| [ | |
| { | |
| "loss": 0.7451, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.02, | |
| "step": 1 | |
| }, | |
| { | |
| "loss": 0.7931, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.04, | |
| "step": 2 | |
| }, | |
| { | |
| "loss": 0.7147, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.05, | |
| "step": 3 | |
| }, | |
| { | |
| "loss": 0.6208, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.07, | |
| "step": 4 | |
| }, | |
| { | |
| "loss": 0.6318, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.09, | |
| "step": 5 | |
| }, | |
| { | |
| "loss": 0.5347, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.11, | |
| "step": 6 | |
| }, | |
| { | |
| "loss": 0.5445, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.12, | |
| "step": 7 | |
| }, | |
| { | |
| "loss": 0.5323, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.14, | |
| "step": 8 | |
| }, | |
| { | |
| "loss": 0.5599, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.16, | |
| "step": 9 | |
| }, | |
| { | |
| "loss": 0.544, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.18, | |
| "step": 10 | |
| }, | |
| { | |
| "loss": 0.495, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.19, | |
| "step": 11 | |
| }, | |
| { | |
| "loss": 0.4648, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.21, | |
| "step": 12 | |
| }, | |
| { | |
| "loss": 0.5144, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.23, | |
| "step": 13 | |
| }, | |
| { | |
| "loss": 0.3049, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.25, | |
| "step": 14 | |
| }, | |
| { | |
| "loss": 0.2755, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.26, | |
| "step": 15 | |
| }, | |
| { | |
| "loss": 0.3953, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.28, | |
| "step": 16 | |
| }, | |
| { | |
| "loss": 0.293, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.3, | |
| "step": 17 | |
| }, | |
| { | |
| "loss": 0.3035, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.32, | |
| "step": 18 | |
| }, | |
| { | |
| "loss": 0.2193, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.33, | |
| "step": 19 | |
| }, | |
| { | |
| "loss": 0.306, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.35, | |
| "step": 20 | |
| }, | |
| { | |
| "loss": 0.3844, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.37, | |
| "step": 21 | |
| }, | |
| { | |
| "loss": 0.4754, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.39, | |
| "step": 22 | |
| }, | |
| { | |
| "loss": 0.247, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.4, | |
| "step": 23 | |
| }, | |
| { | |
| "loss": 0.2831, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.42, | |
| "step": 24 | |
| }, | |
| { | |
| "loss": 0.2395, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.44, | |
| "step": 25 | |
| }, | |
| { | |
| "eval_code_hard_loss": 0.18990269303321838, | |
| "eval_code_hard_score": -0.06281973421573639, | |
| "eval_code_hard_brier_score": 0.06281973421573639, | |
| "eval_code_hard_average_probability": 0.8535524010658264, | |
| "eval_code_hard_accuracy": 0.91, | |
| "eval_code_hard_probabilities": [ | |
| 0.9951574206352234, | |
| 0.6952859163284302, | |
| 0.695438027381897, | |
| 1.0, | |
| 0.9999997615814209, | |
| 0.9999988079071045, | |
| 1.0, | |
| 0.9999998807907104, | |
| 0.9999998807907104, | |
| 0.9022552371025085, | |
| 0.9998458623886108, | |
| 0.9992383718490601, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9705260396003723, | |
| 0.9934076070785522, | |
| 0.9893102645874023, | |
| 0.9892656803131104, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9999995231628418, | |
| 0.9999997615814209, | |
| 0.999992847442627, | |
| 0.9999940395355225, | |
| 0.9999849796295166, | |
| 0.9993022680282593, | |
| 0.9998225569725037, | |
| 0.9997448325157166, | |
| 0.5516940355300903, | |
| 0.6634960770606995, | |
| 0.5038003325462341, | |
| 0.9990589022636414, | |
| 0.9990129470825195, | |
| 0.9992672801017761, | |
| 0.9999462366104126, | |
| 0.9998952150344849, | |
| 0.9999340772628784, | |
| 0.9476701021194458, | |
| 0.9273682832717896, | |
| 0.9702255725860596, | |
| 0.9999784231185913, | |
| 0.9999960660934448, | |
| 0.999988317489624, | |
| 0.6653634309768677, | |
| 0.4677712023258209, | |
| 0.5697581768035889, | |
| 0.8528237342834473, | |
| 0.8113780617713928, | |
| 0.7631744742393494, | |
| 0.4795503616333008, | |
| 0.4501705467700958, | |
| 0.45226800441741943, | |
| 0.9999998807907104, | |
| 0.9999998807907104, | |
| 1.0, | |
| 0.5940175652503967, | |
| 0.5912690162658691, | |
| 0.5215416550636292, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.5378963351249695, | |
| 0.4695456922054291, | |
| 0.8803982734680176, | |
| 0.9970523118972778, | |
| 0.9994822144508362, | |
| 0.9994255304336548, | |
| 0.7984318733215332, | |
| 0.6354186534881592, | |
| 0.8742010593414307, | |
| 0.9999979734420776, | |
| 0.9999983310699463, | |
| 0.999997615814209, | |
| 0.9906328916549683, | |
| 0.9663383364677429, | |
| 0.9692559838294983, | |
| 0.46825510263442993, | |
| 0.6134918928146362, | |
| 0.5529954433441162, | |
| 0.6660357713699341, | |
| 0.9829654097557068, | |
| 0.47695887088775635, | |
| 0.9999682903289795, | |
| 0.9999746084213257, | |
| 0.5424057841300964, | |
| 0.518418550491333, | |
| 0.4060291051864624, | |
| 0.5124395489692688, | |
| 0.9932987689971924, | |
| 0.9838415384292603, | |
| 0.975809633731842, | |
| 0.9949588179588318, | |
| 0.9947303533554077, | |
| 0.9956455230712891, | |
| 0.6032052636146545, | |
| 0.830151379108429, | |
| 0.38333284854888916, | |
| 0.7332690954208374 | |
| ], | |
| "eval_code_hard_runtime": 140.0539, | |
| "eval_code_hard_samples_per_second": 0.714, | |
| "eval_code_hard_steps_per_second": 0.029, | |
| "epoch": 0.44, | |
| "step": 25 | |
| }, | |
| { | |
| "loss": 0.217, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.46, | |
| "step": 26 | |
| }, | |
| { | |
| "loss": 0.2851, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.47, | |
| "step": 27 | |
| }, | |
| { | |
| "loss": 0.2988, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.49, | |
| "step": 28 | |
| }, | |
| { | |
| "loss": 0.1119, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.51, | |
| "step": 29 | |
| }, | |
| { | |
| "loss": 0.1369, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.53, | |
| "step": 30 | |
| }, | |
| { | |
| "loss": 0.2311, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.54, | |
| "step": 31 | |
| }, | |
| { | |
| "loss": 0.2201, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.56, | |
| "step": 32 | |
| }, | |
| { | |
| "loss": 0.1824, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.58, | |
| "step": 33 | |
| }, | |
| { | |
| "loss": 0.1181, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.6, | |
| "step": 34 | |
| }, | |
| { | |
| "loss": 0.0781, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.61, | |
| "step": 35 | |
| }, | |
| { | |
| "loss": 0.1475, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.63, | |
| "step": 36 | |
| }, | |
| { | |
| "loss": 0.198, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.65, | |
| "step": 37 | |
| }, | |
| { | |
| "loss": 0.1074, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.67, | |
| "step": 38 | |
| }, | |
| { | |
| "loss": 0.3816, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.68, | |
| "step": 39 | |
| }, | |
| { | |
| "loss": 0.2657, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.7, | |
| "step": 40 | |
| }, | |
| { | |
| "loss": 0.1525, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.72, | |
| "step": 41 | |
| }, | |
| { | |
| "loss": 0.1426, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.74, | |
| "step": 42 | |
| }, | |
| { | |
| "loss": 0.1578, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.75, | |
| "step": 43 | |
| }, | |
| { | |
| "loss": 0.1234, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.77, | |
| "step": 44 | |
| }, | |
| { | |
| "loss": 0.1591, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.79, | |
| "step": 45 | |
| }, | |
| { | |
| "loss": 0.0388, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.81, | |
| "step": 46 | |
| }, | |
| { | |
| "loss": 0.1186, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.82, | |
| "step": 47 | |
| }, | |
| { | |
| "loss": 0.2242, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.84, | |
| "step": 48 | |
| }, | |
| { | |
| "loss": 0.2245, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.86, | |
| "step": 49 | |
| }, | |
| { | |
| "loss": 0.0825, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.88, | |
| "step": 50 | |
| }, | |
| { | |
| "eval_code_hard_loss": 0.1537313610315323, | |
| "eval_code_hard_score": -0.04667011648416519, | |
| "eval_code_hard_brier_score": 0.04667011648416519, | |
| "eval_code_hard_average_probability": 0.8784838318824768, | |
| "eval_code_hard_accuracy": 0.94, | |
| "eval_code_hard_probabilities": [ | |
| 0.9725497364997864, | |
| 0.7011394500732422, | |
| 0.6710378527641296, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9999716281890869, | |
| 0.9999884366989136, | |
| 0.999969482421875, | |
| 0.48456111550331116, | |
| 0.9981953501701355, | |
| 0.9864223003387451, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9983637928962708, | |
| 0.9989921450614929, | |
| 0.9978587031364441, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9999933242797852, | |
| 0.9999672174453735, | |
| 0.9999806880950928, | |
| 0.9999942779541016, | |
| 0.9999948740005493, | |
| 0.9999951124191284, | |
| 0.9994611144065857, | |
| 0.9993894100189209, | |
| 0.999222993850708, | |
| 0.8560943007469177, | |
| 0.8921459317207336, | |
| 0.7734678387641907, | |
| 0.9890369772911072, | |
| 0.9659588932991028, | |
| 0.9716930985450745, | |
| 0.9913560152053833, | |
| 0.9888952970504761, | |
| 0.9655161499977112, | |
| 0.882901132106781, | |
| 0.8454601168632507, | |
| 0.8470443487167358, | |
| 0.9999980926513672, | |
| 0.9999996423721313, | |
| 0.9999994039535522, | |
| 0.6798665523529053, | |
| 0.49168092012405396, | |
| 0.5368497967720032, | |
| 0.9873051047325134, | |
| 0.9402137994766235, | |
| 0.928394615650177, | |
| 0.35502684116363525, | |
| 0.9124428629875183, | |
| 0.946331799030304, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.6016813516616821, | |
| 0.49267151951789856, | |
| 0.6437432169914246, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.4486885964870453, | |
| 0.640812337398529, | |
| 0.879755437374115, | |
| 0.9916812777519226, | |
| 0.9911965131759644, | |
| 0.9993059635162354, | |
| 0.589796245098114, | |
| 0.7118774652481079, | |
| 0.9268589019775391, | |
| 0.9999767541885376, | |
| 0.9999942779541016, | |
| 0.9999902248382568, | |
| 0.9828012585639954, | |
| 0.9357141256332397, | |
| 0.9222304224967957, | |
| 0.5056607723236084, | |
| 0.7614033818244934, | |
| 0.7656963467597961, | |
| 0.6500656008720398, | |
| 0.8527267575263977, | |
| 0.5424817800521851, | |
| 0.9999986886978149, | |
| 0.999997615814209, | |
| 0.8840383291244507, | |
| 0.5736863613128662, | |
| 0.5833655595779419, | |
| 0.69774329662323, | |
| 0.9973189234733582, | |
| 0.9964283108711243, | |
| 0.9958376884460449, | |
| 0.9838255047798157, | |
| 0.9848774671554565, | |
| 0.9828516244888306, | |
| 0.7398984432220459, | |
| 0.9901463389396667, | |
| 0.3069886267185211, | |
| 0.7398353219032288 | |
| ], | |
| "eval_code_hard_runtime": 140.0621, | |
| "eval_code_hard_samples_per_second": 0.714, | |
| "eval_code_hard_steps_per_second": 0.029, | |
| "epoch": 0.88, | |
| "step": 50 | |
| }, | |
| { | |
| "loss": 0.1986, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.89, | |
| "step": 51 | |
| }, | |
| { | |
| "loss": 0.1768, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.91, | |
| "step": 52 | |
| }, | |
| { | |
| "loss": 0.1257, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.93, | |
| "step": 53 | |
| }, | |
| { | |
| "loss": 0.1255, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.95, | |
| "step": 54 | |
| }, | |
| { | |
| "loss": 0.067, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.96, | |
| "step": 55 | |
| }, | |
| { | |
| "loss": 0.2208, | |
| "learning_rate": 0.0002, | |
| "epoch": 0.98, | |
| "step": 56 | |
| }, | |
| { | |
| "loss": 0.1187, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.0, | |
| "step": 57 | |
| }, | |
| { | |
| "loss": 0.159, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.02, | |
| "step": 58 | |
| }, | |
| { | |
| "loss": 0.0512, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.04, | |
| "step": 59 | |
| }, | |
| { | |
| "loss": 0.0608, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.05, | |
| "step": 60 | |
| }, | |
| { | |
| "loss": 0.1049, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.07, | |
| "step": 61 | |
| }, | |
| { | |
| "loss": 0.0634, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.09, | |
| "step": 62 | |
| }, | |
| { | |
| "loss": 0.0609, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.11, | |
| "step": 63 | |
| }, | |
| { | |
| "loss": 0.0888, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.12, | |
| "step": 64 | |
| }, | |
| { | |
| "loss": 0.1165, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.14, | |
| "step": 65 | |
| }, | |
| { | |
| "loss": 0.017, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.16, | |
| "step": 66 | |
| }, | |
| { | |
| "loss": 0.0504, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.18, | |
| "step": 67 | |
| }, | |
| { | |
| "loss": 0.0958, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.19, | |
| "step": 68 | |
| }, | |
| { | |
| "loss": 0.0276, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.21, | |
| "step": 69 | |
| }, | |
| { | |
| "loss": 0.0394, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.23, | |
| "step": 70 | |
| }, | |
| { | |
| "loss": 0.0398, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.25, | |
| "step": 71 | |
| }, | |
| { | |
| "loss": 0.108, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.26, | |
| "step": 72 | |
| }, | |
| { | |
| "loss": 0.0392, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.28, | |
| "step": 73 | |
| }, | |
| { | |
| "loss": 0.0233, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.3, | |
| "step": 74 | |
| }, | |
| { | |
| "loss": 0.1554, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.32, | |
| "step": 75 | |
| }, | |
| { | |
| "eval_code_hard_loss": 0.13956719636917114, | |
| "eval_code_hard_score": -0.042688366025686264, | |
| "eval_code_hard_brier_score": 0.042688366025686264, | |
| "eval_code_hard_average_probability": 0.9274539947509766, | |
| "eval_code_hard_accuracy": 0.93, | |
| "eval_code_hard_probabilities": [ | |
| 0.9999963045120239, | |
| 0.9699520468711853, | |
| 0.9745141267776489, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.35654664039611816, | |
| 0.886887788772583, | |
| 0.9559746980667114, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9999935626983643, | |
| 0.9999960660934448, | |
| 0.9999923706054688, | |
| 0.9999997615814209, | |
| 1.0, | |
| 0.9999996423721313, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9999995231628418, | |
| 0.9999997615814209, | |
| 0.9998072981834412, | |
| 0.9999998807907104, | |
| 0.999997615814209, | |
| 0.9905921816825867, | |
| 0.9987447261810303, | |
| 0.9492596983909607, | |
| 1.0, | |
| 0.9999996423721313, | |
| 0.9999969005584717, | |
| 1.0, | |
| 0.9999995231628418, | |
| 0.9999992847442627, | |
| 0.9940958023071289, | |
| 0.9881225228309631, | |
| 0.9979448914527893, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9875559210777283, | |
| 0.578984797000885, | |
| 0.643305242061615, | |
| 0.9999995231628418, | |
| 0.9999939203262329, | |
| 0.9999977350234985, | |
| 0.005097602494060993, | |
| 0.9997228980064392, | |
| 0.9975023865699768, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.3311256468296051, | |
| 0.45201003551483154, | |
| 0.7562749981880188, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9942985773086548, | |
| 0.9975347518920898, | |
| 0.9981868863105774, | |
| 0.9999996423721313, | |
| 0.9999785423278809, | |
| 1.0, | |
| 0.46586939692497253, | |
| 0.9903834462165833, | |
| 0.9999986886978149, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9817127585411072, | |
| 0.2876265347003937, | |
| 0.8499608635902405, | |
| 0.61001056432724, | |
| 0.996311604976654, | |
| 0.999902606010437, | |
| 0.9834624528884888, | |
| 0.9992788434028625, | |
| 0.9218541383743286, | |
| 1.0, | |
| 1.0, | |
| 0.9609642028808594, | |
| 0.7825002074241638, | |
| 0.9717795252799988, | |
| 0.9860764741897583, | |
| 0.9999998807907104, | |
| 0.9999990463256836, | |
| 0.9999997615814209, | |
| 0.9999983310699463, | |
| 0.9999971389770508, | |
| 0.9999949932098389, | |
| 0.9969584941864014, | |
| 0.9999991655349731, | |
| 0.15736030042171478, | |
| 0.9994120597839355 | |
| ], | |
| "eval_code_hard_runtime": 140.0578, | |
| "eval_code_hard_samples_per_second": 0.714, | |
| "eval_code_hard_steps_per_second": 0.029, | |
| "epoch": 1.32, | |
| "step": 75 | |
| }, | |
| { | |
| "loss": 0.0227, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.33, | |
| "step": 76 | |
| }, | |
| { | |
| "loss": 0.0849, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.35, | |
| "step": 77 | |
| }, | |
| { | |
| "loss": 0.1094, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.37, | |
| "step": 78 | |
| }, | |
| { | |
| "loss": 0.0689, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.39, | |
| "step": 79 | |
| }, | |
| { | |
| "loss": 0.0418, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.4, | |
| "step": 80 | |
| }, | |
| { | |
| "loss": 0.0886, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.42, | |
| "step": 81 | |
| }, | |
| { | |
| "loss": 0.1833, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.44, | |
| "step": 82 | |
| }, | |
| { | |
| "loss": 0.1213, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.46, | |
| "step": 83 | |
| }, | |
| { | |
| "loss": 0.0424, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.47, | |
| "step": 84 | |
| }, | |
| { | |
| "loss": 0.0194, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.49, | |
| "step": 85 | |
| }, | |
| { | |
| "loss": 0.0229, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.51, | |
| "step": 86 | |
| }, | |
| { | |
| "loss": 0.0231, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.53, | |
| "step": 87 | |
| }, | |
| { | |
| "loss": 0.0263, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.54, | |
| "step": 88 | |
| }, | |
| { | |
| "loss": 0.0619, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.56, | |
| "step": 89 | |
| }, | |
| { | |
| "loss": 0.0365, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.58, | |
| "step": 90 | |
| }, | |
| { | |
| "loss": 0.0523, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.6, | |
| "step": 91 | |
| }, | |
| { | |
| "loss": 0.0172, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.61, | |
| "step": 92 | |
| }, | |
| { | |
| "loss": 0.0455, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.63, | |
| "step": 93 | |
| }, | |
| { | |
| "loss": 0.2764, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.65, | |
| "step": 94 | |
| }, | |
| { | |
| "loss": 0.0225, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.67, | |
| "step": 95 | |
| }, | |
| { | |
| "loss": 0.0133, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.68, | |
| "step": 96 | |
| }, | |
| { | |
| "loss": 0.0188, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.7, | |
| "step": 97 | |
| }, | |
| { | |
| "loss": 0.0207, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.72, | |
| "step": 98 | |
| }, | |
| { | |
| "loss": 0.012, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.74, | |
| "step": 99 | |
| }, | |
| { | |
| "loss": 0.0628, | |
| "learning_rate": 0.0002, | |
| "epoch": 1.75, | |
| "step": 100 | |
| }, | |
| { | |
| "eval_code_hard_loss": 0.1615547239780426, | |
| "eval_code_hard_score": -0.03736421465873718, | |
| "eval_code_hard_brier_score": 0.03736421465873718, | |
| "eval_code_hard_average_probability": 0.9438884854316711, | |
| "eval_code_hard_accuracy": 0.95, | |
| "eval_code_hard_probabilities": [ | |
| 1.0, | |
| 0.9725049734115601, | |
| 0.9244245886802673, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.07730857282876968, | |
| 0.9794886112213135, | |
| 0.999785840511322, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9999511241912842, | |
| 1.0, | |
| 1.0, | |
| 0.9999926090240479, | |
| 1.0, | |
| 0.9999614953994751, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9999998807907104, | |
| 1.0, | |
| 0.9999260902404785, | |
| 0.9998406171798706, | |
| 0.9999992847442627, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9999581575393677, | |
| 0.6513886451721191, | |
| 0.7881560921669006, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.00015713961329311132, | |
| 1.0, | |
| 0.9999954700469971, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.4968324303627014, | |
| 0.40699660778045654, | |
| 0.9514529705047607, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9999997615814209, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9981179237365723, | |
| 0.999592125415802, | |
| 0.9999996423721313, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9999996423721313, | |
| 0.9520946741104126, | |
| 0.9975154399871826, | |
| 0.5100582242012024, | |
| 0.9827112555503845, | |
| 0.9999972581863403, | |
| 0.9999854564666748, | |
| 1.0, | |
| 0.9896236062049866, | |
| 1.0, | |
| 1.0, | |
| 0.8571937680244446, | |
| 0.796758770942688, | |
| 0.9510358572006226, | |
| 0.9994876384735107, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 1.0, | |
| 0.9998961687088013, | |
| 1.0, | |
| 0.10666914284229279, | |
| 1.0 | |
| ], | |
| "eval_code_hard_runtime": 140.0373, | |
| "eval_code_hard_samples_per_second": 0.714, | |
| "eval_code_hard_steps_per_second": 0.029, | |
| "epoch": 1.75, | |
| "step": 100 | |
| }, | |
| { | |
| "train_runtime": 12026.9778, | |
| "train_samples_per_second": 0.266, | |
| "train_steps_per_second": 0.008, | |
| "total_flos": 0.0, | |
| "train_loss": 0.19655362625606357, | |
| "epoch": 1.75, | |
| "step": 100 | |
| } | |
| ] |