| [ |
| { |
| "epoch": 0.02, |
| "learning_rate": 0.0002, |
| "loss": 0.7109, |
| "step": 1 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 0.0002, |
| "loss": 0.7151, |
| "step": 2 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 0.0002, |
| "loss": 0.7036, |
| "step": 3 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 0.0002, |
| "loss": 0.6805, |
| "step": 4 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 0.0002, |
| "loss": 0.6678, |
| "step": 5 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 0.0002, |
| "loss": 0.6692, |
| "step": 6 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 0.0002, |
| "loss": 0.6459, |
| "step": 7 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 0.0002, |
| "loss": 0.6262, |
| "step": 8 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 0.0002, |
| "loss": 0.6574, |
| "step": 9 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 0.0002, |
| "loss": 0.6311, |
| "step": 10 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 0.0002, |
| "loss": 0.5219, |
| "step": 11 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 0.0002, |
| "loss": 0.5933, |
| "step": 12 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 0.0002, |
| "loss": 0.5881, |
| "step": 13 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 0.0002, |
| "loss": 0.6345, |
| "step": 14 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 0.0002, |
| "loss": 0.559, |
| "step": 15 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 0.0002, |
| "loss": 0.5686, |
| "step": 16 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 0.0002, |
| "loss": 0.577, |
| "step": 17 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 0.0002, |
| "loss": 0.4884, |
| "step": 18 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 0.0002, |
| "loss": 0.4593, |
| "step": 19 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 0.0002, |
| "loss": 0.4582, |
| "step": 20 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 0.0002, |
| "loss": 0.428, |
| "step": 21 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 0.0002, |
| "loss": 0.495, |
| "step": 22 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 0.0002, |
| "loss": 0.4683, |
| "step": 23 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 0.0002, |
| "loss": 0.3779, |
| "step": 24 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 0.0002, |
| "loss": 0.5694, |
| "step": 25 |
| }, |
| { |
| "epoch": 0.44, |
| "eval_code_accuracy": 0.81, |
| "eval_code_average_probability": 0.6998826861381531, |
| "eval_code_brier_score": 0.13871249556541443, |
| "eval_code_loss": 0.392591655254364, |
| "eval_code_probabilities": [ |
| 0.9998756647109985, |
| 0.9996539354324341, |
| 0.9968274235725403, |
| 0.867797315120697, |
| 0.604396641254425, |
| 0.8194839358329773, |
| 0.9978196620941162, |
| 0.9973512887954712, |
| 0.8165198564529419, |
| 0.929094672203064, |
| 0.9696783423423767, |
| 0.8749808073043823, |
| 0.9949828386306763, |
| 0.9988803267478943, |
| 0.9974466562271118, |
| 0.5661922693252563, |
| 0.7031828761100769, |
| 0.7817732095718384, |
| 0.7952773571014404, |
| 0.700941801071167, |
| 0.5993526577949524, |
| 0.4928004741668701, |
| 0.7662289142608643, |
| 0.7449039816856384, |
| 0.26918813586235046, |
| 0.23834775388240814, |
| 0.21881021559238434, |
| 0.5174687504768372, |
| 0.48715656995773315, |
| 0.5120030045509338, |
| 0.3874703049659729, |
| 0.46171045303344727, |
| 0.6353215575218201, |
| 0.9995631575584412, |
| 0.9988592863082886, |
| 0.9959748387336731, |
| 0.9921092391014099, |
| 0.9928039908409119, |
| 0.9970149993896484, |
| 0.5361520648002625, |
| 0.6019002795219421, |
| 0.4536624550819397, |
| 0.6162904500961304, |
| 0.5367547869682312, |
| 0.620251476764679, |
| 0.5513021349906921, |
| 0.5809685587882996, |
| 0.5933180451393127, |
| 0.726946234703064, |
| 0.6555966138839722, |
| 0.999262273311615, |
| 0.9996010661125183, |
| 0.9998406171798706, |
| 0.7900382280349731, |
| 0.5323582291603088, |
| 0.5186650156974792, |
| 0.5038493275642395, |
| 0.9872013330459595, |
| 0.5808623433113098, |
| 0.9900431632995605, |
| 0.9054338335990906, |
| 0.8925741910934448, |
| 0.8963204026222229, |
| 0.48921719193458557, |
| 0.565790057182312, |
| 0.544992208480835, |
| 0.6663545370101929, |
| 0.49584266543388367, |
| 0.4396786093711853, |
| 0.5050247311592102, |
| 0.4423765242099762, |
| 0.5252822637557983, |
| 0.5083388686180115, |
| 0.5204058885574341, |
| 0.4985828101634979, |
| 0.9953283071517944, |
| 0.9802095890045166, |
| 0.9042825698852539, |
| 0.9987348914146423, |
| 0.9832821488380432, |
| 0.9921707510948181, |
| 0.5249463319778442, |
| 0.496512770652771, |
| 0.529527485370636, |
| 0.6597751975059509, |
| 0.9941974878311157, |
| 0.8405609130859375, |
| 0.6745859384536743, |
| 0.6913660168647766, |
| 0.4706657826900482, |
| 0.6228535771369934, |
| 0.7330396175384521, |
| 0.5534575581550598, |
| 0.4912925362586975, |
| 0.5196889042854309, |
| 0.5061907768249512, |
| 0.527052640914917, |
| 0.46116262674331665, |
| 0.40944358706474304, |
| 0.4376073181629181 |
| ], |
| "eval_code_runtime": 105.7673, |
| "eval_code_samples_per_second": 0.945, |
| "eval_code_score": -0.13871249556541443, |
| "eval_code_steps_per_second": 0.038, |
| "step": 25 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 0.0002, |
| "loss": 0.3947, |
| "step": 26 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 0.0002, |
| "loss": 0.4236, |
| "step": 27 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 0.0002, |
| "loss": 0.349, |
| "step": 28 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 0.0002, |
| "loss": 0.3414, |
| "step": 29 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 0.0002, |
| "loss": 0.5019, |
| "step": 30 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 0.0002, |
| "loss": 0.3768, |
| "step": 31 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 0.0002, |
| "loss": 0.3902, |
| "step": 32 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 0.0002, |
| "loss": 0.2838, |
| "step": 33 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 0.0002, |
| "loss": 0.1972, |
| "step": 34 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 0.0002, |
| "loss": 0.3796, |
| "step": 35 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 0.0002, |
| "loss": 0.3741, |
| "step": 36 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 0.0002, |
| "loss": 0.2624, |
| "step": 37 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 0.0002, |
| "loss": 0.2505, |
| "step": 38 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 0.0002, |
| "loss": 0.2687, |
| "step": 39 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 0.0002, |
| "loss": 0.2856, |
| "step": 40 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 0.0002, |
| "loss": 0.1638, |
| "step": 41 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 0.0002, |
| "loss": 0.2595, |
| "step": 42 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 0.0002, |
| "loss": 0.2886, |
| "step": 43 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 0.0002, |
| "loss": 0.1604, |
| "step": 44 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 0.0002, |
| "loss": 0.2069, |
| "step": 45 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 0.0002, |
| "loss": 0.2285, |
| "step": 46 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 0.0002, |
| "loss": 0.1975, |
| "step": 47 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 0.0002, |
| "loss": 0.1726, |
| "step": 48 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 0.0002, |
| "loss": 0.0881, |
| "step": 49 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 0.0002, |
| "loss": 0.2712, |
| "step": 50 |
| }, |
| { |
| "epoch": 0.88, |
| "eval_code_accuracy": 0.93, |
| "eval_code_average_probability": 0.8494662642478943, |
| "eval_code_brier_score": 0.06509324163198471, |
| "eval_code_loss": 0.18686418235301971, |
| "eval_code_probabilities": [ |
| 0.9999998807907104, |
| 0.9999940395355225, |
| 0.9999998807907104, |
| 0.999998927116394, |
| 0.9994799494743347, |
| 0.9999616146087646, |
| 1.0, |
| 0.9999617338180542, |
| 0.997948944568634, |
| 0.9999988079071045, |
| 1.0, |
| 0.9999945163726807, |
| 0.9999972581863403, |
| 0.9999998807907104, |
| 0.9972482323646545, |
| 0.8824266195297241, |
| 0.5207855105400085, |
| 0.9811423420906067, |
| 0.9948011636734009, |
| 0.9970309734344482, |
| 0.9518866539001465, |
| 0.9435040950775146, |
| 0.9998762607574463, |
| 0.9999566078186035, |
| 0.5264937281608582, |
| 0.3567848801612854, |
| 0.2969517111778259, |
| 0.5564099550247192, |
| 0.5566191673278809, |
| 0.6488668918609619, |
| 0.2498033493757248, |
| 0.8209301233291626, |
| 0.9682015776634216, |
| 0.999983549118042, |
| 0.998921275138855, |
| 0.9983420372009277, |
| 0.9999994039535522, |
| 1.0, |
| 1.0, |
| 0.9042081236839294, |
| 0.9448812007904053, |
| 0.8440216779708862, |
| 0.9969866871833801, |
| 0.935389518737793, |
| 0.9999938011169434, |
| 0.8614246249198914, |
| 0.9949424862861633, |
| 0.9876769781112671, |
| 0.99991774559021, |
| 0.9980294108390808, |
| 0.9999949932098389, |
| 1.0, |
| 1.0, |
| 0.9999998807907104, |
| 0.6444177627563477, |
| 0.5885342359542847, |
| 0.6353741884231567, |
| 0.9999998807907104, |
| 0.6010631322860718, |
| 0.999997615814209, |
| 0.9806910753250122, |
| 0.9655673503875732, |
| 0.9658817648887634, |
| 0.829046368598938, |
| 0.9939888715744019, |
| 0.8807443380355835, |
| 0.9380287528038025, |
| 0.7914523482322693, |
| 0.3484261929988861, |
| 0.6201522350311279, |
| 0.34111207723617554, |
| 0.7237049341201782, |
| 0.6027003526687622, |
| 0.8865858912467957, |
| 0.632958710193634, |
| 1.0, |
| 0.9925962090492249, |
| 0.9999537467956543, |
| 1.0, |
| 0.999997615814209, |
| 0.9999998807907104, |
| 0.6945015788078308, |
| 0.5074270963668823, |
| 0.7571401000022888, |
| 0.9955568909645081, |
| 0.9997404217720032, |
| 0.9955176711082458, |
| 0.7693279385566711, |
| 0.8310465812683105, |
| 0.5401557683944702, |
| 0.9749163389205933, |
| 0.963936448097229, |
| 0.9640229940414429, |
| 0.5689054131507874, |
| 0.8011016249656677, |
| 0.6685754060745239, |
| 0.6061564087867737, |
| 0.44276222586631775, |
| 0.4091780185699463, |
| 0.7819089889526367 |
| ], |
| "eval_code_runtime": 105.8373, |
| "eval_code_samples_per_second": 0.945, |
| "eval_code_score": -0.06509324163198471, |
| "eval_code_steps_per_second": 0.038, |
| "step": 50 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 0.0002, |
| "loss": 0.181, |
| "step": 51 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 0.0002, |
| "loss": 0.2069, |
| "step": 52 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 0.0002, |
| "loss": 0.2289, |
| "step": 53 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 0.0002, |
| "loss": 0.2987, |
| "step": 54 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 0.0002, |
| "loss": 0.3474, |
| "step": 55 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 0.0002, |
| "loss": 0.284, |
| "step": 56 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 0.0002, |
| "loss": 0.7535, |
| "step": 57 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 0.0002, |
| "loss": 0.1286, |
| "step": 58 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 0.0002, |
| "loss": 0.0816, |
| "step": 59 |
| }, |
| { |
| "epoch": 1.05, |
| "learning_rate": 0.0002, |
| "loss": 0.0607, |
| "step": 60 |
| }, |
| { |
| "epoch": 1.07, |
| "learning_rate": 0.0002, |
| "loss": 0.0772, |
| "step": 61 |
| }, |
| { |
| "epoch": 1.09, |
| "learning_rate": 0.0002, |
| "loss": 0.1129, |
| "step": 62 |
| }, |
| { |
| "epoch": 1.11, |
| "learning_rate": 0.0002, |
| "loss": 0.0509, |
| "step": 63 |
| }, |
| { |
| "epoch": 1.12, |
| "learning_rate": 0.0002, |
| "loss": 0.1928, |
| "step": 64 |
| }, |
| { |
| "epoch": 1.14, |
| "learning_rate": 0.0002, |
| "loss": 0.0945, |
| "step": 65 |
| }, |
| { |
| "epoch": 1.16, |
| "learning_rate": 0.0002, |
| "loss": 0.1268, |
| "step": 66 |
| }, |
| { |
| "epoch": 1.18, |
| "learning_rate": 0.0002, |
| "loss": 0.1186, |
| "step": 67 |
| }, |
| { |
| "epoch": 1.19, |
| "learning_rate": 0.0002, |
| "loss": 0.1123, |
| "step": 68 |
| }, |
| { |
| "epoch": 1.21, |
| "learning_rate": 0.0002, |
| "loss": 0.1512, |
| "step": 69 |
| }, |
| { |
| "epoch": 1.23, |
| "learning_rate": 0.0002, |
| "loss": 0.1303, |
| "step": 70 |
| }, |
| { |
| "epoch": 1.25, |
| "learning_rate": 0.0002, |
| "loss": 0.0441, |
| "step": 71 |
| }, |
| { |
| "epoch": 1.26, |
| "learning_rate": 0.0002, |
| "loss": 0.1391, |
| "step": 72 |
| }, |
| { |
| "epoch": 1.28, |
| "learning_rate": 0.0002, |
| "loss": 0.0685, |
| "step": 73 |
| }, |
| { |
| "epoch": 1.3, |
| "learning_rate": 0.0002, |
| "loss": 0.0762, |
| "step": 74 |
| }, |
| { |
| "epoch": 1.32, |
| "learning_rate": 0.0002, |
| "loss": 0.1011, |
| "step": 75 |
| }, |
| { |
| "epoch": 1.32, |
| "eval_code_accuracy": 0.91, |
| "eval_code_average_probability": 0.889110267162323, |
| "eval_code_brier_score": 0.06802475452423096, |
| "eval_code_loss": 0.21632270514965057, |
| "eval_code_probabilities": [ |
| 0.9983546137809753, |
| 0.8964507579803467, |
| 0.9998843669891357, |
| 1.0, |
| 0.9999983310699463, |
| 0.9999896287918091, |
| 1.0, |
| 1.0, |
| 0.9999992847442627, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9493870139122009, |
| 0.22386565804481506, |
| 0.9954581260681152, |
| 0.9999986886978149, |
| 1.0, |
| 0.9999638795852661, |
| 0.9999364614486694, |
| 1.0, |
| 1.0, |
| 0.14285829663276672, |
| 0.05121216922998428, |
| 0.06197505444288254, |
| 0.8036693334579468, |
| 0.784625768661499, |
| 0.9341561198234558, |
| 0.2160544991493225, |
| 0.9977596998214722, |
| 0.9990569949150085, |
| 0.9999992847442627, |
| 0.5561187863349915, |
| 0.9838144183158875, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999961853027344, |
| 0.999188244342804, |
| 0.9992071986198425, |
| 0.9999998807907104, |
| 0.9616034030914307, |
| 0.9999998807907104, |
| 0.9486005902290344, |
| 0.9999855756759644, |
| 0.9963693618774414, |
| 0.9999858140945435, |
| 0.9972230195999146, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.7309120893478394, |
| 0.8254848122596741, |
| 0.7734106183052063, |
| 1.0, |
| 0.5664382576942444, |
| 1.0, |
| 0.9999998807907104, |
| 0.9999990463256836, |
| 0.9999992847442627, |
| 0.9987679123878479, |
| 0.9999841451644897, |
| 0.9796060919761658, |
| 0.9999995231628418, |
| 0.4220101237297058, |
| 0.517703652381897, |
| 0.8652507066726685, |
| 0.007634077221155167, |
| 0.9999922513961792, |
| 0.9957844614982605, |
| 0.9999879598617554, |
| 0.9737687706947327, |
| 1.0, |
| 0.9995587468147278, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9666233658790588, |
| 0.6209402084350586, |
| 0.978766143321991, |
| 0.9995095729827881, |
| 0.9826872944831848, |
| 1.0, |
| 0.9902674555778503, |
| 0.9968993663787842, |
| 0.8134205937385559, |
| 0.9999436140060425, |
| 0.9293919205665588, |
| 0.9999128580093384, |
| 0.8822279572486877, |
| 0.9978179931640625, |
| 0.9373266100883484, |
| 0.7644922733306885, |
| 0.43501412868499756, |
| 0.4749135673046112, |
| 0.9878302216529846 |
| ], |
| "eval_code_runtime": 105.7625, |
| "eval_code_samples_per_second": 0.946, |
| "eval_code_score": -0.06802475452423096, |
| "eval_code_steps_per_second": 0.038, |
| "step": 75 |
| }, |
| { |
| "epoch": 1.33, |
| "learning_rate": 0.0002, |
| "loss": 0.0665, |
| "step": 76 |
| }, |
| { |
| "epoch": 1.35, |
| "learning_rate": 0.0002, |
| "loss": 0.0808, |
| "step": 77 |
| }, |
| { |
| "epoch": 1.37, |
| "learning_rate": 0.0002, |
| "loss": 0.0637, |
| "step": 78 |
| }, |
| { |
| "epoch": 1.39, |
| "learning_rate": 0.0002, |
| "loss": 0.0323, |
| "step": 79 |
| }, |
| { |
| "epoch": 1.4, |
| "learning_rate": 0.0002, |
| "loss": 0.0441, |
| "step": 80 |
| }, |
| { |
| "epoch": 1.42, |
| "learning_rate": 0.0002, |
| "loss": 0.1698, |
| "step": 81 |
| }, |
| { |
| "epoch": 1.44, |
| "learning_rate": 0.0002, |
| "loss": 0.156, |
| "step": 82 |
| }, |
| { |
| "epoch": 1.46, |
| "learning_rate": 0.0002, |
| "loss": 0.3788, |
| "step": 83 |
| }, |
| { |
| "epoch": 1.47, |
| "learning_rate": 0.0002, |
| "loss": 0.1222, |
| "step": 84 |
| }, |
| { |
| "epoch": 1.49, |
| "learning_rate": 0.0002, |
| "loss": 0.1639, |
| "step": 85 |
| }, |
| { |
| "epoch": 1.51, |
| "learning_rate": 0.0002, |
| "loss": 0.0538, |
| "step": 86 |
| }, |
| { |
| "epoch": 1.53, |
| "learning_rate": 0.0002, |
| "loss": 0.0328, |
| "step": 87 |
| }, |
| { |
| "epoch": 1.54, |
| "learning_rate": 0.0002, |
| "loss": 0.1745, |
| "step": 88 |
| }, |
| { |
| "epoch": 1.56, |
| "learning_rate": 0.0002, |
| "loss": 0.0874, |
| "step": 89 |
| }, |
| { |
| "epoch": 1.58, |
| "learning_rate": 0.0002, |
| "loss": 0.1177, |
| "step": 90 |
| }, |
| { |
| "epoch": 1.6, |
| "learning_rate": 0.0002, |
| "loss": 0.1185, |
| "step": 91 |
| }, |
| { |
| "epoch": 1.61, |
| "learning_rate": 0.0002, |
| "loss": 0.0543, |
| "step": 92 |
| }, |
| { |
| "epoch": 1.63, |
| "learning_rate": 0.0002, |
| "loss": 0.0392, |
| "step": 93 |
| }, |
| { |
| "epoch": 1.65, |
| "learning_rate": 0.0002, |
| "loss": 0.2006, |
| "step": 94 |
| }, |
| { |
| "epoch": 1.67, |
| "learning_rate": 0.0002, |
| "loss": 0.1249, |
| "step": 95 |
| }, |
| { |
| "epoch": 1.68, |
| "learning_rate": 0.0002, |
| "loss": 0.0341, |
| "step": 96 |
| }, |
| { |
| "epoch": 1.7, |
| "learning_rate": 0.0002, |
| "loss": 0.0666, |
| "step": 97 |
| }, |
| { |
| "epoch": 1.72, |
| "learning_rate": 0.0002, |
| "loss": 0.1614, |
| "step": 98 |
| }, |
| { |
| "epoch": 1.74, |
| "learning_rate": 0.0002, |
| "loss": 0.0831, |
| "step": 99 |
| }, |
| { |
| "epoch": 1.75, |
| "learning_rate": 0.0002, |
| "loss": 0.0273, |
| "step": 100 |
| }, |
| { |
| "epoch": 1.75, |
| "eval_code_accuracy": 0.96, |
| "eval_code_average_probability": 0.9304368495941162, |
| "eval_code_brier_score": 0.03531455993652344, |
| "eval_code_loss": 0.10408665984869003, |
| "eval_code_probabilities": [ |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999995231628418, |
| 1.0, |
| 1.0, |
| 0.9999160766601562, |
| 0.9988158941268921, |
| 0.9999899864196777, |
| 0.9999986886978149, |
| 0.9999879598617554, |
| 0.9999833106994629, |
| 0.9999370574951172, |
| 0.9992890357971191, |
| 0.9659052491188049, |
| 0.6254108548164368, |
| 0.997655987739563, |
| 0.9994457364082336, |
| 0.999336302280426, |
| 0.9953096508979797, |
| 0.9922773838043213, |
| 0.9999829530715942, |
| 0.9999885559082031, |
| 0.9871056079864502, |
| 0.9849441647529602, |
| 0.9815734028816223, |
| 0.7201115489006042, |
| 0.6514120101928711, |
| 0.8476458191871643, |
| 0.1292794942855835, |
| 0.9985151886940002, |
| 0.9995579123497009, |
| 1.0, |
| 0.9999856948852539, |
| 0.9999977350234985, |
| 0.9999998807907104, |
| 1.0, |
| 1.0, |
| 0.9999991655349731, |
| 0.9998373985290527, |
| 0.999966025352478, |
| 1.0, |
| 0.9999432563781738, |
| 1.0, |
| 0.997358500957489, |
| 1.0, |
| 0.9999942779541016, |
| 0.9999959468841553, |
| 0.9998447895050049, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9860179424285889, |
| 0.8810455203056335, |
| 0.8414822816848755, |
| 1.0, |
| 0.9645673632621765, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9987167119979858, |
| 0.9995377063751221, |
| 0.969241201877594, |
| 0.9999886751174927, |
| 0.2950672507286072, |
| 0.44002464413642883, |
| 0.8467398881912231, |
| 0.06261960417032242, |
| 0.9991006851196289, |
| 0.9985753297805786, |
| 0.9999589920043945, |
| 0.9881368279457092, |
| 1.0, |
| 0.9875013828277588, |
| 0.9999996423721313, |
| 1.0, |
| 0.9999927282333374, |
| 0.9999997615814209, |
| 0.9641613960266113, |
| 0.5904451012611389, |
| 0.991640567779541, |
| 0.9999064207077026, |
| 0.996203601360321, |
| 0.999998927116394, |
| 0.9644280672073364, |
| 0.9944401979446411, |
| 0.8003358840942383, |
| 0.9963403940200806, |
| 0.9804515242576599, |
| 0.989639163017273, |
| 0.852015495300293, |
| 0.9918712377548218, |
| 0.9663655757904053, |
| 0.8847030997276306, |
| 0.5973413586616516, |
| 0.5435407757759094, |
| 0.811240017414093 |
| ], |
| "eval_code_runtime": 105.7613, |
| "eval_code_samples_per_second": 0.946, |
| "eval_code_score": -0.03531455993652344, |
| "eval_code_steps_per_second": 0.038, |
| "step": 100 |
| }, |
| { |
| "epoch": 1.77, |
| "learning_rate": 0.0002, |
| "loss": 0.0341, |
| "step": 101 |
| }, |
| { |
| "epoch": 1.79, |
| "learning_rate": 0.0002, |
| "loss": 0.0952, |
| "step": 102 |
| }, |
| { |
| "epoch": 1.81, |
| "learning_rate": 0.0002, |
| "loss": 0.0358, |
| "step": 103 |
| }, |
| { |
| "epoch": 1.82, |
| "learning_rate": 0.0002, |
| "loss": 0.1143, |
| "step": 104 |
| }, |
| { |
| "epoch": 1.84, |
| "learning_rate": 0.0002, |
| "loss": 0.07, |
| "step": 105 |
| }, |
| { |
| "epoch": 1.86, |
| "learning_rate": 0.0002, |
| "loss": 0.1232, |
| "step": 106 |
| }, |
| { |
| "epoch": 1.88, |
| "learning_rate": 0.0002, |
| "loss": 0.065, |
| "step": 107 |
| }, |
| { |
| "epoch": 1.89, |
| "learning_rate": 0.0002, |
| "loss": 0.0167, |
| "step": 108 |
| }, |
| { |
| "epoch": 1.91, |
| "learning_rate": 0.0002, |
| "loss": 0.0486, |
| "step": 109 |
| }, |
| { |
| "epoch": 1.93, |
| "learning_rate": 0.0002, |
| "loss": 0.011, |
| "step": 110 |
| }, |
| { |
| "epoch": 1.95, |
| "learning_rate": 0.0002, |
| "loss": 0.0425, |
| "step": 111 |
| }, |
| { |
| "epoch": 1.96, |
| "learning_rate": 0.0002, |
| "loss": 0.1082, |
| "step": 112 |
| }, |
| { |
| "epoch": 1.98, |
| "learning_rate": 0.0002, |
| "loss": 0.0196, |
| "step": 113 |
| }, |
| { |
| "epoch": 2.0, |
| "learning_rate": 0.0002, |
| "loss": 0.0152, |
| "step": 114 |
| }, |
| { |
| "epoch": 2.02, |
| "learning_rate": 0.0002, |
| "loss": 0.012, |
| "step": 115 |
| }, |
| { |
| "epoch": 2.04, |
| "learning_rate": 0.0002, |
| "loss": 0.0186, |
| "step": 116 |
| }, |
| { |
| "epoch": 2.05, |
| "learning_rate": 0.0002, |
| "loss": 0.0186, |
| "step": 117 |
| }, |
| { |
| "epoch": 2.07, |
| "learning_rate": 0.0002, |
| "loss": 0.0238, |
| "step": 118 |
| }, |
| { |
| "epoch": 2.09, |
| "learning_rate": 0.0002, |
| "loss": 0.0097, |
| "step": 119 |
| }, |
| { |
| "epoch": 2.11, |
| "learning_rate": 0.0002, |
| "loss": 0.0007, |
| "step": 120 |
| }, |
| { |
| "epoch": 2.12, |
| "learning_rate": 0.0002, |
| "loss": 0.0409, |
| "step": 121 |
| }, |
| { |
| "epoch": 2.14, |
| "learning_rate": 0.0002, |
| "loss": 0.0238, |
| "step": 122 |
| }, |
| { |
| "epoch": 2.16, |
| "learning_rate": 0.0002, |
| "loss": 0.0045, |
| "step": 123 |
| }, |
| { |
| "epoch": 2.18, |
| "learning_rate": 0.0002, |
| "loss": 0.0008, |
| "step": 124 |
| }, |
| { |
| "epoch": 2.19, |
| "learning_rate": 0.0002, |
| "loss": 0.0307, |
| "step": 125 |
| }, |
| { |
| "epoch": 2.19, |
| "eval_code_accuracy": 0.96, |
| "eval_code_average_probability": 0.9499993324279785, |
| "eval_code_brier_score": 0.028195565566420555, |
| "eval_code_loss": 0.11333052814006805, |
| "eval_code_probabilities": [ |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999997615814209, |
| 1.0, |
| 1.0, |
| 0.9999997615814209, |
| 0.9999997615814209, |
| 0.9999985694885254, |
| 0.9999561309814453, |
| 0.8770335912704468, |
| 0.4641159176826477, |
| 0.9271451234817505, |
| 1.0, |
| 1.0, |
| 0.9999996423721313, |
| 0.9998699426651001, |
| 0.9999997615814209, |
| 0.9999998807907104, |
| 0.9999991655349731, |
| 0.9999380111694336, |
| 0.9999819993972778, |
| 0.7533411979675293, |
| 0.7846699357032776, |
| 0.89438796043396, |
| 0.015444471500813961, |
| 0.999987006187439, |
| 1.0, |
| 1.0, |
| 0.9969304203987122, |
| 0.9999985694885254, |
| 0.9999998807907104, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999716281890869, |
| 0.9999990463256836, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9998655319213867, |
| 1.0, |
| 0.9999997615814209, |
| 0.9999998807907104, |
| 0.9999963045120239, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9978950023651123, |
| 0.7412748336791992, |
| 0.8329392671585083, |
| 1.0, |
| 0.999669075012207, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999936819076538, |
| 0.9999263286590576, |
| 0.9996693134307861, |
| 1.0, |
| 0.9178900122642517, |
| 0.8888350129127502, |
| 0.9699975848197937, |
| 0.014796692878007889, |
| 0.9991298317909241, |
| 0.9968131184577942, |
| 0.9999946355819702, |
| 0.9282417297363281, |
| 1.0, |
| 0.9971838593482971, |
| 1.0, |
| 1.0, |
| 0.9999544620513916, |
| 0.9999992847442627, |
| 0.9987694621086121, |
| 0.8008559346199036, |
| 0.9995920062065125, |
| 1.0, |
| 0.9999116659164429, |
| 1.0, |
| 0.9975823163986206, |
| 0.9999178647994995, |
| 0.9812232255935669, |
| 0.9962044358253479, |
| 0.9992068409919739, |
| 0.9994266033172607, |
| 0.9672200679779053, |
| 0.998022198677063, |
| 0.9971696734428406, |
| 0.9978540539741516, |
| 0.9273316860198975, |
| 0.8601265549659729, |
| 0.4826805889606476 |
| ], |
| "eval_code_runtime": 105.781, |
| "eval_code_samples_per_second": 0.945, |
| "eval_code_score": -0.028195565566420555, |
| "eval_code_steps_per_second": 0.038, |
| "step": 125 |
| }, |
| { |
| "epoch": 2.21, |
| "learning_rate": 0.0002, |
| "loss": 0.0048, |
| "step": 126 |
| }, |
| { |
| "epoch": 2.23, |
| "learning_rate": 0.0002, |
| "loss": 0.0093, |
| "step": 127 |
| }, |
| { |
| "epoch": 2.25, |
| "learning_rate": 0.0002, |
| "loss": 0.0056, |
| "step": 128 |
| }, |
| { |
| "epoch": 2.26, |
| "learning_rate": 0.0002, |
| "loss": 0.0123, |
| "step": 129 |
| }, |
| { |
| "epoch": 2.28, |
| "learning_rate": 0.0002, |
| "loss": 0.0825, |
| "step": 130 |
| }, |
| { |
| "epoch": 2.3, |
| "learning_rate": 0.0002, |
| "loss": 0.0199, |
| "step": 131 |
| }, |
| { |
| "epoch": 2.32, |
| "learning_rate": 0.0002, |
| "loss": 0.025, |
| "step": 132 |
| }, |
| { |
| "epoch": 2.33, |
| "learning_rate": 0.0002, |
| "loss": 0.0477, |
| "step": 133 |
| }, |
| { |
| "epoch": 2.35, |
| "learning_rate": 0.0002, |
| "loss": 0.0403, |
| "step": 134 |
| }, |
| { |
| "epoch": 2.37, |
| "learning_rate": 0.0002, |
| "loss": 0.0209, |
| "step": 135 |
| }, |
| { |
| "epoch": 2.39, |
| "learning_rate": 0.0002, |
| "loss": 0.0004, |
| "step": 136 |
| }, |
| { |
| "epoch": 2.4, |
| "learning_rate": 0.0002, |
| "loss": 0.0003, |
| "step": 137 |
| }, |
| { |
| "epoch": 2.42, |
| "learning_rate": 0.0002, |
| "loss": 0.024, |
| "step": 138 |
| }, |
| { |
| "epoch": 2.44, |
| "learning_rate": 0.0002, |
| "loss": 0.0215, |
| "step": 139 |
| }, |
| { |
| "epoch": 2.46, |
| "learning_rate": 0.0002, |
| "loss": 0.0321, |
| "step": 140 |
| }, |
| { |
| "epoch": 2.47, |
| "learning_rate": 0.0002, |
| "loss": 0.0067, |
| "step": 141 |
| }, |
| { |
| "epoch": 2.49, |
| "learning_rate": 0.0002, |
| "loss": 0.0144, |
| "step": 142 |
| }, |
| { |
| "epoch": 2.51, |
| "learning_rate": 0.0002, |
| "loss": 0.0017, |
| "step": 143 |
| }, |
| { |
| "epoch": 2.53, |
| "learning_rate": 0.0002, |
| "loss": 0.025, |
| "step": 144 |
| }, |
| { |
| "epoch": 2.54, |
| "learning_rate": 0.0002, |
| "loss": 0.0605, |
| "step": 145 |
| }, |
| { |
| "epoch": 2.56, |
| "learning_rate": 0.0002, |
| "loss": 0.0154, |
| "step": 146 |
| }, |
| { |
| "epoch": 2.58, |
| "learning_rate": 0.0002, |
| "loss": 0.002, |
| "step": 147 |
| }, |
| { |
| "epoch": 2.6, |
| "learning_rate": 0.0002, |
| "loss": 0.0012, |
| "step": 148 |
| }, |
| { |
| "epoch": 2.61, |
| "learning_rate": 0.0002, |
| "loss": 0.0066, |
| "step": 149 |
| }, |
| { |
| "epoch": 2.63, |
| "learning_rate": 0.0002, |
| "loss": 0.0065, |
| "step": 150 |
| }, |
| { |
| "epoch": 2.63, |
| "eval_code_accuracy": 0.95, |
| "eval_code_average_probability": 0.9435591101646423, |
| "eval_code_brier_score": 0.03846995532512665, |
| "eval_code_loss": 0.17103993892669678, |
| "eval_code_probabilities": [ |
| 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, |
| 0.9999970197677612, |
| 0.9055734872817993, |
| 0.42537397146224976, |
| 0.9499343633651733, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999983310699463, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999964237213135, |
| 0.9999994039535522, |
| 0.8147205710411072, |
| 0.8424564599990845, |
| 0.9474385976791382, |
| 0.003861561883240938, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.035812485963106155, |
| 0.9962039589881897, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999990463256836, |
| 0.9999998807907104, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999754428863525, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999997615814209, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9926579594612122, |
| 0.7481365203857422, |
| 0.9441766142845154, |
| 1.0, |
| 0.9999790191650391, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999998807907104, |
| 0.9999998807907104, |
| 0.9856025576591492, |
| 1.0, |
| 0.9969327449798584, |
| 0.9738338589668274, |
| 0.9861930012702942, |
| 0.007046875543892384, |
| 0.9993067979812622, |
| 0.9975935816764832, |
| 0.9999728202819824, |
| 0.3661898970603943, |
| 1.0, |
| 0.9999417066574097, |
| 1.0, |
| 1.0, |
| 0.9999967813491821, |
| 1.0, |
| 0.9991926550865173, |
| 0.8271266222000122, |
| 0.9992584586143494, |
| 1.0, |
| 0.9992876648902893, |
| 1.0, |
| 0.9999542236328125, |
| 0.9999967813491821, |
| 0.9966945648193359, |
| 0.9999995231628418, |
| 0.9955873489379883, |
| 0.9999998807907104, |
| 0.8881388306617737, |
| 0.9999102354049683, |
| 0.9994561076164246, |
| 0.9989873766899109, |
| 0.9133449792861938, |
| 0.9107068181037903, |
| 0.9093677401542664 |
| ], |
| "eval_code_runtime": 105.8053, |
| "eval_code_samples_per_second": 0.945, |
| "eval_code_score": -0.03846995532512665, |
| "eval_code_steps_per_second": 0.038, |
| "step": 150 |
| }, |
| { |
| "loss": 0.0046, |
| "learning_rate": 0.0002, |
| "epoch": 2.65, |
| "step": 151 |
| }, |
| { |
| "loss": 0.0011, |
| "learning_rate": 0.0002, |
| "epoch": 2.67, |
| "step": 152 |
| }, |
| { |
| "loss": 0.0004, |
| "learning_rate": 0.0002, |
| "epoch": 2.68, |
| "step": 153 |
| }, |
| { |
| "loss": 0.0009, |
| "learning_rate": 0.0002, |
| "epoch": 2.7, |
| "step": 154 |
| }, |
| { |
| "loss": 0.0167, |
| "learning_rate": 0.0002, |
| "epoch": 2.72, |
| "step": 155 |
| }, |
| { |
| "loss": 0.0033, |
| "learning_rate": 0.0002, |
| "epoch": 2.74, |
| "step": 156 |
| }, |
| { |
| "loss": 0.0004, |
| "learning_rate": 0.0002, |
| "epoch": 2.75, |
| "step": 157 |
| }, |
| { |
| "loss": 0.0013, |
| "learning_rate": 0.0002, |
| "epoch": 2.77, |
| "step": 158 |
| }, |
| { |
| "loss": 0.0003, |
| "learning_rate": 0.0002, |
| "epoch": 2.79, |
| "step": 159 |
| }, |
| { |
| "loss": 0.0005, |
| "learning_rate": 0.0002, |
| "epoch": 2.81, |
| "step": 160 |
| }, |
| { |
| "loss": 0.002, |
| "learning_rate": 0.0002, |
| "epoch": 2.82, |
| "step": 161 |
| }, |
| { |
| "loss": 0.0001, |
| "learning_rate": 0.0002, |
| "epoch": 2.84, |
| "step": 162 |
| }, |
| { |
| "loss": 0.0052, |
| "learning_rate": 0.0002, |
| "epoch": 2.86, |
| "step": 163 |
| }, |
| { |
| "loss": 0.0003, |
| "learning_rate": 0.0002, |
| "epoch": 2.88, |
| "step": 164 |
| }, |
| { |
| "loss": 0.0, |
| "learning_rate": 0.0002, |
| "epoch": 2.89, |
| "step": 165 |
| }, |
| { |
| "loss": 0.0043, |
| "learning_rate": 0.0002, |
| "epoch": 2.91, |
| "step": 166 |
| }, |
| { |
| "loss": 0.0, |
| "learning_rate": 0.0002, |
| "epoch": 2.93, |
| "step": 167 |
| }, |
| { |
| "loss": 0.0011, |
| "learning_rate": 0.0002, |
| "epoch": 2.95, |
| "step": 168 |
| }, |
| { |
| "loss": 0.0024, |
| "learning_rate": 0.0002, |
| "epoch": 2.96, |
| "step": 169 |
| }, |
| { |
| "loss": 0.0, |
| "learning_rate": 0.0002, |
| "epoch": 2.98, |
| "step": 170 |
| }, |
| { |
| "loss": 0.0009, |
| "learning_rate": 0.0002, |
| "epoch": 3.0, |
| "step": 171 |
| }, |
| { |
| "loss": 0.0008, |
| "learning_rate": 0.0002, |
| "epoch": 3.02, |
| "step": 172 |
| }, |
| { |
| "loss": 0.0007, |
| "learning_rate": 0.0002, |
| "epoch": 3.04, |
| "step": 173 |
| }, |
| { |
| "loss": 0.0001, |
| "learning_rate": 0.0002, |
| "epoch": 3.05, |
| "step": 174 |
| }, |
| { |
| "loss": 0.0182, |
| "learning_rate": 0.0002, |
| "epoch": 3.07, |
| "step": 175 |
| }, |
| { |
| "eval_code_loss": 0.24447500705718994, |
| "eval_code_score": -0.04580119252204895, |
| "eval_code_brier_score": 0.04580119252204895, |
| "eval_code_average_probability": 0.9384365677833557, |
| "eval_code_accuracy": 0.95, |
| "eval_code_probabilities": [ |
| 1.0, |
| 0.9999996423721313, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999998807907104, |
| 0.9705566167831421, |
| 0.28825750946998596, |
| 0.990113377571106, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.8121438026428223, |
| 0.8320725560188293, |
| 0.9545630216598511, |
| 0.022090904414653778, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.0003882711462210864, |
| 0.9997542500495911, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999994039535522, |
| 0.9999996423721313, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999974966049194, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9805604219436646, |
| 0.656412661075592, |
| 0.9261404871940613, |
| 1.0, |
| 0.9999086856842041, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999998807907104, |
| 1.0, |
| 0.9739339351654053, |
| 1.0, |
| 0.9995070695877075, |
| 0.9991480112075806, |
| 0.9978482723236084, |
| 0.0005162784364074469, |
| 0.999909520149231, |
| 1.0, |
| 1.0, |
| 0.09138258546590805, |
| 1.0, |
| 0.9999922513961792, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9993377327919006, |
| 0.829042375087738, |
| 0.993126630783081, |
| 1.0, |
| 0.9995529055595398, |
| 1.0, |
| 0.9999717473983765, |
| 0.9999934434890747, |
| 0.9994831085205078, |
| 1.0, |
| 0.9715592861175537, |
| 1.0, |
| 0.7726230025291443, |
| 0.9999821186065674, |
| 0.9998100399971008, |
| 0.9989909529685974, |
| 0.8984997272491455, |
| 0.8994256854057312, |
| 0.9870588183403015 |
| ], |
| "eval_code_runtime": 104.8369, |
| "eval_code_samples_per_second": 0.954, |
| "eval_code_steps_per_second": 0.038, |
| "epoch": 3.07, |
| "step": 175 |
| }, |
| { |
| "loss": 0.0027, |
| "learning_rate": 0.0002, |
| "epoch": 3.09, |
| "step": 176 |
| }, |
| { |
| "loss": 0.0001, |
| "learning_rate": 0.0002, |
| "epoch": 3.11, |
| "step": 177 |
| }, |
| { |
| "loss": 0.0005, |
| "learning_rate": 0.0002, |
| "epoch": 3.12, |
| "step": 178 |
| }, |
| { |
| "loss": 0.0002, |
| "learning_rate": 0.0002, |
| "epoch": 3.14, |
| "step": 179 |
| }, |
| { |
| "loss": 0.0001, |
| "learning_rate": 0.0002, |
| "epoch": 3.16, |
| "step": 180 |
| }, |
| { |
| "loss": 0.0, |
| "learning_rate": 0.0002, |
| "epoch": 3.18, |
| "step": 181 |
| }, |
| { |
| "loss": 0.0005, |
| "learning_rate": 0.0002, |
| "epoch": 3.19, |
| "step": 182 |
| }, |
| { |
| "loss": 0.001, |
| "learning_rate": 0.0002, |
| "epoch": 3.21, |
| "step": 183 |
| }, |
| { |
| "loss": 0.0, |
| "learning_rate": 0.0002, |
| "epoch": 3.23, |
| "step": 184 |
| }, |
| { |
| "loss": 0.0004, |
| "learning_rate": 0.0002, |
| "epoch": 3.25, |
| "step": 185 |
| }, |
| { |
| "loss": 0.0008, |
| "learning_rate": 0.0002, |
| "epoch": 3.26, |
| "step": 186 |
| }, |
| { |
| "loss": 0.0315, |
| "learning_rate": 0.0002, |
| "epoch": 3.28, |
| "step": 187 |
| }, |
| { |
| "loss": 0.0148, |
| "learning_rate": 0.0002, |
| "epoch": 3.3, |
| "step": 188 |
| }, |
| { |
| "loss": 0.0, |
| "learning_rate": 0.0002, |
| "epoch": 3.32, |
| "step": 189 |
| }, |
| { |
| "loss": 0.0016, |
| "learning_rate": 0.0002, |
| "epoch": 3.33, |
| "step": 190 |
| }, |
| { |
| "loss": 0.0053, |
| "learning_rate": 0.0002, |
| "epoch": 3.35, |
| "step": 191 |
| }, |
| { |
| "loss": 0.0001, |
| "learning_rate": 0.0002, |
| "epoch": 3.37, |
| "step": 192 |
| }, |
| { |
| "loss": 0.0009, |
| "learning_rate": 0.0002, |
| "epoch": 3.39, |
| "step": 193 |
| }, |
| { |
| "loss": 0.1843, |
| "learning_rate": 0.0002, |
| "epoch": 3.4, |
| "step": 194 |
| }, |
| { |
| "loss": 0.0001, |
| "learning_rate": 0.0002, |
| "epoch": 3.42, |
| "step": 195 |
| }, |
| { |
| "loss": 0.012, |
| "learning_rate": 0.0002, |
| "epoch": 3.44, |
| "step": 196 |
| }, |
| { |
| "loss": 0.0031, |
| "learning_rate": 0.0002, |
| "epoch": 3.46, |
| "step": 197 |
| }, |
| { |
| "loss": 0.0003, |
| "learning_rate": 0.0002, |
| "epoch": 3.47, |
| "step": 198 |
| }, |
| { |
| "loss": 0.0062, |
| "learning_rate": 0.0002, |
| "epoch": 3.49, |
| "step": 199 |
| }, |
| { |
| "loss": 0.0011, |
| "learning_rate": 0.0002, |
| "epoch": 3.51, |
| "step": 200 |
| }, |
| { |
| "eval_code_loss": 0.20703525841236115, |
| "eval_code_score": -0.036074623465538025, |
| "eval_code_brier_score": 0.036074623465538025, |
| "eval_code_average_probability": 0.9416353702545166, |
| "eval_code_accuracy": 0.96, |
| "eval_code_probabilities": [ |
| 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, |
| 0.9999991655349731, |
| 0.999995231628418, |
| 0.9498335719108582, |
| 0.30828458070755005, |
| 0.987034261226654, |
| 0.999970555305481, |
| 0.9999929666519165, |
| 0.9999411106109619, |
| 0.999946117401123, |
| 0.9999997615814209, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.717918872833252, |
| 0.6784900426864624, |
| 0.8770557641983032, |
| 7.296191597561119e-06, |
| 0.9999998807907104, |
| 0.9999997615814209, |
| 1.0, |
| 0.9999603033065796, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999985694885254, |
| 0.9999916553497314, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999998807907104, |
| 1.0, |
| 1.0, |
| 0.9999997615814209, |
| 0.9999793767929077, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9963011741638184, |
| 0.5810772180557251, |
| 0.9937513470649719, |
| 1.0, |
| 0.9999946355819702, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999998807907104, |
| 0.9999996423721313, |
| 0.999351441860199, |
| 1.0, |
| 0.9979580640792847, |
| 0.9855725169181824, |
| 0.9849231243133545, |
| 0.017488110810518265, |
| 0.999018669128418, |
| 0.9788364768028259, |
| 0.9989467263221741, |
| 0.29412388801574707, |
| 1.0, |
| 0.9881665110588074, |
| 1.0, |
| 1.0, |
| 0.9999973773956299, |
| 1.0, |
| 0.9999986886978149, |
| 0.8509193658828735, |
| 0.9998080134391785, |
| 1.0, |
| 0.9999957084655762, |
| 1.0, |
| 0.9992415904998779, |
| 0.9998071789741516, |
| 0.9979315996170044, |
| 0.9996376037597656, |
| 0.9590405821800232, |
| 0.9996711015701294, |
| 0.5990511775016785, |
| 0.9998351335525513, |
| 0.9997956156730652, |
| 0.9722345471382141, |
| 0.7871147990226746, |
| 0.8062353730201721, |
| 0.8593048453330994 |
| ], |
| "eval_code_runtime": 104.8009, |
| "eval_code_samples_per_second": 0.954, |
| "eval_code_steps_per_second": 0.038, |
| "epoch": 3.51, |
| "step": 200 |
| }, |
| { |
| "loss": 0.0071, |
| "learning_rate": 0.0002, |
| "epoch": 3.53, |
| "step": 201 |
| }, |
| { |
| "loss": 0.0311, |
| "learning_rate": 0.0002, |
| "epoch": 3.54, |
| "step": 202 |
| }, |
| { |
| "loss": 0.0009, |
| "learning_rate": 0.0002, |
| "epoch": 3.56, |
| "step": 203 |
| }, |
| { |
| "loss": 0.0003, |
| "learning_rate": 0.0002, |
| "epoch": 3.58, |
| "step": 204 |
| }, |
| { |
| "loss": 0.0004, |
| "learning_rate": 0.0002, |
| "epoch": 3.6, |
| "step": 205 |
| }, |
| { |
| "loss": 0.0018, |
| "learning_rate": 0.0002, |
| "epoch": 3.61, |
| "step": 206 |
| }, |
| { |
| "loss": 0.0065, |
| "learning_rate": 0.0002, |
| "epoch": 3.63, |
| "step": 207 |
| }, |
| { |
| "loss": 0.001, |
| "learning_rate": 0.0002, |
| "epoch": 3.65, |
| "step": 208 |
| }, |
| { |
| "loss": 0.0, |
| "learning_rate": 0.0002, |
| "epoch": 3.67, |
| "step": 209 |
| }, |
| { |
| "loss": 0.0302, |
| "learning_rate": 0.0002, |
| "epoch": 3.68, |
| "step": 210 |
| }, |
| { |
| "loss": 0.0662, |
| "learning_rate": 0.0002, |
| "epoch": 3.7, |
| "step": 211 |
| }, |
| { |
| "loss": 0.0071, |
| "learning_rate": 0.0002, |
| "epoch": 3.72, |
| "step": 212 |
| }, |
| { |
| "loss": 0.0001, |
| "learning_rate": 0.0002, |
| "epoch": 3.74, |
| "step": 213 |
| }, |
| { |
| "loss": 0.0527, |
| "learning_rate": 0.0002, |
| "epoch": 3.75, |
| "step": 214 |
| }, |
| { |
| "loss": 0.0069, |
| "learning_rate": 0.0002, |
| "epoch": 3.77, |
| "step": 215 |
| }, |
| { |
| "loss": 0.0002, |
| "learning_rate": 0.0002, |
| "epoch": 3.79, |
| "step": 216 |
| }, |
| { |
| "loss": 0.0056, |
| "learning_rate": 0.0002, |
| "epoch": 3.81, |
| "step": 217 |
| }, |
| { |
| "loss": 0.0205, |
| "learning_rate": 0.0002, |
| "epoch": 3.82, |
| "step": 218 |
| }, |
| { |
| "loss": 0.0, |
| "learning_rate": 0.0002, |
| "epoch": 3.84, |
| "step": 219 |
| }, |
| { |
| "loss": 0.0048, |
| "learning_rate": 0.0002, |
| "epoch": 3.86, |
| "step": 220 |
| }, |
| { |
| "loss": 0.0307, |
| "learning_rate": 0.0002, |
| "epoch": 3.88, |
| "step": 221 |
| }, |
| { |
| "loss": 0.0092, |
| "learning_rate": 0.0002, |
| "epoch": 3.89, |
| "step": 222 |
| }, |
| { |
| "loss": 0.0425, |
| "learning_rate": 0.0002, |
| "epoch": 3.91, |
| "step": 223 |
| }, |
| { |
| "loss": 0.0005, |
| "learning_rate": 0.0002, |
| "epoch": 3.93, |
| "step": 224 |
| }, |
| { |
| "loss": 0.0021, |
| "learning_rate": 0.0002, |
| "epoch": 3.95, |
| "step": 225 |
| }, |
| { |
| "eval_code_loss": 0.1442316770553589, |
| "eval_code_score": -0.03884093835949898, |
| "eval_code_brier_score": 0.03884093835949898, |
| "eval_code_average_probability": 0.9314249157905579, |
| "eval_code_accuracy": 0.95, |
| "eval_code_probabilities": [ |
| 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, |
| 0.9999998807907104, |
| 0.9999983310699463, |
| 0.9166925549507141, |
| 0.46314355731010437, |
| 0.9695256948471069, |
| 0.9999979734420776, |
| 0.9999997615814209, |
| 0.9999912977218628, |
| 0.9998043179512024, |
| 0.9999998807907104, |
| 1.0, |
| 0.999934196472168, |
| 0.9998651742935181, |
| 0.9999654293060303, |
| 0.7587816119194031, |
| 0.8365660309791565, |
| 0.8947399258613586, |
| 0.007870269939303398, |
| 0.9999408721923828, |
| 0.9807198643684387, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999988079071045, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999998807907104, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9998663663864136, |
| 0.9656995534896851, |
| 0.935349702835083, |
| 1.0, |
| 0.9999972581863403, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999998807907104, |
| 0.9999998807907104, |
| 0.9999998807907104, |
| 0.9994795918464661, |
| 0.9999734163284302, |
| 0.4865659773349762, |
| 0.5832042694091797, |
| 0.9857308268547058, |
| 0.015312162227928638, |
| 0.999519944190979, |
| 0.7279074788093567, |
| 0.9087716937065125, |
| 0.6908769011497498, |
| 1.0, |
| 0.9999642372131348, |
| 1.0, |
| 1.0, |
| 0.9999998807907104, |
| 1.0, |
| 1.0, |
| 0.8065006136894226, |
| 0.9999997615814209, |
| 1.0, |
| 0.999992847442627, |
| 1.0, |
| 0.998808741569519, |
| 0.9994845390319824, |
| 0.9905843734741211, |
| 0.9999678134918213, |
| 0.9825097322463989, |
| 0.9999821186065674, |
| 0.316112756729126, |
| 0.9999034404754639, |
| 0.9957683086395264, |
| 0.8763726353645325, |
| 0.557384192943573, |
| 0.5571630001068115, |
| 0.936220109462738 |
| ], |
| "eval_code_runtime": 104.8083, |
| "eval_code_samples_per_second": 0.954, |
| "eval_code_steps_per_second": 0.038, |
| "epoch": 3.95, |
| "step": 225 |
| }, |
| { |
| "loss": 0.0037, |
| "learning_rate": 0.0002, |
| "epoch": 3.96, |
| "step": 226 |
| }, |
| { |
| "loss": 0.0061, |
| "learning_rate": 0.0002, |
| "epoch": 3.98, |
| "step": 227 |
| }, |
| { |
| "loss": 0.0211, |
| "learning_rate": 0.0002, |
| "epoch": 4.0, |
| "step": 228 |
| }, |
| { |
| "loss": 0.0028, |
| "learning_rate": 0.0002, |
| "epoch": 4.02, |
| "step": 229 |
| }, |
| { |
| "loss": 0.0009, |
| "learning_rate": 0.0002, |
| "epoch": 4.04, |
| "step": 230 |
| }, |
| { |
| "loss": 0.0146, |
| "learning_rate": 0.0002, |
| "epoch": 4.05, |
| "step": 231 |
| }, |
| { |
| "loss": 0.0155, |
| "learning_rate": 0.0002, |
| "epoch": 4.07, |
| "step": 232 |
| }, |
| { |
| "loss": 0.0008, |
| "learning_rate": 0.0002, |
| "epoch": 4.09, |
| "step": 233 |
| }, |
| { |
| "loss": 0.0008, |
| "learning_rate": 0.0002, |
| "epoch": 4.11, |
| "step": 234 |
| }, |
| { |
| "loss": 0.0001, |
| "learning_rate": 0.0002, |
| "epoch": 4.12, |
| "step": 235 |
| }, |
| { |
| "loss": 0.0004, |
| "learning_rate": 0.0002, |
| "epoch": 4.14, |
| "step": 236 |
| }, |
| { |
| "loss": 0.0073, |
| "learning_rate": 0.0002, |
| "epoch": 4.16, |
| "step": 237 |
| }, |
| { |
| "loss": 0.0005, |
| "learning_rate": 0.0002, |
| "epoch": 4.18, |
| "step": 238 |
| }, |
| { |
| "loss": 0.0011, |
| "learning_rate": 0.0002, |
| "epoch": 4.19, |
| "step": 239 |
| }, |
| { |
| "loss": 0.0004, |
| "learning_rate": 0.0002, |
| "epoch": 4.21, |
| "step": 240 |
| }, |
| { |
| "loss": 0.0001, |
| "learning_rate": 0.0002, |
| "epoch": 4.23, |
| "step": 241 |
| }, |
| { |
| "loss": 0.0005, |
| "learning_rate": 0.0002, |
| "epoch": 4.25, |
| "step": 242 |
| }, |
| { |
| "loss": 0.0005, |
| "learning_rate": 0.0002, |
| "epoch": 4.26, |
| "step": 243 |
| }, |
| { |
| "loss": 0.0009, |
| "learning_rate": 0.0002, |
| "epoch": 4.28, |
| "step": 244 |
| }, |
| { |
| "loss": 0.0026, |
| "learning_rate": 0.0002, |
| "epoch": 4.3, |
| "step": 245 |
| }, |
| { |
| "loss": 0.0003, |
| "learning_rate": 0.0002, |
| "epoch": 4.32, |
| "step": 246 |
| }, |
| { |
| "loss": 0.0, |
| "learning_rate": 0.0002, |
| "epoch": 4.33, |
| "step": 247 |
| }, |
| { |
| "loss": 0.0172, |
| "learning_rate": 0.0002, |
| "epoch": 4.35, |
| "step": 248 |
| }, |
| { |
| "loss": 0.0003, |
| "learning_rate": 0.0002, |
| "epoch": 4.37, |
| "step": 249 |
| }, |
| { |
| "loss": 0.0003, |
| "learning_rate": 0.0002, |
| "epoch": 4.39, |
| "step": 250 |
| }, |
| { |
| "eval_code_loss": 0.1744336634874344, |
| "eval_code_score": -0.03369342163205147, |
| "eval_code_brier_score": 0.03369342163205147, |
| "eval_code_average_probability": 0.9495358467102051, |
| "eval_code_accuracy": 0.97, |
| "eval_code_probabilities": [ |
| 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, |
| 0.9451481699943542, |
| 0.5480159521102905, |
| 0.9964591860771179, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.7053650617599487, |
| 0.8087413311004639, |
| 0.9013814330101013, |
| 0.0008012360776774585, |
| 1.0, |
| 0.9999994039535522, |
| 1.0, |
| 0.9998193383216858, |
| 0.9999997615814209, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999983310699463, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9992625117301941, |
| 0.9287911653518677, |
| 0.872682511806488, |
| 1.0, |
| 0.9996117949485779, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999651908874512, |
| 1.0, |
| 0.9498048424720764, |
| 0.9871782064437866, |
| 0.9963241815567017, |
| 0.0026193673256784678, |
| 0.999943733215332, |
| 0.9986830353736877, |
| 0.9999673366546631, |
| 0.9674615859985352, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999785423278809, |
| 0.9016892313957214, |
| 0.9997751116752625, |
| 1.0, |
| 0.9999997615814209, |
| 1.0, |
| 0.9999815225601196, |
| 0.9999998807907104, |
| 0.9999561309814453, |
| 1.0, |
| 0.9956650137901306, |
| 1.0, |
| 0.08844359964132309, |
| 0.9999833106994629, |
| 0.9990620017051697, |
| 0.9781332612037659, |
| 0.747072160243988, |
| 0.6776840090751648, |
| 0.9581350088119507 |
| ], |
| "eval_code_runtime": 104.7916, |
| "eval_code_samples_per_second": 0.954, |
| "eval_code_steps_per_second": 0.038, |
| "epoch": 4.39, |
| "step": 250 |
| }, |
| { |
| "loss": 0.0001, |
| "learning_rate": 0.0002, |
| "epoch": 4.4, |
| "step": 251 |
| }, |
| { |
| "loss": 0.0008, |
| "learning_rate": 0.0002, |
| "epoch": 4.42, |
| "step": 252 |
| }, |
| { |
| "loss": 0.0004, |
| "learning_rate": 0.0002, |
| "epoch": 4.44, |
| "step": 253 |
| }, |
| { |
| "loss": 0.0034, |
| "learning_rate": 0.0002, |
| "epoch": 4.46, |
| "step": 254 |
| }, |
| { |
| "loss": 0.0006, |
| "learning_rate": 0.0002, |
| "epoch": 4.47, |
| "step": 255 |
| }, |
| { |
| "loss": 0.0, |
| "learning_rate": 0.0002, |
| "epoch": 4.49, |
| "step": 256 |
| }, |
| { |
| "loss": 0.0004, |
| "learning_rate": 0.0002, |
| "epoch": 4.51, |
| "step": 257 |
| }, |
| { |
| "loss": 0.0009, |
| "learning_rate": 0.0002, |
| "epoch": 4.53, |
| "step": 258 |
| }, |
| { |
| "loss": 0.0042, |
| "learning_rate": 0.0002, |
| "epoch": 4.54, |
| "step": 259 |
| }, |
| { |
| "loss": 0.0003, |
| "learning_rate": 0.0002, |
| "epoch": 4.56, |
| "step": 260 |
| }, |
| { |
| "loss": 0.0001, |
| "learning_rate": 0.0002, |
| "epoch": 4.58, |
| "step": 261 |
| }, |
| { |
| "loss": 0.0002, |
| "learning_rate": 0.0002, |
| "epoch": 4.6, |
| "step": 262 |
| }, |
| { |
| "loss": 0.0023, |
| "learning_rate": 0.0002, |
| "epoch": 4.61, |
| "step": 263 |
| }, |
| { |
| "loss": 0.0001, |
| "learning_rate": 0.0002, |
| "epoch": 4.63, |
| "step": 264 |
| }, |
| { |
| "loss": 0.0001, |
| "learning_rate": 0.0002, |
| "epoch": 4.65, |
| "step": 265 |
| }, |
| { |
| "loss": 0.0005, |
| "learning_rate": 0.0002, |
| "epoch": 4.67, |
| "step": 266 |
| }, |
| { |
| "loss": 0.0003, |
| "learning_rate": 0.0002, |
| "epoch": 4.68, |
| "step": 267 |
| }, |
| { |
| "loss": 0.0, |
| "learning_rate": 0.0002, |
| "epoch": 4.7, |
| "step": 268 |
| }, |
| { |
| "loss": 0.0, |
| "learning_rate": 0.0002, |
| "epoch": 4.72, |
| "step": 269 |
| }, |
| { |
| "loss": 0.0, |
| "learning_rate": 0.0002, |
| "epoch": 4.74, |
| "step": 270 |
| }, |
| { |
| "loss": 0.0013, |
| "learning_rate": 0.0002, |
| "epoch": 4.75, |
| "step": 271 |
| }, |
| { |
| "loss": 0.006, |
| "learning_rate": 0.0002, |
| "epoch": 4.77, |
| "step": 272 |
| }, |
| { |
| "loss": 0.0002, |
| "learning_rate": 0.0002, |
| "epoch": 4.79, |
| "step": 273 |
| }, |
| { |
| "loss": 0.0008, |
| "learning_rate": 0.0002, |
| "epoch": 4.81, |
| "step": 274 |
| }, |
| { |
| "loss": 0.0, |
| "learning_rate": 0.0002, |
| "epoch": 4.82, |
| "step": 275 |
| }, |
| { |
| "eval_code_loss": 0.18856269121170044, |
| "eval_code_score": -0.03526609018445015, |
| "eval_code_brier_score": 0.03526609018445015, |
| "eval_code_average_probability": 0.949932873249054, |
| "eval_code_accuracy": 0.96, |
| "eval_code_probabilities": [ |
| 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, |
| 0.9667010307312012, |
| 0.4576479494571686, |
| 0.9961231350898743, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.7199408411979675, |
| 0.8098323345184326, |
| 0.8810715079307556, |
| 0.00041508462163619697, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999997615814209, |
| 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, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999830722808838, |
| 0.9893618226051331, |
| 0.948550820350647, |
| 1.0, |
| 0.9997448325157166, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999991655349731, |
| 1.0, |
| 0.9401667714118958, |
| 0.9961704611778259, |
| 0.9991104006767273, |
| 0.0038628315087407827, |
| 0.9999805688858032, |
| 0.9974443912506104, |
| 0.9999669790267944, |
| 0.9498675465583801, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999998807907104, |
| 0.9562873244285583, |
| 0.9999939203262329, |
| 1.0, |
| 0.9999998807907104, |
| 1.0, |
| 0.9999774694442749, |
| 0.9999997615814209, |
| 0.9999823570251465, |
| 1.0, |
| 0.9991405010223389, |
| 1.0, |
| 0.027186574414372444, |
| 0.9999986886978149, |
| 0.9999057054519653, |
| 0.9804957509040833, |
| 0.7432851791381836, |
| 0.7164115905761719, |
| 0.9146708250045776 |
| ], |
| "eval_code_runtime": 104.8208, |
| "eval_code_samples_per_second": 0.954, |
| "eval_code_steps_per_second": 0.038, |
| "epoch": 4.82, |
| "step": 275 |
| }, |
| { |
| "loss": 0.0018, |
| "learning_rate": 0.0002, |
| "epoch": 4.84, |
| "step": 276 |
| }, |
| { |
| "loss": 0.0001, |
| "learning_rate": 0.0002, |
| "epoch": 4.86, |
| "step": 277 |
| }, |
| { |
| "loss": 0.0005, |
| "learning_rate": 0.0002, |
| "epoch": 4.88, |
| "step": 278 |
| }, |
| { |
| "loss": 0.0019, |
| "learning_rate": 0.0002, |
| "epoch": 4.89, |
| "step": 279 |
| }, |
| { |
| "loss": 0.0009, |
| "learning_rate": 0.0002, |
| "epoch": 4.91, |
| "step": 280 |
| }, |
| { |
| "loss": 0.0, |
| "learning_rate": 0.0002, |
| "epoch": 4.93, |
| "step": 281 |
| }, |
| { |
| "loss": 0.0008, |
| "learning_rate": 0.0002, |
| "epoch": 4.95, |
| "step": 282 |
| }, |
| { |
| "loss": 0.0001, |
| "learning_rate": 0.0002, |
| "epoch": 4.96, |
| "step": 283 |
| }, |
| { |
| "loss": 0.0, |
| "learning_rate": 0.0002, |
| "epoch": 4.98, |
| "step": 284 |
| }, |
| { |
| "loss": 0.0, |
| "learning_rate": 0.0002, |
| "epoch": 5.0, |
| "step": 285 |
| }, |
| { |
| "loss": 0.0021, |
| "learning_rate": 0.0002, |
| "epoch": 5.02, |
| "step": 286 |
| }, |
| { |
| "loss": 0.0, |
| "learning_rate": 0.0002, |
| "epoch": 5.04, |
| "step": 287 |
| }, |
| { |
| "loss": 0.0, |
| "learning_rate": 0.0002, |
| "epoch": 5.05, |
| "step": 288 |
| }, |
| { |
| "loss": 0.0003, |
| "learning_rate": 0.0002, |
| "epoch": 5.07, |
| "step": 289 |
| }, |
| { |
| "loss": 0.0005, |
| "learning_rate": 0.0002, |
| "epoch": 5.09, |
| "step": 290 |
| }, |
| { |
| "loss": 0.0005, |
| "learning_rate": 0.0002, |
| "epoch": 5.11, |
| "step": 291 |
| }, |
| { |
| "loss": 0.0, |
| "learning_rate": 0.0002, |
| "epoch": 5.12, |
| "step": 292 |
| }, |
| { |
| "loss": 0.0, |
| "learning_rate": 0.0002, |
| "epoch": 5.14, |
| "step": 293 |
| }, |
| { |
| "loss": 0.0, |
| "learning_rate": 0.0002, |
| "epoch": 5.16, |
| "step": 294 |
| }, |
| { |
| "loss": 0.0, |
| "learning_rate": 0.0002, |
| "epoch": 5.18, |
| "step": 295 |
| }, |
| { |
| "loss": 0.0, |
| "learning_rate": 0.0002, |
| "epoch": 5.19, |
| "step": 296 |
| }, |
| { |
| "loss": 0.0001, |
| "learning_rate": 0.0002, |
| "epoch": 5.21, |
| "step": 297 |
| }, |
| { |
| "loss": 0.0001, |
| "learning_rate": 0.0002, |
| "epoch": 5.23, |
| "step": 298 |
| }, |
| { |
| "loss": 0.0008, |
| "learning_rate": 0.0002, |
| "epoch": 5.25, |
| "step": 299 |
| }, |
| { |
| "loss": 0.0001, |
| "learning_rate": 0.0002, |
| "epoch": 5.26, |
| "step": 300 |
| }, |
| { |
| "eval_code_loss": 0.17653299868106842, |
| "eval_code_score": -0.035850197076797485, |
| "eval_code_brier_score": 0.035850197076797485, |
| "eval_code_average_probability": 0.949984610080719, |
| "eval_code_accuracy": 0.96, |
| "eval_code_probabilities": [ |
| 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, |
| 0.9785318374633789, |
| 0.42346829175949097, |
| 0.9971503615379333, |
| 1.0, |
| 1.0, |
| 0.9999997615814209, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.7221306562423706, |
| 0.8104501962661743, |
| 0.8751990795135498, |
| 0.0004177717200946063, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999998807907104, |
| 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, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9999911785125732, |
| 0.9935014843940735, |
| 0.9632013440132141, |
| 1.0, |
| 0.9997548460960388, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.999998927116394, |
| 1.0, |
| 0.9625283479690552, |
| 0.9952881336212158, |
| 0.9994906187057495, |
| 0.011133184656500816, |
| 0.9999599456787109, |
| 0.9987699389457703, |
| 0.9999761581420898, |
| 0.9492788314819336, |
| 1.0, |
| 0.9999998807907104, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 0.9763202667236328, |
| 0.9999985694885254, |
| 1.0, |
| 0.9999996423721313, |
| 1.0, |
| 0.9999876022338867, |
| 0.9999998807907104, |
| 0.9999854564666748, |
| 1.0, |
| 0.9988405108451843, |
| 1.0, |
| 0.03199751302599907, |
| 0.9999868869781494, |
| 0.9996381998062134, |
| 0.9803478121757507, |
| 0.6885609030723572, |
| 0.6760653853416443, |
| 0.9665107727050781 |
| ], |
| "eval_code_runtime": 104.8074, |
| "eval_code_samples_per_second": 0.954, |
| "eval_code_steps_per_second": 0.038, |
| "epoch": 5.26, |
| "step": 300 |
| }, |
| { |
| "train_runtime": 13610.3633, |
| "train_samples_per_second": 0.705, |
| "train_steps_per_second": 0.022, |
| "total_flos": 0.0, |
| "train_loss": 0.002647306595269659, |
| "epoch": 5.26, |
| "step": 300 |
| } |
| ] |
