diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,5390 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 0.9999074845036544, + "eval_steps": 100, + "global_step": 2702, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.001850309926912758, + "grad_norm": 2.734375, + "learning_rate": 3.690036900369004e-07, + "loss": 2.8172, + "mean_token_accuracy": 0.7237301587301586, + "mean_token_accuracy_utility": 0.6263164330117172, + "step": 5 + }, + { + "epoch": 0.003700619853825516, + "grad_norm": 2.53125, + "learning_rate": 7.380073800738008e-07, + "loss": 2.8457, + "mean_token_accuracy": 0.7139926739926739, + "mean_token_accuracy_utility": 0.6239560439560438, + "step": 10 + }, + { + "epoch": 0.005550929780738274, + "grad_norm": 2.703125, + "learning_rate": 1.1070110701107011e-06, + "loss": 2.8885, + "mean_token_accuracy": 0.7149023199023197, + "mean_token_accuracy_utility": 0.6203418803418802, + "step": 15 + }, + { + "epoch": 0.007401239707651032, + "grad_norm": 2.4375, + "learning_rate": 1.4760147601476015e-06, + "loss": 2.8439, + "mean_token_accuracy": 0.7276556776556775, + "mean_token_accuracy_utility": 0.6185714285714284, + "step": 20 + }, + { + "epoch": 0.00925154963456379, + "grad_norm": 2.546875, + "learning_rate": 1.845018450184502e-06, + "loss": 2.8526, + "mean_token_accuracy": 0.7135714285714284, + "mean_token_accuracy_utility": 0.6250061050061048, + "step": 25 + }, + { + "epoch": 0.011101859561476548, + "grad_norm": 2.578125, + "learning_rate": 2.2140221402214023e-06, + "loss": 2.8883, + "mean_token_accuracy": 0.7093040293040291, + "mean_token_accuracy_utility": 0.6193223443223441, + "step": 30 + }, + { + "epoch": 0.012952169488389306, + "grad_norm": 2.5, + "learning_rate": 2.5830258302583027e-06, + "loss": 2.8149, + "mean_token_accuracy": 0.7137118437118435, + "mean_token_accuracy_utility": 0.6313003663003661, + "step": 35 + }, + { + "epoch": 0.014802479415302064, + "grad_norm": 2.484375, + "learning_rate": 2.952029520295203e-06, + "loss": 2.7901, + "mean_token_accuracy": 0.7235897435897434, + "mean_token_accuracy_utility": 0.6272762489242514, + "step": 40 + }, + { + "epoch": 0.01665278934221482, + "grad_norm": 2.453125, + "learning_rate": 3.3210332103321034e-06, + "loss": 2.7968, + "mean_token_accuracy": 0.7122588522588522, + "mean_token_accuracy_utility": 0.6385103785103783, + "step": 45 + }, + { + "epoch": 0.01850309926912758, + "grad_norm": 2.421875, + "learning_rate": 3.690036900369004e-06, + "loss": 2.7839, + "mean_token_accuracy": 0.7185964035964034, + "mean_token_accuracy_utility": 0.6287118437118435, + "step": 50 + }, + { + "epoch": 0.020353409196040336, + "grad_norm": 2.34375, + "learning_rate": 4.059040590405905e-06, + "loss": 2.8042, + "mean_token_accuracy": 0.713168498168498, + "mean_token_accuracy_utility": 0.6368240093240092, + "step": 55 + }, + { + "epoch": 0.022203719122953096, + "grad_norm": 2.296875, + "learning_rate": 4.428044280442805e-06, + "loss": 2.8382, + "mean_token_accuracy": 0.7160256410256409, + "mean_token_accuracy_utility": 0.6184126984126983, + "step": 60 + }, + { + "epoch": 0.024054029049865852, + "grad_norm": 2.3125, + "learning_rate": 4.797047970479705e-06, + "loss": 2.7488, + "mean_token_accuracy": 0.7132967032967031, + "mean_token_accuracy_utility": 0.6416824188006075, + "step": 65 + }, + { + "epoch": 0.025904338976778612, + "grad_norm": 2.0625, + "learning_rate": 5.166051660516605e-06, + "loss": 2.7055, + "mean_token_accuracy": 0.7402075702075701, + "mean_token_accuracy_utility": 0.629145299145299, + "step": 70 + }, + { + "epoch": 0.027754648903691368, + "grad_norm": 2.078125, + "learning_rate": 5.535055350553506e-06, + "loss": 2.7952, + "mean_token_accuracy": 0.7170451770451769, + "mean_token_accuracy_utility": 0.6239987789987789, + "step": 75 + }, + { + "epoch": 0.029604958830604128, + "grad_norm": 1.7734375, + "learning_rate": 5.904059040590406e-06, + "loss": 2.8421, + "mean_token_accuracy": 0.713296703296703, + "mean_token_accuracy_utility": 0.6210012210012208, + "step": 80 + }, + { + "epoch": 0.031455268757516884, + "grad_norm": 1.9453125, + "learning_rate": 6.273062730627307e-06, + "loss": 2.748, + "mean_token_accuracy": 0.714145299145299, + "mean_token_accuracy_utility": 0.6345808450953776, + "step": 85 + }, + { + "epoch": 0.03330557868442964, + "grad_norm": 1.90625, + "learning_rate": 6.642066420664207e-06, + "loss": 2.748, + "mean_token_accuracy": 0.727899877899878, + "mean_token_accuracy_utility": 0.626141636141636, + "step": 90 + }, + { + "epoch": 0.0351558886113424, + "grad_norm": 1.7421875, + "learning_rate": 7.011070110701108e-06, + "loss": 2.8165, + "mean_token_accuracy": 0.7236630036630035, + "mean_token_accuracy_utility": 0.6159672670563757, + "step": 95 + }, + { + "epoch": 0.03700619853825516, + "grad_norm": 1.6640625, + "learning_rate": 7.380073800738008e-06, + "loss": 2.7346, + "mean_token_accuracy": 0.7208363858363857, + "mean_token_accuracy_utility": 0.6315201465201462, + "step": 100 + }, + { + "epoch": 0.03700619853825516, + "eval_mean_token_accuracy": 0.7086819498584206, + "eval_reasoning_loss": 1.095815896987915, + "eval_reasoning_runtime": 28.1385, + "eval_reasoning_samples_per_second": 4.584, + "eval_reasoning_steps_per_second": 1.173, + "step": 100 + }, + { + "epoch": 0.03700619853825516, + "eval_mean_token_accuracy": 0.6320687748054061, + "eval_utility_loss": 1.671917200088501, + "eval_utility_runtime": 10.4335, + "eval_utility_samples_per_second": 4.696, + "eval_utility_steps_per_second": 1.246, + "step": 100 + }, + { + "epoch": 0.038856508465167916, + "grad_norm": 1.6953125, + "learning_rate": 7.749077490774908e-06, + "loss": 2.7013, + "mean_token_accuracy": 0.7297496947496948, + "mean_token_accuracy_utility": 0.6395360195360194, + "step": 105 + }, + { + "epoch": 0.04070681839208067, + "grad_norm": 1.5, + "learning_rate": 8.11808118081181e-06, + "loss": 2.7586, + "mean_token_accuracy": 0.7181440781440782, + "mean_token_accuracy_utility": 0.6322832722832721, + "step": 110 + }, + { + "epoch": 0.04255712831899343, + "grad_norm": 1.5546875, + "learning_rate": 8.48708487084871e-06, + "loss": 2.6618, + "mean_token_accuracy": 0.7284554334554334, + "mean_token_accuracy_utility": 0.6384676434676433, + "step": 115 + }, + { + "epoch": 0.04440743824590619, + "grad_norm": 1.53125, + "learning_rate": 8.85608856088561e-06, + "loss": 2.7228, + "mean_token_accuracy": 0.7202319902319901, + "mean_token_accuracy_utility": 0.6311294261294259, + "step": 120 + }, + { + "epoch": 0.04625774817281895, + "grad_norm": 1.515625, + "learning_rate": 9.22509225092251e-06, + "loss": 2.7111, + "mean_token_accuracy": 0.7247863247863245, + "mean_token_accuracy_utility": 0.6272893772893771, + "step": 125 + }, + { + "epoch": 0.048108058099731704, + "grad_norm": 1.453125, + "learning_rate": 9.59409594095941e-06, + "loss": 2.6363, + "mean_token_accuracy": 0.7367155067155066, + "mean_token_accuracy_utility": 0.6390778403168647, + "step": 130 + }, + { + "epoch": 0.04995836802664446, + "grad_norm": 1.4609375, + "learning_rate": 9.963099630996312e-06, + "loss": 2.6715, + "mean_token_accuracy": 0.7265567765567764, + "mean_token_accuracy_utility": 0.6306471306471305, + "step": 135 + }, + { + "epoch": 0.051808677953557224, + "grad_norm": 1.5234375, + "learning_rate": 1.033210332103321e-05, + "loss": 2.6352, + "mean_token_accuracy": 0.7250366300366299, + "mean_token_accuracy_utility": 0.6349511599511597, + "step": 140 + }, + { + "epoch": 0.05365898788046998, + "grad_norm": 1.390625, + "learning_rate": 1.0701107011070112e-05, + "loss": 2.5993, + "mean_token_accuracy": 0.7307814407814408, + "mean_token_accuracy_utility": 0.6437240537240536, + "step": 145 + }, + { + "epoch": 0.055509297807382736, + "grad_norm": 1.296875, + "learning_rate": 1.1070110701107012e-05, + "loss": 2.6087, + "mean_token_accuracy": 0.729261294261294, + "mean_token_accuracy_utility": 0.637271062271062, + "step": 150 + }, + { + "epoch": 0.05735960773429549, + "grad_norm": 1.484375, + "learning_rate": 1.1439114391143913e-05, + "loss": 2.5513, + "mean_token_accuracy": 0.725958485958486, + "mean_token_accuracy_utility": 0.6488766788766787, + "step": 155 + }, + { + "epoch": 0.059209917661208256, + "grad_norm": 1.359375, + "learning_rate": 1.1808118081180812e-05, + "loss": 2.6472, + "mean_token_accuracy": 0.7295909645909644, + "mean_token_accuracy_utility": 0.6322222222222221, + "step": 160 + }, + { + "epoch": 0.06106022758812101, + "grad_norm": 1.2421875, + "learning_rate": 1.2177121771217713e-05, + "loss": 2.6282, + "mean_token_accuracy": 0.7273626373626371, + "mean_token_accuracy_utility": 0.6314835164835164, + "step": 165 + }, + { + "epoch": 0.06291053751503377, + "grad_norm": 1.3046875, + "learning_rate": 1.2546125461254614e-05, + "loss": 2.5541, + "mean_token_accuracy": 0.7285042735042733, + "mean_token_accuracy_utility": 0.6473809523809522, + "step": 170 + }, + { + "epoch": 0.06476084744194653, + "grad_norm": 1.3671875, + "learning_rate": 1.2915129151291515e-05, + "loss": 2.627, + "mean_token_accuracy": 0.7305372405372406, + "mean_token_accuracy_utility": 0.6331135531135528, + "step": 175 + }, + { + "epoch": 0.06661115736885928, + "grad_norm": 1.296875, + "learning_rate": 1.3284132841328414e-05, + "loss": 2.5742, + "mean_token_accuracy": 0.737008547008547, + "mean_token_accuracy_utility": 0.6396031746031744, + "step": 180 + }, + { + "epoch": 0.06846146729577204, + "grad_norm": 1.328125, + "learning_rate": 1.3653136531365315e-05, + "loss": 2.69, + "mean_token_accuracy": 0.7373137973137973, + "mean_token_accuracy_utility": 0.6166910866910864, + "step": 185 + }, + { + "epoch": 0.0703117772226848, + "grad_norm": 1.28125, + "learning_rate": 1.4022140221402215e-05, + "loss": 2.5494, + "mean_token_accuracy": 0.7435347985347984, + "mean_token_accuracy_utility": 0.6370573870573868, + "step": 190 + }, + { + "epoch": 0.07216208714959756, + "grad_norm": 1.3359375, + "learning_rate": 1.4391143911439116e-05, + "loss": 2.6274, + "mean_token_accuracy": 0.7304334554334553, + "mean_token_accuracy_utility": 0.6212017078670652, + "step": 195 + }, + { + "epoch": 0.07401239707651032, + "grad_norm": 1.359375, + "learning_rate": 1.4760147601476015e-05, + "loss": 2.5646, + "mean_token_accuracy": 0.7305250305250306, + "mean_token_accuracy_utility": 0.6369841269841269, + "step": 200 + }, + { + "epoch": 0.07401239707651032, + "eval_mean_token_accuracy": 0.7265697265697266, + "eval_reasoning_loss": 0.9805031418800354, + "eval_reasoning_runtime": 27.7257, + "eval_reasoning_samples_per_second": 4.653, + "eval_reasoning_steps_per_second": 1.19, + "step": 200 + }, + { + "epoch": 0.07401239707651032, + "eval_mean_token_accuracy": 0.6414547546946745, + "eval_utility_loss": 1.5672366619110107, + "eval_utility_runtime": 10.4314, + "eval_utility_samples_per_second": 4.697, + "eval_utility_steps_per_second": 1.246, + "step": 200 + }, + { + "epoch": 0.07586270700342307, + "grad_norm": 1.3828125, + "learning_rate": 1.5129151291512916e-05, + "loss": 2.6058, + "mean_token_accuracy": 0.7363858363858364, + "mean_token_accuracy_utility": 0.6282356532356531, + "step": 205 + }, + { + "epoch": 0.07771301693033583, + "grad_norm": 1.328125, + "learning_rate": 1.5498154981549817e-05, + "loss": 2.4926, + "mean_token_accuracy": 0.7396825396825395, + "mean_token_accuracy_utility": 0.6444383394383392, + "step": 210 + }, + { + "epoch": 0.0795633268572486, + "grad_norm": 1.3515625, + "learning_rate": 1.5867158671586716e-05, + "loss": 2.513, + "mean_token_accuracy": 0.7431318681318682, + "mean_token_accuracy_utility": 0.637605152029542, + "step": 215 + }, + { + "epoch": 0.08141363678416134, + "grad_norm": 1.3046875, + "learning_rate": 1.623616236162362e-05, + "loss": 2.6489, + "mean_token_accuracy": 0.7217704517704515, + "mean_token_accuracy_utility": 0.6312792897884174, + "step": 220 + }, + { + "epoch": 0.08326394671107411, + "grad_norm": 1.28125, + "learning_rate": 1.6605166051660518e-05, + "loss": 2.5264, + "mean_token_accuracy": 0.7343956043956043, + "mean_token_accuracy_utility": 0.6406959706959705, + "step": 225 + }, + { + "epoch": 0.08511425663798686, + "grad_norm": 1.265625, + "learning_rate": 1.697416974169742e-05, + "loss": 2.4749, + "mean_token_accuracy": 0.7450793650793652, + "mean_token_accuracy_utility": 0.6402258852258851, + "step": 230 + }, + { + "epoch": 0.08696456656489962, + "grad_norm": 1.2578125, + "learning_rate": 1.734317343173432e-05, + "loss": 2.5614, + "mean_token_accuracy": 0.7416300366300363, + "mean_token_accuracy_utility": 0.6323504273504272, + "step": 235 + }, + { + "epoch": 0.08881487649181238, + "grad_norm": 1.3359375, + "learning_rate": 1.771217712177122e-05, + "loss": 2.36, + "mean_token_accuracy": 0.7451648351648352, + "mean_token_accuracy_utility": 0.6598412698412697, + "step": 240 + }, + { + "epoch": 0.09066518641872513, + "grad_norm": 1.2578125, + "learning_rate": 1.8081180811808117e-05, + "loss": 2.5768, + "mean_token_accuracy": 0.7321611721611723, + "mean_token_accuracy_utility": 0.6356043956043955, + "step": 245 + }, + { + "epoch": 0.0925154963456379, + "grad_norm": 1.203125, + "learning_rate": 1.845018450184502e-05, + "loss": 2.4927, + "mean_token_accuracy": 0.7371001221001221, + "mean_token_accuracy_utility": 0.6415079365079362, + "step": 250 + }, + { + "epoch": 0.09436580627255066, + "grad_norm": 1.203125, + "learning_rate": 1.8819188191881922e-05, + "loss": 2.4833, + "mean_token_accuracy": 0.74503663003663, + "mean_token_accuracy_utility": 0.6393927634521692, + "step": 255 + }, + { + "epoch": 0.09621611619946341, + "grad_norm": 1.1953125, + "learning_rate": 1.918819188191882e-05, + "loss": 2.4159, + "mean_token_accuracy": 0.7428937728937728, + "mean_token_accuracy_utility": 0.6525335775335773, + "step": 260 + }, + { + "epoch": 0.09806642612637617, + "grad_norm": 1.3359375, + "learning_rate": 1.955719557195572e-05, + "loss": 2.3955, + "mean_token_accuracy": 0.7417826617826616, + "mean_token_accuracy_utility": 0.6487362637362637, + "step": 265 + }, + { + "epoch": 0.09991673605328892, + "grad_norm": 1.3046875, + "learning_rate": 1.9926199261992623e-05, + "loss": 2.5987, + "mean_token_accuracy": 0.7282356532356531, + "mean_token_accuracy_utility": 0.6253296703296701, + "step": 270 + }, + { + "epoch": 0.10176704598020168, + "grad_norm": 1.2421875, + "learning_rate": 1.9999866396188624e-05, + "loss": 2.422, + "mean_token_accuracy": 0.7381440781440781, + "mean_token_accuracy_utility": 0.6473687423687421, + "step": 275 + }, + { + "epoch": 0.10361735590711445, + "grad_norm": 1.140625, + "learning_rate": 1.9999323636823398e-05, + "loss": 2.5184, + "mean_token_accuracy": 0.7480586080586079, + "mean_token_accuracy_utility": 0.6308363858363857, + "step": 280 + }, + { + "epoch": 0.1054676658340272, + "grad_norm": 1.0703125, + "learning_rate": 1.9998363394309497e-05, + "loss": 2.3524, + "mean_token_accuracy": 0.7643101343101344, + "mean_token_accuracy_utility": 0.6419230769230768, + "step": 285 + }, + { + "epoch": 0.10731797576093996, + "grad_norm": 1.0, + "learning_rate": 1.9996985708738146e-05, + "loss": 2.2596, + "mean_token_accuracy": 0.7655860805860807, + "mean_token_accuracy_utility": 0.6554334554334552, + "step": 290 + }, + { + "epoch": 0.10916828568785271, + "grad_norm": 1.0078125, + "learning_rate": 1.999519063762928e-05, + "loss": 2.4145, + "mean_token_accuracy": 0.7460378510378509, + "mean_token_accuracy_utility": 0.642416522553993, + "step": 295 + }, + { + "epoch": 0.11101859561476547, + "grad_norm": 0.95703125, + "learning_rate": 1.9992978255929168e-05, + "loss": 2.3684, + "mean_token_accuracy": 0.7392918192918192, + "mean_token_accuracy_utility": 0.6537667887667885, + "step": 300 + }, + { + "epoch": 0.11101859561476547, + "eval_mean_token_accuracy": 0.7384044060514651, + "eval_reasoning_loss": 0.9072665572166443, + "eval_reasoning_runtime": 27.7285, + "eval_reasoning_samples_per_second": 4.652, + "eval_reasoning_steps_per_second": 1.19, + "step": 300 + }, + { + "epoch": 0.11101859561476547, + "eval_mean_token_accuracy": 0.6499298020316033, + "eval_utility_loss": 1.4782265424728394, + "eval_utility_runtime": 10.4392, + "eval_utility_samples_per_second": 4.694, + "eval_utility_steps_per_second": 1.245, + "step": 300 + }, + { + "epoch": 0.11286890554167824, + "grad_norm": 0.97265625, + "learning_rate": 1.999034865600726e-05, + "loss": 2.469, + "mean_token_accuracy": 0.7296886446886448, + "mean_token_accuracy_utility": 0.647228327228327, + "step": 305 + }, + { + "epoch": 0.11471921546859098, + "grad_norm": 0.984375, + "learning_rate": 1.9987301947652354e-05, + "loss": 2.34, + "mean_token_accuracy": 0.7496947496947498, + "mean_token_accuracy_utility": 0.6489926739926738, + "step": 310 + }, + { + "epoch": 0.11656952539550375, + "grad_norm": 0.9609375, + "learning_rate": 1.998383825806799e-05, + "loss": 2.4057, + "mean_token_accuracy": 0.7502686202686204, + "mean_token_accuracy_utility": 0.643144078144078, + "step": 315 + }, + { + "epoch": 0.11841983532241651, + "grad_norm": 1.0078125, + "learning_rate": 1.9979957731867143e-05, + "loss": 2.4009, + "mean_token_accuracy": 0.7547313797313798, + "mean_token_accuracy_utility": 0.6399328449328447, + "step": 320 + }, + { + "epoch": 0.12027014524932926, + "grad_norm": 0.8984375, + "learning_rate": 1.9975660531066215e-05, + "loss": 2.3321, + "mean_token_accuracy": 0.7339072039072041, + "mean_token_accuracy_utility": 0.6634126984126982, + "step": 325 + }, + { + "epoch": 0.12212045517624202, + "grad_norm": 0.9453125, + "learning_rate": 1.9970946835078227e-05, + "loss": 2.3301, + "mean_token_accuracy": 0.7549328449328448, + "mean_token_accuracy_utility": 0.6485470085470084, + "step": 330 + }, + { + "epoch": 0.12397076510315477, + "grad_norm": 0.90234375, + "learning_rate": 1.9965816840705355e-05, + "loss": 2.3368, + "mean_token_accuracy": 0.7613003663003666, + "mean_token_accuracy_utility": 0.646275946275946, + "step": 335 + }, + { + "epoch": 0.12582107503006754, + "grad_norm": 0.96875, + "learning_rate": 1.9960270762130705e-05, + "loss": 2.379, + "mean_token_accuracy": 0.7510641050424544, + "mean_token_accuracy_utility": 0.6450793650793649, + "step": 340 + }, + { + "epoch": 0.1276713849569803, + "grad_norm": 0.9296875, + "learning_rate": 1.9954308830909372e-05, + "loss": 2.3996, + "mean_token_accuracy": 0.7592490842490843, + "mean_token_accuracy_utility": 0.6360195360195358, + "step": 345 + }, + { + "epoch": 0.12952169488389306, + "grad_norm": 0.875, + "learning_rate": 1.9947931295958778e-05, + "loss": 2.389, + "mean_token_accuracy": 0.7545848595848594, + "mean_token_accuracy_utility": 0.641214896214896, + "step": 350 + }, + { + "epoch": 0.1313720048108058, + "grad_norm": 0.8984375, + "learning_rate": 1.9941138423548266e-05, + "loss": 2.2637, + "mean_token_accuracy": 0.7532600732600734, + "mean_token_accuracy_utility": 0.662161172161172, + "step": 355 + }, + { + "epoch": 0.13322231473771856, + "grad_norm": 0.9453125, + "learning_rate": 1.9933930497287996e-05, + "loss": 2.3603, + "mean_token_accuracy": 0.7554090354090355, + "mean_token_accuracy_utility": 0.6483028083028082, + "step": 360 + }, + { + "epoch": 0.13507262466463132, + "grad_norm": 0.875, + "learning_rate": 1.9926307818117098e-05, + "loss": 2.3191, + "mean_token_accuracy": 0.7553296703296702, + "mean_token_accuracy_utility": 0.6577411477411476, + "step": 365 + }, + { + "epoch": 0.1369229345915441, + "grad_norm": 0.94140625, + "learning_rate": 1.9918270704291104e-05, + "loss": 2.3551, + "mean_token_accuracy": 0.7594871794871796, + "mean_token_accuracy_utility": 0.6467582417582417, + "step": 370 + }, + { + "epoch": 0.13877324451845685, + "grad_norm": 0.88671875, + "learning_rate": 1.9909819491368677e-05, + "loss": 2.3272, + "mean_token_accuracy": 0.7750061050061051, + "mean_token_accuracy_utility": 0.6403663003663002, + "step": 375 + }, + { + "epoch": 0.1406235544453696, + "grad_norm": 0.875, + "learning_rate": 1.990095453219757e-05, + "loss": 2.167, + "mean_token_accuracy": 0.7607326007326007, + "mean_token_accuracy_utility": 0.676263736263736, + "step": 380 + }, + { + "epoch": 0.14247386437228235, + "grad_norm": 0.9453125, + "learning_rate": 1.989167619689993e-05, + "loss": 2.3333, + "mean_token_accuracy": 0.752014652014652, + "mean_token_accuracy_utility": 0.6525060560406609, + "step": 385 + }, + { + "epoch": 0.1443241742991951, + "grad_norm": 0.89453125, + "learning_rate": 1.988198487285682e-05, + "loss": 2.3289, + "mean_token_accuracy": 0.7404578754578753, + "mean_token_accuracy_utility": 0.6670268620268619, + "step": 390 + }, + { + "epoch": 0.14617448422610788, + "grad_norm": 0.91796875, + "learning_rate": 1.9871880964692055e-05, + "loss": 2.3107, + "mean_token_accuracy": 0.7522649572649573, + "mean_token_accuracy_utility": 0.659023199023199, + "step": 395 + }, + { + "epoch": 0.14802479415302064, + "grad_norm": 0.90625, + "learning_rate": 1.9861364894255306e-05, + "loss": 2.3356, + "mean_token_accuracy": 0.7610317460317462, + "mean_token_accuracy_utility": 0.6489682539682537, + "step": 400 + }, + { + "epoch": 0.14802479415302064, + "eval_mean_token_accuracy": 0.7448276124746713, + "eval_reasoning_loss": 0.8728643655776978, + "eval_reasoning_runtime": 27.7443, + "eval_reasoning_samples_per_second": 4.65, + "eval_reasoning_steps_per_second": 1.189, + "step": 400 + }, + { + "epoch": 0.14802479415302064, + "eval_mean_token_accuracy": 0.6572214174158696, + "eval_utility_loss": 1.4576278924942017, + "eval_utility_runtime": 10.4369, + "eval_utility_samples_per_second": 4.695, + "eval_utility_steps_per_second": 1.246, + "step": 400 + }, + { + "epoch": 0.1498751040799334, + "grad_norm": 0.93359375, + "learning_rate": 1.985043710060449e-05, + "loss": 2.2243, + "mean_token_accuracy": 0.7484737484737485, + "mean_token_accuracy_utility": 0.6717277167277165, + "step": 405 + }, + { + "epoch": 0.15172541400684614, + "grad_norm": 0.87109375, + "learning_rate": 1.9839098039987435e-05, + "loss": 2.2841, + "mean_token_accuracy": 0.7629242979242983, + "mean_token_accuracy_utility": 0.6583516483516482, + "step": 410 + }, + { + "epoch": 0.1535757239337589, + "grad_norm": 0.87890625, + "learning_rate": 1.9827348185822834e-05, + "loss": 2.3422, + "mean_token_accuracy": 0.759242979242979, + "mean_token_accuracy_utility": 0.648211233211233, + "step": 415 + }, + { + "epoch": 0.15542603386067166, + "grad_norm": 0.9453125, + "learning_rate": 1.981518802868048e-05, + "loss": 2.2349, + "mean_token_accuracy": 0.7533638583638582, + "mean_token_accuracy_utility": 0.6740354090354088, + "step": 420 + }, + { + "epoch": 0.15727634378758443, + "grad_norm": 0.8515625, + "learning_rate": 1.9802618076260784e-05, + "loss": 2.2703, + "mean_token_accuracy": 0.76519536019536, + "mean_token_accuracy_utility": 0.6586507936507935, + "step": 425 + }, + { + "epoch": 0.1591266537144972, + "grad_norm": 0.890625, + "learning_rate": 1.9789638853373563e-05, + "loss": 2.1796, + "mean_token_accuracy": 0.7605738705738707, + "mean_token_accuracy_utility": 0.6739194139194138, + "step": 430 + }, + { + "epoch": 0.16097696364140993, + "grad_norm": 0.87890625, + "learning_rate": 1.9776250901916168e-05, + "loss": 2.3352, + "mean_token_accuracy": 0.7380769230769232, + "mean_token_accuracy_utility": 0.6620146520146518, + "step": 435 + }, + { + "epoch": 0.1628272735683227, + "grad_norm": 0.8984375, + "learning_rate": 1.9762454780850807e-05, + "loss": 2.2704, + "mean_token_accuracy": 0.7576129426129425, + "mean_token_accuracy_utility": 0.6587423687423686, + "step": 440 + }, + { + "epoch": 0.16467758349523545, + "grad_norm": 0.91015625, + "learning_rate": 1.9748251066181247e-05, + "loss": 2.3133, + "mean_token_accuracy": 0.7630463980463981, + "mean_token_accuracy_utility": 0.6472527472527471, + "step": 445 + }, + { + "epoch": 0.16652789342214822, + "grad_norm": 0.921875, + "learning_rate": 1.973364035092875e-05, + "loss": 2.3992, + "mean_token_accuracy": 0.745103785103785, + "mean_token_accuracy_utility": 0.6436507936507934, + "step": 450 + }, + { + "epoch": 0.16837820334906098, + "grad_norm": 0.875, + "learning_rate": 1.971862324510732e-05, + "loss": 2.3524, + "mean_token_accuracy": 0.7407203907203906, + "mean_token_accuracy_utility": 0.6602319902319901, + "step": 455 + }, + { + "epoch": 0.17022851327597371, + "grad_norm": 0.91015625, + "learning_rate": 1.9703200375698223e-05, + "loss": 2.3692, + "mean_token_accuracy": 0.7462759462759462, + "mean_token_accuracy_utility": 0.6480280830280829, + "step": 460 + }, + { + "epoch": 0.17207882320288648, + "grad_norm": 0.84375, + "learning_rate": 1.968737238662382e-05, + "loss": 2.1892, + "mean_token_accuracy": 0.7690293040293043, + "mean_token_accuracy_utility": 0.6689590323111594, + "step": 465 + }, + { + "epoch": 0.17392913312979924, + "grad_norm": 0.91796875, + "learning_rate": 1.9671139938720678e-05, + "loss": 2.2921, + "mean_token_accuracy": 0.7616910866910866, + "mean_token_accuracy_utility": 0.6520695970695969, + "step": 470 + }, + { + "epoch": 0.175779443056712, + "grad_norm": 0.93359375, + "learning_rate": 1.9654503709711984e-05, + "loss": 2.32, + "mean_token_accuracy": 0.7516849816849817, + "mean_token_accuracy_utility": 0.6562881562881561, + "step": 475 + }, + { + "epoch": 0.17762975298362477, + "grad_norm": 0.8984375, + "learning_rate": 1.963746439417924e-05, + "loss": 2.3225, + "mean_token_accuracy": 0.7676678876678877, + "mean_token_accuracy_utility": 0.6508852258852257, + "step": 480 + }, + { + "epoch": 0.1794800629105375, + "grad_norm": 0.8828125, + "learning_rate": 1.962002270353328e-05, + "loss": 2.2549, + "mean_token_accuracy": 0.7675213675213676, + "mean_token_accuracy_utility": 0.6527594627594626, + "step": 485 + }, + { + "epoch": 0.18133037283745027, + "grad_norm": 0.85546875, + "learning_rate": 1.960217936598454e-05, + "loss": 2.1422, + "mean_token_accuracy": 0.768809523809524, + "mean_token_accuracy_utility": 0.6677228327228325, + "step": 490 + }, + { + "epoch": 0.18318068276436303, + "grad_norm": 0.89453125, + "learning_rate": 1.958393512651269e-05, + "loss": 2.2416, + "mean_token_accuracy": 0.7761477411477412, + "mean_token_accuracy_utility": 0.6483851184091929, + "step": 495 + }, + { + "epoch": 0.1850309926912758, + "grad_norm": 0.83984375, + "learning_rate": 1.956529074683551e-05, + "loss": 2.3034, + "mean_token_accuracy": 0.7571855921855921, + "mean_token_accuracy_utility": 0.6535470085470083, + "step": 500 + }, + { + "epoch": 0.1850309926912758, + "eval_mean_token_accuracy": 0.7476340652811243, + "eval_reasoning_loss": 0.8572344183921814, + "eval_reasoning_runtime": 27.73, + "eval_reasoning_samples_per_second": 4.652, + "eval_reasoning_steps_per_second": 1.19, + "step": 500 + }, + { + "epoch": 0.1850309926912758, + "eval_mean_token_accuracy": 0.6585142576420815, + "eval_utility_loss": 1.4485703706741333, + "eval_utility_runtime": 10.443, + "eval_utility_samples_per_second": 4.692, + "eval_utility_steps_per_second": 1.245, + "step": 500 + }, + { + "epoch": 0.18688130261818856, + "grad_norm": 0.8671875, + "learning_rate": 1.9546247005377065e-05, + "loss": 2.1913, + "mean_token_accuracy": 0.7614102564102565, + "mean_token_accuracy_utility": 0.6697985347985347, + "step": 505 + }, + { + "epoch": 0.18873161254510132, + "grad_norm": 0.8125, + "learning_rate": 1.952680469723526e-05, + "loss": 2.286, + "mean_token_accuracy": 0.7631623931623933, + "mean_token_accuracy_utility": 0.653125763125763, + "step": 510 + }, + { + "epoch": 0.19058192247201405, + "grad_norm": 0.87890625, + "learning_rate": 1.9506964634148597e-05, + "loss": 2.3304, + "mean_token_accuracy": 0.7570390720390722, + "mean_token_accuracy_utility": 0.6504517704517703, + "step": 515 + }, + { + "epoch": 0.19243223239892682, + "grad_norm": 0.91796875, + "learning_rate": 1.9486727644462306e-05, + "loss": 2.2784, + "mean_token_accuracy": 0.7577045177045177, + "mean_token_accuracy_utility": 0.6618681318681318, + "step": 520 + }, + { + "epoch": 0.19428254232583958, + "grad_norm": 0.9609375, + "learning_rate": 1.9466094573093744e-05, + "loss": 2.2625, + "mean_token_accuracy": 0.775964590964591, + "mean_token_accuracy_utility": 0.6486507936507935, + "step": 525 + }, + { + "epoch": 0.19613285225275234, + "grad_norm": 0.87109375, + "learning_rate": 1.9445066281497144e-05, + "loss": 2.2164, + "mean_token_accuracy": 0.7693162393162392, + "mean_token_accuracy_utility": 0.6604151404151402, + "step": 530 + }, + { + "epoch": 0.1979831621796651, + "grad_norm": 0.89453125, + "learning_rate": 1.9423643647627625e-05, + "loss": 2.2971, + "mean_token_accuracy": 0.7547557997557999, + "mean_token_accuracy_utility": 0.6588715500848441, + "step": 535 + }, + { + "epoch": 0.19983347210657784, + "grad_norm": 0.83203125, + "learning_rate": 1.940182756590454e-05, + "loss": 2.19, + "mean_token_accuracy": 0.7603540903540905, + "mean_token_accuracy_utility": 0.6699938949938948, + "step": 540 + }, + { + "epoch": 0.2016837820334906, + "grad_norm": 0.91796875, + "learning_rate": 1.9379618947174155e-05, + "loss": 2.2728, + "mean_token_accuracy": 0.7642918192918196, + "mean_token_accuracy_utility": 0.662197802197802, + "step": 545 + }, + { + "epoch": 0.20353409196040337, + "grad_norm": 0.921875, + "learning_rate": 1.935701871867158e-05, + "loss": 2.2293, + "mean_token_accuracy": 0.7636691086691088, + "mean_token_accuracy_utility": 0.6610012210012208, + "step": 550 + }, + { + "epoch": 0.20538440188731613, + "grad_norm": 0.890625, + "learning_rate": 1.9334027823982103e-05, + "loss": 2.218, + "mean_token_accuracy": 0.764061413061413, + "mean_token_accuracy_utility": 0.6633821733821732, + "step": 555 + }, + { + "epoch": 0.2072347118142289, + "grad_norm": 0.8828125, + "learning_rate": 1.9310647223001752e-05, + "loss": 2.2719, + "mean_token_accuracy": 0.7485042735042735, + "mean_token_accuracy_utility": 0.6627655677655675, + "step": 560 + }, + { + "epoch": 0.20908502174114163, + "grad_norm": 0.8984375, + "learning_rate": 1.9286877891897244e-05, + "loss": 2.2585, + "mean_token_accuracy": 0.7772832722832725, + "mean_token_accuracy_utility": 0.6499572649572648, + "step": 565 + }, + { + "epoch": 0.2109353316680544, + "grad_norm": 0.90234375, + "learning_rate": 1.9262720823065217e-05, + "loss": 2.308, + "mean_token_accuracy": 0.7471184371184373, + "mean_token_accuracy_utility": 0.6587545787545787, + "step": 570 + }, + { + "epoch": 0.21278564159496716, + "grad_norm": 0.83203125, + "learning_rate": 1.923817702509081e-05, + "loss": 2.2527, + "mean_token_accuracy": 0.7661050061050063, + "mean_token_accuracy_utility": 0.6553907203907203, + "step": 575 + }, + { + "epoch": 0.21463595152187992, + "grad_norm": 0.84765625, + "learning_rate": 1.9213247522705532e-05, + "loss": 2.2259, + "mean_token_accuracy": 0.7563492063492064, + "mean_token_accuracy_utility": 0.6694383394383392, + "step": 580 + }, + { + "epoch": 0.21648626144879268, + "grad_norm": 0.875, + "learning_rate": 1.9187933356744504e-05, + "loss": 2.2468, + "mean_token_accuracy": 0.7662026862026863, + "mean_token_accuracy_utility": 0.66015873015873, + "step": 585 + }, + { + "epoch": 0.21833657137570542, + "grad_norm": 0.85546875, + "learning_rate": 1.9162235584102973e-05, + "loss": 2.2189, + "mean_token_accuracy": 0.7604639804639804, + "mean_token_accuracy_utility": 0.661214896214896, + "step": 590 + }, + { + "epoch": 0.22018688130261818, + "grad_norm": 0.87109375, + "learning_rate": 1.9136155277692215e-05, + "loss": 2.2116, + "mean_token_accuracy": 0.7655128205128208, + "mean_token_accuracy_utility": 0.6594566544566542, + "step": 595 + }, + { + "epoch": 0.22203719122953094, + "grad_norm": 0.8984375, + "learning_rate": 1.9109693526394722e-05, + "loss": 2.2601, + "mean_token_accuracy": 0.7620757020757022, + "mean_token_accuracy_utility": 0.653089133089133, + "step": 600 + }, + { + "epoch": 0.22203719122953094, + "eval_mean_token_accuracy": 0.7497344703227057, + "eval_reasoning_loss": 0.848007321357727, + "eval_reasoning_runtime": 27.7433, + "eval_reasoning_samples_per_second": 4.65, + "eval_reasoning_steps_per_second": 1.189, + "step": 600 + }, + { + "epoch": 0.22203719122953094, + "eval_mean_token_accuracy": 0.6597615329099457, + "eval_utility_loss": 1.4425245523452759, + "eval_utility_runtime": 10.4344, + "eval_utility_samples_per_second": 4.696, + "eval_utility_steps_per_second": 1.246, + "step": 600 + }, + { + "epoch": 0.2238875011564437, + "grad_norm": 0.87890625, + "learning_rate": 1.9082851435018743e-05, + "loss": 2.1959, + "mean_token_accuracy": 0.7575030525030524, + "mean_token_accuracy_utility": 0.6734493284493281, + "step": 605 + }, + { + "epoch": 0.22573781108335647, + "grad_norm": 0.83984375, + "learning_rate": 1.905563012425216e-05, + "loss": 2.3294, + "mean_token_accuracy": 0.7663858363858365, + "mean_token_accuracy_utility": 0.6419597069597067, + "step": 610 + }, + { + "epoch": 0.2275881210102692, + "grad_norm": 0.88671875, + "learning_rate": 1.9028030730615696e-05, + "loss": 2.3709, + "mean_token_accuracy": 0.75519536019536, + "mean_token_accuracy_utility": 0.64517094017094, + "step": 615 + }, + { + "epoch": 0.22943843093718197, + "grad_norm": 0.93359375, + "learning_rate": 1.9000054406415467e-05, + "loss": 2.288, + "mean_token_accuracy": 0.7665995115995116, + "mean_token_accuracy_utility": 0.6514468864468863, + "step": 620 + }, + { + "epoch": 0.23128874086409473, + "grad_norm": 0.8515625, + "learning_rate": 1.897170231969486e-05, + "loss": 2.2617, + "mean_token_accuracy": 0.7613797313797315, + "mean_token_accuracy_utility": 0.6609645909645908, + "step": 625 + }, + { + "epoch": 0.2331390507910075, + "grad_norm": 0.87109375, + "learning_rate": 1.8942975654185788e-05, + "loss": 2.2089, + "mean_token_accuracy": 0.765927960927961, + "mean_token_accuracy_utility": 0.6652319902319901, + "step": 630 + }, + { + "epoch": 0.23498936071792026, + "grad_norm": 0.82421875, + "learning_rate": 1.8913875609259246e-05, + "loss": 2.1982, + "mean_token_accuracy": 0.7667704517704518, + "mean_token_accuracy_utility": 0.6630769230769229, + "step": 635 + }, + { + "epoch": 0.23683967064483302, + "grad_norm": 0.87890625, + "learning_rate": 1.8884403399875252e-05, + "loss": 2.2526, + "mean_token_accuracy": 0.757869352869353, + "mean_token_accuracy_utility": 0.6658871683871683, + "step": 640 + }, + { + "epoch": 0.23868998057174576, + "grad_norm": 0.859375, + "learning_rate": 1.8854560256532098e-05, + "loss": 2.1406, + "mean_token_accuracy": 0.7788355533355533, + "mean_token_accuracy_utility": 0.6690415140415139, + "step": 645 + }, + { + "epoch": 0.24054029049865852, + "grad_norm": 0.8515625, + "learning_rate": 1.8824347425215016e-05, + "loss": 2.2403, + "mean_token_accuracy": 0.7518070818070817, + "mean_token_accuracy_utility": 0.6668803418803416, + "step": 650 + }, + { + "epoch": 0.24239060042557128, + "grad_norm": 0.86328125, + "learning_rate": 1.8793766167344115e-05, + "loss": 2.1826, + "mean_token_accuracy": 0.7650732600732603, + "mean_token_accuracy_utility": 0.6655189255189253, + "step": 655 + }, + { + "epoch": 0.24424091035248405, + "grad_norm": 0.921875, + "learning_rate": 1.8762817759721735e-05, + "loss": 2.3232, + "mean_token_accuracy": 0.7635531135531137, + "mean_token_accuracy_utility": 0.645125421698776, + "step": 660 + }, + { + "epoch": 0.2460912202793968, + "grad_norm": 0.875, + "learning_rate": 1.8731503494479132e-05, + "loss": 2.2969, + "mean_token_accuracy": 0.7564529914529916, + "mean_token_accuracy_utility": 0.6526007326007324, + "step": 665 + }, + { + "epoch": 0.24794153020630955, + "grad_norm": 0.8359375, + "learning_rate": 1.869982467902255e-05, + "loss": 2.2874, + "mean_token_accuracy": 0.768742368742369, + "mean_token_accuracy_utility": 0.6486813186813186, + "step": 670 + }, + { + "epoch": 0.2497918401332223, + "grad_norm": 0.8671875, + "learning_rate": 1.8667782635978597e-05, + "loss": 2.3294, + "mean_token_accuracy": 0.7513919413919414, + "mean_token_accuracy_utility": 0.6497533352281044, + "step": 675 + }, + { + "epoch": 0.2516421500601351, + "grad_norm": 0.86328125, + "learning_rate": 1.8635378703139066e-05, + "loss": 2.1555, + "mean_token_accuracy": 0.7744871794871796, + "mean_token_accuracy_utility": 0.6653052503052501, + "step": 680 + }, + { + "epoch": 0.25349245998704784, + "grad_norm": 0.8828125, + "learning_rate": 1.8602614233405047e-05, + "loss": 2.2473, + "mean_token_accuracy": 0.7644139194139197, + "mean_token_accuracy_utility": 0.6566644274602751, + "step": 685 + }, + { + "epoch": 0.2553427699139606, + "grad_norm": 0.89453125, + "learning_rate": 1.8569490594730474e-05, + "loss": 2.2408, + "mean_token_accuracy": 0.7611782661782664, + "mean_token_accuracy_utility": 0.6658547008547007, + "step": 690 + }, + { + "epoch": 0.25719307984087336, + "grad_norm": 0.85546875, + "learning_rate": 1.853600917006497e-05, + "loss": 2.14, + "mean_token_accuracy": 0.7767521367521368, + "mean_token_accuracy_utility": 0.670140415140415, + "step": 695 + }, + { + "epoch": 0.2590433897677861, + "grad_norm": 0.9140625, + "learning_rate": 1.8502171357296144e-05, + "loss": 2.1679, + "mean_token_accuracy": 0.7762942612942613, + "mean_token_accuracy_utility": 0.6643711843711841, + "step": 700 + }, + { + "epoch": 0.2590433897677861, + "eval_mean_token_accuracy": 0.7508611214493569, + "eval_reasoning_loss": 0.8416144847869873, + "eval_reasoning_runtime": 27.7544, + "eval_reasoning_samples_per_second": 4.648, + "eval_reasoning_steps_per_second": 1.189, + "step": 700 + }, + { + "epoch": 0.2590433897677861, + "eval_mean_token_accuracy": 0.6597418579260157, + "eval_utility_loss": 1.4384644031524658, + "eval_utility_runtime": 10.4497, + "eval_utility_samples_per_second": 4.689, + "eval_utility_steps_per_second": 1.244, + "step": 700 + }, + { + "epoch": 0.2608936996946989, + "grad_norm": 0.87890625, + "learning_rate": 1.8467978569191216e-05, + "loss": 2.2653, + "mean_token_accuracy": 0.7617643467643469, + "mean_token_accuracy_utility": 0.6567582417582416, + "step": 705 + }, + { + "epoch": 0.2627440096216116, + "grad_norm": 0.84765625, + "learning_rate": 1.8433432233338027e-05, + "loss": 2.1812, + "mean_token_accuracy": 0.771874236874237, + "mean_token_accuracy_utility": 0.6699084249084247, + "step": 710 + }, + { + "epoch": 0.26459431954852436, + "grad_norm": 0.87890625, + "learning_rate": 1.8398533792085436e-05, + "loss": 2.2335, + "mean_token_accuracy": 0.7629426129426129, + "mean_token_accuracy_utility": 0.6616136574822234, + "step": 715 + }, + { + "epoch": 0.2664446294754371, + "grad_norm": 0.86328125, + "learning_rate": 1.8363284702483106e-05, + "loss": 2.3028, + "mean_token_accuracy": 0.7560805860805861, + "mean_token_accuracy_utility": 0.6505433455433454, + "step": 720 + }, + { + "epoch": 0.2682949394023499, + "grad_norm": 0.87890625, + "learning_rate": 1.832768643622067e-05, + "loss": 2.2481, + "mean_token_accuracy": 0.7721062271062273, + "mean_token_accuracy_utility": 0.6509829059829058, + "step": 725 + }, + { + "epoch": 0.27014524932926265, + "grad_norm": 0.83203125, + "learning_rate": 1.8291740479566286e-05, + "loss": 2.1525, + "mean_token_accuracy": 0.7696336996336998, + "mean_token_accuracy_utility": 0.6757509157509155, + "step": 730 + }, + { + "epoch": 0.2719955592561754, + "grad_norm": 0.86328125, + "learning_rate": 1.825544833330457e-05, + "loss": 2.1781, + "mean_token_accuracy": 0.765995115995116, + "mean_token_accuracy_utility": 0.6678205128205127, + "step": 735 + }, + { + "epoch": 0.2738458691830882, + "grad_norm": 0.89453125, + "learning_rate": 1.8218811512673958e-05, + "loss": 2.2555, + "mean_token_accuracy": 0.7531623931623932, + "mean_token_accuracy_utility": 0.6669169719169716, + "step": 740 + }, + { + "epoch": 0.27569617911000094, + "grad_norm": 0.8203125, + "learning_rate": 1.818183154730344e-05, + "loss": 2.2444, + "mean_token_accuracy": 0.7621916971916973, + "mean_token_accuracy_utility": 0.6587423687423686, + "step": 745 + }, + { + "epoch": 0.2775464890369137, + "grad_norm": 0.87109375, + "learning_rate": 1.8144509981148675e-05, + "loss": 2.232, + "mean_token_accuracy": 0.7663858363858362, + "mean_token_accuracy_utility": 0.6583455433455432, + "step": 750 + }, + { + "epoch": 0.27939679896382646, + "grad_norm": 0.8671875, + "learning_rate": 1.810684837242755e-05, + "loss": 2.3373, + "mean_token_accuracy": 0.7610744810744812, + "mean_token_accuracy_utility": 0.6407328855503296, + "step": 755 + }, + { + "epoch": 0.2812471088907392, + "grad_norm": 0.8984375, + "learning_rate": 1.8068848293555118e-05, + "loss": 2.222, + "mean_token_accuracy": 0.7605677655677655, + "mean_token_accuracy_utility": 0.6656288156288155, + "step": 760 + }, + { + "epoch": 0.28309741881765194, + "grad_norm": 0.8125, + "learning_rate": 1.8030511331077945e-05, + "loss": 2.1907, + "mean_token_accuracy": 0.7641636141636144, + "mean_token_accuracy_utility": 0.6702747252747251, + "step": 765 + }, + { + "epoch": 0.2849477287445647, + "grad_norm": 0.86328125, + "learning_rate": 1.799183908560787e-05, + "loss": 2.1966, + "mean_token_accuracy": 0.7526617826617826, + "mean_token_accuracy_utility": 0.6761416361416359, + "step": 770 + }, + { + "epoch": 0.28679803867147746, + "grad_norm": 0.8828125, + "learning_rate": 1.795283317175518e-05, + "loss": 2.2042, + "mean_token_accuracy": 0.7722832722832724, + "mean_token_accuracy_utility": 0.6603052503052502, + "step": 775 + }, + { + "epoch": 0.2886483485983902, + "grad_norm": 0.87890625, + "learning_rate": 1.7913495218061202e-05, + "loss": 2.2837, + "mean_token_accuracy": 0.7629853479853479, + "mean_token_accuracy_utility": 0.6535897435897434, + "step": 780 + }, + { + "epoch": 0.290498658525303, + "grad_norm": 0.83984375, + "learning_rate": 1.787382686693029e-05, + "loss": 2.2776, + "mean_token_accuracy": 0.766770451770452, + "mean_token_accuracy_utility": 0.6527350427350426, + "step": 785 + }, + { + "epoch": 0.29234896845221575, + "grad_norm": 0.84765625, + "learning_rate": 1.783382977456128e-05, + "loss": 2.1359, + "mean_token_accuracy": 0.7707203907203908, + "mean_token_accuracy_utility": 0.6725396825396823, + "step": 790 + }, + { + "epoch": 0.2941992783791285, + "grad_norm": 0.89453125, + "learning_rate": 1.779350561087833e-05, + "loss": 2.2826, + "mean_token_accuracy": 0.7553968253968255, + "mean_token_accuracy_utility": 0.6610683760683759, + "step": 795 + }, + { + "epoch": 0.2960495883060413, + "grad_norm": 0.91015625, + "learning_rate": 1.775285605946119e-05, + "loss": 2.1529, + "mean_token_accuracy": 0.7740476190476191, + "mean_token_accuracy_utility": 0.6706715506715505, + "step": 800 + }, + { + "epoch": 0.2960495883060413, + "eval_mean_token_accuracy": 0.7518566400919344, + "eval_reasoning_loss": 0.8366485238075256, + "eval_reasoning_runtime": 27.7323, + "eval_reasoning_samples_per_second": 4.652, + "eval_reasoning_steps_per_second": 1.19, + "step": 800 + }, + { + "epoch": 0.2960495883060413, + "eval_mean_token_accuracy": 0.6604523744929808, + "eval_utility_loss": 1.435474157333374, + "eval_utility_runtime": 10.4374, + "eval_utility_samples_per_second": 4.695, + "eval_utility_steps_per_second": 1.246, + "step": 800 + }, + { + "epoch": 0.29789989823295404, + "grad_norm": 0.859375, + "learning_rate": 1.7711882817474922e-05, + "loss": 2.2062, + "mean_token_accuracy": 0.7702503052503056, + "mean_token_accuracy_utility": 0.6617216117216115, + "step": 805 + }, + { + "epoch": 0.2997502081598668, + "grad_norm": 0.86328125, + "learning_rate": 1.7670587595599034e-05, + "loss": 2.2011, + "mean_token_accuracy": 0.7684920634920636, + "mean_token_accuracy_utility": 0.6702564102564101, + "step": 810 + }, + { + "epoch": 0.3016005180867795, + "grad_norm": 0.890625, + "learning_rate": 1.762897211795607e-05, + "loss": 2.2761, + "mean_token_accuracy": 0.7547374847374846, + "mean_token_accuracy_utility": 0.654114774114774, + "step": 815 + }, + { + "epoch": 0.3034508280136923, + "grad_norm": 0.8359375, + "learning_rate": 1.758703812203961e-05, + "loss": 2.2881, + "mean_token_accuracy": 0.7541941391941394, + "mean_token_accuracy_utility": 0.6578327228327228, + "step": 820 + }, + { + "epoch": 0.30530113794060504, + "grad_norm": 0.87890625, + "learning_rate": 1.7544787358641735e-05, + "loss": 2.3264, + "mean_token_accuracy": 0.7441269841269842, + "mean_token_accuracy_utility": 0.6584859584859583, + "step": 825 + }, + { + "epoch": 0.3071514478675178, + "grad_norm": 0.91796875, + "learning_rate": 1.7502221591779932e-05, + "loss": 2.2547, + "mean_token_accuracy": 0.7444749694749696, + "mean_token_accuracy_utility": 0.6685382910465665, + "step": 830 + }, + { + "epoch": 0.30900175779443056, + "grad_norm": 0.85546875, + "learning_rate": 1.7459342598623438e-05, + "loss": 2.2185, + "mean_token_accuracy": 0.7646642246642248, + "mean_token_accuracy_utility": 0.6677777777777776, + "step": 835 + }, + { + "epoch": 0.31085206772134333, + "grad_norm": 0.86328125, + "learning_rate": 1.741615216941905e-05, + "loss": 2.1604, + "mean_token_accuracy": 0.7760561660561661, + "mean_token_accuracy_utility": 0.668217338217338, + "step": 840 + }, + { + "epoch": 0.3127023776482561, + "grad_norm": 0.8359375, + "learning_rate": 1.7372652107416364e-05, + "loss": 2.279, + "mean_token_accuracy": 0.7592612942612944, + "mean_token_accuracy_utility": 0.6571123321123319, + "step": 845 + }, + { + "epoch": 0.31455268757516885, + "grad_norm": 0.8359375, + "learning_rate": 1.7328844228792513e-05, + "loss": 2.1851, + "mean_token_accuracy": 0.7823137973137975, + "mean_token_accuracy_utility": 0.6595115995115993, + "step": 850 + }, + { + "epoch": 0.3164029975020816, + "grad_norm": 0.8671875, + "learning_rate": 1.7284730362576308e-05, + "loss": 2.2524, + "mean_token_accuracy": 0.761886446886447, + "mean_token_accuracy_utility": 0.6588095238095237, + "step": 855 + }, + { + "epoch": 0.3182533074289944, + "grad_norm": 0.88671875, + "learning_rate": 1.7240312350571905e-05, + "loss": 2.3422, + "mean_token_accuracy": 0.7682173382173383, + "mean_token_accuracy_utility": 0.6394200244200243, + "step": 860 + }, + { + "epoch": 0.3201036173559071, + "grad_norm": 0.83984375, + "learning_rate": 1.719559204728188e-05, + "loss": 2.2142, + "mean_token_accuracy": 0.7756959706959708, + "mean_token_accuracy_utility": 0.655115995115995, + "step": 865 + }, + { + "epoch": 0.32195392728281985, + "grad_norm": 0.88671875, + "learning_rate": 1.715057131982983e-05, + "loss": 2.3012, + "mean_token_accuracy": 0.7663064713064712, + "mean_token_accuracy_utility": 0.6465689865689864, + "step": 870 + }, + { + "epoch": 0.3238042372097326, + "grad_norm": 0.8984375, + "learning_rate": 1.710525204788239e-05, + "loss": 2.167, + "mean_token_accuracy": 0.772899877899878, + "mean_token_accuracy_utility": 0.6668890372905564, + "step": 875 + }, + { + "epoch": 0.3256545471366454, + "grad_norm": 0.875, + "learning_rate": 1.7059636123570767e-05, + "loss": 2.2658, + "mean_token_accuracy": 0.7663431013431016, + "mean_token_accuracy_utility": 0.6517338217338216, + "step": 880 + }, + { + "epoch": 0.32750485706355814, + "grad_norm": 0.80859375, + "learning_rate": 1.7013725451411757e-05, + "loss": 2.1919, + "mean_token_accuracy": 0.7793528693528695, + "mean_token_accuracy_utility": 0.661294261294261, + "step": 885 + }, + { + "epoch": 0.3293551669904709, + "grad_norm": 0.84765625, + "learning_rate": 1.696752194822819e-05, + "loss": 2.2566, + "mean_token_accuracy": 0.7771306471306472, + "mean_token_accuracy_utility": 0.6468803418803418, + "step": 890 + }, + { + "epoch": 0.33120547691738367, + "grad_norm": 0.84375, + "learning_rate": 1.692102754306895e-05, + "loss": 2.2469, + "mean_token_accuracy": 0.776807081807082, + "mean_token_accuracy_utility": 0.6463492063492062, + "step": 895 + }, + { + "epoch": 0.33305578684429643, + "grad_norm": 0.8515625, + "learning_rate": 1.6874244177128395e-05, + "loss": 2.148, + "mean_token_accuracy": 0.7748229548229549, + "mean_token_accuracy_utility": 0.6681990231990229, + "step": 900 + }, + { + "epoch": 0.33305578684429643, + "eval_mean_token_accuracy": 0.7523170729053084, + "eval_reasoning_loss": 0.8331823945045471, + "eval_reasoning_runtime": 27.7437, + "eval_reasoning_samples_per_second": 4.65, + "eval_reasoning_steps_per_second": 1.189, + "step": 900 + }, + { + "epoch": 0.33305578684429643, + "eval_mean_token_accuracy": 0.6607885950036365, + "eval_utility_loss": 1.433144211769104, + "eval_utility_runtime": 10.4401, + "eval_utility_samples_per_second": 4.693, + "eval_utility_steps_per_second": 1.245, + "step": 900 + }, + { + "epoch": 0.3349060967712092, + "grad_norm": 0.87109375, + "learning_rate": 1.6827173803665328e-05, + "loss": 2.2904, + "mean_token_accuracy": 0.7511355311355311, + "mean_token_accuracy_utility": 0.6667704517704516, + "step": 905 + }, + { + "epoch": 0.33675640669812196, + "grad_norm": 0.86328125, + "learning_rate": 1.677981838792144e-05, + "loss": 2.2306, + "mean_token_accuracy": 0.7610744810744812, + "mean_token_accuracy_utility": 0.6624053724053722, + "step": 910 + }, + { + "epoch": 0.3386067166250347, + "grad_norm": 0.83984375, + "learning_rate": 1.6732179907039266e-05, + "loss": 2.1361, + "mean_token_accuracy": 0.7669841269841272, + "mean_token_accuracy_utility": 0.6727838827838826, + "step": 915 + }, + { + "epoch": 0.34045702655194743, + "grad_norm": 0.86328125, + "learning_rate": 1.6684260349979637e-05, + "loss": 2.1957, + "mean_token_accuracy": 0.763021978021978, + "mean_token_accuracy_utility": 0.6665628815628813, + "step": 920 + }, + { + "epoch": 0.3423073364788602, + "grad_norm": 0.83203125, + "learning_rate": 1.6636061717438626e-05, + "loss": 2.1017, + "mean_token_accuracy": 0.7771123321123323, + "mean_token_accuracy_utility": 0.6746336996336995, + "step": 925 + }, + { + "epoch": 0.34415764640577295, + "grad_norm": 0.8125, + "learning_rate": 1.6587586021764022e-05, + "loss": 2.1425, + "mean_token_accuracy": 0.7725579975579977, + "mean_token_accuracy_utility": 0.6699901279755461, + "step": 930 + }, + { + "epoch": 0.3460079563326857, + "grad_norm": 0.84375, + "learning_rate": 1.653883528687133e-05, + "loss": 2.2476, + "mean_token_accuracy": 0.7644485824151837, + "mean_token_accuracy_utility": 0.6531803528973843, + "step": 935 + }, + { + "epoch": 0.3478582662595985, + "grad_norm": 0.875, + "learning_rate": 1.6489811548159245e-05, + "loss": 2.2232, + "mean_token_accuracy": 0.7602380952380953, + "mean_token_accuracy_utility": 0.6604578754578754, + "step": 940 + }, + { + "epoch": 0.34970857618651124, + "grad_norm": 0.87109375, + "learning_rate": 1.6440516852424678e-05, + "loss": 2.2434, + "mean_token_accuracy": 0.7594932844932846, + "mean_token_accuracy_utility": 0.6621367521367519, + "step": 945 + }, + { + "epoch": 0.351558886113424, + "grad_norm": 0.8125, + "learning_rate": 1.6390953257777324e-05, + "loss": 2.1483, + "mean_token_accuracy": 0.7724542124542125, + "mean_token_accuracy_utility": 0.6709829059829058, + "step": 950 + }, + { + "epoch": 0.35340919604033677, + "grad_norm": 0.8828125, + "learning_rate": 1.634112283355369e-05, + "loss": 2.3033, + "mean_token_accuracy": 0.7588339438339439, + "mean_token_accuracy_utility": 0.6488095238095235, + "step": 955 + }, + { + "epoch": 0.35525950596724953, + "grad_norm": 0.875, + "learning_rate": 1.6291027660230735e-05, + "loss": 2.1805, + "mean_token_accuracy": 0.7677777777777778, + "mean_token_accuracy_utility": 0.6668015546410607, + "step": 960 + }, + { + "epoch": 0.3571098158941623, + "grad_norm": 0.94140625, + "learning_rate": 1.6240669829338992e-05, + "loss": 2.2921, + "mean_token_accuracy": 0.7516117216117215, + "mean_token_accuracy_utility": 0.6515201465201462, + "step": 965 + }, + { + "epoch": 0.358960125821075, + "grad_norm": 0.84375, + "learning_rate": 1.6190051443375248e-05, + "loss": 2.1649, + "mean_token_accuracy": 0.7676068376068377, + "mean_token_accuracy_utility": 0.6706227106227104, + "step": 970 + }, + { + "epoch": 0.36081043574798777, + "grad_norm": 0.90625, + "learning_rate": 1.6139174615714753e-05, + "loss": 2.2259, + "mean_token_accuracy": 0.7662026862026863, + "mean_token_accuracy_utility": 0.6614733978055776, + "step": 975 + }, + { + "epoch": 0.36266074567490053, + "grad_norm": 0.89453125, + "learning_rate": 1.6088041470523005e-05, + "loss": 2.2743, + "mean_token_accuracy": 0.7533028083028086, + "mean_token_accuracy_utility": 0.6611904761904759, + "step": 980 + }, + { + "epoch": 0.3645110556018133, + "grad_norm": 0.8828125, + "learning_rate": 1.6036654142667043e-05, + "loss": 2.2837, + "mean_token_accuracy": 0.7561111111111114, + "mean_token_accuracy_utility": 0.6583865884211904, + "step": 985 + }, + { + "epoch": 0.36636136552872606, + "grad_norm": 0.85546875, + "learning_rate": 1.598501477762632e-05, + "loss": 2.263, + "mean_token_accuracy": 0.7605860805860806, + "mean_token_accuracy_utility": 0.6579853479853478, + "step": 990 + }, + { + "epoch": 0.3682116754556388, + "grad_norm": 0.84765625, + "learning_rate": 1.5933125531403135e-05, + "loss": 2.3075, + "mean_token_accuracy": 0.7622358920434631, + "mean_token_accuracy_utility": 0.6445665445665444, + "step": 995 + }, + { + "epoch": 0.3700619853825516, + "grad_norm": 0.87890625, + "learning_rate": 1.5880988570432603e-05, + "loss": 2.2596, + "mean_token_accuracy": 0.7767826617826619, + "mean_token_accuracy_utility": 0.6463003663003661, + "step": 1000 + }, + { + "epoch": 0.3700619853825516, + "eval_mean_token_accuracy": 0.7528776234658587, + "eval_reasoning_loss": 0.8303592801094055, + "eval_reasoning_runtime": 27.7483, + "eval_reasoning_samples_per_second": 4.649, + "eval_reasoning_steps_per_second": 1.189, + "step": 1000 + }, + { + "epoch": 0.3700619853825516, + "eval_mean_token_accuracy": 0.6614690317149569, + "eval_utility_loss": 1.4314581155776978, + "eval_utility_runtime": 10.4422, + "eval_utility_samples_per_second": 4.693, + "eval_utility_steps_per_second": 1.245, + "step": 1000 + }, + { + "epoch": 0.37191229530946435, + "grad_norm": 0.8671875, + "learning_rate": 1.582860607149222e-05, + "loss": 2.2095, + "mean_token_accuracy": 0.7639926739926739, + "mean_token_accuracy_utility": 0.662228327228327, + "step": 1005 + }, + { + "epoch": 0.3737626052363771, + "grad_norm": 0.84375, + "learning_rate": 1.5775980221610966e-05, + "loss": 2.2617, + "mean_token_accuracy": 0.7609279609279611, + "mean_token_accuracy_utility": 0.6590907973796387, + "step": 1010 + }, + { + "epoch": 0.3756129151632899, + "grad_norm": 0.84375, + "learning_rate": 1.5723113217978e-05, + "loss": 2.1718, + "mean_token_accuracy": 0.7652747252747254, + "mean_token_accuracy_utility": 0.6763125763125761, + "step": 1015 + }, + { + "epoch": 0.37746322509020264, + "grad_norm": 0.84375, + "learning_rate": 1.567000726785093e-05, + "loss": 2.0829, + "mean_token_accuracy": 0.7757203907203909, + "mean_token_accuracy_utility": 0.6816178266178264, + "step": 1020 + }, + { + "epoch": 0.37931353501711534, + "grad_norm": 0.875, + "learning_rate": 1.561666458846365e-05, + "loss": 2.2744, + "mean_token_accuracy": 0.751050061050061, + "mean_token_accuracy_utility": 0.6600671550671549, + "step": 1025 + }, + { + "epoch": 0.3811638449440281, + "grad_norm": 0.83984375, + "learning_rate": 1.5563087406933762e-05, + "loss": 2.212, + "mean_token_accuracy": 0.7576984126984128, + "mean_token_accuracy_utility": 0.6649206349206347, + "step": 1030 + }, + { + "epoch": 0.38301415487094087, + "grad_norm": 0.875, + "learning_rate": 1.550927796016961e-05, + "loss": 2.2775, + "mean_token_accuracy": 0.7679853479853478, + "mean_token_accuracy_utility": 0.6467826617826616, + "step": 1035 + }, + { + "epoch": 0.38486446479785363, + "grad_norm": 0.88671875, + "learning_rate": 1.5455238494776876e-05, + "loss": 2.2516, + "mean_token_accuracy": 0.7564774114774117, + "mean_token_accuracy_utility": 0.6606837606837604, + "step": 1040 + }, + { + "epoch": 0.3867147747247664, + "grad_norm": 0.8515625, + "learning_rate": 1.5400971266964772e-05, + "loss": 2.1765, + "mean_token_accuracy": 0.7751098901098904, + "mean_token_accuracy_utility": 0.6612531939605109, + "step": 1045 + }, + { + "epoch": 0.38856508465167916, + "grad_norm": 0.828125, + "learning_rate": 1.5346478542451862e-05, + "loss": 2.1546, + "mean_token_accuracy": 0.7601465201465203, + "mean_token_accuracy_utility": 0.6783982991347483, + "step": 1050 + }, + { + "epoch": 0.3904153945785919, + "grad_norm": 0.84765625, + "learning_rate": 1.529176259637145e-05, + "loss": 2.3141, + "mean_token_accuracy": 0.7625579975579977, + "mean_token_accuracy_utility": 0.6494261294261292, + "step": 1055 + }, + { + "epoch": 0.3922657045055047, + "grad_norm": 0.8515625, + "learning_rate": 1.5236825713176584e-05, + "loss": 2.2634, + "mean_token_accuracy": 0.7579120879120881, + "mean_token_accuracy_utility": 0.6571245421245419, + "step": 1060 + }, + { + "epoch": 0.39411601443241745, + "grad_norm": 0.83984375, + "learning_rate": 1.5181670186544706e-05, + "loss": 2.298, + "mean_token_accuracy": 0.7576434676434678, + "mean_token_accuracy_utility": 0.6545506968563962, + "step": 1065 + }, + { + "epoch": 0.3959663243593302, + "grad_norm": 0.83984375, + "learning_rate": 1.5126298319281859e-05, + "loss": 2.2189, + "mean_token_accuracy": 0.7800732600732603, + "mean_token_accuracy_utility": 0.6564896214896213, + "step": 1070 + }, + { + "epoch": 0.3978166342862429, + "grad_norm": 0.859375, + "learning_rate": 1.5070712423226552e-05, + "loss": 2.1591, + "mean_token_accuracy": 0.7723687423687424, + "mean_token_accuracy_utility": 0.6704273504273502, + "step": 1075 + }, + { + "epoch": 0.3996669442131557, + "grad_norm": 0.84375, + "learning_rate": 1.5014914819153252e-05, + "loss": 2.1617, + "mean_token_accuracy": 0.7552197802197801, + "mean_token_accuracy_utility": 0.6753113553113551, + "step": 1080 + }, + { + "epoch": 0.40151725414006845, + "grad_norm": 0.83203125, + "learning_rate": 1.4958907836675467e-05, + "loss": 2.2267, + "mean_token_accuracy": 0.7616544566544567, + "mean_token_accuracy_utility": 0.6627350427350425, + "step": 1085 + }, + { + "epoch": 0.4033675640669812, + "grad_norm": 0.87109375, + "learning_rate": 1.490269381414849e-05, + "loss": 2.2135, + "mean_token_accuracy": 0.7647619047619048, + "mean_token_accuracy_utility": 0.6602747252747252, + "step": 1090 + }, + { + "epoch": 0.405217873993894, + "grad_norm": 0.87109375, + "learning_rate": 1.484627509857178e-05, + "loss": 2.1484, + "mean_token_accuracy": 0.7669963369963373, + "mean_token_accuracy_utility": 0.6744261294261291, + "step": 1095 + }, + { + "epoch": 0.40706818392080674, + "grad_norm": 0.87890625, + "learning_rate": 1.4789654045490957e-05, + "loss": 2.2, + "mean_token_accuracy": 0.7731074481074482, + "mean_token_accuracy_utility": 0.6573931623931621, + "step": 1100 + }, + { + "epoch": 0.40706818392080674, + "eval_mean_token_accuracy": 0.7531588237470588, + "eval_reasoning_loss": 0.8281946778297424, + "eval_reasoning_runtime": 27.7375, + "eval_reasoning_samples_per_second": 4.651, + "eval_reasoning_steps_per_second": 1.19, + "step": 1100 + }, + { + "epoch": 0.40706818392080674, + "eval_mean_token_accuracy": 0.6614770274049655, + "eval_utility_loss": 1.4300123453140259, + "eval_utility_runtime": 10.4367, + "eval_utility_samples_per_second": 4.695, + "eval_utility_steps_per_second": 1.246, + "step": 1100 + }, + { + "epoch": 0.4089184938477195, + "grad_norm": 0.8984375, + "learning_rate": 1.4732833018899468e-05, + "loss": 2.2515, + "mean_token_accuracy": 0.7592918192918192, + "mean_token_accuracy_utility": 0.6640109890109888, + "step": 1105 + }, + { + "epoch": 0.41076880377463226, + "grad_norm": 0.84375, + "learning_rate": 1.4675814391139875e-05, + "loss": 2.2107, + "mean_token_accuracy": 0.7621794871794874, + "mean_token_accuracy_utility": 0.6625091575091574, + "step": 1110 + }, + { + "epoch": 0.412619113701545, + "grad_norm": 0.859375, + "learning_rate": 1.4618600542804819e-05, + "loss": 2.2848, + "mean_token_accuracy": 0.7436874236874236, + "mean_token_accuracy_utility": 0.6574114774114773, + "step": 1115 + }, + { + "epoch": 0.4144694236284578, + "grad_norm": 0.8515625, + "learning_rate": 1.4561193862637621e-05, + "loss": 2.1546, + "mean_token_accuracy": 0.767014652014652, + "mean_token_accuracy_utility": 0.6792307692307691, + "step": 1120 + }, + { + "epoch": 0.4163197335553705, + "grad_norm": 0.8671875, + "learning_rate": 1.4503596747432554e-05, + "loss": 2.2639, + "mean_token_accuracy": 0.7653846153846154, + "mean_token_accuracy_utility": 0.6551763903147985, + "step": 1125 + }, + { + "epoch": 0.41817004348228326, + "grad_norm": 0.82421875, + "learning_rate": 1.4445811601934763e-05, + "loss": 2.1342, + "mean_token_accuracy": 0.7680586080586083, + "mean_token_accuracy_utility": 0.6712332112332111, + "step": 1130 + }, + { + "epoch": 0.420020353409196, + "grad_norm": 0.82421875, + "learning_rate": 1.4387840838739875e-05, + "loss": 2.2651, + "mean_token_accuracy": 0.7585103785103786, + "mean_token_accuracy_utility": 0.6640476190476189, + "step": 1135 + }, + { + "epoch": 0.4218706633361088, + "grad_norm": 0.8671875, + "learning_rate": 1.4329686878193271e-05, + "loss": 2.2034, + "mean_token_accuracy": 0.758882783882784, + "mean_token_accuracy_utility": 0.6656255822246211, + "step": 1140 + }, + { + "epoch": 0.42372097326302155, + "grad_norm": 0.85546875, + "learning_rate": 1.4271352148289025e-05, + "loss": 2.2354, + "mean_token_accuracy": 0.7775274725274726, + "mean_token_accuracy_utility": 0.6527899877899876, + "step": 1145 + }, + { + "epoch": 0.4255712831899343, + "grad_norm": 0.8671875, + "learning_rate": 1.421283908456854e-05, + "loss": 2.2717, + "mean_token_accuracy": 0.7802442002442004, + "mean_token_accuracy_utility": 0.6443040293040291, + "step": 1150 + }, + { + "epoch": 0.4274215931168471, + "grad_norm": 0.85546875, + "learning_rate": 1.4154150130018867e-05, + "loss": 2.1793, + "mean_token_accuracy": 0.7755189255189257, + "mean_token_accuracy_utility": 0.6620512820512819, + "step": 1155 + }, + { + "epoch": 0.42927190304375984, + "grad_norm": 0.87890625, + "learning_rate": 1.4095287734970678e-05, + "loss": 2.2006, + "mean_token_accuracy": 0.7683211233211233, + "mean_token_accuracy_utility": 0.6614346764346763, + "step": 1160 + }, + { + "epoch": 0.4311222129706726, + "grad_norm": 0.87109375, + "learning_rate": 1.4036254356996004e-05, + "loss": 2.2816, + "mean_token_accuracy": 0.7593772893772893, + "mean_token_accuracy_utility": 0.6554761904761903, + "step": 1165 + }, + { + "epoch": 0.43297252289758537, + "grad_norm": 0.8671875, + "learning_rate": 1.3977052460805597e-05, + "loss": 2.2651, + "mean_token_accuracy": 0.7703479853479857, + "mean_token_accuracy_utility": 0.648089133089133, + "step": 1170 + }, + { + "epoch": 0.43482283282449813, + "grad_norm": 0.8828125, + "learning_rate": 1.3917684518146044e-05, + "loss": 2.1634, + "mean_token_accuracy": 0.7676495726495727, + "mean_token_accuracy_utility": 0.6684554334554333, + "step": 1175 + }, + { + "epoch": 0.43667314275141084, + "grad_norm": 0.8828125, + "learning_rate": 1.3858153007696552e-05, + "loss": 2.2111, + "mean_token_accuracy": 0.7652014652014651, + "mean_token_accuracy_utility": 0.6614774114774111, + "step": 1180 + }, + { + "epoch": 0.4385234526783236, + "grad_norm": 0.84375, + "learning_rate": 1.3798460414965475e-05, + "loss": 2.1385, + "mean_token_accuracy": 0.7784859584859586, + "mean_token_accuracy_utility": 0.6669291819291818, + "step": 1185 + }, + { + "epoch": 0.44037376260523636, + "grad_norm": 0.85546875, + "learning_rate": 1.3738609232186537e-05, + "loss": 2.2555, + "mean_token_accuracy": 0.7521184371184373, + "mean_token_accuracy_utility": 0.6597680097680096, + "step": 1190 + }, + { + "epoch": 0.4422240725321491, + "grad_norm": 0.8125, + "learning_rate": 1.3678601958214779e-05, + "loss": 2.1219, + "mean_token_accuracy": 0.7757814407814411, + "mean_token_accuracy_utility": 0.67021978021978, + "step": 1195 + }, + { + "epoch": 0.4440743824590619, + "grad_norm": 0.83984375, + "learning_rate": 1.3618441098422215e-05, + "loss": 2.1938, + "mean_token_accuracy": 0.7782234432234435, + "mean_token_accuracy_utility": 0.6550671550671548, + "step": 1200 + }, + { + "epoch": 0.4440743824590619, + "eval_mean_token_accuracy": 0.7536638742521092, + "eval_reasoning_loss": 0.8265872001647949, + "eval_reasoning_runtime": 27.7425, + "eval_reasoning_samples_per_second": 4.65, + "eval_reasoning_steps_per_second": 1.19, + "step": 1200 + }, + { + "epoch": 0.4440743824590619, + "eval_mean_token_accuracy": 0.6616615742151917, + "eval_utility_loss": 1.429228663444519, + "eval_utility_runtime": 10.4376, + "eval_utility_samples_per_second": 4.695, + "eval_utility_steps_per_second": 1.246, + "step": 1200 + }, + { + "epoch": 0.44592469238597465, + "grad_norm": 0.875, + "learning_rate": 1.3558129164593256e-05, + "loss": 2.2801, + "mean_token_accuracy": 0.7629487179487181, + "mean_token_accuracy_utility": 0.6505300133489432, + "step": 1205 + }, + { + "epoch": 0.4477750023128874, + "grad_norm": 0.82421875, + "learning_rate": 1.349766867481982e-05, + "loss": 2.1352, + "mean_token_accuracy": 0.7590659340659343, + "mean_token_accuracy_utility": 0.6805982905982905, + "step": 1210 + }, + { + "epoch": 0.4496253122398002, + "grad_norm": 0.8671875, + "learning_rate": 1.3437062153396201e-05, + "loss": 2.1621, + "mean_token_accuracy": 0.7549755799755801, + "mean_token_accuracy_utility": 0.6763492063492061, + "step": 1215 + }, + { + "epoch": 0.45147562216671294, + "grad_norm": 0.89453125, + "learning_rate": 1.337631213071369e-05, + "loss": 2.4311, + "mean_token_accuracy": 0.7546031746031746, + "mean_token_accuracy_utility": 0.6337545787545786, + "step": 1220 + }, + { + "epoch": 0.4533259320936257, + "grad_norm": 0.87890625, + "learning_rate": 1.331542114315491e-05, + "loss": 2.2266, + "mean_token_accuracy": 0.7701587301587302, + "mean_token_accuracy_utility": 0.6566281463311165, + "step": 1225 + }, + { + "epoch": 0.4551762420205384, + "grad_norm": 0.83984375, + "learning_rate": 1.325439173298793e-05, + "loss": 2.1955, + "mean_token_accuracy": 0.7792612942612946, + "mean_token_accuracy_utility": 0.6564590964590963, + "step": 1230 + }, + { + "epoch": 0.4570265519474512, + "grad_norm": 0.984375, + "learning_rate": 1.3193226448260128e-05, + "loss": 2.2666, + "mean_token_accuracy": 0.7604334554334555, + "mean_token_accuracy_utility": 0.6555094905094904, + "step": 1235 + }, + { + "epoch": 0.45887686187436394, + "grad_norm": 0.875, + "learning_rate": 1.3131927842691793e-05, + "loss": 2.2876, + "mean_token_accuracy": 0.7582905982905983, + "mean_token_accuracy_utility": 0.6488095238095236, + "step": 1240 + }, + { + "epoch": 0.4607271718012767, + "grad_norm": 0.859375, + "learning_rate": 1.3070498475569507e-05, + "loss": 2.2915, + "mean_token_accuracy": 0.7447069597069597, + "mean_token_accuracy_utility": 0.6600366300366298, + "step": 1245 + }, + { + "epoch": 0.46257748172818947, + "grad_norm": 0.921875, + "learning_rate": 1.3008940911639302e-05, + "loss": 2.2358, + "mean_token_accuracy": 0.7579609279609277, + "mean_token_accuracy_utility": 0.6722583247161217, + "step": 1250 + }, + { + "epoch": 0.46442779165510223, + "grad_norm": 0.84375, + "learning_rate": 1.2947257720999577e-05, + "loss": 2.354, + "mean_token_accuracy": 0.7688339438339439, + "mean_token_accuracy_utility": 0.6419352869352869, + "step": 1255 + }, + { + "epoch": 0.466278101582015, + "grad_norm": 0.89453125, + "learning_rate": 1.2885451478993777e-05, + "loss": 2.3551, + "mean_token_accuracy": 0.7534432234432235, + "mean_token_accuracy_utility": 0.6444505494505492, + "step": 1260 + }, + { + "epoch": 0.46812841150892776, + "grad_norm": 0.85546875, + "learning_rate": 1.282352476610289e-05, + "loss": 2.2567, + "mean_token_accuracy": 0.7637057387057387, + "mean_token_accuracy_utility": 0.6557020757020755, + "step": 1265 + }, + { + "epoch": 0.4699787214358405, + "grad_norm": 0.8125, + "learning_rate": 1.2761480167837705e-05, + "loss": 2.1314, + "mean_token_accuracy": 0.7788278388278389, + "mean_token_accuracy_utility": 0.6665750915750913, + "step": 1270 + }, + { + "epoch": 0.4718290313627533, + "grad_norm": 0.890625, + "learning_rate": 1.2699320274630847e-05, + "loss": 2.123, + "mean_token_accuracy": 0.770763125763126, + "mean_token_accuracy_utility": 0.6758424908424907, + "step": 1275 + }, + { + "epoch": 0.47367934128966604, + "grad_norm": 0.84765625, + "learning_rate": 1.263704768172864e-05, + "loss": 2.3406, + "mean_token_accuracy": 0.7496398046398047, + "mean_token_accuracy_utility": 0.6487057387057386, + "step": 1280 + }, + { + "epoch": 0.47552965121657875, + "grad_norm": 0.86328125, + "learning_rate": 1.257466498908276e-05, + "loss": 2.1233, + "mean_token_accuracy": 0.7696520146520147, + "mean_token_accuracy_utility": 0.6760989010989009, + "step": 1285 + }, + { + "epoch": 0.4773799611434915, + "grad_norm": 0.8984375, + "learning_rate": 1.2512174801241657e-05, + "loss": 2.1618, + "mean_token_accuracy": 0.778766788766789, + "mean_token_accuracy_utility": 0.6605921855921856, + "step": 1290 + }, + { + "epoch": 0.4792302710704043, + "grad_norm": 0.859375, + "learning_rate": 1.2449579727241834e-05, + "loss": 2.1865, + "mean_token_accuracy": 0.765805860805861, + "mean_token_accuracy_utility": 0.6651709401709399, + "step": 1295 + }, + { + "epoch": 0.48108058099731704, + "grad_norm": 0.89453125, + "learning_rate": 1.2386882380498918e-05, + "loss": 2.2356, + "mean_token_accuracy": 0.76485347985348, + "mean_token_accuracy_utility": 0.6589987789987789, + "step": 1300 + }, + { + "epoch": 0.48108058099731704, + "eval_mean_token_accuracy": 0.7541173423526363, + "eval_reasoning_loss": 0.825336217880249, + "eval_reasoning_runtime": 27.7404, + "eval_reasoning_samples_per_second": 4.65, + "eval_reasoning_steps_per_second": 1.19, + "step": 1300 + }, + { + "epoch": 0.48108058099731704, + "eval_mean_token_accuracy": 0.6615803428928964, + "eval_utility_loss": 1.4282413721084595, + "eval_utility_runtime": 10.4358, + "eval_utility_samples_per_second": 4.695, + "eval_utility_steps_per_second": 1.246, + "step": 1300 + }, + { + "epoch": 0.4829308909242298, + "grad_norm": 0.82421875, + "learning_rate": 1.2324085378698529e-05, + "loss": 2.1602, + "mean_token_accuracy": 0.7859645909645911, + "mean_token_accuracy_utility": 0.6591758241758241, + "step": 1305 + }, + { + "epoch": 0.48478120085114257, + "grad_norm": 0.87109375, + "learning_rate": 1.2261191343687e-05, + "loss": 2.1713, + "mean_token_accuracy": 0.765866910866911, + "mean_token_accuracy_utility": 0.667216117216117, + "step": 1310 + }, + { + "epoch": 0.48663151077805533, + "grad_norm": 0.85546875, + "learning_rate": 1.219820290136192e-05, + "loss": 2.2267, + "mean_token_accuracy": 0.7812515262515264, + "mean_token_accuracy_utility": 0.6549694749694749, + "step": 1315 + }, + { + "epoch": 0.4884818207049681, + "grad_norm": 0.85546875, + "learning_rate": 1.2135122681562481e-05, + "loss": 2.2381, + "mean_token_accuracy": 0.7616361416361417, + "mean_token_accuracy_utility": 0.661080586080586, + "step": 1320 + }, + { + "epoch": 0.49033213063188086, + "grad_norm": 0.8515625, + "learning_rate": 1.2071953317959692e-05, + "loss": 2.1668, + "mean_token_accuracy": 0.7698270066735724, + "mean_token_accuracy_utility": 0.6677899877899877, + "step": 1325 + }, + { + "epoch": 0.4921824405587936, + "grad_norm": 0.83984375, + "learning_rate": 1.2008697447946421e-05, + "loss": 2.1632, + "mean_token_accuracy": 0.767625152625153, + "mean_token_accuracy_utility": 0.673223443223443, + "step": 1330 + }, + { + "epoch": 0.49403275048570633, + "grad_norm": 0.8671875, + "learning_rate": 1.1945357712527273e-05, + "loss": 2.2506, + "mean_token_accuracy": 0.7661233211233214, + "mean_token_accuracy_utility": 0.6574786324786324, + "step": 1335 + }, + { + "epoch": 0.4958830604126191, + "grad_norm": 0.84765625, + "learning_rate": 1.1881936756208329e-05, + "loss": 2.1609, + "mean_token_accuracy": 0.768956043956044, + "mean_token_accuracy_utility": 0.6658669108669106, + "step": 1340 + }, + { + "epoch": 0.49773337033953186, + "grad_norm": 0.8828125, + "learning_rate": 1.1818437226886738e-05, + "loss": 2.2523, + "mean_token_accuracy": 0.7704822954822956, + "mean_token_accuracy_utility": 0.6520329670329668, + "step": 1345 + }, + { + "epoch": 0.4995836802664446, + "grad_norm": 0.859375, + "learning_rate": 1.1754861775740163e-05, + "loss": 2.1092, + "mean_token_accuracy": 0.7687667887667887, + "mean_token_accuracy_utility": 0.6789621489621489, + "step": 1350 + }, + { + "epoch": 0.5014339901933574, + "grad_norm": 0.84765625, + "learning_rate": 1.1691213057116082e-05, + "loss": 2.2274, + "mean_token_accuracy": 0.7741880341880344, + "mean_token_accuracy_utility": 0.6536874236874234, + "step": 1355 + }, + { + "epoch": 0.5032843001202701, + "grad_norm": 0.87109375, + "learning_rate": 1.1627493728420978e-05, + "loss": 2.168, + "mean_token_accuracy": 0.7680586080586081, + "mean_token_accuracy_utility": 0.6720233620732468, + "step": 1360 + }, + { + "epoch": 0.5051346100471829, + "grad_norm": 0.8359375, + "learning_rate": 1.1563706450009391e-05, + "loss": 2.2877, + "mean_token_accuracy": 0.7658974358974358, + "mean_token_accuracy_utility": 0.6508974358974358, + "step": 1365 + }, + { + "epoch": 0.5069849199740957, + "grad_norm": 0.84375, + "learning_rate": 1.1499853885072827e-05, + "loss": 2.1888, + "mean_token_accuracy": 0.7831562881562883, + "mean_token_accuracy_utility": 0.6529609279609278, + "step": 1370 + }, + { + "epoch": 0.5088352299010084, + "grad_norm": 0.86328125, + "learning_rate": 1.1435938699528586e-05, + "loss": 2.2777, + "mean_token_accuracy": 0.7497741147741148, + "mean_token_accuracy_utility": 0.6596459096459095, + "step": 1375 + }, + { + "epoch": 0.5106855398279212, + "grad_norm": 0.890625, + "learning_rate": 1.137196356190845e-05, + "loss": 2.3127, + "mean_token_accuracy": 0.7618498168498171, + "mean_token_accuracy_utility": 0.6485046748714568, + "step": 1380 + }, + { + "epoch": 0.5125358497548339, + "grad_norm": 0.875, + "learning_rate": 1.1307931143247268e-05, + "loss": 2.1445, + "mean_token_accuracy": 0.7647985347985349, + "mean_token_accuracy_utility": 0.674249084249084, + "step": 1385 + }, + { + "epoch": 0.5143861596817467, + "grad_norm": 0.83984375, + "learning_rate": 1.1243844116971433e-05, + "loss": 2.2309, + "mean_token_accuracy": 0.7685103785103786, + "mean_token_accuracy_utility": 0.6577533577533576, + "step": 1390 + }, + { + "epoch": 0.5162364696086594, + "grad_norm": 0.86328125, + "learning_rate": 1.1179705158787276e-05, + "loss": 2.1969, + "mean_token_accuracy": 0.7596703296703298, + "mean_token_accuracy_utility": 0.663191290293002, + "step": 1395 + }, + { + "epoch": 0.5180867795355723, + "grad_norm": 0.90234375, + "learning_rate": 1.1115516946569333e-05, + "loss": 2.2667, + "mean_token_accuracy": 0.7700122100122101, + "mean_token_accuracy_utility": 0.6534188034188032, + "step": 1400 + }, + { + "epoch": 0.5180867795355723, + "eval_mean_token_accuracy": 0.75409492468316, + "eval_reasoning_loss": 0.8243395090103149, + "eval_reasoning_runtime": 27.7485, + "eval_reasoning_samples_per_second": 4.649, + "eval_reasoning_steps_per_second": 1.189, + "step": 1400 + }, + { + "epoch": 0.5180867795355723, + "eval_mean_token_accuracy": 0.6617996657173093, + "eval_utility_loss": 1.4277063608169556, + "eval_utility_runtime": 10.4358, + "eval_utility_samples_per_second": 4.695, + "eval_utility_steps_per_second": 1.246, + "step": 1400 + }, + { + "epoch": 0.519937089462485, + "grad_norm": 0.87890625, + "learning_rate": 1.105128216024857e-05, + "loss": 2.1818, + "mean_token_accuracy": 0.7749511599511599, + "mean_token_accuracy_utility": 0.66669851836835, + "step": 1405 + }, + { + "epoch": 0.5217873993893978, + "grad_norm": 0.81640625, + "learning_rate": 1.0987003481700456e-05, + "loss": 2.1806, + "mean_token_accuracy": 0.7742307692307694, + "mean_token_accuracy_utility": 0.666263736263736, + "step": 1410 + }, + { + "epoch": 0.5236377093163105, + "grad_norm": 0.88671875, + "learning_rate": 1.092268359463302e-05, + "loss": 2.18, + "mean_token_accuracy": 0.7779026256585866, + "mean_token_accuracy_utility": 0.6604945054945053, + "step": 1415 + }, + { + "epoch": 0.5254880192432232, + "grad_norm": 0.8828125, + "learning_rate": 1.0858325184474796e-05, + "loss": 2.203, + "mean_token_accuracy": 0.7699267399267399, + "mean_token_accuracy_utility": 0.6616849816849816, + "step": 1420 + }, + { + "epoch": 0.527338329170136, + "grad_norm": 0.86328125, + "learning_rate": 1.0793930938262689e-05, + "loss": 2.2736, + "mean_token_accuracy": 0.7585958485958486, + "mean_token_accuracy_utility": 0.6553113553113552, + "step": 1425 + }, + { + "epoch": 0.5291886390970487, + "grad_norm": 0.86328125, + "learning_rate": 1.0729503544529814e-05, + "loss": 2.235, + "mean_token_accuracy": 0.7692979242979245, + "mean_token_accuracy_utility": 0.6564224664224663, + "step": 1430 + }, + { + "epoch": 0.5310389490239615, + "grad_norm": 0.86328125, + "learning_rate": 1.0665045693193226e-05, + "loss": 2.2728, + "mean_token_accuracy": 0.7679731379731379, + "mean_token_accuracy_utility": 0.6506654456654455, + "step": 1435 + }, + { + "epoch": 0.5328892589508742, + "grad_norm": 0.7890625, + "learning_rate": 1.0600560075441617e-05, + "loss": 2.2008, + "mean_token_accuracy": 0.780775335775336, + "mean_token_accuracy_utility": 0.6557081807081804, + "step": 1440 + }, + { + "epoch": 0.5347395688777871, + "grad_norm": 0.83984375, + "learning_rate": 1.0536049383622966e-05, + "loss": 2.2148, + "mean_token_accuracy": 0.7574786324786326, + "mean_token_accuracy_utility": 0.6677228327228326, + "step": 1445 + }, + { + "epoch": 0.5365898788046998, + "grad_norm": 0.8359375, + "learning_rate": 1.047151631113212e-05, + "loss": 2.1447, + "mean_token_accuracy": 0.7728998778998781, + "mean_token_accuracy_utility": 0.6679853479853477, + "step": 1450 + }, + { + "epoch": 0.5384401887316126, + "grad_norm": 0.89453125, + "learning_rate": 1.0406963552298332e-05, + "loss": 2.1575, + "mean_token_accuracy": 0.7813858363858366, + "mean_token_accuracy_utility": 0.6616788766788765, + "step": 1455 + }, + { + "epoch": 0.5402904986585253, + "grad_norm": 0.88671875, + "learning_rate": 1.034239380227281e-05, + "loss": 2.1748, + "mean_token_accuracy": 0.7783882783882786, + "mean_token_accuracy_utility": 0.663238043227901, + "step": 1460 + }, + { + "epoch": 0.542140808585438, + "grad_norm": 0.86328125, + "learning_rate": 1.0277809756916134e-05, + "loss": 2.3, + "mean_token_accuracy": 0.7587057387057385, + "mean_token_accuracy_utility": 0.6523076923076921, + "step": 1465 + }, + { + "epoch": 0.5439911185123508, + "grad_norm": 0.859375, + "learning_rate": 1.0213214112685747e-05, + "loss": 2.2058, + "mean_token_accuracy": 0.7712515262515265, + "mean_token_accuracy_utility": 0.66020757020757, + "step": 1470 + }, + { + "epoch": 0.5458414284392635, + "grad_norm": 0.83984375, + "learning_rate": 1.0148609566523358e-05, + "loss": 2.1973, + "mean_token_accuracy": 0.7618559218559221, + "mean_token_accuracy_utility": 0.6665934065934065, + "step": 1475 + }, + { + "epoch": 0.5476917383661764, + "grad_norm": 0.86328125, + "learning_rate": 1.0083998815742335e-05, + "loss": 2.1435, + "mean_token_accuracy": 0.7617643467643467, + "mean_token_accuracy_utility": 0.6813125763125761, + "step": 1480 + }, + { + "epoch": 0.5495420482930891, + "grad_norm": 0.859375, + "learning_rate": 1.0019384557915099e-05, + "loss": 2.1916, + "mean_token_accuracy": 0.7664652014652015, + "mean_token_accuracy_utility": 0.6625152625152623, + "step": 1485 + }, + { + "epoch": 0.5513923582200019, + "grad_norm": 0.82421875, + "learning_rate": 9.9547694907605e-06, + "loss": 2.1605, + "mean_token_accuracy": 0.7672222222222222, + "mean_token_accuracy_utility": 0.6675274725274724, + "step": 1490 + }, + { + "epoch": 0.5532426681469146, + "grad_norm": 0.85546875, + "learning_rate": 9.890156312031165e-06, + "loss": 2.2461, + "mean_token_accuracy": 0.7711599511599514, + "mean_token_accuracy_utility": 0.6608032563382358, + "step": 1495 + }, + { + "epoch": 0.5550929780738274, + "grad_norm": 0.87109375, + "learning_rate": 9.825547719400889e-06, + "loss": 2.2284, + "mean_token_accuracy": 0.7737789987789988, + "mean_token_accuracy_utility": 0.6507264957264955, + "step": 1500 + }, + { + "epoch": 0.5550929780738274, + "eval_mean_token_accuracy": 0.7540486746369099, + "eval_reasoning_loss": 0.8237046003341675, + "eval_reasoning_runtime": 27.7951, + "eval_reasoning_samples_per_second": 4.641, + "eval_reasoning_steps_per_second": 1.187, + "step": 1500 + }, + { + "epoch": 0.5550929780738274, + "eval_mean_token_accuracy": 0.6617808810831401, + "eval_utility_loss": 1.4274048805236816, + "eval_utility_runtime": 10.4543, + "eval_utility_samples_per_second": 4.687, + "eval_utility_steps_per_second": 1.244, + "step": 1500 + }, + { + "epoch": 0.5569432880007401, + "grad_norm": 0.82421875, + "learning_rate": 9.760946410351988e-06, + "loss": 2.2928, + "mean_token_accuracy": 0.7558974358974359, + "mean_token_accuracy_utility": 0.6524420024420022, + "step": 1505 + }, + { + "epoch": 0.5587935979276529, + "grad_norm": 0.9453125, + "learning_rate": 9.696355082062679e-06, + "loss": 2.3201, + "mean_token_accuracy": 0.7626373626373627, + "mean_token_accuracy_utility": 0.6397741147741146, + "step": 1510 + }, + { + "epoch": 0.5606439078545656, + "grad_norm": 0.80859375, + "learning_rate": 9.631776431294475e-06, + "loss": 2.2493, + "mean_token_accuracy": 0.7673382173382175, + "mean_token_accuracy_utility": 0.6582234432234431, + "step": 1515 + }, + { + "epoch": 0.5624942177814783, + "grad_norm": 0.8515625, + "learning_rate": 9.567213154279582e-06, + "loss": 2.1607, + "mean_token_accuracy": 0.7754945054945057, + "mean_token_accuracy_utility": 0.6627411477411476, + "step": 1520 + }, + { + "epoch": 0.5643445277083912, + "grad_norm": 0.8671875, + "learning_rate": 9.502667946608332e-06, + "loss": 2.1579, + "mean_token_accuracy": 0.7675091575091576, + "mean_token_accuracy_utility": 0.6729914529914527, + "step": 1525 + }, + { + "epoch": 0.5661948376353039, + "grad_norm": 0.89453125, + "learning_rate": 9.43814350311666e-06, + "loss": 2.1831, + "mean_token_accuracy": 0.7651465201465203, + "mean_token_accuracy_utility": 0.6702672776707511, + "step": 1530 + }, + { + "epoch": 0.5680451475622167, + "grad_norm": 0.859375, + "learning_rate": 9.37364251777355e-06, + "loss": 2.2366, + "mean_token_accuracy": 0.7546275946275948, + "mean_token_accuracy_utility": 0.6634493284493281, + "step": 1535 + }, + { + "epoch": 0.5698954574891294, + "grad_norm": 0.87890625, + "learning_rate": 9.309167683568597e-06, + "loss": 2.326, + "mean_token_accuracy": 0.7610073260073262, + "mean_token_accuracy_utility": 0.6436752136752134, + "step": 1540 + }, + { + "epoch": 0.5717457674160422, + "grad_norm": 0.796875, + "learning_rate": 9.244721692399545e-06, + "loss": 2.141, + "mean_token_accuracy": 0.7757509157509157, + "mean_token_accuracy_utility": 0.6676068376068374, + "step": 1545 + }, + { + "epoch": 0.5735960773429549, + "grad_norm": 0.8359375, + "learning_rate": 9.180307234959918e-06, + "loss": 2.1239, + "mean_token_accuracy": 0.7657020757020757, + "mean_token_accuracy_utility": 0.6743711843711842, + "step": 1550 + }, + { + "epoch": 0.5754463872698677, + "grad_norm": 0.84765625, + "learning_rate": 9.115927000626665e-06, + "loss": 2.1728, + "mean_token_accuracy": 0.7645177045177045, + "mean_token_accuracy_utility": 0.6652869352869353, + "step": 1555 + }, + { + "epoch": 0.5772966971967805, + "grad_norm": 0.86328125, + "learning_rate": 9.051583677347879e-06, + "loss": 2.2419, + "mean_token_accuracy": 0.7645604395604397, + "mean_token_accuracy_utility": 0.6531135531135531, + "step": 1560 + }, + { + "epoch": 0.5791470071236933, + "grad_norm": 0.87890625, + "learning_rate": 8.987279951530586e-06, + "loss": 2.1526, + "mean_token_accuracy": 0.76489010989011, + "mean_token_accuracy_utility": 0.6717094017094015, + "step": 1565 + }, + { + "epoch": 0.580997317050606, + "grad_norm": 0.8359375, + "learning_rate": 8.923018507928564e-06, + "loss": 2.2959, + "mean_token_accuracy": 0.7592429792429793, + "mean_token_accuracy_utility": 0.6522710622710621, + "step": 1570 + }, + { + "epoch": 0.5828476269775187, + "grad_norm": 0.83984375, + "learning_rate": 8.85880202953026e-06, + "loss": 2.1487, + "mean_token_accuracy": 0.7690354090354093, + "mean_token_accuracy_utility": 0.6723992673992674, + "step": 1575 + }, + { + "epoch": 0.5846979369044315, + "grad_norm": 0.8671875, + "learning_rate": 8.79463319744677e-06, + "loss": 2.2499, + "mean_token_accuracy": 0.7518009768009769, + "mean_token_accuracy_utility": 0.6655494505494505, + "step": 1580 + }, + { + "epoch": 0.5865482468313442, + "grad_norm": 0.85546875, + "learning_rate": 8.730514690799916e-06, + "loss": 2.1463, + "mean_token_accuracy": 0.7706593406593409, + "mean_token_accuracy_utility": 0.6696031746031746, + "step": 1585 + }, + { + "epoch": 0.588398556758257, + "grad_norm": 0.89453125, + "learning_rate": 8.666449186610353e-06, + "loss": 2.1927, + "mean_token_accuracy": 0.7747496947496949, + "mean_token_accuracy_utility": 0.6626739926739925, + "step": 1590 + }, + { + "epoch": 0.5902488666851697, + "grad_norm": 0.81640625, + "learning_rate": 8.60243935968585e-06, + "loss": 2.1543, + "mean_token_accuracy": 0.7713003663003665, + "mean_token_accuracy_utility": 0.6692979242979241, + "step": 1595 + }, + { + "epoch": 0.5920991766120826, + "grad_norm": 0.86328125, + "learning_rate": 8.538487882509568e-06, + "loss": 2.1652, + "mean_token_accuracy": 0.769029304029304, + "mean_token_accuracy_utility": 0.6703418803418801, + "step": 1600 + }, + { + "epoch": 0.5920991766120826, + "eval_mean_token_accuracy": 0.754452192687487, + "eval_reasoning_loss": 0.8232352137565613, + "eval_reasoning_runtime": 27.8055, + "eval_reasoning_samples_per_second": 4.639, + "eval_reasoning_steps_per_second": 1.187, + "step": 1600 + }, + { + "epoch": 0.5920991766120826, + "eval_mean_token_accuracy": 0.6619560355762814, + "eval_utility_loss": 1.4270882606506348, + "eval_utility_runtime": 10.4537, + "eval_utility_samples_per_second": 4.687, + "eval_utility_steps_per_second": 1.244, + "step": 1600 + }, + { + "epoch": 0.5939494865389953, + "grad_norm": 0.8984375, + "learning_rate": 8.474597425128501e-06, + "loss": 2.2368, + "mean_token_accuracy": 0.772967032967033, + "mean_token_accuracy_utility": 0.6529670329670327, + "step": 1605 + }, + { + "epoch": 0.5957997964659081, + "grad_norm": 0.86328125, + "learning_rate": 8.410770655042003e-06, + "loss": 2.1383, + "mean_token_accuracy": 0.7688888888888891, + "mean_token_accuracy_utility": 0.6741147741147742, + "step": 1610 + }, + { + "epoch": 0.5976501063928208, + "grad_norm": 0.8359375, + "learning_rate": 8.347010237090408e-06, + "loss": 2.2546, + "mean_token_accuracy": 0.762863247863248, + "mean_token_accuracy_utility": 0.6569317140919653, + "step": 1615 + }, + { + "epoch": 0.5995004163197336, + "grad_norm": 0.87890625, + "learning_rate": 8.283318833343773e-06, + "loss": 2.1978, + "mean_token_accuracy": 0.7574114774114775, + "mean_token_accuracy_utility": 0.6710195360195359, + "step": 1620 + }, + { + "epoch": 0.6013507262466463, + "grad_norm": 0.90234375, + "learning_rate": 8.219699102990735e-06, + "loss": 2.2578, + "mean_token_accuracy": 0.761434129513637, + "mean_token_accuracy_utility": 0.6586691086691085, + "step": 1625 + }, + { + "epoch": 0.603201036173559, + "grad_norm": 0.875, + "learning_rate": 8.156153702227484e-06, + "loss": 2.1725, + "mean_token_accuracy": 0.7697069597069597, + "mean_token_accuracy_utility": 0.67003663003663, + "step": 1630 + }, + { + "epoch": 0.6050513461004718, + "grad_norm": 0.90625, + "learning_rate": 8.092685284146865e-06, + "loss": 2.2529, + "mean_token_accuracy": 0.7492490842490842, + "mean_token_accuracy_utility": 0.6644331842407876, + "step": 1635 + }, + { + "epoch": 0.6069016560273846, + "grad_norm": 0.875, + "learning_rate": 8.029296498627608e-06, + "loss": 2.2797, + "mean_token_accuracy": 0.7646153846153847, + "mean_token_accuracy_utility": 0.653982606526188, + "step": 1640 + }, + { + "epoch": 0.6087519659542974, + "grad_norm": 0.81640625, + "learning_rate": 7.965989992223693e-06, + "loss": 2.1285, + "mean_token_accuracy": 0.7712881562881565, + "mean_token_accuracy_utility": 0.6770940170940168, + "step": 1645 + }, + { + "epoch": 0.6106022758812101, + "grad_norm": 0.90625, + "learning_rate": 7.90276840805385e-06, + "loss": 2.2868, + "mean_token_accuracy": 0.7585653235653238, + "mean_token_accuracy_utility": 0.6547987142453555, + "step": 1650 + }, + { + "epoch": 0.6124525858081229, + "grad_norm": 0.85546875, + "learning_rate": 7.839634385691214e-06, + "loss": 2.1757, + "mean_token_accuracy": 0.7680769230769233, + "mean_token_accuracy_utility": 0.6672100122100122, + "step": 1655 + }, + { + "epoch": 0.6143028957350356, + "grad_norm": 0.8359375, + "learning_rate": 7.776590561053117e-06, + "loss": 2.1443, + "mean_token_accuracy": 0.7611294261294261, + "mean_token_accuracy_utility": 0.679304029304029, + "step": 1660 + }, + { + "epoch": 0.6161532056619484, + "grad_norm": 0.84765625, + "learning_rate": 7.713639566291028e-06, + "loss": 2.2964, + "mean_token_accuracy": 0.75001221001221, + "mean_token_accuracy_utility": 0.6550732600732599, + "step": 1665 + }, + { + "epoch": 0.6180035155888611, + "grad_norm": 0.8671875, + "learning_rate": 7.650784029680662e-06, + "loss": 2.1703, + "mean_token_accuracy": 0.7833760683760685, + "mean_token_accuracy_utility": 0.659175824175824, + "step": 1670 + }, + { + "epoch": 0.6198538255157738, + "grad_norm": 0.84765625, + "learning_rate": 7.58802657551225e-06, + "loss": 2.1694, + "mean_token_accuracy": 0.7774297924297928, + "mean_token_accuracy_utility": 0.6607264957264956, + "step": 1675 + }, + { + "epoch": 0.6217041354426867, + "grad_norm": 0.86328125, + "learning_rate": 7.52536982398097e-06, + "loss": 2.2122, + "mean_token_accuracy": 0.7634615384615384, + "mean_token_accuracy_utility": 0.6655372405372404, + "step": 1680 + }, + { + "epoch": 0.6235544453695994, + "grad_norm": 0.87890625, + "learning_rate": 7.46281639107755e-06, + "loss": 2.3831, + "mean_token_accuracy": 0.7460321006130588, + "mean_token_accuracy_utility": 0.6466483516483514, + "step": 1685 + }, + { + "epoch": 0.6254047552965122, + "grad_norm": 0.84375, + "learning_rate": 7.400368888479048e-06, + "loss": 2.1837, + "mean_token_accuracy": 0.7547252747252748, + "mean_token_accuracy_utility": 0.6766910866910865, + "step": 1690 + }, + { + "epoch": 0.6272550652234249, + "grad_norm": 0.8984375, + "learning_rate": 7.3380299234398076e-06, + "loss": 2.3235, + "mean_token_accuracy": 0.757142857142857, + "mean_token_accuracy_utility": 0.6490720390720389, + "step": 1695 + }, + { + "epoch": 0.6291053751503377, + "grad_norm": 0.859375, + "learning_rate": 7.275802098682612e-06, + "loss": 2.2404, + "mean_token_accuracy": 0.7529059829059828, + "mean_token_accuracy_utility": 0.6660012210012207, + "step": 1700 + }, + { + "epoch": 0.6291053751503377, + "eval_mean_token_accuracy": 0.7542780748663105, + "eval_reasoning_loss": 0.822817325592041, + "eval_reasoning_runtime": 27.7974, + "eval_reasoning_samples_per_second": 4.641, + "eval_reasoning_steps_per_second": 1.187, + "step": 1700 + }, + { + "epoch": 0.6291053751503377, + "eval_mean_token_accuracy": 0.6619123735223249, + "eval_utility_loss": 1.4267956018447876, + "eval_utility_runtime": 10.4595, + "eval_utility_samples_per_second": 4.685, + "eval_utility_steps_per_second": 1.243, + "step": 1700 + }, + { + "epoch": 0.6309556850772504, + "grad_norm": 0.84375, + "learning_rate": 7.213688012290004e-06, + "loss": 2.1367, + "mean_token_accuracy": 0.7613736263736265, + "mean_token_accuracy_utility": 0.6782295482295481, + "step": 1705 + }, + { + "epoch": 0.6328059950041632, + "grad_norm": 0.85546875, + "learning_rate": 7.151690257595826e-06, + "loss": 2.1551, + "mean_token_accuracy": 0.7687667887667888, + "mean_token_accuracy_utility": 0.6706105006105004, + "step": 1710 + }, + { + "epoch": 0.6346563049310759, + "grad_norm": 0.83984375, + "learning_rate": 7.089811423076936e-06, + "loss": 2.1785, + "mean_token_accuracy": 0.7852014652014654, + "mean_token_accuracy_utility": 0.6578571428571427, + "step": 1715 + }, + { + "epoch": 0.6365066148579888, + "grad_norm": 0.8515625, + "learning_rate": 7.028054092245134e-06, + "loss": 2.2202, + "mean_token_accuracy": 0.7588583638583639, + "mean_token_accuracy_utility": 0.6629914529914528, + "step": 1720 + }, + { + "epoch": 0.6383569247849015, + "grad_norm": 0.83984375, + "learning_rate": 6.966420843539321e-06, + "loss": 2.1464, + "mean_token_accuracy": 0.7560134310134308, + "mean_token_accuracy_utility": 0.6792735042735041, + "step": 1725 + }, + { + "epoch": 0.6402072347118142, + "grad_norm": 0.8046875, + "learning_rate": 6.90491425021781e-06, + "loss": 2.225, + "mean_token_accuracy": 0.7596581196581198, + "mean_token_accuracy_utility": 0.6674664224664223, + "step": 1730 + }, + { + "epoch": 0.642057544638727, + "grad_norm": 0.875, + "learning_rate": 6.843536880250914e-06, + "loss": 2.1986, + "mean_token_accuracy": 0.7666239316239318, + "mean_token_accuracy_utility": 0.6649694749694748, + "step": 1735 + }, + { + "epoch": 0.6439078545656397, + "grad_norm": 0.90625, + "learning_rate": 6.7822912962137225e-06, + "loss": 2.1464, + "mean_token_accuracy": 0.7611355311355312, + "mean_token_accuracy_utility": 0.6720207570207568, + "step": 1740 + }, + { + "epoch": 0.6457581644925525, + "grad_norm": 0.84765625, + "learning_rate": 6.721180055179113e-06, + "loss": 2.1908, + "mean_token_accuracy": 0.7637734010627106, + "mean_token_accuracy_utility": 0.6713919413919411, + "step": 1745 + }, + { + "epoch": 0.6476084744194652, + "grad_norm": 0.91015625, + "learning_rate": 6.660205708610987e-06, + "loss": 2.1627, + "mean_token_accuracy": 0.755903540903541, + "mean_token_accuracy_utility": 0.6797557997557996, + "step": 1750 + }, + { + "epoch": 0.649458784346378, + "grad_norm": 0.85546875, + "learning_rate": 6.599370802257755e-06, + "loss": 2.1479, + "mean_token_accuracy": 0.7669841269841271, + "mean_token_accuracy_utility": 0.6741575091575089, + "step": 1755 + }, + { + "epoch": 0.6513090942732908, + "grad_norm": 0.87109375, + "learning_rate": 6.5386778760460316e-06, + "loss": 2.2475, + "mean_token_accuracy": 0.776764346764347, + "mean_token_accuracy_utility": 0.6529242979242978, + "step": 1760 + }, + { + "epoch": 0.6531594042002036, + "grad_norm": 0.84375, + "learning_rate": 6.478129463974598e-06, + "loss": 2.2927, + "mean_token_accuracy": 0.7552319902319902, + "mean_token_accuracy_utility": 0.6555250305250303, + "step": 1765 + }, + { + "epoch": 0.6550097141271163, + "grad_norm": 0.83203125, + "learning_rate": 6.417728094008613e-06, + "loss": 2.1218, + "mean_token_accuracy": 0.7752564102564103, + "mean_token_accuracy_utility": 0.6716361416361414, + "step": 1770 + }, + { + "epoch": 0.6568600240540291, + "grad_norm": 0.8125, + "learning_rate": 6.357476287974051e-06, + "loss": 2.1287, + "mean_token_accuracy": 0.7644139194139197, + "mean_token_accuracy_utility": 0.675177045177045, + "step": 1775 + }, + { + "epoch": 0.6587103339809418, + "grad_norm": 0.84765625, + "learning_rate": 6.297376561452428e-06, + "loss": 2.1647, + "mean_token_accuracy": 0.7733150183150184, + "mean_token_accuracy_utility": 0.6666499846746758, + "step": 1780 + }, + { + "epoch": 0.6605606439078545, + "grad_norm": 0.83203125, + "learning_rate": 6.237431423675764e-06, + "loss": 2.2321, + "mean_token_accuracy": 0.7657326007326011, + "mean_token_accuracy_utility": 0.6613919413919411, + "step": 1785 + }, + { + "epoch": 0.6624109538347673, + "grad_norm": 0.859375, + "learning_rate": 6.177643377421827e-06, + "loss": 2.2235, + "mean_token_accuracy": 0.7686874236874238, + "mean_token_accuracy_utility": 0.657185592185592, + "step": 1790 + }, + { + "epoch": 0.66426126376168, + "grad_norm": 0.859375, + "learning_rate": 6.118014918909633e-06, + "loss": 2.2113, + "mean_token_accuracy": 0.7624786324786326, + "mean_token_accuracy_utility": 0.6613204204007241, + "step": 1795 + }, + { + "epoch": 0.6661115736885929, + "grad_norm": 0.87890625, + "learning_rate": 6.058548537695225e-06, + "loss": 2.2142, + "mean_token_accuracy": 0.77998778998779, + "mean_token_accuracy_utility": 0.6561599511599511, + "step": 1800 + }, + { + "epoch": 0.6661115736885929, + "eval_mean_token_accuracy": 0.7545317427670372, + "eval_reasoning_loss": 0.8226430416107178, + "eval_reasoning_runtime": 27.8047, + "eval_reasoning_samples_per_second": 4.639, + "eval_reasoning_steps_per_second": 1.187, + "step": 1800 + }, + { + "epoch": 0.6661115736885929, + "eval_mean_token_accuracy": 0.6621237006567289, + "eval_utility_loss": 1.4266791343688965, + "eval_utility_runtime": 10.4584, + "eval_utility_samples_per_second": 4.685, + "eval_utility_steps_per_second": 1.243, + "step": 1800 + }, + { + "epoch": 0.6679618836155056, + "grad_norm": 0.84375, + "learning_rate": 5.999246716567737e-06, + "loss": 2.1596, + "mean_token_accuracy": 0.7563492063492065, + "mean_token_accuracy_utility": 0.6754456654456652, + "step": 1805 + }, + { + "epoch": 0.6698121935424184, + "grad_norm": 0.84765625, + "learning_rate": 5.940111931445731e-06, + "loss": 2.3029, + "mean_token_accuracy": 0.7472466422466423, + "mean_token_accuracy_utility": 0.6607081807081805, + "step": 1810 + }, + { + "epoch": 0.6716625034693311, + "grad_norm": 0.8046875, + "learning_rate": 5.881146651273825e-06, + "loss": 2.123, + "mean_token_accuracy": 0.7790109890109893, + "mean_token_accuracy_utility": 0.6680036630036628, + "step": 1815 + }, + { + "epoch": 0.6735128133962439, + "grad_norm": 0.83984375, + "learning_rate": 5.822353337919616e-06, + "loss": 2.1542, + "mean_token_accuracy": 0.768937728937729, + "mean_token_accuracy_utility": 0.6639987789987788, + "step": 1820 + }, + { + "epoch": 0.6753631233231566, + "grad_norm": 0.8203125, + "learning_rate": 5.763734446070892e-06, + "loss": 2.076, + "mean_token_accuracy": 0.7813614163614165, + "mean_token_accuracy_utility": 0.6735103785103784, + "step": 1825 + }, + { + "epoch": 0.6772134332500694, + "grad_norm": 0.85546875, + "learning_rate": 5.705292423133133e-06, + "loss": 2.1989, + "mean_token_accuracy": 0.7816544566544568, + "mean_token_accuracy_utility": 0.6533211233211231, + "step": 1830 + }, + { + "epoch": 0.6790637431769821, + "grad_norm": 0.89453125, + "learning_rate": 5.647029709127355e-06, + "loss": 2.2723, + "mean_token_accuracy": 0.7595360195360197, + "mean_token_accuracy_utility": 0.6523626373626372, + "step": 1835 + }, + { + "epoch": 0.6809140531038949, + "grad_norm": 0.8359375, + "learning_rate": 5.5889487365882065e-06, + "loss": 2.165, + "mean_token_accuracy": 0.7708363858363858, + "mean_token_accuracy_utility": 0.6658730158730157, + "step": 1840 + }, + { + "epoch": 0.6827643630308077, + "grad_norm": 0.84375, + "learning_rate": 5.531051930462437e-06, + "loss": 2.1931, + "mean_token_accuracy": 0.7717277167277169, + "mean_token_accuracy_utility": 0.6633760683760682, + "step": 1845 + }, + { + "epoch": 0.6846146729577204, + "grad_norm": 0.8671875, + "learning_rate": 5.4733417080076325e-06, + "loss": 2.1951, + "mean_token_accuracy": 0.7735775335775339, + "mean_token_accuracy_utility": 0.6584065934065932, + "step": 1850 + }, + { + "epoch": 0.6864649828846332, + "grad_norm": 0.88671875, + "learning_rate": 5.415820478691301e-06, + "loss": 2.1198, + "mean_token_accuracy": 0.7727472527472529, + "mean_token_accuracy_utility": 0.6729609279609278, + "step": 1855 + }, + { + "epoch": 0.6883152928115459, + "grad_norm": 0.84765625, + "learning_rate": 5.358490644090263e-06, + "loss": 2.2164, + "mean_token_accuracy": 0.760689865689866, + "mean_token_accuracy_utility": 0.6621245421245421, + "step": 1860 + }, + { + "epoch": 0.6901656027384587, + "grad_norm": 0.86328125, + "learning_rate": 5.3013545977904005e-06, + "loss": 2.3052, + "mean_token_accuracy": 0.7536813186813187, + "mean_token_accuracy_utility": 0.6498473748473746, + "step": 1865 + }, + { + "epoch": 0.6920159126653714, + "grad_norm": 0.8515625, + "learning_rate": 5.244414725286717e-06, + "loss": 2.2745, + "mean_token_accuracy": 0.7553357753357755, + "mean_token_accuracy_utility": 0.6606532356532355, + "step": 1870 + }, + { + "epoch": 0.6938662225922843, + "grad_norm": 0.83203125, + "learning_rate": 5.187673403883721e-06, + "loss": 2.1815, + "mean_token_accuracy": 0.7764468864468868, + "mean_token_accuracy_utility": 0.6597008547008545, + "step": 1875 + }, + { + "epoch": 0.695716532519197, + "grad_norm": 0.83984375, + "learning_rate": 5.131133002596199e-06, + "loss": 2.194, + "mean_token_accuracy": 0.7810439560439562, + "mean_token_accuracy_utility": 0.6528977976061627, + "step": 1880 + }, + { + "epoch": 0.6975668424461097, + "grad_norm": 0.85546875, + "learning_rate": 5.074795882050293e-06, + "loss": 2.1825, + "mean_token_accuracy": 0.7623260073260074, + "mean_token_accuracy_utility": 0.6715995115995115, + "step": 1885 + }, + { + "epoch": 0.6994171523730225, + "grad_norm": 0.86328125, + "learning_rate": 5.018664394384942e-06, + "loss": 2.2523, + "mean_token_accuracy": 0.7756837606837608, + "mean_token_accuracy_utility": 0.6493711843711842, + "step": 1890 + }, + { + "epoch": 0.7012674622999352, + "grad_norm": 0.8515625, + "learning_rate": 4.9627408831536705e-06, + "loss": 2.2182, + "mean_token_accuracy": 0.7690048840048844, + "mean_token_accuracy_utility": 0.6605189255189254, + "step": 1895 + }, + { + "epoch": 0.703117772226848, + "grad_norm": 0.90625, + "learning_rate": 4.907027683226761e-06, + "loss": 2.1435, + "mean_token_accuracy": 0.7735133429106265, + "mean_token_accuracy_utility": 0.6691086691086691, + "step": 1900 + }, + { + "epoch": 0.703117772226848, + "eval_mean_token_accuracy": 0.7543409749292104, + "eval_reasoning_loss": 0.8224542140960693, + "eval_reasoning_runtime": 27.8043, + "eval_reasoning_samples_per_second": 4.64, + "eval_reasoning_steps_per_second": 1.187, + "step": 1900 + }, + { + "epoch": 0.703117772226848, + "eval_mean_token_accuracy": 0.662034473644425, + "eval_utility_loss": 1.426527976989746, + "eval_utility_runtime": 10.4638, + "eval_utility_samples_per_second": 4.683, + "eval_utility_steps_per_second": 1.242, + "step": 1900 + }, + { + "epoch": 0.7049680821537607, + "grad_norm": 0.81640625, + "learning_rate": 4.85152712069375e-06, + "loss": 2.2042, + "mean_token_accuracy": 0.7674786324786329, + "mean_token_accuracy_utility": 0.6624786324786323, + "step": 1905 + }, + { + "epoch": 0.7068183920806735, + "grad_norm": 0.87890625, + "learning_rate": 4.7962415127663265e-06, + "loss": 2.2396, + "mean_token_accuracy": 0.7704517704517707, + "mean_token_accuracy_utility": 0.6564407814407812, + "step": 1910 + }, + { + "epoch": 0.7086687020075862, + "grad_norm": 0.9453125, + "learning_rate": 4.74117316768158e-06, + "loss": 2.248, + "mean_token_accuracy": 0.7622161172161172, + "mean_token_accuracy_utility": 0.6607936507936507, + "step": 1915 + }, + { + "epoch": 0.7105190119344991, + "grad_norm": 0.83203125, + "learning_rate": 4.686324384605629e-06, + "loss": 2.2125, + "mean_token_accuracy": 0.7616239316239317, + "mean_token_accuracy_utility": 0.6656288156288154, + "step": 1920 + }, + { + "epoch": 0.7123693218614118, + "grad_norm": 0.890625, + "learning_rate": 4.631697453537623e-06, + "loss": 2.2391, + "mean_token_accuracy": 0.7648412698412701, + "mean_token_accuracy_utility": 0.6597619047619047, + "step": 1925 + }, + { + "epoch": 0.7142196317883246, + "grad_norm": 0.83203125, + "learning_rate": 4.577294655214144e-06, + "loss": 2.1911, + "mean_token_accuracy": 0.7613431013431015, + "mean_token_accuracy_utility": 0.6682242231828627, + "step": 1930 + }, + { + "epoch": 0.7160699417152373, + "grad_norm": 0.84375, + "learning_rate": 4.523118261013969e-06, + "loss": 2.2407, + "mean_token_accuracy": 0.7605372405372405, + "mean_token_accuracy_utility": 0.6632297333997701, + "step": 1935 + }, + { + "epoch": 0.71792025164215, + "grad_norm": 0.796875, + "learning_rate": 4.469170532863254e-06, + "loss": 2.2483, + "mean_token_accuracy": 0.7506654456654458, + "mean_token_accuracy_utility": 0.6673321123321122, + "step": 1940 + }, + { + "epoch": 0.7197705615690628, + "grad_norm": 0.84375, + "learning_rate": 4.415453723141081e-06, + "loss": 2.3412, + "mean_token_accuracy": 0.7561050061050062, + "mean_token_accuracy_utility": 0.6527472527472526, + "step": 1945 + }, + { + "epoch": 0.7216208714959755, + "grad_norm": 0.8125, + "learning_rate": 4.361970074585426e-06, + "loss": 2.0709, + "mean_token_accuracy": 0.7752319902319905, + "mean_token_accuracy_utility": 0.6849511599511596, + "step": 1950 + }, + { + "epoch": 0.7234711814228884, + "grad_norm": 0.91015625, + "learning_rate": 4.308721820199529e-06, + "loss": 2.256, + "mean_token_accuracy": 0.7598229548229548, + "mean_token_accuracy_utility": 0.6574908424908423, + "step": 1955 + }, + { + "epoch": 0.7253214913498011, + "grad_norm": 0.86328125, + "learning_rate": 4.255711183158635e-06, + "loss": 2.1629, + "mean_token_accuracy": 0.7693894993894995, + "mean_token_accuracy_utility": 0.6692551892551891, + "step": 1960 + }, + { + "epoch": 0.7271718012767139, + "grad_norm": 0.84765625, + "learning_rate": 4.2029403767172175e-06, + "loss": 2.2382, + "mean_token_accuracy": 0.7657326007326007, + "mean_token_accuracy_utility": 0.6589194139194137, + "step": 1965 + }, + { + "epoch": 0.7290221112036266, + "grad_norm": 0.87109375, + "learning_rate": 4.150411604116531e-06, + "loss": 2.2574, + "mean_token_accuracy": 0.7630647130647132, + "mean_token_accuracy_utility": 0.6554578754578753, + "step": 1970 + }, + { + "epoch": 0.7308724211305394, + "grad_norm": 0.828125, + "learning_rate": 4.098127058492652e-06, + "loss": 2.1496, + "mean_token_accuracy": 0.7616727716727718, + "mean_token_accuracy_utility": 0.6744261294261292, + "step": 1975 + }, + { + "epoch": 0.7327227310574521, + "grad_norm": 0.84375, + "learning_rate": 4.0460889227849e-06, + "loss": 2.2243, + "mean_token_accuracy": 0.7622649572649574, + "mean_token_accuracy_utility": 0.6622954822954823, + "step": 1980 + }, + { + "epoch": 0.7345730409843649, + "grad_norm": 0.86328125, + "learning_rate": 3.9942993696447045e-06, + "loss": 2.2387, + "mean_token_accuracy": 0.7669536019536021, + "mean_token_accuracy_utility": 0.6604090354090352, + "step": 1985 + }, + { + "epoch": 0.7364233509112776, + "grad_norm": 0.828125, + "learning_rate": 3.942760561344877e-06, + "loss": 2.1439, + "mean_token_accuracy": 0.7604029304029304, + "mean_token_accuracy_utility": 0.6778510378510376, + "step": 1990 + }, + { + "epoch": 0.7382736608381903, + "grad_norm": 0.828125, + "learning_rate": 3.891474649689362e-06, + "loss": 2.2111, + "mean_token_accuracy": 0.7557081807081809, + "mean_token_accuracy_utility": 0.6673260073260072, + "step": 1995 + }, + { + "epoch": 0.7401239707651032, + "grad_norm": 0.85546875, + "learning_rate": 3.840443775923365e-06, + "loss": 2.1273, + "mean_token_accuracy": 0.7636752136752138, + "mean_token_accuracy_utility": 0.677130647130647, + "step": 2000 + }, + { + "epoch": 0.7401239707651032, + "eval_mean_token_accuracy": 0.7545354427707368, + "eval_reasoning_loss": 0.8224493861198425, + "eval_reasoning_runtime": 27.8097, + "eval_reasoning_samples_per_second": 4.639, + "eval_reasoning_steps_per_second": 1.187, + "step": 2000 + }, + { + "epoch": 0.7401239707651032, + "eval_mean_token_accuracy": 0.6618090580343939, + "eval_utility_loss": 1.4265598058700562, + "eval_utility_runtime": 10.4559, + "eval_utility_samples_per_second": 4.686, + "eval_utility_steps_per_second": 1.243, + "step": 2000 + }, + { + "epoch": 0.7419742806920159, + "grad_norm": 0.85546875, + "learning_rate": 3.7896700706439826e-06, + "loss": 2.2895, + "mean_token_accuracy": 0.7648778998778999, + "mean_token_accuracy_utility": 0.6490555028100982, + "step": 2005 + }, + { + "epoch": 0.7438245906189287, + "grad_norm": 0.84765625, + "learning_rate": 3.7391556537112282e-06, + "loss": 2.141, + "mean_token_accuracy": 0.7741880341880344, + "mean_token_accuracy_utility": 0.6662515262515261, + "step": 2010 + }, + { + "epoch": 0.7456749005458414, + "grad_norm": 0.84765625, + "learning_rate": 3.6889026341595378e-06, + "loss": 2.2072, + "mean_token_accuracy": 0.7767582417582419, + "mean_token_accuracy_utility": 0.6558730158730157, + "step": 2015 + }, + { + "epoch": 0.7475252104727542, + "grad_norm": 0.84375, + "learning_rate": 3.6389131101096953e-06, + "loss": 2.2036, + "mean_token_accuracy": 0.7685225885225886, + "mean_token_accuracy_utility": 0.659200244200244, + "step": 2020 + }, + { + "epoch": 0.7493755203996669, + "grad_norm": 0.83984375, + "learning_rate": 3.5891891686812597e-06, + "loss": 2.1545, + "mean_token_accuracy": 0.7576491393150013, + "mean_token_accuracy_utility": 0.6729238826941032, + "step": 2025 + }, + { + "epoch": 0.7512258303265797, + "grad_norm": 0.828125, + "learning_rate": 3.5397328859054138e-06, + "loss": 2.1244, + "mean_token_accuracy": 0.7617643467643469, + "mean_token_accuracy_utility": 0.6818742368742368, + "step": 2030 + }, + { + "epoch": 0.7530761402534925, + "grad_norm": 0.875, + "learning_rate": 3.490546326638273e-06, + "loss": 2.2311, + "mean_token_accuracy": 0.7556004384746521, + "mean_token_accuracy_utility": 0.6669780219780218, + "step": 2035 + }, + { + "epoch": 0.7549264501804053, + "grad_norm": 0.8828125, + "learning_rate": 3.441631544474705e-06, + "loss": 2.1758, + "mean_token_accuracy": 0.7800915750915753, + "mean_token_accuracy_utility": 0.660164835164835, + "step": 2040 + }, + { + "epoch": 0.756776760107318, + "grad_norm": 0.8671875, + "learning_rate": 3.3929905816625653e-06, + "loss": 2.3432, + "mean_token_accuracy": 0.766050061050061, + "mean_token_accuracy_utility": 0.6328876678876677, + "step": 2045 + }, + { + "epoch": 0.7586270700342307, + "grad_norm": 0.9140625, + "learning_rate": 3.344625469017445e-06, + "loss": 2.1597, + "mean_token_accuracy": 0.7613858363858365, + "mean_token_accuracy_utility": 0.6784004884004882, + "step": 2050 + }, + { + "epoch": 0.7604773799611435, + "grad_norm": 0.8671875, + "learning_rate": 3.2965382258378674e-06, + "loss": 2.142, + "mean_token_accuracy": 0.7816117216117218, + "mean_token_accuracy_utility": 0.6646092796092794, + "step": 2055 + }, + { + "epoch": 0.7623276898880562, + "grad_norm": 0.859375, + "learning_rate": 3.248730859821002e-06, + "loss": 2.2386, + "mean_token_accuracy": 0.763943833943834, + "mean_token_accuracy_utility": 0.6640537240537239, + "step": 2060 + }, + { + "epoch": 0.764177999814969, + "grad_norm": 0.88671875, + "learning_rate": 3.2012053669788136e-06, + "loss": 2.2184, + "mean_token_accuracy": 0.7610195360195361, + "mean_token_accuracy_utility": 0.6599511599511596, + "step": 2065 + }, + { + "epoch": 0.7660283097418817, + "grad_norm": 0.81640625, + "learning_rate": 3.1539637315547524e-06, + "loss": 2.0742, + "mean_token_accuracy": 0.7690293040293043, + "mean_token_accuracy_utility": 0.6834126984126982, + "step": 2070 + }, + { + "epoch": 0.7678786196687946, + "grad_norm": 0.84375, + "learning_rate": 3.1070079259408934e-06, + "loss": 2.1542, + "mean_token_accuracy": 0.7583394383394383, + "mean_token_accuracy_utility": 0.676263736263736, + "step": 2075 + }, + { + "epoch": 0.7697289295957073, + "grad_norm": 0.8515625, + "learning_rate": 3.0603399105955966e-06, + "loss": 2.2929, + "mean_token_accuracy": 0.7646459096459097, + "mean_token_accuracy_utility": 0.6539316239316237, + "step": 2080 + }, + { + "epoch": 0.7715792395226201, + "grad_norm": 0.90234375, + "learning_rate": 3.0139616339616394e-06, + "loss": 2.1881, + "mean_token_accuracy": 0.7685470085470086, + "mean_token_accuracy_utility": 0.6674314104643042, + "step": 2085 + }, + { + "epoch": 0.7734295494495328, + "grad_norm": 0.8203125, + "learning_rate": 2.9678750323848893e-06, + "loss": 2.1378, + "mean_token_accuracy": 0.7706959706959707, + "mean_token_accuracy_utility": 0.6755494505494503, + "step": 2090 + }, + { + "epoch": 0.7752798593764455, + "grad_norm": 0.87890625, + "learning_rate": 2.922082030033446e-06, + "loss": 2.1928, + "mean_token_accuracy": 0.7534676434676435, + "mean_token_accuracy_utility": 0.670197102090106, + "step": 2095 + }, + { + "epoch": 0.7771301693033583, + "grad_norm": 0.87109375, + "learning_rate": 2.8765845388172955e-06, + "loss": 2.2124, + "mean_token_accuracy": 0.7737240537240538, + "mean_token_accuracy_utility": 0.655880744784455, + "step": 2100 + }, + { + "epoch": 0.7771301693033583, + "eval_mean_token_accuracy": 0.7545335927688874, + "eval_reasoning_loss": 0.8224275708198547, + "eval_reasoning_runtime": 27.8205, + "eval_reasoning_samples_per_second": 4.637, + "eval_reasoning_steps_per_second": 1.186, + "step": 2100 + }, + { + "epoch": 0.7771301693033583, + "eval_mean_token_accuracy": 0.6619198629350189, + "eval_utility_loss": 1.4264271259307861, + "eval_utility_runtime": 10.4655, + "eval_utility_samples_per_second": 4.682, + "eval_utility_steps_per_second": 1.242, + "step": 2100 + }, + { + "epoch": 0.778980479230271, + "grad_norm": 0.8515625, + "learning_rate": 2.831384458308518e-06, + "loss": 2.1621, + "mean_token_accuracy": 0.779023199023199, + "mean_token_accuracy_utility": 0.6593406593406592, + "step": 2105 + }, + { + "epoch": 0.7808307891571838, + "grad_norm": 0.82421875, + "learning_rate": 2.7864836756619407e-06, + "loss": 2.2147, + "mean_token_accuracy": 0.7678510378510379, + "mean_token_accuracy_utility": 0.6560989010989011, + "step": 2110 + }, + { + "epoch": 0.7826810990840966, + "grad_norm": 0.875, + "learning_rate": 2.741884065536373e-06, + "loss": 2.1959, + "mean_token_accuracy": 0.7810256410256412, + "mean_token_accuracy_utility": 0.6546092796092794, + "step": 2115 + }, + { + "epoch": 0.7845314090110094, + "grad_norm": 0.84765625, + "learning_rate": 2.6975874900163223e-06, + "loss": 2.235, + "mean_token_accuracy": 0.7669474969474971, + "mean_token_accuracy_utility": 0.6569719169719168, + "step": 2120 + }, + { + "epoch": 0.7863817189379221, + "grad_norm": 0.8359375, + "learning_rate": 2.6535957985342653e-06, + "loss": 2.1397, + "mean_token_accuracy": 0.7729181929181931, + "mean_token_accuracy_utility": 0.6716178266178265, + "step": 2125 + }, + { + "epoch": 0.7882320288648349, + "grad_norm": 0.890625, + "learning_rate": 2.6099108277934105e-06, + "loss": 2.1194, + "mean_token_accuracy": 0.7810561660561662, + "mean_token_accuracy_utility": 0.6675364596445064, + "step": 2130 + }, + { + "epoch": 0.7900823387917476, + "grad_norm": 0.828125, + "learning_rate": 2.5665344016910367e-06, + "loss": 2.0647, + "mean_token_accuracy": 0.7735409035409034, + "mean_token_accuracy_utility": 0.678235653235653, + "step": 2135 + }, + { + "epoch": 0.7919326487186604, + "grad_norm": 0.84375, + "learning_rate": 2.523468331242329e-06, + "loss": 2.129, + "mean_token_accuracy": 0.7656349206349209, + "mean_token_accuracy_utility": 0.6754151404151403, + "step": 2140 + }, + { + "epoch": 0.7937829586455731, + "grad_norm": 0.8984375, + "learning_rate": 2.4807144145047734e-06, + "loss": 2.1923, + "mean_token_accuracy": 0.7529914529914529, + "mean_token_accuracy_utility": 0.6723626373626372, + "step": 2145 + }, + { + "epoch": 0.7956332685724858, + "grad_norm": 0.85546875, + "learning_rate": 2.438274436503074e-06, + "loss": 2.1609, + "mean_token_accuracy": 0.7800976800976801, + "mean_token_accuracy_utility": 0.6617298864395986, + "step": 2150 + }, + { + "epoch": 0.7974835784993987, + "grad_norm": 0.828125, + "learning_rate": 2.396150169154644e-06, + "loss": 2.2838, + "mean_token_accuracy": 0.7600793650793651, + "mean_token_accuracy_utility": 0.6541025641025638, + "step": 2155 + }, + { + "epoch": 0.7993338884263114, + "grad_norm": 0.87890625, + "learning_rate": 2.3543433711956197e-06, + "loss": 2.2226, + "mean_token_accuracy": 0.7615750915750918, + "mean_token_accuracy_utility": 0.6617704517704516, + "step": 2160 + }, + { + "epoch": 0.8011841983532242, + "grad_norm": 0.84765625, + "learning_rate": 2.3128557881074153e-06, + "loss": 2.2034, + "mean_token_accuracy": 0.7720024420024421, + "mean_token_accuracy_utility": 0.6659462759462758, + "step": 2165 + }, + { + "epoch": 0.8030345082801369, + "grad_norm": 0.875, + "learning_rate": 2.271689152043873e-06, + "loss": 2.2192, + "mean_token_accuracy": 0.7640354090354091, + "mean_token_accuracy_utility": 0.6588095238095237, + "step": 2170 + }, + { + "epoch": 0.8048848182070497, + "grad_norm": 0.89453125, + "learning_rate": 2.230845181758928e-06, + "loss": 2.2147, + "mean_token_accuracy": 0.7481746031746033, + "mean_token_accuracy_utility": 0.674242979242979, + "step": 2175 + }, + { + "epoch": 0.8067351281339624, + "grad_norm": 0.88671875, + "learning_rate": 2.1903255825348533e-06, + "loss": 2.3004, + "mean_token_accuracy": 0.7742429792429795, + "mean_token_accuracy_utility": 0.6433455433455432, + "step": 2180 + }, + { + "epoch": 0.8085854380608752, + "grad_norm": 0.8984375, + "learning_rate": 2.150132046111054e-06, + "loss": 2.1447, + "mean_token_accuracy": 0.7642673992673992, + "mean_token_accuracy_utility": 0.6739316239316236, + "step": 2185 + }, + { + "epoch": 0.810435747987788, + "grad_norm": 0.88671875, + "learning_rate": 2.1102662506134506e-06, + "loss": 2.282, + "mean_token_accuracy": 0.7498717948717948, + "mean_token_accuracy_utility": 0.6562210012210011, + "step": 2190 + }, + { + "epoch": 0.8122860579147008, + "grad_norm": 0.8671875, + "learning_rate": 2.0707298604843964e-06, + "loss": 2.2306, + "mean_token_accuracy": 0.7679304029304029, + "mean_token_accuracy_utility": 0.6576800976800974, + "step": 2195 + }, + { + "epoch": 0.8141363678416135, + "grad_norm": 0.87890625, + "learning_rate": 2.03152452641321e-06, + "loss": 2.1785, + "mean_token_accuracy": 0.7783699633699634, + "mean_token_accuracy_utility": 0.6562942612942612, + "step": 2200 + }, + { + "epoch": 0.8141363678416135, + "eval_mean_token_accuracy": 0.7545428427781371, + "eval_reasoning_loss": 0.8223759531974792, + "eval_reasoning_runtime": 27.8404, + "eval_reasoning_samples_per_second": 4.634, + "eval_reasoning_steps_per_second": 1.185, + "step": 2200 + }, + { + "epoch": 0.8141363678416135, + "eval_mean_token_accuracy": 0.6619626346392147, + "eval_utility_loss": 1.4265111684799194, + "eval_utility_runtime": 10.4557, + "eval_utility_samples_per_second": 4.686, + "eval_utility_steps_per_second": 1.243, + "step": 2200 + }, + { + "epoch": 0.8159866777685262, + "grad_norm": 0.84375, + "learning_rate": 1.9926518852672294e-06, + "loss": 2.1329, + "mean_token_accuracy": 0.764761904761905, + "mean_token_accuracy_utility": 0.6750183150183149, + "step": 2205 + }, + { + "epoch": 0.817836987695439, + "grad_norm": 0.85546875, + "learning_rate": 1.9541135600234917e-06, + "loss": 2.1485, + "mean_token_accuracy": 0.7752380952380953, + "mean_token_accuracy_utility": 0.6705006105006103, + "step": 2210 + }, + { + "epoch": 0.8196872976223517, + "grad_norm": 0.83203125, + "learning_rate": 1.9159111597009584e-06, + "loss": 2.2147, + "mean_token_accuracy": 0.7742368742368743, + "mean_token_accuracy_utility": 0.6570268620268618, + "step": 2215 + }, + { + "epoch": 0.8215376075492645, + "grad_norm": 0.890625, + "learning_rate": 1.8780462792933473e-06, + "loss": 2.2442, + "mean_token_accuracy": 0.775915750915751, + "mean_token_accuracy_utility": 0.6475457875457875, + "step": 2220 + }, + { + "epoch": 0.8233879174761772, + "grad_norm": 0.8828125, + "learning_rate": 1.8405204997025394e-06, + "loss": 2.2716, + "mean_token_accuracy": 0.7447680097680096, + "mean_token_accuracy_utility": 0.6616422466422465, + "step": 2225 + }, + { + "epoch": 0.82523822740309, + "grad_norm": 0.86328125, + "learning_rate": 1.8033353876725578e-06, + "loss": 2.2178, + "mean_token_accuracy": 0.7495482295482295, + "mean_token_accuracy_utility": 0.6728937728937728, + "step": 2230 + }, + { + "epoch": 0.8270885373300028, + "grad_norm": 0.86328125, + "learning_rate": 1.766492495724178e-06, + "loss": 2.2129, + "mean_token_accuracy": 0.7724847374847377, + "mean_token_accuracy_utility": 0.657258852258852, + "step": 2235 + }, + { + "epoch": 0.8289388472569156, + "grad_norm": 0.875, + "learning_rate": 1.7299933620900945e-06, + "loss": 2.1881, + "mean_token_accuracy": 0.760915750915751, + "mean_token_accuracy_utility": 0.6684065934065933, + "step": 2240 + }, + { + "epoch": 0.8307891571838283, + "grad_norm": 0.84375, + "learning_rate": 1.6938395106507034e-06, + "loss": 2.0872, + "mean_token_accuracy": 0.7895543345543348, + "mean_token_accuracy_utility": 0.6696336996336993, + "step": 2245 + }, + { + "epoch": 0.832639467110741, + "grad_norm": 0.83984375, + "learning_rate": 1.658032450870467e-06, + "loss": 2.2115, + "mean_token_accuracy": 0.7676434676434676, + "mean_token_accuracy_utility": 0.6597069597069596, + "step": 2250 + }, + { + "epoch": 0.8344897770376538, + "grad_norm": 0.8359375, + "learning_rate": 1.622573677734911e-06, + "loss": 2.1579, + "mean_token_accuracy": 0.7789438339438339, + "mean_token_accuracy_utility": 0.6628510378510377, + "step": 2255 + }, + { + "epoch": 0.8363400869645665, + "grad_norm": 0.87890625, + "learning_rate": 1.587464671688187e-06, + "loss": 2.1319, + "mean_token_accuracy": 0.7603052503052503, + "mean_token_accuracy_utility": 0.6774786324786323, + "step": 2260 + }, + { + "epoch": 0.8381903968914793, + "grad_norm": 0.8515625, + "learning_rate": 1.552706898571288e-06, + "loss": 2.1875, + "mean_token_accuracy": 0.7578388278388278, + "mean_token_accuracy_utility": 0.6691882657464409, + "step": 2265 + }, + { + "epoch": 0.840040706818392, + "grad_norm": 0.890625, + "learning_rate": 1.5183018095608138e-06, + "loss": 2.1996, + "mean_token_accuracy": 0.7605827474329176, + "mean_token_accuracy_utility": 0.6661844367015097, + "step": 2270 + }, + { + "epoch": 0.8418910167453049, + "grad_norm": 0.82421875, + "learning_rate": 1.4842508411084145e-06, + "loss": 2.0852, + "mean_token_accuracy": 0.7748351648351649, + "mean_token_accuracy_utility": 0.6854273504273503, + "step": 2275 + }, + { + "epoch": 0.8437413266722176, + "grad_norm": 0.8828125, + "learning_rate": 1.4505554148807954e-06, + "loss": 2.3427, + "mean_token_accuracy": 0.7576862026862028, + "mean_token_accuracy_utility": 0.6580830280830279, + "step": 2280 + }, + { + "epoch": 0.8455916365991304, + "grad_norm": 0.85546875, + "learning_rate": 1.4172169377003775e-06, + "loss": 2.168, + "mean_token_accuracy": 0.7654029304029305, + "mean_token_accuracy_utility": 0.6774542124542122, + "step": 2285 + }, + { + "epoch": 0.8474419465260431, + "grad_norm": 0.859375, + "learning_rate": 1.3842368014865414e-06, + "loss": 2.1758, + "mean_token_accuracy": 0.7708363858363858, + "mean_token_accuracy_utility": 0.6704321638367426, + "step": 2290 + }, + { + "epoch": 0.8492922564529559, + "grad_norm": 0.8984375, + "learning_rate": 1.3516163831975337e-06, + "loss": 2.2074, + "mean_token_accuracy": 0.7722344322344324, + "mean_token_accuracy_utility": 0.6598778998778997, + "step": 2295 + }, + { + "epoch": 0.8511425663798686, + "grad_norm": 0.9375, + "learning_rate": 1.3193570447729642e-06, + "loss": 2.2162, + "mean_token_accuracy": 0.7583028083028084, + "mean_token_accuracy_utility": 0.6666468577419067, + "step": 2300 + }, + { + "epoch": 0.8511425663798686, + "eval_mean_token_accuracy": 0.7544947427300368, + "eval_reasoning_loss": 0.8223364949226379, + "eval_reasoning_runtime": 27.7295, + "eval_reasoning_samples_per_second": 4.652, + "eval_reasoning_steps_per_second": 1.19, + "step": 2300 + }, + { + "epoch": 0.8511425663798686, + "eval_mean_token_accuracy": 0.6618405345171142, + "eval_utility_loss": 1.4264107942581177, + "eval_utility_runtime": 10.4276, + "eval_utility_samples_per_second": 4.699, + "eval_utility_steps_per_second": 1.247, + "step": 2300 + }, + { + "epoch": 0.8529928763067813, + "grad_norm": 0.8828125, + "learning_rate": 1.2874601330769488e-06, + "loss": 2.2343, + "mean_token_accuracy": 0.7560927960927961, + "mean_token_accuracy_utility": 0.665127737306407, + "step": 2305 + }, + { + "epoch": 0.8548431862336942, + "grad_norm": 0.828125, + "learning_rate": 1.255926979841876e-06, + "loss": 2.1945, + "mean_token_accuracy": 0.774847374847375, + "mean_token_accuracy_utility": 0.6619047619047617, + "step": 2310 + }, + { + "epoch": 0.8566934961606069, + "grad_norm": 0.859375, + "learning_rate": 1.224758901612796e-06, + "loss": 2.2382, + "mean_token_accuracy": 0.7701282051282053, + "mean_token_accuracy_utility": 0.6548962148962147, + "step": 2315 + }, + { + "epoch": 0.8585438060875197, + "grad_norm": 0.86328125, + "learning_rate": 1.1939571996924738e-06, + "loss": 2.1969, + "mean_token_accuracy": 0.7777655677655677, + "mean_token_accuracy_utility": 0.6601343101343101, + "step": 2320 + }, + { + "epoch": 0.8603941160144324, + "grad_norm": 0.83203125, + "learning_rate": 1.1635231600870334e-06, + "loss": 2.2628, + "mean_token_accuracy": 0.7561599511599512, + "mean_token_accuracy_utility": 0.6533211233211231, + "step": 2325 + }, + { + "epoch": 0.8622444259413452, + "grad_norm": 0.82421875, + "learning_rate": 1.1334580534522932e-06, + "loss": 2.1805, + "mean_token_accuracy": 0.7649145299145299, + "mean_token_accuracy_utility": 0.669065934065934, + "step": 2330 + }, + { + "epoch": 0.8640947358682579, + "grad_norm": 0.85546875, + "learning_rate": 1.1037631350406874e-06, + "loss": 2.1588, + "mean_token_accuracy": 0.7927106227106226, + "mean_token_accuracy_utility": 0.6527577667505907, + "step": 2335 + }, + { + "epoch": 0.8659450457951707, + "grad_norm": 0.8671875, + "learning_rate": 1.0744396446488781e-06, + "loss": 2.2309, + "mean_token_accuracy": 0.7623260073260075, + "mean_token_accuracy_utility": 0.6615689865689863, + "step": 2340 + }, + { + "epoch": 0.8677953557220834, + "grad_norm": 0.86328125, + "learning_rate": 1.0454888065659775e-06, + "loss": 2.2529, + "mean_token_accuracy": 0.7505067155067155, + "mean_token_accuracy_utility": 0.6604395604395603, + "step": 2345 + }, + { + "epoch": 0.8696456656489963, + "grad_norm": 0.8671875, + "learning_rate": 1.0169118295224488e-06, + "loss": 2.3307, + "mean_token_accuracy": 0.7554395604395603, + "mean_token_accuracy_utility": 0.6505067155067155, + "step": 2350 + }, + { + "epoch": 0.871495975575909, + "grad_norm": 0.85546875, + "learning_rate": 9.887099066396178e-07, + "loss": 2.2333, + "mean_token_accuracy": 0.7552075702075702, + "mean_token_accuracy_utility": 0.6622100122100121, + "step": 2355 + }, + { + "epoch": 0.8733462855028217, + "grad_norm": 0.890625, + "learning_rate": 9.608842153798903e-07, + "loss": 2.2772, + "mean_token_accuracy": 0.7497435897435898, + "mean_token_accuracy_utility": 0.6652635977883501, + "step": 2360 + }, + { + "epoch": 0.8751965954297345, + "grad_norm": 0.8515625, + "learning_rate": 9.33435917497556e-07, + "loss": 2.1567, + "mean_token_accuracy": 0.7688400488400489, + "mean_token_accuracy_utility": 0.6681135531135529, + "step": 2365 + }, + { + "epoch": 0.8770469053566472, + "grad_norm": 0.8515625, + "learning_rate": 9.063661589903116e-07, + "loss": 2.1665, + "mean_token_accuracy": 0.7672405372405373, + "mean_token_accuracy_utility": 0.6686495895168189, + "step": 2370 + }, + { + "epoch": 0.87889721528356, + "grad_norm": 0.8515625, + "learning_rate": 8.796760700513984e-07, + "loss": 2.1697, + "mean_token_accuracy": 0.7526007326007325, + "mean_token_accuracy_utility": 0.6819227994227994, + "step": 2375 + }, + { + "epoch": 0.8807475252104727, + "grad_norm": 0.83984375, + "learning_rate": 8.533667650224253e-07, + "loss": 2.2144, + "mean_token_accuracy": 0.7666971916971919, + "mean_token_accuracy_utility": 0.6609157509157508, + "step": 2380 + }, + { + "epoch": 0.8825978351373855, + "grad_norm": 0.88671875, + "learning_rate": 8.274393423468385e-07, + "loss": 2.3034, + "mean_token_accuracy": 0.7512393162393162, + "mean_token_accuracy_utility": 0.6583516483516481, + "step": 2385 + }, + { + "epoch": 0.8844481450642983, + "grad_norm": 0.859375, + "learning_rate": 8.018948845240538e-07, + "loss": 2.2697, + "mean_token_accuracy": 0.7598412698412698, + "mean_token_accuracy_utility": 0.6563492063492061, + "step": 2390 + }, + { + "epoch": 0.8862984549912111, + "grad_norm": 0.90625, + "learning_rate": 7.767344580642821e-07, + "loss": 2.2586, + "mean_token_accuracy": 0.7684004884004885, + "mean_token_accuracy_utility": 0.6523260073260072, + "step": 2395 + }, + { + "epoch": 0.8881487649181238, + "grad_norm": 0.8359375, + "learning_rate": 7.519591134439753e-07, + "loss": 2.2411, + "mean_token_accuracy": 0.7680708180708182, + "mean_token_accuracy_utility": 0.6576373626373625, + "step": 2400 + }, + { + "epoch": 0.8881487649181238, + "eval_mean_token_accuracy": 0.7544630750513104, + "eval_reasoning_loss": 0.8223586082458496, + "eval_reasoning_runtime": 27.7469, + "eval_reasoning_samples_per_second": 4.649, + "eval_reasoning_steps_per_second": 1.189, + "step": 2400 + }, + { + "epoch": 0.8881487649181238, + "eval_mean_token_accuracy": 0.6624204489946552, + "eval_utility_loss": 1.4264180660247803, + "eval_utility_runtime": 10.4582, + "eval_utility_samples_per_second": 4.685, + "eval_utility_steps_per_second": 1.243, + "step": 2400 + }, + { + "epoch": 0.8899990748450366, + "grad_norm": 0.84765625, + "learning_rate": 7.275698850619861e-07, + "loss": 2.235, + "mean_token_accuracy": 0.7691636141636142, + "mean_token_accuracy_utility": 0.6574236874236872, + "step": 2405 + }, + { + "epoch": 0.8918493847719493, + "grad_norm": 0.86328125, + "learning_rate": 7.035677911963712e-07, + "loss": 2.126, + "mean_token_accuracy": 0.7774725274725276, + "mean_token_accuracy_utility": 0.6715567765567764, + "step": 2410 + }, + { + "epoch": 0.893699694698862, + "grad_norm": 0.83203125, + "learning_rate": 6.799538339618838e-07, + "loss": 2.1321, + "mean_token_accuracy": 0.7607570207570207, + "mean_token_accuracy_utility": 0.6794505494505494, + "step": 2415 + }, + { + "epoch": 0.8955500046257748, + "grad_norm": 0.82421875, + "learning_rate": 6.567289992681258e-07, + "loss": 2.1893, + "mean_token_accuracy": 0.7710927960927961, + "mean_token_accuracy_utility": 0.6596459096459095, + "step": 2420 + }, + { + "epoch": 0.8974003145526875, + "grad_norm": 0.87890625, + "learning_rate": 6.33894256778399e-07, + "loss": 2.2644, + "mean_token_accuracy": 0.7607081807081808, + "mean_token_accuracy_utility": 0.656062271062271, + "step": 2425 + }, + { + "epoch": 0.8992506244796004, + "grad_norm": 0.91015625, + "learning_rate": 6.114505598692011e-07, + "loss": 2.2413, + "mean_token_accuracy": 0.7624786324786327, + "mean_token_accuracy_utility": 0.6565873015873014, + "step": 2430 + }, + { + "epoch": 0.9011009344065131, + "grad_norm": 0.83984375, + "learning_rate": 5.893988455904387e-07, + "loss": 2.1501, + "mean_token_accuracy": 0.7771123321123323, + "mean_token_accuracy_utility": 0.6650324081575961, + "step": 2435 + }, + { + "epoch": 0.9029512443334259, + "grad_norm": 0.82421875, + "learning_rate": 5.677400346262918e-07, + "loss": 2.2225, + "mean_token_accuracy": 0.7724542124542128, + "mean_token_accuracy_utility": 0.6567460317460315, + "step": 2440 + }, + { + "epoch": 0.9048015542603386, + "grad_norm": 0.86328125, + "learning_rate": 5.464750312567835e-07, + "loss": 2.0648, + "mean_token_accuracy": 0.7891025641025644, + "mean_token_accuracy_utility": 0.6747581449769537, + "step": 2445 + }, + { + "epoch": 0.9066518641872514, + "grad_norm": 0.8515625, + "learning_rate": 5.256047233200201e-07, + "loss": 2.2244, + "mean_token_accuracy": 0.7660561660561662, + "mean_token_accuracy_utility": 0.6590842490842489, + "step": 2450 + }, + { + "epoch": 0.9085021741141641, + "grad_norm": 0.82421875, + "learning_rate": 5.051299821751254e-07, + "loss": 2.0982, + "mean_token_accuracy": 0.761965811965812, + "mean_token_accuracy_utility": 0.6856593406593404, + "step": 2455 + }, + { + "epoch": 0.9103524840410768, + "grad_norm": 0.859375, + "learning_rate": 4.850516626658585e-07, + "loss": 2.28, + "mean_token_accuracy": 0.7545787545787548, + "mean_token_accuracy_utility": 0.6622771672771671, + "step": 2460 + }, + { + "epoch": 0.9122027939679896, + "grad_norm": 0.86328125, + "learning_rate": 4.653706030849214e-07, + "loss": 2.1914, + "mean_token_accuracy": 0.7651465201465202, + "mean_token_accuracy_utility": 0.6661538461538459, + "step": 2465 + }, + { + "epoch": 0.9140531038949024, + "grad_norm": 0.8828125, + "learning_rate": 4.4608762513896455e-07, + "loss": 2.2316, + "mean_token_accuracy": 0.7483333333333334, + "mean_token_accuracy_utility": 0.6697191697191694, + "step": 2470 + }, + { + "epoch": 0.9159034138218152, + "grad_norm": 0.84765625, + "learning_rate": 4.2720353391427547e-07, + "loss": 2.1493, + "mean_token_accuracy": 0.7710073260073261, + "mean_token_accuracy_utility": 0.6650724180034523, + "step": 2475 + }, + { + "epoch": 0.9177537237487279, + "grad_norm": 0.80078125, + "learning_rate": 4.087191178431682e-07, + "loss": 2.149, + "mean_token_accuracy": 0.7607753357753358, + "mean_token_accuracy_utility": 0.6747741147741146, + "step": 2480 + }, + { + "epoch": 0.9196040336756407, + "grad_norm": 0.83203125, + "learning_rate": 3.9063514867105914e-07, + "loss": 2.2972, + "mean_token_accuracy": 0.7439682539682538, + "mean_token_accuracy_utility": 0.6640048840048839, + "step": 2485 + }, + { + "epoch": 0.9214543436025534, + "grad_norm": 0.8515625, + "learning_rate": 3.729523814242608e-07, + "loss": 2.2168, + "mean_token_accuracy": 0.7636080586080587, + "mean_token_accuracy_utility": 0.6628021978021976, + "step": 2490 + }, + { + "epoch": 0.9233046535294662, + "grad_norm": 0.8203125, + "learning_rate": 3.5567155437843725e-07, + "loss": 2.1075, + "mean_token_accuracy": 0.771935286935287, + "mean_token_accuracy_utility": 0.6766910866910866, + "step": 2495 + }, + { + "epoch": 0.9251549634563789, + "grad_norm": 0.9453125, + "learning_rate": 3.3879338902779945e-07, + "loss": 2.1709, + "mean_token_accuracy": 0.7821978021978023, + "mean_token_accuracy_utility": 0.652167277167277, + "step": 2500 + }, + { + "epoch": 0.9251549634563789, + "eval_mean_token_accuracy": 0.75450932509756, + "eval_reasoning_loss": 0.8223300576210022, + "eval_reasoning_runtime": 27.7381, + "eval_reasoning_samples_per_second": 4.651, + "eval_reasoning_steps_per_second": 1.19, + "step": 2500 + }, + { + "epoch": 0.9251549634563789, + "eval_mean_token_accuracy": 0.661952736044815, + "eval_utility_loss": 1.426324486732483, + "eval_utility_runtime": 10.4467, + "eval_utility_samples_per_second": 4.69, + "eval_utility_steps_per_second": 1.244, + "step": 2500 + }, + { + "epoch": 0.9270052733832918, + "grad_norm": 0.85546875, + "learning_rate": 3.223185900549686e-07, + "loss": 2.1386, + "mean_token_accuracy": 0.7610683760683761, + "mean_token_accuracy_utility": 0.6803846153846151, + "step": 2505 + }, + { + "epoch": 0.9288555833102045, + "grad_norm": 0.89453125, + "learning_rate": 3.0624784530156384e-07, + "loss": 2.2633, + "mean_token_accuracy": 0.7639072039072039, + "mean_token_accuracy_utility": 0.657258852258852, + "step": 2510 + }, + { + "epoch": 0.9307058932371172, + "grad_norm": 0.82421875, + "learning_rate": 2.905818257394799e-07, + "loss": 2.0388, + "mean_token_accuracy": 0.7794932844932846, + "mean_token_accuracy_utility": 0.688260073260073, + "step": 2515 + }, + { + "epoch": 0.93255620316403, + "grad_norm": 0.85546875, + "learning_rate": 2.753211854428728e-07, + "loss": 2.1488, + "mean_token_accuracy": 0.7787301587301588, + "mean_token_accuracy_utility": 0.6624297924297922, + "step": 2520 + }, + { + "epoch": 0.9344065130909427, + "grad_norm": 0.84765625, + "learning_rate": 2.604665615608526e-07, + "loss": 2.2945, + "mean_token_accuracy": 0.7641517867105334, + "mean_token_accuracy_utility": 0.6483032065051306, + "step": 2525 + }, + { + "epoch": 0.9362568230178555, + "grad_norm": 0.86328125, + "learning_rate": 2.460185742908816e-07, + "loss": 2.1859, + "mean_token_accuracy": 0.7532905982905984, + "mean_token_accuracy_utility": 0.6721245421245419, + "step": 2530 + }, + { + "epoch": 0.9381071329447682, + "grad_norm": 0.8671875, + "learning_rate": 2.3197782685288385e-07, + "loss": 2.2633, + "mean_token_accuracy": 0.7602380952380953, + "mean_token_accuracy_utility": 0.6547435897435896, + "step": 2535 + }, + { + "epoch": 0.939957442871681, + "grad_norm": 0.81640625, + "learning_rate": 2.1834490546405186e-07, + "loss": 2.1884, + "mean_token_accuracy": 0.7644749694749695, + "mean_token_accuracy_utility": 0.6680455179867885, + "step": 2540 + }, + { + "epoch": 0.9418077527985937, + "grad_norm": 0.83984375, + "learning_rate": 2.0512037931437855e-07, + "loss": 2.2728, + "mean_token_accuracy": 0.7575579975579977, + "mean_token_accuracy_utility": 0.6558669108669106, + "step": 2545 + }, + { + "epoch": 0.9436580627255066, + "grad_norm": 0.859375, + "learning_rate": 1.9230480054288958e-07, + "loss": 2.1533, + "mean_token_accuracy": 0.7816483516483517, + "mean_token_accuracy_utility": 0.6623870573870573, + "step": 2550 + }, + { + "epoch": 0.9455083726524193, + "grad_norm": 0.87109375, + "learning_rate": 1.7989870421459498e-07, + "loss": 2.1223, + "mean_token_accuracy": 0.7732295482295484, + "mean_token_accuracy_utility": 0.6732016435533865, + "step": 2555 + }, + { + "epoch": 0.9473586825793321, + "grad_norm": 0.86328125, + "learning_rate": 1.6790260829814053e-07, + "loss": 2.1297, + "mean_token_accuracy": 0.7583150183150185, + "mean_token_accuracy_utility": 0.6855250305250303, + "step": 2560 + }, + { + "epoch": 0.9492089925062448, + "grad_norm": 0.82421875, + "learning_rate": 1.5631701364419492e-07, + "loss": 2.0219, + "mean_token_accuracy": 0.7890842490842491, + "mean_token_accuracy_utility": 0.6833394383394381, + "step": 2565 + }, + { + "epoch": 0.9510593024331575, + "grad_norm": 0.83203125, + "learning_rate": 1.4514240396452438e-07, + "loss": 2.2477, + "mean_token_accuracy": 0.7726984126984129, + "mean_token_accuracy_utility": 0.6523565323565321, + "step": 2570 + }, + { + "epoch": 0.9529096123600703, + "grad_norm": 0.82421875, + "learning_rate": 1.3437924581181205e-07, + "loss": 2.1129, + "mean_token_accuracy": 0.7819474969474972, + "mean_token_accuracy_utility": 0.6665628815628815, + "step": 2575 + }, + { + "epoch": 0.954759922286983, + "grad_norm": 0.92578125, + "learning_rate": 1.2402798856016474e-07, + "loss": 2.2132, + "mean_token_accuracy": 0.7512820512820513, + "mean_token_accuracy_utility": 0.6739987789987789, + "step": 2580 + }, + { + "epoch": 0.9566102322138959, + "grad_norm": 0.85546875, + "learning_rate": 1.1408906438636236e-07, + "loss": 2.2925, + "mean_token_accuracy": 0.77489010989011, + "mean_token_accuracy_utility": 0.6464774114774113, + "step": 2585 + }, + { + "epoch": 0.9584605421408086, + "grad_norm": 0.85546875, + "learning_rate": 1.045628882518046e-07, + "loss": 2.1244, + "mean_token_accuracy": 0.7775030525030526, + "mean_token_accuracy_utility": 0.6698046398046396, + "step": 2590 + }, + { + "epoch": 0.9603108520677214, + "grad_norm": 0.8515625, + "learning_rate": 9.544985788519589e-08, + "loss": 2.2013, + "mean_token_accuracy": 0.7552442002442004, + "mean_token_accuracy_utility": 0.6779603707590447, + "step": 2595 + }, + { + "epoch": 0.9621611619946341, + "grad_norm": 0.859375, + "learning_rate": 8.675035376593088e-08, + "loss": 2.2471, + "mean_token_accuracy": 0.7656715506715508, + "mean_token_accuracy_utility": 0.6532234432234431, + "step": 2600 + }, + { + "epoch": 0.9621611619946341, + "eval_mean_token_accuracy": 0.7545872428225371, + "eval_reasoning_loss": 0.8223090767860413, + "eval_reasoning_runtime": 27.7693, + "eval_reasoning_samples_per_second": 4.645, + "eval_reasoning_steps_per_second": 1.188, + "step": 2600 + }, + { + "epoch": 0.9621611619946341, + "eval_mean_token_accuracy": 0.662000203907553, + "eval_utility_loss": 1.4264215230941772, + "eval_utility_runtime": 10.4446, + "eval_utility_samples_per_second": 4.691, + "eval_utility_steps_per_second": 1.245, + "step": 2600 + }, + { + "epoch": 0.9640114719215469, + "grad_norm": 0.8203125, + "learning_rate": 7.846473910821162e-08, + "loss": 2.1683, + "mean_token_accuracy": 0.760848595848596, + "mean_token_accuracy_utility": 0.6717149106090674, + "step": 2605 + }, + { + "epoch": 0.9658617818484596, + "grad_norm": 0.875, + "learning_rate": 7.059335984588634e-08, + "loss": 2.1865, + "mean_token_accuracy": 0.7841880341880344, + "mean_token_accuracy_utility": 0.6532362062881869, + "step": 2610 + }, + { + "epoch": 0.9677120917753724, + "grad_norm": 0.86328125, + "learning_rate": 6.313654461800322e-08, + "loss": 2.2174, + "mean_token_accuracy": 0.7492857142857142, + "mean_token_accuracy_utility": 0.676196581196581, + "step": 2615 + }, + { + "epoch": 0.9695624017022851, + "grad_norm": 0.82421875, + "learning_rate": 5.609460475509032e-08, + "loss": 2.1222, + "mean_token_accuracy": 0.776880341880342, + "mean_token_accuracy_utility": 0.6656593406593404, + "step": 2620 + }, + { + "epoch": 0.9714127116291978, + "grad_norm": 0.84375, + "learning_rate": 4.9467834266154756e-08, + "loss": 2.1802, + "mean_token_accuracy": 0.7569597069597069, + "mean_token_accuracy_utility": 0.6725396825396823, + "step": 2625 + }, + { + "epoch": 0.9732630215561107, + "grad_norm": 0.82421875, + "learning_rate": 4.325650982641039e-08, + "loss": 2.2752, + "mean_token_accuracy": 0.7637362637362637, + "mean_token_accuracy_utility": 0.6598079896449887, + "step": 2630 + }, + { + "epoch": 0.9751133314830234, + "grad_norm": 0.86328125, + "learning_rate": 3.746089076572701e-08, + "loss": 2.3106, + "mean_token_accuracy": 0.7700549450549452, + "mean_token_accuracy_utility": 0.6395054945054943, + "step": 2635 + }, + { + "epoch": 0.9769636414099362, + "grad_norm": 0.84375, + "learning_rate": 3.208121905779904e-08, + "loss": 2.143, + "mean_token_accuracy": 0.7801159951159953, + "mean_token_accuracy_utility": 0.662148962148962, + "step": 2640 + }, + { + "epoch": 0.9788139513368489, + "grad_norm": 0.90625, + "learning_rate": 2.711771931004692e-08, + "loss": 2.2312, + "mean_token_accuracy": 0.7701221001221, + "mean_token_accuracy_utility": 0.655201465201465, + "step": 2645 + }, + { + "epoch": 0.9806642612637617, + "grad_norm": 0.83984375, + "learning_rate": 2.257059875423795e-08, + "loss": 2.2609, + "mean_token_accuracy": 0.7483455433455435, + "mean_token_accuracy_utility": 0.6668070818070816, + "step": 2650 + }, + { + "epoch": 0.9825145711906744, + "grad_norm": 0.85546875, + "learning_rate": 1.8440047237832105e-08, + "loss": 2.0578, + "mean_token_accuracy": 0.7902380952380953, + "mean_token_accuracy_utility": 0.6688461538461536, + "step": 2655 + }, + { + "epoch": 0.9843648811175872, + "grad_norm": 0.84765625, + "learning_rate": 1.472623721606059e-08, + "loss": 2.1455, + "mean_token_accuracy": 0.7715750915750919, + "mean_token_accuracy_utility": 0.6751098901098899, + "step": 2660 + }, + { + "epoch": 0.9862151910445, + "grad_norm": 0.90234375, + "learning_rate": 1.1429323744720499e-08, + "loss": 2.2023, + "mean_token_accuracy": 0.7663919413919414, + "mean_token_accuracy_utility": 0.6606776556776556, + "step": 2665 + }, + { + "epoch": 0.9880655009714127, + "grad_norm": 0.84765625, + "learning_rate": 8.549444473702207e-09, + "loss": 2.2777, + "mean_token_accuracy": 0.752014652014652, + "mean_token_accuracy_utility": 0.6577960927960926, + "step": 2670 + }, + { + "epoch": 0.9899158108983255, + "grad_norm": 0.86328125, + "learning_rate": 6.086719641246186e-09, + "loss": 2.1224, + "mean_token_accuracy": 0.7816483516483519, + "mean_token_accuracy_utility": 0.6655067155067151, + "step": 2675 + }, + { + "epoch": 0.9917661208252382, + "grad_norm": 0.82421875, + "learning_rate": 4.041252068918145e-09, + "loss": 2.1489, + "mean_token_accuracy": 0.7794932844932847, + "mean_token_accuracy_utility": 0.6603235653235651, + "step": 2680 + }, + { + "epoch": 0.993616430752151, + "grad_norm": 0.85546875, + "learning_rate": 2.4131271573191172e-09, + "loss": 2.3229, + "mean_token_accuracy": 0.7709340659340661, + "mean_token_accuracy_utility": 0.6439682539682539, + "step": 2685 + }, + { + "epoch": 0.9954667406790637, + "grad_norm": 0.8515625, + "learning_rate": 1.2024128825172121e-09, + "loss": 2.0435, + "mean_token_accuracy": 0.7706959706959706, + "mean_token_accuracy_utility": 0.6877411477411476, + "step": 2690 + }, + { + "epoch": 0.9973170506059765, + "grad_norm": 0.91796875, + "learning_rate": 4.0915979321320967e-10, + "loss": 2.1105, + "mean_token_accuracy": 0.7629304029304028, + "mean_token_accuracy_utility": 0.676086691086691, + "step": 2695 + }, + { + "epoch": 0.9991673605328892, + "grad_norm": 0.8828125, + "learning_rate": 3.3401008625588706e-11, + "loss": 2.2387, + "mean_token_accuracy": 0.7707326007326009, + "mean_token_accuracy_utility": 0.654084249084249, + "step": 2700 + }, + { + "epoch": 0.9991673605328892, + "eval_mean_token_accuracy": 0.7546075928428875, + "eval_reasoning_loss": 0.8222787380218506, + "eval_reasoning_runtime": 27.7804, + "eval_reasoning_samples_per_second": 4.644, + "eval_reasoning_steps_per_second": 1.188, + "step": 2700 + }, + { + "epoch": 0.9991673605328892, + "eval_mean_token_accuracy": 0.6620748361669153, + "eval_utility_loss": 1.4264225959777832, + "eval_utility_runtime": 10.4545, + "eval_utility_samples_per_second": 4.687, + "eval_utility_steps_per_second": 1.243, + "step": 2700 + }, + { + "epoch": 0.9999074845036544, + "mean_token_accuracy": 0.7956654456654455, + "mean_token_accuracy_utility": 0.6721916971916972, + "step": 2702, + "total_flos": 0.0, + "train_loss": 2.266656749783226, + "train_runtime": 29910.3238, + "train_samples_per_second": 0.723, + "train_steps_per_second": 0.09 + } + ], + "logging_steps": 5, + "max_steps": 2702, + "num_input_tokens_seen": 0, + "num_train_epochs": 1, + "save_steps": 500, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": false, + "should_training_stop": false + }, + "attributes": {} + } + }, + "total_flos": 0.0, + "train_batch_size": 2, + "trial_name": null, + "trial_params": null +}