Invalid JSON: Unexpected non-whitespace character after JSONat line 1595, column 2
| [ | |
| { | |
| "loss": 0.7559, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.02, | |
| "step": 1 | |
| }, | |
| { | |
| "loss": 0.809, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.04, | |
| "step": 2 | |
| }, | |
| { | |
| "loss": 0.6615, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.05, | |
| "step": 3 | |
| }, | |
| { | |
| "loss": 0.6488, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.07, | |
| "step": 4 | |
| }, | |
| { | |
| "loss": 0.6381, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.09, | |
| "step": 5 | |
| }, | |
| { | |
| "loss": 0.5315, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.11, | |
| "step": 6 | |
| }, | |
| { | |
| "loss": 0.5713, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.12, | |
| "step": 7 | |
| }, | |
| { | |
| "loss": 0.5663, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.14, | |
| "step": 8 | |
| }, | |
| { | |
| "loss": 0.6267, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.16, | |
| "step": 9 | |
| }, | |
| { | |
| "loss": 0.5685, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.18, | |
| "step": 10 | |
| }, | |
| { | |
| "loss": 0.501, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.19, | |
| "step": 11 | |
| }, | |
| { | |
| "loss": 0.5033, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.21, | |
| "step": 12 | |
| }, | |
| { | |
| "loss": 0.5903, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.23, | |
| "step": 13 | |
| }, | |
| { | |
| "loss": 0.4598, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.25, | |
| "step": 14 | |
| }, | |
| { | |
| "loss": 0.4332, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.26, | |
| "step": 15 | |
| }, | |
| { | |
| "loss": 0.4749, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.28, | |
| "step": 16 | |
| }, | |
| { | |
| "loss": 0.3193, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.3, | |
| "step": 17 | |
| }, | |
| { | |
| "loss": 0.4061, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.32, | |
| "step": 18 | |
| }, | |
| { | |
| "loss": 0.3541, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.33, | |
| "step": 19 | |
| }, | |
| { | |
| "loss": 0.3414, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.35, | |
| "step": 20 | |
| }, | |
| { | |
| "loss": 0.3453, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.37, | |
| "step": 21 | |
| }, | |
| { | |
| "loss": 0.5322, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.39, | |
| "step": 22 | |
| }, | |
| { | |
| "loss": 0.3092, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.4, | |
| "step": 23 | |
| }, | |
| { | |
| "loss": 0.2027, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.42, | |
| "step": 24 | |
| }, | |
| { | |
| "loss": 0.403, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.44, | |
| "step": 25 | |
| }, | |
| { | |
| "eval_code_hard_loss": 0.24356038868427277, | |
| "eval_code_hard_score": -0.08314146846532822, | |
| "eval_code_hard_brier_score": 0.08314146846532822, | |
| "eval_code_hard_average_probability": 0.8104196786880493, | |
| "eval_code_hard_accuracy": 0.88, | |
| "eval_code_hard_probabilities": [ | |
| 0.7802631855010986, | |
| 0.5728839039802551, | |
| 0.601148784160614, | |
| 0.9993261098861694, | |
| 0.9960917830467224, | |
| 0.9883104562759399, | |
| 0.9998921155929565, | |
| 0.9998250603675842, | |
| 0.9999023675918579, | |
| 0.7125715017318726, | |
| 0.9519601464271545, | |
| 0.8720795512199402, | |
| 1.0, | |
| 0.9999998807907104, | |
| 0.9999998807907104, | |
| 0.9911274313926697, | |
| 0.6485933065414429, | |
| 0.831139326095581, | |
| 0.9780414700508118, | |
| 0.9999822378158569, | |
| 0.9999316930770874, | |
| 0.9999449253082275, | |
| 0.9999632835388184, | |
| 0.9999346733093262, | |
| 0.9998100399971008, | |
| 0.9998506307601929, | |
| 0.999704897403717, | |
| 0.9949617385864258, | |
| 0.9998635053634644, | |
| 0.9998619556427002, | |
| 0.6420595645904541, | |
| 0.7948806881904602, | |
| 0.6511540412902832, | |
| 0.9534466862678528, | |
| 0.9455068707466125, | |
| 0.944064736366272, | |
| 0.9278716444969177, | |
| 0.8786247968673706, | |
| 0.9867132902145386, | |
| 0.9512923359870911, | |
| 0.975769579410553, | |
| 0.9804350137710571, | |
| 0.9998563528060913, | |
| 0.9999313354492188, | |
| 0.9999279975891113, | |
| 0.8713787198066711, | |
| 0.5257866382598877, | |
| 0.5062138438224792, | |
| 0.7719318270683289, | |
| 0.42283594608306885, | |
| 0.43223893642425537, | |
| 0.5629215836524963, | |
| 0.9509612321853638, | |
| 0.6529775857925415, | |
| 0.998943030834198, | |
| 0.9988754391670227, | |
| 0.9993382096290588, | |
| 0.4989044666290283, | |
| 0.5084826946258545, | |
| 0.5146058797836304, | |
| 0.9999768733978271, | |
| 0.9999709129333496, | |
| 0.9999724626541138, | |
| 0.6513444781303406, | |
| 0.5117835402488708, | |
| 0.8250595331192017, | |
| 0.9398683309555054, | |
| 0.9772916436195374, | |
| 0.9429591298103333, | |
| 0.9575245380401611, | |
| 0.8619183301925659, | |
| 0.5832539796829224, | |
| 0.9970285296440125, | |
| 0.999936580657959, | |
| 0.9999572038650513, | |
| 0.9991602897644043, | |
| 0.8450718522071838, | |
| 0.7737744450569153, | |
| 0.4931277930736542, | |
| 0.483172208070755, | |
| 0.5014657974243164, | |
| 0.5353665947914124, | |
| 0.434240460395813, | |
| 0.516578197479248, | |
| 0.9958915710449219, | |
| 0.9994507431983948, | |
| 0.5593408942222595, | |
| 0.4825429320335388, | |
| 0.4861961901187897, | |
| 0.49605879187583923, | |
| 0.9998632669448853, | |
| 0.9998589754104614, | |
| 0.9998570680618286, | |
| 0.39110851287841797, | |
| 0.38484665751457214, | |
| 0.3077014684677124, | |
| 0.5388085246086121, | |
| 0.6023823618888855, | |
| 0.5493093729019165, | |
| 0.6540540456771851 | |
| ], | |
| "eval_code_hard_runtime": 110.6525, | |
| "eval_code_hard_samples_per_second": 0.904, | |
| "eval_code_hard_steps_per_second": 0.036, | |
| "epoch": 0.44, | |
| "step": 25 | |
| }, | |
| { | |
| "loss": 0.2619, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.46, | |
| "step": 26 | |
| }, | |
| { | |
| "loss": 0.2734, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.47, | |
| "step": 27 | |
| }, | |
| { | |
| "loss": 0.4004, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.49, | |
| "step": 28 | |
| }, | |
| { | |
| "loss": 0.1422, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.51, | |
| "step": 29 | |
| }, | |
| { | |
| "loss": 0.2156, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.53, | |
| "step": 30 | |
| }, | |
| { | |
| "loss": 0.2327, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.54, | |
| "step": 31 | |
| }, | |
| { | |
| "loss": 0.1781, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.56, | |
| "step": 32 | |
| }, | |
| { | |
| "loss": 0.2628, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.58, | |
| "step": 33 | |
| }, | |
| { | |
| "loss": 0.3038, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.6, | |
| "step": 34 | |
| }, | |
| { | |
| "loss": 0.2291, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.61, | |
| "step": 35 | |
| }, | |
| { | |
| "loss": 0.1074, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.63, | |
| "step": 36 | |
| }, | |
| { | |
| "loss": 0.2432, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.65, | |
| "step": 37 | |
| }, | |
| { | |
| "loss": 0.1269, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.67, | |
| "step": 38 | |
| }, | |
| { | |
| "loss": 0.2828, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.68, | |
| "step": 39 | |
| }, | |
| { | |
| "loss": 0.2679, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.7, | |
| "step": 40 | |
| }, | |
| { | |
| "loss": 0.1949, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.72, | |
| "step": 41 | |
| }, | |
| { | |
| "loss": 0.2657, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.74, | |
| "step": 42 | |
| }, | |
| { | |
| "loss": 0.1775, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.75, | |
| "step": 43 | |
| }, | |
| { | |
| "loss": 0.2224, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.77, | |
| "step": 44 | |
| }, | |
| { | |
| "loss": 0.2793, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.79, | |
| "step": 45 | |
| }, | |
| { | |
| "loss": 0.0357, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.81, | |
| "step": 46 | |
| }, | |
| { | |
| "loss": 0.1634, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.82, | |
| "step": 47 | |
| }, | |
| { | |
| "loss": 0.2606, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.84, | |
| "step": 48 | |
| }, | |
| { | |
| "loss": 0.2214, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.86, | |
| "step": 49 | |
| }, | |
| { | |
| "loss": 0.0889, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.88, | |
| "step": 50 | |
| }, | |
| { | |
| "eval_code_hard_loss": 0.16699519753456116, | |
| "eval_code_hard_score": -0.05673101916909218, | |
| "eval_code_hard_brier_score": 0.05673101916909218, | |
| "eval_code_hard_average_probability": 0.8761882781982422, | |
| "eval_code_hard_accuracy": 0.93, | |
| "eval_code_hard_probabilities": [ | |
| 0.9874073266983032, | |
| 0.7980256080627441, | |
| 0.9544906616210938, | |
| 0.9999760389328003, | |
| 0.9833183288574219, | |
| 0.9734002947807312, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.8710820078849792, | |
| 0.9463728070259094, | |
| 0.9430120587348938, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9999483823776245, | |
| 0.9996401071548462, | |
| 0.9999982118606567, | |
| 0.6661297082901001, | |
| 0.9998952150344849, | |
| 0.9999295473098755, | |
| 0.9999996423721313, | |
| 1.0, | |
| 0.9999994039535522, | |
| 0.9999994039535522, | |
| 0.9999171495437622, | |
| 0.9984815716743469, | |
| 0.9999889135360718, | |
| 0.9999997615814209, | |
| 0.9999967813491821, | |
| 0.997412383556366, | |
| 0.9992191791534424, | |
| 0.9149346947669983, | |
| 0.9999977350234985, | |
| 0.9999892711639404, | |
| 0.9999914169311523, | |
| 1.0, | |
| 0.9999998807907104, | |
| 1.0, | |
| 0.9768458604812622, | |
| 0.9713558554649353, | |
| 0.9916350841522217, | |
| 0.9999990463256836, | |
| 1.0, | |
| 1.0, | |
| 0.7556970119476318, | |
| 0.5051810145378113, | |
| 0.443418949842453, | |
| 0.9936054944992065, | |
| 0.3165128529071808, | |
| 0.3557163178920746, | |
| 0.5657868385314941, | |
| 0.9981420040130615, | |
| 0.5603527426719666, | |
| 0.9821581840515137, | |
| 0.9779713749885559, | |
| 0.9982566237449646, | |
| 0.5955212712287903, | |
| 0.5466123819351196, | |
| 0.5278903245925903, | |
| 0.999467670917511, | |
| 0.9992781281471252, | |
| 0.9993084669113159, | |
| 0.9714255332946777, | |
| 0.5806359648704529, | |
| 0.9231281876564026, | |
| 0.9999953508377075, | |
| 1.0, | |
| 0.9999998807907104, | |
| 0.996260941028595, | |
| 0.9989031553268433, | |
| 0.9639995098114014, | |
| 0.9999957084655762, | |
| 1.0, | |
| 0.9999998807907104, | |
| 0.9980409741401672, | |
| 0.8687159419059753, | |
| 0.821354866027832, | |
| 0.4796599745750427, | |
| 0.4628332555294037, | |
| 0.5007415413856506, | |
| 0.3400116562843323, | |
| 0.6277292966842651, | |
| 0.5096919536590576, | |
| 0.9999779462814331, | |
| 0.9994074106216431, | |
| 0.596235454082489, | |
| 0.5236307382583618, | |
| 0.44950583577156067, | |
| 0.5437645316123962, | |
| 0.9999998807907104, | |
| 0.9999997615814209, | |
| 0.9999998807907104, | |
| 0.9999169111251831, | |
| 0.9999459981918335, | |
| 0.9999337196350098, | |
| 0.9902818202972412, | |
| 0.9997207522392273, | |
| 0.5093064308166504, | |
| 0.870800793170929 | |
| ], | |
| "eval_code_hard_runtime": 110.6115, | |
| "eval_code_hard_samples_per_second": 0.904, | |
| "eval_code_hard_steps_per_second": 0.036, | |
| "epoch": 0.88, | |
| "step": 50 | |
| }, | |
| { | |
| "loss": 0.3248, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.89, | |
| "step": 51 | |
| }, | |
| { | |
| "loss": 0.1904, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.91, | |
| "step": 52 | |
| }, | |
| { | |
| "loss": 0.185, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.93, | |
| "step": 53 | |
| }, | |
| { | |
| "loss": 0.0932, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.95, | |
| "step": 54 | |
| }, | |
| { | |
| "loss": 0.1643, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.96, | |
| "step": 55 | |
| }, | |
| { | |
| "loss": 0.2504, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.98, | |
| "step": 56 | |
| }, | |
| { | |
| "loss": 0.17, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.0, | |
| "step": 57 | |
| }, | |
| { | |
| "loss": 0.1939, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.02, | |
| "step": 58 | |
| }, | |
| { | |
| "loss": 0.1383, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.04, | |
| "step": 59 | |
| }, | |
| { | |
| "loss": 0.149, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.05, | |
| "step": 60 | |
| }, | |
| { | |
| "loss": 0.0699, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.07, | |
| "step": 61 | |
| }, | |
| { | |
| "loss": 0.1258, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.09, | |
| "step": 62 | |
| }, | |
| { | |
| "loss": 0.0563, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.11, | |
| "step": 63 | |
| }, | |
| { | |
| "loss": 0.1171, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.12, | |
| "step": 64 | |
| }, | |
| { | |
| "loss": 0.1197, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.14, | |
| "step": 65 | |
| }, | |
| { | |
| "loss": 0.0812, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.16, | |
| "step": 66 | |
| }, | |
| { | |
| "loss": 0.1488, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.18, | |
| "step": 67 | |
| }, | |
| { | |
| "loss": 0.1459, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.19, | |
| "step": 68 | |
| }, | |
| { | |
| "loss": 0.0691, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.21, | |
| "step": 69 | |
| }, | |
| { | |
| "loss": 0.0529, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.23, | |
| "step": 70 | |
| }, | |
| { | |
| "loss": 0.0443, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.25, | |
| "step": 71 | |
| }, | |
| { | |
| "loss": 0.214, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.26, | |
| "step": 72 | |
| }, | |
| { | |
| "loss": 0.0409, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.28, | |
| "step": 73 | |
| }, | |
| { | |
| "loss": 0.0357, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.3, | |
| "step": 74 | |
| }, | |
| { | |
| "loss": 0.0676, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.32, | |
| "step": 75 | |
| }, | |
| { | |
| "eval_code_hard_loss": 0.15125365555286407, | |
| "eval_code_hard_score": -0.047643136233091354, | |
| "eval_code_hard_brier_score": 0.047643136233091354, | |
| "eval_code_hard_average_probability": 0.9072666764259338, | |
| "eval_code_hard_accuracy": 0.94, | |
| "eval_code_hard_probabilities": [ | |
| 0.9881106615066528, | |
| 0.67057204246521, | |
| 0.9074558615684509, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9999914169311523, | |
| 0.9999996423721313, | |
| 0.9999991655349731, | |
| 0.8589997291564941, | |
| 0.9643713235855103, | |
| 0.9970460534095764, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9999948740005493, | |
| 0.9998992681503296, | |
| 0.9999996423721313, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9999997615814209, | |
| 1.0, | |
| 0.9999988079071045, | |
| 0.9999991655349731, | |
| 0.9999972581863403, | |
| 0.9980295300483704, | |
| 0.9999752044677734, | |
| 0.9991430044174194, | |
| 0.999230146408081, | |
| 0.9863813519477844, | |
| 0.9973239898681641, | |
| 0.8935987949371338, | |
| 1.0, | |
| 0.9999996423721313, | |
| 0.9999998807907104, | |
| 0.9999157190322876, | |
| 0.9997393488883972, | |
| 0.999843955039978, | |
| 0.8906758427619934, | |
| 0.7957176566123962, | |
| 0.8944621682167053, | |
| 0.9999850988388062, | |
| 0.9999994039535522, | |
| 0.9999984502792358, | |
| 0.8667024970054626, | |
| 0.5891015529632568, | |
| 0.5897384881973267, | |
| 0.9999762773513794, | |
| 0.920762836933136, | |
| 0.8804107904434204, | |
| 0.037825245410203934, | |
| 0.9987059831619263, | |
| 0.9998573064804077, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.5703369379043579, | |
| 0.6154165863990784, | |
| 0.47148749232292175, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9997634291648865, | |
| 0.9999960660934448, | |
| 0.9999247789382935, | |
| 0.9999982118606567, | |
| 0.9999983310699463, | |
| 0.7629109025001526, | |
| 0.9476748108863831, | |
| 0.7605304718017578, | |
| 0.999997615814209, | |
| 0.9999995231628418, | |
| 0.9999983310699463, | |
| 0.9999594688415527, | |
| 0.9989461302757263, | |
| 0.99898761510849, | |
| 0.5305535793304443, | |
| 0.7419759631156921, | |
| 0.9985262155532837, | |
| 0.9876091480255127, | |
| 0.9818170070648193, | |
| 0.8375343084335327, | |
| 0.9999997615814209, | |
| 0.9999995231628418, | |
| 0.07670325040817261, | |
| 0.6875349283218384, | |
| 0.35709840059280396, | |
| 0.3490428626537323, | |
| 0.99744713306427, | |
| 0.9997475743293762, | |
| 0.9991705417633057, | |
| 0.9985999464988708, | |
| 0.9992005228996277, | |
| 0.9987180233001709, | |
| 0.9907884001731873, | |
| 0.9931132197380066, | |
| 0.35269367694854736, | |
| 0.9993346333503723 | |
| ], | |
| "eval_code_hard_runtime": 110.7336, | |
| "eval_code_hard_samples_per_second": 0.903, | |
| "eval_code_hard_steps_per_second": 0.036, | |
| "epoch": 1.32, | |
| "step": 75 | |
| }, | |
| { | |
| "loss": 0.0635, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.33, | |
| "step": 76 | |
| }, | |
| { | |
| "loss": 0.0923, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.35, | |
| "step": 77 | |
| }, | |
| { | |
| "loss": 0.1026, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.37, | |
| "step": 78 | |
| }, | |
| { | |
| "loss": 0.0674, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.39, | |
| "step": 79 | |
| }, | |
| { | |
| "loss": 0.0493, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.4, | |
| "step": 80 | |
| }, | |
| { | |
| "loss": 0.1209, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.42, | |
| "step": 81 | |
| }, | |
| { | |
| "loss": 0.5175, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.44, | |
| "step": 82 | |
| }, | |
| { | |
| "loss": 0.1547, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.46, | |
| "step": 83 | |
| }, | |
| { | |
| "loss": 0.1185, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.47, | |
| "step": 84 | |
| }, | |
| { | |
| "loss": 0.0621, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.49, | |
| "step": 85 | |
| }, | |
| { | |
| "loss": 0.07, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.51, | |
| "step": 86 | |
| }, | |
| { | |
| "loss": 0.092, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.53, | |
| "step": 87 | |
| }, | |
| { | |
| "loss": 0.0422, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.54, | |
| "step": 88 | |
| }, | |
| { | |
| "loss": 0.1117, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.56, | |
| "step": 89 | |
| }, | |
| { | |
| "loss": 0.0654, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.58, | |
| "step": 90 | |
| }, | |
| { | |
| "loss": 0.1185, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.6, | |
| "step": 91 | |
| }, | |
| { | |
| "loss": 0.1308, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.61, | |
| "step": 92 | |
| }, | |
| { | |
| "loss": 0.1293, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.63, | |
| "step": 93 | |
| }, | |
| { | |
| "loss": 0.1071, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.65, | |
| "step": 94 | |
| }, | |
| { | |
| "loss": 0.0593, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.67, | |
| "step": 95 | |
| }, | |
| { | |
| "loss": 0.0494, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.68, | |
| "step": 96 | |
| }, | |
| { | |
| "loss": 0.0641, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.7, | |
| "step": 97 | |
| }, | |
| { | |
| "loss": 0.0815, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.72, | |
| "step": 98 | |
| }, | |
| { | |
| "loss": 0.0747, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.74, | |
| "step": 99 | |
| }, | |
| { | |
| "loss": 0.0776, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.75, | |
| "step": 100 | |
| }, | |
| { | |
| "eval_code_hard_loss": 0.09847890585660934, | |
| "eval_code_hard_score": -0.035213593393564224, | |
| "eval_code_hard_brier_score": 0.035213593393564224, | |
| "eval_code_hard_average_probability": 0.9248704314231873, | |
| "eval_code_hard_accuracy": 0.94, | |
| "eval_code_hard_probabilities": [ | |
| 0.9995959401130676, | |
| 0.823664665222168, | |
| 0.9978653788566589, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.7222961187362671, | |
| 0.9800193309783936, | |
| 0.9987350106239319, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.999994158744812, | |
| 1.0, | |
| 0.303742915391922, | |
| 0.9999943971633911, | |
| 0.9999933242797852, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9999995231628418, | |
| 0.9999909400939941, | |
| 0.9999997615814209, | |
| 1.0, | |
| 1.0, | |
| 0.999997615814209, | |
| 1.0, | |
| 0.9949355721473694, | |
| 1.0, | |
| 0.9999887943267822, | |
| 0.999993085861206, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9999736547470093, | |
| 0.9997861981391907, | |
| 0.9999978542327881, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9989193677902222, | |
| 0.6210824847221375, | |
| 0.8116233944892883, | |
| 0.9997373223304749, | |
| 0.8275205492973328, | |
| 0.7434841394424438, | |
| 0.27361589670181274, | |
| 0.9993396401405334, | |
| 0.9998682737350464, | |
| 0.999985933303833, | |
| 0.9999828338623047, | |
| 0.9999997615814209, | |
| 0.39626461267471313, | |
| 0.5240136384963989, | |
| 0.6546356081962585, | |
| 0.9999992847442627, | |
| 0.9999991655349731, | |
| 0.9999990463256836, | |
| 0.9999998807907104, | |
| 0.9612408876419067, | |
| 0.9947011470794678, | |
| 0.9999994039535522, | |
| 1.0, | |
| 1.0, | |
| 0.9999735355377197, | |
| 0.9999927282333374, | |
| 0.9999716281890869, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9999995231628418, | |
| 0.9988247752189636, | |
| 0.9937974810600281, | |
| 0.49393144249916077, | |
| 0.4931151568889618, | |
| 0.7545955777168274, | |
| 0.918387770652771, | |
| 0.9984613656997681, | |
| 0.9579277634620667, | |
| 1.0, | |
| 0.9999995231628418, | |
| 0.7278538942337036, | |
| 0.5116866827011108, | |
| 0.9212220311164856, | |
| 0.7694430947303772, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9999998807907104, | |
| 1.0, | |
| 1.0, | |
| 0.9999598264694214, | |
| 0.9999998807907104, | |
| 0.3216085433959961, | |
| 0.999714195728302 | |
| ], | |
| "eval_code_hard_runtime": 110.6096, | |
| "eval_code_hard_samples_per_second": 0.904, | |
| "eval_code_hard_steps_per_second": 0.036, | |
| "epoch": 1.75, | |
| "step": 100 | |
| }, | |
| { | |
| "loss": 0.0619, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.77, | |
| "step": 101 | |
| }, | |
| { | |
| "loss": 0.0526, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.79, | |
| "step": 102 | |
| }, | |
| { | |
| "loss": 0.1902, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.81, | |
| "step": 103 | |
| }, | |
| { | |
| "loss": 0.2676, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.82, | |
| "step": 104 | |
| }, | |
| { | |
| "loss": 0.1422, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.84, | |
| "step": 105 | |
| }, | |
| { | |
| "loss": 0.0869, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.86, | |
| "step": 106 | |
| }, | |
| { | |
| "loss": 0.1517, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.88, | |
| "step": 107 | |
| }, | |
| { | |
| "loss": 0.0282, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.89, | |
| "step": 108 | |
| }, | |
| { | |
| "loss": 0.0516, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.91, | |
| "step": 109 | |
| }, | |
| { | |
| "loss": 0.0479, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.93, | |
| "step": 110 | |
| }, | |
| { | |
| "loss": 0.108, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.95, | |
| "step": 111 | |
| }, | |
| { | |
| "loss": 0.0833, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.96, | |
| "step": 112 | |
| }, | |
| { | |
| "loss": 0.1826, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.98, | |
| "step": 113 | |
| }, | |
| { | |
| "loss": 0.071, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.0, | |
| "step": 114 | |
| }, | |
| { | |
| "loss": 0.0619, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.02, | |
| "step": 115 | |
| }, | |
| { | |
| "loss": 0.0764, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.04, | |
| "step": 116 | |
| }, | |
| { | |
| "loss": 0.0009, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.05, | |
| "step": 117 | |
| }, | |
| { | |
| "loss": 0.0175, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.07, | |
| "step": 118 | |
| }, | |
| { | |
| "loss": 0.0216, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.09, | |
| "step": 119 | |
| }, | |
| { | |
| "loss": 0.0278, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.11, | |
| "step": 120 | |
| }, | |
| { | |
| "loss": 0.0665, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.12, | |
| "step": 121 | |
| }, | |
| { | |
| "loss": 0.0205, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.14, | |
| "step": 122 | |
| }, | |
| { | |
| "loss": 0.0052, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.16, | |
| "step": 123 | |
| }, | |
| { | |
| "loss": 0.0635, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.18, | |
| "step": 124 | |
| }, | |
| { | |
| "loss": 0.0132, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.19, | |
| "step": 125 | |
| }, | |
| { | |
| "eval_code_hard_loss": 0.05377120152115822, | |
| "eval_code_hard_score": -0.015216915868222713, | |
| "eval_code_hard_brier_score": 0.015216915868222713, | |
| "eval_code_hard_average_probability": 0.9605898261070251, | |
| "eval_code_hard_accuracy": 0.99, | |
| "eval_code_hard_probabilities": [ | |
| 0.9999129772186279, | |
| 0.9065911173820496, | |
| 0.9998986721038818, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.8372133374214172, | |
| 0.9947729706764221, | |
| 0.999957799911499, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9999998807907104, | |
| 1.0, | |
| 0.9999995231628418, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9999923706054688, | |
| 0.9862022995948792, | |
| 1.0, | |
| 1.0, | |
| 0.9995848536491394, | |
| 0.9999656677246094, | |
| 0.9963919520378113, | |
| 1.0, | |
| 0.9999942779541016, | |
| 0.999997615814209, | |
| 0.9999997615814209, | |
| 0.9999994039535522, | |
| 0.9999994039535522, | |
| 0.9998449087142944, | |
| 0.996722400188446, | |
| 0.9999366998672485, | |
| 0.9999986886978149, | |
| 0.9999994039535522, | |
| 0.9999995231628418, | |
| 0.9986975193023682, | |
| 0.6720824837684631, | |
| 0.8636956810951233, | |
| 0.9999996423721313, | |
| 0.9801883697509766, | |
| 0.9772735238075256, | |
| 0.22563716769218445, | |
| 0.9999665021896362, | |
| 0.9957283139228821, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.928858757019043, | |
| 0.7347054481506348, | |
| 0.8363421559333801, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9992239475250244, | |
| 0.9996882677078247, | |
| 0.9999969005584717, | |
| 1.0, | |
| 1.0, | |
| 0.9928505420684814, | |
| 0.9979680180549622, | |
| 0.9968575239181519, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9999997615814209, | |
| 0.9999778270721436, | |
| 0.9998857975006104, | |
| 0.5365374684333801, | |
| 0.6125951409339905, | |
| 0.9207331538200378, | |
| 0.9939044117927551, | |
| 0.9994903802871704, | |
| 0.9971593618392944, | |
| 1.0, | |
| 1.0, | |
| 0.8098579049110413, | |
| 0.5238723158836365, | |
| 0.9532552361488342, | |
| 0.8978452682495117, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9999998807907104, | |
| 1.0, | |
| 1.0, | |
| 0.9999985694885254, | |
| 1.0, | |
| 0.8971565961837769, | |
| 0.9999786615371704 | |
| ], | |
| "eval_code_hard_runtime": 110.7219, | |
| "eval_code_hard_samples_per_second": 0.903, | |
| "eval_code_hard_steps_per_second": 0.036, | |
| "epoch": 2.19, | |
| "step": 125 | |
| }, | |
| { | |
| "loss": 0.0251, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.21, | |
| "step": 126 | |
| }, | |
| { | |
| "loss": 0.0102, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.23, | |
| "step": 127 | |
| }, | |
| { | |
| "loss": 0.0179, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.25, | |
| "step": 128 | |
| }, | |
| { | |
| "loss": 0.1329, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.26, | |
| "step": 129 | |
| }, | |
| { | |
| "loss": 0.0082, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.28, | |
| "step": 130 | |
| }, | |
| { | |
| "loss": 0.0325, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.3, | |
| "step": 131 | |
| }, | |
| { | |
| "loss": 0.0094, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.32, | |
| "step": 132 | |
| }, | |
| { | |
| "loss": 0.004, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.33, | |
| "step": 133 | |
| }, | |
| { | |
| "loss": 0.0099, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.35, | |
| "step": 134 | |
| }, | |
| { | |
| "loss": 0.0454, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.37, | |
| "step": 135 | |
| }, | |
| { | |
| "loss": 0.0342, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.39, | |
| "step": 136 | |
| }, | |
| { | |
| "loss": 0.0161, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.4, | |
| "step": 137 | |
| }, | |
| { | |
| "loss": 0.0686, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.42, | |
| "step": 138 | |
| }, | |
| { | |
| "loss": 0.0404, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.44, | |
| "step": 139 | |
| }, | |
| { | |
| "loss": 0.0311, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.46, | |
| "step": 140 | |
| }, | |
| { | |
| "loss": 0.0426, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.47, | |
| "step": 141 | |
| }, | |
| { | |
| "loss": 0.013, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.49, | |
| "step": 142 | |
| }, | |
| { | |
| "loss": 0.0209, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.51, | |
| "step": 143 | |
| }, | |
| { | |
| "loss": 0.0441, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.53, | |
| "step": 144 | |
| }, | |
| { | |
| "loss": 0.0414, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.54, | |
| "step": 145 | |
| }, | |
| { | |
| "loss": 0.0022, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.56, | |
| "step": 146 | |
| }, | |
| { | |
| "loss": 0.0098, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.58, | |
| "step": 147 | |
| }, | |
| { | |
| "loss": 0.0076, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.6, | |
| "step": 148 | |
| }, | |
| { | |
| "loss": 0.0242, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.61, | |
| "step": 149 | |
| }, | |
| { | |
| "loss": 0.002, | |
| "learning_rate": 0.0002, | |
| "epoch": 2.63, | |
| "step": 150 | |
| }, | |
| { | |
| "eval_code_hard_loss": 0.06451280415058136, | |
| "eval_code_hard_score": -0.019366176798939705, | |
| "eval_code_hard_brier_score": 0.019366176798939705, | |
| "eval_code_hard_average_probability": 0.957671582698822, | |
| "eval_code_hard_accuracy": 0.97, | |
| "eval_code_hard_probabilities": [ | |
| 0.9999499320983887, | |
| 0.8330075144767761, | |
| 0.9997716546058655, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.4579625427722931, | |
| 0.9998301267623901, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9999998807907104, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9999909400939941, | |
| 1.0, | |
| 1.0, | |
| 0.9999996423721313, | |
| 1.0, | |
| 0.9973661303520203, | |
| 1.0, | |
| 0.9999998807907104, | |
| 1.0, | |
| 0.9999984502792358, | |
| 0.9999953508377075, | |
| 0.9999986886978149, | |
| 1.0, | |
| 0.9999997615814209, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9999028444290161, | |
| 0.7898390293121338, | |
| 0.8942784070968628, | |
| 1.0, | |
| 0.9987273812294006, | |
| 0.9987801909446716, | |
| 0.14106149971485138, | |
| 1.0, | |
| 0.9999027252197266, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.6841090321540833, | |
| 0.7871870398521423, | |
| 0.896818220615387, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.999897837638855, | |
| 0.9999984502792358, | |
| 0.9999983310699463, | |
| 1.0, | |
| 1.0, | |
| 0.9256799817085266, | |
| 0.9996222257614136, | |
| 0.9997231364250183, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9999997615814209, | |
| 0.9999966621398926, | |
| 0.5312544107437134, | |
| 0.583741307258606, | |
| 0.9588684439659119, | |
| 0.9209092259407043, | |
| 0.999733030796051, | |
| 0.9926868677139282, | |
| 1.0, | |
| 1.0, | |
| 0.9422075748443604, | |
| 0.4956146478652954, | |
| 0.9603027105331421, | |
| 0.9903588891029358, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.988089382648468, | |
| 0.9999991655349731 | |
| ], | |
| "eval_code_hard_runtime": 110.671, | |
| "eval_code_hard_samples_per_second": 0.904, | |
| "eval_code_hard_steps_per_second": 0.036, | |
| "epoch": 2.63, | |
| "step": 150 | |
| }, | |
| { | |
| "train_runtime": 14190.0736, | |
| "train_samples_per_second": 0.338, | |
| "train_steps_per_second": 0.011, | |
| "total_flos": 0.0, | |
| "train_loss": 0.17637818212776135, | |
| "epoch": 2.63, | |
| "step": 150 | |
| } | |
| ]] |