| [ |
| { |
| "loss": 0.7004, |
| "learning_rate": 0.0002, |
| "epoch": 0.05, |
| "step": 1 |
| }, |
| { |
| "loss": 0.7322, |
| "learning_rate": 0.0002, |
| "epoch": 0.11, |
| "step": 2 |
| }, |
| { |
| "loss": 0.6932, |
| "learning_rate": 0.0002, |
| "epoch": 0.16, |
| "step": 3 |
| }, |
| { |
| "loss": 0.7077, |
| "learning_rate": 0.0002, |
| "epoch": 0.21, |
| "step": 4 |
| }, |
| { |
| "loss": 0.683, |
| "learning_rate": 0.0002, |
| "epoch": 0.26, |
| "step": 5 |
| }, |
| { |
| "loss": 0.7367, |
| "learning_rate": 0.0002, |
| "epoch": 0.32, |
| "step": 6 |
| }, |
| { |
| "loss": 0.6696, |
| "learning_rate": 0.0002, |
| "epoch": 0.37, |
| "step": 7 |
| }, |
| { |
| "loss": 0.6816, |
| "learning_rate": 0.0002, |
| "epoch": 0.42, |
| "step": 8 |
| }, |
| { |
| "loss": 0.7249, |
| "learning_rate": 0.0002, |
| "epoch": 0.47, |
| "step": 9 |
| }, |
| { |
| "loss": 0.7143, |
| "learning_rate": 0.0002, |
| "epoch": 0.53, |
| "step": 10 |
| }, |
| { |
| "loss": 0.6898, |
| "learning_rate": 0.0002, |
| "epoch": 0.58, |
| "step": 11 |
| }, |
| { |
| "loss": 0.6921, |
| "learning_rate": 0.0002, |
| "epoch": 0.63, |
| "step": 12 |
| }, |
| { |
| "loss": 0.6883, |
| "learning_rate": 0.0002, |
| "epoch": 0.68, |
| "step": 13 |
| }, |
| { |
| "loss": 0.6973, |
| "learning_rate": 0.0002, |
| "epoch": 0.74, |
| "step": 14 |
| }, |
| { |
| "loss": 0.7143, |
| "learning_rate": 0.0002, |
| "epoch": 0.79, |
| "step": 15 |
| }, |
| { |
| "loss": 0.6668, |
| "learning_rate": 0.0002, |
| "epoch": 0.84, |
| "step": 16 |
| }, |
| { |
| "loss": 0.7046, |
| "learning_rate": 0.0002, |
| "epoch": 0.89, |
| "step": 17 |
| }, |
| { |
| "loss": 0.6849, |
| "learning_rate": 0.0002, |
| "epoch": 0.95, |
| "step": 18 |
| }, |
| { |
| "loss": 0.6756, |
| "learning_rate": 0.0002, |
| "epoch": 1.0, |
| "step": 19 |
| }, |
| { |
| "loss": 0.6962, |
| "learning_rate": 0.0002, |
| "epoch": 1.05, |
| "step": 20 |
| }, |
| { |
| "loss": 0.648, |
| "learning_rate": 0.0002, |
| "epoch": 1.11, |
| "step": 21 |
| }, |
| { |
| "loss": 0.7082, |
| "learning_rate": 0.0002, |
| "epoch": 1.16, |
| "step": 22 |
| }, |
| { |
| "loss": 0.6706, |
| "learning_rate": 0.0002, |
| "epoch": 1.21, |
| "step": 23 |
| }, |
| { |
| "loss": 0.7363, |
| "learning_rate": 0.0002, |
| "epoch": 1.26, |
| "step": 24 |
| }, |
| { |
| "loss": 0.6666, |
| "learning_rate": 0.0002, |
| "epoch": 1.32, |
| "step": 25 |
| }, |
| { |
| "eval_code_is_correct_loss": 0.6788089275360107, |
| "eval_code_is_correct_score": -0.2422185093164444, |
| "eval_code_is_correct_brier_score": 0.2422185093164444, |
| "eval_code_is_correct_average_probability": 0.5103297829627991, |
| "eval_code_is_correct_accuracy": 0.56, |
| "eval_code_is_correct_probabilities": [ |
| 0.5360161066055298, |
| 0.5297089219093323, |
| 0.5712761282920837, |
| 0.42497631907463074, |
| 0.4960133731365204, |
| 0.5085265636444092, |
| 0.5326896905899048, |
| 0.49355220794677734, |
| 0.46554645895957947, |
| 0.42035338282585144, |
| 0.4887016713619232, |
| 0.43004417419433594, |
| 0.47448185086250305, |
| 0.5101436972618103, |
| 0.5531357526779175, |
| 0.5649287700653076, |
| 0.5931124091148376, |
| 0.45721137523651123, |
| 0.5910964012145996, |
| 0.5548848509788513, |
| 0.5117404460906982, |
| 0.5237535834312439, |
| 0.5005296468734741, |
| 0.4387059807777405, |
| 0.4531331956386566, |
| 0.4623565077781677, |
| 0.5529483556747437, |
| 0.5602871775627136, |
| 0.5290127992630005, |
| 0.48502641916275024, |
| 0.5573620200157166, |
| 0.5443336963653564, |
| 0.43417370319366455, |
| 0.571002721786499, |
| 0.5364949107170105, |
| 0.4765365421772003, |
| 0.5763438940048218, |
| 0.5167781114578247, |
| 0.4962172210216522, |
| 0.45051077008247375, |
| 0.5816547274589539, |
| 0.42688706517219543, |
| 0.5689108371734619, |
| 0.4981440007686615, |
| 0.5417700409889221, |
| 0.5569514036178589, |
| 0.47263190150260925, |
| 0.4741884469985962, |
| 0.4526112377643585, |
| 0.48570743203163147, |
| 0.46209368109703064, |
| 0.5971618294715881, |
| 0.4813869595527649, |
| 0.4937862157821655, |
| 0.5675031542778015, |
| 0.5264382362365723, |
| 0.47280600666999817, |
| 0.5295760035514832, |
| 0.4977070987224579, |
| 0.5580363869667053, |
| 0.5485188364982605, |
| 0.49979132413864136, |
| 0.5156800150871277, |
| 0.5864205956459045, |
| 0.5867674350738525, |
| 0.5723587274551392, |
| 0.5590366125106812, |
| 0.49105167388916016, |
| 0.5177514553070068, |
| 0.44017159938812256, |
| 0.6072307229042053, |
| 0.5222448110580444, |
| 0.5204848051071167, |
| 0.5057241916656494, |
| 0.4764748811721802, |
| 0.5142781734466553, |
| 0.5178312659263611, |
| 0.4987412095069885, |
| 0.5352400541305542, |
| 0.4117741882801056, |
| 0.5038720965385437, |
| 0.5204145312309265, |
| 0.4230634272098541, |
| 0.45601940155029297, |
| 0.5815764665603638, |
| 0.5216283202171326, |
| 0.5572896599769592, |
| 0.4941575825214386, |
| 0.4914657771587372, |
| 0.545746386051178, |
| 0.5276685953140259, |
| 0.4675300419330597, |
| 0.47495853900909424, |
| 0.4216349720954895, |
| 0.6105045676231384, |
| 0.4074248671531677, |
| 0.48389774560928345, |
| 0.43395885825157166, |
| 0.554872453212738, |
| 0.5081204175949097 |
| ], |
| "eval_code_is_correct_runtime": 67.7492, |
| "eval_code_is_correct_samples_per_second": 1.476, |
| "eval_code_is_correct_steps_per_second": 0.059, |
| "epoch": 1.32, |
| "step": 25 |
| }, |
| { |
| "loss": 0.6611, |
| "learning_rate": 0.0002, |
| "epoch": 1.37, |
| "step": 26 |
| }, |
| { |
| "loss": 0.6909, |
| "learning_rate": 0.0002, |
| "epoch": 1.42, |
| "step": 27 |
| }, |
| { |
| "loss": 0.6235, |
| "learning_rate": 0.0002, |
| "epoch": 1.47, |
| "step": 28 |
| }, |
| { |
| "loss": 0.7308, |
| "learning_rate": 0.0002, |
| "epoch": 1.53, |
| "step": 29 |
| }, |
| { |
| "loss": 0.7595, |
| "learning_rate": 0.0002, |
| "epoch": 1.58, |
| "step": 30 |
| }, |
| { |
| "loss": 0.6369, |
| "learning_rate": 0.0002, |
| "epoch": 1.63, |
| "step": 31 |
| }, |
| { |
| "loss": 0.6853, |
| "learning_rate": 0.0002, |
| "epoch": 1.68, |
| "step": 32 |
| }, |
| { |
| "loss": 0.6916, |
| "learning_rate": 0.0002, |
| "epoch": 1.74, |
| "step": 33 |
| }, |
| { |
| "loss": 0.5824, |
| "learning_rate": 0.0002, |
| "epoch": 1.79, |
| "step": 34 |
| }, |
| { |
| "loss": 0.6258, |
| "learning_rate": 0.0002, |
| "epoch": 1.84, |
| "step": 35 |
| }, |
| { |
| "loss": 0.7912, |
| "learning_rate": 0.0002, |
| "epoch": 1.89, |
| "step": 36 |
| }, |
| { |
| "loss": 0.6382, |
| "learning_rate": 0.0002, |
| "epoch": 1.95, |
| "step": 37 |
| }, |
| { |
| "loss": 0.9123, |
| "learning_rate": 0.0002, |
| "epoch": 2.0, |
| "step": 38 |
| }, |
| { |
| "loss": 0.5432, |
| "learning_rate": 0.0002, |
| "epoch": 2.05, |
| "step": 39 |
| }, |
| { |
| "loss": 0.6325, |
| "learning_rate": 0.0002, |
| "epoch": 2.11, |
| "step": 40 |
| }, |
| { |
| "loss": 0.6731, |
| "learning_rate": 0.0002, |
| "epoch": 2.16, |
| "step": 41 |
| }, |
| { |
| "loss": 0.5489, |
| "learning_rate": 0.0002, |
| "epoch": 2.21, |
| "step": 42 |
| }, |
| { |
| "loss": 0.6372, |
| "learning_rate": 0.0002, |
| "epoch": 2.26, |
| "step": 43 |
| }, |
| { |
| "loss": 0.7034, |
| "learning_rate": 0.0002, |
| "epoch": 2.32, |
| "step": 44 |
| }, |
| { |
| "loss": 0.512, |
| "learning_rate": 0.0002, |
| "epoch": 2.37, |
| "step": 45 |
| }, |
| { |
| "loss": 0.6945, |
| "learning_rate": 0.0002, |
| "epoch": 2.42, |
| "step": 46 |
| }, |
| { |
| "loss": 0.5883, |
| "learning_rate": 0.0002, |
| "epoch": 2.47, |
| "step": 47 |
| }, |
| { |
| "loss": 0.5311, |
| "learning_rate": 0.0002, |
| "epoch": 2.53, |
| "step": 48 |
| }, |
| { |
| "loss": 0.5292, |
| "learning_rate": 0.0002, |
| "epoch": 2.58, |
| "step": 49 |
| }, |
| { |
| "loss": 0.5541, |
| "learning_rate": 0.0002, |
| "epoch": 2.63, |
| "step": 50 |
| }, |
| { |
| "eval_code_is_correct_loss": 0.5940853357315063, |
| "eval_code_is_correct_score": -0.19792629778385162, |
| "eval_code_is_correct_brier_score": 0.19792629778385162, |
| "eval_code_is_correct_average_probability": 0.6069728136062622, |
| "eval_code_is_correct_accuracy": 0.69, |
| "eval_code_is_correct_probabilities": [ |
| 0.824921190738678, |
| 0.46786367893218994, |
| 0.6803945302963257, |
| 0.3338681757450104, |
| 0.4255507290363312, |
| 0.8192625641822815, |
| 0.6870703101158142, |
| 0.5385801196098328, |
| 0.4014917314052582, |
| 0.2137792408466339, |
| 0.5761697888374329, |
| 0.2802768051624298, |
| 0.4230016767978668, |
| 0.7144221067428589, |
| 0.702923059463501, |
| 0.7705851197242737, |
| 0.7801277041435242, |
| 0.18786321580410004, |
| 0.8821263909339905, |
| 0.7419917583465576, |
| 0.8037750720977783, |
| 0.7353634238243103, |
| 0.525610625743866, |
| 0.199602872133255, |
| 0.3059692978858948, |
| 0.354174941778183, |
| 0.8103994727134705, |
| 0.7111828327178955, |
| 0.7084280252456665, |
| 0.5410431623458862, |
| 0.7837337255477905, |
| 0.6090495586395264, |
| 0.12445675581693649, |
| 0.7169621586799622, |
| 0.8234235048294067, |
| 0.8735712766647339, |
| 0.7732588052749634, |
| 0.43631064891815186, |
| 0.5443180799484253, |
| 0.348981112241745, |
| 0.8432489633560181, |
| 0.32388389110565186, |
| 0.9644874334335327, |
| 0.6631470918655396, |
| 0.7456516623497009, |
| 0.7492141127586365, |
| 0.4318337142467499, |
| 0.46674731373786926, |
| 0.4367634654045105, |
| 0.4380766749382019, |
| 0.39654305577278137, |
| 0.9517335295677185, |
| 0.36624813079833984, |
| 0.645854115486145, |
| 0.8562455177307129, |
| 0.7007656693458557, |
| 0.6245970726013184, |
| 0.7546803951263428, |
| 0.6691403388977051, |
| 0.8129599094390869, |
| 0.5258075594902039, |
| 0.6954044103622437, |
| 0.36306533217430115, |
| 0.9693009853363037, |
| 0.6357976198196411, |
| 0.7374253869056702, |
| 0.5900733470916748, |
| 0.540191650390625, |
| 0.6071694493293762, |
| 0.3094453513622284, |
| 0.8251391649246216, |
| 0.8399093151092529, |
| 0.8712323307991028, |
| 0.5785478949546814, |
| 0.5815758109092712, |
| 0.7164744138717651, |
| 0.5060900449752808, |
| 0.6983495950698853, |
| 0.7523359656333923, |
| 0.14946384727954865, |
| 0.7415146827697754, |
| 0.34309232234954834, |
| 0.23558610677719116, |
| 0.43409526348114014, |
| 0.7354661226272583, |
| 0.7268248200416565, |
| 0.7648622989654541, |
| 0.7701873779296875, |
| 0.4304506778717041, |
| 0.7484250068664551, |
| 0.7701525688171387, |
| 0.3247123658657074, |
| 0.6498774886131287, |
| 0.34104210138320923, |
| 0.9773394465446472, |
| 0.27353885769844055, |
| 0.747333824634552, |
| 0.5845430493354797, |
| 0.8978924751281738, |
| 0.6838365793228149 |
| ], |
| "eval_code_is_correct_runtime": 67.7443, |
| "eval_code_is_correct_samples_per_second": 1.476, |
| "eval_code_is_correct_steps_per_second": 0.059, |
| "epoch": 2.63, |
| "step": 50 |
| }, |
| { |
| "loss": 0.5214, |
| "learning_rate": 0.0002, |
| "epoch": 2.68, |
| "step": 51 |
| }, |
| { |
| "loss": 0.7301, |
| "learning_rate": 0.0002, |
| "epoch": 2.74, |
| "step": 52 |
| }, |
| { |
| "loss": 0.53, |
| "learning_rate": 0.0002, |
| "epoch": 2.79, |
| "step": 53 |
| }, |
| { |
| "loss": 0.5241, |
| "learning_rate": 0.0002, |
| "epoch": 2.84, |
| "step": 54 |
| }, |
| { |
| "loss": 0.4484, |
| "learning_rate": 0.0002, |
| "epoch": 2.89, |
| "step": 55 |
| }, |
| { |
| "loss": 0.5665, |
| "learning_rate": 0.0002, |
| "epoch": 2.95, |
| "step": 56 |
| }, |
| { |
| "loss": 0.4578, |
| "learning_rate": 0.0002, |
| "epoch": 3.0, |
| "step": 57 |
| }, |
| { |
| "loss": 0.4425, |
| "learning_rate": 0.0002, |
| "epoch": 3.05, |
| "step": 58 |
| }, |
| { |
| "loss": 0.4053, |
| "learning_rate": 0.0002, |
| "epoch": 3.11, |
| "step": 59 |
| }, |
| { |
| "loss": 0.4992, |
| "learning_rate": 0.0002, |
| "epoch": 3.16, |
| "step": 60 |
| }, |
| { |
| "loss": 0.2353, |
| "learning_rate": 0.0002, |
| "epoch": 3.21, |
| "step": 61 |
| }, |
| { |
| "loss": 0.3383, |
| "learning_rate": 0.0002, |
| "epoch": 3.26, |
| "step": 62 |
| }, |
| { |
| "loss": 0.6835, |
| "learning_rate": 0.0002, |
| "epoch": 3.32, |
| "step": 63 |
| }, |
| { |
| "loss": 0.6729, |
| "learning_rate": 0.0002, |
| "epoch": 3.37, |
| "step": 64 |
| }, |
| { |
| "loss": 0.6045, |
| "learning_rate": 0.0002, |
| "epoch": 3.42, |
| "step": 65 |
| }, |
| { |
| "loss": 0.324, |
| "learning_rate": 0.0002, |
| "epoch": 3.47, |
| "step": 66 |
| }, |
| { |
| "loss": 0.3726, |
| "learning_rate": 0.0002, |
| "epoch": 3.53, |
| "step": 67 |
| }, |
| { |
| "loss": 0.3529, |
| "learning_rate": 0.0002, |
| "epoch": 3.58, |
| "step": 68 |
| }, |
| { |
| "loss": 0.4642, |
| "learning_rate": 0.0002, |
| "epoch": 3.63, |
| "step": 69 |
| }, |
| { |
| "loss": 0.4211, |
| "learning_rate": 0.0002, |
| "epoch": 3.68, |
| "step": 70 |
| }, |
| { |
| "loss": 0.298, |
| "learning_rate": 0.0002, |
| "epoch": 3.74, |
| "step": 71 |
| }, |
| { |
| "loss": 0.3618, |
| "learning_rate": 0.0002, |
| "epoch": 3.79, |
| "step": 72 |
| }, |
| { |
| "loss": 0.4728, |
| "learning_rate": 0.0002, |
| "epoch": 3.84, |
| "step": 73 |
| }, |
| { |
| "loss": 0.402, |
| "learning_rate": 0.0002, |
| "epoch": 3.89, |
| "step": 74 |
| }, |
| { |
| "loss": 0.4046, |
| "learning_rate": 0.0002, |
| "epoch": 3.95, |
| "step": 75 |
| }, |
| { |
| "eval_code_is_correct_loss": 0.47848281264305115, |
| "eval_code_is_correct_score": -0.15475420653820038, |
| "eval_code_is_correct_brier_score": 0.15475420653820038, |
| "eval_code_is_correct_average_probability": 0.6844885945320129, |
| "eval_code_is_correct_accuracy": 0.76, |
| "eval_code_is_correct_probabilities": [ |
| 0.9683555960655212, |
| 0.44921836256980896, |
| 0.7495419383049011, |
| 0.2939276099205017, |
| 0.34451109170913696, |
| 0.9546052813529968, |
| 0.6665733456611633, |
| 0.6185269951820374, |
| 0.7960335612297058, |
| 0.24308358132839203, |
| 0.6751048564910889, |
| 0.3198447823524475, |
| 0.6380735635757446, |
| 0.9478679895401001, |
| 0.5936516523361206, |
| 0.6253194212913513, |
| 0.8495998978614807, |
| 0.3085791766643524, |
| 0.7716754674911499, |
| 0.7831162214279175, |
| 0.9598362445831299, |
| 0.5667487382888794, |
| 0.5415951609611511, |
| 0.20470955967903137, |
| 0.5040990710258484, |
| 0.3724431097507477, |
| 0.8682514429092407, |
| 0.7767794728279114, |
| 0.7808045744895935, |
| 0.5015738010406494, |
| 0.7318475842475891, |
| 0.7234196066856384, |
| 0.4821881353855133, |
| 0.6735180616378784, |
| 0.8447625637054443, |
| 0.9881181120872498, |
| 0.7274037599563599, |
| 0.5772131085395813, |
| 0.6710129380226135, |
| 0.3252182602882385, |
| 0.8553206920623779, |
| 0.2691892981529236, |
| 0.9974000453948975, |
| 0.8067336082458496, |
| 0.7142802476882935, |
| 0.7670422792434692, |
| 0.36208221316337585, |
| 0.9332921504974365, |
| 0.8860767483711243, |
| 0.40401703119277954, |
| 0.35290971398353577, |
| 0.997400164604187, |
| 0.4895634353160858, |
| 0.8298471570014954, |
| 0.8764240741729736, |
| 0.6197875142097473, |
| 0.8558983206748962, |
| 0.8380407691001892, |
| 0.8485222458839417, |
| 0.555997908115387, |
| 0.26674920320510864, |
| 0.8279209733009338, |
| 0.41571998596191406, |
| 0.9991807341575623, |
| 0.8207216858863831, |
| 0.7336632609367371, |
| 0.6954337358474731, |
| 0.6488922238349915, |
| 0.922136127948761, |
| 0.4455629289150238, |
| 0.8186019062995911, |
| 0.9768170714378357, |
| 0.9885281324386597, |
| 0.8712007999420166, |
| 0.8978477716445923, |
| 0.8873462677001953, |
| 0.6821259260177612, |
| 0.8775141835212708, |
| 0.7657942771911621, |
| 0.2309243381023407, |
| 0.8537100553512573, |
| 0.13663502037525177, |
| 0.1426939070224762, |
| 0.6065592169761658, |
| 0.7732885479927063, |
| 0.7179746031761169, |
| 0.8297063112258911, |
| 0.9848968982696533, |
| 0.24758392572402954, |
| 0.757599949836731, |
| 0.9647876620292664, |
| 0.4051612615585327, |
| 0.8224207758903503, |
| 0.8955028653144836, |
| 0.9978285431861877, |
| 0.4562136232852936, |
| 0.9620097875595093, |
| 0.9017798900604248, |
| 0.8915098905563354, |
| 0.6497373580932617 |
| ], |
| "eval_code_is_correct_runtime": 67.7494, |
| "eval_code_is_correct_samples_per_second": 1.476, |
| "eval_code_is_correct_steps_per_second": 0.059, |
| "epoch": 3.95, |
| "step": 75 |
| }, |
| { |
| "loss": 0.3116, |
| "learning_rate": 0.0002, |
| "epoch": 4.0, |
| "step": 76 |
| }, |
| { |
| "loss": 0.3615, |
| "learning_rate": 0.0002, |
| "epoch": 4.05, |
| "step": 77 |
| }, |
| { |
| "loss": 0.2709, |
| "learning_rate": 0.0002, |
| "epoch": 4.11, |
| "step": 78 |
| }, |
| { |
| "loss": 0.3665, |
| "learning_rate": 0.0002, |
| "epoch": 4.16, |
| "step": 79 |
| }, |
| { |
| "loss": 0.1072, |
| "learning_rate": 0.0002, |
| "epoch": 4.21, |
| "step": 80 |
| }, |
| { |
| "loss": 0.2752, |
| "learning_rate": 0.0002, |
| "epoch": 4.26, |
| "step": 81 |
| }, |
| { |
| "loss": 0.1764, |
| "learning_rate": 0.0002, |
| "epoch": 4.32, |
| "step": 82 |
| }, |
| { |
| "loss": 0.1275, |
| "learning_rate": 0.0002, |
| "epoch": 4.37, |
| "step": 83 |
| }, |
| { |
| "loss": 0.3218, |
| "learning_rate": 0.0002, |
| "epoch": 4.42, |
| "step": 84 |
| }, |
| { |
| "loss": 0.3217, |
| "learning_rate": 0.0002, |
| "epoch": 4.47, |
| "step": 85 |
| }, |
| { |
| "loss": 0.3257, |
| "learning_rate": 0.0002, |
| "epoch": 4.53, |
| "step": 86 |
| }, |
| { |
| "loss": 0.2085, |
| "learning_rate": 0.0002, |
| "epoch": 4.58, |
| "step": 87 |
| }, |
| { |
| "loss": 0.5161, |
| "learning_rate": 0.0002, |
| "epoch": 4.63, |
| "step": 88 |
| }, |
| { |
| "loss": 0.3032, |
| "learning_rate": 0.0002, |
| "epoch": 4.68, |
| "step": 89 |
| }, |
| { |
| "loss": 0.1597, |
| "learning_rate": 0.0002, |
| "epoch": 4.74, |
| "step": 90 |
| }, |
| { |
| "loss": 0.3244, |
| "learning_rate": 0.0002, |
| "epoch": 4.79, |
| "step": 91 |
| }, |
| { |
| "loss": 0.2538, |
| "learning_rate": 0.0002, |
| "epoch": 4.84, |
| "step": 92 |
| }, |
| { |
| "loss": 0.3986, |
| "learning_rate": 0.0002, |
| "epoch": 4.89, |
| "step": 93 |
| }, |
| { |
| "loss": 0.1784, |
| "learning_rate": 0.0002, |
| "epoch": 4.95, |
| "step": 94 |
| }, |
| { |
| "loss": 0.0902, |
| "learning_rate": 0.0002, |
| "epoch": 5.0, |
| "step": 95 |
| }, |
| { |
| "loss": 0.1785, |
| "learning_rate": 0.0002, |
| "epoch": 5.05, |
| "step": 96 |
| }, |
| { |
| "loss": 0.1223, |
| "learning_rate": 0.0002, |
| "epoch": 5.11, |
| "step": 97 |
| }, |
| { |
| "loss": 0.1473, |
| "learning_rate": 0.0002, |
| "epoch": 5.16, |
| "step": 98 |
| }, |
| { |
| "loss": 0.1615, |
| "learning_rate": 0.0002, |
| "epoch": 5.21, |
| "step": 99 |
| }, |
| { |
| "loss": 0.1691, |
| "learning_rate": 0.0002, |
| "epoch": 5.26, |
| "step": 100 |
| }, |
| { |
| "eval_code_is_correct_loss": 0.7131536602973938, |
| "eval_code_is_correct_score": -0.21241626143455505, |
| "eval_code_is_correct_brier_score": 0.21241626143455505, |
| "eval_code_is_correct_average_probability": 0.7060017585754395, |
| "eval_code_is_correct_accuracy": 0.73, |
| "eval_code_is_correct_probabilities": [ |
| 0.9998742341995239, |
| 0.24125896394252777, |
| 0.5266091823577881, |
| 0.9446415901184082, |
| 0.658671498298645, |
| 0.999782145023346, |
| 0.4476185441017151, |
| 0.9995893836021423, |
| 0.9987817406654358, |
| 0.9846014976501465, |
| 0.6241582632064819, |
| 0.9147050380706787, |
| 0.9970441460609436, |
| 0.999207079410553, |
| 0.03215882182121277, |
| 0.08953133225440979, |
| 0.923194944858551, |
| 0.9589815735816956, |
| 0.2918163537979126, |
| 0.7663164734840393, |
| 0.9991129040718079, |
| 0.12000331282615662, |
| 0.15813884139060974, |
| 0.8554588556289673, |
| 0.9006542563438416, |
| 0.9951829314231873, |
| 0.48046576976776123, |
| 0.2053559571504593, |
| 0.8127390146255493, |
| 0.9593634605407715, |
| 0.11686976253986359, |
| 0.124169260263443, |
| 0.9974563717842102, |
| 0.04802563041448593, |
| 0.8246001601219177, |
| 0.9987502098083496, |
| 0.015473408624529839, |
| 0.0645429939031601, |
| 0.49971136450767517, |
| 0.9124832153320312, |
| 0.39771828055381775, |
| 0.7872201204299927, |
| 0.9997814297676086, |
| 0.987456738948822, |
| 0.1562322974205017, |
| 0.7515577673912048, |
| 0.8418968319892883, |
| 0.9998055100440979, |
| 0.9988735318183899, |
| 0.8924939036369324, |
| 0.9951292276382446, |
| 0.999900221824646, |
| 0.9838736653327942, |
| 0.994502067565918, |
| 0.7390810251235962, |
| 0.04368113726377487, |
| 0.9977819323539734, |
| 0.9505625367164612, |
| 0.9955796599388123, |
| 0.06847873330116272, |
| 0.001323648844845593, |
| 0.9979519248008728, |
| 0.5310106873512268, |
| 0.9994561076164246, |
| 0.3053685426712036, |
| 0.8901605606079102, |
| 0.6555763483047485, |
| 0.9988672733306885, |
| 0.9997801184654236, |
| 0.9971674084663391, |
| 0.5509296655654907, |
| 0.9994674324989319, |
| 0.9999033212661743, |
| 0.9959565997123718, |
| 0.99689781665802, |
| 0.994744062423706, |
| 0.6479522585868835, |
| 0.9884534478187561, |
| 0.8599462509155273, |
| 0.8442027568817139, |
| 0.9853453040122986, |
| 0.015567261725664139, |
| 0.17351527512073517, |
| 0.9978513717651367, |
| 0.29119962453842163, |
| 0.5376598834991455, |
| 0.06629227846860886, |
| 0.9998811483383179, |
| 0.030417898669838905, |
| 0.8476421236991882, |
| 0.9996863603591919, |
| 0.185418501496315, |
| 0.9999172687530518, |
| 0.9991611242294312, |
| 0.9999512434005737, |
| 0.9985194802284241, |
| 0.9991263747215271, |
| 0.9992461204528809, |
| 0.5890063047409058, |
| 0.5609598159790039 |
| ], |
| "eval_code_is_correct_runtime": 67.7473, |
| "eval_code_is_correct_samples_per_second": 1.476, |
| "eval_code_is_correct_steps_per_second": 0.059, |
| "epoch": 5.26, |
| "step": 100 |
| }, |
| { |
| "loss": 0.5527, |
| "learning_rate": 0.0002, |
| "epoch": 5.32, |
| "step": 101 |
| }, |
| { |
| "loss": 0.5535, |
| "learning_rate": 0.0002, |
| "epoch": 5.37, |
| "step": 102 |
| }, |
| { |
| "loss": 0.4444, |
| "learning_rate": 0.0002, |
| "epoch": 5.42, |
| "step": 103 |
| }, |
| { |
| "loss": 0.1897, |
| "learning_rate": 0.0002, |
| "epoch": 5.47, |
| "step": 104 |
| }, |
| { |
| "loss": 0.015, |
| "learning_rate": 0.0002, |
| "epoch": 5.53, |
| "step": 105 |
| }, |
| { |
| "loss": 0.6378, |
| "learning_rate": 0.0002, |
| "epoch": 5.58, |
| "step": 106 |
| }, |
| { |
| "loss": 0.5394, |
| "learning_rate": 0.0002, |
| "epoch": 5.63, |
| "step": 107 |
| }, |
| { |
| "loss": 0.3359, |
| "learning_rate": 0.0002, |
| "epoch": 5.68, |
| "step": 108 |
| }, |
| { |
| "loss": 0.2657, |
| "learning_rate": 0.0002, |
| "epoch": 5.74, |
| "step": 109 |
| }, |
| { |
| "loss": 0.2404, |
| "learning_rate": 0.0002, |
| "epoch": 5.79, |
| "step": 110 |
| }, |
| { |
| "loss": 0.5007, |
| "learning_rate": 0.0002, |
| "epoch": 5.84, |
| "step": 111 |
| }, |
| { |
| "loss": 0.5109, |
| "learning_rate": 0.0002, |
| "epoch": 5.89, |
| "step": 112 |
| }, |
| { |
| "loss": 0.1953, |
| "learning_rate": 0.0002, |
| "epoch": 5.95, |
| "step": 113 |
| }, |
| { |
| "loss": 0.0324, |
| "learning_rate": 0.0002, |
| "epoch": 6.0, |
| "step": 114 |
| }, |
| { |
| "loss": 0.3503, |
| "learning_rate": 0.0002, |
| "epoch": 6.05, |
| "step": 115 |
| }, |
| { |
| "loss": 0.2595, |
| "learning_rate": 0.0002, |
| "epoch": 6.11, |
| "step": 116 |
| }, |
| { |
| "loss": 0.4104, |
| "learning_rate": 0.0002, |
| "epoch": 6.16, |
| "step": 117 |
| }, |
| { |
| "loss": 0.1498, |
| "learning_rate": 0.0002, |
| "epoch": 6.21, |
| "step": 118 |
| }, |
| { |
| "loss": 0.1297, |
| "learning_rate": 0.0002, |
| "epoch": 6.26, |
| "step": 119 |
| }, |
| { |
| "loss": 0.1723, |
| "learning_rate": 0.0002, |
| "epoch": 6.32, |
| "step": 120 |
| }, |
| { |
| "loss": 0.1331, |
| "learning_rate": 0.0002, |
| "epoch": 6.37, |
| "step": 121 |
| }, |
| { |
| "loss": 0.2478, |
| "learning_rate": 0.0002, |
| "epoch": 6.42, |
| "step": 122 |
| }, |
| { |
| "loss": 0.3555, |
| "learning_rate": 0.0002, |
| "epoch": 6.47, |
| "step": 123 |
| }, |
| { |
| "loss": 0.2095, |
| "learning_rate": 0.0002, |
| "epoch": 6.53, |
| "step": 124 |
| }, |
| { |
| "loss": 0.1896, |
| "learning_rate": 0.0002, |
| "epoch": 6.58, |
| "step": 125 |
| }, |
| { |
| "eval_code_is_correct_loss": 0.4567447602748871, |
| "eval_code_is_correct_score": -0.14007602632045746, |
| "eval_code_is_correct_brier_score": 0.14007602632045746, |
| "eval_code_is_correct_average_probability": 0.7727714776992798, |
| "eval_code_is_correct_accuracy": 0.82, |
| "eval_code_is_correct_probabilities": [ |
| 0.9987679123878479, |
| 0.4806418716907501, |
| 0.7848430871963501, |
| 0.3105872571468353, |
| 0.5165024399757385, |
| 0.998405396938324, |
| 0.8343289494514465, |
| 0.9901788234710693, |
| 0.9813113808631897, |
| 0.7574440240859985, |
| 0.7910118699073792, |
| 0.06982726603746414, |
| 0.9809978604316711, |
| 0.9973791837692261, |
| 0.5600104928016663, |
| 0.7603288888931274, |
| 0.9818860292434692, |
| 0.3341038227081299, |
| 0.9702696800231934, |
| 0.9806032776832581, |
| 0.9957588315010071, |
| 0.5696340203285217, |
| 0.7143341898918152, |
| 0.16743437945842743, |
| 0.7721751928329468, |
| 0.1173228770494461, |
| 0.8101537823677063, |
| 0.9562052488327026, |
| 0.906458854675293, |
| 0.8179009556770325, |
| 0.7743387222290039, |
| 0.8904827237129211, |
| 0.8322084546089172, |
| 0.8095942139625549, |
| 0.9716352224349976, |
| 0.9985573887825012, |
| 0.1670643389225006, |
| 0.748282253742218, |
| 0.7800930142402649, |
| 0.14132891595363617, |
| 0.8936496376991272, |
| 0.36719489097595215, |
| 0.9985307455062866, |
| 0.9869412779808044, |
| 0.6673361659049988, |
| 0.9174671769142151, |
| 0.9842150807380676, |
| 0.997148334980011, |
| 0.985522985458374, |
| 0.8499152064323425, |
| 0.9198179244995117, |
| 0.9993763566017151, |
| 0.6242245435714722, |
| 0.9891676902770996, |
| 0.9344865083694458, |
| 0.8665608763694763, |
| 0.9534940719604492, |
| 0.9868246912956238, |
| 0.989677906036377, |
| 0.6916028261184692, |
| 0.05886725336313248, |
| 0.9953218102455139, |
| 0.9273397922515869, |
| 0.9996368885040283, |
| 0.12487627565860748, |
| 0.9797862768173218, |
| 0.9216504693031311, |
| 0.9686106443405151, |
| 0.997191846370697, |
| 0.9759156703948975, |
| 0.7190247178077698, |
| 0.9861569404602051, |
| 0.9982544779777527, |
| 0.9947845339775085, |
| 0.9661282896995544, |
| 0.9940819144248962, |
| 0.7725122570991516, |
| 0.9937759637832642, |
| 0.8750337362289429, |
| 0.15745404362678528, |
| 0.9853953123092651, |
| 0.0993090569972992, |
| 0.07436255365610123, |
| 0.9853988885879517, |
| 0.7725264430046082, |
| 0.4874012768268585, |
| 0.1029394119977951, |
| 0.9995567202568054, |
| 0.013778586871922016, |
| 0.9781476855278015, |
| 0.9989705085754395, |
| 0.26722535490989685, |
| 0.9956694841384888, |
| 0.9968309998512268, |
| 0.9996675252914429, |
| 0.82625812292099, |
| 0.9978654980659485, |
| 0.9962685108184814, |
| 0.9435150027275085, |
| 0.7261157631874084 |
| ], |
| "eval_code_is_correct_runtime": 67.7467, |
| "eval_code_is_correct_samples_per_second": 1.476, |
| "eval_code_is_correct_steps_per_second": 0.059, |
| "epoch": 6.58, |
| "step": 125 |
| }, |
| { |
| "loss": 0.1246, |
| "learning_rate": 0.0002, |
| "epoch": 6.63, |
| "step": 126 |
| }, |
| { |
| "loss": 0.1277, |
| "learning_rate": 0.0002, |
| "epoch": 6.68, |
| "step": 127 |
| }, |
| { |
| "loss": 0.0844, |
| "learning_rate": 0.0002, |
| "epoch": 6.74, |
| "step": 128 |
| }, |
| { |
| "loss": 0.1287, |
| "learning_rate": 0.0002, |
| "epoch": 6.79, |
| "step": 129 |
| }, |
| { |
| "loss": 0.0997, |
| "learning_rate": 0.0002, |
| "epoch": 6.84, |
| "step": 130 |
| }, |
| { |
| "loss": 0.0979, |
| "learning_rate": 0.0002, |
| "epoch": 6.89, |
| "step": 131 |
| }, |
| { |
| "loss": 0.0433, |
| "learning_rate": 0.0002, |
| "epoch": 6.95, |
| "step": 132 |
| }, |
| { |
| "loss": 0.2242, |
| "learning_rate": 0.0002, |
| "epoch": 7.0, |
| "step": 133 |
| }, |
| { |
| "loss": 0.0423, |
| "learning_rate": 0.0002, |
| "epoch": 7.05, |
| "step": 134 |
| }, |
| { |
| "loss": 0.1313, |
| "learning_rate": 0.0002, |
| "epoch": 7.11, |
| "step": 135 |
| }, |
| { |
| "loss": 0.1776, |
| "learning_rate": 0.0002, |
| "epoch": 7.16, |
| "step": 136 |
| }, |
| { |
| "loss": 0.0513, |
| "learning_rate": 0.0002, |
| "epoch": 7.21, |
| "step": 137 |
| }, |
| { |
| "loss": 0.0166, |
| "learning_rate": 0.0002, |
| "epoch": 7.26, |
| "step": 138 |
| }, |
| { |
| "loss": 0.1085, |
| "learning_rate": 0.0002, |
| "epoch": 7.32, |
| "step": 139 |
| }, |
| { |
| "loss": 0.1359, |
| "learning_rate": 0.0002, |
| "epoch": 7.37, |
| "step": 140 |
| }, |
| { |
| "loss": 0.3451, |
| "learning_rate": 0.0002, |
| "epoch": 7.42, |
| "step": 141 |
| }, |
| { |
| "loss": 0.3083, |
| "learning_rate": 0.0002, |
| "epoch": 7.47, |
| "step": 142 |
| }, |
| { |
| "loss": 0.0116, |
| "learning_rate": 0.0002, |
| "epoch": 7.53, |
| "step": 143 |
| }, |
| { |
| "loss": 0.0504, |
| "learning_rate": 0.0002, |
| "epoch": 7.58, |
| "step": 144 |
| }, |
| { |
| "loss": 0.049, |
| "learning_rate": 0.0002, |
| "epoch": 7.63, |
| "step": 145 |
| }, |
| { |
| "loss": 0.0906, |
| "learning_rate": 0.0002, |
| "epoch": 7.68, |
| "step": 146 |
| }, |
| { |
| "loss": 0.1394, |
| "learning_rate": 0.0002, |
| "epoch": 7.74, |
| "step": 147 |
| }, |
| { |
| "loss": 0.0492, |
| "learning_rate": 0.0002, |
| "epoch": 7.79, |
| "step": 148 |
| }, |
| { |
| "loss": 0.0513, |
| "learning_rate": 0.0002, |
| "epoch": 7.84, |
| "step": 149 |
| }, |
| { |
| "loss": 0.0248, |
| "learning_rate": 0.0002, |
| "epoch": 7.89, |
| "step": 150 |
| }, |
| { |
| "eval_code_is_correct_loss": 0.8317287564277649, |
| "eval_code_is_correct_score": -0.16410262882709503, |
| "eval_code_is_correct_brier_score": 0.16410262882709503, |
| "eval_code_is_correct_average_probability": 0.7992724776268005, |
| "eval_code_is_correct_accuracy": 0.81, |
| "eval_code_is_correct_probabilities": [ |
| 0.9999188184738159, |
| 0.9416216611862183, |
| 0.9813385009765625, |
| 0.16267889738082886, |
| 0.521327793598175, |
| 0.9999769926071167, |
| 0.9942213296890259, |
| 0.999913215637207, |
| 0.9997870326042175, |
| 0.9944077730178833, |
| 0.9233108758926392, |
| 0.006665369495749474, |
| 0.994364857673645, |
| 0.99979168176651, |
| 0.13727723062038422, |
| 0.8380267024040222, |
| 0.999836802482605, |
| 0.6078657507896423, |
| 0.9981252551078796, |
| 0.9990299940109253, |
| 0.9997946619987488, |
| 0.7654120326042175, |
| 0.9162885546684265, |
| 0.032905738800764084, |
| 0.8361085653305054, |
| 0.43142011761665344, |
| 0.9244018793106079, |
| 0.9802189469337463, |
| 0.9970953464508057, |
| 0.28045544028282166, |
| 0.9703474640846252, |
| 0.9523987174034119, |
| 0.9937481880187988, |
| 0.6890085935592651, |
| 0.9996067881584167, |
| 0.9997078776359558, |
| 0.018358726054430008, |
| 0.7913146615028381, |
| 0.9718425869941711, |
| 0.051817361265420914, |
| 0.9650564193725586, |
| 0.039147596806287766, |
| 0.9999592304229736, |
| 0.9997337460517883, |
| 0.9772196412086487, |
| 0.9988301396369934, |
| 0.006023730617016554, |
| 0.9997532963752747, |
| 0.9996465444564819, |
| 0.8749299645423889, |
| 0.9997159838676453, |
| 0.999969482421875, |
| 0.47189682722091675, |
| 0.9998407363891602, |
| 0.9996135830879211, |
| 0.992363691329956, |
| 0.9981967806816101, |
| 0.999813973903656, |
| 0.9998838901519775, |
| 0.5446488857269287, |
| 0.0003066202625632286, |
| 0.999956488609314, |
| 0.9962054491043091, |
| 0.99986732006073, |
| 0.9834662079811096, |
| 0.9997921586036682, |
| 0.9902730584144592, |
| 0.9985108971595764, |
| 0.9999505281448364, |
| 0.9998027682304382, |
| 0.9723180532455444, |
| 0.9994682669639587, |
| 0.9999501705169678, |
| 0.9999010562896729, |
| 0.9973589777946472, |
| 0.9999006986618042, |
| 0.9960615038871765, |
| 0.9998500347137451, |
| 0.9986233711242676, |
| 0.008311701007187366, |
| 0.9997389912605286, |
| 0.017625167965888977, |
| 0.0008023975533433259, |
| 0.9993647933006287, |
| 0.9838919639587402, |
| 0.43106237053871155, |
| 0.0007903283112682402, |
| 0.9999732971191406, |
| 0.00020203986787237227, |
| 0.9997217059135437, |
| 0.9999573230743408, |
| 0.001714992569759488, |
| 0.9998447895050049, |
| 0.9997473359107971, |
| 0.9999877214431763, |
| 0.9988928437232971, |
| 0.9996260404586792, |
| 0.9999154806137085, |
| 0.9983515739440918, |
| 0.9881762266159058 |
| ], |
| "eval_code_is_correct_runtime": 67.7436, |
| "eval_code_is_correct_samples_per_second": 1.476, |
| "eval_code_is_correct_steps_per_second": 0.059, |
| "epoch": 7.89, |
| "step": 150 |
| }, |
| { |
| "train_runtime": 7451.7189, |
| "train_samples_per_second": 0.644, |
| "train_steps_per_second": 0.02, |
| "total_flos": 0.0, |
| "train_loss": 0.41071448507408304, |
| "epoch": 7.89, |
| "step": 150 |
| } |
| ] |
