192, "Component": 193, "Component_entities": 194, "Component_sound": 195, "Components": 196, "Composite_number": 197, "Conceded_state_of_affairs": 198, "Concept_1": 199, "Concept_2": 200, "Concepts": 201, "Concessive": 202, "Condition": 203, "Configuration": 204, "Conflict": 205, "Connected_item": 206, "Connected_locations": 207, "Connection": 208, "Connections": 209, "Connector": 210, "Conqueror": 211, "Consequence": 212, "Consequences": 213, "Consideration": 214, "Constancy": 215, "Constant_location": 216, "Constituent_parts": 217, "Constituents": 218, "Construction": 219, "Consumer": 220, "Container": 221, "Container_portal": 222, "Container_possessor": 223, "Containing_event": 224, "Containing_object": 225, "Containing_text": 226, "Content": 227, "Contents": 228, "Context": 229, "Context_of_acquaintance": 230, "Contract_basis": 231, "Contrary_circumstances": 232, "Contrast": 233, "Contrast_set": 234, "Contrasting_alternative": 235, "Contrastive_context": 236, "Control": 237, "Control_group": 238, "Control_group_member": 239, "Controlled_entity": 240, "Controlling_act": 241, "Controlling_entity": 242, "Controlling_situation": 243, "Controlling_variable": 244, "Conveyed_emotion": 245, "Convict": 246, "Conviction": 247, "Convoy": 248, "Cook": 249, "Coordinated_event": 250, "Copy": 251, "Correlated_event": 252, "Correlated_variable": 253, "Cotheme": 254, "Count": 255, "Counter_actor": 256, "Countertransfer": 257, "Country_of_origin": 258, "Course": 259, "Court": 260, "Created_entity": 261, "Creator": 262, "Crime": 263, "Crime_jurisdiction": 264, "Criteria": 265, "Criterion": 266, "Crop": 267, "Culture": 268, "Current_country": 269, "Current_instance": 270, "Current_jurisdiction": 271, "Current_leadership": 272, "Current_order": 273, "Customer": 274, "Danger": 275, "Dangerous_entity": 276, "Dangerous_situation": 277, "Data": 278, "Debtor": 279, "Deceased": 280, "Deceiver": 281, "Deception": 282, "Decision": 283, "Decision_maker": 284, "Decorated_individual": 285, "Decoration": 286, "Decoration_descriptor": 287, "Dedicated_entity": 288, "Dedicated_type": 289, "Defendant": 290, "Defender": 291, "Defense": 292, "Defining_event": 293, "Deformer": 294, "Degree": 295, "Degree_of_involvement": 296, "Deixis": 297, "Deliverer": 298, "Delivery_device": 299, "Demands": 300, "Denominator": 301, "Denoted_quantity": 302, "Dependent": 303, "Dependent_entity": 304, "Dependent_individual": 305, "Dependent_situation": 306, "Dependent_state": 307, "Dependent_variable": 308, "Depictive": 309, "Depictive_of_represented": 310, "Described_entity": 311, "Descriptor": 312, "Designer": 313, "Desirability": 314, "Desirable": 315, "Desirable_action": 316, "Desirable_situation": 317, "Desired_goal": 318, "Desired_state_of_affairs": 319, "Destination_event": 320, "Destination_time": 321, "Destroyer": 322, "Determinant": 323, "Detonator": 324, "Developer": 325, "Deviation": 326, "Device": 327, "Difference": 328, "Differentiating_fact": 329, "Dimension": 330, "Direction": 331, "Directness": 332, "Disaster": 333, "Discourse_salient_country": 334, "Disease": 335, "Display": 336, "Dispute": 337, "Distance": 338, "Distant_profiled_region": 339, "Distributor": 340, "Document": 341, "Documents": 342, "Dodger": 343, "Domain": 344, "Domain_of_relevance": 345, "Donor": 346, "Dough": 347, "Driver": 348, "Dryee": 349, "Duration": 350, "Duration_of_activity": 351, "Duration_of_final_state": 352, "Duration_of_state": 353, "Duty": 354, "Earner": 355, "Earnings": 356, "Eclipsed": 357, "Economy": 358, "Effect": 359, "Effort": 360, "Egg": 361, "Ego": 362, "Elapsed_time": 363, "Electricity": 364, "Element": 365, "Emission": 366, "Emitter": 367, "Emotion": 368, "Emotional_state": 369, "Empathy_target": 370, "Employee": 371, "Employer": 372, "Employment": 373, "Enabled_action": 374, "Enabled_situation": 375, "Enclosed_region": 376, "End_point": 377, "End_time": 378, "Endangering_act": 379, "Endeavor": 380, "Endpoints": 381, "Enemy": 382, "Entities": 383, "Entity": 384, "Entity_1": 385, "Entity_2": 386, "Entourage": 387, "Entry_path": 388, "Equivalence": 389, "Escapee": 390, "Estimation": 391, "Ethnicity": 392, "Etymological_root": 393, "Evader": 394, "Evaluation": 395, "Evaluator": 396, "Evaluee": 397, "Event": 398, "Event_description": 399, "Events": 400, "Eventuality": 401, "Evidence": 402, "Examination": 403, "Examinee": 404, "Examiner": 405, "Excess": 406, "Exchange_rate": 407, "Exchange_service": 408, "Exchanger_1": 409, "Exchanger_2": 410, "Exchangers": 411, "Excreta": 412, "Excreter": 413, "Executed": 414, "Execution": 415, "Executioner": 416, "Executive_authority": 417, "Exerter": 418, "Existing_member": 419, "Expectation": 420, "Expected_entity": 421, "Expected_event": 422, "Expected_request": 423, "Experience": 424, "Experiencer": 425, "Experiment": 426, "Experimenter": 427, "Explanation": 428, "Explosive": 429, "Exporter": 430, "Exporting_area": 431, "Expression": 432, "Expression_1": 433, "Expression_2": 434, "Expressor": 435, "Extent": 436, "Extent_of_acclaim": 437, "Exterior": 438, "External_cause": 439, "Facility": 440, "Fact": 441, "Factor": 442, "Factory": 443, "Factual_situation": 444, "Faculty": 445, "Falling_event": 446, "Fastener": 447, "Father": 448, "Feature": 449, "Fidelity": 450, "Field": 451, "Fight": 452, "Fighter": 453, "Figure": 454, "Figures": 455, "Final_category": 456, "Final_correlate": 457, "Final_element": 458, "Final_number": 459, "Final_quality": 460, "Final_situation": 461, "Final_state": 462, "Final_subevent": 463, "Final_temperature": 464, "Final_value": 465, "Financial_instrument": 466, "Finding": 467, "Fine": 468, "Fire": 469, "Fire_break": 470, "Firearm": 471, "Firefighter": 472, "Firing_point": 473, "Fixed_location": 474, "Flame": 475, "Flammables": 476, "Fluid": 477, "Focal_entity": 478, "Focal_feature": 479, "Focal_occasion": 480, "Focal_participant": 481, "Focal_time": 482, "Following_distance": 483, "Food": 484, "Forbidden": 485, "Force": 486, "Foreground_event": 487, "Forgery": 488, "Forgoer": 489, "Form": 490, "Formal_realization": 491, "Formational_cause": 492, "Frequency": 493, "Frequency_of_success": 494, "Frequency_of_use": 495, "Front": 496, "Fuel": 497, "Fugitive": 498, "Function": 499, "Future_phenomenon": 500, "Gambler": 501, "Game": 502, "Garment": 503, "Gatherer": 504, "Genre": 505, "Gift": 506, "Gizmo": 507, "Goal": 508, "Goal_area": 509, "Goods": 510, "Governed": 511, "Governing_authority": 512, "Gradable_attribute": 513, "Graph": 514, "Grinder": 515, "Grinding_cause": 516, "Ground": 517, "Grounds": 518, "Group": 519, "Grower": 520, "Guardian": 521, "Guest": 522, "Hair": 523, "Hair_property": 524, "Handle": 525, "Harmful_event": 526, "Harmful_situation": 527, "Healer": 528, "Hearer": 529, "Heat_source": 530, "Heating_instrument": 531, "Hedged_content": 532, "Helper": 533, "Hiatus": 534, "Hidden_object": 535, "Hiding_place": 536, "Hindrance": 537, "Holder": 538, "Holding_location": 539, "Homeland": 540, "Honored_entity": 541, "Host": 542, "Host_location": 543, "Hot_Cold_source": 544, "Hot_cold_source": 545, "Hunter": 546, "Hypothesis": 547, "Hypothetical_event": 548, "Idea": 549, "Identity": 550, "Idiosyncrasy": 551, "Illicit_activity": 552, "Illicit_organization": 553, "Image": 554, "Impactee": 555, "Impactor": 556, "Impactors": 557, "Importer": 558, "Importing_area": 559, "Imposed_purpose": 560, "Impression": 561, "Incident": 562, "Increment": 563, "Independent_variable": 564, "Indicated": 565, "Indicated_entity": 566, "Indicated_resident": 567, "Indicator": 568, "Individual": 569, "Individual_1": 570, "Individual_2": 571, "Individuals": 572, "Infected_entity": 573, "Infection": 574, "Infection_cause": 575, "Inference": 576, "Influence": 577, "Influencing_entity": 578, "Influencing_situation": 579, "Influencing_variable": 580, "Information": 581, "Information_source": 582, "Informer": 583, "Infrastructure": 584, "Ingestibles": 585, "Ingestor": 586, "Ingredients": 587, "Inherent_purpose": 588, "Initial_category": 589, "Initial_correlate": 590, "Initial_duration": 591, "Initial_element": 592, "Initial_number": 593, "Initial_quality": 594, "Initial_set": 595, "Initial_situation": 596, "Initial_size": 597, "Initial_state": 598, "Initial_subevent": 599, "Initial_temperature": 600, "Initial_value": 601, "Injured_party": 602, "Injuring_entity": 603, "Injury": 604, "Inmates": 605, "Input": 606, "Inspector": 607, "Instance": 608, "Instance_prop": 609, "Institution": 610, "Instrument": 611, "Insulator": 612, "Intended_academic_credit": 613, "Intended_event": 614, "Intended_goal": 615, "Intended_perceiver": 616, "Intended_referent": 617, "Intended_wearing_location": 618, "Interceptor": 619, "Interest": 620, "Interested_party": 621, "Interior": 622, "Interjection": 623, "Interlocutor": 624, "Interlocutor_1": 625, "Interlocutor_2": 626, "Interlocutors": 627, "Intermediary": 628, "Intermediate_value": 629, "Internal_cause": 630, "Interval": 631, "Intervention": 632, "Intoxicant": 633, "Invader": 634, "Invasion_act": 635, "Investigator": 636, "Issue": 637, "Issuer": 638, "Item": 639, "Item_1": 640, "Item_2": 641, "Item_property": 642, "Items": 643, "Iteration_count": 644, "Iterations": 645, "Itinerary": 646, "Joker": 647, "Journey": 648, "Judge": 649, "Judges": 650, "Judgment": 651, "Judicial_body": 652, "Jurisdiction": 653, "Jury": 654, "Key": 655, "Killer": 656, "Kindler": 657, "Knot": 658, "Knowledge": 659, "Known_resident": 660, "Label": 661, "Land": 662, "Landmark": 663, "Landmark_entity": 664, "Landmark_event": 665, "Landmark_feature": 666, "Landmark_occasion": 667, "Landmark_period": 668, "Landmark_time": 669, "Language": 670, "Larger_purpose": 671, "Law": 672, "Leader": 673, "Legal_basis": 674, "Lender": 675, "Length": 676, "Lessee": 677, "Lessor": 678, "Level": 679, "Lifestyle": 680, "Light": 681, "Limit1": 682, "Limit2": 683, "Limits": 684, "Limits_of_consideration": 685, "Liquid": 686, "Loan": 687, "Loaner": 688, "Locale": 689, "Location": 690, "Location_of_appearance": 691, "Location_of_assailant": 692, "Location_of_confinement": 693, "Location_of_event": 694, "Location_of_protagonist": 695, "Location_of_representation": 696, "Location_of_sound_source": 697, "Locus": 698, "Lodger": 699, "Loser": 700, "Loyal_side": 701, "Main_assertion": 702, "Maleficiary": 703, "Manifestation_of_bias": 704, "Manipulator": 705, "Manner": 706, "Manner_descriptor": 707, "Margin": 708, "Marker": 709, "Market": 710, "Market_type": 711, "Mass": 712, "Mass_theme": 713, "Material": 714, "Matter": 715, "Meaning": 716, "Means": 717, "Means_of_communication": 718, "Means_of_gathering": 719, "Means_of_motion": 720, "Means_of_propulsion": 721, "Means_of_transfer": 722, "Measure": 723, "Measurement": 724, "Medic": 725, "Medical_center": 726, "Medical_condition": 727, "Medical_professional": 728, "Medication": 729, "Medium": 730, "Member": 731, "Member_1": 732, "Member_2": 733, "Members": 734, "Mental_content": 735, "Message": 736, "Method": 737, "Military_organization": 738, "Miner": 739, "Mineral": 740, "Misdeed": 741, "Mitigating_situation": 742, "Mode_of_transfer": 743, "Mode_of_transportation": 744, "Money": 745, "Money_owner": 746, "Mother": 747, "Multiplier": 748, "Name": 749, "Name_source": 750, "Namesake": 751, "Nationality": 752, "Near_profiled_region": 753, "Negated_proposition": 754, "Negative_consequences": 755, "Network": 756, "New": 757, "New_area": 758, "New_duration": 759, "New_idea": 760, "New_leader": 761, "New_member": 762, "New_order": 763, "New_punishment": 764, "New_situation": 765, "New_status": 766, "New_tool": 767, "Next_subevent": 768, "Nodes": 769, "Noise_maker": 770, "Noisy_event": 771, "Norm": 772, "Normal_location": 773, "Number": 774, "Number_of_possibilities": 775, "Numbers": 776, "Numerator": 777, "Object": 778, "Obligated_party": 779, "Obligation": 780, "Obligator": 781, "Obscuring_medium": 782, "Obstruction": 783, "Occasion": 784, "Occupant": 785, "Odds": 786, "Offender": 787, "Offense": 788, "Offerer": 789, "Official": 790, "Offspring": 791, "Old": 792, "Old_leader": 793, "Old_order": 794, "Old_tool": 795, "Operator": 796, "Opinion": 797, "Opponent": 798, "Opportunity": 799, "Opposing_force": 800, "Opposite_possibility": 801, "Option_1": 802, "Option_2": 803, "Options": 804, "Ordinality": 805, "Organism": 806, "Organization": 807, "Orientation": 808, "Orientational_location": 809, "Origin": 810, "Original": 811, "Original_body": 812, "Original_context": 813, "Original_owner": 814, "Original_path": 815, "Original_punishment": 816, "Other": 817, "Outcome": 818, "Output": 819, "Owed_party": 820, "Owner": 821, "Paradigm": 822, "Parameter": 823, "Parent_organization": 824, "Parents": 825, "Part": 826, "Part_1": 827, "Part_2": 828, "Part_of_form": 829, "Part_prop": 830, "Participant": 831, "Participant1": 832, "Participant2": 833, "Participant_1": 834, "Participant_2": 835, "Participants": 836, "Particular_iteration": 837, "Parties": 838, "Partner_1": 839, "Partner_2": 840, "Partners": 841, "Parts": 842, "Party": 843, "Party_1": 844, "Party_2": 845, "Path": 846, "Path_of_gaze": 847, "Path_shape": 848, "Patient": 849, "Patrol": 850, "Pattern": 851, "Payer": 852, "Payment": 853, "Payment_status": 854, "Penal_institution": 855, "Perceiver": 856, "Perceiver_agentive": 857, "Perceiver_passive": 858, "Percept": 859, "Perceptual_source": 860, "Performance": 861, "Performer": 862, "Performer1": 863, "Performer2": 864, "Period": 865, "Period_of_existence": 866, "Period_of_iterations": 867, "Periodicity": 868, "Perpetrator": 869, "Persistent_characteristic": 870, "Person": 871, "Personnel": 872, "Phase": 873, "Phenomena": 874, "Phenomenon": 875, "Phenomenon_1": 876, "Phenomenon_2": 877, "Picture": 878, "Piece": 879, "Piece_prop": 880, "Pieces": 881, "Place": 882, "Place_of_employment": 883, "Place_of_operation": 884, "Plan": 885, "Plant": 886, "Player": 887, "Plea": 888, "Point_of_contact": 889, "Point_of_view": 890, "Political_region": 891, "Population": 892, "Portal": 893, "Portion": 894, "Position": 895, "Position_qualifier": 896, "Positive_consequence": 897, "Possession": 898, "Possessor": 899, "Possibilities": 900, "Possible_event": 901, "Possible_sentence": 902, "Post": 903, "Post_state": 904, "Potential_hindrance": 905, "Potential_observer": 906, "Potential_recipient": 907, "Power_source": 908, "Practice": 909, "Practitioner": 910, "Pragmatic_marker": 911, "Prank": 912, "Precept": 913, "Precipitation": 914, "Precision": 915, "Predicate": 916, "Predictive_phenomenon": 917, "Preparatory_act": 918, "Preparatory_phase": 919, "Presentation": 920, "Preventing_cause": 921, "Previous_context": 922, "Previous_instance": 923, "Price": 924, "Principle": 925, "Prior_state": 926, "Prison": 927, "Prisoner": 928, "Prize": 929, "Problem": 930, "Process": 931, "Produced_food": 932, "Producer": 933, "Product": 934, "Production": 935, "Products": 936, "Professional": 937, "Profiled_alternative": 938, "Profiled_attribute": 939, "Profiled_event": 940, "Profiled_item": 941, "Profiled_possibility": 942, "Profiled_region": 943, "Project": 944, "Projectile": 945, "Property": 946, "Proposal": 947, "Proposed_action": 948, "Proposition": 949, "Proprietor": 950, "Prosecution": 951, "Protagonist": 952, "Protagonist_1": 953, "Protagonist_2": 954, "Protagonists": 955, "Protection": 956, "Protester": 957, "Publisher": 958, "Punishment": 959, "Purchaser": 960, "Purpose": 961, "Purpose_of_donor": 962, "Purpose_of_event": 963, "Purpose_of_theme": 964, "Pursuer": 965, "Q_prop": 966, "Qualification": 967, "Quality": 968, "Quantity": 969, "Question": 970, "Questioner": 971, "Quote": 972, "Range_of_options": 973, "Rank": 974, "Rate": 975, "Ratifier": 976, "Ratio": 977, "Re-encoding": 978, "Reader": 979, "Reason": 980, "Rebel": 981, "Recipient": 982, "Recipients": 983, "Reciprocation": 984, "Record": 985, "Recorder": 986, "Reference": 987, "Reference_category": 988, "Reference_occasion": 989, "Reference_point": 990, "Reference_quantity": 991, "Referent": 992, "Refreshment": 993, "Region_quantification": 994, "Regulated_entity": 995, "Related_event": 996, "Relation": 997, "Relation_type": 998, "Relationship": 999, "Relative_location": 1000, "Relative_time": 1001, "Relatives": 1002, "Relay": 1003, "Relevant_feature": 1004, "Reliability": 1005, "Religion": 1006, "Remainder": 1007, "Reported_fact": 1008, "Representation": 1009, "Representative": 1010, "Represented": 1011, "Required": 1012, "Required_entity": 1013, "Required_individual": 1014, "Required_situation": 1015, "Requirement": 1016, "Requirements": 1017, "Researcher": 1018, "Residence": 1019, "Resident": 1020, "Residents": 1021, "Resistant_surface": 1022, "Resisting_entity": 1023, "Resource": 1024, "Resource_controller": 1025, "Responder": 1026, "Responding_entity": 1027, "Response": 1028, "Response_action": 1029, "Responsible_party": 1030, "Rest": 1031, "Restrictor": 1032, "Result": 1033, "Result_size": 1034, "Resultant_action": 1035, "Resultant_configuration": 1036, "Resultant_situation": 1037, "Resulting_action": 1038, "Rider": 1039, "Right": 1040, "Road": 1041, "Roadway": 1042, "Role": 1043, "Role_of_focal_participant": 1044, "Rope": 1045, "Route": 1046, "Rule": 1047, "Salient_activity": 1048, "Salient_entity": 1049, "Salient_event": 1050, "Scale": 1051, "Scene": 1052, "Scheduled_time": 1053, "Score": 1054, "Script": 1055, "Searcher": 1056, "Seat": 1057, "Seat_of_emotion": 1058, "Security": 1059, "Seeker": 1060, "Selector": 1061, "Self_mover": 1062, "Seller": 1063, "Sender": 1064, "Sensory_attribute": 1065, "Sent_item": 1066, "Sentence": 1067, "Sentient_entity": 1068, "Service_action": 1069, "Service_provider": 1070, "Services": 1071, "Set": 1072, "Set_of_instances": 1073, "Severity": 1074, "Shape": 1075, "Shape_prop": 1076, "Shareholder": 1077, "Sharpness": 1078, "Shopper": 1079, "Side": 1080, "Side_1": 1081, "Side_2": 1082, "Side_effects": 1083, "Sides": 1084, "Sign": 1085, "Sign_1": 1086, "Sign_2": 1087, "Signatory": 1088, "Signature": 1089, "Signs": 1090, "Simulated_entity": 1091, "Situation": 1092, "Size": 1093, "Size_change": 1094, "Skill": 1095, "Sleep_state": 1096, "Sleeper": 1097, "Social_actor": 1098, "Social_event": 1099, "Society": 1100, "Sought_entity": 1101, "Sought_phenomenon": 1102, "Sound": 1103, "Sound_maker": 1104, "Sound_source": 1105, "Source": 1106, "Source_currency": 1107, "Source_emitter": 1108, "Source_material": 1109, "Source_of_authority": 1110, "Source_of_information": 1111, "Source_of_legal_authority": 1112, "Source_of_relationship": 1113, "Source_representation": 1114, "Source_symbol": 1115, "Speaker": 1116, "Specialty": 1117, "Specification": 1118, "Specified_content": 1119, "Specified_entity": 1120, "Speed": 1121, "Standard": 1122, "Standard_attribute": 1123, "Standard_item": 1124, "Standing": 1125, "Start_time": 1126, "Starting_point": 1127, "State": 1128, "State_of_affairs": 1129, "State_of_represented": 1130, "Static_object": 1131, "Station": 1132, "Status": 1133, "Stimulus": 1134, "Stock": 1135, "Strength": 1136, "Student": 1137, "Studio": 1138, "Stuff": 1139, "Style": 1140, "Sub-source": 1141, "Subevent": 1142, "Subgroup": 1143, "Subject": 1144, "Subjects": 1145, "Submittor": 1146, "Subordinate": 1147, "Subpart": 1148, "Subpart_property": 1149, "Subregion": 1150, "Subregion_bodypart": 1151, "Subset": 1152, "Substance": 1153, "Subtype": 1154, "Sufferer": 1155, "Sum_1": 1156, "Sum_2": 1157, "Summary": 1158, "Superior": 1159, "Superset": 1160, "Supplier": 1161, "Supply": 1162, "Support": 1163, "Supported": 1164, "Supporter": 1165, "Surface": 1166, "Surrenderer": 1167, "Survivor": 1168, "Suspect": 1169, "Symptom": 1170, "System": 1171, "Systemic_effect": 1172, "Target": 1173, "Target_currency": 1174, "Target_group": 1175, "Target_location": 1176, "Target_representation": 1177, "Target_symbol": 1178, "Target_value": 1179, "Targeted": 1180, "Task": 1181, "Teacher": 1182, "Team": 1183, "Temperature": 1184, "Temperature_change": 1185, "Temperature_goal": 1186, "Temperature_setting": 1187, "Temperature_start": 1188, "Temporal_profile": 1189, "Temporary_group": 1190, "Term": 1191, "Term1": 1192, "Term2": 1193, "Term_1": 1194, "Term_2": 1195, "Term_of_sentence": 1196, "Terms": 1197, "Terrorist": 1198, "Tested_property": 1199, "Tester": 1200, "Text": 1201, "Textual_location": 1202, "Theme": 1203, "Theme_1": 1204, "Theme_2": 1205, "Themes": 1206, "Time": 1207, "Time_length": 1208, "Time_of_creation": 1209, "Time_of_event": 1210, "Time_of_eventuality": 1211, "Time_of_implementation": 1212, "Time_of_phenomenon": 1213, "Time_period": 1214, "Time_span": 1215, "Timespan": 1216, "Timetable": 1217, "Title": 1218, "Tool": 1219, "Tools": 1220, "Topic": 1221, "Topical_entity": 1222, "Total": 1223, "Tourist": 1224, "Toxic_substance": 1225, "Toxin_source": 1226, "Trajector": 1227, "Trajector_entity": 1228, "Trajector_event": 1229, "Trajector_period": 1230, "Transferors": 1231, "Transitional_period": 1232, "Transport_means": 1233, "Trap": 1234, "Travel_means": 1235, "Traveler": 1236, "Traveller": 1237, "Treated_individual": 1238, "Treatment": 1239, "Treatment_group": 1240, "Trigger": 1241, "Type": 1242, "Type_of_possession": 1243, "Type_property": 1244, "Uncertain_situation": 1245, "Unconfirmed_content": 1246, "Undertaking": 1247, "Undesirable_consequence": 1248, "Undesirable_event": 1249, "Undesirable_location": 1250, "Undesirable_situation": 1251, "Unit": 1252, "Unresolved_referent": 1253, "Unwanted_characteristics": 1254, "Unwanted_entity": 1255, "Use": 1256, "Useful_location": 1257, "User": 1258, "Value": 1259, "Value_1": 1260, "Value_2": 1261, "Value_range": 1262, "Valued_entity": 1263, "Vantage_point": 1264, "Variable": 1265, "Vehicle": 1266, "Vehicles": 1267, "Venue": 1268, "Victim": 1269, "Viewpoint": 1270, "Visitor": 1271, "Vividness": 1272, "Voice": 1273, "Volume": 1274, "Weapon": 1275, "Wearer": 1276, "Weather": 1277, "Whole": 1278, "Whole_patient": 1279, "Wielder": 1280, "Winner": 1281, "Witness": 1282, "Work": 1283, "Wrong": 1284, "Wrongdoer": 1285}, "base_model": "microsoft/deberta-v3-large"}

tokenizer_config.json

@@ -0,0 +1,23 @@
+{
+  "add_prefix_space": true,
+  "backend": "tokenizers",
+  "bos_token": "[CLS]",
+  "cls_token": "[CLS]",
+  "do_lower_case": false,
+  "eos_token": "[SEP]",
+  "extra_special_tokens": [
+    "<t>",
+    "</t>"
+  ],
+  "is_local": false,
+  "local_files_only": false,
+  "mask_token": "[MASK]",
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "split_by_punct": false,
+  "tokenizer_class": "DebertaV2Tokenizer",
+  "unk_id": 3,
+  "unk_token": "[UNK]",
+  "vocab_type": "spm"
+}