| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 9.99996056937818, | |
| "global_step": 63400, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 3e-06, | |
| "loss": 6.9272, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 6e-06, | |
| "loss": 6.9063, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9e-06, | |
| "loss": 6.8492, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.2e-05, | |
| "loss": 6.724, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.5e-05, | |
| "loss": 6.534, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.8e-05, | |
| "loss": 6.3145, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.1e-05, | |
| "loss": 6.1061, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.4e-05, | |
| "loss": 5.8852, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.7000000000000002e-05, | |
| "loss": 5.6446, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3e-05, | |
| "loss": 5.4451, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.9976152623211448e-05, | |
| "loss": 5.2905, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.9952305246422896e-05, | |
| "loss": 5.0606, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.9928457869634343e-05, | |
| "loss": 4.8638, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.9904610492845787e-05, | |
| "loss": 4.713, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.9880763116057235e-05, | |
| "loss": 4.4828, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.9856915739268682e-05, | |
| "loss": 4.5133, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.983306836248013e-05, | |
| "loss": 4.2172, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.9809220985691574e-05, | |
| "loss": 4.1656, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.978537360890302e-05, | |
| "loss": 4.1704, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.976152623211447e-05, | |
| "loss": 4.0979, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.9737678855325916e-05, | |
| "loss": 3.9012, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.971383147853736e-05, | |
| "loss": 3.9303, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.9689984101748808e-05, | |
| "loss": 3.946, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.9666136724960255e-05, | |
| "loss": 3.9126, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.9642289348171703e-05, | |
| "loss": 3.8422, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.9618441971383147e-05, | |
| "loss": 3.6961, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.9594594594594598e-05, | |
| "loss": 3.7898, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.9570747217806042e-05, | |
| "loss": 3.721, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.954689984101749e-05, | |
| "loss": 3.8564, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.9523052464228933e-05, | |
| "loss": 3.7442, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.9499205087440384e-05, | |
| "loss": 3.84, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.9475357710651828e-05, | |
| "loss": 3.7246, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.9451510333863276e-05, | |
| "loss": 3.7482, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.942766295707472e-05, | |
| "loss": 3.5869, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.940381558028617e-05, | |
| "loss": 3.5348, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.9379968203497615e-05, | |
| "loss": 3.6747, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.9356120826709062e-05, | |
| "loss": 3.5978, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.933227344992051e-05, | |
| "loss": 3.6665, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.9308426073131957e-05, | |
| "loss": 3.7422, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.92845786963434e-05, | |
| "loss": 3.4649, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.926073131955485e-05, | |
| "loss": 3.5158, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 2.9236883942766296e-05, | |
| "loss": 3.4809, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 2.9213036565977744e-05, | |
| "loss": 3.59, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 2.918918918918919e-05, | |
| "loss": 3.444, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 2.9165341812400635e-05, | |
| "loss": 3.5945, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 2.9141494435612086e-05, | |
| "loss": 3.4512, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 2.911764705882353e-05, | |
| "loss": 3.5605, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 2.9093799682034978e-05, | |
| "loss": 3.6303, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 2.9069952305246422e-05, | |
| "loss": 3.5535, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 2.9046104928457873e-05, | |
| "loss": 3.4535, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 2.9022257551669317e-05, | |
| "loss": 3.5228, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 2.8998410174880764e-05, | |
| "loss": 3.4336, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 2.8974562798092212e-05, | |
| "loss": 3.5632, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 2.895071542130366e-05, | |
| "loss": 3.5295, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 2.8926868044515103e-05, | |
| "loss": 3.5078, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 2.890302066772655e-05, | |
| "loss": 3.4095, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 2.8879173290937998e-05, | |
| "loss": 3.4464, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 2.8855325914149446e-05, | |
| "loss": 3.5273, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 2.883147853736089e-05, | |
| "loss": 3.5008, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 2.8807631160572337e-05, | |
| "loss": 3.5109, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 2.8783783783783785e-05, | |
| "loss": 3.4497, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.8759936406995232e-05, | |
| "loss": 3.5214, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.8736089030206676e-05, | |
| "loss": 3.3884, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.8712241653418124e-05, | |
| "loss": 3.4611, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 2.868839427662957e-05, | |
| "loss": 3.4954, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.866454689984102e-05, | |
| "loss": 3.3634, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 2.8640699523052463e-05, | |
| "loss": 3.3091, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 2.8616852146263914e-05, | |
| "loss": 3.2928, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 2.8593004769475358e-05, | |
| "loss": 3.522, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 2.8569157392686805e-05, | |
| "loss": 3.4484, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 2.854531001589825e-05, | |
| "loss": 3.4451, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 2.85214626391097e-05, | |
| "loss": 3.4004, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 2.8497615262321144e-05, | |
| "loss": 3.3544, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 2.8473767885532592e-05, | |
| "loss": 3.3632, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 2.8449920508744036e-05, | |
| "loss": 3.4666, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 2.8426073131955487e-05, | |
| "loss": 3.3472, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 2.8402225755166934e-05, | |
| "loss": 3.3627, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 2.8378378378378378e-05, | |
| "loss": 3.3888, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 2.8354531001589826e-05, | |
| "loss": 3.3849, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 2.8330683624801273e-05, | |
| "loss": 3.4325, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 2.830683624801272e-05, | |
| "loss": 3.4523, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.8282988871224165e-05, | |
| "loss": 3.4083, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 2.8259141494435616e-05, | |
| "loss": 3.3159, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 2.823529411764706e-05, | |
| "loss": 3.3498, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 2.8211446740858507e-05, | |
| "loss": 3.4931, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 2.818759936406995e-05, | |
| "loss": 3.4344, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 2.8163751987281402e-05, | |
| "loss": 3.4875, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 2.8139904610492846e-05, | |
| "loss": 3.2618, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 2.8116057233704294e-05, | |
| "loss": 3.3632, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 2.8092209856915738e-05, | |
| "loss": 3.332, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 2.806836248012719e-05, | |
| "loss": 3.4005, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 2.8044515103338633e-05, | |
| "loss": 3.4338, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 2.802066772655008e-05, | |
| "loss": 3.4034, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 2.7996820349761528e-05, | |
| "loss": 3.2532, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 2.7972972972972975e-05, | |
| "loss": 3.2847, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.794912559618442e-05, | |
| "loss": 3.5397, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.7925278219395867e-05, | |
| "loss": 3.3293, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.7901430842607314e-05, | |
| "loss": 3.2783, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.7877583465818762e-05, | |
| "loss": 3.376, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.7853736089030206e-05, | |
| "loss": 3.3117, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.7829888712241653e-05, | |
| "loss": 3.3216, | |
| "step": 5050 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.78060413354531e-05, | |
| "loss": 3.4992, | |
| "step": 5100 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 2.7782193958664548e-05, | |
| "loss": 3.4038, | |
| "step": 5150 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 2.7758346581875992e-05, | |
| "loss": 3.2956, | |
| "step": 5200 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 2.773449920508744e-05, | |
| "loss": 3.3612, | |
| "step": 5250 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.7710651828298887e-05, | |
| "loss": 3.3029, | |
| "step": 5300 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.7686804451510335e-05, | |
| "loss": 3.4093, | |
| "step": 5350 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.766295707472178e-05, | |
| "loss": 3.2937, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.763910969793323e-05, | |
| "loss": 3.2035, | |
| "step": 5450 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 2.7615262321144677e-05, | |
| "loss": 3.3679, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 2.759141494435612e-05, | |
| "loss": 3.3374, | |
| "step": 5550 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 2.756756756756757e-05, | |
| "loss": 3.2496, | |
| "step": 5600 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 2.7543720190779016e-05, | |
| "loss": 3.2862, | |
| "step": 5650 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 2.7519872813990464e-05, | |
| "loss": 3.2177, | |
| "step": 5700 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 2.7496025437201908e-05, | |
| "loss": 3.229, | |
| "step": 5750 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 2.7472178060413355e-05, | |
| "loss": 3.3426, | |
| "step": 5800 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 2.7448330683624803e-05, | |
| "loss": 3.2579, | |
| "step": 5850 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.742448330683625e-05, | |
| "loss": 3.2497, | |
| "step": 5900 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.7400635930047694e-05, | |
| "loss": 3.3506, | |
| "step": 5950 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.7376788553259142e-05, | |
| "loss": 3.2882, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.735294117647059e-05, | |
| "loss": 3.3079, | |
| "step": 6050 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 2.7329093799682037e-05, | |
| "loss": 3.3407, | |
| "step": 6100 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 2.730524642289348e-05, | |
| "loss": 3.2765, | |
| "step": 6150 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.728139904610493e-05, | |
| "loss": 3.3539, | |
| "step": 6200 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 2.7257551669316376e-05, | |
| "loss": 3.3298, | |
| "step": 6250 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 2.7233704292527823e-05, | |
| "loss": 3.3286, | |
| "step": 6300 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 2.7209856915739267e-05, | |
| "loss": 3.3775, | |
| "step": 6350 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 2.7186009538950718e-05, | |
| "loss": 3.2558, | |
| "step": 6400 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 2.7162162162162162e-05, | |
| "loss": 3.1356, | |
| "step": 6450 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 2.713831478537361e-05, | |
| "loss": 3.2696, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 2.7114467408585054e-05, | |
| "loss": 3.4071, | |
| "step": 6550 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 2.7090620031796505e-05, | |
| "loss": 3.3242, | |
| "step": 6600 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 2.706677265500795e-05, | |
| "loss": 3.2378, | |
| "step": 6650 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 2.7042925278219396e-05, | |
| "loss": 3.41, | |
| "step": 6700 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 2.701907790143084e-05, | |
| "loss": 3.2521, | |
| "step": 6750 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 2.699523052464229e-05, | |
| "loss": 3.276, | |
| "step": 6800 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 2.6971383147853735e-05, | |
| "loss": 3.3381, | |
| "step": 6850 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 2.6947535771065183e-05, | |
| "loss": 3.2429, | |
| "step": 6900 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 2.692368839427663e-05, | |
| "loss": 3.247, | |
| "step": 6950 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 2.6899841017488078e-05, | |
| "loss": 3.2207, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 2.6875993640699522e-05, | |
| "loss": 3.2676, | |
| "step": 7050 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 2.685214626391097e-05, | |
| "loss": 3.2761, | |
| "step": 7100 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 2.682829888712242e-05, | |
| "loss": 3.2047, | |
| "step": 7150 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 2.6804451510333864e-05, | |
| "loss": 3.2238, | |
| "step": 7200 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 2.6780604133545312e-05, | |
| "loss": 3.2134, | |
| "step": 7250 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 2.6756756756756756e-05, | |
| "loss": 3.1721, | |
| "step": 7300 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 2.6732909379968207e-05, | |
| "loss": 3.2021, | |
| "step": 7350 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 2.670906200317965e-05, | |
| "loss": 3.2336, | |
| "step": 7400 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 2.6685214626391098e-05, | |
| "loss": 3.2834, | |
| "step": 7450 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 2.6661367249602546e-05, | |
| "loss": 3.1427, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 2.6637519872813993e-05, | |
| "loss": 3.2783, | |
| "step": 7550 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 2.6613672496025437e-05, | |
| "loss": 3.1762, | |
| "step": 7600 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 2.6589825119236885e-05, | |
| "loss": 3.1464, | |
| "step": 7650 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 2.6565977742448332e-05, | |
| "loss": 3.2136, | |
| "step": 7700 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 2.654213036565978e-05, | |
| "loss": 3.318, | |
| "step": 7750 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 2.6518282988871224e-05, | |
| "loss": 3.2813, | |
| "step": 7800 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 2.649443561208267e-05, | |
| "loss": 3.2116, | |
| "step": 7850 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 2.647058823529412e-05, | |
| "loss": 3.3032, | |
| "step": 7900 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 2.6446740858505566e-05, | |
| "loss": 3.1659, | |
| "step": 7950 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 2.642289348171701e-05, | |
| "loss": 3.1673, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 2.6399046104928458e-05, | |
| "loss": 3.1739, | |
| "step": 8050 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 2.6375198728139905e-05, | |
| "loss": 3.2324, | |
| "step": 8100 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 2.6351351351351353e-05, | |
| "loss": 3.2337, | |
| "step": 8150 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 2.6327503974562797e-05, | |
| "loss": 3.3776, | |
| "step": 8200 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 2.6303656597774248e-05, | |
| "loss": 3.1979, | |
| "step": 8250 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 2.6279809220985692e-05, | |
| "loss": 3.2378, | |
| "step": 8300 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 2.625596184419714e-05, | |
| "loss": 3.1871, | |
| "step": 8350 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 2.6232114467408583e-05, | |
| "loss": 3.2388, | |
| "step": 8400 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 2.6208267090620034e-05, | |
| "loss": 3.1731, | |
| "step": 8450 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 2.6184419713831478e-05, | |
| "loss": 3.111, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 2.6160572337042926e-05, | |
| "loss": 3.0985, | |
| "step": 8550 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 2.613672496025437e-05, | |
| "loss": 3.175, | |
| "step": 8600 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 2.611287758346582e-05, | |
| "loss": 3.0125, | |
| "step": 8650 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 2.6089030206677265e-05, | |
| "loss": 3.0734, | |
| "step": 8700 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 2.6065182829888712e-05, | |
| "loss": 3.181, | |
| "step": 8750 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 2.604133545310016e-05, | |
| "loss": 3.1494, | |
| "step": 8800 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 2.6017488076311607e-05, | |
| "loss": 3.1585, | |
| "step": 8850 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 2.5993640699523055e-05, | |
| "loss": 3.1531, | |
| "step": 8900 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 2.59697933227345e-05, | |
| "loss": 3.0955, | |
| "step": 8950 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 2.594594594594595e-05, | |
| "loss": 3.2459, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 2.5922098569157394e-05, | |
| "loss": 3.0333, | |
| "step": 9050 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 2.589825119236884e-05, | |
| "loss": 3.1748, | |
| "step": 9100 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 2.5874403815580285e-05, | |
| "loss": 3.2446, | |
| "step": 9150 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 2.5850556438791736e-05, | |
| "loss": 3.1582, | |
| "step": 9200 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 2.582670906200318e-05, | |
| "loss": 3.0357, | |
| "step": 9250 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 2.5802861685214628e-05, | |
| "loss": 3.1443, | |
| "step": 9300 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 2.5779014308426072e-05, | |
| "loss": 3.2152, | |
| "step": 9350 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 2.5755166931637523e-05, | |
| "loss": 3.0451, | |
| "step": 9400 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 2.5731319554848967e-05, | |
| "loss": 3.1334, | |
| "step": 9450 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 2.5707472178060414e-05, | |
| "loss": 3.0432, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 2.5683624801271858e-05, | |
| "loss": 3.0226, | |
| "step": 9550 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 2.565977742448331e-05, | |
| "loss": 3.0225, | |
| "step": 9600 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 2.5635930047694753e-05, | |
| "loss": 3.0502, | |
| "step": 9650 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 2.56120826709062e-05, | |
| "loss": 3.0431, | |
| "step": 9700 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 2.5588235294117648e-05, | |
| "loss": 3.0514, | |
| "step": 9750 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 2.5564387917329096e-05, | |
| "loss": 2.9757, | |
| "step": 9800 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 2.554054054054054e-05, | |
| "loss": 2.9868, | |
| "step": 9850 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 2.5516693163751987e-05, | |
| "loss": 3.0799, | |
| "step": 9900 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 2.5492845786963435e-05, | |
| "loss": 3.0556, | |
| "step": 9950 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 2.5468998410174882e-05, | |
| "loss": 3.0771, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 2.5445151033386326e-05, | |
| "loss": 2.9782, | |
| "step": 10050 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 2.5421303656597774e-05, | |
| "loss": 2.9809, | |
| "step": 10100 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 2.539745627980922e-05, | |
| "loss": 3.019, | |
| "step": 10150 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 2.537360890302067e-05, | |
| "loss": 3.0336, | |
| "step": 10200 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 2.5349761526232113e-05, | |
| "loss": 2.9084, | |
| "step": 10250 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 2.5325914149443564e-05, | |
| "loss": 2.9777, | |
| "step": 10300 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 2.5302066772655008e-05, | |
| "loss": 2.9971, | |
| "step": 10350 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 2.5278219395866455e-05, | |
| "loss": 3.0934, | |
| "step": 10400 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 2.5254372019077903e-05, | |
| "loss": 2.9911, | |
| "step": 10450 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 2.523052464228935e-05, | |
| "loss": 2.8821, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 2.5206677265500798e-05, | |
| "loss": 3.0607, | |
| "step": 10550 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 2.5182829888712242e-05, | |
| "loss": 2.9407, | |
| "step": 10600 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 2.515898251192369e-05, | |
| "loss": 2.9614, | |
| "step": 10650 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 2.5135135135135137e-05, | |
| "loss": 2.9679, | |
| "step": 10700 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 2.5111287758346584e-05, | |
| "loss": 2.9488, | |
| "step": 10750 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 2.5087440381558028e-05, | |
| "loss": 3.0865, | |
| "step": 10800 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 2.5063593004769476e-05, | |
| "loss": 3.0614, | |
| "step": 10850 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 2.5039745627980923e-05, | |
| "loss": 2.9559, | |
| "step": 10900 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 2.501589825119237e-05, | |
| "loss": 2.9049, | |
| "step": 10950 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 2.4992050874403815e-05, | |
| "loss": 2.9864, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 2.4968203497615266e-05, | |
| "loss": 2.9728, | |
| "step": 11050 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 2.494435612082671e-05, | |
| "loss": 2.8598, | |
| "step": 11100 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 2.4920508744038157e-05, | |
| "loss": 2.9705, | |
| "step": 11150 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 2.48966613672496e-05, | |
| "loss": 2.872, | |
| "step": 11200 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 2.4872813990461052e-05, | |
| "loss": 2.9734, | |
| "step": 11250 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 2.4848966613672496e-05, | |
| "loss": 2.9332, | |
| "step": 11300 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 2.4825119236883944e-05, | |
| "loss": 2.9188, | |
| "step": 11350 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 2.4801271860095388e-05, | |
| "loss": 2.8664, | |
| "step": 11400 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 2.477742448330684e-05, | |
| "loss": 2.8799, | |
| "step": 11450 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 2.4753577106518283e-05, | |
| "loss": 2.9415, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 2.472972972972973e-05, | |
| "loss": 2.9911, | |
| "step": 11550 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 2.4705882352941174e-05, | |
| "loss": 2.9031, | |
| "step": 11600 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 2.4682034976152625e-05, | |
| "loss": 2.9432, | |
| "step": 11650 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 2.465818759936407e-05, | |
| "loss": 2.9868, | |
| "step": 11700 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 2.4634340222575517e-05, | |
| "loss": 2.8311, | |
| "step": 11750 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 2.4610492845786964e-05, | |
| "loss": 2.8431, | |
| "step": 11800 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 2.458664546899841e-05, | |
| "loss": 2.9702, | |
| "step": 11850 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 2.4562798092209856e-05, | |
| "loss": 2.9432, | |
| "step": 11900 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 2.4538950715421303e-05, | |
| "loss": 2.904, | |
| "step": 11950 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 2.451510333863275e-05, | |
| "loss": 2.899, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 2.4491255961844198e-05, | |
| "loss": 2.7878, | |
| "step": 12050 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 2.4467408585055646e-05, | |
| "loss": 2.876, | |
| "step": 12100 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 2.444356120826709e-05, | |
| "loss": 2.9902, | |
| "step": 12150 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 2.441971383147854e-05, | |
| "loss": 2.8072, | |
| "step": 12200 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 2.4395866454689985e-05, | |
| "loss": 2.7914, | |
| "step": 12250 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 2.4372019077901432e-05, | |
| "loss": 2.9975, | |
| "step": 12300 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 2.4348171701112876e-05, | |
| "loss": 2.9566, | |
| "step": 12350 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 2.4324324324324327e-05, | |
| "loss": 2.9045, | |
| "step": 12400 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 2.430047694753577e-05, | |
| "loss": 2.9696, | |
| "step": 12450 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 2.427662957074722e-05, | |
| "loss": 3.03, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 2.4252782193958666e-05, | |
| "loss": 2.8536, | |
| "step": 12550 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 2.4228934817170114e-05, | |
| "loss": 2.9715, | |
| "step": 12600 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 2.4205087440381558e-05, | |
| "loss": 2.8773, | |
| "step": 12650 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 2.4181240063593005e-05, | |
| "loss": 2.8083, | |
| "step": 12700 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 2.4157392686804453e-05, | |
| "loss": 2.8994, | |
| "step": 12750 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 2.41335453100159e-05, | |
| "loss": 2.8305, | |
| "step": 12800 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 2.4109697933227344e-05, | |
| "loss": 2.9153, | |
| "step": 12850 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 2.4085850556438792e-05, | |
| "loss": 2.8469, | |
| "step": 12900 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 2.406200317965024e-05, | |
| "loss": 2.8301, | |
| "step": 12950 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 2.4038155802861687e-05, | |
| "loss": 2.8205, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 2.401430842607313e-05, | |
| "loss": 2.8669, | |
| "step": 13050 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 2.399046104928458e-05, | |
| "loss": 2.7455, | |
| "step": 13100 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 2.3966613672496026e-05, | |
| "loss": 2.8466, | |
| "step": 13150 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 2.3942766295707473e-05, | |
| "loss": 2.7977, | |
| "step": 13200 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 2.3918918918918917e-05, | |
| "loss": 2.7967, | |
| "step": 13250 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 2.3895071542130368e-05, | |
| "loss": 2.8632, | |
| "step": 13300 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 2.3871224165341812e-05, | |
| "loss": 2.8509, | |
| "step": 13350 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 2.384737678855326e-05, | |
| "loss": 2.7735, | |
| "step": 13400 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 2.3823529411764704e-05, | |
| "loss": 2.7421, | |
| "step": 13450 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 2.3799682034976155e-05, | |
| "loss": 2.8944, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 2.37758346581876e-05, | |
| "loss": 2.812, | |
| "step": 13550 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 2.3751987281399046e-05, | |
| "loss": 2.7978, | |
| "step": 13600 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 2.372813990461049e-05, | |
| "loss": 2.8839, | |
| "step": 13650 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 2.370429252782194e-05, | |
| "loss": 2.8741, | |
| "step": 13700 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 2.368044515103339e-05, | |
| "loss": 2.812, | |
| "step": 13750 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 2.3656597774244833e-05, | |
| "loss": 2.8138, | |
| "step": 13800 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 2.3632750397456284e-05, | |
| "loss": 2.7536, | |
| "step": 13850 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 2.3608903020667728e-05, | |
| "loss": 2.9393, | |
| "step": 13900 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 2.3585055643879175e-05, | |
| "loss": 2.8055, | |
| "step": 13950 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 2.356120826709062e-05, | |
| "loss": 2.7866, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 2.353736089030207e-05, | |
| "loss": 2.7212, | |
| "step": 14050 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 2.3513513513513514e-05, | |
| "loss": 2.901, | |
| "step": 14100 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 2.348966613672496e-05, | |
| "loss": 2.9047, | |
| "step": 14150 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 2.3465818759936406e-05, | |
| "loss": 2.8909, | |
| "step": 14200 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 2.3441971383147857e-05, | |
| "loss": 2.7759, | |
| "step": 14250 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 2.34181240063593e-05, | |
| "loss": 2.8714, | |
| "step": 14300 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 2.3394276629570748e-05, | |
| "loss": 2.7804, | |
| "step": 14350 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 2.3370429252782192e-05, | |
| "loss": 2.8279, | |
| "step": 14400 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 2.3346581875993643e-05, | |
| "loss": 2.7939, | |
| "step": 14450 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 2.3322734499205087e-05, | |
| "loss": 2.7902, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 2.3298887122416535e-05, | |
| "loss": 2.7452, | |
| "step": 14550 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 2.3275039745627982e-05, | |
| "loss": 2.7765, | |
| "step": 14600 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 2.325119236883943e-05, | |
| "loss": 2.7643, | |
| "step": 14650 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 2.3227344992050874e-05, | |
| "loss": 2.8241, | |
| "step": 14700 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 2.320349761526232e-05, | |
| "loss": 2.6875, | |
| "step": 14750 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 2.317965023847377e-05, | |
| "loss": 2.7444, | |
| "step": 14800 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 2.3155802861685216e-05, | |
| "loss": 2.8183, | |
| "step": 14850 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 2.313195548489666e-05, | |
| "loss": 2.7885, | |
| "step": 14900 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 2.3108108108108108e-05, | |
| "loss": 2.7358, | |
| "step": 14950 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 2.3084260731319555e-05, | |
| "loss": 2.7983, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 2.3060413354531003e-05, | |
| "loss": 2.7354, | |
| "step": 15050 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 2.3036565977742447e-05, | |
| "loss": 2.7596, | |
| "step": 15100 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 2.3012718600953894e-05, | |
| "loss": 2.7021, | |
| "step": 15150 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 2.2988871224165342e-05, | |
| "loss": 2.7264, | |
| "step": 15200 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 2.296502384737679e-05, | |
| "loss": 2.7476, | |
| "step": 15250 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 2.2941176470588233e-05, | |
| "loss": 2.8127, | |
| "step": 15300 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 2.2917329093799684e-05, | |
| "loss": 2.73, | |
| "step": 15350 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 2.289348171701113e-05, | |
| "loss": 2.7201, | |
| "step": 15400 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 2.2869634340222576e-05, | |
| "loss": 2.8273, | |
| "step": 15450 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 2.2845786963434023e-05, | |
| "loss": 2.8538, | |
| "step": 15500 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 2.282193958664547e-05, | |
| "loss": 2.7814, | |
| "step": 15550 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 2.2798092209856918e-05, | |
| "loss": 2.7943, | |
| "step": 15600 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 2.2774244833068362e-05, | |
| "loss": 2.8791, | |
| "step": 15650 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 2.275039745627981e-05, | |
| "loss": 2.6577, | |
| "step": 15700 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 2.2726550079491257e-05, | |
| "loss": 2.7085, | |
| "step": 15750 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 2.2702702702702705e-05, | |
| "loss": 2.8721, | |
| "step": 15800 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 2.267885532591415e-05, | |
| "loss": 2.7809, | |
| "step": 15850 | |
| }, | |
| { | |
| "epoch": 2.51, | |
| "learning_rate": 2.26550079491256e-05, | |
| "loss": 2.8199, | |
| "step": 15900 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 2.2631160572337044e-05, | |
| "loss": 2.7886, | |
| "step": 15950 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 2.260731319554849e-05, | |
| "loss": 2.6889, | |
| "step": 16000 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 2.2583465818759935e-05, | |
| "loss": 2.8305, | |
| "step": 16050 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 2.2559618441971386e-05, | |
| "loss": 2.6675, | |
| "step": 16100 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 2.253577106518283e-05, | |
| "loss": 2.7808, | |
| "step": 16150 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 2.2511923688394278e-05, | |
| "loss": 2.878, | |
| "step": 16200 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 2.2488076311605722e-05, | |
| "loss": 2.7607, | |
| "step": 16250 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 2.2464228934817173e-05, | |
| "loss": 2.7103, | |
| "step": 16300 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 2.2440381558028617e-05, | |
| "loss": 2.7305, | |
| "step": 16350 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 2.2416534181240064e-05, | |
| "loss": 2.7234, | |
| "step": 16400 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 2.2392686804451508e-05, | |
| "loss": 2.8669, | |
| "step": 16450 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 2.236883942766296e-05, | |
| "loss": 2.7355, | |
| "step": 16500 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 2.2344992050874403e-05, | |
| "loss": 2.8589, | |
| "step": 16550 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 2.232114467408585e-05, | |
| "loss": 2.7384, | |
| "step": 16600 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 2.2297297297297298e-05, | |
| "loss": 2.7677, | |
| "step": 16650 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 2.2273449920508746e-05, | |
| "loss": 2.7464, | |
| "step": 16700 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 2.224960254372019e-05, | |
| "loss": 2.7476, | |
| "step": 16750 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 2.2225755166931637e-05, | |
| "loss": 2.6909, | |
| "step": 16800 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 2.2201907790143085e-05, | |
| "loss": 2.7507, | |
| "step": 16850 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 2.2178060413354532e-05, | |
| "loss": 2.7785, | |
| "step": 16900 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 2.2154213036565976e-05, | |
| "loss": 2.7384, | |
| "step": 16950 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 2.2130365659777424e-05, | |
| "loss": 2.8127, | |
| "step": 17000 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 2.210651828298887e-05, | |
| "loss": 2.6794, | |
| "step": 17050 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 2.208267090620032e-05, | |
| "loss": 2.732, | |
| "step": 17100 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 2.2058823529411766e-05, | |
| "loss": 2.6593, | |
| "step": 17150 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 2.203497615262321e-05, | |
| "loss": 2.8902, | |
| "step": 17200 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 2.201112877583466e-05, | |
| "loss": 2.6455, | |
| "step": 17250 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 2.1987281399046105e-05, | |
| "loss": 2.6208, | |
| "step": 17300 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 2.1963434022257553e-05, | |
| "loss": 2.8468, | |
| "step": 17350 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 2.1939586645469e-05, | |
| "loss": 2.7659, | |
| "step": 17400 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 2.1915739268680448e-05, | |
| "loss": 2.6662, | |
| "step": 17450 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 2.1891891891891892e-05, | |
| "loss": 2.7181, | |
| "step": 17500 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 2.186804451510334e-05, | |
| "loss": 2.8185, | |
| "step": 17550 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 2.1844197138314787e-05, | |
| "loss": 2.7554, | |
| "step": 17600 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 2.1820349761526234e-05, | |
| "loss": 2.7064, | |
| "step": 17650 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 2.1796502384737678e-05, | |
| "loss": 2.7318, | |
| "step": 17700 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 2.1772655007949126e-05, | |
| "loss": 2.7196, | |
| "step": 17750 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 2.1748807631160573e-05, | |
| "loss": 2.5457, | |
| "step": 17800 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 2.172496025437202e-05, | |
| "loss": 2.6836, | |
| "step": 17850 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 2.1701112877583465e-05, | |
| "loss": 2.685, | |
| "step": 17900 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 2.1677265500794912e-05, | |
| "loss": 2.749, | |
| "step": 17950 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 2.165341812400636e-05, | |
| "loss": 2.781, | |
| "step": 18000 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 2.1629570747217807e-05, | |
| "loss": 2.7667, | |
| "step": 18050 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 2.160572337042925e-05, | |
| "loss": 2.7558, | |
| "step": 18100 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 2.1581875993640702e-05, | |
| "loss": 2.6881, | |
| "step": 18150 | |
| }, | |
| { | |
| "epoch": 2.87, | |
| "learning_rate": 2.1558028616852146e-05, | |
| "loss": 2.7462, | |
| "step": 18200 | |
| }, | |
| { | |
| "epoch": 2.88, | |
| "learning_rate": 2.1534181240063594e-05, | |
| "loss": 2.6983, | |
| "step": 18250 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 2.1510333863275038e-05, | |
| "loss": 2.6547, | |
| "step": 18300 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 2.148648648648649e-05, | |
| "loss": 2.65, | |
| "step": 18350 | |
| }, | |
| { | |
| "epoch": 2.9, | |
| "learning_rate": 2.1462639109697933e-05, | |
| "loss": 2.7827, | |
| "step": 18400 | |
| }, | |
| { | |
| "epoch": 2.91, | |
| "learning_rate": 2.143879173290938e-05, | |
| "loss": 2.6902, | |
| "step": 18450 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 2.1414944356120824e-05, | |
| "loss": 2.6752, | |
| "step": 18500 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 2.1391096979332275e-05, | |
| "loss": 2.6049, | |
| "step": 18550 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 2.136724960254372e-05, | |
| "loss": 2.8715, | |
| "step": 18600 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 2.1343402225755167e-05, | |
| "loss": 2.6845, | |
| "step": 18650 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 2.1319554848966614e-05, | |
| "loss": 2.7181, | |
| "step": 18700 | |
| }, | |
| { | |
| "epoch": 2.96, | |
| "learning_rate": 2.129570747217806e-05, | |
| "loss": 2.688, | |
| "step": 18750 | |
| }, | |
| { | |
| "epoch": 2.97, | |
| "learning_rate": 2.127186009538951e-05, | |
| "loss": 2.7634, | |
| "step": 18800 | |
| }, | |
| { | |
| "epoch": 2.97, | |
| "learning_rate": 2.1248012718600953e-05, | |
| "loss": 2.7003, | |
| "step": 18850 | |
| }, | |
| { | |
| "epoch": 2.98, | |
| "learning_rate": 2.1224165341812404e-05, | |
| "loss": 2.6903, | |
| "step": 18900 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 2.1200317965023848e-05, | |
| "loss": 2.635, | |
| "step": 18950 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 2.1176470588235296e-05, | |
| "loss": 2.6414, | |
| "step": 19000 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 2.115262321144674e-05, | |
| "loss": 2.7126, | |
| "step": 19050 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 2.112877583465819e-05, | |
| "loss": 2.5964, | |
| "step": 19100 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 2.1104928457869635e-05, | |
| "loss": 2.7598, | |
| "step": 19150 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 2.1081081081081082e-05, | |
| "loss": 2.6722, | |
| "step": 19200 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 2.1057233704292526e-05, | |
| "loss": 2.684, | |
| "step": 19250 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 2.1033386327503977e-05, | |
| "loss": 2.6479, | |
| "step": 19300 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 2.100953895071542e-05, | |
| "loss": 2.7599, | |
| "step": 19350 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 2.098569157392687e-05, | |
| "loss": 2.8154, | |
| "step": 19400 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 2.0961844197138316e-05, | |
| "loss": 2.6285, | |
| "step": 19450 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 2.0937996820349764e-05, | |
| "loss": 2.6614, | |
| "step": 19500 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 2.0914149443561208e-05, | |
| "loss": 2.5899, | |
| "step": 19550 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 2.0890302066772655e-05, | |
| "loss": 2.6255, | |
| "step": 19600 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 2.0866454689984103e-05, | |
| "loss": 2.7297, | |
| "step": 19650 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 2.084260731319555e-05, | |
| "loss": 2.5885, | |
| "step": 19700 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 2.0818759936406994e-05, | |
| "loss": 2.6466, | |
| "step": 19750 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 2.0794912559618442e-05, | |
| "loss": 2.6435, | |
| "step": 19800 | |
| }, | |
| { | |
| "epoch": 3.13, | |
| "learning_rate": 2.077106518282989e-05, | |
| "loss": 2.6621, | |
| "step": 19850 | |
| }, | |
| { | |
| "epoch": 3.14, | |
| "learning_rate": 2.0747217806041337e-05, | |
| "loss": 2.6563, | |
| "step": 19900 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 2.072337042925278e-05, | |
| "loss": 2.636, | |
| "step": 19950 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 2.0699523052464228e-05, | |
| "loss": 2.5775, | |
| "step": 20000 | |
| }, | |
| { | |
| "epoch": 3.16, | |
| "learning_rate": 2.0675675675675676e-05, | |
| "loss": 2.6366, | |
| "step": 20050 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 2.0651828298887123e-05, | |
| "loss": 2.6649, | |
| "step": 20100 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "learning_rate": 2.0627980922098567e-05, | |
| "loss": 2.7077, | |
| "step": 20150 | |
| }, | |
| { | |
| "epoch": 3.19, | |
| "learning_rate": 2.0604133545310018e-05, | |
| "loss": 2.6725, | |
| "step": 20200 | |
| }, | |
| { | |
| "epoch": 3.19, | |
| "learning_rate": 2.0580286168521462e-05, | |
| "loss": 2.7058, | |
| "step": 20250 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "learning_rate": 2.055643879173291e-05, | |
| "loss": 2.6434, | |
| "step": 20300 | |
| }, | |
| { | |
| "epoch": 3.21, | |
| "learning_rate": 2.0532591414944354e-05, | |
| "loss": 2.6279, | |
| "step": 20350 | |
| }, | |
| { | |
| "epoch": 3.22, | |
| "learning_rate": 2.0508744038155805e-05, | |
| "loss": 2.7532, | |
| "step": 20400 | |
| }, | |
| { | |
| "epoch": 3.23, | |
| "learning_rate": 2.0484896661367252e-05, | |
| "loss": 2.6562, | |
| "step": 20450 | |
| }, | |
| { | |
| "epoch": 3.23, | |
| "learning_rate": 2.0461049284578696e-05, | |
| "loss": 2.679, | |
| "step": 20500 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "learning_rate": 2.0437201907790144e-05, | |
| "loss": 2.6441, | |
| "step": 20550 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "learning_rate": 2.041335453100159e-05, | |
| "loss": 2.7591, | |
| "step": 20600 | |
| }, | |
| { | |
| "epoch": 3.26, | |
| "learning_rate": 2.038950715421304e-05, | |
| "loss": 2.73, | |
| "step": 20650 | |
| }, | |
| { | |
| "epoch": 3.26, | |
| "learning_rate": 2.0365659777424483e-05, | |
| "loss": 2.6668, | |
| "step": 20700 | |
| }, | |
| { | |
| "epoch": 3.27, | |
| "learning_rate": 2.034181240063593e-05, | |
| "loss": 2.5595, | |
| "step": 20750 | |
| }, | |
| { | |
| "epoch": 3.28, | |
| "learning_rate": 2.0317965023847378e-05, | |
| "loss": 2.5686, | |
| "step": 20800 | |
| }, | |
| { | |
| "epoch": 3.29, | |
| "learning_rate": 2.0294117647058825e-05, | |
| "loss": 2.6029, | |
| "step": 20850 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 2.027027027027027e-05, | |
| "loss": 2.5819, | |
| "step": 20900 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 2.024642289348172e-05, | |
| "loss": 2.638, | |
| "step": 20950 | |
| }, | |
| { | |
| "epoch": 3.31, | |
| "learning_rate": 2.0222575516693164e-05, | |
| "loss": 2.625, | |
| "step": 21000 | |
| }, | |
| { | |
| "epoch": 3.32, | |
| "learning_rate": 2.019872813990461e-05, | |
| "loss": 2.6363, | |
| "step": 21050 | |
| }, | |
| { | |
| "epoch": 3.33, | |
| "learning_rate": 2.0174880763116056e-05, | |
| "loss": 2.8446, | |
| "step": 21100 | |
| }, | |
| { | |
| "epoch": 3.34, | |
| "learning_rate": 2.0151033386327507e-05, | |
| "loss": 2.6304, | |
| "step": 21150 | |
| }, | |
| { | |
| "epoch": 3.34, | |
| "learning_rate": 2.012718600953895e-05, | |
| "loss": 2.5041, | |
| "step": 21200 | |
| }, | |
| { | |
| "epoch": 3.35, | |
| "learning_rate": 2.0103338632750398e-05, | |
| "loss": 2.6228, | |
| "step": 21250 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 2.0079491255961842e-05, | |
| "loss": 2.5432, | |
| "step": 21300 | |
| }, | |
| { | |
| "epoch": 3.37, | |
| "learning_rate": 2.0055643879173293e-05, | |
| "loss": 2.7547, | |
| "step": 21350 | |
| }, | |
| { | |
| "epoch": 3.38, | |
| "learning_rate": 2.0031796502384737e-05, | |
| "loss": 2.634, | |
| "step": 21400 | |
| }, | |
| { | |
| "epoch": 3.38, | |
| "learning_rate": 2.0007949125596185e-05, | |
| "loss": 2.5405, | |
| "step": 21450 | |
| }, | |
| { | |
| "epoch": 3.39, | |
| "learning_rate": 1.9984101748807632e-05, | |
| "loss": 2.6386, | |
| "step": 21500 | |
| }, | |
| { | |
| "epoch": 3.4, | |
| "learning_rate": 1.996025437201908e-05, | |
| "loss": 2.6188, | |
| "step": 21550 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 1.9936406995230524e-05, | |
| "loss": 2.6959, | |
| "step": 21600 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 1.991255961844197e-05, | |
| "loss": 2.6742, | |
| "step": 21650 | |
| }, | |
| { | |
| "epoch": 3.42, | |
| "learning_rate": 1.988871224165342e-05, | |
| "loss": 2.5569, | |
| "step": 21700 | |
| }, | |
| { | |
| "epoch": 3.43, | |
| "learning_rate": 1.9864864864864866e-05, | |
| "loss": 2.5727, | |
| "step": 21750 | |
| }, | |
| { | |
| "epoch": 3.44, | |
| "learning_rate": 1.984101748807631e-05, | |
| "loss": 2.5024, | |
| "step": 21800 | |
| }, | |
| { | |
| "epoch": 3.45, | |
| "learning_rate": 1.9817170111287758e-05, | |
| "loss": 2.6301, | |
| "step": 21850 | |
| }, | |
| { | |
| "epoch": 3.45, | |
| "learning_rate": 1.9793322734499205e-05, | |
| "loss": 2.539, | |
| "step": 21900 | |
| }, | |
| { | |
| "epoch": 3.46, | |
| "learning_rate": 1.9769475357710653e-05, | |
| "loss": 2.513, | |
| "step": 21950 | |
| }, | |
| { | |
| "epoch": 3.47, | |
| "learning_rate": 1.9745627980922097e-05, | |
| "loss": 2.6151, | |
| "step": 22000 | |
| }, | |
| { | |
| "epoch": 3.48, | |
| "learning_rate": 1.9721780604133544e-05, | |
| "loss": 2.5928, | |
| "step": 22050 | |
| }, | |
| { | |
| "epoch": 3.49, | |
| "learning_rate": 1.9697933227344995e-05, | |
| "loss": 2.69, | |
| "step": 22100 | |
| }, | |
| { | |
| "epoch": 3.49, | |
| "learning_rate": 1.967408585055644e-05, | |
| "loss": 2.5951, | |
| "step": 22150 | |
| }, | |
| { | |
| "epoch": 3.5, | |
| "learning_rate": 1.9650238473767887e-05, | |
| "loss": 2.6194, | |
| "step": 22200 | |
| }, | |
| { | |
| "epoch": 3.51, | |
| "learning_rate": 1.9626391096979334e-05, | |
| "loss": 2.7138, | |
| "step": 22250 | |
| }, | |
| { | |
| "epoch": 3.52, | |
| "learning_rate": 1.960254372019078e-05, | |
| "loss": 2.6472, | |
| "step": 22300 | |
| }, | |
| { | |
| "epoch": 3.53, | |
| "learning_rate": 1.9578696343402226e-05, | |
| "loss": 2.6166, | |
| "step": 22350 | |
| }, | |
| { | |
| "epoch": 3.53, | |
| "learning_rate": 1.9554848966613673e-05, | |
| "loss": 2.6573, | |
| "step": 22400 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 1.953100158982512e-05, | |
| "loss": 2.531, | |
| "step": 22450 | |
| }, | |
| { | |
| "epoch": 3.55, | |
| "learning_rate": 1.9507154213036568e-05, | |
| "loss": 2.5752, | |
| "step": 22500 | |
| }, | |
| { | |
| "epoch": 3.56, | |
| "learning_rate": 1.9483306836248012e-05, | |
| "loss": 2.5375, | |
| "step": 22550 | |
| }, | |
| { | |
| "epoch": 3.56, | |
| "learning_rate": 1.945945945945946e-05, | |
| "loss": 2.5569, | |
| "step": 22600 | |
| }, | |
| { | |
| "epoch": 3.57, | |
| "learning_rate": 1.9435612082670907e-05, | |
| "loss": 2.6506, | |
| "step": 22650 | |
| }, | |
| { | |
| "epoch": 3.58, | |
| "learning_rate": 1.9411764705882355e-05, | |
| "loss": 2.5785, | |
| "step": 22700 | |
| }, | |
| { | |
| "epoch": 3.59, | |
| "learning_rate": 1.93879173290938e-05, | |
| "loss": 2.6245, | |
| "step": 22750 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "learning_rate": 1.9364069952305246e-05, | |
| "loss": 2.6386, | |
| "step": 22800 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "learning_rate": 1.9340222575516694e-05, | |
| "loss": 2.5858, | |
| "step": 22850 | |
| }, | |
| { | |
| "epoch": 3.61, | |
| "learning_rate": 1.931637519872814e-05, | |
| "loss": 2.6831, | |
| "step": 22900 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 1.9292527821939585e-05, | |
| "loss": 2.5829, | |
| "step": 22950 | |
| }, | |
| { | |
| "epoch": 3.63, | |
| "learning_rate": 1.9268680445151036e-05, | |
| "loss": 2.5942, | |
| "step": 23000 | |
| }, | |
| { | |
| "epoch": 3.64, | |
| "learning_rate": 1.924483306836248e-05, | |
| "loss": 2.4712, | |
| "step": 23050 | |
| }, | |
| { | |
| "epoch": 3.64, | |
| "learning_rate": 1.9220985691573928e-05, | |
| "loss": 2.663, | |
| "step": 23100 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 1.9197138314785372e-05, | |
| "loss": 2.68, | |
| "step": 23150 | |
| }, | |
| { | |
| "epoch": 3.66, | |
| "learning_rate": 1.9173290937996823e-05, | |
| "loss": 2.5576, | |
| "step": 23200 | |
| }, | |
| { | |
| "epoch": 3.67, | |
| "learning_rate": 1.9149443561208267e-05, | |
| "loss": 2.5646, | |
| "step": 23250 | |
| }, | |
| { | |
| "epoch": 3.68, | |
| "learning_rate": 1.9125596184419714e-05, | |
| "loss": 2.6577, | |
| "step": 23300 | |
| }, | |
| { | |
| "epoch": 3.68, | |
| "learning_rate": 1.9101748807631158e-05, | |
| "loss": 2.6625, | |
| "step": 23350 | |
| }, | |
| { | |
| "epoch": 3.69, | |
| "learning_rate": 1.907790143084261e-05, | |
| "loss": 2.5941, | |
| "step": 23400 | |
| }, | |
| { | |
| "epoch": 3.7, | |
| "learning_rate": 1.9054054054054053e-05, | |
| "loss": 2.4865, | |
| "step": 23450 | |
| }, | |
| { | |
| "epoch": 3.71, | |
| "learning_rate": 1.90302066772655e-05, | |
| "loss": 2.5481, | |
| "step": 23500 | |
| }, | |
| { | |
| "epoch": 3.71, | |
| "learning_rate": 1.9006359300476948e-05, | |
| "loss": 2.5746, | |
| "step": 23550 | |
| }, | |
| { | |
| "epoch": 3.72, | |
| "learning_rate": 1.8982511923688396e-05, | |
| "loss": 2.6428, | |
| "step": 23600 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 1.895866454689984e-05, | |
| "loss": 2.55, | |
| "step": 23650 | |
| }, | |
| { | |
| "epoch": 3.74, | |
| "learning_rate": 1.8934817170111287e-05, | |
| "loss": 2.5699, | |
| "step": 23700 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 1.8910969793322738e-05, | |
| "loss": 2.6442, | |
| "step": 23750 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 1.8887122416534182e-05, | |
| "loss": 2.5961, | |
| "step": 23800 | |
| }, | |
| { | |
| "epoch": 3.76, | |
| "learning_rate": 1.886327503974563e-05, | |
| "loss": 2.6291, | |
| "step": 23850 | |
| }, | |
| { | |
| "epoch": 3.77, | |
| "learning_rate": 1.8839427662957074e-05, | |
| "loss": 2.6604, | |
| "step": 23900 | |
| }, | |
| { | |
| "epoch": 3.78, | |
| "learning_rate": 1.8815580286168525e-05, | |
| "loss": 2.676, | |
| "step": 23950 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 1.879173290937997e-05, | |
| "loss": 2.6272, | |
| "step": 24000 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 1.8767885532591416e-05, | |
| "loss": 2.6494, | |
| "step": 24050 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 1.874403815580286e-05, | |
| "loss": 2.5214, | |
| "step": 24100 | |
| }, | |
| { | |
| "epoch": 3.81, | |
| "learning_rate": 1.872019077901431e-05, | |
| "loss": 2.6031, | |
| "step": 24150 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "learning_rate": 1.8696343402225755e-05, | |
| "loss": 2.6587, | |
| "step": 24200 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "learning_rate": 1.8672496025437203e-05, | |
| "loss": 2.5222, | |
| "step": 24250 | |
| }, | |
| { | |
| "epoch": 3.83, | |
| "learning_rate": 1.864864864864865e-05, | |
| "loss": 2.6324, | |
| "step": 24300 | |
| }, | |
| { | |
| "epoch": 3.84, | |
| "learning_rate": 1.8624801271860098e-05, | |
| "loss": 2.5523, | |
| "step": 24350 | |
| }, | |
| { | |
| "epoch": 3.85, | |
| "learning_rate": 1.860095389507154e-05, | |
| "loss": 2.5159, | |
| "step": 24400 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 1.857710651828299e-05, | |
| "loss": 2.6001, | |
| "step": 24450 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 1.8553259141494437e-05, | |
| "loss": 2.5723, | |
| "step": 24500 | |
| }, | |
| { | |
| "epoch": 3.87, | |
| "learning_rate": 1.8529411764705884e-05, | |
| "loss": 2.5136, | |
| "step": 24550 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 1.8505564387917328e-05, | |
| "loss": 2.5974, | |
| "step": 24600 | |
| }, | |
| { | |
| "epoch": 3.89, | |
| "learning_rate": 1.8481717011128776e-05, | |
| "loss": 2.4753, | |
| "step": 24650 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "learning_rate": 1.8457869634340223e-05, | |
| "loss": 2.5379, | |
| "step": 24700 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "learning_rate": 1.843402225755167e-05, | |
| "loss": 2.5857, | |
| "step": 24750 | |
| }, | |
| { | |
| "epoch": 3.91, | |
| "learning_rate": 1.8410174880763115e-05, | |
| "loss": 2.5175, | |
| "step": 24800 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 1.8386327503974562e-05, | |
| "loss": 2.617, | |
| "step": 24850 | |
| }, | |
| { | |
| "epoch": 3.93, | |
| "learning_rate": 1.836248012718601e-05, | |
| "loss": 2.5201, | |
| "step": 24900 | |
| }, | |
| { | |
| "epoch": 3.94, | |
| "learning_rate": 1.8338632750397457e-05, | |
| "loss": 2.6352, | |
| "step": 24950 | |
| }, | |
| { | |
| "epoch": 3.94, | |
| "learning_rate": 1.83147853736089e-05, | |
| "loss": 2.5455, | |
| "step": 25000 | |
| }, | |
| { | |
| "epoch": 3.95, | |
| "learning_rate": 1.8290937996820352e-05, | |
| "loss": 2.5444, | |
| "step": 25050 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "learning_rate": 1.8267090620031796e-05, | |
| "loss": 2.6286, | |
| "step": 25100 | |
| }, | |
| { | |
| "epoch": 3.97, | |
| "learning_rate": 1.8243243243243244e-05, | |
| "loss": 2.6376, | |
| "step": 25150 | |
| }, | |
| { | |
| "epoch": 3.97, | |
| "learning_rate": 1.8219395866454688e-05, | |
| "loss": 2.4895, | |
| "step": 25200 | |
| }, | |
| { | |
| "epoch": 3.98, | |
| "learning_rate": 1.819554848966614e-05, | |
| "loss": 2.4588, | |
| "step": 25250 | |
| }, | |
| { | |
| "epoch": 3.99, | |
| "learning_rate": 1.8171701112877583e-05, | |
| "loss": 2.5429, | |
| "step": 25300 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 1.814785373608903e-05, | |
| "loss": 2.5472, | |
| "step": 25350 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 1.8124006359300478e-05, | |
| "loss": 2.3872, | |
| "step": 25400 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 1.8100158982511925e-05, | |
| "loss": 2.4872, | |
| "step": 25450 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "learning_rate": 1.8076311605723373e-05, | |
| "loss": 2.4989, | |
| "step": 25500 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "learning_rate": 1.8052464228934817e-05, | |
| "loss": 2.5616, | |
| "step": 25550 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "learning_rate": 1.8028616852146264e-05, | |
| "loss": 2.544, | |
| "step": 25600 | |
| }, | |
| { | |
| "epoch": 4.05, | |
| "learning_rate": 1.800476947535771e-05, | |
| "loss": 2.5039, | |
| "step": 25650 | |
| }, | |
| { | |
| "epoch": 4.05, | |
| "learning_rate": 1.798092209856916e-05, | |
| "loss": 2.5312, | |
| "step": 25700 | |
| }, | |
| { | |
| "epoch": 4.06, | |
| "learning_rate": 1.7957074721780603e-05, | |
| "loss": 2.5458, | |
| "step": 25750 | |
| }, | |
| { | |
| "epoch": 4.07, | |
| "learning_rate": 1.7933227344992054e-05, | |
| "loss": 2.4705, | |
| "step": 25800 | |
| }, | |
| { | |
| "epoch": 4.08, | |
| "learning_rate": 1.7909379968203498e-05, | |
| "loss": 2.5152, | |
| "step": 25850 | |
| }, | |
| { | |
| "epoch": 4.09, | |
| "learning_rate": 1.7885532591414946e-05, | |
| "loss": 2.552, | |
| "step": 25900 | |
| }, | |
| { | |
| "epoch": 4.09, | |
| "learning_rate": 1.786168521462639e-05, | |
| "loss": 2.4723, | |
| "step": 25950 | |
| }, | |
| { | |
| "epoch": 4.1, | |
| "learning_rate": 1.783783783783784e-05, | |
| "loss": 2.6591, | |
| "step": 26000 | |
| }, | |
| { | |
| "epoch": 4.11, | |
| "learning_rate": 1.7813990461049285e-05, | |
| "loss": 2.5466, | |
| "step": 26050 | |
| }, | |
| { | |
| "epoch": 4.12, | |
| "learning_rate": 1.7790143084260732e-05, | |
| "loss": 2.5868, | |
| "step": 26100 | |
| }, | |
| { | |
| "epoch": 4.12, | |
| "learning_rate": 1.7766295707472176e-05, | |
| "loss": 2.5455, | |
| "step": 26150 | |
| }, | |
| { | |
| "epoch": 4.13, | |
| "learning_rate": 1.7742448330683627e-05, | |
| "loss": 2.5525, | |
| "step": 26200 | |
| }, | |
| { | |
| "epoch": 4.14, | |
| "learning_rate": 1.771860095389507e-05, | |
| "loss": 2.6624, | |
| "step": 26250 | |
| }, | |
| { | |
| "epoch": 4.15, | |
| "learning_rate": 1.769475357710652e-05, | |
| "loss": 2.5108, | |
| "step": 26300 | |
| }, | |
| { | |
| "epoch": 4.16, | |
| "learning_rate": 1.7670906200317966e-05, | |
| "loss": 2.5807, | |
| "step": 26350 | |
| }, | |
| { | |
| "epoch": 4.16, | |
| "learning_rate": 1.7647058823529414e-05, | |
| "loss": 2.5022, | |
| "step": 26400 | |
| }, | |
| { | |
| "epoch": 4.17, | |
| "learning_rate": 1.7623211446740858e-05, | |
| "loss": 2.5007, | |
| "step": 26450 | |
| }, | |
| { | |
| "epoch": 4.18, | |
| "learning_rate": 1.7599364069952305e-05, | |
| "loss": 2.6456, | |
| "step": 26500 | |
| }, | |
| { | |
| "epoch": 4.19, | |
| "learning_rate": 1.7575516693163753e-05, | |
| "loss": 2.5869, | |
| "step": 26550 | |
| }, | |
| { | |
| "epoch": 4.2, | |
| "learning_rate": 1.75516693163752e-05, | |
| "loss": 2.5292, | |
| "step": 26600 | |
| }, | |
| { | |
| "epoch": 4.2, | |
| "learning_rate": 1.7527821939586644e-05, | |
| "loss": 2.3513, | |
| "step": 26650 | |
| }, | |
| { | |
| "epoch": 4.21, | |
| "learning_rate": 1.750397456279809e-05, | |
| "loss": 2.5386, | |
| "step": 26700 | |
| }, | |
| { | |
| "epoch": 4.22, | |
| "learning_rate": 1.748012718600954e-05, | |
| "loss": 2.4587, | |
| "step": 26750 | |
| }, | |
| { | |
| "epoch": 4.23, | |
| "learning_rate": 1.7456279809220987e-05, | |
| "loss": 2.5152, | |
| "step": 26800 | |
| }, | |
| { | |
| "epoch": 4.24, | |
| "learning_rate": 1.743243243243243e-05, | |
| "loss": 2.5277, | |
| "step": 26850 | |
| }, | |
| { | |
| "epoch": 4.24, | |
| "learning_rate": 1.7408585055643878e-05, | |
| "loss": 2.549, | |
| "step": 26900 | |
| }, | |
| { | |
| "epoch": 4.25, | |
| "learning_rate": 1.7384737678855326e-05, | |
| "loss": 2.4883, | |
| "step": 26950 | |
| }, | |
| { | |
| "epoch": 4.26, | |
| "learning_rate": 1.7360890302066773e-05, | |
| "loss": 2.5428, | |
| "step": 27000 | |
| }, | |
| { | |
| "epoch": 4.27, | |
| "learning_rate": 1.733704292527822e-05, | |
| "loss": 2.5253, | |
| "step": 27050 | |
| }, | |
| { | |
| "epoch": 4.27, | |
| "learning_rate": 1.7313195548489668e-05, | |
| "loss": 2.5487, | |
| "step": 27100 | |
| }, | |
| { | |
| "epoch": 4.28, | |
| "learning_rate": 1.7289348171701116e-05, | |
| "loss": 2.4613, | |
| "step": 27150 | |
| }, | |
| { | |
| "epoch": 4.29, | |
| "learning_rate": 1.726550079491256e-05, | |
| "loss": 2.5073, | |
| "step": 27200 | |
| }, | |
| { | |
| "epoch": 4.3, | |
| "learning_rate": 1.7241653418124007e-05, | |
| "loss": 2.4965, | |
| "step": 27250 | |
| }, | |
| { | |
| "epoch": 4.31, | |
| "learning_rate": 1.7217806041335455e-05, | |
| "loss": 2.4334, | |
| "step": 27300 | |
| }, | |
| { | |
| "epoch": 4.31, | |
| "learning_rate": 1.7193958664546902e-05, | |
| "loss": 2.4705, | |
| "step": 27350 | |
| }, | |
| { | |
| "epoch": 4.32, | |
| "learning_rate": 1.7170111287758346e-05, | |
| "loss": 2.5386, | |
| "step": 27400 | |
| }, | |
| { | |
| "epoch": 4.33, | |
| "learning_rate": 1.7146263910969794e-05, | |
| "loss": 2.5268, | |
| "step": 27450 | |
| }, | |
| { | |
| "epoch": 4.34, | |
| "learning_rate": 1.712241653418124e-05, | |
| "loss": 2.4655, | |
| "step": 27500 | |
| }, | |
| { | |
| "epoch": 4.35, | |
| "learning_rate": 1.709856915739269e-05, | |
| "loss": 2.485, | |
| "step": 27550 | |
| }, | |
| { | |
| "epoch": 4.35, | |
| "learning_rate": 1.7074721780604133e-05, | |
| "loss": 2.5323, | |
| "step": 27600 | |
| }, | |
| { | |
| "epoch": 4.36, | |
| "learning_rate": 1.705087440381558e-05, | |
| "loss": 2.3916, | |
| "step": 27650 | |
| }, | |
| { | |
| "epoch": 4.37, | |
| "learning_rate": 1.7027027027027028e-05, | |
| "loss": 2.4557, | |
| "step": 27700 | |
| }, | |
| { | |
| "epoch": 4.38, | |
| "learning_rate": 1.7003179650238475e-05, | |
| "loss": 2.2644, | |
| "step": 27750 | |
| }, | |
| { | |
| "epoch": 4.38, | |
| "learning_rate": 1.697933227344992e-05, | |
| "loss": 2.4271, | |
| "step": 27800 | |
| }, | |
| { | |
| "epoch": 4.39, | |
| "learning_rate": 1.695548489666137e-05, | |
| "loss": 2.4928, | |
| "step": 27850 | |
| }, | |
| { | |
| "epoch": 4.4, | |
| "learning_rate": 1.6931637519872814e-05, | |
| "loss": 2.5587, | |
| "step": 27900 | |
| }, | |
| { | |
| "epoch": 4.41, | |
| "learning_rate": 1.690779014308426e-05, | |
| "loss": 2.4824, | |
| "step": 27950 | |
| }, | |
| { | |
| "epoch": 4.42, | |
| "learning_rate": 1.6883942766295706e-05, | |
| "loss": 2.5206, | |
| "step": 28000 | |
| }, | |
| { | |
| "epoch": 4.42, | |
| "learning_rate": 1.6860095389507157e-05, | |
| "loss": 2.4376, | |
| "step": 28050 | |
| }, | |
| { | |
| "epoch": 4.43, | |
| "learning_rate": 1.68362480127186e-05, | |
| "loss": 2.4982, | |
| "step": 28100 | |
| }, | |
| { | |
| "epoch": 4.44, | |
| "learning_rate": 1.6812400635930048e-05, | |
| "loss": 2.4095, | |
| "step": 28150 | |
| }, | |
| { | |
| "epoch": 4.45, | |
| "learning_rate": 1.6788553259141492e-05, | |
| "loss": 2.4715, | |
| "step": 28200 | |
| }, | |
| { | |
| "epoch": 4.46, | |
| "learning_rate": 1.6764705882352943e-05, | |
| "loss": 2.5444, | |
| "step": 28250 | |
| }, | |
| { | |
| "epoch": 4.46, | |
| "learning_rate": 1.6740858505564387e-05, | |
| "loss": 2.4082, | |
| "step": 28300 | |
| }, | |
| { | |
| "epoch": 4.47, | |
| "learning_rate": 1.6717011128775835e-05, | |
| "loss": 2.4436, | |
| "step": 28350 | |
| }, | |
| { | |
| "epoch": 4.48, | |
| "learning_rate": 1.669316375198728e-05, | |
| "loss": 2.5415, | |
| "step": 28400 | |
| }, | |
| { | |
| "epoch": 4.49, | |
| "learning_rate": 1.666931637519873e-05, | |
| "loss": 2.5421, | |
| "step": 28450 | |
| }, | |
| { | |
| "epoch": 4.5, | |
| "learning_rate": 1.6645468998410174e-05, | |
| "loss": 2.5493, | |
| "step": 28500 | |
| }, | |
| { | |
| "epoch": 4.5, | |
| "learning_rate": 1.662162162162162e-05, | |
| "loss": 2.556, | |
| "step": 28550 | |
| }, | |
| { | |
| "epoch": 4.51, | |
| "learning_rate": 1.659777424483307e-05, | |
| "loss": 2.4381, | |
| "step": 28600 | |
| }, | |
| { | |
| "epoch": 4.52, | |
| "learning_rate": 1.6573926868044516e-05, | |
| "loss": 2.6628, | |
| "step": 28650 | |
| }, | |
| { | |
| "epoch": 4.53, | |
| "learning_rate": 1.6550079491255964e-05, | |
| "loss": 2.503, | |
| "step": 28700 | |
| }, | |
| { | |
| "epoch": 4.53, | |
| "learning_rate": 1.6526232114467408e-05, | |
| "loss": 2.5354, | |
| "step": 28750 | |
| }, | |
| { | |
| "epoch": 4.54, | |
| "learning_rate": 1.650238473767886e-05, | |
| "loss": 2.4082, | |
| "step": 28800 | |
| }, | |
| { | |
| "epoch": 4.55, | |
| "learning_rate": 1.6478537360890303e-05, | |
| "loss": 2.4708, | |
| "step": 28850 | |
| }, | |
| { | |
| "epoch": 4.56, | |
| "learning_rate": 1.645468998410175e-05, | |
| "loss": 2.441, | |
| "step": 28900 | |
| }, | |
| { | |
| "epoch": 4.57, | |
| "learning_rate": 1.6430842607313194e-05, | |
| "loss": 2.4419, | |
| "step": 28950 | |
| }, | |
| { | |
| "epoch": 4.57, | |
| "learning_rate": 1.6406995230524645e-05, | |
| "loss": 2.5285, | |
| "step": 29000 | |
| }, | |
| { | |
| "epoch": 4.58, | |
| "learning_rate": 1.638314785373609e-05, | |
| "loss": 2.4596, | |
| "step": 29050 | |
| }, | |
| { | |
| "epoch": 4.59, | |
| "learning_rate": 1.6359300476947537e-05, | |
| "loss": 2.5179, | |
| "step": 29100 | |
| }, | |
| { | |
| "epoch": 4.6, | |
| "learning_rate": 1.6335453100158984e-05, | |
| "loss": 2.4328, | |
| "step": 29150 | |
| }, | |
| { | |
| "epoch": 4.61, | |
| "learning_rate": 1.631160572337043e-05, | |
| "loss": 2.4823, | |
| "step": 29200 | |
| }, | |
| { | |
| "epoch": 4.61, | |
| "learning_rate": 1.6287758346581876e-05, | |
| "loss": 2.4784, | |
| "step": 29250 | |
| }, | |
| { | |
| "epoch": 4.62, | |
| "learning_rate": 1.6263910969793323e-05, | |
| "loss": 2.4365, | |
| "step": 29300 | |
| }, | |
| { | |
| "epoch": 4.63, | |
| "learning_rate": 1.624006359300477e-05, | |
| "loss": 2.4199, | |
| "step": 29350 | |
| }, | |
| { | |
| "epoch": 4.64, | |
| "learning_rate": 1.6216216216216218e-05, | |
| "loss": 2.5584, | |
| "step": 29400 | |
| }, | |
| { | |
| "epoch": 4.65, | |
| "learning_rate": 1.6192368839427662e-05, | |
| "loss": 2.4681, | |
| "step": 29450 | |
| }, | |
| { | |
| "epoch": 4.65, | |
| "learning_rate": 1.616852146263911e-05, | |
| "loss": 2.6135, | |
| "step": 29500 | |
| }, | |
| { | |
| "epoch": 4.66, | |
| "learning_rate": 1.6144674085850557e-05, | |
| "loss": 2.4771, | |
| "step": 29550 | |
| }, | |
| { | |
| "epoch": 4.67, | |
| "learning_rate": 1.6120826709062005e-05, | |
| "loss": 2.4149, | |
| "step": 29600 | |
| }, | |
| { | |
| "epoch": 4.68, | |
| "learning_rate": 1.609697933227345e-05, | |
| "loss": 2.5233, | |
| "step": 29650 | |
| }, | |
| { | |
| "epoch": 4.68, | |
| "learning_rate": 1.6073131955484896e-05, | |
| "loss": 2.5135, | |
| "step": 29700 | |
| }, | |
| { | |
| "epoch": 4.69, | |
| "learning_rate": 1.6049284578696344e-05, | |
| "loss": 2.458, | |
| "step": 29750 | |
| }, | |
| { | |
| "epoch": 4.7, | |
| "learning_rate": 1.602543720190779e-05, | |
| "loss": 2.4923, | |
| "step": 29800 | |
| }, | |
| { | |
| "epoch": 4.71, | |
| "learning_rate": 1.6001589825119235e-05, | |
| "loss": 2.508, | |
| "step": 29850 | |
| }, | |
| { | |
| "epoch": 4.72, | |
| "learning_rate": 1.5977742448330686e-05, | |
| "loss": 2.4119, | |
| "step": 29900 | |
| }, | |
| { | |
| "epoch": 4.72, | |
| "learning_rate": 1.595389507154213e-05, | |
| "loss": 2.491, | |
| "step": 29950 | |
| }, | |
| { | |
| "epoch": 4.73, | |
| "learning_rate": 1.5930047694753578e-05, | |
| "loss": 2.437, | |
| "step": 30000 | |
| }, | |
| { | |
| "epoch": 4.74, | |
| "learning_rate": 1.5906200317965022e-05, | |
| "loss": 2.4738, | |
| "step": 30050 | |
| }, | |
| { | |
| "epoch": 4.75, | |
| "learning_rate": 1.5882352941176473e-05, | |
| "loss": 2.3932, | |
| "step": 30100 | |
| }, | |
| { | |
| "epoch": 4.76, | |
| "learning_rate": 1.5858505564387917e-05, | |
| "loss": 2.4635, | |
| "step": 30150 | |
| }, | |
| { | |
| "epoch": 4.76, | |
| "learning_rate": 1.5834658187599364e-05, | |
| "loss": 2.4331, | |
| "step": 30200 | |
| }, | |
| { | |
| "epoch": 4.77, | |
| "learning_rate": 1.5810810810810808e-05, | |
| "loss": 2.4436, | |
| "step": 30250 | |
| }, | |
| { | |
| "epoch": 4.78, | |
| "learning_rate": 1.578696343402226e-05, | |
| "loss": 2.4471, | |
| "step": 30300 | |
| }, | |
| { | |
| "epoch": 4.79, | |
| "learning_rate": 1.5763116057233707e-05, | |
| "loss": 2.5101, | |
| "step": 30350 | |
| }, | |
| { | |
| "epoch": 4.79, | |
| "learning_rate": 1.573926868044515e-05, | |
| "loss": 2.4395, | |
| "step": 30400 | |
| }, | |
| { | |
| "epoch": 4.8, | |
| "learning_rate": 1.5715421303656598e-05, | |
| "loss": 2.5001, | |
| "step": 30450 | |
| }, | |
| { | |
| "epoch": 4.81, | |
| "learning_rate": 1.5691573926868046e-05, | |
| "loss": 2.5082, | |
| "step": 30500 | |
| }, | |
| { | |
| "epoch": 4.82, | |
| "learning_rate": 1.5667726550079493e-05, | |
| "loss": 2.4692, | |
| "step": 30550 | |
| }, | |
| { | |
| "epoch": 4.83, | |
| "learning_rate": 1.5643879173290937e-05, | |
| "loss": 2.3858, | |
| "step": 30600 | |
| }, | |
| { | |
| "epoch": 4.83, | |
| "learning_rate": 1.5620031796502388e-05, | |
| "loss": 2.5594, | |
| "step": 30650 | |
| }, | |
| { | |
| "epoch": 4.84, | |
| "learning_rate": 1.5596184419713832e-05, | |
| "loss": 2.4556, | |
| "step": 30700 | |
| }, | |
| { | |
| "epoch": 4.85, | |
| "learning_rate": 1.557233704292528e-05, | |
| "loss": 2.5237, | |
| "step": 30750 | |
| }, | |
| { | |
| "epoch": 4.86, | |
| "learning_rate": 1.5548489666136724e-05, | |
| "loss": 2.582, | |
| "step": 30800 | |
| }, | |
| { | |
| "epoch": 4.87, | |
| "learning_rate": 1.5524642289348175e-05, | |
| "loss": 2.4763, | |
| "step": 30850 | |
| }, | |
| { | |
| "epoch": 4.87, | |
| "learning_rate": 1.550079491255962e-05, | |
| "loss": 2.4364, | |
| "step": 30900 | |
| }, | |
| { | |
| "epoch": 4.88, | |
| "learning_rate": 1.5476947535771066e-05, | |
| "loss": 2.44, | |
| "step": 30950 | |
| }, | |
| { | |
| "epoch": 4.89, | |
| "learning_rate": 1.545310015898251e-05, | |
| "loss": 2.5285, | |
| "step": 31000 | |
| }, | |
| { | |
| "epoch": 4.9, | |
| "learning_rate": 1.542925278219396e-05, | |
| "loss": 2.4616, | |
| "step": 31050 | |
| }, | |
| { | |
| "epoch": 4.91, | |
| "learning_rate": 1.5405405405405405e-05, | |
| "loss": 2.5149, | |
| "step": 31100 | |
| }, | |
| { | |
| "epoch": 4.91, | |
| "learning_rate": 1.5381558028616853e-05, | |
| "loss": 2.4926, | |
| "step": 31150 | |
| }, | |
| { | |
| "epoch": 4.92, | |
| "learning_rate": 1.5357710651828297e-05, | |
| "loss": 2.4282, | |
| "step": 31200 | |
| }, | |
| { | |
| "epoch": 4.93, | |
| "learning_rate": 1.5333863275039748e-05, | |
| "loss": 2.4479, | |
| "step": 31250 | |
| }, | |
| { | |
| "epoch": 4.94, | |
| "learning_rate": 1.531001589825119e-05, | |
| "loss": 2.517, | |
| "step": 31300 | |
| }, | |
| { | |
| "epoch": 4.94, | |
| "learning_rate": 1.528616852146264e-05, | |
| "loss": 2.5034, | |
| "step": 31350 | |
| }, | |
| { | |
| "epoch": 4.95, | |
| "learning_rate": 1.5262321144674087e-05, | |
| "loss": 2.4819, | |
| "step": 31400 | |
| }, | |
| { | |
| "epoch": 4.96, | |
| "learning_rate": 1.5238473767885534e-05, | |
| "loss": 2.4886, | |
| "step": 31450 | |
| }, | |
| { | |
| "epoch": 4.97, | |
| "learning_rate": 1.5214626391096978e-05, | |
| "loss": 2.4712, | |
| "step": 31500 | |
| }, | |
| { | |
| "epoch": 4.98, | |
| "learning_rate": 1.5190779014308427e-05, | |
| "loss": 2.5013, | |
| "step": 31550 | |
| }, | |
| { | |
| "epoch": 4.98, | |
| "learning_rate": 1.5166931637519871e-05, | |
| "loss": 2.5287, | |
| "step": 31600 | |
| }, | |
| { | |
| "epoch": 4.99, | |
| "learning_rate": 1.514308426073132e-05, | |
| "loss": 2.4269, | |
| "step": 31650 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 1.5119236883942765e-05, | |
| "loss": 2.4604, | |
| "step": 31700 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 1.5095389507154214e-05, | |
| "loss": 2.3029, | |
| "step": 31750 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "learning_rate": 1.5071542130365658e-05, | |
| "loss": 2.4492, | |
| "step": 31800 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "learning_rate": 1.5047694753577107e-05, | |
| "loss": 2.4764, | |
| "step": 31850 | |
| }, | |
| { | |
| "epoch": 5.03, | |
| "learning_rate": 1.5023847376788551e-05, | |
| "loss": 2.5057, | |
| "step": 31900 | |
| }, | |
| { | |
| "epoch": 5.04, | |
| "learning_rate": 1.5e-05, | |
| "loss": 2.4618, | |
| "step": 31950 | |
| }, | |
| { | |
| "epoch": 5.05, | |
| "learning_rate": 1.4976152623211448e-05, | |
| "loss": 2.4304, | |
| "step": 32000 | |
| }, | |
| { | |
| "epoch": 5.06, | |
| "learning_rate": 1.4952305246422894e-05, | |
| "loss": 2.4029, | |
| "step": 32050 | |
| }, | |
| { | |
| "epoch": 5.06, | |
| "learning_rate": 1.4928457869634341e-05, | |
| "loss": 2.2965, | |
| "step": 32100 | |
| }, | |
| { | |
| "epoch": 5.07, | |
| "learning_rate": 1.4904610492845787e-05, | |
| "loss": 2.3965, | |
| "step": 32150 | |
| }, | |
| { | |
| "epoch": 5.08, | |
| "learning_rate": 1.4880763116057234e-05, | |
| "loss": 2.3282, | |
| "step": 32200 | |
| }, | |
| { | |
| "epoch": 5.09, | |
| "learning_rate": 1.485691573926868e-05, | |
| "loss": 2.4409, | |
| "step": 32250 | |
| }, | |
| { | |
| "epoch": 5.09, | |
| "learning_rate": 1.4833068362480128e-05, | |
| "loss": 2.4228, | |
| "step": 32300 | |
| }, | |
| { | |
| "epoch": 5.1, | |
| "learning_rate": 1.4809220985691573e-05, | |
| "loss": 2.5412, | |
| "step": 32350 | |
| }, | |
| { | |
| "epoch": 5.11, | |
| "learning_rate": 1.4785373608903021e-05, | |
| "loss": 2.374, | |
| "step": 32400 | |
| }, | |
| { | |
| "epoch": 5.12, | |
| "learning_rate": 1.4761526232114467e-05, | |
| "loss": 2.3942, | |
| "step": 32450 | |
| }, | |
| { | |
| "epoch": 5.13, | |
| "learning_rate": 1.4737678855325914e-05, | |
| "loss": 2.3872, | |
| "step": 32500 | |
| }, | |
| { | |
| "epoch": 5.13, | |
| "learning_rate": 1.471383147853736e-05, | |
| "loss": 2.4669, | |
| "step": 32550 | |
| }, | |
| { | |
| "epoch": 5.14, | |
| "learning_rate": 1.4689984101748807e-05, | |
| "loss": 2.549, | |
| "step": 32600 | |
| }, | |
| { | |
| "epoch": 5.15, | |
| "learning_rate": 1.4666136724960255e-05, | |
| "loss": 2.453, | |
| "step": 32650 | |
| }, | |
| { | |
| "epoch": 5.16, | |
| "learning_rate": 1.46422893481717e-05, | |
| "loss": 2.3588, | |
| "step": 32700 | |
| }, | |
| { | |
| "epoch": 5.17, | |
| "learning_rate": 1.4618441971383148e-05, | |
| "loss": 2.5884, | |
| "step": 32750 | |
| }, | |
| { | |
| "epoch": 5.17, | |
| "learning_rate": 1.4594594594594596e-05, | |
| "loss": 2.442, | |
| "step": 32800 | |
| }, | |
| { | |
| "epoch": 5.18, | |
| "learning_rate": 1.4570747217806043e-05, | |
| "loss": 2.3182, | |
| "step": 32850 | |
| }, | |
| { | |
| "epoch": 5.19, | |
| "learning_rate": 1.4546899841017489e-05, | |
| "loss": 2.4851, | |
| "step": 32900 | |
| }, | |
| { | |
| "epoch": 5.2, | |
| "learning_rate": 1.4523052464228936e-05, | |
| "loss": 2.3914, | |
| "step": 32950 | |
| }, | |
| { | |
| "epoch": 5.21, | |
| "learning_rate": 1.4499205087440382e-05, | |
| "loss": 2.4582, | |
| "step": 33000 | |
| }, | |
| { | |
| "epoch": 5.21, | |
| "learning_rate": 1.447535771065183e-05, | |
| "loss": 2.3891, | |
| "step": 33050 | |
| }, | |
| { | |
| "epoch": 5.22, | |
| "learning_rate": 1.4451510333863275e-05, | |
| "loss": 2.4047, | |
| "step": 33100 | |
| }, | |
| { | |
| "epoch": 5.23, | |
| "learning_rate": 1.4427662957074723e-05, | |
| "loss": 2.4761, | |
| "step": 33150 | |
| }, | |
| { | |
| "epoch": 5.24, | |
| "learning_rate": 1.4403815580286169e-05, | |
| "loss": 2.4103, | |
| "step": 33200 | |
| }, | |
| { | |
| "epoch": 5.24, | |
| "learning_rate": 1.4379968203497616e-05, | |
| "loss": 2.4628, | |
| "step": 33250 | |
| }, | |
| { | |
| "epoch": 5.25, | |
| "learning_rate": 1.4356120826709062e-05, | |
| "loss": 2.3495, | |
| "step": 33300 | |
| }, | |
| { | |
| "epoch": 5.26, | |
| "learning_rate": 1.433227344992051e-05, | |
| "loss": 2.4002, | |
| "step": 33350 | |
| }, | |
| { | |
| "epoch": 5.27, | |
| "learning_rate": 1.4308426073131957e-05, | |
| "loss": 2.3719, | |
| "step": 33400 | |
| }, | |
| { | |
| "epoch": 5.28, | |
| "learning_rate": 1.4284578696343403e-05, | |
| "loss": 2.4141, | |
| "step": 33450 | |
| }, | |
| { | |
| "epoch": 5.28, | |
| "learning_rate": 1.426073131955485e-05, | |
| "loss": 2.4345, | |
| "step": 33500 | |
| }, | |
| { | |
| "epoch": 5.29, | |
| "learning_rate": 1.4236883942766296e-05, | |
| "loss": 2.5029, | |
| "step": 33550 | |
| }, | |
| { | |
| "epoch": 5.3, | |
| "learning_rate": 1.4213036565977743e-05, | |
| "loss": 2.4863, | |
| "step": 33600 | |
| }, | |
| { | |
| "epoch": 5.31, | |
| "learning_rate": 1.4189189189189189e-05, | |
| "loss": 2.4066, | |
| "step": 33650 | |
| }, | |
| { | |
| "epoch": 5.32, | |
| "learning_rate": 1.4165341812400637e-05, | |
| "loss": 2.4462, | |
| "step": 33700 | |
| }, | |
| { | |
| "epoch": 5.32, | |
| "learning_rate": 1.4141494435612082e-05, | |
| "loss": 2.3888, | |
| "step": 33750 | |
| }, | |
| { | |
| "epoch": 5.33, | |
| "learning_rate": 1.411764705882353e-05, | |
| "loss": 2.4463, | |
| "step": 33800 | |
| }, | |
| { | |
| "epoch": 5.34, | |
| "learning_rate": 1.4093799682034976e-05, | |
| "loss": 2.4239, | |
| "step": 33850 | |
| }, | |
| { | |
| "epoch": 5.35, | |
| "learning_rate": 1.4069952305246423e-05, | |
| "loss": 2.4442, | |
| "step": 33900 | |
| }, | |
| { | |
| "epoch": 5.35, | |
| "learning_rate": 1.4046104928457869e-05, | |
| "loss": 2.4302, | |
| "step": 33950 | |
| }, | |
| { | |
| "epoch": 5.36, | |
| "learning_rate": 1.4022257551669316e-05, | |
| "loss": 2.3692, | |
| "step": 34000 | |
| }, | |
| { | |
| "epoch": 5.37, | |
| "learning_rate": 1.3998410174880764e-05, | |
| "loss": 2.3628, | |
| "step": 34050 | |
| }, | |
| { | |
| "epoch": 5.38, | |
| "learning_rate": 1.397456279809221e-05, | |
| "loss": 2.5001, | |
| "step": 34100 | |
| }, | |
| { | |
| "epoch": 5.39, | |
| "learning_rate": 1.3950715421303657e-05, | |
| "loss": 2.4981, | |
| "step": 34150 | |
| }, | |
| { | |
| "epoch": 5.39, | |
| "learning_rate": 1.3926868044515103e-05, | |
| "loss": 2.4526, | |
| "step": 34200 | |
| }, | |
| { | |
| "epoch": 5.4, | |
| "learning_rate": 1.390302066772655e-05, | |
| "loss": 2.3627, | |
| "step": 34250 | |
| }, | |
| { | |
| "epoch": 5.41, | |
| "learning_rate": 1.3879173290937996e-05, | |
| "loss": 2.3423, | |
| "step": 34300 | |
| }, | |
| { | |
| "epoch": 5.42, | |
| "learning_rate": 1.3855325914149444e-05, | |
| "loss": 2.3372, | |
| "step": 34350 | |
| }, | |
| { | |
| "epoch": 5.43, | |
| "learning_rate": 1.383147853736089e-05, | |
| "loss": 2.3553, | |
| "step": 34400 | |
| }, | |
| { | |
| "epoch": 5.43, | |
| "learning_rate": 1.3807631160572339e-05, | |
| "loss": 2.424, | |
| "step": 34450 | |
| }, | |
| { | |
| "epoch": 5.44, | |
| "learning_rate": 1.3783783783783784e-05, | |
| "loss": 2.3868, | |
| "step": 34500 | |
| }, | |
| { | |
| "epoch": 5.45, | |
| "learning_rate": 1.3759936406995232e-05, | |
| "loss": 2.4128, | |
| "step": 34550 | |
| }, | |
| { | |
| "epoch": 5.46, | |
| "learning_rate": 1.3736089030206678e-05, | |
| "loss": 2.3561, | |
| "step": 34600 | |
| }, | |
| { | |
| "epoch": 5.47, | |
| "learning_rate": 1.3712241653418125e-05, | |
| "loss": 2.5233, | |
| "step": 34650 | |
| }, | |
| { | |
| "epoch": 5.47, | |
| "learning_rate": 1.3688394276629571e-05, | |
| "loss": 2.4911, | |
| "step": 34700 | |
| }, | |
| { | |
| "epoch": 5.48, | |
| "learning_rate": 1.3664546899841018e-05, | |
| "loss": 2.387, | |
| "step": 34750 | |
| }, | |
| { | |
| "epoch": 5.49, | |
| "learning_rate": 1.3640699523052466e-05, | |
| "loss": 2.4505, | |
| "step": 34800 | |
| }, | |
| { | |
| "epoch": 5.5, | |
| "learning_rate": 1.3616852146263912e-05, | |
| "loss": 2.4076, | |
| "step": 34850 | |
| }, | |
| { | |
| "epoch": 5.5, | |
| "learning_rate": 1.3593004769475359e-05, | |
| "loss": 2.3071, | |
| "step": 34900 | |
| }, | |
| { | |
| "epoch": 5.51, | |
| "learning_rate": 1.3569157392686805e-05, | |
| "loss": 2.531, | |
| "step": 34950 | |
| }, | |
| { | |
| "epoch": 5.52, | |
| "learning_rate": 1.3545310015898252e-05, | |
| "loss": 2.5236, | |
| "step": 35000 | |
| }, | |
| { | |
| "epoch": 5.53, | |
| "learning_rate": 1.3521462639109698e-05, | |
| "loss": 2.4052, | |
| "step": 35050 | |
| }, | |
| { | |
| "epoch": 5.54, | |
| "learning_rate": 1.3497615262321146e-05, | |
| "loss": 2.4237, | |
| "step": 35100 | |
| }, | |
| { | |
| "epoch": 5.54, | |
| "learning_rate": 1.3473767885532591e-05, | |
| "loss": 2.4589, | |
| "step": 35150 | |
| }, | |
| { | |
| "epoch": 5.55, | |
| "learning_rate": 1.3449920508744039e-05, | |
| "loss": 2.3325, | |
| "step": 35200 | |
| }, | |
| { | |
| "epoch": 5.56, | |
| "learning_rate": 1.3426073131955485e-05, | |
| "loss": 2.3614, | |
| "step": 35250 | |
| }, | |
| { | |
| "epoch": 5.57, | |
| "learning_rate": 1.3402225755166932e-05, | |
| "loss": 2.4497, | |
| "step": 35300 | |
| }, | |
| { | |
| "epoch": 5.58, | |
| "learning_rate": 1.3378378378378378e-05, | |
| "loss": 2.4434, | |
| "step": 35350 | |
| }, | |
| { | |
| "epoch": 5.58, | |
| "learning_rate": 1.3354531001589825e-05, | |
| "loss": 2.4159, | |
| "step": 35400 | |
| }, | |
| { | |
| "epoch": 5.59, | |
| "learning_rate": 1.3330683624801273e-05, | |
| "loss": 2.4629, | |
| "step": 35450 | |
| }, | |
| { | |
| "epoch": 5.6, | |
| "learning_rate": 1.3306836248012719e-05, | |
| "loss": 2.383, | |
| "step": 35500 | |
| }, | |
| { | |
| "epoch": 5.61, | |
| "learning_rate": 1.3282988871224166e-05, | |
| "loss": 2.498, | |
| "step": 35550 | |
| }, | |
| { | |
| "epoch": 5.62, | |
| "learning_rate": 1.3259141494435612e-05, | |
| "loss": 2.3559, | |
| "step": 35600 | |
| }, | |
| { | |
| "epoch": 5.62, | |
| "learning_rate": 1.323529411764706e-05, | |
| "loss": 2.389, | |
| "step": 35650 | |
| }, | |
| { | |
| "epoch": 5.63, | |
| "learning_rate": 1.3211446740858505e-05, | |
| "loss": 2.367, | |
| "step": 35700 | |
| }, | |
| { | |
| "epoch": 5.64, | |
| "learning_rate": 1.3187599364069953e-05, | |
| "loss": 2.2969, | |
| "step": 35750 | |
| }, | |
| { | |
| "epoch": 5.65, | |
| "learning_rate": 1.3163751987281398e-05, | |
| "loss": 2.395, | |
| "step": 35800 | |
| }, | |
| { | |
| "epoch": 5.65, | |
| "learning_rate": 1.3139904610492846e-05, | |
| "loss": 2.3354, | |
| "step": 35850 | |
| }, | |
| { | |
| "epoch": 5.66, | |
| "learning_rate": 1.3116057233704292e-05, | |
| "loss": 2.3289, | |
| "step": 35900 | |
| }, | |
| { | |
| "epoch": 5.67, | |
| "learning_rate": 1.3092209856915739e-05, | |
| "loss": 2.4246, | |
| "step": 35950 | |
| }, | |
| { | |
| "epoch": 5.68, | |
| "learning_rate": 1.3068362480127185e-05, | |
| "loss": 2.5001, | |
| "step": 36000 | |
| }, | |
| { | |
| "epoch": 5.69, | |
| "learning_rate": 1.3044515103338632e-05, | |
| "loss": 2.5762, | |
| "step": 36050 | |
| }, | |
| { | |
| "epoch": 5.69, | |
| "learning_rate": 1.302066772655008e-05, | |
| "loss": 2.358, | |
| "step": 36100 | |
| }, | |
| { | |
| "epoch": 5.7, | |
| "learning_rate": 1.2996820349761527e-05, | |
| "loss": 2.422, | |
| "step": 36150 | |
| }, | |
| { | |
| "epoch": 5.71, | |
| "learning_rate": 1.2972972972972975e-05, | |
| "loss": 2.4942, | |
| "step": 36200 | |
| }, | |
| { | |
| "epoch": 5.72, | |
| "learning_rate": 1.294912559618442e-05, | |
| "loss": 2.2875, | |
| "step": 36250 | |
| }, | |
| { | |
| "epoch": 5.73, | |
| "learning_rate": 1.2925278219395868e-05, | |
| "loss": 2.3661, | |
| "step": 36300 | |
| }, | |
| { | |
| "epoch": 5.73, | |
| "learning_rate": 1.2901430842607314e-05, | |
| "loss": 2.405, | |
| "step": 36350 | |
| }, | |
| { | |
| "epoch": 5.74, | |
| "learning_rate": 1.2877583465818761e-05, | |
| "loss": 2.407, | |
| "step": 36400 | |
| }, | |
| { | |
| "epoch": 5.75, | |
| "learning_rate": 1.2853736089030207e-05, | |
| "loss": 2.4231, | |
| "step": 36450 | |
| }, | |
| { | |
| "epoch": 5.76, | |
| "learning_rate": 1.2829888712241655e-05, | |
| "loss": 2.3957, | |
| "step": 36500 | |
| }, | |
| { | |
| "epoch": 5.76, | |
| "learning_rate": 1.28060413354531e-05, | |
| "loss": 2.4263, | |
| "step": 36550 | |
| }, | |
| { | |
| "epoch": 5.77, | |
| "learning_rate": 1.2782193958664548e-05, | |
| "loss": 2.4255, | |
| "step": 36600 | |
| }, | |
| { | |
| "epoch": 5.78, | |
| "learning_rate": 1.2758346581875994e-05, | |
| "loss": 2.3842, | |
| "step": 36650 | |
| }, | |
| { | |
| "epoch": 5.79, | |
| "learning_rate": 1.2734499205087441e-05, | |
| "loss": 2.3538, | |
| "step": 36700 | |
| }, | |
| { | |
| "epoch": 5.8, | |
| "learning_rate": 1.2710651828298887e-05, | |
| "loss": 2.4627, | |
| "step": 36750 | |
| }, | |
| { | |
| "epoch": 5.8, | |
| "learning_rate": 1.2686804451510334e-05, | |
| "loss": 2.4949, | |
| "step": 36800 | |
| }, | |
| { | |
| "epoch": 5.81, | |
| "learning_rate": 1.2662957074721782e-05, | |
| "loss": 2.3725, | |
| "step": 36850 | |
| }, | |
| { | |
| "epoch": 5.82, | |
| "learning_rate": 1.2639109697933228e-05, | |
| "loss": 2.4256, | |
| "step": 36900 | |
| }, | |
| { | |
| "epoch": 5.83, | |
| "learning_rate": 1.2615262321144675e-05, | |
| "loss": 2.5716, | |
| "step": 36950 | |
| }, | |
| { | |
| "epoch": 5.84, | |
| "learning_rate": 1.2591414944356121e-05, | |
| "loss": 2.4394, | |
| "step": 37000 | |
| }, | |
| { | |
| "epoch": 5.84, | |
| "learning_rate": 1.2567567567567568e-05, | |
| "loss": 2.4445, | |
| "step": 37050 | |
| }, | |
| { | |
| "epoch": 5.85, | |
| "learning_rate": 1.2543720190779014e-05, | |
| "loss": 2.4336, | |
| "step": 37100 | |
| }, | |
| { | |
| "epoch": 5.86, | |
| "learning_rate": 1.2519872813990462e-05, | |
| "loss": 2.4754, | |
| "step": 37150 | |
| }, | |
| { | |
| "epoch": 5.87, | |
| "learning_rate": 1.2496025437201907e-05, | |
| "loss": 2.4117, | |
| "step": 37200 | |
| }, | |
| { | |
| "epoch": 5.88, | |
| "learning_rate": 1.2472178060413355e-05, | |
| "loss": 2.3351, | |
| "step": 37250 | |
| }, | |
| { | |
| "epoch": 5.88, | |
| "learning_rate": 1.24483306836248e-05, | |
| "loss": 2.2414, | |
| "step": 37300 | |
| }, | |
| { | |
| "epoch": 5.89, | |
| "learning_rate": 1.2424483306836248e-05, | |
| "loss": 2.4939, | |
| "step": 37350 | |
| }, | |
| { | |
| "epoch": 5.9, | |
| "learning_rate": 1.2400635930047694e-05, | |
| "loss": 2.5218, | |
| "step": 37400 | |
| }, | |
| { | |
| "epoch": 5.91, | |
| "learning_rate": 1.2376788553259141e-05, | |
| "loss": 2.4034, | |
| "step": 37450 | |
| }, | |
| { | |
| "epoch": 5.91, | |
| "learning_rate": 1.2352941176470587e-05, | |
| "loss": 2.3499, | |
| "step": 37500 | |
| }, | |
| { | |
| "epoch": 5.92, | |
| "learning_rate": 1.2329093799682035e-05, | |
| "loss": 2.3668, | |
| "step": 37550 | |
| }, | |
| { | |
| "epoch": 5.93, | |
| "learning_rate": 1.2305246422893482e-05, | |
| "loss": 2.3756, | |
| "step": 37600 | |
| }, | |
| { | |
| "epoch": 5.94, | |
| "learning_rate": 1.2281399046104928e-05, | |
| "loss": 2.3595, | |
| "step": 37650 | |
| }, | |
| { | |
| "epoch": 5.95, | |
| "learning_rate": 1.2257551669316375e-05, | |
| "loss": 2.5006, | |
| "step": 37700 | |
| }, | |
| { | |
| "epoch": 5.95, | |
| "learning_rate": 1.2233704292527823e-05, | |
| "loss": 2.3216, | |
| "step": 37750 | |
| }, | |
| { | |
| "epoch": 5.96, | |
| "learning_rate": 1.220985691573927e-05, | |
| "loss": 2.3741, | |
| "step": 37800 | |
| }, | |
| { | |
| "epoch": 5.97, | |
| "learning_rate": 1.2186009538950716e-05, | |
| "loss": 2.4229, | |
| "step": 37850 | |
| }, | |
| { | |
| "epoch": 5.98, | |
| "learning_rate": 1.2162162162162164e-05, | |
| "loss": 2.4669, | |
| "step": 37900 | |
| }, | |
| { | |
| "epoch": 5.99, | |
| "learning_rate": 1.213831478537361e-05, | |
| "loss": 2.3883, | |
| "step": 37950 | |
| }, | |
| { | |
| "epoch": 5.99, | |
| "learning_rate": 1.2114467408585057e-05, | |
| "loss": 2.3252, | |
| "step": 38000 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 1.2090620031796503e-05, | |
| "loss": 2.3869, | |
| "step": 38050 | |
| }, | |
| { | |
| "epoch": 6.01, | |
| "learning_rate": 1.206677265500795e-05, | |
| "loss": 2.3895, | |
| "step": 38100 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 1.2042925278219396e-05, | |
| "loss": 2.3348, | |
| "step": 38150 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "learning_rate": 1.2019077901430843e-05, | |
| "loss": 2.3672, | |
| "step": 38200 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "learning_rate": 1.199523052464229e-05, | |
| "loss": 2.4028, | |
| "step": 38250 | |
| }, | |
| { | |
| "epoch": 6.04, | |
| "learning_rate": 1.1971383147853737e-05, | |
| "loss": 2.2395, | |
| "step": 38300 | |
| }, | |
| { | |
| "epoch": 6.05, | |
| "learning_rate": 1.1947535771065184e-05, | |
| "loss": 2.3821, | |
| "step": 38350 | |
| }, | |
| { | |
| "epoch": 6.06, | |
| "learning_rate": 1.192368839427663e-05, | |
| "loss": 2.3787, | |
| "step": 38400 | |
| }, | |
| { | |
| "epoch": 6.06, | |
| "learning_rate": 1.1899841017488077e-05, | |
| "loss": 2.3736, | |
| "step": 38450 | |
| }, | |
| { | |
| "epoch": 6.07, | |
| "learning_rate": 1.1875993640699523e-05, | |
| "loss": 2.2846, | |
| "step": 38500 | |
| }, | |
| { | |
| "epoch": 6.08, | |
| "learning_rate": 1.185214626391097e-05, | |
| "loss": 2.3708, | |
| "step": 38550 | |
| }, | |
| { | |
| "epoch": 6.09, | |
| "learning_rate": 1.1828298887122416e-05, | |
| "loss": 2.2383, | |
| "step": 38600 | |
| }, | |
| { | |
| "epoch": 6.1, | |
| "learning_rate": 1.1804451510333864e-05, | |
| "loss": 2.4514, | |
| "step": 38650 | |
| }, | |
| { | |
| "epoch": 6.1, | |
| "learning_rate": 1.178060413354531e-05, | |
| "loss": 2.3704, | |
| "step": 38700 | |
| }, | |
| { | |
| "epoch": 6.11, | |
| "learning_rate": 1.1756756756756757e-05, | |
| "loss": 2.3119, | |
| "step": 38750 | |
| }, | |
| { | |
| "epoch": 6.12, | |
| "learning_rate": 1.1732909379968203e-05, | |
| "loss": 2.2737, | |
| "step": 38800 | |
| }, | |
| { | |
| "epoch": 6.13, | |
| "learning_rate": 1.170906200317965e-05, | |
| "loss": 2.3966, | |
| "step": 38850 | |
| }, | |
| { | |
| "epoch": 6.14, | |
| "learning_rate": 1.1685214626391096e-05, | |
| "loss": 2.3318, | |
| "step": 38900 | |
| }, | |
| { | |
| "epoch": 6.14, | |
| "learning_rate": 1.1661367249602544e-05, | |
| "loss": 2.4453, | |
| "step": 38950 | |
| }, | |
| { | |
| "epoch": 6.15, | |
| "learning_rate": 1.1637519872813991e-05, | |
| "loss": 2.3973, | |
| "step": 39000 | |
| }, | |
| { | |
| "epoch": 6.16, | |
| "learning_rate": 1.1613672496025437e-05, | |
| "loss": 2.3625, | |
| "step": 39050 | |
| }, | |
| { | |
| "epoch": 6.17, | |
| "learning_rate": 1.1589825119236884e-05, | |
| "loss": 2.3174, | |
| "step": 39100 | |
| }, | |
| { | |
| "epoch": 6.18, | |
| "learning_rate": 1.156597774244833e-05, | |
| "loss": 2.4294, | |
| "step": 39150 | |
| }, | |
| { | |
| "epoch": 6.18, | |
| "learning_rate": 1.1542130365659778e-05, | |
| "loss": 2.3756, | |
| "step": 39200 | |
| }, | |
| { | |
| "epoch": 6.19, | |
| "learning_rate": 1.1518282988871223e-05, | |
| "loss": 2.4119, | |
| "step": 39250 | |
| }, | |
| { | |
| "epoch": 6.2, | |
| "learning_rate": 1.1494435612082671e-05, | |
| "loss": 2.442, | |
| "step": 39300 | |
| }, | |
| { | |
| "epoch": 6.21, | |
| "learning_rate": 1.1470588235294117e-05, | |
| "loss": 2.3622, | |
| "step": 39350 | |
| }, | |
| { | |
| "epoch": 6.21, | |
| "learning_rate": 1.1446740858505566e-05, | |
| "loss": 2.4294, | |
| "step": 39400 | |
| }, | |
| { | |
| "epoch": 6.22, | |
| "learning_rate": 1.1422893481717012e-05, | |
| "loss": 2.3, | |
| "step": 39450 | |
| }, | |
| { | |
| "epoch": 6.23, | |
| "learning_rate": 1.1399046104928459e-05, | |
| "loss": 2.4134, | |
| "step": 39500 | |
| }, | |
| { | |
| "epoch": 6.24, | |
| "learning_rate": 1.1375198728139905e-05, | |
| "loss": 2.3861, | |
| "step": 39550 | |
| }, | |
| { | |
| "epoch": 6.25, | |
| "learning_rate": 1.1351351351351352e-05, | |
| "loss": 2.3675, | |
| "step": 39600 | |
| }, | |
| { | |
| "epoch": 6.25, | |
| "learning_rate": 1.13275039745628e-05, | |
| "loss": 2.3885, | |
| "step": 39650 | |
| }, | |
| { | |
| "epoch": 6.26, | |
| "learning_rate": 1.1303656597774246e-05, | |
| "loss": 2.36, | |
| "step": 39700 | |
| }, | |
| { | |
| "epoch": 6.27, | |
| "learning_rate": 1.1279809220985693e-05, | |
| "loss": 2.3284, | |
| "step": 39750 | |
| }, | |
| { | |
| "epoch": 6.28, | |
| "learning_rate": 1.1255961844197139e-05, | |
| "loss": 2.4368, | |
| "step": 39800 | |
| }, | |
| { | |
| "epoch": 6.29, | |
| "learning_rate": 1.1232114467408586e-05, | |
| "loss": 2.3698, | |
| "step": 39850 | |
| }, | |
| { | |
| "epoch": 6.29, | |
| "learning_rate": 1.1208267090620032e-05, | |
| "loss": 2.28, | |
| "step": 39900 | |
| }, | |
| { | |
| "epoch": 6.3, | |
| "learning_rate": 1.118441971383148e-05, | |
| "loss": 2.4025, | |
| "step": 39950 | |
| }, | |
| { | |
| "epoch": 6.31, | |
| "learning_rate": 1.1160572337042925e-05, | |
| "loss": 2.4193, | |
| "step": 40000 | |
| }, | |
| { | |
| "epoch": 6.32, | |
| "learning_rate": 1.1136724960254373e-05, | |
| "loss": 2.268, | |
| "step": 40050 | |
| }, | |
| { | |
| "epoch": 6.32, | |
| "learning_rate": 1.1112877583465819e-05, | |
| "loss": 2.4122, | |
| "step": 40100 | |
| }, | |
| { | |
| "epoch": 6.33, | |
| "learning_rate": 1.1089030206677266e-05, | |
| "loss": 2.3692, | |
| "step": 40150 | |
| }, | |
| { | |
| "epoch": 6.34, | |
| "learning_rate": 1.1065182829888712e-05, | |
| "loss": 2.3528, | |
| "step": 40200 | |
| }, | |
| { | |
| "epoch": 6.35, | |
| "learning_rate": 1.104133545310016e-05, | |
| "loss": 2.3224, | |
| "step": 40250 | |
| }, | |
| { | |
| "epoch": 6.36, | |
| "learning_rate": 1.1017488076311605e-05, | |
| "loss": 2.4246, | |
| "step": 40300 | |
| }, | |
| { | |
| "epoch": 6.36, | |
| "learning_rate": 1.0993640699523053e-05, | |
| "loss": 2.3906, | |
| "step": 40350 | |
| }, | |
| { | |
| "epoch": 6.37, | |
| "learning_rate": 1.09697933227345e-05, | |
| "loss": 2.4014, | |
| "step": 40400 | |
| }, | |
| { | |
| "epoch": 6.38, | |
| "learning_rate": 1.0945945945945946e-05, | |
| "loss": 2.3314, | |
| "step": 40450 | |
| }, | |
| { | |
| "epoch": 6.39, | |
| "learning_rate": 1.0922098569157393e-05, | |
| "loss": 2.366, | |
| "step": 40500 | |
| }, | |
| { | |
| "epoch": 6.4, | |
| "learning_rate": 1.0898251192368839e-05, | |
| "loss": 2.3566, | |
| "step": 40550 | |
| }, | |
| { | |
| "epoch": 6.4, | |
| "learning_rate": 1.0874403815580287e-05, | |
| "loss": 2.3871, | |
| "step": 40600 | |
| }, | |
| { | |
| "epoch": 6.41, | |
| "learning_rate": 1.0850556438791732e-05, | |
| "loss": 2.276, | |
| "step": 40650 | |
| }, | |
| { | |
| "epoch": 6.42, | |
| "learning_rate": 1.082670906200318e-05, | |
| "loss": 2.3321, | |
| "step": 40700 | |
| }, | |
| { | |
| "epoch": 6.43, | |
| "learning_rate": 1.0802861685214626e-05, | |
| "loss": 2.3235, | |
| "step": 40750 | |
| }, | |
| { | |
| "epoch": 6.44, | |
| "learning_rate": 1.0779014308426073e-05, | |
| "loss": 2.4327, | |
| "step": 40800 | |
| }, | |
| { | |
| "epoch": 6.44, | |
| "learning_rate": 1.0755166931637519e-05, | |
| "loss": 2.2971, | |
| "step": 40850 | |
| }, | |
| { | |
| "epoch": 6.45, | |
| "learning_rate": 1.0731319554848966e-05, | |
| "loss": 2.4482, | |
| "step": 40900 | |
| }, | |
| { | |
| "epoch": 6.46, | |
| "learning_rate": 1.0707472178060412e-05, | |
| "loss": 2.3656, | |
| "step": 40950 | |
| }, | |
| { | |
| "epoch": 6.47, | |
| "learning_rate": 1.068362480127186e-05, | |
| "loss": 2.433, | |
| "step": 41000 | |
| }, | |
| { | |
| "epoch": 6.47, | |
| "learning_rate": 1.0659777424483307e-05, | |
| "loss": 2.4306, | |
| "step": 41050 | |
| }, | |
| { | |
| "epoch": 6.48, | |
| "learning_rate": 1.0635930047694755e-05, | |
| "loss": 2.3191, | |
| "step": 41100 | |
| }, | |
| { | |
| "epoch": 6.49, | |
| "learning_rate": 1.0612082670906202e-05, | |
| "loss": 2.4315, | |
| "step": 41150 | |
| }, | |
| { | |
| "epoch": 6.5, | |
| "learning_rate": 1.0588235294117648e-05, | |
| "loss": 2.3453, | |
| "step": 41200 | |
| }, | |
| { | |
| "epoch": 6.51, | |
| "learning_rate": 1.0564387917329095e-05, | |
| "loss": 2.2802, | |
| "step": 41250 | |
| }, | |
| { | |
| "epoch": 6.51, | |
| "learning_rate": 1.0540540540540541e-05, | |
| "loss": 2.3409, | |
| "step": 41300 | |
| }, | |
| { | |
| "epoch": 6.52, | |
| "learning_rate": 1.0516693163751989e-05, | |
| "loss": 2.354, | |
| "step": 41350 | |
| }, | |
| { | |
| "epoch": 6.53, | |
| "learning_rate": 1.0492845786963434e-05, | |
| "loss": 2.3923, | |
| "step": 41400 | |
| }, | |
| { | |
| "epoch": 6.54, | |
| "learning_rate": 1.0468998410174882e-05, | |
| "loss": 2.3687, | |
| "step": 41450 | |
| }, | |
| { | |
| "epoch": 6.55, | |
| "learning_rate": 1.0445151033386328e-05, | |
| "loss": 2.375, | |
| "step": 41500 | |
| }, | |
| { | |
| "epoch": 6.55, | |
| "learning_rate": 1.0421303656597775e-05, | |
| "loss": 2.3809, | |
| "step": 41550 | |
| }, | |
| { | |
| "epoch": 6.56, | |
| "learning_rate": 1.0397456279809221e-05, | |
| "loss": 2.3784, | |
| "step": 41600 | |
| }, | |
| { | |
| "epoch": 6.57, | |
| "learning_rate": 1.0373608903020668e-05, | |
| "loss": 2.3258, | |
| "step": 41650 | |
| }, | |
| { | |
| "epoch": 6.58, | |
| "learning_rate": 1.0349761526232114e-05, | |
| "loss": 2.3209, | |
| "step": 41700 | |
| }, | |
| { | |
| "epoch": 6.59, | |
| "learning_rate": 1.0325914149443562e-05, | |
| "loss": 2.3521, | |
| "step": 41750 | |
| }, | |
| { | |
| "epoch": 6.59, | |
| "learning_rate": 1.0302066772655009e-05, | |
| "loss": 2.3647, | |
| "step": 41800 | |
| }, | |
| { | |
| "epoch": 6.6, | |
| "learning_rate": 1.0278219395866455e-05, | |
| "loss": 2.3021, | |
| "step": 41850 | |
| }, | |
| { | |
| "epoch": 6.61, | |
| "learning_rate": 1.0254372019077902e-05, | |
| "loss": 2.415, | |
| "step": 41900 | |
| }, | |
| { | |
| "epoch": 6.62, | |
| "learning_rate": 1.0230524642289348e-05, | |
| "loss": 2.3468, | |
| "step": 41950 | |
| }, | |
| { | |
| "epoch": 6.62, | |
| "learning_rate": 1.0206677265500796e-05, | |
| "loss": 2.4086, | |
| "step": 42000 | |
| }, | |
| { | |
| "epoch": 6.63, | |
| "learning_rate": 1.0182829888712241e-05, | |
| "loss": 2.379, | |
| "step": 42050 | |
| }, | |
| { | |
| "epoch": 6.64, | |
| "learning_rate": 1.0158982511923689e-05, | |
| "loss": 2.453, | |
| "step": 42100 | |
| }, | |
| { | |
| "epoch": 6.65, | |
| "learning_rate": 1.0135135135135135e-05, | |
| "loss": 2.4528, | |
| "step": 42150 | |
| }, | |
| { | |
| "epoch": 6.66, | |
| "learning_rate": 1.0111287758346582e-05, | |
| "loss": 2.3831, | |
| "step": 42200 | |
| }, | |
| { | |
| "epoch": 6.66, | |
| "learning_rate": 1.0087440381558028e-05, | |
| "loss": 2.2717, | |
| "step": 42250 | |
| }, | |
| { | |
| "epoch": 6.67, | |
| "learning_rate": 1.0063593004769475e-05, | |
| "loss": 2.4052, | |
| "step": 42300 | |
| }, | |
| { | |
| "epoch": 6.68, | |
| "learning_rate": 1.0039745627980921e-05, | |
| "loss": 2.4027, | |
| "step": 42350 | |
| }, | |
| { | |
| "epoch": 6.69, | |
| "learning_rate": 1.0015898251192369e-05, | |
| "loss": 2.3474, | |
| "step": 42400 | |
| }, | |
| { | |
| "epoch": 6.7, | |
| "learning_rate": 9.992050874403816e-06, | |
| "loss": 2.319, | |
| "step": 42450 | |
| }, | |
| { | |
| "epoch": 6.7, | |
| "learning_rate": 9.968203497615262e-06, | |
| "loss": 2.3748, | |
| "step": 42500 | |
| }, | |
| { | |
| "epoch": 6.71, | |
| "learning_rate": 9.94435612082671e-06, | |
| "loss": 2.3052, | |
| "step": 42550 | |
| }, | |
| { | |
| "epoch": 6.72, | |
| "learning_rate": 9.920508744038155e-06, | |
| "loss": 2.3753, | |
| "step": 42600 | |
| }, | |
| { | |
| "epoch": 6.73, | |
| "learning_rate": 9.896661367249603e-06, | |
| "loss": 2.2557, | |
| "step": 42650 | |
| }, | |
| { | |
| "epoch": 6.73, | |
| "learning_rate": 9.872813990461048e-06, | |
| "loss": 2.3367, | |
| "step": 42700 | |
| }, | |
| { | |
| "epoch": 6.74, | |
| "learning_rate": 9.848966613672498e-06, | |
| "loss": 2.2875, | |
| "step": 42750 | |
| }, | |
| { | |
| "epoch": 6.75, | |
| "learning_rate": 9.825119236883943e-06, | |
| "loss": 2.2972, | |
| "step": 42800 | |
| }, | |
| { | |
| "epoch": 6.76, | |
| "learning_rate": 9.80127186009539e-06, | |
| "loss": 2.4598, | |
| "step": 42850 | |
| }, | |
| { | |
| "epoch": 6.77, | |
| "learning_rate": 9.777424483306837e-06, | |
| "loss": 2.2605, | |
| "step": 42900 | |
| }, | |
| { | |
| "epoch": 6.77, | |
| "learning_rate": 9.753577106518284e-06, | |
| "loss": 2.3722, | |
| "step": 42950 | |
| }, | |
| { | |
| "epoch": 6.78, | |
| "learning_rate": 9.72972972972973e-06, | |
| "loss": 2.3536, | |
| "step": 43000 | |
| }, | |
| { | |
| "epoch": 6.79, | |
| "learning_rate": 9.705882352941177e-06, | |
| "loss": 2.3247, | |
| "step": 43050 | |
| }, | |
| { | |
| "epoch": 6.8, | |
| "learning_rate": 9.682034976152623e-06, | |
| "loss": 2.3724, | |
| "step": 43100 | |
| }, | |
| { | |
| "epoch": 6.81, | |
| "learning_rate": 9.65818759936407e-06, | |
| "loss": 2.3119, | |
| "step": 43150 | |
| }, | |
| { | |
| "epoch": 6.81, | |
| "learning_rate": 9.634340222575518e-06, | |
| "loss": 2.3488, | |
| "step": 43200 | |
| }, | |
| { | |
| "epoch": 6.82, | |
| "learning_rate": 9.610492845786964e-06, | |
| "loss": 2.4406, | |
| "step": 43250 | |
| }, | |
| { | |
| "epoch": 6.83, | |
| "learning_rate": 9.586645468998411e-06, | |
| "loss": 2.3001, | |
| "step": 43300 | |
| }, | |
| { | |
| "epoch": 6.84, | |
| "learning_rate": 9.562798092209857e-06, | |
| "loss": 2.3593, | |
| "step": 43350 | |
| }, | |
| { | |
| "epoch": 6.85, | |
| "learning_rate": 9.538950715421305e-06, | |
| "loss": 2.3887, | |
| "step": 43400 | |
| }, | |
| { | |
| "epoch": 6.85, | |
| "learning_rate": 9.51510333863275e-06, | |
| "loss": 2.3598, | |
| "step": 43450 | |
| }, | |
| { | |
| "epoch": 6.86, | |
| "learning_rate": 9.491255961844198e-06, | |
| "loss": 2.3472, | |
| "step": 43500 | |
| }, | |
| { | |
| "epoch": 6.87, | |
| "learning_rate": 9.467408585055644e-06, | |
| "loss": 2.3077, | |
| "step": 43550 | |
| }, | |
| { | |
| "epoch": 6.88, | |
| "learning_rate": 9.443561208267091e-06, | |
| "loss": 2.3358, | |
| "step": 43600 | |
| }, | |
| { | |
| "epoch": 6.88, | |
| "learning_rate": 9.419713831478537e-06, | |
| "loss": 2.3872, | |
| "step": 43650 | |
| }, | |
| { | |
| "epoch": 6.89, | |
| "learning_rate": 9.395866454689984e-06, | |
| "loss": 2.2983, | |
| "step": 43700 | |
| }, | |
| { | |
| "epoch": 6.9, | |
| "learning_rate": 9.37201907790143e-06, | |
| "loss": 2.4493, | |
| "step": 43750 | |
| }, | |
| { | |
| "epoch": 6.91, | |
| "learning_rate": 9.348171701112878e-06, | |
| "loss": 2.3481, | |
| "step": 43800 | |
| }, | |
| { | |
| "epoch": 6.92, | |
| "learning_rate": 9.324324324324325e-06, | |
| "loss": 2.3392, | |
| "step": 43850 | |
| }, | |
| { | |
| "epoch": 6.92, | |
| "learning_rate": 9.30047694753577e-06, | |
| "loss": 2.4189, | |
| "step": 43900 | |
| }, | |
| { | |
| "epoch": 6.93, | |
| "learning_rate": 9.276629570747218e-06, | |
| "loss": 2.3721, | |
| "step": 43950 | |
| }, | |
| { | |
| "epoch": 6.94, | |
| "learning_rate": 9.252782193958664e-06, | |
| "loss": 2.2599, | |
| "step": 44000 | |
| }, | |
| { | |
| "epoch": 6.95, | |
| "learning_rate": 9.228934817170112e-06, | |
| "loss": 2.3492, | |
| "step": 44050 | |
| }, | |
| { | |
| "epoch": 6.96, | |
| "learning_rate": 9.205087440381557e-06, | |
| "loss": 2.3938, | |
| "step": 44100 | |
| }, | |
| { | |
| "epoch": 6.96, | |
| "learning_rate": 9.181240063593005e-06, | |
| "loss": 2.4319, | |
| "step": 44150 | |
| }, | |
| { | |
| "epoch": 6.97, | |
| "learning_rate": 9.15739268680445e-06, | |
| "loss": 2.3865, | |
| "step": 44200 | |
| }, | |
| { | |
| "epoch": 6.98, | |
| "learning_rate": 9.133545310015898e-06, | |
| "loss": 2.3202, | |
| "step": 44250 | |
| }, | |
| { | |
| "epoch": 6.99, | |
| "learning_rate": 9.109697933227344e-06, | |
| "loss": 2.3791, | |
| "step": 44300 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 9.085850556438791e-06, | |
| "loss": 2.3508, | |
| "step": 44350 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 9.062003179650239e-06, | |
| "loss": 2.3603, | |
| "step": 44400 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 9.038155802861686e-06, | |
| "loss": 2.3675, | |
| "step": 44450 | |
| }, | |
| { | |
| "epoch": 7.02, | |
| "learning_rate": 9.014308426073132e-06, | |
| "loss": 2.3226, | |
| "step": 44500 | |
| }, | |
| { | |
| "epoch": 7.03, | |
| "learning_rate": 8.99046104928458e-06, | |
| "loss": 2.2309, | |
| "step": 44550 | |
| }, | |
| { | |
| "epoch": 7.03, | |
| "learning_rate": 8.966613672496027e-06, | |
| "loss": 2.3336, | |
| "step": 44600 | |
| }, | |
| { | |
| "epoch": 7.04, | |
| "learning_rate": 8.942766295707473e-06, | |
| "loss": 2.3969, | |
| "step": 44650 | |
| }, | |
| { | |
| "epoch": 7.05, | |
| "learning_rate": 8.91891891891892e-06, | |
| "loss": 2.3952, | |
| "step": 44700 | |
| }, | |
| { | |
| "epoch": 7.06, | |
| "learning_rate": 8.895071542130366e-06, | |
| "loss": 2.2318, | |
| "step": 44750 | |
| }, | |
| { | |
| "epoch": 7.07, | |
| "learning_rate": 8.871224165341814e-06, | |
| "loss": 2.3633, | |
| "step": 44800 | |
| }, | |
| { | |
| "epoch": 7.07, | |
| "learning_rate": 8.84737678855326e-06, | |
| "loss": 2.3753, | |
| "step": 44850 | |
| }, | |
| { | |
| "epoch": 7.08, | |
| "learning_rate": 8.823529411764707e-06, | |
| "loss": 2.3284, | |
| "step": 44900 | |
| }, | |
| { | |
| "epoch": 7.09, | |
| "learning_rate": 8.799682034976153e-06, | |
| "loss": 2.3755, | |
| "step": 44950 | |
| }, | |
| { | |
| "epoch": 7.1, | |
| "learning_rate": 8.7758346581876e-06, | |
| "loss": 2.2437, | |
| "step": 45000 | |
| }, | |
| { | |
| "epoch": 7.11, | |
| "learning_rate": 8.751987281399046e-06, | |
| "loss": 2.3282, | |
| "step": 45050 | |
| }, | |
| { | |
| "epoch": 7.11, | |
| "learning_rate": 8.728139904610493e-06, | |
| "loss": 2.2998, | |
| "step": 45100 | |
| }, | |
| { | |
| "epoch": 7.12, | |
| "learning_rate": 8.704292527821939e-06, | |
| "loss": 2.2286, | |
| "step": 45150 | |
| }, | |
| { | |
| "epoch": 7.13, | |
| "learning_rate": 8.680445151033387e-06, | |
| "loss": 2.3136, | |
| "step": 45200 | |
| }, | |
| { | |
| "epoch": 7.14, | |
| "learning_rate": 8.656597774244834e-06, | |
| "loss": 2.3727, | |
| "step": 45250 | |
| }, | |
| { | |
| "epoch": 7.15, | |
| "learning_rate": 8.63275039745628e-06, | |
| "loss": 2.4005, | |
| "step": 45300 | |
| }, | |
| { | |
| "epoch": 7.15, | |
| "learning_rate": 8.608903020667727e-06, | |
| "loss": 2.2586, | |
| "step": 45350 | |
| }, | |
| { | |
| "epoch": 7.16, | |
| "learning_rate": 8.585055643879173e-06, | |
| "loss": 2.3769, | |
| "step": 45400 | |
| }, | |
| { | |
| "epoch": 7.17, | |
| "learning_rate": 8.56120826709062e-06, | |
| "loss": 2.2427, | |
| "step": 45450 | |
| }, | |
| { | |
| "epoch": 7.18, | |
| "learning_rate": 8.537360890302066e-06, | |
| "loss": 2.4139, | |
| "step": 45500 | |
| }, | |
| { | |
| "epoch": 7.18, | |
| "learning_rate": 8.513513513513514e-06, | |
| "loss": 2.3047, | |
| "step": 45550 | |
| }, | |
| { | |
| "epoch": 7.19, | |
| "learning_rate": 8.48966613672496e-06, | |
| "loss": 2.4186, | |
| "step": 45600 | |
| }, | |
| { | |
| "epoch": 7.2, | |
| "learning_rate": 8.465818759936407e-06, | |
| "loss": 2.3452, | |
| "step": 45650 | |
| }, | |
| { | |
| "epoch": 7.21, | |
| "learning_rate": 8.441971383147853e-06, | |
| "loss": 2.2185, | |
| "step": 45700 | |
| }, | |
| { | |
| "epoch": 7.22, | |
| "learning_rate": 8.4181240063593e-06, | |
| "loss": 2.3871, | |
| "step": 45750 | |
| }, | |
| { | |
| "epoch": 7.22, | |
| "learning_rate": 8.394276629570746e-06, | |
| "loss": 2.2726, | |
| "step": 45800 | |
| }, | |
| { | |
| "epoch": 7.23, | |
| "learning_rate": 8.370429252782194e-06, | |
| "loss": 2.3655, | |
| "step": 45850 | |
| }, | |
| { | |
| "epoch": 7.24, | |
| "learning_rate": 8.34658187599364e-06, | |
| "loss": 2.2998, | |
| "step": 45900 | |
| }, | |
| { | |
| "epoch": 7.25, | |
| "learning_rate": 8.322734499205087e-06, | |
| "loss": 2.3761, | |
| "step": 45950 | |
| }, | |
| { | |
| "epoch": 7.26, | |
| "learning_rate": 8.298887122416534e-06, | |
| "loss": 2.4071, | |
| "step": 46000 | |
| }, | |
| { | |
| "epoch": 7.26, | |
| "learning_rate": 8.275039745627982e-06, | |
| "loss": 2.2446, | |
| "step": 46050 | |
| }, | |
| { | |
| "epoch": 7.27, | |
| "learning_rate": 8.25119236883943e-06, | |
| "loss": 2.4368, | |
| "step": 46100 | |
| }, | |
| { | |
| "epoch": 7.28, | |
| "learning_rate": 8.227344992050875e-06, | |
| "loss": 2.3106, | |
| "step": 46150 | |
| }, | |
| { | |
| "epoch": 7.29, | |
| "learning_rate": 8.203497615262323e-06, | |
| "loss": 2.2454, | |
| "step": 46200 | |
| }, | |
| { | |
| "epoch": 7.29, | |
| "learning_rate": 8.179650238473768e-06, | |
| "loss": 2.3197, | |
| "step": 46250 | |
| }, | |
| { | |
| "epoch": 7.3, | |
| "learning_rate": 8.155802861685216e-06, | |
| "loss": 2.291, | |
| "step": 46300 | |
| }, | |
| { | |
| "epoch": 7.31, | |
| "learning_rate": 8.131955484896662e-06, | |
| "loss": 2.3262, | |
| "step": 46350 | |
| }, | |
| { | |
| "epoch": 7.32, | |
| "learning_rate": 8.108108108108109e-06, | |
| "loss": 2.3377, | |
| "step": 46400 | |
| }, | |
| { | |
| "epoch": 7.33, | |
| "learning_rate": 8.084260731319555e-06, | |
| "loss": 2.3217, | |
| "step": 46450 | |
| }, | |
| { | |
| "epoch": 7.33, | |
| "learning_rate": 8.060413354531002e-06, | |
| "loss": 2.2479, | |
| "step": 46500 | |
| }, | |
| { | |
| "epoch": 7.34, | |
| "learning_rate": 8.036565977742448e-06, | |
| "loss": 2.2902, | |
| "step": 46550 | |
| }, | |
| { | |
| "epoch": 7.35, | |
| "learning_rate": 8.012718600953896e-06, | |
| "loss": 2.2538, | |
| "step": 46600 | |
| }, | |
| { | |
| "epoch": 7.36, | |
| "learning_rate": 7.988871224165343e-06, | |
| "loss": 2.3246, | |
| "step": 46650 | |
| }, | |
| { | |
| "epoch": 7.37, | |
| "learning_rate": 7.965023847376789e-06, | |
| "loss": 2.3713, | |
| "step": 46700 | |
| }, | |
| { | |
| "epoch": 7.37, | |
| "learning_rate": 7.941176470588236e-06, | |
| "loss": 2.3611, | |
| "step": 46750 | |
| }, | |
| { | |
| "epoch": 7.38, | |
| "learning_rate": 7.917329093799682e-06, | |
| "loss": 2.3627, | |
| "step": 46800 | |
| }, | |
| { | |
| "epoch": 7.39, | |
| "learning_rate": 7.89348171701113e-06, | |
| "loss": 2.2824, | |
| "step": 46850 | |
| }, | |
| { | |
| "epoch": 7.4, | |
| "learning_rate": 7.869634340222575e-06, | |
| "loss": 2.3895, | |
| "step": 46900 | |
| }, | |
| { | |
| "epoch": 7.41, | |
| "learning_rate": 7.845786963434023e-06, | |
| "loss": 2.2528, | |
| "step": 46950 | |
| }, | |
| { | |
| "epoch": 7.41, | |
| "learning_rate": 7.821939586645469e-06, | |
| "loss": 2.3639, | |
| "step": 47000 | |
| }, | |
| { | |
| "epoch": 7.42, | |
| "learning_rate": 7.798092209856916e-06, | |
| "loss": 2.3498, | |
| "step": 47050 | |
| }, | |
| { | |
| "epoch": 7.43, | |
| "learning_rate": 7.774244833068362e-06, | |
| "loss": 2.3634, | |
| "step": 47100 | |
| }, | |
| { | |
| "epoch": 7.44, | |
| "learning_rate": 7.75039745627981e-06, | |
| "loss": 2.277, | |
| "step": 47150 | |
| }, | |
| { | |
| "epoch": 7.44, | |
| "learning_rate": 7.726550079491255e-06, | |
| "loss": 2.333, | |
| "step": 47200 | |
| }, | |
| { | |
| "epoch": 7.45, | |
| "learning_rate": 7.702702702702703e-06, | |
| "loss": 2.3236, | |
| "step": 47250 | |
| }, | |
| { | |
| "epoch": 7.46, | |
| "learning_rate": 7.678855325914148e-06, | |
| "loss": 2.3351, | |
| "step": 47300 | |
| }, | |
| { | |
| "epoch": 7.47, | |
| "learning_rate": 7.655007949125596e-06, | |
| "loss": 2.1992, | |
| "step": 47350 | |
| }, | |
| { | |
| "epoch": 7.48, | |
| "learning_rate": 7.631160572337043e-06, | |
| "loss": 2.3276, | |
| "step": 47400 | |
| }, | |
| { | |
| "epoch": 7.48, | |
| "learning_rate": 7.607313195548489e-06, | |
| "loss": 2.3495, | |
| "step": 47450 | |
| }, | |
| { | |
| "epoch": 7.49, | |
| "learning_rate": 7.583465818759936e-06, | |
| "loss": 2.3136, | |
| "step": 47500 | |
| }, | |
| { | |
| "epoch": 7.5, | |
| "learning_rate": 7.559618441971382e-06, | |
| "loss": 2.3916, | |
| "step": 47550 | |
| }, | |
| { | |
| "epoch": 7.51, | |
| "learning_rate": 7.535771065182829e-06, | |
| "loss": 2.3319, | |
| "step": 47600 | |
| }, | |
| { | |
| "epoch": 7.52, | |
| "learning_rate": 7.511923688394276e-06, | |
| "loss": 2.2989, | |
| "step": 47650 | |
| }, | |
| { | |
| "epoch": 7.52, | |
| "learning_rate": 7.488076311605724e-06, | |
| "loss": 2.2922, | |
| "step": 47700 | |
| }, | |
| { | |
| "epoch": 7.53, | |
| "learning_rate": 7.4642289348171706e-06, | |
| "loss": 2.2579, | |
| "step": 47750 | |
| }, | |
| { | |
| "epoch": 7.54, | |
| "learning_rate": 7.440381558028617e-06, | |
| "loss": 2.25, | |
| "step": 47800 | |
| }, | |
| { | |
| "epoch": 7.55, | |
| "learning_rate": 7.416534181240064e-06, | |
| "loss": 2.3738, | |
| "step": 47850 | |
| }, | |
| { | |
| "epoch": 7.56, | |
| "learning_rate": 7.3926868044515104e-06, | |
| "loss": 2.2627, | |
| "step": 47900 | |
| }, | |
| { | |
| "epoch": 7.56, | |
| "learning_rate": 7.368839427662957e-06, | |
| "loss": 2.309, | |
| "step": 47950 | |
| }, | |
| { | |
| "epoch": 7.57, | |
| "learning_rate": 7.344992050874404e-06, | |
| "loss": 2.3631, | |
| "step": 48000 | |
| }, | |
| { | |
| "epoch": 7.58, | |
| "learning_rate": 7.32114467408585e-06, | |
| "loss": 2.2956, | |
| "step": 48050 | |
| }, | |
| { | |
| "epoch": 7.59, | |
| "learning_rate": 7.297297297297298e-06, | |
| "loss": 2.2407, | |
| "step": 48100 | |
| }, | |
| { | |
| "epoch": 7.59, | |
| "learning_rate": 7.2734499205087444e-06, | |
| "loss": 2.3228, | |
| "step": 48150 | |
| }, | |
| { | |
| "epoch": 7.6, | |
| "learning_rate": 7.249602543720191e-06, | |
| "loss": 2.231, | |
| "step": 48200 | |
| }, | |
| { | |
| "epoch": 7.61, | |
| "learning_rate": 7.225755166931638e-06, | |
| "loss": 2.352, | |
| "step": 48250 | |
| }, | |
| { | |
| "epoch": 7.62, | |
| "learning_rate": 7.201907790143084e-06, | |
| "loss": 2.3881, | |
| "step": 48300 | |
| }, | |
| { | |
| "epoch": 7.63, | |
| "learning_rate": 7.178060413354531e-06, | |
| "loss": 2.3211, | |
| "step": 48350 | |
| }, | |
| { | |
| "epoch": 7.63, | |
| "learning_rate": 7.154213036565978e-06, | |
| "loss": 2.3425, | |
| "step": 48400 | |
| }, | |
| { | |
| "epoch": 7.64, | |
| "learning_rate": 7.130365659777425e-06, | |
| "loss": 2.3098, | |
| "step": 48450 | |
| }, | |
| { | |
| "epoch": 7.65, | |
| "learning_rate": 7.106518282988872e-06, | |
| "loss": 2.3842, | |
| "step": 48500 | |
| }, | |
| { | |
| "epoch": 7.66, | |
| "learning_rate": 7.082670906200318e-06, | |
| "loss": 2.3383, | |
| "step": 48550 | |
| }, | |
| { | |
| "epoch": 7.67, | |
| "learning_rate": 7.058823529411765e-06, | |
| "loss": 2.3951, | |
| "step": 48600 | |
| }, | |
| { | |
| "epoch": 7.67, | |
| "learning_rate": 7.0349761526232116e-06, | |
| "loss": 2.3078, | |
| "step": 48650 | |
| }, | |
| { | |
| "epoch": 7.68, | |
| "learning_rate": 7.011128775834658e-06, | |
| "loss": 2.4464, | |
| "step": 48700 | |
| }, | |
| { | |
| "epoch": 7.69, | |
| "learning_rate": 6.987281399046105e-06, | |
| "loss": 2.3558, | |
| "step": 48750 | |
| }, | |
| { | |
| "epoch": 7.7, | |
| "learning_rate": 6.9634340222575514e-06, | |
| "loss": 2.2287, | |
| "step": 48800 | |
| }, | |
| { | |
| "epoch": 7.71, | |
| "learning_rate": 6.939586645468998e-06, | |
| "loss": 2.3043, | |
| "step": 48850 | |
| }, | |
| { | |
| "epoch": 7.71, | |
| "learning_rate": 6.915739268680445e-06, | |
| "loss": 2.3101, | |
| "step": 48900 | |
| }, | |
| { | |
| "epoch": 7.72, | |
| "learning_rate": 6.891891891891892e-06, | |
| "loss": 2.3032, | |
| "step": 48950 | |
| }, | |
| { | |
| "epoch": 7.73, | |
| "learning_rate": 6.868044515103339e-06, | |
| "loss": 2.2677, | |
| "step": 49000 | |
| }, | |
| { | |
| "epoch": 7.74, | |
| "learning_rate": 6.8441971383147854e-06, | |
| "loss": 2.3642, | |
| "step": 49050 | |
| }, | |
| { | |
| "epoch": 7.74, | |
| "learning_rate": 6.820349761526233e-06, | |
| "loss": 2.3649, | |
| "step": 49100 | |
| }, | |
| { | |
| "epoch": 7.75, | |
| "learning_rate": 6.7965023847376795e-06, | |
| "loss": 2.3617, | |
| "step": 49150 | |
| }, | |
| { | |
| "epoch": 7.76, | |
| "learning_rate": 6.772655007949126e-06, | |
| "loss": 2.3569, | |
| "step": 49200 | |
| }, | |
| { | |
| "epoch": 7.77, | |
| "learning_rate": 6.748807631160573e-06, | |
| "loss": 2.297, | |
| "step": 49250 | |
| }, | |
| { | |
| "epoch": 7.78, | |
| "learning_rate": 6.7249602543720194e-06, | |
| "loss": 2.2902, | |
| "step": 49300 | |
| }, | |
| { | |
| "epoch": 7.78, | |
| "learning_rate": 6.701112877583466e-06, | |
| "loss": 2.3593, | |
| "step": 49350 | |
| }, | |
| { | |
| "epoch": 7.79, | |
| "learning_rate": 6.677265500794913e-06, | |
| "loss": 2.2884, | |
| "step": 49400 | |
| }, | |
| { | |
| "epoch": 7.8, | |
| "learning_rate": 6.653418124006359e-06, | |
| "loss": 2.4, | |
| "step": 49450 | |
| }, | |
| { | |
| "epoch": 7.81, | |
| "learning_rate": 6.629570747217806e-06, | |
| "loss": 2.2576, | |
| "step": 49500 | |
| }, | |
| { | |
| "epoch": 7.82, | |
| "learning_rate": 6.6057233704292526e-06, | |
| "loss": 2.2758, | |
| "step": 49550 | |
| }, | |
| { | |
| "epoch": 7.82, | |
| "learning_rate": 6.581875993640699e-06, | |
| "loss": 2.3434, | |
| "step": 49600 | |
| }, | |
| { | |
| "epoch": 7.83, | |
| "learning_rate": 6.558028616852146e-06, | |
| "loss": 2.2259, | |
| "step": 49650 | |
| }, | |
| { | |
| "epoch": 7.84, | |
| "learning_rate": 6.5341812400635925e-06, | |
| "loss": 2.3104, | |
| "step": 49700 | |
| }, | |
| { | |
| "epoch": 7.85, | |
| "learning_rate": 6.51033386327504e-06, | |
| "loss": 2.2976, | |
| "step": 49750 | |
| }, | |
| { | |
| "epoch": 7.85, | |
| "learning_rate": 6.486486486486487e-06, | |
| "loss": 2.34, | |
| "step": 49800 | |
| }, | |
| { | |
| "epoch": 7.86, | |
| "learning_rate": 6.462639109697934e-06, | |
| "loss": 2.3814, | |
| "step": 49850 | |
| }, | |
| { | |
| "epoch": 7.87, | |
| "learning_rate": 6.438791732909381e-06, | |
| "loss": 2.3021, | |
| "step": 49900 | |
| }, | |
| { | |
| "epoch": 7.88, | |
| "learning_rate": 6.414944356120827e-06, | |
| "loss": 2.418, | |
| "step": 49950 | |
| }, | |
| { | |
| "epoch": 7.89, | |
| "learning_rate": 6.391096979332274e-06, | |
| "loss": 2.3411, | |
| "step": 50000 | |
| }, | |
| { | |
| "epoch": 7.89, | |
| "learning_rate": 6.3672496025437205e-06, | |
| "loss": 2.3213, | |
| "step": 50050 | |
| }, | |
| { | |
| "epoch": 7.9, | |
| "learning_rate": 6.343402225755167e-06, | |
| "loss": 2.3401, | |
| "step": 50100 | |
| }, | |
| { | |
| "epoch": 7.91, | |
| "learning_rate": 6.319554848966614e-06, | |
| "loss": 2.2525, | |
| "step": 50150 | |
| }, | |
| { | |
| "epoch": 7.92, | |
| "learning_rate": 6.2957074721780604e-06, | |
| "loss": 2.271, | |
| "step": 50200 | |
| }, | |
| { | |
| "epoch": 7.93, | |
| "learning_rate": 6.271860095389507e-06, | |
| "loss": 2.425, | |
| "step": 50250 | |
| }, | |
| { | |
| "epoch": 7.93, | |
| "learning_rate": 6.248012718600954e-06, | |
| "loss": 2.4018, | |
| "step": 50300 | |
| }, | |
| { | |
| "epoch": 7.94, | |
| "learning_rate": 6.2241653418124e-06, | |
| "loss": 2.3339, | |
| "step": 50350 | |
| }, | |
| { | |
| "epoch": 7.95, | |
| "learning_rate": 6.200317965023847e-06, | |
| "loss": 2.3432, | |
| "step": 50400 | |
| }, | |
| { | |
| "epoch": 7.96, | |
| "learning_rate": 6.176470588235294e-06, | |
| "loss": 2.3247, | |
| "step": 50450 | |
| }, | |
| { | |
| "epoch": 7.97, | |
| "learning_rate": 6.152623211446741e-06, | |
| "loss": 2.3498, | |
| "step": 50500 | |
| }, | |
| { | |
| "epoch": 7.97, | |
| "learning_rate": 6.128775834658188e-06, | |
| "loss": 2.2981, | |
| "step": 50550 | |
| }, | |
| { | |
| "epoch": 7.98, | |
| "learning_rate": 6.104928457869635e-06, | |
| "loss": 2.3491, | |
| "step": 50600 | |
| }, | |
| { | |
| "epoch": 7.99, | |
| "learning_rate": 6.081081081081082e-06, | |
| "loss": 2.3754, | |
| "step": 50650 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 6.057233704292528e-06, | |
| "loss": 2.2337, | |
| "step": 50700 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 6.033386327503975e-06, | |
| "loss": 2.3989, | |
| "step": 50750 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 6.009538950715422e-06, | |
| "loss": 2.2889, | |
| "step": 50800 | |
| }, | |
| { | |
| "epoch": 8.02, | |
| "learning_rate": 5.985691573926868e-06, | |
| "loss": 2.3615, | |
| "step": 50850 | |
| }, | |
| { | |
| "epoch": 8.03, | |
| "learning_rate": 5.961844197138315e-06, | |
| "loss": 2.2809, | |
| "step": 50900 | |
| }, | |
| { | |
| "epoch": 8.04, | |
| "learning_rate": 5.9379968203497616e-06, | |
| "loss": 2.2556, | |
| "step": 50950 | |
| }, | |
| { | |
| "epoch": 8.04, | |
| "learning_rate": 5.914149443561208e-06, | |
| "loss": 2.3127, | |
| "step": 51000 | |
| }, | |
| { | |
| "epoch": 8.05, | |
| "learning_rate": 5.890302066772655e-06, | |
| "loss": 2.323, | |
| "step": 51050 | |
| }, | |
| { | |
| "epoch": 8.06, | |
| "learning_rate": 5.8664546899841014e-06, | |
| "loss": 2.2539, | |
| "step": 51100 | |
| }, | |
| { | |
| "epoch": 8.07, | |
| "learning_rate": 5.842607313195548e-06, | |
| "loss": 2.3405, | |
| "step": 51150 | |
| }, | |
| { | |
| "epoch": 8.08, | |
| "learning_rate": 5.8187599364069955e-06, | |
| "loss": 2.3322, | |
| "step": 51200 | |
| }, | |
| { | |
| "epoch": 8.08, | |
| "learning_rate": 5.794912559618442e-06, | |
| "loss": 2.3013, | |
| "step": 51250 | |
| }, | |
| { | |
| "epoch": 8.09, | |
| "learning_rate": 5.771065182829889e-06, | |
| "loss": 2.2925, | |
| "step": 51300 | |
| }, | |
| { | |
| "epoch": 8.1, | |
| "learning_rate": 5.7472178060413354e-06, | |
| "loss": 2.2064, | |
| "step": 51350 | |
| }, | |
| { | |
| "epoch": 8.11, | |
| "learning_rate": 5.723370429252783e-06, | |
| "loss": 2.3501, | |
| "step": 51400 | |
| }, | |
| { | |
| "epoch": 8.12, | |
| "learning_rate": 5.6995230524642295e-06, | |
| "loss": 2.2522, | |
| "step": 51450 | |
| }, | |
| { | |
| "epoch": 8.12, | |
| "learning_rate": 5.675675675675676e-06, | |
| "loss": 2.4012, | |
| "step": 51500 | |
| }, | |
| { | |
| "epoch": 8.13, | |
| "learning_rate": 5.651828298887123e-06, | |
| "loss": 2.2995, | |
| "step": 51550 | |
| }, | |
| { | |
| "epoch": 8.14, | |
| "learning_rate": 5.627980922098569e-06, | |
| "loss": 2.3444, | |
| "step": 51600 | |
| }, | |
| { | |
| "epoch": 8.15, | |
| "learning_rate": 5.604133545310016e-06, | |
| "loss": 2.3466, | |
| "step": 51650 | |
| }, | |
| { | |
| "epoch": 8.15, | |
| "learning_rate": 5.580286168521463e-06, | |
| "loss": 2.2761, | |
| "step": 51700 | |
| }, | |
| { | |
| "epoch": 8.16, | |
| "learning_rate": 5.556438791732909e-06, | |
| "loss": 2.2975, | |
| "step": 51750 | |
| }, | |
| { | |
| "epoch": 8.17, | |
| "learning_rate": 5.532591414944356e-06, | |
| "loss": 2.2724, | |
| "step": 51800 | |
| }, | |
| { | |
| "epoch": 8.18, | |
| "learning_rate": 5.5087440381558026e-06, | |
| "loss": 2.3443, | |
| "step": 51850 | |
| }, | |
| { | |
| "epoch": 8.19, | |
| "learning_rate": 5.48489666136725e-06, | |
| "loss": 2.2775, | |
| "step": 51900 | |
| }, | |
| { | |
| "epoch": 8.19, | |
| "learning_rate": 5.461049284578697e-06, | |
| "loss": 2.2304, | |
| "step": 51950 | |
| }, | |
| { | |
| "epoch": 8.2, | |
| "learning_rate": 5.437201907790143e-06, | |
| "loss": 2.2836, | |
| "step": 52000 | |
| }, | |
| { | |
| "epoch": 8.21, | |
| "learning_rate": 5.41335453100159e-06, | |
| "loss": 2.319, | |
| "step": 52050 | |
| }, | |
| { | |
| "epoch": 8.22, | |
| "learning_rate": 5.3895071542130366e-06, | |
| "loss": 2.3162, | |
| "step": 52100 | |
| }, | |
| { | |
| "epoch": 8.23, | |
| "learning_rate": 5.365659777424483e-06, | |
| "loss": 2.3399, | |
| "step": 52150 | |
| }, | |
| { | |
| "epoch": 8.23, | |
| "learning_rate": 5.34181240063593e-06, | |
| "loss": 2.2956, | |
| "step": 52200 | |
| }, | |
| { | |
| "epoch": 8.24, | |
| "learning_rate": 5.317965023847377e-06, | |
| "loss": 2.2469, | |
| "step": 52250 | |
| }, | |
| { | |
| "epoch": 8.25, | |
| "learning_rate": 5.294117647058824e-06, | |
| "loss": 2.2996, | |
| "step": 52300 | |
| }, | |
| { | |
| "epoch": 8.26, | |
| "learning_rate": 5.2702702702702705e-06, | |
| "loss": 2.2981, | |
| "step": 52350 | |
| }, | |
| { | |
| "epoch": 8.26, | |
| "learning_rate": 5.246422893481717e-06, | |
| "loss": 2.2523, | |
| "step": 52400 | |
| }, | |
| { | |
| "epoch": 8.27, | |
| "learning_rate": 5.222575516693164e-06, | |
| "loss": 2.3944, | |
| "step": 52450 | |
| }, | |
| { | |
| "epoch": 8.28, | |
| "learning_rate": 5.1987281399046104e-06, | |
| "loss": 2.2972, | |
| "step": 52500 | |
| }, | |
| { | |
| "epoch": 8.29, | |
| "learning_rate": 5.174880763116057e-06, | |
| "loss": 2.2167, | |
| "step": 52550 | |
| }, | |
| { | |
| "epoch": 8.3, | |
| "learning_rate": 5.1510333863275045e-06, | |
| "loss": 2.2483, | |
| "step": 52600 | |
| }, | |
| { | |
| "epoch": 8.3, | |
| "learning_rate": 5.127186009538951e-06, | |
| "loss": 2.2149, | |
| "step": 52650 | |
| }, | |
| { | |
| "epoch": 8.31, | |
| "learning_rate": 5.103338632750398e-06, | |
| "loss": 2.3079, | |
| "step": 52700 | |
| }, | |
| { | |
| "epoch": 8.32, | |
| "learning_rate": 5.079491255961844e-06, | |
| "loss": 2.3098, | |
| "step": 52750 | |
| }, | |
| { | |
| "epoch": 8.33, | |
| "learning_rate": 5.055643879173291e-06, | |
| "loss": 2.3029, | |
| "step": 52800 | |
| }, | |
| { | |
| "epoch": 8.34, | |
| "learning_rate": 5.031796502384738e-06, | |
| "loss": 2.2854, | |
| "step": 52850 | |
| }, | |
| { | |
| "epoch": 8.34, | |
| "learning_rate": 5.007949125596184e-06, | |
| "loss": 2.2957, | |
| "step": 52900 | |
| }, | |
| { | |
| "epoch": 8.35, | |
| "learning_rate": 4.984101748807631e-06, | |
| "loss": 2.2847, | |
| "step": 52950 | |
| }, | |
| { | |
| "epoch": 8.36, | |
| "learning_rate": 4.9602543720190776e-06, | |
| "loss": 2.2588, | |
| "step": 53000 | |
| }, | |
| { | |
| "epoch": 8.37, | |
| "learning_rate": 4.936406995230524e-06, | |
| "loss": 2.4674, | |
| "step": 53050 | |
| }, | |
| { | |
| "epoch": 8.38, | |
| "learning_rate": 4.912559618441972e-06, | |
| "loss": 2.3095, | |
| "step": 53100 | |
| }, | |
| { | |
| "epoch": 8.38, | |
| "learning_rate": 4.888712241653418e-06, | |
| "loss": 2.2204, | |
| "step": 53150 | |
| }, | |
| { | |
| "epoch": 8.39, | |
| "learning_rate": 4.864864864864865e-06, | |
| "loss": 2.3547, | |
| "step": 53200 | |
| }, | |
| { | |
| "epoch": 8.4, | |
| "learning_rate": 4.8410174880763115e-06, | |
| "loss": 2.3369, | |
| "step": 53250 | |
| }, | |
| { | |
| "epoch": 8.41, | |
| "learning_rate": 4.817170111287759e-06, | |
| "loss": 2.1745, | |
| "step": 53300 | |
| }, | |
| { | |
| "epoch": 8.41, | |
| "learning_rate": 4.793322734499206e-06, | |
| "loss": 2.2104, | |
| "step": 53350 | |
| }, | |
| { | |
| "epoch": 8.42, | |
| "learning_rate": 4.769475357710652e-06, | |
| "loss": 2.2942, | |
| "step": 53400 | |
| }, | |
| { | |
| "epoch": 8.43, | |
| "learning_rate": 4.745627980922099e-06, | |
| "loss": 2.3286, | |
| "step": 53450 | |
| }, | |
| { | |
| "epoch": 8.44, | |
| "learning_rate": 4.7217806041335455e-06, | |
| "loss": 2.3778, | |
| "step": 53500 | |
| }, | |
| { | |
| "epoch": 8.45, | |
| "learning_rate": 4.697933227344992e-06, | |
| "loss": 2.4332, | |
| "step": 53550 | |
| }, | |
| { | |
| "epoch": 8.45, | |
| "learning_rate": 4.674085850556439e-06, | |
| "loss": 2.3262, | |
| "step": 53600 | |
| }, | |
| { | |
| "epoch": 8.46, | |
| "learning_rate": 4.650238473767885e-06, | |
| "loss": 2.3026, | |
| "step": 53650 | |
| }, | |
| { | |
| "epoch": 8.47, | |
| "learning_rate": 4.626391096979332e-06, | |
| "loss": 2.3919, | |
| "step": 53700 | |
| }, | |
| { | |
| "epoch": 8.48, | |
| "learning_rate": 4.602543720190779e-06, | |
| "loss": 2.2402, | |
| "step": 53750 | |
| }, | |
| { | |
| "epoch": 8.49, | |
| "learning_rate": 4.578696343402225e-06, | |
| "loss": 2.3166, | |
| "step": 53800 | |
| }, | |
| { | |
| "epoch": 8.49, | |
| "learning_rate": 4.554848966613672e-06, | |
| "loss": 2.3128, | |
| "step": 53850 | |
| }, | |
| { | |
| "epoch": 8.5, | |
| "learning_rate": 4.531001589825119e-06, | |
| "loss": 2.3067, | |
| "step": 53900 | |
| }, | |
| { | |
| "epoch": 8.51, | |
| "learning_rate": 4.507154213036566e-06, | |
| "loss": 2.1824, | |
| "step": 53950 | |
| }, | |
| { | |
| "epoch": 8.52, | |
| "learning_rate": 4.4833068362480135e-06, | |
| "loss": 2.2984, | |
| "step": 54000 | |
| }, | |
| { | |
| "epoch": 8.53, | |
| "learning_rate": 4.45945945945946e-06, | |
| "loss": 2.265, | |
| "step": 54050 | |
| }, | |
| { | |
| "epoch": 8.53, | |
| "learning_rate": 4.435612082670907e-06, | |
| "loss": 2.2982, | |
| "step": 54100 | |
| }, | |
| { | |
| "epoch": 8.54, | |
| "learning_rate": 4.411764705882353e-06, | |
| "loss": 2.2818, | |
| "step": 54150 | |
| }, | |
| { | |
| "epoch": 8.55, | |
| "learning_rate": 4.3879173290938e-06, | |
| "loss": 2.3345, | |
| "step": 54200 | |
| }, | |
| { | |
| "epoch": 8.56, | |
| "learning_rate": 4.364069952305247e-06, | |
| "loss": 2.2896, | |
| "step": 54250 | |
| }, | |
| { | |
| "epoch": 8.56, | |
| "learning_rate": 4.340222575516693e-06, | |
| "loss": 2.2741, | |
| "step": 54300 | |
| }, | |
| { | |
| "epoch": 8.57, | |
| "learning_rate": 4.31637519872814e-06, | |
| "loss": 2.3498, | |
| "step": 54350 | |
| }, | |
| { | |
| "epoch": 8.58, | |
| "learning_rate": 4.2925278219395865e-06, | |
| "loss": 2.2106, | |
| "step": 54400 | |
| }, | |
| { | |
| "epoch": 8.59, | |
| "learning_rate": 4.268680445151033e-06, | |
| "loss": 2.2493, | |
| "step": 54450 | |
| }, | |
| { | |
| "epoch": 8.6, | |
| "learning_rate": 4.24483306836248e-06, | |
| "loss": 2.2853, | |
| "step": 54500 | |
| }, | |
| { | |
| "epoch": 8.6, | |
| "learning_rate": 4.2209856915739264e-06, | |
| "loss": 2.3935, | |
| "step": 54550 | |
| }, | |
| { | |
| "epoch": 8.61, | |
| "learning_rate": 4.197138314785373e-06, | |
| "loss": 2.3058, | |
| "step": 54600 | |
| }, | |
| { | |
| "epoch": 8.62, | |
| "learning_rate": 4.17329093799682e-06, | |
| "loss": 2.2131, | |
| "step": 54650 | |
| }, | |
| { | |
| "epoch": 8.63, | |
| "learning_rate": 4.149443561208267e-06, | |
| "loss": 2.2303, | |
| "step": 54700 | |
| }, | |
| { | |
| "epoch": 8.64, | |
| "learning_rate": 4.125596184419715e-06, | |
| "loss": 2.3927, | |
| "step": 54750 | |
| }, | |
| { | |
| "epoch": 8.64, | |
| "learning_rate": 4.101748807631161e-06, | |
| "loss": 2.2371, | |
| "step": 54800 | |
| }, | |
| { | |
| "epoch": 8.65, | |
| "learning_rate": 4.077901430842608e-06, | |
| "loss": 2.2474, | |
| "step": 54850 | |
| }, | |
| { | |
| "epoch": 8.66, | |
| "learning_rate": 4.0540540540540545e-06, | |
| "loss": 2.3602, | |
| "step": 54900 | |
| }, | |
| { | |
| "epoch": 8.67, | |
| "learning_rate": 4.030206677265501e-06, | |
| "loss": 2.2762, | |
| "step": 54950 | |
| }, | |
| { | |
| "epoch": 8.68, | |
| "learning_rate": 4.006359300476948e-06, | |
| "loss": 2.368, | |
| "step": 55000 | |
| }, | |
| { | |
| "epoch": 8.68, | |
| "learning_rate": 3.982511923688394e-06, | |
| "loss": 2.2163, | |
| "step": 55050 | |
| }, | |
| { | |
| "epoch": 8.69, | |
| "learning_rate": 3.958664546899841e-06, | |
| "loss": 2.3402, | |
| "step": 55100 | |
| }, | |
| { | |
| "epoch": 8.7, | |
| "learning_rate": 3.934817170111288e-06, | |
| "loss": 2.3111, | |
| "step": 55150 | |
| }, | |
| { | |
| "epoch": 8.71, | |
| "learning_rate": 3.910969793322734e-06, | |
| "loss": 2.3387, | |
| "step": 55200 | |
| }, | |
| { | |
| "epoch": 8.71, | |
| "learning_rate": 3.887122416534181e-06, | |
| "loss": 2.2679, | |
| "step": 55250 | |
| }, | |
| { | |
| "epoch": 8.72, | |
| "learning_rate": 3.8632750397456276e-06, | |
| "loss": 2.2446, | |
| "step": 55300 | |
| }, | |
| { | |
| "epoch": 8.73, | |
| "learning_rate": 3.839427662957074e-06, | |
| "loss": 2.3467, | |
| "step": 55350 | |
| }, | |
| { | |
| "epoch": 8.74, | |
| "learning_rate": 3.815580286168522e-06, | |
| "loss": 2.2937, | |
| "step": 55400 | |
| }, | |
| { | |
| "epoch": 8.75, | |
| "learning_rate": 3.791732909379968e-06, | |
| "loss": 2.2471, | |
| "step": 55450 | |
| }, | |
| { | |
| "epoch": 8.75, | |
| "learning_rate": 3.7678855325914145e-06, | |
| "loss": 2.3274, | |
| "step": 55500 | |
| }, | |
| { | |
| "epoch": 8.76, | |
| "learning_rate": 3.744038155802862e-06, | |
| "loss": 2.3025, | |
| "step": 55550 | |
| }, | |
| { | |
| "epoch": 8.77, | |
| "learning_rate": 3.7201907790143086e-06, | |
| "loss": 2.25, | |
| "step": 55600 | |
| }, | |
| { | |
| "epoch": 8.78, | |
| "learning_rate": 3.6963434022257552e-06, | |
| "loss": 2.2376, | |
| "step": 55650 | |
| }, | |
| { | |
| "epoch": 8.79, | |
| "learning_rate": 3.672496025437202e-06, | |
| "loss": 2.3342, | |
| "step": 55700 | |
| }, | |
| { | |
| "epoch": 8.79, | |
| "learning_rate": 3.648648648648649e-06, | |
| "loss": 2.3249, | |
| "step": 55750 | |
| }, | |
| { | |
| "epoch": 8.8, | |
| "learning_rate": 3.6248012718600955e-06, | |
| "loss": 2.2627, | |
| "step": 55800 | |
| }, | |
| { | |
| "epoch": 8.81, | |
| "learning_rate": 3.600953895071542e-06, | |
| "loss": 2.2669, | |
| "step": 55850 | |
| }, | |
| { | |
| "epoch": 8.82, | |
| "learning_rate": 3.577106518282989e-06, | |
| "loss": 2.2734, | |
| "step": 55900 | |
| }, | |
| { | |
| "epoch": 8.82, | |
| "learning_rate": 3.553259141494436e-06, | |
| "loss": 2.4094, | |
| "step": 55950 | |
| }, | |
| { | |
| "epoch": 8.83, | |
| "learning_rate": 3.5294117647058825e-06, | |
| "loss": 2.3159, | |
| "step": 56000 | |
| }, | |
| { | |
| "epoch": 8.84, | |
| "learning_rate": 3.505564387917329e-06, | |
| "loss": 2.2596, | |
| "step": 56050 | |
| }, | |
| { | |
| "epoch": 8.85, | |
| "learning_rate": 3.4817170111287757e-06, | |
| "loss": 2.2068, | |
| "step": 56100 | |
| }, | |
| { | |
| "epoch": 8.86, | |
| "learning_rate": 3.4578696343402224e-06, | |
| "loss": 2.3419, | |
| "step": 56150 | |
| }, | |
| { | |
| "epoch": 8.86, | |
| "learning_rate": 3.4340222575516694e-06, | |
| "loss": 2.2312, | |
| "step": 56200 | |
| }, | |
| { | |
| "epoch": 8.87, | |
| "learning_rate": 3.4101748807631165e-06, | |
| "loss": 2.2964, | |
| "step": 56250 | |
| }, | |
| { | |
| "epoch": 8.88, | |
| "learning_rate": 3.386327503974563e-06, | |
| "loss": 2.3362, | |
| "step": 56300 | |
| }, | |
| { | |
| "epoch": 8.89, | |
| "learning_rate": 3.3624801271860097e-06, | |
| "loss": 2.1478, | |
| "step": 56350 | |
| }, | |
| { | |
| "epoch": 8.9, | |
| "learning_rate": 3.3386327503974563e-06, | |
| "loss": 2.2711, | |
| "step": 56400 | |
| }, | |
| { | |
| "epoch": 8.9, | |
| "learning_rate": 3.314785373608903e-06, | |
| "loss": 2.216, | |
| "step": 56450 | |
| }, | |
| { | |
| "epoch": 8.91, | |
| "learning_rate": 3.2909379968203496e-06, | |
| "loss": 2.2763, | |
| "step": 56500 | |
| }, | |
| { | |
| "epoch": 8.92, | |
| "learning_rate": 3.2670906200317962e-06, | |
| "loss": 2.2942, | |
| "step": 56550 | |
| }, | |
| { | |
| "epoch": 8.93, | |
| "learning_rate": 3.2432432432432437e-06, | |
| "loss": 2.3179, | |
| "step": 56600 | |
| }, | |
| { | |
| "epoch": 8.94, | |
| "learning_rate": 3.2193958664546903e-06, | |
| "loss": 2.3063, | |
| "step": 56650 | |
| }, | |
| { | |
| "epoch": 8.94, | |
| "learning_rate": 3.195548489666137e-06, | |
| "loss": 2.3186, | |
| "step": 56700 | |
| }, | |
| { | |
| "epoch": 8.95, | |
| "learning_rate": 3.1717011128775836e-06, | |
| "loss": 2.2831, | |
| "step": 56750 | |
| }, | |
| { | |
| "epoch": 8.96, | |
| "learning_rate": 3.1478537360890302e-06, | |
| "loss": 2.3537, | |
| "step": 56800 | |
| }, | |
| { | |
| "epoch": 8.97, | |
| "learning_rate": 3.124006359300477e-06, | |
| "loss": 2.2919, | |
| "step": 56850 | |
| }, | |
| { | |
| "epoch": 8.97, | |
| "learning_rate": 3.1001589825119235e-06, | |
| "loss": 2.3836, | |
| "step": 56900 | |
| }, | |
| { | |
| "epoch": 8.98, | |
| "learning_rate": 3.0763116057233705e-06, | |
| "loss": 2.4449, | |
| "step": 56950 | |
| }, | |
| { | |
| "epoch": 8.99, | |
| "learning_rate": 3.0524642289348176e-06, | |
| "loss": 2.2795, | |
| "step": 57000 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "learning_rate": 3.028616852146264e-06, | |
| "loss": 2.2668, | |
| "step": 57050 | |
| }, | |
| { | |
| "epoch": 9.01, | |
| "learning_rate": 3.004769475357711e-06, | |
| "loss": 2.2844, | |
| "step": 57100 | |
| }, | |
| { | |
| "epoch": 9.01, | |
| "learning_rate": 2.9809220985691575e-06, | |
| "loss": 2.2807, | |
| "step": 57150 | |
| }, | |
| { | |
| "epoch": 9.02, | |
| "learning_rate": 2.957074721780604e-06, | |
| "loss": 2.3082, | |
| "step": 57200 | |
| }, | |
| { | |
| "epoch": 9.03, | |
| "learning_rate": 2.9332273449920507e-06, | |
| "loss": 2.2464, | |
| "step": 57250 | |
| }, | |
| { | |
| "epoch": 9.04, | |
| "learning_rate": 2.9093799682034978e-06, | |
| "loss": 2.3599, | |
| "step": 57300 | |
| }, | |
| { | |
| "epoch": 9.05, | |
| "learning_rate": 2.8855325914149444e-06, | |
| "loss": 2.2516, | |
| "step": 57350 | |
| }, | |
| { | |
| "epoch": 9.05, | |
| "learning_rate": 2.8616852146263915e-06, | |
| "loss": 2.2633, | |
| "step": 57400 | |
| }, | |
| { | |
| "epoch": 9.06, | |
| "learning_rate": 2.837837837837838e-06, | |
| "loss": 2.364, | |
| "step": 57450 | |
| }, | |
| { | |
| "epoch": 9.07, | |
| "learning_rate": 2.8139904610492847e-06, | |
| "loss": 2.2732, | |
| "step": 57500 | |
| }, | |
| { | |
| "epoch": 9.08, | |
| "learning_rate": 2.7901430842607313e-06, | |
| "loss": 2.2697, | |
| "step": 57550 | |
| }, | |
| { | |
| "epoch": 9.09, | |
| "learning_rate": 2.766295707472178e-06, | |
| "loss": 2.3099, | |
| "step": 57600 | |
| }, | |
| { | |
| "epoch": 9.09, | |
| "learning_rate": 2.742448330683625e-06, | |
| "loss": 2.2531, | |
| "step": 57650 | |
| }, | |
| { | |
| "epoch": 9.1, | |
| "learning_rate": 2.7186009538950716e-06, | |
| "loss": 2.3208, | |
| "step": 57700 | |
| }, | |
| { | |
| "epoch": 9.11, | |
| "learning_rate": 2.6947535771065183e-06, | |
| "loss": 2.2014, | |
| "step": 57750 | |
| }, | |
| { | |
| "epoch": 9.12, | |
| "learning_rate": 2.670906200317965e-06, | |
| "loss": 2.3493, | |
| "step": 57800 | |
| }, | |
| { | |
| "epoch": 9.12, | |
| "learning_rate": 2.647058823529412e-06, | |
| "loss": 2.2841, | |
| "step": 57850 | |
| }, | |
| { | |
| "epoch": 9.13, | |
| "learning_rate": 2.6232114467408586e-06, | |
| "loss": 2.2263, | |
| "step": 57900 | |
| }, | |
| { | |
| "epoch": 9.14, | |
| "learning_rate": 2.5993640699523052e-06, | |
| "loss": 2.3054, | |
| "step": 57950 | |
| }, | |
| { | |
| "epoch": 9.15, | |
| "learning_rate": 2.5755166931637523e-06, | |
| "loss": 2.2863, | |
| "step": 58000 | |
| }, | |
| { | |
| "epoch": 9.16, | |
| "learning_rate": 2.551669316375199e-06, | |
| "loss": 2.2468, | |
| "step": 58050 | |
| }, | |
| { | |
| "epoch": 9.16, | |
| "learning_rate": 2.5278219395866455e-06, | |
| "loss": 2.3116, | |
| "step": 58100 | |
| }, | |
| { | |
| "epoch": 9.17, | |
| "learning_rate": 2.503974562798092e-06, | |
| "loss": 2.3623, | |
| "step": 58150 | |
| }, | |
| { | |
| "epoch": 9.18, | |
| "learning_rate": 2.4801271860095388e-06, | |
| "loss": 2.2285, | |
| "step": 58200 | |
| }, | |
| { | |
| "epoch": 9.19, | |
| "learning_rate": 2.456279809220986e-06, | |
| "loss": 2.3856, | |
| "step": 58250 | |
| }, | |
| { | |
| "epoch": 9.2, | |
| "learning_rate": 2.4324324324324325e-06, | |
| "loss": 2.2852, | |
| "step": 58300 | |
| }, | |
| { | |
| "epoch": 9.2, | |
| "learning_rate": 2.4085850556438795e-06, | |
| "loss": 2.315, | |
| "step": 58350 | |
| }, | |
| { | |
| "epoch": 9.21, | |
| "learning_rate": 2.384737678855326e-06, | |
| "loss": 2.3652, | |
| "step": 58400 | |
| }, | |
| { | |
| "epoch": 9.22, | |
| "learning_rate": 2.3608903020667728e-06, | |
| "loss": 2.2811, | |
| "step": 58450 | |
| }, | |
| { | |
| "epoch": 9.23, | |
| "learning_rate": 2.3370429252782194e-06, | |
| "loss": 2.2823, | |
| "step": 58500 | |
| }, | |
| { | |
| "epoch": 9.24, | |
| "learning_rate": 2.313195548489666e-06, | |
| "loss": 2.2714, | |
| "step": 58550 | |
| }, | |
| { | |
| "epoch": 9.24, | |
| "learning_rate": 2.2893481717011127e-06, | |
| "loss": 2.2866, | |
| "step": 58600 | |
| }, | |
| { | |
| "epoch": 9.25, | |
| "learning_rate": 2.2655007949125597e-06, | |
| "loss": 2.2227, | |
| "step": 58650 | |
| }, | |
| { | |
| "epoch": 9.26, | |
| "learning_rate": 2.2416534181240068e-06, | |
| "loss": 2.3044, | |
| "step": 58700 | |
| }, | |
| { | |
| "epoch": 9.27, | |
| "learning_rate": 2.2178060413354534e-06, | |
| "loss": 2.3071, | |
| "step": 58750 | |
| }, | |
| { | |
| "epoch": 9.27, | |
| "learning_rate": 2.1939586645469e-06, | |
| "loss": 2.1963, | |
| "step": 58800 | |
| }, | |
| { | |
| "epoch": 9.28, | |
| "learning_rate": 2.1701112877583466e-06, | |
| "loss": 2.281, | |
| "step": 58850 | |
| }, | |
| { | |
| "epoch": 9.29, | |
| "learning_rate": 2.1462639109697933e-06, | |
| "loss": 2.2891, | |
| "step": 58900 | |
| }, | |
| { | |
| "epoch": 9.3, | |
| "learning_rate": 2.12241653418124e-06, | |
| "loss": 2.3513, | |
| "step": 58950 | |
| }, | |
| { | |
| "epoch": 9.31, | |
| "learning_rate": 2.0985691573926865e-06, | |
| "loss": 2.2369, | |
| "step": 59000 | |
| }, | |
| { | |
| "epoch": 9.31, | |
| "learning_rate": 2.0747217806041336e-06, | |
| "loss": 2.2101, | |
| "step": 59050 | |
| }, | |
| { | |
| "epoch": 9.32, | |
| "learning_rate": 2.0508744038155806e-06, | |
| "loss": 2.2356, | |
| "step": 59100 | |
| }, | |
| { | |
| "epoch": 9.33, | |
| "learning_rate": 2.0270270270270273e-06, | |
| "loss": 2.2986, | |
| "step": 59150 | |
| }, | |
| { | |
| "epoch": 9.34, | |
| "learning_rate": 2.003179650238474e-06, | |
| "loss": 2.3052, | |
| "step": 59200 | |
| }, | |
| { | |
| "epoch": 9.35, | |
| "learning_rate": 1.9793322734499205e-06, | |
| "loss": 2.2017, | |
| "step": 59250 | |
| }, | |
| { | |
| "epoch": 9.35, | |
| "learning_rate": 1.955484896661367e-06, | |
| "loss": 2.3372, | |
| "step": 59300 | |
| }, | |
| { | |
| "epoch": 9.36, | |
| "learning_rate": 1.9316375198728138e-06, | |
| "loss": 2.2929, | |
| "step": 59350 | |
| }, | |
| { | |
| "epoch": 9.37, | |
| "learning_rate": 1.907790143084261e-06, | |
| "loss": 2.2468, | |
| "step": 59400 | |
| }, | |
| { | |
| "epoch": 9.38, | |
| "learning_rate": 1.8839427662957072e-06, | |
| "loss": 2.316, | |
| "step": 59450 | |
| }, | |
| { | |
| "epoch": 9.38, | |
| "learning_rate": 1.8600953895071543e-06, | |
| "loss": 2.3643, | |
| "step": 59500 | |
| }, | |
| { | |
| "epoch": 9.39, | |
| "learning_rate": 1.836248012718601e-06, | |
| "loss": 2.2971, | |
| "step": 59550 | |
| }, | |
| { | |
| "epoch": 9.4, | |
| "learning_rate": 1.8124006359300478e-06, | |
| "loss": 2.3568, | |
| "step": 59600 | |
| }, | |
| { | |
| "epoch": 9.41, | |
| "learning_rate": 1.7885532591414946e-06, | |
| "loss": 2.1878, | |
| "step": 59650 | |
| }, | |
| { | |
| "epoch": 9.42, | |
| "learning_rate": 1.7647058823529412e-06, | |
| "loss": 2.3828, | |
| "step": 59700 | |
| }, | |
| { | |
| "epoch": 9.42, | |
| "learning_rate": 1.7408585055643879e-06, | |
| "loss": 2.2528, | |
| "step": 59750 | |
| }, | |
| { | |
| "epoch": 9.43, | |
| "learning_rate": 1.7170111287758347e-06, | |
| "loss": 2.329, | |
| "step": 59800 | |
| }, | |
| { | |
| "epoch": 9.44, | |
| "learning_rate": 1.6931637519872815e-06, | |
| "loss": 2.3648, | |
| "step": 59850 | |
| }, | |
| { | |
| "epoch": 9.45, | |
| "learning_rate": 1.6693163751987282e-06, | |
| "loss": 2.3221, | |
| "step": 59900 | |
| }, | |
| { | |
| "epoch": 9.46, | |
| "learning_rate": 1.6454689984101748e-06, | |
| "loss": 2.1805, | |
| "step": 59950 | |
| }, | |
| { | |
| "epoch": 9.46, | |
| "learning_rate": 1.6216216216216219e-06, | |
| "loss": 2.3446, | |
| "step": 60000 | |
| }, | |
| { | |
| "epoch": 9.47, | |
| "learning_rate": 1.5977742448330685e-06, | |
| "loss": 2.42, | |
| "step": 60050 | |
| }, | |
| { | |
| "epoch": 9.48, | |
| "learning_rate": 1.5739268680445151e-06, | |
| "loss": 2.1562, | |
| "step": 60100 | |
| }, | |
| { | |
| "epoch": 9.49, | |
| "learning_rate": 1.5500794912559617e-06, | |
| "loss": 2.1505, | |
| "step": 60150 | |
| }, | |
| { | |
| "epoch": 9.5, | |
| "learning_rate": 1.5262321144674088e-06, | |
| "loss": 2.2464, | |
| "step": 60200 | |
| }, | |
| { | |
| "epoch": 9.5, | |
| "learning_rate": 1.5023847376788554e-06, | |
| "loss": 2.3115, | |
| "step": 60250 | |
| }, | |
| { | |
| "epoch": 9.51, | |
| "learning_rate": 1.478537360890302e-06, | |
| "loss": 2.0721, | |
| "step": 60300 | |
| }, | |
| { | |
| "epoch": 9.52, | |
| "learning_rate": 1.4546899841017489e-06, | |
| "loss": 2.27, | |
| "step": 60350 | |
| }, | |
| { | |
| "epoch": 9.53, | |
| "learning_rate": 1.4308426073131957e-06, | |
| "loss": 2.3262, | |
| "step": 60400 | |
| }, | |
| { | |
| "epoch": 9.53, | |
| "learning_rate": 1.4069952305246424e-06, | |
| "loss": 2.3878, | |
| "step": 60450 | |
| }, | |
| { | |
| "epoch": 9.54, | |
| "learning_rate": 1.383147853736089e-06, | |
| "loss": 2.2035, | |
| "step": 60500 | |
| }, | |
| { | |
| "epoch": 9.55, | |
| "learning_rate": 1.3593004769475358e-06, | |
| "loss": 2.2805, | |
| "step": 60550 | |
| }, | |
| { | |
| "epoch": 9.56, | |
| "learning_rate": 1.3354531001589825e-06, | |
| "loss": 2.2035, | |
| "step": 60600 | |
| }, | |
| { | |
| "epoch": 9.57, | |
| "learning_rate": 1.3116057233704293e-06, | |
| "loss": 2.225, | |
| "step": 60650 | |
| }, | |
| { | |
| "epoch": 9.57, | |
| "learning_rate": 1.2877583465818761e-06, | |
| "loss": 2.2499, | |
| "step": 60700 | |
| }, | |
| { | |
| "epoch": 9.58, | |
| "learning_rate": 1.2639109697933228e-06, | |
| "loss": 2.2492, | |
| "step": 60750 | |
| }, | |
| { | |
| "epoch": 9.59, | |
| "learning_rate": 1.2400635930047694e-06, | |
| "loss": 2.2315, | |
| "step": 60800 | |
| }, | |
| { | |
| "epoch": 9.6, | |
| "learning_rate": 1.2162162162162162e-06, | |
| "loss": 2.2754, | |
| "step": 60850 | |
| }, | |
| { | |
| "epoch": 9.61, | |
| "learning_rate": 1.192368839427663e-06, | |
| "loss": 2.2571, | |
| "step": 60900 | |
| }, | |
| { | |
| "epoch": 9.61, | |
| "learning_rate": 1.1685214626391097e-06, | |
| "loss": 2.3155, | |
| "step": 60950 | |
| }, | |
| { | |
| "epoch": 9.62, | |
| "learning_rate": 1.1446740858505563e-06, | |
| "loss": 2.2825, | |
| "step": 61000 | |
| }, | |
| { | |
| "epoch": 9.63, | |
| "learning_rate": 1.1208267090620034e-06, | |
| "loss": 2.2699, | |
| "step": 61050 | |
| }, | |
| { | |
| "epoch": 9.64, | |
| "learning_rate": 1.09697933227345e-06, | |
| "loss": 2.2182, | |
| "step": 61100 | |
| }, | |
| { | |
| "epoch": 9.65, | |
| "learning_rate": 1.0731319554848966e-06, | |
| "loss": 2.2889, | |
| "step": 61150 | |
| }, | |
| { | |
| "epoch": 9.65, | |
| "learning_rate": 1.0492845786963433e-06, | |
| "loss": 2.2569, | |
| "step": 61200 | |
| }, | |
| { | |
| "epoch": 9.66, | |
| "learning_rate": 1.0254372019077903e-06, | |
| "loss": 2.2386, | |
| "step": 61250 | |
| }, | |
| { | |
| "epoch": 9.67, | |
| "learning_rate": 1.001589825119237e-06, | |
| "loss": 2.3111, | |
| "step": 61300 | |
| }, | |
| { | |
| "epoch": 9.68, | |
| "learning_rate": 9.777424483306836e-07, | |
| "loss": 2.2391, | |
| "step": 61350 | |
| }, | |
| { | |
| "epoch": 9.68, | |
| "learning_rate": 9.538950715421304e-07, | |
| "loss": 2.2852, | |
| "step": 61400 | |
| }, | |
| { | |
| "epoch": 9.69, | |
| "learning_rate": 9.300476947535771e-07, | |
| "loss": 2.2392, | |
| "step": 61450 | |
| }, | |
| { | |
| "epoch": 9.7, | |
| "learning_rate": 9.062003179650239e-07, | |
| "loss": 2.3655, | |
| "step": 61500 | |
| }, | |
| { | |
| "epoch": 9.71, | |
| "learning_rate": 8.823529411764706e-07, | |
| "loss": 2.2298, | |
| "step": 61550 | |
| }, | |
| { | |
| "epoch": 9.72, | |
| "learning_rate": 8.585055643879174e-07, | |
| "loss": 2.2856, | |
| "step": 61600 | |
| }, | |
| { | |
| "epoch": 9.72, | |
| "learning_rate": 8.346581875993641e-07, | |
| "loss": 2.2957, | |
| "step": 61650 | |
| }, | |
| { | |
| "epoch": 9.73, | |
| "learning_rate": 8.108108108108109e-07, | |
| "loss": 2.3311, | |
| "step": 61700 | |
| }, | |
| { | |
| "epoch": 9.74, | |
| "learning_rate": 7.869634340222576e-07, | |
| "loss": 2.2878, | |
| "step": 61750 | |
| }, | |
| { | |
| "epoch": 9.75, | |
| "learning_rate": 7.631160572337044e-07, | |
| "loss": 2.3333, | |
| "step": 61800 | |
| }, | |
| { | |
| "epoch": 9.76, | |
| "learning_rate": 7.39268680445151e-07, | |
| "loss": 2.2988, | |
| "step": 61850 | |
| }, | |
| { | |
| "epoch": 9.76, | |
| "learning_rate": 7.154213036565979e-07, | |
| "loss": 2.3403, | |
| "step": 61900 | |
| }, | |
| { | |
| "epoch": 9.77, | |
| "learning_rate": 6.915739268680445e-07, | |
| "loss": 2.3127, | |
| "step": 61950 | |
| }, | |
| { | |
| "epoch": 9.78, | |
| "learning_rate": 6.677265500794912e-07, | |
| "loss": 2.2709, | |
| "step": 62000 | |
| }, | |
| { | |
| "epoch": 9.79, | |
| "learning_rate": 6.438791732909381e-07, | |
| "loss": 2.2767, | |
| "step": 62050 | |
| }, | |
| { | |
| "epoch": 9.79, | |
| "learning_rate": 6.200317965023847e-07, | |
| "loss": 2.2643, | |
| "step": 62100 | |
| }, | |
| { | |
| "epoch": 9.8, | |
| "learning_rate": 5.961844197138315e-07, | |
| "loss": 2.1819, | |
| "step": 62150 | |
| }, | |
| { | |
| "epoch": 9.81, | |
| "learning_rate": 5.723370429252782e-07, | |
| "loss": 2.3761, | |
| "step": 62200 | |
| }, | |
| { | |
| "epoch": 9.82, | |
| "learning_rate": 5.48489666136725e-07, | |
| "loss": 2.1973, | |
| "step": 62250 | |
| }, | |
| { | |
| "epoch": 9.83, | |
| "learning_rate": 5.246422893481716e-07, | |
| "loss": 2.3058, | |
| "step": 62300 | |
| }, | |
| { | |
| "epoch": 9.83, | |
| "learning_rate": 5.007949125596185e-07, | |
| "loss": 2.3225, | |
| "step": 62350 | |
| }, | |
| { | |
| "epoch": 9.84, | |
| "learning_rate": 4.769475357710652e-07, | |
| "loss": 2.2697, | |
| "step": 62400 | |
| }, | |
| { | |
| "epoch": 9.85, | |
| "learning_rate": 4.5310015898251194e-07, | |
| "loss": 2.2113, | |
| "step": 62450 | |
| }, | |
| { | |
| "epoch": 9.86, | |
| "learning_rate": 4.292527821939587e-07, | |
| "loss": 2.3822, | |
| "step": 62500 | |
| }, | |
| { | |
| "epoch": 9.87, | |
| "learning_rate": 4.0540540540540546e-07, | |
| "loss": 2.3111, | |
| "step": 62550 | |
| }, | |
| { | |
| "epoch": 9.87, | |
| "learning_rate": 3.815580286168522e-07, | |
| "loss": 2.3057, | |
| "step": 62600 | |
| }, | |
| { | |
| "epoch": 9.88, | |
| "learning_rate": 3.5771065182829893e-07, | |
| "loss": 2.2624, | |
| "step": 62650 | |
| }, | |
| { | |
| "epoch": 9.89, | |
| "learning_rate": 3.338632750397456e-07, | |
| "loss": 2.2167, | |
| "step": 62700 | |
| }, | |
| { | |
| "epoch": 9.9, | |
| "learning_rate": 3.1001589825119235e-07, | |
| "loss": 2.3353, | |
| "step": 62750 | |
| }, | |
| { | |
| "epoch": 9.91, | |
| "learning_rate": 2.861685214626391e-07, | |
| "loss": 2.3511, | |
| "step": 62800 | |
| }, | |
| { | |
| "epoch": 9.91, | |
| "learning_rate": 2.623211446740858e-07, | |
| "loss": 2.2372, | |
| "step": 62850 | |
| }, | |
| { | |
| "epoch": 9.92, | |
| "learning_rate": 2.384737678855326e-07, | |
| "loss": 2.2875, | |
| "step": 62900 | |
| }, | |
| { | |
| "epoch": 9.93, | |
| "learning_rate": 2.1462639109697934e-07, | |
| "loss": 2.246, | |
| "step": 62950 | |
| }, | |
| { | |
| "epoch": 9.94, | |
| "learning_rate": 1.907790143084261e-07, | |
| "loss": 2.2529, | |
| "step": 63000 | |
| }, | |
| { | |
| "epoch": 9.94, | |
| "learning_rate": 1.669316375198728e-07, | |
| "loss": 2.2318, | |
| "step": 63050 | |
| }, | |
| { | |
| "epoch": 9.95, | |
| "learning_rate": 1.4308426073131954e-07, | |
| "loss": 2.3183, | |
| "step": 63100 | |
| }, | |
| { | |
| "epoch": 9.96, | |
| "learning_rate": 1.192368839427663e-07, | |
| "loss": 2.3006, | |
| "step": 63150 | |
| }, | |
| { | |
| "epoch": 9.97, | |
| "learning_rate": 9.538950715421305e-08, | |
| "loss": 2.2809, | |
| "step": 63200 | |
| }, | |
| { | |
| "epoch": 9.98, | |
| "learning_rate": 7.154213036565977e-08, | |
| "loss": 2.2341, | |
| "step": 63250 | |
| }, | |
| { | |
| "epoch": 9.98, | |
| "learning_rate": 4.7694753577106525e-08, | |
| "loss": 2.3625, | |
| "step": 63300 | |
| }, | |
| { | |
| "epoch": 9.99, | |
| "learning_rate": 2.3847376788553262e-08, | |
| "loss": 2.2389, | |
| "step": 63350 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "learning_rate": 0.0, | |
| "loss": 2.307, | |
| "step": 63400 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "step": 63400, | |
| "total_flos": 2.706173459807863e+18, | |
| "train_loss": 2.6413619818221132, | |
| "train_runtime": 44032.3597, | |
| "train_samples_per_second": 46.076, | |
| "train_steps_per_second": 1.44 | |
| } | |
| ], | |
| "max_steps": 63400, | |
| "num_train_epochs": 10, | |
| "total_flos": 2.706173459807863e+18, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |