Invalid JSON: Unexpected non-whitespace character after JSONat line 1067, column 2
| [ | |
| { | |
| "loss": 0.6932, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.02, | |
| "step": 1 | |
| }, | |
| { | |
| "loss": 0.6489, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.04, | |
| "step": 2 | |
| }, | |
| { | |
| "loss": 0.5662, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.05, | |
| "step": 3 | |
| }, | |
| { | |
| "loss": 0.5949, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.07, | |
| "step": 4 | |
| }, | |
| { | |
| "loss": 0.6451, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.09, | |
| "step": 5 | |
| }, | |
| { | |
| "loss": 0.5372, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.11, | |
| "step": 6 | |
| }, | |
| { | |
| "loss": 0.6065, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.12, | |
| "step": 7 | |
| }, | |
| { | |
| "loss": 0.6478, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.14, | |
| "step": 8 | |
| }, | |
| { | |
| "loss": 0.676, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.16, | |
| "step": 9 | |
| }, | |
| { | |
| "loss": 0.5954, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.18, | |
| "step": 10 | |
| }, | |
| { | |
| "loss": 0.5809, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.19, | |
| "step": 11 | |
| }, | |
| { | |
| "loss": 0.5934, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.21, | |
| "step": 12 | |
| }, | |
| { | |
| "loss": 0.6806, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.23, | |
| "step": 13 | |
| }, | |
| { | |
| "loss": 0.5866, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.25, | |
| "step": 14 | |
| }, | |
| { | |
| "loss": 0.6057, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.26, | |
| "step": 15 | |
| }, | |
| { | |
| "loss": 0.6071, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.28, | |
| "step": 16 | |
| }, | |
| { | |
| "loss": 0.484, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.3, | |
| "step": 17 | |
| }, | |
| { | |
| "loss": 0.5456, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.32, | |
| "step": 18 | |
| }, | |
| { | |
| "loss": 0.5866, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.33, | |
| "step": 19 | |
| }, | |
| { | |
| "loss": 0.5685, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.35, | |
| "step": 20 | |
| }, | |
| { | |
| "loss": 0.561, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.37, | |
| "step": 21 | |
| }, | |
| { | |
| "loss": 0.6703, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.39, | |
| "step": 22 | |
| }, | |
| { | |
| "loss": 0.5549, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.4, | |
| "step": 23 | |
| }, | |
| { | |
| "loss": 0.5696, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.42, | |
| "step": 24 | |
| }, | |
| { | |
| "loss": 0.6033, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.44, | |
| "step": 25 | |
| }, | |
| { | |
| "eval_code_hard_loss": 0.5722607374191284, | |
| "eval_code_hard_score": -0.20501708984375, | |
| "eval_code_hard_brier_score": 0.20501708984375, | |
| "eval_code_hard_average_probability": 0.5791293978691101, | |
| "eval_code_hard_accuracy": 0.72, | |
| "eval_code_hard_probabilities": [ | |
| 0.5183141827583313, | |
| 0.5020153522491455, | |
| 0.4965677261352539, | |
| 0.9698847532272339, | |
| 0.9712179899215698, | |
| 0.9694114327430725, | |
| 0.5282353162765503, | |
| 0.537986695766449, | |
| 0.539186954498291, | |
| 0.5343082547187805, | |
| 0.5109472274780273, | |
| 0.5189101099967957, | |
| 0.9725697040557861, | |
| 0.9698337316513062, | |
| 0.9678789973258972, | |
| 0.4937111735343933, | |
| 0.4989865720272064, | |
| 0.47172561287879944, | |
| 0.9673305153846741, | |
| 0.974334180355072, | |
| 0.9758511781692505, | |
| 0.5213514566421509, | |
| 0.5089063048362732, | |
| 0.5075064301490784, | |
| 0.528186023235321, | |
| 0.5266287922859192, | |
| 0.5207107067108154, | |
| 0.6874252557754517, | |
| 0.5859951376914978, | |
| 0.5912247896194458, | |
| 0.4992046058177948, | |
| 0.49837902188301086, | |
| 0.48732590675354004, | |
| 0.5029973983764648, | |
| 0.5089282989501953, | |
| 0.5102050304412842, | |
| 0.49744459986686707, | |
| 0.5039994120597839, | |
| 0.5496022701263428, | |
| 0.4445340037345886, | |
| 0.4402852952480316, | |
| 0.4423198997974396, | |
| 0.5244624018669128, | |
| 0.5203747749328613, | |
| 0.510670006275177, | |
| 0.5222905874252319, | |
| 0.49989885091781616, | |
| 0.5032917261123657, | |
| 0.5169360041618347, | |
| 0.4835692048072815, | |
| 0.49604979157447815, | |
| 0.48959624767303467, | |
| 0.474219411611557, | |
| 0.39609599113464355, | |
| 0.965065598487854, | |
| 0.9651297926902771, | |
| 0.9658015966415405, | |
| 0.4790042042732239, | |
| 0.5015706419944763, | |
| 0.5015318393707275, | |
| 0.9697352051734924, | |
| 0.969887912273407, | |
| 0.9698120951652527, | |
| 0.5096650123596191, | |
| 0.515565037727356, | |
| 0.50871741771698, | |
| 0.5117788910865784, | |
| 0.47279125452041626, | |
| 0.5129478573799133, | |
| 0.513483464717865, | |
| 0.45352092385292053, | |
| 0.4934040606021881, | |
| 0.5126722455024719, | |
| 0.5389389991760254, | |
| 0.517637312412262, | |
| 0.535071849822998, | |
| 0.5183783769607544, | |
| 0.5209357142448425, | |
| 0.501355767250061, | |
| 0.49917733669281006, | |
| 0.4997139871120453, | |
| 0.4962727725505829, | |
| 0.4922448694705963, | |
| 0.49315178394317627, | |
| 0.5435120463371277, | |
| 0.5534088611602783, | |
| 0.5032275319099426, | |
| 0.4997408390045166, | |
| 0.5043116807937622, | |
| 0.5037261843681335, | |
| 0.5321357846260071, | |
| 0.5255170464515686, | |
| 0.528411328792572, | |
| 0.5442461371421814, | |
| 0.5378051996231079, | |
| 0.5345613956451416, | |
| 0.5022143125534058, | |
| 0.5025732517242432, | |
| 0.5056655406951904, | |
| 0.497100830078125 | |
| ], | |
| "eval_code_hard_runtime": 53.2886, | |
| "eval_code_hard_samples_per_second": 1.877, | |
| "eval_code_hard_steps_per_second": 0.038, | |
| "epoch": 0.44, | |
| "step": 25 | |
| }, | |
| { | |
| "loss": 0.537, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.46, | |
| "step": 26 | |
| }, | |
| { | |
| "loss": 0.56, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.47, | |
| "step": 27 | |
| }, | |
| { | |
| "loss": 0.6412, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.49, | |
| "step": 28 | |
| }, | |
| { | |
| "loss": 0.5359, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.51, | |
| "step": 29 | |
| }, | |
| { | |
| "loss": 0.5129, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.53, | |
| "step": 30 | |
| }, | |
| { | |
| "loss": 0.5104, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.54, | |
| "step": 31 | |
| }, | |
| { | |
| "loss": 0.4987, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.56, | |
| "step": 32 | |
| }, | |
| { | |
| "loss": 0.4875, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.58, | |
| "step": 33 | |
| }, | |
| { | |
| "loss": 0.4075, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.6, | |
| "step": 34 | |
| }, | |
| { | |
| "loss": 0.4638, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.61, | |
| "step": 35 | |
| }, | |
| { | |
| "loss": 0.4016, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.63, | |
| "step": 36 | |
| }, | |
| { | |
| "loss": 0.5376, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.65, | |
| "step": 37 | |
| }, | |
| { | |
| "loss": 0.5298, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.67, | |
| "step": 38 | |
| }, | |
| { | |
| "loss": 0.4862, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.68, | |
| "step": 39 | |
| }, | |
| { | |
| "loss": 0.5096, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.7, | |
| "step": 40 | |
| }, | |
| { | |
| "loss": 0.4659, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.72, | |
| "step": 41 | |
| }, | |
| { | |
| "loss": 0.4683, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.74, | |
| "step": 42 | |
| }, | |
| { | |
| "loss": 0.427, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.75, | |
| "step": 43 | |
| }, | |
| { | |
| "loss": 0.4062, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.77, | |
| "step": 44 | |
| }, | |
| { | |
| "loss": 0.4772, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.79, | |
| "step": 45 | |
| }, | |
| { | |
| "loss": 0.3566, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.81, | |
| "step": 46 | |
| }, | |
| { | |
| "loss": 0.4675, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.82, | |
| "step": 47 | |
| }, | |
| { | |
| "loss": 0.4146, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.84, | |
| "step": 48 | |
| }, | |
| { | |
| "loss": 0.4207, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.86, | |
| "step": 49 | |
| }, | |
| { | |
| "loss": 0.3584, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.88, | |
| "step": 50 | |
| }, | |
| { | |
| "eval_code_hard_loss": 0.34301868081092834, | |
| "eval_code_hard_score": -0.11159273982048035, | |
| "eval_code_hard_brier_score": 0.11159273982048035, | |
| "eval_code_hard_average_probability": 0.7284997701644897, | |
| "eval_code_hard_accuracy": 0.86, | |
| "eval_code_hard_probabilities": [ | |
| 0.619510293006897, | |
| 0.6040221452713013, | |
| 0.5958019495010376, | |
| 0.9492889046669006, | |
| 0.9177025556564331, | |
| 0.9289835095405579, | |
| 0.8927240371704102, | |
| 0.9211738705635071, | |
| 0.9168058037757874, | |
| 0.7706602215766907, | |
| 0.7466457486152649, | |
| 0.8367064595222473, | |
| 0.9997155070304871, | |
| 0.9995123147964478, | |
| 0.9982605576515198, | |
| 0.4536142945289612, | |
| 0.5211822986602783, | |
| 0.4859596788883209, | |
| 0.9396563172340393, | |
| 0.9814699292182922, | |
| 0.9855539202690125, | |
| 0.9185433387756348, | |
| 0.9342010617256165, | |
| 0.9058659076690674, | |
| 0.8620021939277649, | |
| 0.8177945017814636, | |
| 0.9235488176345825, | |
| 0.9499996304512024, | |
| 0.9158856868743896, | |
| 0.9167284369468689, | |
| 0.4879337251186371, | |
| 0.43694543838500977, | |
| 0.4652228355407715, | |
| 0.896939218044281, | |
| 0.8988721966743469, | |
| 0.9008588790893555, | |
| 0.6782506108283997, | |
| 0.6183739304542542, | |
| 0.7814574837684631, | |
| 0.740852952003479, | |
| 0.6149900555610657, | |
| 0.6410658359527588, | |
| 0.828216552734375, | |
| 0.7742620706558228, | |
| 0.7144817113876343, | |
| 0.5286781191825867, | |
| 0.4971938729286194, | |
| 0.506715714931488, | |
| 0.7012039422988892, | |
| 0.3713959753513336, | |
| 0.4451392889022827, | |
| 0.5213230848312378, | |
| 0.495220422744751, | |
| 0.47032222151756287, | |
| 0.995394766330719, | |
| 0.9957503080368042, | |
| 0.9961684346199036, | |
| 0.39281144738197327, | |
| 0.5100150108337402, | |
| 0.47976943850517273, | |
| 0.9990038275718689, | |
| 0.9989190101623535, | |
| 0.9989215135574341, | |
| 0.5996841192245483, | |
| 0.5701720118522644, | |
| 0.58901047706604, | |
| 0.9164248108863831, | |
| 0.9526078104972839, | |
| 0.9073719382286072, | |
| 0.7042233943939209, | |
| 0.5729785561561584, | |
| 0.5182032585144043, | |
| 0.8634644150733948, | |
| 0.9283011555671692, | |
| 0.9007238745689392, | |
| 0.7439146637916565, | |
| 0.6496825814247131, | |
| 0.6525353789329529, | |
| 0.5044839382171631, | |
| 0.4995957612991333, | |
| 0.5234301090240479, | |
| 0.5146706104278564, | |
| 0.6355378031730652, | |
| 0.5025534629821777, | |
| 0.9130427837371826, | |
| 0.8997641205787659, | |
| 0.5774441957473755, | |
| 0.519719123840332, | |
| 0.5050529837608337, | |
| 0.5099009275436401, | |
| 0.8784401416778564, | |
| 0.886016845703125, | |
| 0.8805269598960876, | |
| 0.8014693856239319, | |
| 0.7851292490959167, | |
| 0.8030146956443787, | |
| 0.5338838696479797, | |
| 0.5891290903091431, | |
| 0.3894357979297638, | |
| 0.5362494587898254 | |
| ], | |
| "eval_code_hard_runtime": 53.3003, | |
| "eval_code_hard_samples_per_second": 1.876, | |
| "eval_code_hard_steps_per_second": 0.038, | |
| "epoch": 0.88, | |
| "step": 50 | |
| }, | |
| { | |
| "loss": 0.4189, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.89, | |
| "step": 51 | |
| }, | |
| { | |
| "loss": 0.3505, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.91, | |
| "step": 52 | |
| }, | |
| { | |
| "loss": 0.3945, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.93, | |
| "step": 53 | |
| }, | |
| { | |
| "loss": 0.4302, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.95, | |
| "step": 54 | |
| }, | |
| { | |
| "loss": 0.3197, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.96, | |
| "step": 55 | |
| }, | |
| { | |
| "loss": 0.3941, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 0.98, | |
| "step": 56 | |
| }, | |
| { | |
| "loss": 0.2922, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.0, | |
| "step": 57 | |
| }, | |
| { | |
| "loss": 0.4051, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.02, | |
| "step": 58 | |
| }, | |
| { | |
| "loss": 0.3588, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.04, | |
| "step": 59 | |
| }, | |
| { | |
| "loss": 0.3388, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.05, | |
| "step": 60 | |
| }, | |
| { | |
| "loss": 0.3442, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.07, | |
| "step": 61 | |
| }, | |
| { | |
| "loss": 0.2811, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.09, | |
| "step": 62 | |
| }, | |
| { | |
| "loss": 0.2847, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.11, | |
| "step": 63 | |
| }, | |
| { | |
| "loss": 0.3596, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.12, | |
| "step": 64 | |
| }, | |
| { | |
| "loss": 0.4253, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.14, | |
| "step": 65 | |
| }, | |
| { | |
| "loss": 0.2929, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.16, | |
| "step": 66 | |
| }, | |
| { | |
| "loss": 0.2756, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.18, | |
| "step": 67 | |
| }, | |
| { | |
| "loss": 0.3144, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.19, | |
| "step": 68 | |
| }, | |
| { | |
| "loss": 0.2823, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.21, | |
| "step": 69 | |
| }, | |
| { | |
| "loss": 0.2129, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.23, | |
| "step": 70 | |
| }, | |
| { | |
| "loss": 0.2722, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.25, | |
| "step": 71 | |
| }, | |
| { | |
| "loss": 0.2964, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.26, | |
| "step": 72 | |
| }, | |
| { | |
| "loss": 0.3038, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.28, | |
| "step": 73 | |
| }, | |
| { | |
| "loss": 0.2398, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.3, | |
| "step": 74 | |
| }, | |
| { | |
| "loss": 0.2923, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.32, | |
| "step": 75 | |
| }, | |
| { | |
| "eval_code_hard_loss": 0.24489852786064148, | |
| "eval_code_hard_score": -0.08631329238414764, | |
| "eval_code_hard_brier_score": 0.08631329238414764, | |
| "eval_code_hard_average_probability": 0.8100240230560303, | |
| "eval_code_hard_accuracy": 0.85, | |
| "eval_code_hard_probabilities": [ | |
| 0.7177616953849792, | |
| 0.795508623123169, | |
| 0.7857334613800049, | |
| 0.9991870522499084, | |
| 0.9987917542457581, | |
| 0.9994052648544312, | |
| 0.9846364259719849, | |
| 0.9944690465927124, | |
| 0.9703215956687927, | |
| 0.778681755065918, | |
| 0.971013069152832, | |
| 0.9788711071014404, | |
| 0.9999998807907104, | |
| 1.0, | |
| 1.0, | |
| 0.35785171389579773, | |
| 0.4882768392562866, | |
| 0.6737111806869507, | |
| 0.9996606111526489, | |
| 0.9999253749847412, | |
| 0.9999637603759766, | |
| 0.9994186162948608, | |
| 0.9990422129631042, | |
| 0.9974905252456665, | |
| 0.9995825886726379, | |
| 0.9953965544700623, | |
| 0.9997794032096863, | |
| 0.9983587861061096, | |
| 0.9997310042381287, | |
| 0.9997231364250183, | |
| 0.4219929277896881, | |
| 0.2606010437011719, | |
| 0.6489297151565552, | |
| 0.9547426104545593, | |
| 0.9553850889205933, | |
| 0.950897216796875, | |
| 0.9628025889396667, | |
| 0.9581644535064697, | |
| 0.8295655846595764, | |
| 0.9866873621940613, | |
| 0.9280973672866821, | |
| 0.9797408580780029, | |
| 0.9973568916320801, | |
| 0.9953936338424683, | |
| 0.9935207962989807, | |
| 0.5510905385017395, | |
| 0.49630528688430786, | |
| 0.4779393970966339, | |
| 0.8343468308448792, | |
| 0.43401989340782166, | |
| 0.3891345262527466, | |
| 0.4458746612071991, | |
| 0.6388282179832458, | |
| 0.8165429830551147, | |
| 0.9992231130599976, | |
| 0.9993274211883545, | |
| 0.9994113445281982, | |
| 0.41094741225242615, | |
| 0.5163617134094238, | |
| 0.488020658493042, | |
| 0.999996542930603, | |
| 0.9999948740005493, | |
| 0.9999948740005493, | |
| 0.6086482405662537, | |
| 0.6147478222846985, | |
| 0.6657524108886719, | |
| 0.9949613809585571, | |
| 0.9949374198913574, | |
| 0.9945763349533081, | |
| 0.7646587491035461, | |
| 0.817954957485199, | |
| 0.6410450339317322, | |
| 0.9997159838676453, | |
| 0.9999265670776367, | |
| 0.9998390674591064, | |
| 0.9317758083343506, | |
| 0.8353602290153503, | |
| 0.8241412043571472, | |
| 0.501185417175293, | |
| 0.4842207431793213, | |
| 0.6102815270423889, | |
| 0.4974021911621094, | |
| 0.6316254138946533, | |
| 0.4609917998313904, | |
| 0.9999159574508667, | |
| 0.999816358089447, | |
| 0.6341696381568909, | |
| 0.5534595847129822, | |
| 0.5341170430183411, | |
| 0.5336964130401611, | |
| 0.9986201524734497, | |
| 0.9990990161895752, | |
| 0.9992678761482239, | |
| 0.9444563984870911, | |
| 0.9397106170654297, | |
| 0.9423726201057434, | |
| 0.6565976142883301, | |
| 0.8141821026802063, | |
| 0.16233490407466888, | |
| 0.6133077144622803 | |
| ], | |
| "eval_code_hard_runtime": 53.2911, | |
| "eval_code_hard_samples_per_second": 1.876, | |
| "eval_code_hard_steps_per_second": 0.038, | |
| "epoch": 1.32, | |
| "step": 75 | |
| }, | |
| { | |
| "loss": 0.3058, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.33, | |
| "step": 76 | |
| }, | |
| { | |
| "loss": 0.3695, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.35, | |
| "step": 77 | |
| }, | |
| { | |
| "loss": 0.1812, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.37, | |
| "step": 78 | |
| }, | |
| { | |
| "loss": 0.2968, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.39, | |
| "step": 79 | |
| }, | |
| { | |
| "loss": 0.3167, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.4, | |
| "step": 80 | |
| }, | |
| { | |
| "loss": 0.488, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.42, | |
| "step": 81 | |
| }, | |
| { | |
| "loss": 0.3406, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.44, | |
| "step": 82 | |
| }, | |
| { | |
| "loss": 0.3298, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.46, | |
| "step": 83 | |
| }, | |
| { | |
| "loss": 0.3004, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.47, | |
| "step": 84 | |
| }, | |
| { | |
| "loss": 0.1752, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.49, | |
| "step": 85 | |
| }, | |
| { | |
| "loss": 0.2629, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.51, | |
| "step": 86 | |
| }, | |
| { | |
| "loss": 0.2125, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.53, | |
| "step": 87 | |
| }, | |
| { | |
| "loss": 0.2029, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.54, | |
| "step": 88 | |
| }, | |
| { | |
| "loss": 0.2783, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.56, | |
| "step": 89 | |
| }, | |
| { | |
| "loss": 0.207, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.58, | |
| "step": 90 | |
| }, | |
| { | |
| "loss": 0.2904, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.6, | |
| "step": 91 | |
| }, | |
| { | |
| "loss": 0.1934, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.61, | |
| "step": 92 | |
| }, | |
| { | |
| "loss": 0.3917, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.63, | |
| "step": 93 | |
| }, | |
| { | |
| "loss": 0.194, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.65, | |
| "step": 94 | |
| }, | |
| { | |
| "loss": 0.2205, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.67, | |
| "step": 95 | |
| }, | |
| { | |
| "loss": 0.1669, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.68, | |
| "step": 96 | |
| }, | |
| { | |
| "loss": 0.2966, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.7, | |
| "step": 97 | |
| }, | |
| { | |
| "loss": 0.2516, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.72, | |
| "step": 98 | |
| }, | |
| { | |
| "loss": 0.2276, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.74, | |
| "step": 99 | |
| }, | |
| { | |
| "loss": 0.1814, | |
| "learning_rate": 7.2e-05, | |
| "epoch": 1.75, | |
| "step": 100 | |
| }, | |
| { | |
| "eval_code_hard_loss": 0.2296922653913498, | |
| "eval_code_hard_score": -0.08056112378835678, | |
| "eval_code_hard_brier_score": 0.08056112378835678, | |
| "eval_code_hard_average_probability": 0.8206403255462646, | |
| "eval_code_hard_accuracy": 0.84, | |
| "eval_code_hard_probabilities": [ | |
| 0.7861270308494568, | |
| 0.8348590731620789, | |
| 0.8349205851554871, | |
| 0.9997000694274902, | |
| 0.9995880722999573, | |
| 0.9998056292533875, | |
| 0.9992101192474365, | |
| 0.9999064207077026, | |
| 0.9989945292472839, | |
| 0.8230730891227722, | |
| 0.9760466814041138, | |
| 0.9709921479225159, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.460991233587265, | |
| 0.6684326529502869, | |
| 0.957171618938446, | |
| 0.999259889125824, | |
| 0.9999682903289795, | |
| 0.9999852180480957, | |
| 0.9999778270721436, | |
| 0.9999755620956421, | |
| 0.9997803568840027, | |
| 0.9999879598617554, | |
| 0.9971315860748291, | |
| 0.9999914169311523, | |
| 0.999528169631958, | |
| 0.9998747110366821, | |
| 0.9998290538787842, | |
| 0.4468780755996704, | |
| 0.2516433894634247, | |
| 0.6790833473205566, | |
| 0.9911589026451111, | |
| 0.9917107820510864, | |
| 0.9915744662284851, | |
| 0.9812775254249573, | |
| 0.9739009141921997, | |
| 0.7593258023262024, | |
| 0.9627566933631897, | |
| 0.8079734444618225, | |
| 0.943427562713623, | |
| 0.9999816417694092, | |
| 0.9999567270278931, | |
| 0.9999591112136841, | |
| 0.5994269847869873, | |
| 0.48037776350975037, | |
| 0.4851032495498657, | |
| 0.9079890847206116, | |
| 0.4937061071395874, | |
| 0.4938817620277405, | |
| 0.4849710762500763, | |
| 0.7558672428131104, | |
| 0.8851609230041504, | |
| 0.9924831986427307, | |
| 0.9915366172790527, | |
| 0.9959695339202881, | |
| 0.4088115990161896, | |
| 0.5256107449531555, | |
| 0.4217263460159302, | |
| 0.999980092048645, | |
| 0.999962568283081, | |
| 0.9999624490737915, | |
| 0.7795777916908264, | |
| 0.6225330829620361, | |
| 0.70358806848526, | |
| 0.9998552799224854, | |
| 0.9998051524162292, | |
| 0.9998724460601807, | |
| 0.40740594267845154, | |
| 0.7388888597488403, | |
| 0.4775027632713318, | |
| 0.9999308586120605, | |
| 0.9999823570251465, | |
| 0.9999585151672363, | |
| 0.9347867369651794, | |
| 0.5755781531333923, | |
| 0.6215885877609253, | |
| 0.5002992153167725, | |
| 0.5065785050392151, | |
| 0.699110209941864, | |
| 0.472739577293396, | |
| 0.6727705597877502, | |
| 0.4803294241428375, | |
| 0.9999979734420776, | |
| 0.9999915361404419, | |
| 0.8218505382537842, | |
| 0.5483625531196594, | |
| 0.5385692119598389, | |
| 0.49461352825164795, | |
| 0.9993742108345032, | |
| 0.99980229139328, | |
| 0.9998019337654114, | |
| 0.993417501449585, | |
| 0.9916191101074219, | |
| 0.9953900575637817, | |
| 0.7289006114006042, | |
| 0.8122584223747253, | |
| 0.2769394814968109, | |
| 0.6622205972671509 | |
| ], | |
| "eval_code_hard_runtime": 53.2227, | |
| "eval_code_hard_samples_per_second": 1.879, | |
| "eval_code_hard_steps_per_second": 0.038, | |
| "epoch": 1.75, | |
| "step": 100 | |
| }, | |
| { | |
| "train_runtime": 4004.592, | |
| "train_samples_per_second": 0.799, | |
| "train_steps_per_second": 0.025, | |
| "total_flos": 0.0, | |
| "train_loss": 0.417534539103508, | |
| "epoch": 1.75, | |
| "step": 100 | |
| } | |
| ]] |