{ "adaptation": "LoRA", "dataset_passes": 3.0016, "evaluation_note": "Trajectory PPL uses the fixed online validation sample every 250 updates. Endpoint PPL uses the larger frozen endpoint evaluation and is reported separately.", "model": "Qwen2.5-1.5B", "objective": "continued pretraining", "online_trajectory_last_step": 23250, "regimes": { "degraded": { "arms": { "native_accelerated": { "endpoint": { "gpu_hours": 1.689647, "peak_vram_gib": 5.53625, "retention_ppl": 37.672249, "validation_ppl": 22.347558 }, "label": "Native accelerated", "points": [ { "gpu_hours": 0.027239, "step": 250, "validation_ppl": 12.684407 }, { "gpu_hours": 0.044513, "step": 500, "validation_ppl": 13.78947 }, { "gpu_hours": 0.061749, "step": 750, "validation_ppl": 14.397174 }, { "gpu_hours": 0.078904, "step": 1000, "validation_ppl": 15.064611 }, { "gpu_hours": 0.096193, "step": 1250, "validation_ppl": 15.554332 }, { "gpu_hours": 0.113292, "step": 1500, "validation_ppl": 16.181663 }, { "gpu_hours": 0.130554, "step": 1750, "validation_ppl": 16.402589 }, { "gpu_hours": 0.147902, "step": 2000, "validation_ppl": 16.770689 }, { "gpu_hours": 0.165448, "step": 2250, "validation_ppl": 17.11491 }, { "gpu_hours": 0.182877, "step": 2500, "validation_ppl": 17.325501 }, { "gpu_hours": 0.200164, "step": 2750, "validation_ppl": 17.570781 }, { "gpu_hours": 0.217439, "step": 3000, "validation_ppl": 17.616517 }, { "gpu_hours": 0.234811, "step": 3250, "validation_ppl": 17.819468 }, { "gpu_hours": 0.252025, "step": 3500, "validation_ppl": 18.272793 }, { "gpu_hours": 0.269069, "step": 3750, "validation_ppl": 18.317636 }, { "gpu_hours": 0.286131, "step": 4000, "validation_ppl": 18.519761 }, { "gpu_hours": 0.303465, "step": 4250, "validation_ppl": 18.596418 }, { "gpu_hours": 0.320929, "step": 4500, "validation_ppl": 18.639144 }, { "gpu_hours": 0.338336, "step": 4750, "validation_ppl": 18.624141 }, { "gpu_hours": 0.355767, "step": 5000, "validation_ppl": 18.712513 }, { "gpu_hours": 0.373073, "step": 5250, "validation_ppl": 19.34744 }, { "gpu_hours": 0.390808, "step": 5500, "validation_ppl": 19.097152 }, { "gpu_hours": 0.409124, "step": 5750, "validation_ppl": 19.261673 }, { "gpu_hours": 0.427033, "step": 6000, "validation_ppl": 19.401468 }, { "gpu_hours": 0.445151, "step": 6250, "validation_ppl": 19.180127 }, { "gpu_hours": 0.463167, "step": 6500, "validation_ppl": 19.358813 }, { "gpu_hours": 0.481442, "step": 6750, "validation_ppl": 19.444537 }, { "gpu_hours": 0.499815, "step": 7000, "validation_ppl": 19.390136 }, { "gpu_hours": 0.517721, "step": 7250, "validation_ppl": 19.302207 }, { "gpu_hours": 0.53583, "step": 7500, "validation_ppl": 19.355071 }, { "gpu_hours": 0.553962, "step": 7750, "validation_ppl": 19.714659 }, { "gpu_hours": 0.572006, "step": 8000, "validation_ppl": 19.868105 }, { "gpu_hours": 0.590267, "step": 8250, "validation_ppl": 19.602235 }, { "gpu_hours": 0.608329, "step": 8500, "validation_ppl": 20.04868 }, { "gpu_hours": 0.626209, "step": 8750, "validation_ppl": 19.677626 }, { "gpu_hours": 0.644125, "step": 9000, "validation_ppl": 19.996268 }, { "gpu_hours": 0.662113, "step": 9250, "validation_ppl": 20.076554 }, { "gpu_hours": 0.680126, "step": 9500, "validation_ppl": 20.171967 }, { "gpu_hours": 0.698156, "step": 9750, "validation_ppl": 19.870365 }, { "gpu_hours": 0.716321, "step": 10000, "validation_ppl": 19.947783 }, { "gpu_hours": 0.734287, "step": 10250, "validation_ppl": 19.729529 }, { "gpu_hours": 0.752366, "step": 10500, "validation_ppl": 19.906704 }, { "gpu_hours": 0.770371, "step": 10750, "validation_ppl": 20.039402 }, { "gpu_hours": 0.788556, "step": 11000, "validation_ppl": 20.281648 }, { "gpu_hours": 0.806483, "step": 11250, "validation_ppl": 20.236537 }, { "gpu_hours": 0.82425, "step": 11500, "validation_ppl": 20.215826 }, { "gpu_hours": 0.842254, "step": 11750, "validation_ppl": 20.284947 }, { "gpu_hours": 0.860253, "step": 12000, "validation_ppl": 20.241919 }, { "gpu_hours": 0.8784, "step": 12250, "validation_ppl": 20.388338 }, { "gpu_hours": 0.896558, "step": 12500, "validation_ppl": 20.369685 }, { "gpu_hours": 0.91463, "step": 12750, "validation_ppl": 20.153082 }, { "gpu_hours": 0.932553, "step": 13000, "validation_ppl": 20.211409 }, { "gpu_hours": 0.950404, "step": 13250, "validation_ppl": 20.397701 }, { "gpu_hours": 0.968395, "step": 13500, "validation_ppl": 20.808735 }, { "gpu_hours": 0.986532, "step": 13750, "validation_ppl": 20.344808 }, { "gpu_hours": 1.00464, "step": 14000, "validation_ppl": 20.396145 }, { "gpu_hours": 1.022258, "step": 14250, "validation_ppl": 20.389079 }, { "gpu_hours": 1.039935, "step": 14500, "validation_ppl": 20.521114 }, { "gpu_hours": 1.057862, "step": 14750, "validation_ppl": 20.467688 }, { "gpu_hours": 1.07577, "step": 15000, "validation_ppl": 20.478095 }, { "gpu_hours": 1.094138, "step": 15250, "validation_ppl": 20.52358 }, { "gpu_hours": 1.112049, "step": 15500, "validation_ppl": 20.609683 }, { "gpu_hours": 1.129737, "step": 15750, "validation_ppl": 20.529803 }, { "gpu_hours": 1.147621, "step": 16000, "validation_ppl": 21.365394 }, { "gpu_hours": 1.165667, "step": 16250, "validation_ppl": 20.536154 }, { "gpu_hours": 1.183713, "step": 16500, "validation_ppl": 20.617442 }, { "gpu_hours": 1.201831, "step": 16750, "validation_ppl": 20.524765 }, { "gpu_hours": 1.219627, "step": 17000, "validation_ppl": 20.34999 }, { "gpu_hours": 1.237551, "step": 17250, "validation_ppl": 20.834304 }, { "gpu_hours": 1.255426, "step": 17500, "validation_ppl": 20.556917 }, { "gpu_hours": 1.273445, "step": 17750, "validation_ppl": 20.431766 }, { "gpu_hours": 1.291649, "step": 18000, "validation_ppl": 20.413949 }, { "gpu_hours": 1.309632, "step": 18250, "validation_ppl": 21.004487 }, { "gpu_hours": 1.327551, "step": 18500, "validation_ppl": 20.673769 }, { "gpu_hours": 1.345419, "step": 18750, "validation_ppl": 20.448295 }, { "gpu_hours": 1.363379, "step": 19000, "validation_ppl": 20.634294 }, { "gpu_hours": 1.381534, "step": 19250, "validation_ppl": 20.413692 }, { "gpu_hours": 1.399526, "step": 19500, "validation_ppl": 20.736514 }, { "gpu_hours": 1.417447, "step": 19750, "validation_ppl": 20.896089 }, { "gpu_hours": 1.435331, "step": 20000, "validation_ppl": 20.717532 }, { "gpu_hours": 1.45328, "step": 20250, "validation_ppl": 20.749096 }, { "gpu_hours": 1.471279, "step": 20500, "validation_ppl": 20.28023 }, { "gpu_hours": 1.489658, "step": 20750, "validation_ppl": 20.448973 }, { "gpu_hours": 1.50782, "step": 21000, "validation_ppl": 20.532413 }, { "gpu_hours": 1.525625, "step": 21250, "validation_ppl": 20.265847 }, { "gpu_hours": 1.543862, "step": 21500, "validation_ppl": 20.88415 }, { "gpu_hours": 1.561664, "step": 21750, "validation_ppl": 20.500609 }, { "gpu_hours": 1.579793, "step": 22000, "validation_ppl": 20.291007 }, { "gpu_hours": 1.597905, "step": 22250, "validation_ppl": 20.728187 }, { "gpu_hours": 1.615712, "step": 22500, "validation_ppl": 20.574111 }, { "gpu_hours": 1.633446, "step": 22750, "validation_ppl": 20.747238 }, { "gpu_hours": 1.651404, "step": 23000, "validation_ppl": 20.760644 }, { "gpu_hours": 1.669414, "step": 23250, "validation_ppl": 20.497021 } ] }, "native_eager": { "endpoint": { "gpu_hours": 3.852763, "peak_vram_gib": 5.625956, "retention_ppl": 37.082547, "validation_ppl": 22.215321 }, "label": "Native eager", "points": [ { "gpu_hours": 0.04463, "step": 250, "validation_ppl": 12.662124 }, { "gpu_hours": 0.082635, "step": 500, "validation_ppl": 13.759752 }, { "gpu_hours": 0.121217, "step": 750, "validation_ppl": 14.233356 }, { "gpu_hours": 0.158641, "step": 1000, "validation_ppl": 15.003746 }, { "gpu_hours": 0.197142, "step": 1250, "validation_ppl": 15.514076 }, { "gpu_hours": 0.233844, "step": 1500, "validation_ppl": 15.984854 }, { "gpu_hours": 0.273379, "step": 1750, "validation_ppl": 16.398714 }, { "gpu_hours": 0.315766, "step": 2000, "validation_ppl": 16.831655 }, { "gpu_hours": 0.357706, "step": 2250, "validation_ppl": 17.328139 }, { "gpu_hours": 0.399706, "step": 2500, "validation_ppl": 17.574231 }, { "gpu_hours": 0.441646, "step": 2750, "validation_ppl": 18.200601 }, { "gpu_hours": 0.483496, "step": 3000, "validation_ppl": 18.024774 }, { "gpu_hours": 0.522097, "step": 3250, "validation_ppl": 17.902093 }, { "gpu_hours": 0.558705, "step": 3500, "validation_ppl": 18.220487 }, { "gpu_hours": 0.600481, "step": 3750, "validation_ppl": 18.214473 }, { "gpu_hours": 0.641237, "step": 4000, "validation_ppl": 18.524194 }, { "gpu_hours": 0.679614, "step": 4250, "validation_ppl": 18.40288 }, { "gpu_hours": 0.712373, "step": 4500, "validation_ppl": 18.755113 }, { "gpu_hours": 0.751009, "step": 4750, "validation_ppl": 18.81934 }, { "gpu_hours": 0.79453, "step": 5000, "validation_ppl": 18.734366 }, { "gpu_hours": 0.836955, "step": 5250, "validation_ppl": 19.096413 }, { "gpu_hours": 0.879345, "step": 5500, "validation_ppl": 18.922341 }, { "gpu_hours": 0.921617, "step": 5750, "validation_ppl": 19.098148 }, { "gpu_hours": 0.964274, "step": 6000, "validation_ppl": 18.961285 }, { "gpu_hours": 1.007208, "step": 6250, "validation_ppl": 19.181751 }, { "gpu_hours": 1.04934, "step": 6500, "validation_ppl": 19.426248 }, { "gpu_hours": 1.091805, "step": 6750, "validation_ppl": 19.333078 }, { "gpu_hours": 1.133341, "step": 7000, "validation_ppl": 19.510162 }, { "gpu_hours": 1.174872, "step": 7250, "validation_ppl": 19.681569 }, { "gpu_hours": 1.216021, "step": 7500, "validation_ppl": 19.316248 }, { "gpu_hours": 1.257293, "step": 7750, "validation_ppl": 19.723232 }, { "gpu_hours": 1.299029, "step": 8000, "validation_ppl": 19.859849 }, { "gpu_hours": 1.340183, "step": 8250, "validation_ppl": 19.687243 }, { "gpu_hours": 1.3816, "step": 8500, "validation_ppl": 20.049969 }, { "gpu_hours": 1.423122, "step": 8750, "validation_ppl": 20.067478 }, { "gpu_hours": 1.4653, "step": 9000, "validation_ppl": 20.166911 }, { "gpu_hours": 1.507664, "step": 9250, "validation_ppl": 20.243791 }, { "gpu_hours": 1.550056, "step": 9500, "validation_ppl": 20.217506 }, { "gpu_hours": 1.592691, "step": 9750, "validation_ppl": 20.001642 }, { "gpu_hours": 1.635308, "step": 10000, "validation_ppl": 20.063002 }, { "gpu_hours": 1.677378, "step": 10250, "validation_ppl": 19.618557 }, { "gpu_hours": 1.719978, "step": 10500, "validation_ppl": 19.860132 }, { "gpu_hours": 1.762328, "step": 10750, "validation_ppl": 20.011102 }, { "gpu_hours": 1.80424, "step": 11000, "validation_ppl": 19.959388 }, { "gpu_hours": 1.845746, "step": 11250, "validation_ppl": 20.252553 }, { "gpu_hours": 1.888245, "step": 11500, "validation_ppl": 20.001442 }, { "gpu_hours": 1.929543, "step": 11750, "validation_ppl": 20.074495 }, { "gpu_hours": 1.973498, "step": 12000, "validation_ppl": 20.033874 }, { "gpu_hours": 2.016707, "step": 12250, "validation_ppl": 20.055302 }, { "gpu_hours": 2.053229, "step": 12500, "validation_ppl": 20.25794 }, { "gpu_hours": 2.089294, "step": 12750, "validation_ppl": 20.117372 }, { "gpu_hours": 2.125506, "step": 13000, "validation_ppl": 20.272024 }, { "gpu_hours": 2.166684, "step": 13250, "validation_ppl": 21.035977 }, { "gpu_hours": 2.208256, "step": 13500, "validation_ppl": 20.577014 }, { "gpu_hours": 2.250296, "step": 13750, "validation_ppl": 20.162012 }, { "gpu_hours": 2.292802, "step": 14000, "validation_ppl": 20.398508 }, { "gpu_hours": 2.33489, "step": 14250, "validation_ppl": 20.448372 }, { "gpu_hours": 2.376763, "step": 14500, "validation_ppl": 20.313053 }, { "gpu_hours": 2.418475, "step": 14750, "validation_ppl": 20.377013 }, { "gpu_hours": 2.459666, "step": 15000, "validation_ppl": 20.358141 }, { "gpu_hours": 2.501448, "step": 15250, "validation_ppl": 20.214246 }, { "gpu_hours": 2.540899, "step": 15500, "validation_ppl": 20.496261 }, { "gpu_hours": 2.582493, "step": 15750, "validation_ppl": 20.462521 }, { "gpu_hours": 2.622248, "step": 16000, "validation_ppl": 20.429465 }, { "gpu_hours": 2.662926, "step": 16250, "validation_ppl": 20.614523 }, { "gpu_hours": 2.704275, "step": 16500, "validation_ppl": 20.457296 }, { "gpu_hours": 2.745526, "step": 16750, "validation_ppl": 20.45664 }, { "gpu_hours": 2.787008, "step": 17000, "validation_ppl": 20.57667 }, { "gpu_hours": 2.828085, "step": 17250, "validation_ppl": 20.40175 }, { "gpu_hours": 2.870026, "step": 17500, "validation_ppl": 20.868157 }, { "gpu_hours": 2.911285, "step": 17750, "validation_ppl": 20.653816 }, { "gpu_hours": 2.951985, "step": 18000, "validation_ppl": 20.472704 }, { "gpu_hours": 2.99349, "step": 18250, "validation_ppl": 20.799289 }, { "gpu_hours": 3.034186, "step": 18500, "validation_ppl": 20.874356 }, { "gpu_hours": 3.075583, "step": 18750, "validation_ppl": 20.541666 }, { "gpu_hours": 3.116712, "step": 19000, "validation_ppl": 21.031225 }, { "gpu_hours": 3.158044, "step": 19250, "validation_ppl": 21.450072 }, { "gpu_hours": 3.199401, "step": 19500, "validation_ppl": 20.698723 }, { "gpu_hours": 3.240692, "step": 19750, "validation_ppl": 21.27471 }, { "gpu_hours": 3.282515, "step": 20000, "validation_ppl": 21.431734 }, { "gpu_hours": 3.323201, "step": 20250, "validation_ppl": 21.240055 }, { "gpu_hours": 3.364642, "step": 20500, "validation_ppl": 20.68268 }, { "gpu_hours": 3.405702, "step": 20750, "validation_ppl": 20.829397 }, { "gpu_hours": 3.447065, "step": 21000, "validation_ppl": 20.587428 }, { "gpu_hours": 3.488452, "step": 21250, "validation_ppl": 20.520114 }, { "gpu_hours": 3.529063, "step": 21500, "validation_ppl": 20.4676 }, { "gpu_hours": 3.569803, "step": 21750, "validation_ppl": 20.454159 }, { "gpu_hours": 3.610621, "step": 22000, "validation_ppl": 20.656367 }, { "gpu_hours": 3.651617, "step": 22250, "validation_ppl": 20.518236 }, { "gpu_hours": 3.692078, "step": 22500, "validation_ppl": 20.616375 }, { "gpu_hours": 3.733134, "step": 22750, "validation_ppl": 20.636457 }, { "gpu_hours": 3.773618, "step": 23000, "validation_ppl": 20.713748 }, { "gpu_hours": 3.814621, "step": 23250, "validation_ppl": 20.688717 } ] }, "optimizer_plus": { "endpoint": { "gpu_hours": 1.354574, "peak_vram_gib": 5.706615, "retention_ppl": 22.765814, "validation_ppl": 14.154056 }, "label": "Optimizer+", "points": [ { "gpu_hours": 0.042656, "step": 250, "validation_ppl": 12.023211 }, { "gpu_hours": 0.056705, "step": 500, "validation_ppl": 11.870104 }, { "gpu_hours": 0.070696, "step": 750, "validation_ppl": 11.792682 }, { "gpu_hours": 0.084834, "step": 1000, "validation_ppl": 11.766762 }, { "gpu_hours": 0.09885, "step": 1250, "validation_ppl": 11.730245 }, { "gpu_hours": 0.113, "step": 1500, "validation_ppl": 11.735504 }, { "gpu_hours": 0.127074, "step": 1750, "validation_ppl": 11.686048 }, { "gpu_hours": 0.141241, "step": 2000, "validation_ppl": 11.694711 }, { "gpu_hours": 0.155297, "step": 2250, "validation_ppl": 11.688396 }, { "gpu_hours": 0.169443, "step": 2500, "validation_ppl": 11.67965 }, { "gpu_hours": 0.183559, "step": 2750, "validation_ppl": 11.659471 }, { "gpu_hours": 0.197486, "step": 3000, "validation_ppl": 11.666934 }, { "gpu_hours": 0.211469, "step": 3250, "validation_ppl": 11.642633 }, { "gpu_hours": 0.225427, "step": 3500, "validation_ppl": 11.656295 }, { "gpu_hours": 0.23947, "step": 3750, "validation_ppl": 11.676142 }, { "gpu_hours": 0.253403, "step": 4000, "validation_ppl": 11.670709 }, { "gpu_hours": 0.267377, "step": 4250, "validation_ppl": 11.681838 }, { "gpu_hours": 0.28127, "step": 4500, "validation_ppl": 11.675106 }, { "gpu_hours": 0.295347, "step": 4750, "validation_ppl": 11.651926 }, { "gpu_hours": 0.309372, "step": 5000, "validation_ppl": 11.637979 }, { "gpu_hours": 0.323473, "step": 5250, "validation_ppl": 11.640443 }, { "gpu_hours": 0.337398, "step": 5500, "validation_ppl": 11.638485 }, { "gpu_hours": 0.351397, "step": 5750, "validation_ppl": 11.629174 }, { "gpu_hours": 0.365359, "step": 6000, "validation_ppl": 11.651343 }, { "gpu_hours": 0.379475, "step": 6250, "validation_ppl": 11.637018 }, { "gpu_hours": 0.393511, "step": 6500, "validation_ppl": 11.642292 }, { "gpu_hours": 0.407554, "step": 6750, "validation_ppl": 11.642163 }, { "gpu_hours": 0.421639, "step": 7000, "validation_ppl": 11.63388 }, { "gpu_hours": 0.435657, "step": 7250, "validation_ppl": 11.632932 }, { "gpu_hours": 0.449723, "step": 7500, "validation_ppl": 11.633925 }, { "gpu_hours": 0.463655, "step": 7750, "validation_ppl": 11.648548 }, { "gpu_hours": 0.477756, "step": 8000, "validation_ppl": 11.769839 }, { "gpu_hours": 0.491779, "step": 8250, "validation_ppl": 11.828709 }, { "gpu_hours": 0.50602, "step": 8500, "validation_ppl": 11.887996 }, { "gpu_hours": 0.520033, "step": 8750, "validation_ppl": 11.936734 }, { "gpu_hours": 0.534207, "step": 9000, "validation_ppl": 11.938894 }, { "gpu_hours": 0.548301, "step": 9250, "validation_ppl": 11.954633 }, { "gpu_hours": 0.562442, "step": 9500, "validation_ppl": 11.990864 }, { "gpu_hours": 0.576619, "step": 9750, "validation_ppl": 11.99845 }, { "gpu_hours": 0.590663, "step": 10000, "validation_ppl": 11.991966 }, { "gpu_hours": 0.604837, "step": 10250, "validation_ppl": 12.008301 }, { "gpu_hours": 0.618942, "step": 10500, "validation_ppl": 12.032106 }, { "gpu_hours": 0.633123, "step": 10750, "validation_ppl": 12.06095 }, { "gpu_hours": 0.647177, "step": 11000, "validation_ppl": 12.092788 }, { "gpu_hours": 0.661333, "step": 11250, "validation_ppl": 12.078376 }, { "gpu_hours": 0.675463, "step": 11500, "validation_ppl": 12.110392 }, { "gpu_hours": 0.689628, "step": 11750, "validation_ppl": 12.083847 }, { "gpu_hours": 0.703715, "step": 12000, "validation_ppl": 12.118003 }, { "gpu_hours": 0.717831, "step": 12250, "validation_ppl": 12.154415 }, { "gpu_hours": 0.731912, "step": 12500, "validation_ppl": 12.124527 }, { "gpu_hours": 0.746111, "step": 12750, "validation_ppl": 12.133445 }, { "gpu_hours": 0.760266, "step": 13000, "validation_ppl": 12.15903 }, { "gpu_hours": 0.774359, "step": 13250, "validation_ppl": 12.175813 }, { "gpu_hours": 0.788523, "step": 13500, "validation_ppl": 12.171542 }, { "gpu_hours": 0.802623, "step": 13750, "validation_ppl": 12.148186 }, { "gpu_hours": 0.816797, "step": 14000, "validation_ppl": 12.176939 }, { "gpu_hours": 0.830864, "step": 14250, "validation_ppl": 12.162336 }, { "gpu_hours": 0.845039, "step": 14500, "validation_ppl": 12.178179 }, { "gpu_hours": 0.858874, "step": 14750, "validation_ppl": 12.178017 }, { "gpu_hours": 0.873092, "step": 15000, "validation_ppl": 12.194117 }, { "gpu_hours": 0.887138, "step": 15250, "validation_ppl": 12.173459 }, { "gpu_hours": 0.901328, "step": 15500, "validation_ppl": 12.157871 }, { "gpu_hours": 0.915378, "step": 15750, "validation_ppl": 12.813738 }, { "gpu_hours": 0.929563, "step": 16000, "validation_ppl": 12.839028 }, { "gpu_hours": 0.943736, "step": 16250, "validation_ppl": 12.955758 }, { "gpu_hours": 0.95775, "step": 16500, "validation_ppl": 12.985218 }, { "gpu_hours": 0.971922, "step": 16750, "validation_ppl": 13.033136 }, { "gpu_hours": 0.985977, "step": 17000, "validation_ppl": 13.03194 }, { "gpu_hours": 1.000137, "step": 17250, "validation_ppl": 13.126797 }, { "gpu_hours": 1.01419, "step": 17500, "validation_ppl": 13.085913 }, { "gpu_hours": 1.028352, "step": 17750, "validation_ppl": 13.171921 }, { "gpu_hours": 1.042398, "step": 18000, "validation_ppl": 13.169214 }, { "gpu_hours": 1.056537, "step": 18250, "validation_ppl": 13.085237 }, { "gpu_hours": 1.070632, "step": 18500, "validation_ppl": 13.172105 }, { "gpu_hours": 1.084809, "step": 18750, "validation_ppl": 13.187941 }, { "gpu_hours": 1.098896, "step": 19000, "validation_ppl": 13.123476 }, { "gpu_hours": 1.113095, "step": 19250, "validation_ppl": 13.166942 }, { "gpu_hours": 1.12729, "step": 19500, "validation_ppl": 13.174437 }, { "gpu_hours": 1.141397, "step": 19750, "validation_ppl": 13.24014 }, { "gpu_hours": 1.15553, "step": 20000, "validation_ppl": 13.188531 }, { "gpu_hours": 1.169654, "step": 20250, "validation_ppl": 13.190211 }, { "gpu_hours": 1.183818, "step": 20500, "validation_ppl": 13.155586 }, { "gpu_hours": 1.197884, "step": 20750, "validation_ppl": 13.226564 }, { "gpu_hours": 1.212089, "step": 21000, "validation_ppl": 13.222988 }, { "gpu_hours": 1.226126, "step": 21250, "validation_ppl": 13.177869 }, { "gpu_hours": 1.240308, "step": 21500, "validation_ppl": 13.164353 }, { "gpu_hours": 1.254334, "step": 21750, "validation_ppl": 13.156128 }, { "gpu_hours": 1.268451, "step": 22000, "validation_ppl": 13.200795 }, { "gpu_hours": 1.282381, "step": 22250, "validation_ppl": 13.182933 }, { "gpu_hours": 1.296374, "step": 22500, "validation_ppl": 13.182205 }, { "gpu_hours": 1.310393, "step": 22750, "validation_ppl": 13.175115 }, { "gpu_hours": 1.324388, "step": 23000, "validation_ppl": 13.205885 }, { "gpu_hours": 1.338468, "step": 23250, "validation_ppl": 13.213687 } ] } }, "learning_rate": "1e-3" }, "healthy": { "arms": { "native_accelerated": { "endpoint": { "gpu_hours": 1.664412, "peak_vram_gib": 5.53625, "retention_ppl": 19.668349, "validation_ppl": 12.478814 }, "label": "Native accelerated", "points": [ { "gpu_hours": 0.025985, "step": 250, "validation_ppl": 11.405286 }, { "gpu_hours": 0.042717, "step": 500, "validation_ppl": 11.404767 }, { "gpu_hours": 0.059765, "step": 750, "validation_ppl": 11.393016 }, { "gpu_hours": 0.076914, "step": 1000, "validation_ppl": 11.400139 }, { "gpu_hours": 0.094258, "step": 1250, "validation_ppl": 11.401661 }, { "gpu_hours": 0.111431, "step": 1500, "validation_ppl": 11.409783 }, { "gpu_hours": 0.128663, "step": 1750, "validation_ppl": 11.394039 }, { "gpu_hours": 0.146155, "step": 2000, "validation_ppl": 11.402828 }, { "gpu_hours": 0.163704, "step": 2250, "validation_ppl": 11.398439 }, { "gpu_hours": 0.180987, "step": 2500, "validation_ppl": 11.400459 }, { "gpu_hours": 0.198307, "step": 2750, "validation_ppl": 11.393644 }, { "gpu_hours": 0.215532, "step": 3000, "validation_ppl": 11.404609 }, { "gpu_hours": 0.232708, "step": 3250, "validation_ppl": 11.398461 }, { "gpu_hours": 0.249977, "step": 3500, "validation_ppl": 11.405359 }, { "gpu_hours": 0.267239, "step": 3750, "validation_ppl": 11.406715 }, { "gpu_hours": 0.284685, "step": 4000, "validation_ppl": 11.407547 }, { "gpu_hours": 0.301741, "step": 4250, "validation_ppl": 11.402286 }, { "gpu_hours": 0.319044, "step": 4500, "validation_ppl": 11.417902 }, { "gpu_hours": 0.336261, "step": 4750, "validation_ppl": 11.415966 }, { "gpu_hours": 0.353653, "step": 5000, "validation_ppl": 11.407485 }, { "gpu_hours": 0.370878, "step": 5250, "validation_ppl": 11.406973 }, { "gpu_hours": 0.388031, "step": 5500, "validation_ppl": 11.414854 }, { "gpu_hours": 0.405234, "step": 5750, "validation_ppl": 11.41349 }, { "gpu_hours": 0.422431, "step": 6000, "validation_ppl": 11.422962 }, { "gpu_hours": 0.439555, "step": 6250, "validation_ppl": 11.423879 }, { "gpu_hours": 0.456305, "step": 6500, "validation_ppl": 11.42896 }, { "gpu_hours": 0.472995, "step": 6750, "validation_ppl": 11.424676 }, { "gpu_hours": 0.489855, "step": 7000, "validation_ppl": 11.4208 }, { "gpu_hours": 0.50686, "step": 7250, "validation_ppl": 11.417315 }, { "gpu_hours": 0.523914, "step": 7500, "validation_ppl": 11.419178 }, { "gpu_hours": 0.540798, "step": 7750, "validation_ppl": 11.423064 }, { "gpu_hours": 0.557658, "step": 8000, "validation_ppl": 11.445744 }, { "gpu_hours": 0.57474, "step": 8250, "validation_ppl": 11.46499 }, { "gpu_hours": 0.591518, "step": 8500, "validation_ppl": 11.492891 }, { "gpu_hours": 0.608506, "step": 8750, "validation_ppl": 11.51176 }, { "gpu_hours": 0.625432, "step": 9000, "validation_ppl": 11.520939 }, { "gpu_hours": 0.64241, "step": 9250, "validation_ppl": 11.521744 }, { "gpu_hours": 0.659387, "step": 9500, "validation_ppl": 11.534695 }, { "gpu_hours": 0.676328, "step": 9750, "validation_ppl": 11.553262 }, { "gpu_hours": 0.693371, "step": 10000, "validation_ppl": 11.554811 }, { "gpu_hours": 0.710342, "step": 10250, "validation_ppl": 11.557776 }, { "gpu_hours": 0.727426, "step": 10500, "validation_ppl": 11.5771 }, { "gpu_hours": 0.744347, "step": 10750, "validation_ppl": 11.567237 }, { "gpu_hours": 0.761291, "step": 11000, "validation_ppl": 11.581565 }, { "gpu_hours": 0.778923, "step": 11250, "validation_ppl": 11.575882 }, { "gpu_hours": 0.796466, "step": 11500, "validation_ppl": 11.592805 }, { "gpu_hours": 0.814358, "step": 11750, "validation_ppl": 11.588739 }, { "gpu_hours": 0.83199, "step": 12000, "validation_ppl": 11.59114 }, { "gpu_hours": 0.849589, "step": 12250, "validation_ppl": 11.587696 }, { "gpu_hours": 0.867297, "step": 12500, "validation_ppl": 11.592596 }, { "gpu_hours": 0.884959, "step": 12750, "validation_ppl": 11.593775 }, { "gpu_hours": 0.902646, "step": 13000, "validation_ppl": 11.598191 }, { "gpu_hours": 0.920183, "step": 13250, "validation_ppl": 11.608364 }, { "gpu_hours": 0.937891, "step": 13500, "validation_ppl": 11.602494 }, { "gpu_hours": 0.955623, "step": 13750, "validation_ppl": 11.617823 }, { "gpu_hours": 0.973324, "step": 14000, "validation_ppl": 11.617914 }, { "gpu_hours": 0.991186, "step": 14250, "validation_ppl": 11.612829 }, { "gpu_hours": 1.008921, "step": 14500, "validation_ppl": 11.618496 }, { "gpu_hours": 1.026853, "step": 14750, "validation_ppl": 11.614331 }, { "gpu_hours": 1.04475, "step": 15000, "validation_ppl": 11.627494 }, { "gpu_hours": 1.062779, "step": 15250, "validation_ppl": 11.618778 }, { "gpu_hours": 1.081014, "step": 15500, "validation_ppl": 11.601528 }, { "gpu_hours": 1.099014, "step": 15750, "validation_ppl": 11.778979 }, { "gpu_hours": 1.117117, "step": 16000, "validation_ppl": 11.816577 }, { "gpu_hours": 1.13529, "step": 16250, "validation_ppl": 11.850641 }, { "gpu_hours": 1.153472, "step": 16500, "validation_ppl": 11.874936 }, { "gpu_hours": 1.171759, "step": 16750, "validation_ppl": 11.899977 }, { "gpu_hours": 1.189877, "step": 17000, "validation_ppl": 11.922784 }, { "gpu_hours": 1.208044, "step": 17250, "validation_ppl": 11.931353 }, { "gpu_hours": 1.22603, "step": 17500, "validation_ppl": 11.936193 }, { "gpu_hours": 1.243946, "step": 17750, "validation_ppl": 11.983799 }, { "gpu_hours": 1.262188, "step": 18000, "validation_ppl": 11.953 }, { "gpu_hours": 1.280368, "step": 18250, "validation_ppl": 11.950041 }, { "gpu_hours": 1.298651, "step": 18500, "validation_ppl": 11.986956 }, { "gpu_hours": 1.316888, "step": 18750, "validation_ppl": 11.987243 }, { "gpu_hours": 1.335054, "step": 19000, "validation_ppl": 11.999921 }, { "gpu_hours": 1.353375, "step": 19250, "validation_ppl": 11.994513 }, { "gpu_hours": 1.371733, "step": 19500, "validation_ppl": 11.999788 }, { "gpu_hours": 1.390225, "step": 19750, "validation_ppl": 11.991851 }, { "gpu_hours": 1.408395, "step": 20000, "validation_ppl": 12.010345 }, { "gpu_hours": 1.42657, "step": 20250, "validation_ppl": 12.034359 }, { "gpu_hours": 1.444715, "step": 20500, "validation_ppl": 12.035014 }, { "gpu_hours": 1.462672, "step": 20750, "validation_ppl": 12.038767 }, { "gpu_hours": 1.480752, "step": 21000, "validation_ppl": 12.049597 }, { "gpu_hours": 1.498864, "step": 21250, "validation_ppl": 12.048206 }, { "gpu_hours": 1.516982, "step": 21500, "validation_ppl": 12.039251 }, { "gpu_hours": 1.535108, "step": 21750, "validation_ppl": 12.052251 }, { "gpu_hours": 1.553329, "step": 22000, "validation_ppl": 12.055254 }, { "gpu_hours": 1.571585, "step": 22250, "validation_ppl": 12.047669 }, { "gpu_hours": 1.589728, "step": 22500, "validation_ppl": 12.050624 }, { "gpu_hours": 1.607983, "step": 22750, "validation_ppl": 12.051797 }, { "gpu_hours": 1.626122, "step": 23000, "validation_ppl": 12.073673 }, { "gpu_hours": 1.644205, "step": 23250, "validation_ppl": 12.091457 } ] }, "native_eager": { "endpoint": { "gpu_hours": 3.539082, "peak_vram_gib": 5.625956, "retention_ppl": 19.694716, "validation_ppl": 12.481506 }, "label": "Native eager", "points": [ { "gpu_hours": 0.042535, "step": 250, "validation_ppl": 11.406957 }, { "gpu_hours": 0.075121, "step": 500, "validation_ppl": 11.405866 }, { "gpu_hours": 0.114032, "step": 750, "validation_ppl": 11.392941 }, { "gpu_hours": 0.152329, "step": 1000, "validation_ppl": 11.406284 }, { "gpu_hours": 0.191237, "step": 1250, "validation_ppl": 11.404003 }, { "gpu_hours": 0.229826, "step": 1500, "validation_ppl": 11.410226 }, { "gpu_hours": 0.267585, "step": 1750, "validation_ppl": 11.396181 }, { "gpu_hours": 0.305976, "step": 2000, "validation_ppl": 11.403244 }, { "gpu_hours": 0.343801, "step": 2250, "validation_ppl": 11.401986 }, { "gpu_hours": 0.382496, "step": 2500, "validation_ppl": 11.400716 }, { "gpu_hours": 0.417951, "step": 2750, "validation_ppl": 11.393522 }, { "gpu_hours": 0.446633, "step": 3000, "validation_ppl": 11.403588 }, { "gpu_hours": 0.479024, "step": 3250, "validation_ppl": 11.397658 }, { "gpu_hours": 0.508164, "step": 3500, "validation_ppl": 11.402211 }, { "gpu_hours": 0.540673, "step": 3750, "validation_ppl": 11.406869 }, { "gpu_hours": 0.568321, "step": 4000, "validation_ppl": 11.406361 }, { "gpu_hours": 0.596531, "step": 4250, "validation_ppl": 11.408427 }, { "gpu_hours": 0.635679, "step": 4500, "validation_ppl": 11.41644 }, { "gpu_hours": 0.674375, "step": 4750, "validation_ppl": 11.416002 }, { "gpu_hours": 0.712524, "step": 5000, "validation_ppl": 11.408753 }, { "gpu_hours": 0.751543, "step": 5250, "validation_ppl": 11.409139 }, { "gpu_hours": 0.788284, "step": 5500, "validation_ppl": 11.414413 }, { "gpu_hours": 0.819773, "step": 5750, "validation_ppl": 11.415753 }, { "gpu_hours": 0.851374, "step": 6000, "validation_ppl": 11.423701 }, { "gpu_hours": 0.887717, "step": 6250, "validation_ppl": 11.425467 }, { "gpu_hours": 0.918523, "step": 6500, "validation_ppl": 11.424127 }, { "gpu_hours": 0.946875, "step": 6750, "validation_ppl": 11.424485 }, { "gpu_hours": 0.977023, "step": 7000, "validation_ppl": 11.41904 }, { "gpu_hours": 1.016523, "step": 7250, "validation_ppl": 11.426081 }, { "gpu_hours": 1.056978, "step": 7500, "validation_ppl": 11.420957 }, { "gpu_hours": 1.09528, "step": 7750, "validation_ppl": 11.415942 }, { "gpu_hours": 1.127246, "step": 8000, "validation_ppl": 11.442775 }, { "gpu_hours": 1.168018, "step": 8250, "validation_ppl": 11.476139 }, { "gpu_hours": 1.207375, "step": 8500, "validation_ppl": 11.494482 }, { "gpu_hours": 1.24629, "step": 8750, "validation_ppl": 11.508531 }, { "gpu_hours": 1.284336, "step": 9000, "validation_ppl": 11.525347 }, { "gpu_hours": 1.323057, "step": 9250, "validation_ppl": 11.524089 }, { "gpu_hours": 1.362104, "step": 9500, "validation_ppl": 11.540688 }, { "gpu_hours": 1.401175, "step": 9750, "validation_ppl": 11.552882 }, { "gpu_hours": 1.440188, "step": 10000, "validation_ppl": 11.556206 }, { "gpu_hours": 1.478591, "step": 10250, "validation_ppl": 11.56587 }, { "gpu_hours": 1.517242, "step": 10500, "validation_ppl": 11.577958 }, { "gpu_hours": 1.554551, "step": 10750, "validation_ppl": 11.5708 }, { "gpu_hours": 1.592123, "step": 11000, "validation_ppl": 11.58701 }, { "gpu_hours": 1.629886, "step": 11250, "validation_ppl": 11.577134 }, { "gpu_hours": 1.667978, "step": 11500, "validation_ppl": 11.595841 }, { "gpu_hours": 1.70694, "step": 11750, "validation_ppl": 11.59246 }, { "gpu_hours": 1.74576, "step": 12000, "validation_ppl": 11.591029 }, { "gpu_hours": 1.784626, "step": 12250, "validation_ppl": 11.59155 }, { "gpu_hours": 1.823614, "step": 12500, "validation_ppl": 11.595956 }, { "gpu_hours": 1.86238, "step": 12750, "validation_ppl": 11.603462 }, { "gpu_hours": 1.901757, "step": 13000, "validation_ppl": 11.594383 }, { "gpu_hours": 1.940208, "step": 13250, "validation_ppl": 11.607962 }, { "gpu_hours": 1.979155, "step": 13500, "validation_ppl": 11.605093 }, { "gpu_hours": 2.01837, "step": 13750, "validation_ppl": 11.621323 }, { "gpu_hours": 2.057721, "step": 14000, "validation_ppl": 11.618689 }, { "gpu_hours": 2.096768, "step": 14250, "validation_ppl": 11.620716 }, { "gpu_hours": 2.135866, "step": 14500, "validation_ppl": 11.622663 }, { "gpu_hours": 2.175353, "step": 14750, "validation_ppl": 11.615415 }, { "gpu_hours": 2.214093, "step": 15000, "validation_ppl": 11.628441 }, { "gpu_hours": 2.252209, "step": 15250, "validation_ppl": 11.62299 }, { "gpu_hours": 2.291181, "step": 15500, "validation_ppl": 11.608507 }, { "gpu_hours": 2.330375, "step": 15750, "validation_ppl": 11.784691 }, { "gpu_hours": 2.369579, "step": 16000, "validation_ppl": 11.825221 }, { "gpu_hours": 2.408203, "step": 16250, "validation_ppl": 11.853606 }, { "gpu_hours": 2.447524, "step": 16500, "validation_ppl": 11.882178 }, { "gpu_hours": 2.48674, "step": 16750, "validation_ppl": 11.901029 }, { "gpu_hours": 2.526041, "step": 17000, "validation_ppl": 11.923916 }, { "gpu_hours": 2.565156, "step": 17250, "validation_ppl": 11.94451 }, { "gpu_hours": 2.604236, "step": 17500, "validation_ppl": 11.945364 }, { "gpu_hours": 2.64312, "step": 17750, "validation_ppl": 11.986219 }, { "gpu_hours": 2.681683, "step": 18000, "validation_ppl": 11.957776 }, { "gpu_hours": 2.720618, "step": 18250, "validation_ppl": 11.956991 }, { "gpu_hours": 2.759884, "step": 18500, "validation_ppl": 11.99415 }, { "gpu_hours": 2.798836, "step": 18750, "validation_ppl": 11.993971 }, { "gpu_hours": 2.838045, "step": 19000, "validation_ppl": 12.001953 }, { "gpu_hours": 2.876437, "step": 19250, "validation_ppl": 11.994161 }, { "gpu_hours": 2.913642, "step": 19500, "validation_ppl": 12.003353 }, { "gpu_hours": 2.953209, "step": 19750, "validation_ppl": 11.995229 }, { "gpu_hours": 2.992333, "step": 20000, "validation_ppl": 12.012382 }, { "gpu_hours": 3.031505, "step": 20250, "validation_ppl": 12.027163 }, { "gpu_hours": 3.070625, "step": 20500, "validation_ppl": 12.036336 }, { "gpu_hours": 3.109015, "step": 20750, "validation_ppl": 12.04682 }, { "gpu_hours": 3.148201, "step": 21000, "validation_ppl": 12.053556 }, { "gpu_hours": 3.187708, "step": 21250, "validation_ppl": 12.050111 }, { "gpu_hours": 3.226804, "step": 21500, "validation_ppl": 12.042012 }, { "gpu_hours": 3.265293, "step": 21750, "validation_ppl": 12.05537 }, { "gpu_hours": 3.304241, "step": 22000, "validation_ppl": 12.057427 }, { "gpu_hours": 3.342703, "step": 22250, "validation_ppl": 12.05373 }, { "gpu_hours": 3.382811, "step": 22500, "validation_ppl": 12.052097 }, { "gpu_hours": 3.422955, "step": 22750, "validation_ppl": 12.056641 }, { "gpu_hours": 3.462377, "step": 23000, "validation_ppl": 12.07983 }, { "gpu_hours": 3.502233, "step": 23250, "validation_ppl": 12.09381 } ] }, "optimizer_plus": { "endpoint": { "gpu_hours": 1.364616, "peak_vram_gib": 5.706615, "retention_ppl": 19.713538, "validation_ppl": 12.482374 }, "label": "Optimizer+", "points": [ { "gpu_hours": 0.043009, "step": 250, "validation_ppl": 11.407138 }, { "gpu_hours": 0.057102, "step": 500, "validation_ppl": 11.401976 }, { "gpu_hours": 0.071159, "step": 750, "validation_ppl": 11.394001 }, { "gpu_hours": 0.085364, "step": 1000, "validation_ppl": 11.403073 }, { "gpu_hours": 0.099456, "step": 1250, "validation_ppl": 11.399612 }, { "gpu_hours": 0.113655, "step": 1500, "validation_ppl": 11.40331 }, { "gpu_hours": 0.127764, "step": 1750, "validation_ppl": 11.393104 }, { "gpu_hours": 0.141979, "step": 2000, "validation_ppl": 11.402239 }, { "gpu_hours": 0.156092, "step": 2250, "validation_ppl": 11.397856 }, { "gpu_hours": 0.170371, "step": 2500, "validation_ppl": 11.405422 }, { "gpu_hours": 0.184611, "step": 2750, "validation_ppl": 11.39386 }, { "gpu_hours": 0.198765, "step": 3000, "validation_ppl": 11.40392 }, { "gpu_hours": 0.213011, "step": 3250, "validation_ppl": 11.395147 }, { "gpu_hours": 0.227155, "step": 3500, "validation_ppl": 11.406258 }, { "gpu_hours": 0.241423, "step": 3750, "validation_ppl": 11.404724 }, { "gpu_hours": 0.255562, "step": 4000, "validation_ppl": 11.403555 }, { "gpu_hours": 0.2698, "step": 4250, "validation_ppl": 11.408627 }, { "gpu_hours": 0.283971, "step": 4500, "validation_ppl": 11.417476 }, { "gpu_hours": 0.298203, "step": 4750, "validation_ppl": 11.413352 }, { "gpu_hours": 0.312346, "step": 5000, "validation_ppl": 11.406765 }, { "gpu_hours": 0.326662, "step": 5250, "validation_ppl": 11.40539 }, { "gpu_hours": 0.340824, "step": 5500, "validation_ppl": 11.415054 }, { "gpu_hours": 0.355049, "step": 5750, "validation_ppl": 11.415902 }, { "gpu_hours": 0.36932, "step": 6000, "validation_ppl": 11.419179 }, { "gpu_hours": 0.383446, "step": 6250, "validation_ppl": 11.421063 }, { "gpu_hours": 0.397686, "step": 6500, "validation_ppl": 11.425465 }, { "gpu_hours": 0.411816, "step": 6750, "validation_ppl": 11.423501 }, { "gpu_hours": 0.426006, "step": 7000, "validation_ppl": 11.419573 }, { "gpu_hours": 0.440142, "step": 7250, "validation_ppl": 11.421921 }, { "gpu_hours": 0.45435, "step": 7500, "validation_ppl": 11.419329 }, { "gpu_hours": 0.468506, "step": 7750, "validation_ppl": 11.416301 }, { "gpu_hours": 0.482711, "step": 8000, "validation_ppl": 11.446627 }, { "gpu_hours": 0.496815, "step": 8250, "validation_ppl": 11.468579 }, { "gpu_hours": 0.511025, "step": 8500, "validation_ppl": 11.497569 }, { "gpu_hours": 0.52519, "step": 8750, "validation_ppl": 11.516332 }, { "gpu_hours": 0.539422, "step": 9000, "validation_ppl": 11.523182 }, { "gpu_hours": 0.553667, "step": 9250, "validation_ppl": 11.526433 }, { "gpu_hours": 0.567801, "step": 9500, "validation_ppl": 11.538704 }, { "gpu_hours": 0.582087, "step": 9750, "validation_ppl": 11.556174 }, { "gpu_hours": 0.596236, "step": 10000, "validation_ppl": 11.555204 }, { "gpu_hours": 0.610449, "step": 10250, "validation_ppl": 11.557243 }, { "gpu_hours": 0.624549, "step": 10500, "validation_ppl": 11.57466 }, { "gpu_hours": 0.638795, "step": 10750, "validation_ppl": 11.571354 }, { "gpu_hours": 0.652947, "step": 11000, "validation_ppl": 11.580189 }, { "gpu_hours": 0.667204, "step": 11250, "validation_ppl": 11.581091 }, { "gpu_hours": 0.681354, "step": 11500, "validation_ppl": 11.592229 }, { "gpu_hours": 0.695623, "step": 11750, "validation_ppl": 11.587701 }, { "gpu_hours": 0.709804, "step": 12000, "validation_ppl": 11.589158 }, { "gpu_hours": 0.724042, "step": 12250, "validation_ppl": 11.585227 }, { "gpu_hours": 0.738293, "step": 12500, "validation_ppl": 11.593194 }, { "gpu_hours": 0.752408, "step": 12750, "validation_ppl": 11.595264 }, { "gpu_hours": 0.766687, "step": 13000, "validation_ppl": 11.595239 }, { "gpu_hours": 0.780852, "step": 13250, "validation_ppl": 11.610109 }, { "gpu_hours": 0.795151, "step": 13500, "validation_ppl": 11.600882 }, { "gpu_hours": 0.809328, "step": 13750, "validation_ppl": 11.61989 }, { "gpu_hours": 0.823416, "step": 14000, "validation_ppl": 11.617077 }, { "gpu_hours": 0.83737, "step": 14250, "validation_ppl": 11.611581 }, { "gpu_hours": 0.851486, "step": 14500, "validation_ppl": 11.620875 }, { "gpu_hours": 0.865661, "step": 14750, "validation_ppl": 11.621274 }, { "gpu_hours": 0.87987, "step": 15000, "validation_ppl": 11.629876 }, { "gpu_hours": 0.89401, "step": 15250, "validation_ppl": 11.622007 }, { "gpu_hours": 0.908245, "step": 15500, "validation_ppl": 11.608879 }, { "gpu_hours": 0.922541, "step": 15750, "validation_ppl": 11.780285 }, { "gpu_hours": 0.936643, "step": 16000, "validation_ppl": 11.822397 }, { "gpu_hours": 0.950843, "step": 16250, "validation_ppl": 11.854213 }, { "gpu_hours": 0.964943, "step": 16500, "validation_ppl": 11.87449 }, { "gpu_hours": 0.979128, "step": 16750, "validation_ppl": 11.901323 }, { "gpu_hours": 0.993242, "step": 17000, "validation_ppl": 11.929923 }, { "gpu_hours": 1.007453, "step": 17250, "validation_ppl": 11.938926 }, { "gpu_hours": 1.021546, "step": 17500, "validation_ppl": 11.938016 }, { "gpu_hours": 1.035768, "step": 17750, "validation_ppl": 11.984598 }, { "gpu_hours": 1.049875, "step": 18000, "validation_ppl": 11.954585 }, { "gpu_hours": 1.064093, "step": 18250, "validation_ppl": 11.951178 }, { "gpu_hours": 1.078189, "step": 18500, "validation_ppl": 11.993453 }, { "gpu_hours": 1.092407, "step": 18750, "validation_ppl": 11.983857 }, { "gpu_hours": 1.10653, "step": 19000, "validation_ppl": 12.004513 }, { "gpu_hours": 1.120743, "step": 19250, "validation_ppl": 11.999836 }, { "gpu_hours": 1.134966, "step": 19500, "validation_ppl": 12.006501 }, { "gpu_hours": 1.149067, "step": 19750, "validation_ppl": 11.994401 }, { "gpu_hours": 1.16327, "step": 20000, "validation_ppl": 12.014768 }, { "gpu_hours": 1.177388, "step": 20250, "validation_ppl": 12.031222 }, { "gpu_hours": 1.191622, "step": 20500, "validation_ppl": 12.043713 }, { "gpu_hours": 1.205707, "step": 20750, "validation_ppl": 12.045033 }, { "gpu_hours": 1.219929, "step": 21000, "validation_ppl": 12.053403 }, { "gpu_hours": 1.234034, "step": 21250, "validation_ppl": 12.048235 }, { "gpu_hours": 1.248268, "step": 21500, "validation_ppl": 12.043004 }, { "gpu_hours": 1.262404, "step": 21750, "validation_ppl": 12.046041 }, { "gpu_hours": 1.276595, "step": 22000, "validation_ppl": 12.06017 }, { "gpu_hours": 1.290725, "step": 22250, "validation_ppl": 12.042998 }, { "gpu_hours": 1.304903, "step": 22500, "validation_ppl": 12.050848 }, { "gpu_hours": 1.319124, "step": 22750, "validation_ppl": 12.049497 }, { "gpu_hours": 1.333278, "step": 23000, "validation_ppl": 12.077974 }, { "gpu_hours": 1.347505, "step": 23250, "validation_ppl": 12.091615 } ] } }, "learning_rate": "5e-5" } }, "schema_version": "1.0", "scope": "Sanitized online validation and process GPU-hour trajectory. No model weights, raw data, controller thresholds, or actuator coefficients.", "seed": 4007, "status": "complete", "steps": 23450, "study": "OptimizerPlus-RC106-Pilot-3Epoch-AdamW", "training_tokens": 24012800 }