{ "best_metric": null, "best_model_checkpoint": null, "epoch": 33.373263546863605, "global_step": 6441500, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 4.99875e-05, "loss": 7.6867, "step": 500 }, { "epoch": 0.0, "learning_rate": 4.9975e-05, "loss": 7.1846, "step": 1000 }, { "epoch": 0.0, "learning_rate": 4.99625e-05, "loss": 7.0603, "step": 1500 }, { "epoch": 0.01, "learning_rate": 4.995e-05, "loss": 7.0014, "step": 2000 }, { "epoch": 0.01, "learning_rate": 4.99375e-05, "loss": 6.9367, "step": 2500 }, { "epoch": 0.01, "learning_rate": 4.992500000000001e-05, "loss": 6.885, "step": 3000 }, { "epoch": 0.01, "learning_rate": 4.99125e-05, "loss": 6.8455, "step": 3500 }, { "epoch": 0.01, "learning_rate": 4.99e-05, "loss": 6.8174, "step": 4000 }, { "epoch": 0.01, "learning_rate": 4.9887500000000006e-05, "loss": 6.783, "step": 4500 }, { "epoch": 0.01, "learning_rate": 4.9875000000000006e-05, "loss": 6.7767, "step": 5000 }, { "epoch": 0.01, "learning_rate": 4.98625e-05, "loss": 6.7321, "step": 5500 }, { "epoch": 0.02, "learning_rate": 4.9850000000000006e-05, "loss": 6.7322, "step": 6000 }, { "epoch": 0.02, "learning_rate": 4.9837500000000005e-05, "loss": 6.7188, "step": 6500 }, { "epoch": 0.02, "learning_rate": 4.9825000000000005e-05, "loss": 6.6962, "step": 7000 }, { "epoch": 0.02, "learning_rate": 4.98125e-05, "loss": 6.6968, "step": 7500 }, { "epoch": 0.02, "learning_rate": 4.9800000000000004e-05, "loss": 6.6752, "step": 8000 }, { "epoch": 0.02, "learning_rate": 4.9787500000000004e-05, "loss": 6.6468, "step": 8500 }, { "epoch": 0.02, "learning_rate": 4.9775000000000004e-05, "loss": 6.6489, "step": 9000 }, { "epoch": 0.03, "learning_rate": 4.9762500000000003e-05, "loss": 6.6365, "step": 9500 }, { "epoch": 0.03, "learning_rate": 4.975e-05, "loss": 6.6432, "step": 10000 }, { "epoch": 0.03, "learning_rate": 4.97375e-05, "loss": 6.6361, "step": 10500 }, { "epoch": 0.03, "learning_rate": 4.9725e-05, "loss": 6.62, "step": 11000 }, { "epoch": 0.03, "learning_rate": 4.97125e-05, "loss": 6.609, "step": 11500 }, { "epoch": 0.03, "learning_rate": 4.97e-05, "loss": 6.6101, "step": 12000 }, { "epoch": 0.03, "learning_rate": 4.96875e-05, "loss": 6.6071, "step": 12500 }, { "epoch": 0.03, "learning_rate": 4.967500000000001e-05, "loss": 6.5958, "step": 13000 }, { "epoch": 0.04, "learning_rate": 4.96625e-05, "loss": 6.5976, "step": 13500 }, { "epoch": 0.04, "learning_rate": 4.965e-05, "loss": 6.5817, "step": 14000 }, { "epoch": 0.04, "learning_rate": 4.96375e-05, "loss": 6.578, "step": 14500 }, { "epoch": 0.04, "learning_rate": 4.962500000000001e-05, "loss": 6.5599, "step": 15000 }, { "epoch": 0.04, "learning_rate": 4.96125e-05, "loss": 6.5581, "step": 15500 }, { "epoch": 0.04, "learning_rate": 4.96e-05, "loss": 6.5662, "step": 16000 }, { "epoch": 0.04, "learning_rate": 4.9587500000000006e-05, "loss": 6.5393, "step": 16500 }, { "epoch": 0.05, "learning_rate": 4.9575000000000006e-05, "loss": 6.5344, "step": 17000 }, { "epoch": 0.05, "learning_rate": 4.95625e-05, "loss": 6.5428, "step": 17500 }, { "epoch": 0.05, "learning_rate": 4.9550000000000005e-05, "loss": 6.5357, "step": 18000 }, { "epoch": 0.05, "learning_rate": 4.9537500000000005e-05, "loss": 6.5356, "step": 18500 }, { "epoch": 0.05, "learning_rate": 4.9525000000000004e-05, "loss": 6.5201, "step": 19000 }, { "epoch": 0.05, "learning_rate": 4.95125e-05, "loss": 6.53, "step": 19500 }, { "epoch": 0.05, "learning_rate": 4.9500000000000004e-05, "loss": 6.5217, "step": 20000 }, { "epoch": 0.05, "learning_rate": 4.9487500000000003e-05, "loss": 6.5248, "step": 20500 }, { "epoch": 0.06, "learning_rate": 4.9475e-05, "loss": 6.5032, "step": 21000 }, { "epoch": 0.06, "learning_rate": 4.94625e-05, "loss": 6.4978, "step": 21500 }, { "epoch": 0.06, "learning_rate": 4.945e-05, "loss": 6.5092, "step": 22000 }, { "epoch": 0.06, "learning_rate": 4.94375e-05, "loss": 6.4926, "step": 22500 }, { "epoch": 0.06, "learning_rate": 4.9425e-05, "loss": 6.4769, "step": 23000 }, { "epoch": 0.06, "learning_rate": 4.94125e-05, "loss": 6.5088, "step": 23500 }, { "epoch": 0.06, "learning_rate": 4.94e-05, "loss": 6.4877, "step": 24000 }, { "epoch": 0.07, "learning_rate": 4.93875e-05, "loss": 6.4715, "step": 24500 }, { "epoch": 0.07, "learning_rate": 4.937500000000001e-05, "loss": 6.4829, "step": 25000 }, { "epoch": 0.07, "learning_rate": 4.93625e-05, "loss": 6.4644, "step": 25500 }, { "epoch": 0.07, "learning_rate": 4.935e-05, "loss": 6.4721, "step": 26000 }, { "epoch": 0.07, "learning_rate": 4.93375e-05, "loss": 6.4588, "step": 26500 }, { "epoch": 0.07, "learning_rate": 4.9325000000000006e-05, "loss": 6.4411, "step": 27000 }, { "epoch": 0.07, "learning_rate": 4.93125e-05, "loss": 6.4545, "step": 27500 }, { "epoch": 0.07, "learning_rate": 4.93e-05, "loss": 6.4487, "step": 28000 }, { "epoch": 0.08, "learning_rate": 4.9287500000000005e-05, "loss": 6.4497, "step": 28500 }, { "epoch": 0.08, "learning_rate": 4.9275000000000005e-05, "loss": 6.4515, "step": 29000 }, { "epoch": 0.08, "learning_rate": 4.92625e-05, "loss": 6.456, "step": 29500 }, { "epoch": 0.08, "learning_rate": 4.9250000000000004e-05, "loss": 6.4447, "step": 30000 }, { "epoch": 0.08, "learning_rate": 4.9237500000000004e-05, "loss": 6.4427, "step": 30500 }, { "epoch": 0.08, "learning_rate": 4.9225000000000004e-05, "loss": 6.4583, "step": 31000 }, { "epoch": 0.08, "learning_rate": 4.9212500000000004e-05, "loss": 6.4362, "step": 31500 }, { "epoch": 0.09, "learning_rate": 4.92e-05, "loss": 6.4414, "step": 32000 }, { "epoch": 0.09, "learning_rate": 4.91875e-05, "loss": 6.4324, "step": 32500 }, { "epoch": 0.09, "learning_rate": 4.9175e-05, "loss": 6.4299, "step": 33000 }, { "epoch": 0.09, "learning_rate": 4.91625e-05, "loss": 6.4252, "step": 33500 }, { "epoch": 0.09, "learning_rate": 4.915e-05, "loss": 6.4328, "step": 34000 }, { "epoch": 0.09, "learning_rate": 4.91375e-05, "loss": 6.422, "step": 34500 }, { "epoch": 0.09, "learning_rate": 4.9125e-05, "loss": 6.4237, "step": 35000 }, { "epoch": 0.09, "learning_rate": 4.91125e-05, "loss": 6.4364, "step": 35500 }, { "epoch": 0.1, "learning_rate": 4.91e-05, "loss": 6.41, "step": 36000 }, { "epoch": 0.1, "learning_rate": 4.90875e-05, "loss": 6.4199, "step": 36500 }, { "epoch": 0.1, "learning_rate": 4.907500000000001e-05, "loss": 6.392, "step": 37000 }, { "epoch": 0.1, "learning_rate": 4.90625e-05, "loss": 6.4086, "step": 37500 }, { "epoch": 0.1, "learning_rate": 4.905e-05, "loss": 6.3977, "step": 38000 }, { "epoch": 0.1, "learning_rate": 4.9037500000000006e-05, "loss": 6.4058, "step": 38500 }, { "epoch": 0.1, "learning_rate": 4.9025000000000006e-05, "loss": 6.3978, "step": 39000 }, { "epoch": 0.11, "learning_rate": 4.90125e-05, "loss": 6.3982, "step": 39500 }, { "epoch": 0.11, "learning_rate": 4.9e-05, "loss": 6.4008, "step": 40000 }, { "epoch": 0.11, "learning_rate": 4.8987500000000005e-05, "loss": 6.4041, "step": 40500 }, { "epoch": 0.11, "learning_rate": 4.8975000000000005e-05, "loss": 6.3773, "step": 41000 }, { "epoch": 0.11, "learning_rate": 4.89625e-05, "loss": 6.3782, "step": 41500 }, { "epoch": 0.11, "learning_rate": 4.8950000000000004e-05, "loss": 6.3867, "step": 42000 }, { "epoch": 0.11, "learning_rate": 4.8937500000000004e-05, "loss": 6.3814, "step": 42500 }, { "epoch": 0.11, "learning_rate": 4.8925e-05, "loss": 6.4073, "step": 43000 }, { "epoch": 0.12, "learning_rate": 4.89125e-05, "loss": 6.3844, "step": 43500 }, { "epoch": 0.12, "learning_rate": 4.89e-05, "loss": 6.3753, "step": 44000 }, { "epoch": 0.12, "learning_rate": 4.88875e-05, "loss": 6.3733, "step": 44500 }, { "epoch": 0.12, "learning_rate": 4.8875e-05, "loss": 6.3738, "step": 45000 }, { "epoch": 0.12, "learning_rate": 4.88625e-05, "loss": 6.3584, "step": 45500 }, { "epoch": 0.12, "learning_rate": 4.885e-05, "loss": 6.3686, "step": 46000 }, { "epoch": 0.12, "learning_rate": 4.88375e-05, "loss": 6.3623, "step": 46500 }, { "epoch": 0.13, "learning_rate": 4.8825e-05, "loss": 6.3848, "step": 47000 }, { "epoch": 0.13, "learning_rate": 4.88125e-05, "loss": 6.3679, "step": 47500 }, { "epoch": 0.13, "learning_rate": 4.88e-05, "loss": 6.38, "step": 48000 }, { "epoch": 0.13, "learning_rate": 4.87875e-05, "loss": 6.3657, "step": 48500 }, { "epoch": 0.13, "learning_rate": 4.8775000000000007e-05, "loss": 6.3655, "step": 49000 }, { "epoch": 0.13, "learning_rate": 4.87625e-05, "loss": 6.3701, "step": 49500 }, { "epoch": 0.13, "learning_rate": 4.875e-05, "loss": 6.3604, "step": 50000 }, { "epoch": 0.13, "learning_rate": 4.8737500000000006e-05, "loss": 6.3553, "step": 50500 }, { "epoch": 0.14, "learning_rate": 4.8725000000000005e-05, "loss": 6.3585, "step": 51000 }, { "epoch": 0.14, "learning_rate": 4.87125e-05, "loss": 6.3579, "step": 51500 }, { "epoch": 0.14, "learning_rate": 4.87e-05, "loss": 6.3513, "step": 52000 }, { "epoch": 0.14, "learning_rate": 4.8687500000000004e-05, "loss": 6.3474, "step": 52500 }, { "epoch": 0.14, "learning_rate": 4.8675000000000004e-05, "loss": 6.3496, "step": 53000 }, { "epoch": 0.14, "learning_rate": 4.86625e-05, "loss": 6.3577, "step": 53500 }, { "epoch": 0.14, "learning_rate": 4.8650000000000003e-05, "loss": 6.3742, "step": 54000 }, { "epoch": 0.15, "learning_rate": 4.86375e-05, "loss": 6.3531, "step": 54500 }, { "epoch": 0.15, "learning_rate": 4.8625e-05, "loss": 6.3359, "step": 55000 }, { "epoch": 0.15, "learning_rate": 4.86125e-05, "loss": 6.3448, "step": 55500 }, { "epoch": 0.15, "learning_rate": 4.86e-05, "loss": 6.3423, "step": 56000 }, { "epoch": 0.15, "learning_rate": 4.85875e-05, "loss": 6.3387, "step": 56500 }, { "epoch": 0.15, "learning_rate": 4.8575e-05, "loss": 6.348, "step": 57000 }, { "epoch": 0.15, "learning_rate": 4.85625e-05, "loss": 6.3353, "step": 57500 }, { "epoch": 0.15, "learning_rate": 4.855e-05, "loss": 6.3578, "step": 58000 }, { "epoch": 0.16, "learning_rate": 4.85375e-05, "loss": 6.3299, "step": 58500 }, { "epoch": 0.16, "learning_rate": 4.8525e-05, "loss": 6.346, "step": 59000 }, { "epoch": 0.16, "learning_rate": 4.85125e-05, "loss": 6.3558, "step": 59500 }, { "epoch": 0.16, "learning_rate": 4.85e-05, "loss": 6.3303, "step": 60000 }, { "epoch": 0.16, "learning_rate": 4.84875e-05, "loss": 6.3116, "step": 60500 }, { "epoch": 0.16, "learning_rate": 4.8475000000000006e-05, "loss": 6.3285, "step": 61000 }, { "epoch": 0.16, "learning_rate": 4.84625e-05, "loss": 6.3503, "step": 61500 }, { "epoch": 0.17, "learning_rate": 4.845e-05, "loss": 6.3286, "step": 62000 }, { "epoch": 0.17, "learning_rate": 4.8437500000000005e-05, "loss": 6.3232, "step": 62500 }, { "epoch": 0.17, "learning_rate": 4.8425000000000005e-05, "loss": 6.3221, "step": 63000 }, { "epoch": 0.17, "learning_rate": 4.8412500000000004e-05, "loss": 6.3399, "step": 63500 }, { "epoch": 0.17, "learning_rate": 4.8400000000000004e-05, "loss": 6.3282, "step": 64000 }, { "epoch": 0.17, "learning_rate": 4.8387500000000004e-05, "loss": 6.3168, "step": 64500 }, { "epoch": 0.17, "learning_rate": 4.8375000000000004e-05, "loss": 6.3096, "step": 65000 }, { "epoch": 0.17, "learning_rate": 4.83625e-05, "loss": 6.3125, "step": 65500 }, { "epoch": 0.18, "learning_rate": 4.835e-05, "loss": 6.3242, "step": 66000 }, { "epoch": 0.18, "learning_rate": 4.83375e-05, "loss": 6.2921, "step": 66500 }, { "epoch": 0.18, "learning_rate": 4.8325e-05, "loss": 6.3323, "step": 67000 }, { "epoch": 0.18, "learning_rate": 4.83125e-05, "loss": 6.3232, "step": 67500 }, { "epoch": 0.18, "learning_rate": 4.83e-05, "loss": 6.3089, "step": 68000 }, { "epoch": 0.18, "learning_rate": 4.82875e-05, "loss": 6.3083, "step": 68500 }, { "epoch": 0.18, "learning_rate": 4.8275e-05, "loss": 6.3072, "step": 69000 }, { "epoch": 0.19, "learning_rate": 4.826250000000001e-05, "loss": 6.3043, "step": 69500 }, { "epoch": 0.19, "learning_rate": 4.825e-05, "loss": 6.3035, "step": 70000 }, { "epoch": 0.19, "learning_rate": 4.82375e-05, "loss": 6.3174, "step": 70500 }, { "epoch": 0.19, "learning_rate": 4.822500000000001e-05, "loss": 6.3086, "step": 71000 }, { "epoch": 0.19, "learning_rate": 4.8212500000000006e-05, "loss": 6.2835, "step": 71500 }, { "epoch": 0.19, "learning_rate": 4.82e-05, "loss": 6.303, "step": 72000 }, { "epoch": 0.19, "learning_rate": 4.81875e-05, "loss": 6.3227, "step": 72500 }, { "epoch": 0.19, "learning_rate": 4.8175000000000005e-05, "loss": 6.3044, "step": 73000 }, { "epoch": 0.2, "learning_rate": 4.8162500000000005e-05, "loss": 6.3235, "step": 73500 }, { "epoch": 0.2, "learning_rate": 4.815e-05, "loss": 6.3012, "step": 74000 }, { "epoch": 0.2, "learning_rate": 4.8137500000000005e-05, "loss": 6.3047, "step": 74500 }, { "epoch": 0.2, "learning_rate": 4.8125000000000004e-05, "loss": 6.2825, "step": 75000 }, { "epoch": 0.2, "learning_rate": 4.8112500000000004e-05, "loss": 6.2836, "step": 75500 }, { "epoch": 0.2, "learning_rate": 4.8100000000000004e-05, "loss": 6.2975, "step": 76000 }, { "epoch": 0.2, "learning_rate": 4.80875e-05, "loss": 6.2972, "step": 76500 }, { "epoch": 0.2, "learning_rate": 4.8075e-05, "loss": 6.3025, "step": 77000 }, { "epoch": 0.21, "learning_rate": 4.80625e-05, "loss": 6.2771, "step": 77500 }, { "epoch": 0.21, "learning_rate": 4.805e-05, "loss": 6.3021, "step": 78000 }, { "epoch": 0.21, "learning_rate": 4.80375e-05, "loss": 6.2933, "step": 78500 }, { "epoch": 0.21, "learning_rate": 4.8025e-05, "loss": 6.293, "step": 79000 }, { "epoch": 0.21, "learning_rate": 4.80125e-05, "loss": 6.2867, "step": 79500 }, { "epoch": 0.21, "learning_rate": 4.8e-05, "loss": 6.2955, "step": 80000 }, { "epoch": 0.21, "learning_rate": 4.79875e-05, "loss": 6.3036, "step": 80500 }, { "epoch": 0.22, "learning_rate": 4.7975e-05, "loss": 6.2936, "step": 81000 }, { "epoch": 0.22, "learning_rate": 4.796250000000001e-05, "loss": 6.2663, "step": 81500 }, { "epoch": 0.22, "learning_rate": 4.795e-05, "loss": 6.3006, "step": 82000 }, { "epoch": 0.22, "learning_rate": 4.79375e-05, "loss": 6.2898, "step": 82500 }, { "epoch": 0.22, "learning_rate": 4.7925000000000006e-05, "loss": 6.2857, "step": 83000 }, { "epoch": 0.22, "learning_rate": 4.7912500000000006e-05, "loss": 6.3123, "step": 83500 }, { "epoch": 0.22, "learning_rate": 4.79e-05, "loss": 6.281, "step": 84000 }, { "epoch": 0.22, "learning_rate": 4.78875e-05, "loss": 6.2991, "step": 84500 }, { "epoch": 0.23, "learning_rate": 4.7875000000000005e-05, "loss": 6.2727, "step": 85000 }, { "epoch": 0.23, "learning_rate": 4.7862500000000005e-05, "loss": 6.2679, "step": 85500 }, { "epoch": 0.23, "learning_rate": 4.785e-05, "loss": 6.2813, "step": 86000 }, { "epoch": 0.23, "learning_rate": 4.7837500000000004e-05, "loss": 6.2848, "step": 86500 }, { "epoch": 0.23, "learning_rate": 4.7825000000000004e-05, "loss": 6.2623, "step": 87000 }, { "epoch": 0.23, "learning_rate": 4.7812500000000003e-05, "loss": 6.2836, "step": 87500 }, { "epoch": 0.23, "learning_rate": 4.78e-05, "loss": 6.2578, "step": 88000 }, { "epoch": 0.24, "learning_rate": 4.77875e-05, "loss": 6.2692, "step": 88500 }, { "epoch": 0.24, "learning_rate": 4.7775e-05, "loss": 6.277, "step": 89000 }, { "epoch": 0.24, "learning_rate": 4.77625e-05, "loss": 6.2761, "step": 89500 }, { "epoch": 0.24, "learning_rate": 4.775e-05, "loss": 6.2653, "step": 90000 }, { "epoch": 0.24, "learning_rate": 4.77375e-05, "loss": 6.2724, "step": 90500 }, { "epoch": 0.24, "learning_rate": 4.7725e-05, "loss": 6.2613, "step": 91000 }, { "epoch": 0.24, "learning_rate": 4.771250000000001e-05, "loss": 6.2635, "step": 91500 }, { "epoch": 0.24, "learning_rate": 4.77e-05, "loss": 6.254, "step": 92000 }, { "epoch": 0.25, "learning_rate": 4.76875e-05, "loss": 6.2754, "step": 92500 }, { "epoch": 0.25, "learning_rate": 4.7675e-05, "loss": 6.2731, "step": 93000 }, { "epoch": 0.25, "learning_rate": 4.7662500000000007e-05, "loss": 6.2535, "step": 93500 }, { "epoch": 0.25, "learning_rate": 4.765e-05, "loss": 6.258, "step": 94000 }, { "epoch": 0.25, "learning_rate": 4.76375e-05, "loss": 6.2562, "step": 94500 }, { "epoch": 0.25, "learning_rate": 4.7625000000000006e-05, "loss": 6.2643, "step": 95000 }, { "epoch": 0.25, "learning_rate": 4.7612500000000005e-05, "loss": 6.2518, "step": 95500 }, { "epoch": 0.26, "learning_rate": 4.76e-05, "loss": 6.2529, "step": 96000 }, { "epoch": 0.26, "learning_rate": 4.7587500000000005e-05, "loss": 6.2518, "step": 96500 }, { "epoch": 0.26, "learning_rate": 4.7575000000000004e-05, "loss": 6.2556, "step": 97000 }, { "epoch": 0.26, "learning_rate": 4.7562500000000004e-05, "loss": 6.2406, "step": 97500 }, { "epoch": 0.26, "learning_rate": 4.755e-05, "loss": 6.2499, "step": 98000 }, { "epoch": 0.26, "learning_rate": 4.7537500000000004e-05, "loss": 6.257, "step": 98500 }, { "epoch": 0.26, "learning_rate": 4.7525e-05, "loss": 6.2704, "step": 99000 }, { "epoch": 0.26, "learning_rate": 4.75125e-05, "loss": 6.2595, "step": 99500 }, { "epoch": 0.27, "learning_rate": 4.75e-05, "loss": 6.2472, "step": 100000 }, { "epoch": 0.27, "learning_rate": 4.74875e-05, "loss": 6.2507, "step": 100500 }, { "epoch": 0.27, "learning_rate": 4.7475e-05, "loss": 6.2301, "step": 101000 }, { "epoch": 0.27, "learning_rate": 4.74625e-05, "loss": 6.2448, "step": 101500 }, { "epoch": 0.27, "learning_rate": 4.745e-05, "loss": 6.24, "step": 102000 }, { "epoch": 0.27, "learning_rate": 4.74375e-05, "loss": 6.2396, "step": 102500 }, { "epoch": 0.27, "learning_rate": 4.7425e-05, "loss": 6.2501, "step": 103000 }, { "epoch": 0.28, "learning_rate": 4.741250000000001e-05, "loss": 6.2499, "step": 103500 }, { "epoch": 0.28, "learning_rate": 4.74e-05, "loss": 6.2324, "step": 104000 }, { "epoch": 0.28, "learning_rate": 4.73875e-05, "loss": 6.2461, "step": 104500 }, { "epoch": 0.28, "learning_rate": 4.7375e-05, "loss": 6.2677, "step": 105000 }, { "epoch": 0.28, "learning_rate": 4.7362500000000006e-05, "loss": 6.2472, "step": 105500 }, { "epoch": 0.28, "learning_rate": 4.735e-05, "loss": 6.243, "step": 106000 }, { "epoch": 0.28, "learning_rate": 4.73375e-05, "loss": 6.2458, "step": 106500 }, { "epoch": 0.28, "learning_rate": 4.7325000000000005e-05, "loss": 6.2454, "step": 107000 }, { "epoch": 0.29, "learning_rate": 4.7312500000000005e-05, "loss": 6.241, "step": 107500 }, { "epoch": 0.29, "learning_rate": 4.73e-05, "loss": 6.2302, "step": 108000 }, { "epoch": 0.29, "learning_rate": 4.7287500000000004e-05, "loss": 6.2298, "step": 108500 }, { "epoch": 0.29, "learning_rate": 4.7275000000000004e-05, "loss": 6.2312, "step": 109000 }, { "epoch": 0.29, "learning_rate": 4.7262500000000004e-05, "loss": 6.2472, "step": 109500 }, { "epoch": 0.29, "learning_rate": 4.7249999999999997e-05, "loss": 6.2305, "step": 110000 }, { "epoch": 0.29, "learning_rate": 4.72375e-05, "loss": 6.243, "step": 110500 }, { "epoch": 0.3, "learning_rate": 4.7225e-05, "loss": 6.2212, "step": 111000 }, { "epoch": 0.3, "learning_rate": 4.72125e-05, "loss": 6.2331, "step": 111500 }, { "epoch": 0.3, "learning_rate": 4.72e-05, "loss": 6.2457, "step": 112000 }, { "epoch": 0.3, "learning_rate": 4.71875e-05, "loss": 6.2292, "step": 112500 }, { "epoch": 0.3, "learning_rate": 4.7175e-05, "loss": 6.2321, "step": 113000 }, { "epoch": 0.3, "learning_rate": 4.71625e-05, "loss": 6.2288, "step": 113500 }, { "epoch": 0.3, "learning_rate": 4.715e-05, "loss": 6.2307, "step": 114000 }, { "epoch": 0.3, "learning_rate": 4.71375e-05, "loss": 6.2144, "step": 114500 }, { "epoch": 0.31, "learning_rate": 4.7125e-05, "loss": 6.2273, "step": 115000 }, { "epoch": 0.31, "learning_rate": 4.711250000000001e-05, "loss": 6.2154, "step": 115500 }, { "epoch": 0.31, "learning_rate": 4.71e-05, "loss": 6.2294, "step": 116000 }, { "epoch": 0.31, "learning_rate": 4.70875e-05, "loss": 6.2184, "step": 116500 }, { "epoch": 0.31, "learning_rate": 4.7075e-05, "loss": 6.2182, "step": 117000 }, { "epoch": 0.31, "learning_rate": 4.7062500000000006e-05, "loss": 6.2353, "step": 117500 }, { "epoch": 0.31, "learning_rate": 4.705e-05, "loss": 6.2445, "step": 118000 }, { "epoch": 0.32, "learning_rate": 4.70375e-05, "loss": 6.2154, "step": 118500 }, { "epoch": 0.32, "learning_rate": 4.7025000000000005e-05, "loss": 6.2292, "step": 119000 }, { "epoch": 0.32, "learning_rate": 4.7012500000000004e-05, "loss": 6.2221, "step": 119500 }, { "epoch": 0.32, "learning_rate": 4.7e-05, "loss": 6.2265, "step": 120000 }, { "epoch": 0.32, "learning_rate": 4.6987500000000004e-05, "loss": 6.2149, "step": 120500 }, { "epoch": 0.32, "learning_rate": 4.6975000000000003e-05, "loss": 6.2332, "step": 121000 }, { "epoch": 0.32, "learning_rate": 4.69625e-05, "loss": 6.2135, "step": 121500 }, { "epoch": 0.32, "learning_rate": 4.695e-05, "loss": 6.2126, "step": 122000 }, { "epoch": 0.33, "learning_rate": 4.69375e-05, "loss": 6.2141, "step": 122500 }, { "epoch": 0.33, "learning_rate": 4.6925e-05, "loss": 6.222, "step": 123000 }, { "epoch": 0.33, "learning_rate": 4.69125e-05, "loss": 6.2017, "step": 123500 }, { "epoch": 0.33, "learning_rate": 4.69e-05, "loss": 6.2106, "step": 124000 }, { "epoch": 0.33, "learning_rate": 4.68875e-05, "loss": 6.1869, "step": 124500 }, { "epoch": 0.33, "learning_rate": 4.6875e-05, "loss": 6.212, "step": 125000 }, { "epoch": 0.33, "learning_rate": 4.68625e-05, "loss": 6.2164, "step": 125500 }, { "epoch": 0.34, "learning_rate": 4.685000000000001e-05, "loss": 6.2007, "step": 126000 }, { "epoch": 0.34, "learning_rate": 4.68375e-05, "loss": 6.2198, "step": 126500 }, { "epoch": 0.34, "learning_rate": 4.6825e-05, "loss": 6.2191, "step": 127000 }, { "epoch": 0.34, "learning_rate": 4.6812500000000006e-05, "loss": 6.227, "step": 127500 }, { "epoch": 0.34, "learning_rate": 4.6800000000000006e-05, "loss": 6.2136, "step": 128000 }, { "epoch": 0.34, "learning_rate": 4.67875e-05, "loss": 6.2052, "step": 128500 }, { "epoch": 0.34, "learning_rate": 4.6775000000000005e-05, "loss": 6.2189, "step": 129000 }, { "epoch": 0.34, "learning_rate": 4.6762500000000005e-05, "loss": 6.2116, "step": 129500 }, { "epoch": 0.35, "learning_rate": 4.6750000000000005e-05, "loss": 6.2319, "step": 130000 }, { "epoch": 0.35, "learning_rate": 4.67375e-05, "loss": 6.2082, "step": 130500 }, { "epoch": 0.35, "learning_rate": 4.6725000000000004e-05, "loss": 6.2176, "step": 131000 }, { "epoch": 0.35, "learning_rate": 4.6712500000000004e-05, "loss": 6.2143, "step": 131500 }, { "epoch": 0.35, "learning_rate": 4.6700000000000003e-05, "loss": 6.2003, "step": 132000 }, { "epoch": 0.35, "learning_rate": 4.66875e-05, "loss": 6.1726, "step": 132500 }, { "epoch": 0.35, "learning_rate": 4.6675e-05, "loss": 6.2192, "step": 133000 }, { "epoch": 0.36, "learning_rate": 4.66625e-05, "loss": 6.208, "step": 133500 }, { "epoch": 0.36, "learning_rate": 4.665e-05, "loss": 6.2095, "step": 134000 }, { "epoch": 0.36, "learning_rate": 4.66375e-05, "loss": 6.1838, "step": 134500 }, { "epoch": 0.36, "learning_rate": 4.6625e-05, "loss": 6.1931, "step": 135000 }, { "epoch": 0.36, "learning_rate": 4.66125e-05, "loss": 6.2061, "step": 135500 }, { "epoch": 0.36, "learning_rate": 4.660000000000001e-05, "loss": 6.2037, "step": 136000 }, { "epoch": 0.36, "learning_rate": 4.65875e-05, "loss": 6.1895, "step": 136500 }, { "epoch": 0.36, "learning_rate": 4.6575e-05, "loss": 6.185, "step": 137000 }, { "epoch": 0.37, "learning_rate": 4.65625e-05, "loss": 6.1963, "step": 137500 }, { "epoch": 0.37, "learning_rate": 4.655000000000001e-05, "loss": 6.1765, "step": 138000 }, { "epoch": 0.37, "learning_rate": 4.65375e-05, "loss": 6.2025, "step": 138500 }, { "epoch": 0.37, "learning_rate": 4.6525e-05, "loss": 6.1965, "step": 139000 }, { "epoch": 0.37, "learning_rate": 4.6512500000000006e-05, "loss": 6.1961, "step": 139500 }, { "epoch": 0.37, "learning_rate": 4.6500000000000005e-05, "loss": 6.1813, "step": 140000 }, { "epoch": 0.37, "learning_rate": 4.64875e-05, "loss": 6.1752, "step": 140500 }, { "epoch": 0.38, "learning_rate": 4.6475000000000005e-05, "loss": 6.1696, "step": 141000 }, { "epoch": 0.38, "learning_rate": 4.6462500000000005e-05, "loss": 6.1798, "step": 141500 }, { "epoch": 0.38, "learning_rate": 4.6450000000000004e-05, "loss": 6.1865, "step": 142000 }, { "epoch": 0.38, "learning_rate": 4.64375e-05, "loss": 6.1812, "step": 142500 }, { "epoch": 0.38, "learning_rate": 4.6425000000000004e-05, "loss": 6.1554, "step": 143000 }, { "epoch": 0.38, "learning_rate": 4.64125e-05, "loss": 6.1616, "step": 143500 }, { "epoch": 0.38, "learning_rate": 4.64e-05, "loss": 6.1788, "step": 144000 }, { "epoch": 0.38, "learning_rate": 4.63875e-05, "loss": 6.1679, "step": 144500 }, { "epoch": 0.39, "learning_rate": 4.6375e-05, "loss": 6.1535, "step": 145000 }, { "epoch": 0.39, "learning_rate": 4.63625e-05, "loss": 6.164, "step": 145500 }, { "epoch": 0.39, "learning_rate": 4.635e-05, "loss": 6.1527, "step": 146000 }, { "epoch": 0.39, "learning_rate": 4.63375e-05, "loss": 6.1572, "step": 146500 }, { "epoch": 0.39, "learning_rate": 4.6325e-05, "loss": 6.1587, "step": 147000 }, { "epoch": 0.39, "learning_rate": 4.63125e-05, "loss": 6.1409, "step": 147500 }, { "epoch": 0.39, "learning_rate": 4.630000000000001e-05, "loss": 6.1641, "step": 148000 }, { "epoch": 0.4, "learning_rate": 4.62875e-05, "loss": 6.1566, "step": 148500 }, { "epoch": 0.4, "learning_rate": 4.6275e-05, "loss": 6.1608, "step": 149000 }, { "epoch": 0.4, "learning_rate": 4.6262500000000006e-05, "loss": 6.1438, "step": 149500 }, { "epoch": 0.4, "learning_rate": 4.6250000000000006e-05, "loss": 6.1559, "step": 150000 }, { "epoch": 0.4, "learning_rate": 4.62375e-05, "loss": 6.1411, "step": 150500 }, { "epoch": 0.4, "learning_rate": 4.6225e-05, "loss": 6.1565, "step": 151000 }, { "epoch": 0.4, "learning_rate": 4.6212500000000005e-05, "loss": 6.1352, "step": 151500 }, { "epoch": 0.4, "learning_rate": 4.6200000000000005e-05, "loss": 6.1581, "step": 152000 }, { "epoch": 0.41, "learning_rate": 4.61875e-05, "loss": 6.1319, "step": 152500 }, { "epoch": 0.41, "learning_rate": 4.6175000000000004e-05, "loss": 6.1177, "step": 153000 }, { "epoch": 0.41, "learning_rate": 4.6162500000000004e-05, "loss": 6.1324, "step": 153500 }, { "epoch": 0.41, "learning_rate": 4.6150000000000004e-05, "loss": 6.1204, "step": 154000 }, { "epoch": 0.41, "learning_rate": 4.61375e-05, "loss": 6.1234, "step": 154500 }, { "epoch": 0.41, "learning_rate": 4.6125e-05, "loss": 6.1216, "step": 155000 }, { "epoch": 0.41, "learning_rate": 4.61125e-05, "loss": 6.1022, "step": 155500 }, { "epoch": 0.42, "learning_rate": 4.61e-05, "loss": 6.1066, "step": 156000 }, { "epoch": 0.42, "learning_rate": 4.60875e-05, "loss": 6.1129, "step": 156500 }, { "epoch": 0.42, "learning_rate": 4.6075e-05, "loss": 6.0909, "step": 157000 }, { "epoch": 0.42, "learning_rate": 4.60625e-05, "loss": 6.0852, "step": 157500 }, { "epoch": 0.42, "learning_rate": 4.605e-05, "loss": 6.0783, "step": 158000 }, { "epoch": 0.42, "learning_rate": 4.60375e-05, "loss": 6.0566, "step": 158500 }, { "epoch": 0.42, "learning_rate": 4.6025e-05, "loss": 6.0559, "step": 159000 }, { "epoch": 0.42, "learning_rate": 4.60125e-05, "loss": 6.0384, "step": 159500 }, { "epoch": 0.43, "learning_rate": 4.600000000000001e-05, "loss": 6.044, "step": 160000 }, { "epoch": 0.43, "learning_rate": 4.59875e-05, "loss": 6.0359, "step": 160500 }, { "epoch": 0.43, "learning_rate": 4.5975e-05, "loss": 6.0204, "step": 161000 }, { "epoch": 0.43, "learning_rate": 4.5962500000000006e-05, "loss": 6.0422, "step": 161500 }, { "epoch": 0.43, "learning_rate": 4.5950000000000006e-05, "loss": 6.0262, "step": 162000 }, { "epoch": 0.43, "learning_rate": 4.59375e-05, "loss": 6.0279, "step": 162500 }, { "epoch": 0.43, "learning_rate": 4.5925e-05, "loss": 6.005, "step": 163000 }, { "epoch": 0.44, "learning_rate": 4.5912500000000005e-05, "loss": 6.0022, "step": 163500 }, { "epoch": 0.44, "learning_rate": 4.5900000000000004e-05, "loss": 5.9959, "step": 164000 }, { "epoch": 0.44, "learning_rate": 4.58875e-05, "loss": 5.9914, "step": 164500 }, { "epoch": 0.44, "learning_rate": 4.5875000000000004e-05, "loss": 5.9793, "step": 165000 }, { "epoch": 0.44, "learning_rate": 4.5862500000000003e-05, "loss": 5.9703, "step": 165500 }, { "epoch": 0.44, "learning_rate": 4.585e-05, "loss": 5.9709, "step": 166000 }, { "epoch": 0.44, "learning_rate": 4.58375e-05, "loss": 5.9543, "step": 166500 }, { "epoch": 0.44, "learning_rate": 4.5825e-05, "loss": 5.9628, "step": 167000 }, { "epoch": 0.45, "learning_rate": 4.58125e-05, "loss": 5.9688, "step": 167500 }, { "epoch": 0.45, "learning_rate": 4.58e-05, "loss": 5.9328, "step": 168000 }, { "epoch": 0.45, "learning_rate": 4.57875e-05, "loss": 5.9424, "step": 168500 }, { "epoch": 0.45, "learning_rate": 4.5775e-05, "loss": 5.9221, "step": 169000 }, { "epoch": 0.45, "learning_rate": 4.57625e-05, "loss": 5.9372, "step": 169500 }, { "epoch": 0.45, "learning_rate": 4.575e-05, "loss": 5.941, "step": 170000 }, { "epoch": 0.45, "learning_rate": 4.57375e-05, "loss": 5.9076, "step": 170500 }, { "epoch": 0.46, "learning_rate": 4.5725e-05, "loss": 5.927, "step": 171000 }, { "epoch": 0.46, "learning_rate": 4.57125e-05, "loss": 5.9184, "step": 171500 }, { "epoch": 0.46, "learning_rate": 4.5700000000000006e-05, "loss": 5.8988, "step": 172000 }, { "epoch": 0.46, "learning_rate": 4.56875e-05, "loss": 5.9116, "step": 172500 }, { "epoch": 0.46, "learning_rate": 4.5675e-05, "loss": 5.8822, "step": 173000 }, { "epoch": 0.46, "learning_rate": 4.5662500000000005e-05, "loss": 5.8843, "step": 173500 }, { "epoch": 0.46, "learning_rate": 4.5650000000000005e-05, "loss": 5.8552, "step": 174000 }, { "epoch": 0.46, "learning_rate": 4.56375e-05, "loss": 5.8856, "step": 174500 }, { "epoch": 0.47, "learning_rate": 4.5625e-05, "loss": 5.8581, "step": 175000 }, { "epoch": 0.47, "learning_rate": 4.5612500000000004e-05, "loss": 5.8598, "step": 175500 }, { "epoch": 0.47, "learning_rate": 4.5600000000000004e-05, "loss": 5.8468, "step": 176000 }, { "epoch": 0.47, "learning_rate": 4.55875e-05, "loss": 5.8423, "step": 176500 }, { "epoch": 0.47, "learning_rate": 4.5575e-05, "loss": 5.8264, "step": 177000 }, { "epoch": 0.47, "learning_rate": 4.55625e-05, "loss": 5.8408, "step": 177500 }, { "epoch": 0.47, "learning_rate": 4.555e-05, "loss": 5.8315, "step": 178000 }, { "epoch": 0.48, "learning_rate": 4.55375e-05, "loss": 5.8038, "step": 178500 }, { "epoch": 0.48, "learning_rate": 4.5525e-05, "loss": 5.8139, "step": 179000 }, { "epoch": 0.48, "learning_rate": 4.55125e-05, "loss": 5.8035, "step": 179500 }, { "epoch": 0.48, "learning_rate": 4.55e-05, "loss": 5.7955, "step": 180000 }, { "epoch": 0.48, "learning_rate": 4.54875e-05, "loss": 5.8036, "step": 180500 }, { "epoch": 0.48, "learning_rate": 4.5475e-05, "loss": 5.7897, "step": 181000 }, { "epoch": 0.48, "learning_rate": 4.54625e-05, "loss": 5.7969, "step": 181500 }, { "epoch": 0.48, "learning_rate": 4.545000000000001e-05, "loss": 5.7728, "step": 182000 }, { "epoch": 0.49, "learning_rate": 4.54375e-05, "loss": 5.7942, "step": 182500 }, { "epoch": 0.49, "learning_rate": 4.5425e-05, "loss": 5.7591, "step": 183000 }, { "epoch": 0.49, "learning_rate": 4.54125e-05, "loss": 5.7735, "step": 183500 }, { "epoch": 0.49, "learning_rate": 4.5400000000000006e-05, "loss": 5.7549, "step": 184000 }, { "epoch": 0.49, "learning_rate": 4.53875e-05, "loss": 5.7349, "step": 184500 }, { "epoch": 0.49, "learning_rate": 4.5375e-05, "loss": 5.7535, "step": 185000 }, { "epoch": 0.49, "learning_rate": 4.5362500000000005e-05, "loss": 5.7399, "step": 185500 }, { "epoch": 0.5, "learning_rate": 4.5350000000000005e-05, "loss": 5.7354, "step": 186000 }, { "epoch": 0.5, "learning_rate": 4.53375e-05, "loss": 5.7241, "step": 186500 }, { "epoch": 0.5, "learning_rate": 4.5325000000000004e-05, "loss": 5.7265, "step": 187000 }, { "epoch": 0.5, "learning_rate": 4.5312500000000004e-05, "loss": 5.7175, "step": 187500 }, { "epoch": 0.5, "learning_rate": 4.53e-05, "loss": 5.6938, "step": 188000 }, { "epoch": 0.5, "learning_rate": 4.52875e-05, "loss": 5.7273, "step": 188500 }, { "epoch": 0.5, "learning_rate": 4.5275e-05, "loss": 5.693, "step": 189000 }, { "epoch": 0.5, "learning_rate": 4.52625e-05, "loss": 5.7199, "step": 189500 }, { "epoch": 0.51, "learning_rate": 4.525e-05, "loss": 5.6914, "step": 190000 }, { "epoch": 0.51, "learning_rate": 4.52375e-05, "loss": 5.7079, "step": 190500 }, { "epoch": 0.51, "learning_rate": 4.5225e-05, "loss": 5.6993, "step": 191000 }, { "epoch": 0.51, "learning_rate": 4.52125e-05, "loss": 5.6905, "step": 191500 }, { "epoch": 0.51, "learning_rate": 4.52e-05, "loss": 5.6906, "step": 192000 }, { "epoch": 0.51, "learning_rate": 4.518750000000001e-05, "loss": 5.6776, "step": 192500 }, { "epoch": 0.51, "learning_rate": 4.5175e-05, "loss": 5.656, "step": 193000 }, { "epoch": 0.52, "learning_rate": 4.51625e-05, "loss": 5.6694, "step": 193500 }, { "epoch": 0.52, "learning_rate": 4.5150000000000006e-05, "loss": 5.6717, "step": 194000 }, { "epoch": 0.52, "learning_rate": 4.5137500000000006e-05, "loss": 5.6654, "step": 194500 }, { "epoch": 0.52, "learning_rate": 4.5125e-05, "loss": 5.6541, "step": 195000 }, { "epoch": 0.52, "learning_rate": 4.51125e-05, "loss": 5.6553, "step": 195500 }, { "epoch": 0.52, "learning_rate": 4.5100000000000005e-05, "loss": 5.6695, "step": 196000 }, { "epoch": 0.52, "learning_rate": 4.5087500000000005e-05, "loss": 5.653, "step": 196500 }, { "epoch": 0.52, "learning_rate": 4.5075e-05, "loss": 5.6323, "step": 197000 }, { "epoch": 0.53, "learning_rate": 4.5062500000000004e-05, "loss": 5.6519, "step": 197500 }, { "epoch": 0.53, "learning_rate": 4.5050000000000004e-05, "loss": 5.6514, "step": 198000 }, { "epoch": 0.53, "learning_rate": 4.5037500000000004e-05, "loss": 5.6298, "step": 198500 }, { "epoch": 0.53, "learning_rate": 4.5025000000000003e-05, "loss": 5.6335, "step": 199000 }, { "epoch": 0.53, "learning_rate": 4.50125e-05, "loss": 5.6355, "step": 199500 }, { "epoch": 0.53, "learning_rate": 4.5e-05, "loss": 5.6297, "step": 200000 }, { "epoch": 0.53, "learning_rate": 4.49875e-05, "loss": 5.6118, "step": 200500 }, { "epoch": 0.54, "learning_rate": 4.4975e-05, "loss": 5.6259, "step": 201000 }, { "epoch": 0.54, "learning_rate": 4.49625e-05, "loss": 5.6393, "step": 201500 }, { "epoch": 0.54, "learning_rate": 4.495e-05, "loss": 5.6335, "step": 202000 }, { "epoch": 0.54, "learning_rate": 4.49375e-05, "loss": 5.6075, "step": 202500 }, { "epoch": 0.54, "learning_rate": 4.4925e-05, "loss": 5.5999, "step": 203000 }, { "epoch": 0.54, "learning_rate": 4.49125e-05, "loss": 5.6214, "step": 203500 }, { "epoch": 0.54, "learning_rate": 4.49e-05, "loss": 5.6006, "step": 204000 }, { "epoch": 0.54, "learning_rate": 4.488750000000001e-05, "loss": 5.6088, "step": 204500 }, { "epoch": 0.55, "learning_rate": 4.4875e-05, "loss": 5.5965, "step": 205000 }, { "epoch": 0.55, "learning_rate": 4.48625e-05, "loss": 5.6017, "step": 205500 }, { "epoch": 0.55, "learning_rate": 4.4850000000000006e-05, "loss": 5.6059, "step": 206000 }, { "epoch": 0.55, "learning_rate": 4.4837500000000006e-05, "loss": 5.5798, "step": 206500 }, { "epoch": 0.55, "learning_rate": 4.4825e-05, "loss": 5.5669, "step": 207000 }, { "epoch": 0.55, "learning_rate": 4.4812500000000005e-05, "loss": 5.5845, "step": 207500 }, { "epoch": 0.55, "learning_rate": 4.4800000000000005e-05, "loss": 5.5704, "step": 208000 }, { "epoch": 0.56, "learning_rate": 4.4787500000000004e-05, "loss": 5.572, "step": 208500 }, { "epoch": 0.56, "learning_rate": 4.4775e-05, "loss": 5.5886, "step": 209000 }, { "epoch": 0.56, "learning_rate": 4.4762500000000004e-05, "loss": 5.5666, "step": 209500 }, { "epoch": 0.56, "learning_rate": 4.4750000000000004e-05, "loss": 5.5597, "step": 210000 }, { "epoch": 0.56, "learning_rate": 4.47375e-05, "loss": 5.5866, "step": 210500 }, { "epoch": 0.56, "learning_rate": 4.4725e-05, "loss": 5.5591, "step": 211000 }, { "epoch": 0.56, "learning_rate": 4.47125e-05, "loss": 5.5661, "step": 211500 }, { "epoch": 0.56, "learning_rate": 4.47e-05, "loss": 5.559, "step": 212000 }, { "epoch": 0.57, "learning_rate": 4.46875e-05, "loss": 5.5626, "step": 212500 }, { "epoch": 0.57, "learning_rate": 4.4675e-05, "loss": 5.5575, "step": 213000 }, { "epoch": 0.57, "learning_rate": 4.46625e-05, "loss": 5.538, "step": 213500 }, { "epoch": 0.57, "learning_rate": 4.465e-05, "loss": 5.5356, "step": 214000 }, { "epoch": 0.57, "learning_rate": 4.463750000000001e-05, "loss": 5.5333, "step": 214500 }, { "epoch": 0.57, "learning_rate": 4.4625e-05, "loss": 5.5309, "step": 215000 }, { "epoch": 0.57, "learning_rate": 4.46125e-05, "loss": 5.5488, "step": 215500 }, { "epoch": 0.58, "learning_rate": 4.46e-05, "loss": 5.5315, "step": 216000 }, { "epoch": 0.58, "learning_rate": 4.4587500000000006e-05, "loss": 5.5161, "step": 216500 }, { "epoch": 0.58, "learning_rate": 4.4575e-05, "loss": 5.5233, "step": 217000 }, { "epoch": 0.58, "learning_rate": 4.45625e-05, "loss": 5.5219, "step": 217500 }, { "epoch": 0.58, "learning_rate": 4.4550000000000005e-05, "loss": 5.5323, "step": 218000 }, { "epoch": 0.58, "learning_rate": 4.4537500000000005e-05, "loss": 5.5141, "step": 218500 }, { "epoch": 0.58, "learning_rate": 4.4525e-05, "loss": 5.5096, "step": 219000 }, { "epoch": 0.58, "learning_rate": 4.4512500000000005e-05, "loss": 5.5089, "step": 219500 }, { "epoch": 0.59, "learning_rate": 4.4500000000000004e-05, "loss": 5.029, "step": 220000 }, { "epoch": 0.59, "learning_rate": 4.4487500000000004e-05, "loss": 3.8462, "step": 220500 }, { "epoch": 0.59, "learning_rate": 4.4475e-05, "loss": 3.7946, "step": 221000 }, { "epoch": 0.59, "learning_rate": 4.44625e-05, "loss": 3.7581, "step": 221500 }, { "epoch": 0.59, "learning_rate": 4.445e-05, "loss": 3.7255, "step": 222000 }, { "epoch": 0.59, "learning_rate": 4.44375e-05, "loss": 3.6826, "step": 222500 }, { "epoch": 0.59, "learning_rate": 4.4425e-05, "loss": 3.67, "step": 223000 }, { "epoch": 0.6, "learning_rate": 4.44125e-05, "loss": 3.6521, "step": 223500 }, { "epoch": 0.6, "learning_rate": 4.44e-05, "loss": 3.6438, "step": 224000 }, { "epoch": 0.6, "learning_rate": 4.43875e-05, "loss": 3.64, "step": 224500 }, { "epoch": 0.6, "learning_rate": 4.4375e-05, "loss": 3.6093, "step": 225000 }, { "epoch": 0.6, "learning_rate": 4.43625e-05, "loss": 3.5801, "step": 225500 }, { "epoch": 0.6, "learning_rate": 4.435e-05, "loss": 3.573, "step": 226000 }, { "epoch": 0.6, "learning_rate": 4.433750000000001e-05, "loss": 3.5784, "step": 226500 }, { "epoch": 0.6, "learning_rate": 4.4325e-05, "loss": 3.5298, "step": 227000 }, { "epoch": 0.61, "learning_rate": 4.43125e-05, "loss": 3.5173, "step": 227500 }, { "epoch": 0.61, "learning_rate": 4.43e-05, "loss": 3.5157, "step": 228000 }, { "epoch": 0.61, "learning_rate": 4.4287500000000006e-05, "loss": 3.5081, "step": 228500 }, { "epoch": 0.61, "learning_rate": 4.4275e-05, "loss": 3.4888, "step": 229000 }, { "epoch": 0.61, "learning_rate": 4.42625e-05, "loss": 3.4874, "step": 229500 }, { "epoch": 0.61, "learning_rate": 4.4250000000000005e-05, "loss": 3.4704, "step": 230000 }, { "epoch": 0.61, "learning_rate": 4.4237500000000005e-05, "loss": 3.4507, "step": 230500 }, { "epoch": 0.61, "learning_rate": 4.4225e-05, "loss": 3.4603, "step": 231000 }, { "epoch": 0.62, "learning_rate": 4.4212500000000004e-05, "loss": 3.4259, "step": 231500 }, { "epoch": 0.62, "learning_rate": 4.4200000000000004e-05, "loss": 3.4063, "step": 232000 }, { "epoch": 0.62, "learning_rate": 4.4187500000000003e-05, "loss": 3.4067, "step": 232500 }, { "epoch": 0.62, "learning_rate": 4.4174999999999996e-05, "loss": 3.3912, "step": 233000 }, { "epoch": 0.62, "learning_rate": 4.41625e-05, "loss": 3.3861, "step": 233500 }, { "epoch": 0.62, "learning_rate": 4.415e-05, "loss": 3.3665, "step": 234000 }, { "epoch": 0.62, "learning_rate": 4.41375e-05, "loss": 3.3573, "step": 234500 }, { "epoch": 0.63, "learning_rate": 4.4125e-05, "loss": 3.3546, "step": 235000 }, { "epoch": 0.63, "learning_rate": 4.41125e-05, "loss": 3.3359, "step": 235500 }, { "epoch": 0.63, "learning_rate": 4.41e-05, "loss": 3.3183, "step": 236000 }, { "epoch": 0.63, "learning_rate": 4.40875e-05, "loss": 3.3209, "step": 236500 }, { "epoch": 0.63, "learning_rate": 4.4075e-05, "loss": 3.3176, "step": 237000 }, { "epoch": 0.63, "learning_rate": 4.40625e-05, "loss": 3.2999, "step": 237500 }, { "epoch": 0.63, "learning_rate": 4.405e-05, "loss": 3.2913, "step": 238000 }, { "epoch": 0.63, "learning_rate": 4.4037500000000007e-05, "loss": 3.2913, "step": 238500 }, { "epoch": 0.64, "learning_rate": 4.4025e-05, "loss": 3.301, "step": 239000 }, { "epoch": 0.64, "learning_rate": 4.40125e-05, "loss": 3.276, "step": 239500 }, { "epoch": 0.64, "learning_rate": 4.4000000000000006e-05, "loss": 3.2546, "step": 240000 }, { "epoch": 0.64, "learning_rate": 4.3987500000000005e-05, "loss": 3.2657, "step": 240500 }, { "epoch": 0.64, "learning_rate": 4.3975e-05, "loss": 3.2539, "step": 241000 }, { "epoch": 0.64, "learning_rate": 4.39625e-05, "loss": 3.2537, "step": 241500 }, { "epoch": 0.64, "learning_rate": 4.3950000000000004e-05, "loss": 3.2329, "step": 242000 }, { "epoch": 0.65, "learning_rate": 4.3937500000000004e-05, "loss": 3.222, "step": 242500 }, { "epoch": 0.65, "learning_rate": 4.3925e-05, "loss": 3.2174, "step": 243000 }, { "epoch": 0.65, "learning_rate": 4.3912500000000004e-05, "loss": 3.2071, "step": 243500 }, { "epoch": 0.65, "learning_rate": 4.39e-05, "loss": 3.2006, "step": 244000 }, { "epoch": 0.65, "learning_rate": 4.38875e-05, "loss": 3.2143, "step": 244500 }, { "epoch": 0.65, "learning_rate": 4.3875e-05, "loss": 3.2037, "step": 245000 }, { "epoch": 0.65, "learning_rate": 4.38625e-05, "loss": 3.1687, "step": 245500 }, { "epoch": 0.65, "learning_rate": 4.385e-05, "loss": 3.1657, "step": 246000 }, { "epoch": 0.66, "learning_rate": 4.38375e-05, "loss": 3.1516, "step": 246500 }, { "epoch": 0.66, "learning_rate": 4.3825e-05, "loss": 3.1635, "step": 247000 }, { "epoch": 0.66, "learning_rate": 4.38125e-05, "loss": 3.1436, "step": 247500 }, { "epoch": 0.66, "learning_rate": 4.38e-05, "loss": 3.1416, "step": 248000 }, { "epoch": 0.66, "learning_rate": 4.37875e-05, "loss": 3.1304, "step": 248500 }, { "epoch": 0.66, "learning_rate": 4.3775e-05, "loss": 3.1449, "step": 249000 }, { "epoch": 0.66, "learning_rate": 4.37625e-05, "loss": 3.1358, "step": 249500 }, { "epoch": 0.67, "learning_rate": 4.375e-05, "loss": 3.1025, "step": 250000 }, { "epoch": 0.67, "learning_rate": 4.3737500000000006e-05, "loss": 3.0953, "step": 250500 }, { "epoch": 0.67, "learning_rate": 4.3725000000000006e-05, "loss": 3.1176, "step": 251000 }, { "epoch": 0.67, "learning_rate": 4.37125e-05, "loss": 3.1093, "step": 251500 }, { "epoch": 0.67, "learning_rate": 4.3700000000000005e-05, "loss": 3.0782, "step": 252000 }, { "epoch": 0.67, "learning_rate": 4.3687500000000005e-05, "loss": 3.0807, "step": 252500 }, { "epoch": 0.67, "learning_rate": 4.3675000000000005e-05, "loss": 3.0986, "step": 253000 }, { "epoch": 0.67, "learning_rate": 4.36625e-05, "loss": 3.0871, "step": 253500 }, { "epoch": 0.68, "learning_rate": 4.3650000000000004e-05, "loss": 3.0836, "step": 254000 }, { "epoch": 0.68, "learning_rate": 4.3637500000000004e-05, "loss": 3.0776, "step": 254500 }, { "epoch": 0.68, "learning_rate": 4.3625e-05, "loss": 3.0501, "step": 255000 }, { "epoch": 0.68, "learning_rate": 4.36125e-05, "loss": 3.0478, "step": 255500 }, { "epoch": 0.68, "learning_rate": 4.36e-05, "loss": 3.063, "step": 256000 }, { "epoch": 0.68, "learning_rate": 4.35875e-05, "loss": 3.0485, "step": 256500 }, { "epoch": 0.68, "learning_rate": 4.3575e-05, "loss": 3.039, "step": 257000 }, { "epoch": 0.69, "learning_rate": 4.35625e-05, "loss": 3.0399, "step": 257500 }, { "epoch": 0.69, "learning_rate": 4.355e-05, "loss": 3.0462, "step": 258000 }, { "epoch": 0.69, "learning_rate": 4.35375e-05, "loss": 3.016, "step": 258500 }, { "epoch": 0.69, "learning_rate": 4.352500000000001e-05, "loss": 3.0264, "step": 259000 }, { "epoch": 0.69, "learning_rate": 4.35125e-05, "loss": 3.0172, "step": 259500 }, { "epoch": 0.69, "learning_rate": 4.35e-05, "loss": 3.0174, "step": 260000 }, { "epoch": 0.69, "learning_rate": 4.34875e-05, "loss": 3.0193, "step": 260500 }, { "epoch": 0.69, "learning_rate": 4.3475000000000006e-05, "loss": 3.0045, "step": 261000 }, { "epoch": 0.7, "learning_rate": 4.34625e-05, "loss": 2.992, "step": 261500 }, { "epoch": 0.7, "learning_rate": 4.345e-05, "loss": 2.9872, "step": 262000 }, { "epoch": 0.7, "learning_rate": 4.3437500000000006e-05, "loss": 2.9796, "step": 262500 }, { "epoch": 0.7, "learning_rate": 4.3425000000000005e-05, "loss": 2.983, "step": 263000 }, { "epoch": 0.7, "learning_rate": 4.34125e-05, "loss": 2.9638, "step": 263500 }, { "epoch": 0.7, "learning_rate": 4.3400000000000005e-05, "loss": 2.9642, "step": 264000 }, { "epoch": 0.7, "learning_rate": 4.3387500000000004e-05, "loss": 2.9713, "step": 264500 }, { "epoch": 0.71, "learning_rate": 4.3375000000000004e-05, "loss": 2.9688, "step": 265000 }, { "epoch": 0.71, "learning_rate": 4.3362500000000004e-05, "loss": 2.9707, "step": 265500 }, { "epoch": 0.71, "learning_rate": 4.335e-05, "loss": 2.9515, "step": 266000 }, { "epoch": 0.71, "learning_rate": 4.33375e-05, "loss": 2.9666, "step": 266500 }, { "epoch": 0.71, "learning_rate": 4.3325e-05, "loss": 2.9451, "step": 267000 }, { "epoch": 0.71, "learning_rate": 4.33125e-05, "loss": 2.9395, "step": 267500 }, { "epoch": 0.71, "learning_rate": 4.33e-05, "loss": 2.9481, "step": 268000 }, { "epoch": 0.71, "learning_rate": 4.32875e-05, "loss": 2.9353, "step": 268500 }, { "epoch": 0.72, "learning_rate": 4.3275e-05, "loss": 2.9408, "step": 269000 }, { "epoch": 0.72, "learning_rate": 4.32625e-05, "loss": 2.9235, "step": 269500 }, { "epoch": 0.72, "learning_rate": 4.325e-05, "loss": 2.9386, "step": 270000 }, { "epoch": 0.72, "learning_rate": 4.32375e-05, "loss": 2.9225, "step": 270500 }, { "epoch": 0.72, "learning_rate": 4.322500000000001e-05, "loss": 2.9167, "step": 271000 }, { "epoch": 0.72, "learning_rate": 4.32125e-05, "loss": 2.9175, "step": 271500 }, { "epoch": 0.72, "learning_rate": 4.32e-05, "loss": 2.9094, "step": 272000 }, { "epoch": 0.73, "learning_rate": 4.3187500000000006e-05, "loss": 2.8969, "step": 272500 }, { "epoch": 0.73, "learning_rate": 4.3175000000000006e-05, "loss": 2.895, "step": 273000 }, { "epoch": 0.73, "learning_rate": 4.31625e-05, "loss": 2.894, "step": 273500 }, { "epoch": 0.73, "learning_rate": 4.315e-05, "loss": 2.9094, "step": 274000 }, { "epoch": 0.73, "learning_rate": 4.3137500000000005e-05, "loss": 2.8607, "step": 274500 }, { "epoch": 0.73, "learning_rate": 4.3125000000000005e-05, "loss": 2.8984, "step": 275000 }, { "epoch": 0.73, "learning_rate": 4.31125e-05, "loss": 2.874, "step": 275500 }, { "epoch": 0.73, "learning_rate": 4.3100000000000004e-05, "loss": 2.8666, "step": 276000 }, { "epoch": 0.74, "learning_rate": 4.3087500000000004e-05, "loss": 2.875, "step": 276500 }, { "epoch": 0.74, "learning_rate": 4.3075000000000003e-05, "loss": 2.8697, "step": 277000 }, { "epoch": 0.74, "learning_rate": 4.30625e-05, "loss": 2.8631, "step": 277500 }, { "epoch": 0.74, "learning_rate": 4.305e-05, "loss": 2.8653, "step": 278000 }, { "epoch": 0.74, "learning_rate": 4.30375e-05, "loss": 2.8503, "step": 278500 }, { "epoch": 0.74, "learning_rate": 4.3025e-05, "loss": 2.8411, "step": 279000 }, { "epoch": 0.74, "learning_rate": 4.30125e-05, "loss": 2.8503, "step": 279500 }, { "epoch": 0.75, "learning_rate": 4.3e-05, "loss": 2.8376, "step": 280000 }, { "epoch": 0.75, "learning_rate": 4.29875e-05, "loss": 2.8559, "step": 280500 }, { "epoch": 0.75, "learning_rate": 4.2975e-05, "loss": 2.8281, "step": 281000 }, { "epoch": 0.75, "learning_rate": 4.29625e-05, "loss": 2.8581, "step": 281500 }, { "epoch": 0.75, "learning_rate": 4.295e-05, "loss": 2.8443, "step": 282000 }, { "epoch": 0.75, "learning_rate": 4.29375e-05, "loss": 2.8445, "step": 282500 }, { "epoch": 0.75, "learning_rate": 4.2925000000000007e-05, "loss": 2.8334, "step": 283000 }, { "epoch": 0.75, "learning_rate": 4.29125e-05, "loss": 2.8145, "step": 283500 }, { "epoch": 0.76, "learning_rate": 4.29e-05, "loss": 2.8045, "step": 284000 }, { "epoch": 0.76, "learning_rate": 4.2887500000000006e-05, "loss": 2.8287, "step": 284500 }, { "epoch": 0.76, "learning_rate": 4.2875000000000005e-05, "loss": 2.825, "step": 285000 }, { "epoch": 0.76, "learning_rate": 4.28625e-05, "loss": 2.7911, "step": 285500 }, { "epoch": 0.76, "learning_rate": 4.285e-05, "loss": 2.8179, "step": 286000 }, { "epoch": 0.76, "learning_rate": 4.2837500000000004e-05, "loss": 2.8169, "step": 286500 }, { "epoch": 0.76, "learning_rate": 4.2825000000000004e-05, "loss": 2.804, "step": 287000 }, { "epoch": 0.77, "learning_rate": 4.28125e-05, "loss": 2.8004, "step": 287500 }, { "epoch": 0.77, "learning_rate": 4.2800000000000004e-05, "loss": 2.7825, "step": 288000 }, { "epoch": 0.77, "learning_rate": 4.27875e-05, "loss": 2.8022, "step": 288500 }, { "epoch": 0.77, "learning_rate": 4.2775e-05, "loss": 2.786, "step": 289000 }, { "epoch": 0.77, "learning_rate": 4.27625e-05, "loss": 2.785, "step": 289500 }, { "epoch": 0.77, "learning_rate": 4.275e-05, "loss": 2.7969, "step": 290000 }, { "epoch": 0.77, "learning_rate": 4.27375e-05, "loss": 2.778, "step": 290500 }, { "epoch": 0.77, "learning_rate": 4.2725e-05, "loss": 2.7817, "step": 291000 }, { "epoch": 0.78, "learning_rate": 4.27125e-05, "loss": 2.7574, "step": 291500 }, { "epoch": 0.78, "learning_rate": 4.27e-05, "loss": 2.7665, "step": 292000 }, { "epoch": 0.78, "learning_rate": 4.26875e-05, "loss": 2.7662, "step": 292500 }, { "epoch": 0.78, "learning_rate": 4.2675e-05, "loss": 2.7485, "step": 293000 }, { "epoch": 0.78, "learning_rate": 4.26625e-05, "loss": 2.7519, "step": 293500 }, { "epoch": 0.78, "learning_rate": 4.265e-05, "loss": 2.7609, "step": 294000 }, { "epoch": 0.78, "learning_rate": 4.26375e-05, "loss": 2.7625, "step": 294500 }, { "epoch": 0.79, "learning_rate": 4.2625000000000006e-05, "loss": 2.7462, "step": 295000 }, { "epoch": 0.79, "learning_rate": 4.26125e-05, "loss": 2.7601, "step": 295500 }, { "epoch": 0.79, "learning_rate": 4.26e-05, "loss": 2.7493, "step": 296000 }, { "epoch": 0.79, "learning_rate": 4.2587500000000005e-05, "loss": 2.7811, "step": 296500 }, { "epoch": 0.79, "learning_rate": 4.2575000000000005e-05, "loss": 2.7433, "step": 297000 }, { "epoch": 0.79, "learning_rate": 4.25625e-05, "loss": 2.7384, "step": 297500 }, { "epoch": 0.79, "learning_rate": 4.2550000000000004e-05, "loss": 2.7464, "step": 298000 }, { "epoch": 0.79, "learning_rate": 4.2537500000000004e-05, "loss": 2.7285, "step": 298500 }, { "epoch": 0.8, "learning_rate": 4.2525000000000004e-05, "loss": 2.7112, "step": 299000 }, { "epoch": 0.8, "learning_rate": 4.2512499999999997e-05, "loss": 2.754, "step": 299500 }, { "epoch": 0.8, "learning_rate": 4.25e-05, "loss": 2.7363, "step": 300000 }, { "epoch": 0.8, "learning_rate": 4.24875e-05, "loss": 2.7124, "step": 300500 }, { "epoch": 0.8, "learning_rate": 4.2475e-05, "loss": 2.7425, "step": 301000 }, { "epoch": 0.8, "learning_rate": 4.24625e-05, "loss": 2.7313, "step": 301500 }, { "epoch": 0.8, "learning_rate": 4.245e-05, "loss": 2.7198, "step": 302000 }, { "epoch": 0.81, "learning_rate": 4.24375e-05, "loss": 2.7103, "step": 302500 }, { "epoch": 0.81, "learning_rate": 4.2425e-05, "loss": 2.7109, "step": 303000 }, { "epoch": 0.81, "learning_rate": 4.24125e-05, "loss": 2.7108, "step": 303500 }, { "epoch": 0.81, "learning_rate": 4.24e-05, "loss": 2.7079, "step": 304000 }, { "epoch": 0.81, "learning_rate": 4.23875e-05, "loss": 2.7113, "step": 304500 }, { "epoch": 0.81, "learning_rate": 4.237500000000001e-05, "loss": 2.7172, "step": 305000 }, { "epoch": 0.81, "learning_rate": 4.23625e-05, "loss": 2.7079, "step": 305500 }, { "epoch": 0.81, "learning_rate": 4.235e-05, "loss": 2.6975, "step": 306000 }, { "epoch": 0.82, "learning_rate": 4.23375e-05, "loss": 2.7066, "step": 306500 }, { "epoch": 0.82, "learning_rate": 4.2325000000000006e-05, "loss": 2.6851, "step": 307000 }, { "epoch": 0.82, "learning_rate": 4.23125e-05, "loss": 2.6908, "step": 307500 }, { "epoch": 0.82, "learning_rate": 4.23e-05, "loss": 2.7071, "step": 308000 }, { "epoch": 0.82, "learning_rate": 4.2287500000000005e-05, "loss": 2.688, "step": 308500 }, { "epoch": 0.82, "learning_rate": 4.2275000000000004e-05, "loss": 2.6749, "step": 309000 }, { "epoch": 0.82, "learning_rate": 4.22625e-05, "loss": 2.7008, "step": 309500 }, { "epoch": 0.83, "learning_rate": 4.2250000000000004e-05, "loss": 2.6924, "step": 310000 }, { "epoch": 0.83, "learning_rate": 4.2237500000000003e-05, "loss": 2.6662, "step": 310500 }, { "epoch": 0.83, "learning_rate": 4.2225e-05, "loss": 2.6782, "step": 311000 }, { "epoch": 0.83, "learning_rate": 4.2212499999999996e-05, "loss": 2.6759, "step": 311500 }, { "epoch": 0.83, "learning_rate": 4.22e-05, "loss": 2.6685, "step": 312000 }, { "epoch": 0.83, "learning_rate": 4.21875e-05, "loss": 2.6608, "step": 312500 }, { "epoch": 0.83, "learning_rate": 4.2175e-05, "loss": 2.6701, "step": 313000 }, { "epoch": 0.83, "learning_rate": 4.21625e-05, "loss": 2.6692, "step": 313500 }, { "epoch": 0.84, "learning_rate": 4.215e-05, "loss": 2.6651, "step": 314000 }, { "epoch": 0.84, "learning_rate": 4.21375e-05, "loss": 2.6538, "step": 314500 }, { "epoch": 0.84, "learning_rate": 4.2125e-05, "loss": 2.6418, "step": 315000 }, { "epoch": 0.84, "learning_rate": 4.211250000000001e-05, "loss": 2.6721, "step": 315500 }, { "epoch": 0.84, "learning_rate": 4.21e-05, "loss": 2.6647, "step": 316000 }, { "epoch": 0.84, "learning_rate": 4.20875e-05, "loss": 2.6654, "step": 316500 }, { "epoch": 0.84, "learning_rate": 4.2075000000000006e-05, "loss": 2.6771, "step": 317000 }, { "epoch": 0.85, "learning_rate": 4.2062500000000006e-05, "loss": 2.6411, "step": 317500 }, { "epoch": 0.85, "learning_rate": 4.205e-05, "loss": 2.6448, "step": 318000 }, { "epoch": 0.85, "learning_rate": 4.20375e-05, "loss": 2.6575, "step": 318500 }, { "epoch": 0.85, "learning_rate": 4.2025000000000005e-05, "loss": 2.6497, "step": 319000 }, { "epoch": 0.85, "learning_rate": 4.2012500000000005e-05, "loss": 2.6546, "step": 319500 }, { "epoch": 0.85, "learning_rate": 4.2e-05, "loss": 2.635, "step": 320000 }, { "epoch": 0.85, "learning_rate": 4.1987500000000004e-05, "loss": 2.6405, "step": 320500 }, { "epoch": 0.85, "learning_rate": 4.1975000000000004e-05, "loss": 2.6307, "step": 321000 }, { "epoch": 0.86, "learning_rate": 4.1962500000000004e-05, "loss": 2.6273, "step": 321500 }, { "epoch": 0.86, "learning_rate": 4.195e-05, "loss": 2.6334, "step": 322000 }, { "epoch": 0.86, "learning_rate": 4.19375e-05, "loss": 2.6354, "step": 322500 }, { "epoch": 0.86, "learning_rate": 4.1925e-05, "loss": 2.6518, "step": 323000 }, { "epoch": 0.86, "learning_rate": 4.19125e-05, "loss": 2.6418, "step": 323500 }, { "epoch": 0.86, "learning_rate": 4.19e-05, "loss": 2.6429, "step": 324000 }, { "epoch": 0.86, "learning_rate": 4.18875e-05, "loss": 2.6221, "step": 324500 }, { "epoch": 0.87, "learning_rate": 4.1875e-05, "loss": 2.6335, "step": 325000 }, { "epoch": 0.87, "learning_rate": 4.18625e-05, "loss": 2.6268, "step": 325500 }, { "epoch": 0.87, "learning_rate": 4.185e-05, "loss": 2.6208, "step": 326000 }, { "epoch": 0.87, "learning_rate": 4.18375e-05, "loss": 2.6226, "step": 326500 }, { "epoch": 0.87, "learning_rate": 4.1825e-05, "loss": 2.616, "step": 327000 }, { "epoch": 0.87, "learning_rate": 4.181250000000001e-05, "loss": 2.6168, "step": 327500 }, { "epoch": 0.87, "learning_rate": 4.18e-05, "loss": 2.6099, "step": 328000 }, { "epoch": 0.87, "learning_rate": 4.17875e-05, "loss": 2.6059, "step": 328500 }, { "epoch": 0.88, "learning_rate": 4.1775000000000006e-05, "loss": 2.6031, "step": 329000 }, { "epoch": 0.88, "learning_rate": 4.1762500000000005e-05, "loss": 2.6082, "step": 329500 }, { "epoch": 0.88, "learning_rate": 4.175e-05, "loss": 2.6354, "step": 330000 }, { "epoch": 0.88, "learning_rate": 4.1737500000000005e-05, "loss": 2.6227, "step": 330500 }, { "epoch": 0.88, "learning_rate": 4.1725000000000005e-05, "loss": 2.6006, "step": 331000 }, { "epoch": 0.88, "learning_rate": 4.1712500000000004e-05, "loss": 2.5903, "step": 331500 }, { "epoch": 0.88, "learning_rate": 4.17e-05, "loss": 2.6214, "step": 332000 }, { "epoch": 0.89, "learning_rate": 4.1687500000000004e-05, "loss": 2.6046, "step": 332500 }, { "epoch": 0.89, "learning_rate": 4.1675e-05, "loss": 2.616, "step": 333000 }, { "epoch": 0.89, "learning_rate": 4.16625e-05, "loss": 2.6017, "step": 333500 }, { "epoch": 0.89, "learning_rate": 4.165e-05, "loss": 2.6047, "step": 334000 }, { "epoch": 0.89, "learning_rate": 4.16375e-05, "loss": 2.5918, "step": 334500 }, { "epoch": 0.89, "learning_rate": 4.1625e-05, "loss": 2.5877, "step": 335000 }, { "epoch": 0.89, "learning_rate": 4.16125e-05, "loss": 2.5792, "step": 335500 }, { "epoch": 0.89, "learning_rate": 4.16e-05, "loss": 2.5839, "step": 336000 }, { "epoch": 0.9, "learning_rate": 4.15875e-05, "loss": 2.592, "step": 336500 }, { "epoch": 0.9, "learning_rate": 4.1575e-05, "loss": 2.5888, "step": 337000 }, { "epoch": 0.9, "learning_rate": 4.156250000000001e-05, "loss": 2.5922, "step": 337500 }, { "epoch": 0.9, "learning_rate": 4.155e-05, "loss": 2.5633, "step": 338000 }, { "epoch": 0.9, "learning_rate": 4.15375e-05, "loss": 2.5687, "step": 338500 }, { "epoch": 0.9, "learning_rate": 4.1525e-05, "loss": 2.5722, "step": 339000 }, { "epoch": 0.9, "learning_rate": 4.1512500000000006e-05, "loss": 2.5619, "step": 339500 }, { "epoch": 0.91, "learning_rate": 4.15e-05, "loss": 2.5692, "step": 340000 }, { "epoch": 0.91, "learning_rate": 4.14875e-05, "loss": 2.576, "step": 340500 }, { "epoch": 0.91, "learning_rate": 4.1475000000000005e-05, "loss": 2.5658, "step": 341000 }, { "epoch": 0.91, "learning_rate": 4.1462500000000005e-05, "loss": 2.5544, "step": 341500 }, { "epoch": 0.91, "learning_rate": 4.145e-05, "loss": 2.5614, "step": 342000 }, { "epoch": 0.91, "learning_rate": 4.1437500000000004e-05, "loss": 2.5658, "step": 342500 }, { "epoch": 0.91, "learning_rate": 4.1425000000000004e-05, "loss": 2.5822, "step": 343000 }, { "epoch": 0.91, "learning_rate": 4.1412500000000004e-05, "loss": 2.5711, "step": 343500 }, { "epoch": 0.92, "learning_rate": 4.14e-05, "loss": 2.5636, "step": 344000 }, { "epoch": 0.92, "learning_rate": 4.13875e-05, "loss": 2.5537, "step": 344500 }, { "epoch": 0.92, "learning_rate": 4.1375e-05, "loss": 2.5551, "step": 345000 }, { "epoch": 0.92, "learning_rate": 4.13625e-05, "loss": 2.552, "step": 345500 }, { "epoch": 0.92, "learning_rate": 4.135e-05, "loss": 2.5492, "step": 346000 }, { "epoch": 0.92, "learning_rate": 4.13375e-05, "loss": 2.5462, "step": 346500 }, { "epoch": 0.92, "learning_rate": 4.1325e-05, "loss": 2.5584, "step": 347000 }, { "epoch": 0.93, "learning_rate": 4.13125e-05, "loss": 2.5375, "step": 347500 }, { "epoch": 0.93, "learning_rate": 4.13e-05, "loss": 2.5352, "step": 348000 }, { "epoch": 0.93, "learning_rate": 4.12875e-05, "loss": 2.5451, "step": 348500 }, { "epoch": 0.93, "learning_rate": 4.1275e-05, "loss": 2.5498, "step": 349000 }, { "epoch": 0.93, "learning_rate": 4.126250000000001e-05, "loss": 2.5661, "step": 349500 }, { "epoch": 0.93, "learning_rate": 4.125e-05, "loss": 2.5478, "step": 350000 }, { "epoch": 0.93, "learning_rate": 4.12375e-05, "loss": 2.5319, "step": 350500 }, { "epoch": 0.93, "learning_rate": 4.1225e-05, "loss": 2.5126, "step": 351000 }, { "epoch": 0.94, "learning_rate": 4.1212500000000006e-05, "loss": 2.5356, "step": 351500 }, { "epoch": 0.94, "learning_rate": 4.12e-05, "loss": 2.5311, "step": 352000 }, { "epoch": 0.94, "learning_rate": 4.11875e-05, "loss": 2.5173, "step": 352500 }, { "epoch": 0.94, "learning_rate": 4.1175000000000005e-05, "loss": 2.5397, "step": 353000 }, { "epoch": 0.94, "learning_rate": 4.1162500000000004e-05, "loss": 2.534, "step": 353500 }, { "epoch": 0.94, "learning_rate": 4.115e-05, "loss": 2.5334, "step": 354000 }, { "epoch": 0.94, "learning_rate": 4.1137500000000004e-05, "loss": 2.5418, "step": 354500 }, { "epoch": 0.95, "learning_rate": 4.1125000000000004e-05, "loss": 2.5035, "step": 355000 }, { "epoch": 0.95, "learning_rate": 4.11125e-05, "loss": 2.5204, "step": 355500 }, { "epoch": 0.95, "learning_rate": 4.11e-05, "loss": 2.5291, "step": 356000 }, { "epoch": 0.95, "learning_rate": 4.10875e-05, "loss": 2.5219, "step": 356500 }, { "epoch": 0.95, "learning_rate": 4.1075e-05, "loss": 2.5343, "step": 357000 }, { "epoch": 0.95, "learning_rate": 4.10625e-05, "loss": 2.5204, "step": 357500 }, { "epoch": 0.95, "learning_rate": 4.105e-05, "loss": 2.5367, "step": 358000 }, { "epoch": 0.95, "learning_rate": 4.10375e-05, "loss": 2.5109, "step": 358500 }, { "epoch": 0.96, "learning_rate": 4.1025e-05, "loss": 2.503, "step": 359000 }, { "epoch": 0.96, "learning_rate": 4.10125e-05, "loss": 2.5091, "step": 359500 }, { "epoch": 0.96, "learning_rate": 4.1e-05, "loss": 2.5077, "step": 360000 }, { "epoch": 0.96, "learning_rate": 4.09875e-05, "loss": 2.5048, "step": 360500 }, { "epoch": 0.96, "learning_rate": 4.0975e-05, "loss": 2.5094, "step": 361000 }, { "epoch": 0.96, "learning_rate": 4.0962500000000006e-05, "loss": 2.5081, "step": 361500 }, { "epoch": 0.96, "learning_rate": 4.095e-05, "loss": 2.5179, "step": 362000 }, { "epoch": 0.97, "learning_rate": 4.09375e-05, "loss": 2.5085, "step": 362500 }, { "epoch": 0.97, "learning_rate": 4.0925000000000005e-05, "loss": 2.4981, "step": 363000 }, { "epoch": 0.97, "learning_rate": 4.0912500000000005e-05, "loss": 2.5063, "step": 363500 }, { "epoch": 0.97, "learning_rate": 4.09e-05, "loss": 2.4912, "step": 364000 }, { "epoch": 0.97, "learning_rate": 4.08875e-05, "loss": 2.504, "step": 364500 }, { "epoch": 0.97, "learning_rate": 4.0875000000000004e-05, "loss": 2.4879, "step": 365000 }, { "epoch": 0.97, "learning_rate": 4.0862500000000004e-05, "loss": 2.4963, "step": 365500 }, { "epoch": 0.97, "learning_rate": 4.085e-05, "loss": 2.4805, "step": 366000 }, { "epoch": 0.98, "learning_rate": 4.08375e-05, "loss": 2.4837, "step": 366500 }, { "epoch": 0.98, "learning_rate": 4.0825e-05, "loss": 2.47, "step": 367000 }, { "epoch": 0.98, "learning_rate": 4.08125e-05, "loss": 2.4897, "step": 367500 }, { "epoch": 0.98, "learning_rate": 4.08e-05, "loss": 2.5125, "step": 368000 }, { "epoch": 0.98, "learning_rate": 4.07875e-05, "loss": 2.4935, "step": 368500 }, { "epoch": 0.98, "learning_rate": 4.0775e-05, "loss": 2.5015, "step": 369000 }, { "epoch": 0.98, "learning_rate": 4.07625e-05, "loss": 2.4978, "step": 369500 }, { "epoch": 0.99, "learning_rate": 4.075e-05, "loss": 2.473, "step": 370000 }, { "epoch": 0.99, "learning_rate": 4.07375e-05, "loss": 2.4917, "step": 370500 }, { "epoch": 0.99, "learning_rate": 4.0725e-05, "loss": 2.4829, "step": 371000 }, { "epoch": 0.99, "learning_rate": 4.07125e-05, "loss": 2.4811, "step": 371500 }, { "epoch": 0.99, "learning_rate": 4.07e-05, "loss": 2.4719, "step": 372000 }, { "epoch": 0.99, "learning_rate": 4.06875e-05, "loss": 2.4773, "step": 372500 }, { "epoch": 0.99, "learning_rate": 4.0675e-05, "loss": 2.473, "step": 373000 }, { "epoch": 0.99, "learning_rate": 4.0662500000000006e-05, "loss": 2.4769, "step": 373500 }, { "epoch": 1.0, "learning_rate": 4.065e-05, "loss": 2.4836, "step": 374000 }, { "epoch": 1.0, "learning_rate": 4.06375e-05, "loss": 2.4698, "step": 374500 }, { "epoch": 1.0, "learning_rate": 4.0625000000000005e-05, "loss": 2.4745, "step": 375000 }, { "epoch": 1.0, "learning_rate": 4.0612500000000005e-05, "loss": 2.4671, "step": 375500 }, { "epoch": 1.0, "learning_rate": 4.0600000000000004e-05, "loss": 2.459, "step": 376000 }, { "epoch": 1.0, "learning_rate": 4.05875e-05, "loss": 2.4796, "step": 376500 }, { "epoch": 1.0, "learning_rate": 4.0575000000000004e-05, "loss": 2.4632, "step": 377000 }, { "epoch": 1.01, "learning_rate": 4.0562500000000003e-05, "loss": 2.4795, "step": 377500 }, { "epoch": 1.01, "learning_rate": 4.055e-05, "loss": 2.4721, "step": 378000 }, { "epoch": 1.01, "learning_rate": 4.05375e-05, "loss": 2.4688, "step": 378500 }, { "epoch": 1.01, "learning_rate": 4.0525e-05, "loss": 2.4687, "step": 379000 }, { "epoch": 1.01, "learning_rate": 4.05125e-05, "loss": 2.4731, "step": 379500 }, { "epoch": 1.01, "learning_rate": 4.05e-05, "loss": 2.4587, "step": 380000 }, { "epoch": 1.01, "learning_rate": 4.04875e-05, "loss": 2.4563, "step": 380500 }, { "epoch": 1.01, "learning_rate": 4.0475e-05, "loss": 2.4452, "step": 381000 }, { "epoch": 1.02, "learning_rate": 4.04625e-05, "loss": 2.4485, "step": 381500 }, { "epoch": 1.02, "learning_rate": 4.045000000000001e-05, "loss": 2.4391, "step": 382000 }, { "epoch": 1.02, "learning_rate": 4.04375e-05, "loss": 2.4454, "step": 382500 }, { "epoch": 1.02, "learning_rate": 4.0425e-05, "loss": 2.4424, "step": 383000 }, { "epoch": 1.02, "learning_rate": 4.0412500000000007e-05, "loss": 2.4457, "step": 383500 }, { "epoch": 1.02, "learning_rate": 4.0400000000000006e-05, "loss": 2.4538, "step": 384000 }, { "epoch": 1.02, "learning_rate": 4.03875e-05, "loss": 2.4452, "step": 384500 }, { "epoch": 1.02, "learning_rate": 4.0375e-05, "loss": 2.4463, "step": 385000 }, { "epoch": 1.03, "learning_rate": 4.0362500000000005e-05, "loss": 2.4397, "step": 385500 }, { "epoch": 1.03, "learning_rate": 4.0350000000000005e-05, "loss": 2.4534, "step": 386000 }, { "epoch": 1.03, "learning_rate": 4.03375e-05, "loss": 2.4505, "step": 386500 }, { "epoch": 1.03, "learning_rate": 4.0325000000000004e-05, "loss": 2.4465, "step": 387000 }, { "epoch": 1.03, "learning_rate": 4.0312500000000004e-05, "loss": 2.454, "step": 387500 }, { "epoch": 1.03, "learning_rate": 4.0300000000000004e-05, "loss": 2.4533, "step": 388000 }, { "epoch": 1.03, "learning_rate": 4.0287500000000003e-05, "loss": 2.4241, "step": 388500 }, { "epoch": 1.04, "learning_rate": 4.0275e-05, "loss": 2.446, "step": 389000 }, { "epoch": 1.04, "learning_rate": 4.02625e-05, "loss": 2.4397, "step": 389500 }, { "epoch": 1.04, "learning_rate": 4.025e-05, "loss": 2.4267, "step": 390000 }, { "epoch": 1.04, "learning_rate": 4.02375e-05, "loss": 2.432, "step": 390500 }, { "epoch": 1.04, "learning_rate": 4.0225e-05, "loss": 2.4354, "step": 391000 }, { "epoch": 1.04, "learning_rate": 4.02125e-05, "loss": 2.4336, "step": 391500 }, { "epoch": 1.04, "learning_rate": 4.02e-05, "loss": 2.4242, "step": 392000 }, { "epoch": 1.04, "learning_rate": 4.01875e-05, "loss": 2.4432, "step": 392500 }, { "epoch": 1.05, "learning_rate": 4.0175e-05, "loss": 2.4293, "step": 393000 }, { "epoch": 1.05, "learning_rate": 4.01625e-05, "loss": 2.4166, "step": 393500 }, { "epoch": 1.05, "learning_rate": 4.015000000000001e-05, "loss": 2.4341, "step": 394000 }, { "epoch": 1.05, "learning_rate": 4.01375e-05, "loss": 2.4378, "step": 394500 }, { "epoch": 1.05, "learning_rate": 4.0125e-05, "loss": 2.4231, "step": 395000 }, { "epoch": 1.05, "learning_rate": 4.0112500000000006e-05, "loss": 2.432, "step": 395500 }, { "epoch": 1.05, "learning_rate": 4.0100000000000006e-05, "loss": 2.421, "step": 396000 }, { "epoch": 1.06, "learning_rate": 4.00875e-05, "loss": 2.4254, "step": 396500 }, { "epoch": 1.06, "learning_rate": 4.0075e-05, "loss": 2.4004, "step": 397000 }, { "epoch": 1.06, "learning_rate": 4.0062500000000005e-05, "loss": 2.4091, "step": 397500 }, { "epoch": 1.06, "learning_rate": 4.0050000000000004e-05, "loss": 2.4111, "step": 398000 }, { "epoch": 1.06, "learning_rate": 4.00375e-05, "loss": 2.4197, "step": 398500 }, { "epoch": 1.06, "learning_rate": 4.0025000000000004e-05, "loss": 2.4259, "step": 399000 }, { "epoch": 1.06, "learning_rate": 4.0012500000000004e-05, "loss": 2.4171, "step": 399500 }, { "epoch": 1.06, "learning_rate": 4e-05, "loss": 2.4234, "step": 400000 }, { "epoch": 1.07, "learning_rate": 3.99875e-05, "loss": 2.3982, "step": 400500 }, { "epoch": 1.07, "learning_rate": 3.9975e-05, "loss": 2.4105, "step": 401000 }, { "epoch": 1.07, "learning_rate": 3.99625e-05, "loss": 2.3903, "step": 401500 }, { "epoch": 1.07, "learning_rate": 3.995e-05, "loss": 2.4219, "step": 402000 }, { "epoch": 1.07, "learning_rate": 3.99375e-05, "loss": 2.4063, "step": 402500 }, { "epoch": 1.07, "learning_rate": 3.9925e-05, "loss": 2.422, "step": 403000 }, { "epoch": 1.07, "learning_rate": 3.99125e-05, "loss": 2.4119, "step": 403500 }, { "epoch": 1.08, "learning_rate": 3.99e-05, "loss": 2.4075, "step": 404000 }, { "epoch": 1.08, "learning_rate": 3.98875e-05, "loss": 2.412, "step": 404500 }, { "epoch": 1.08, "learning_rate": 3.9875e-05, "loss": 2.4131, "step": 405000 }, { "epoch": 1.08, "learning_rate": 3.98625e-05, "loss": 2.4164, "step": 405500 }, { "epoch": 1.08, "learning_rate": 3.9850000000000006e-05, "loss": 2.4042, "step": 406000 }, { "epoch": 1.08, "learning_rate": 3.98375e-05, "loss": 2.3947, "step": 406500 }, { "epoch": 1.08, "learning_rate": 3.9825e-05, "loss": 2.4073, "step": 407000 }, { "epoch": 1.08, "learning_rate": 3.9812500000000005e-05, "loss": 2.3958, "step": 407500 }, { "epoch": 1.09, "learning_rate": 3.9800000000000005e-05, "loss": 2.3927, "step": 408000 }, { "epoch": 1.09, "learning_rate": 3.97875e-05, "loss": 2.3917, "step": 408500 }, { "epoch": 1.09, "learning_rate": 3.9775e-05, "loss": 2.3917, "step": 409000 }, { "epoch": 1.09, "learning_rate": 3.9762500000000004e-05, "loss": 2.4083, "step": 409500 }, { "epoch": 1.09, "learning_rate": 3.9750000000000004e-05, "loss": 2.4097, "step": 410000 }, { "epoch": 1.09, "learning_rate": 3.97375e-05, "loss": 2.3883, "step": 410500 }, { "epoch": 1.09, "learning_rate": 3.9725e-05, "loss": 2.4005, "step": 411000 }, { "epoch": 1.1, "learning_rate": 3.97125e-05, "loss": 2.3891, "step": 411500 }, { "epoch": 1.1, "learning_rate": 3.97e-05, "loss": 2.3918, "step": 412000 }, { "epoch": 1.1, "learning_rate": 3.96875e-05, "loss": 2.3954, "step": 412500 }, { "epoch": 1.1, "learning_rate": 3.9675e-05, "loss": 2.3926, "step": 413000 }, { "epoch": 1.1, "learning_rate": 3.96625e-05, "loss": 2.3865, "step": 413500 }, { "epoch": 1.1, "learning_rate": 3.965e-05, "loss": 2.3904, "step": 414000 }, { "epoch": 1.1, "learning_rate": 3.96375e-05, "loss": 2.3777, "step": 414500 }, { "epoch": 1.1, "learning_rate": 3.9625e-05, "loss": 2.3675, "step": 415000 }, { "epoch": 1.11, "learning_rate": 3.96125e-05, "loss": 2.3774, "step": 415500 }, { "epoch": 1.11, "learning_rate": 3.960000000000001e-05, "loss": 2.3752, "step": 416000 }, { "epoch": 1.11, "learning_rate": 3.95875e-05, "loss": 2.3914, "step": 416500 }, { "epoch": 1.11, "learning_rate": 3.9575e-05, "loss": 2.3902, "step": 417000 }, { "epoch": 1.11, "learning_rate": 3.95625e-05, "loss": 2.3813, "step": 417500 }, { "epoch": 1.11, "learning_rate": 3.9550000000000006e-05, "loss": 2.3736, "step": 418000 }, { "epoch": 1.11, "learning_rate": 3.95375e-05, "loss": 2.3977, "step": 418500 }, { "epoch": 1.12, "learning_rate": 3.9525e-05, "loss": 2.3747, "step": 419000 }, { "epoch": 1.12, "learning_rate": 3.9512500000000005e-05, "loss": 2.3807, "step": 419500 }, { "epoch": 1.12, "learning_rate": 3.9500000000000005e-05, "loss": 2.3775, "step": 420000 }, { "epoch": 1.12, "learning_rate": 3.94875e-05, "loss": 2.3651, "step": 420500 }, { "epoch": 1.12, "learning_rate": 3.9475000000000004e-05, "loss": 2.3737, "step": 421000 }, { "epoch": 1.12, "learning_rate": 3.9462500000000004e-05, "loss": 2.3572, "step": 421500 }, { "epoch": 1.12, "learning_rate": 3.9450000000000003e-05, "loss": 2.3932, "step": 422000 }, { "epoch": 1.12, "learning_rate": 3.9437499999999996e-05, "loss": 2.3756, "step": 422500 }, { "epoch": 1.13, "learning_rate": 3.9425e-05, "loss": 2.3666, "step": 423000 }, { "epoch": 1.13, "learning_rate": 3.94125e-05, "loss": 2.3805, "step": 423500 }, { "epoch": 1.13, "learning_rate": 3.94e-05, "loss": 2.3599, "step": 424000 }, { "epoch": 1.13, "learning_rate": 3.93875e-05, "loss": 2.351, "step": 424500 }, { "epoch": 1.13, "learning_rate": 3.9375e-05, "loss": 2.3588, "step": 425000 }, { "epoch": 1.13, "learning_rate": 3.93625e-05, "loss": 2.3555, "step": 425500 }, { "epoch": 1.13, "learning_rate": 3.935e-05, "loss": 2.3864, "step": 426000 }, { "epoch": 1.14, "learning_rate": 3.93375e-05, "loss": 2.364, "step": 426500 }, { "epoch": 1.14, "learning_rate": 3.9325e-05, "loss": 2.3624, "step": 427000 }, { "epoch": 1.14, "learning_rate": 3.93125e-05, "loss": 2.3671, "step": 427500 }, { "epoch": 1.14, "learning_rate": 3.9300000000000007e-05, "loss": 2.3682, "step": 428000 }, { "epoch": 1.14, "learning_rate": 3.92875e-05, "loss": 2.3714, "step": 428500 }, { "epoch": 1.14, "learning_rate": 3.9275e-05, "loss": 2.3663, "step": 429000 }, { "epoch": 1.14, "learning_rate": 3.92625e-05, "loss": 2.3773, "step": 429500 }, { "epoch": 1.14, "learning_rate": 3.9250000000000005e-05, "loss": 2.356, "step": 430000 }, { "epoch": 1.15, "learning_rate": 3.92375e-05, "loss": 2.3484, "step": 430500 }, { "epoch": 1.15, "learning_rate": 3.9225e-05, "loss": 2.3633, "step": 431000 }, { "epoch": 1.15, "learning_rate": 3.9212500000000004e-05, "loss": 2.3472, "step": 431500 }, { "epoch": 1.15, "learning_rate": 3.9200000000000004e-05, "loss": 2.3681, "step": 432000 }, { "epoch": 1.15, "learning_rate": 3.91875e-05, "loss": 2.3504, "step": 432500 }, { "epoch": 1.15, "learning_rate": 3.9175000000000004e-05, "loss": 2.3737, "step": 433000 }, { "epoch": 1.15, "learning_rate": 3.91625e-05, "loss": 2.3634, "step": 433500 }, { "epoch": 1.16, "learning_rate": 3.915e-05, "loss": 2.3504, "step": 434000 }, { "epoch": 1.16, "learning_rate": 3.9137499999999996e-05, "loss": 2.3657, "step": 434500 }, { "epoch": 1.16, "learning_rate": 3.9125e-05, "loss": 2.3336, "step": 435000 }, { "epoch": 1.16, "learning_rate": 3.91125e-05, "loss": 2.3582, "step": 435500 }, { "epoch": 1.16, "learning_rate": 3.91e-05, "loss": 2.3522, "step": 436000 }, { "epoch": 1.16, "learning_rate": 3.90875e-05, "loss": 2.3434, "step": 436500 }, { "epoch": 1.16, "learning_rate": 3.9075e-05, "loss": 2.35, "step": 437000 }, { "epoch": 1.16, "learning_rate": 3.90625e-05, "loss": 2.3426, "step": 437500 }, { "epoch": 1.17, "learning_rate": 3.905e-05, "loss": 2.3389, "step": 438000 }, { "epoch": 1.17, "learning_rate": 3.903750000000001e-05, "loss": 2.37, "step": 438500 }, { "epoch": 1.17, "learning_rate": 3.9025e-05, "loss": 2.3338, "step": 439000 }, { "epoch": 1.17, "learning_rate": 3.90125e-05, "loss": 2.3386, "step": 439500 }, { "epoch": 1.17, "learning_rate": 3.9000000000000006e-05, "loss": 2.3395, "step": 440000 }, { "epoch": 1.17, "learning_rate": 3.8987500000000006e-05, "loss": 2.3545, "step": 440500 }, { "epoch": 1.17, "learning_rate": 3.8975e-05, "loss": 2.3462, "step": 441000 }, { "epoch": 1.18, "learning_rate": 3.8962500000000005e-05, "loss": 2.3458, "step": 441500 }, { "epoch": 1.18, "learning_rate": 3.8950000000000005e-05, "loss": 2.3264, "step": 442000 }, { "epoch": 1.18, "learning_rate": 3.8937500000000005e-05, "loss": 2.3422, "step": 442500 }, { "epoch": 1.18, "learning_rate": 3.8925e-05, "loss": 2.3245, "step": 443000 }, { "epoch": 1.18, "learning_rate": 3.8912500000000004e-05, "loss": 2.3318, "step": 443500 }, { "epoch": 1.18, "learning_rate": 3.8900000000000004e-05, "loss": 2.3324, "step": 444000 }, { "epoch": 1.18, "learning_rate": 3.88875e-05, "loss": 2.3555, "step": 444500 }, { "epoch": 1.18, "learning_rate": 3.8875e-05, "loss": 2.3365, "step": 445000 }, { "epoch": 1.19, "learning_rate": 3.88625e-05, "loss": 2.3243, "step": 445500 }, { "epoch": 1.19, "learning_rate": 3.885e-05, "loss": 2.3212, "step": 446000 }, { "epoch": 1.19, "learning_rate": 3.88375e-05, "loss": 2.335, "step": 446500 }, { "epoch": 1.19, "learning_rate": 3.8825e-05, "loss": 2.3405, "step": 447000 }, { "epoch": 1.19, "learning_rate": 3.88125e-05, "loss": 2.3401, "step": 447500 }, { "epoch": 1.19, "learning_rate": 3.88e-05, "loss": 2.3499, "step": 448000 }, { "epoch": 1.19, "learning_rate": 3.878750000000001e-05, "loss": 2.3136, "step": 448500 }, { "epoch": 1.2, "learning_rate": 3.8775e-05, "loss": 2.3489, "step": 449000 }, { "epoch": 1.2, "learning_rate": 3.87625e-05, "loss": 2.3327, "step": 449500 }, { "epoch": 1.2, "learning_rate": 3.875e-05, "loss": 2.3352, "step": 450000 }, { "epoch": 1.2, "learning_rate": 3.8737500000000006e-05, "loss": 2.3175, "step": 450500 }, { "epoch": 1.2, "learning_rate": 3.8725e-05, "loss": 2.3568, "step": 451000 }, { "epoch": 1.2, "learning_rate": 3.87125e-05, "loss": 2.3011, "step": 451500 }, { "epoch": 1.2, "learning_rate": 3.8700000000000006e-05, "loss": 2.3112, "step": 452000 }, { "epoch": 1.2, "learning_rate": 3.8687500000000005e-05, "loss": 2.3219, "step": 452500 }, { "epoch": 1.21, "learning_rate": 3.8675e-05, "loss": 2.3214, "step": 453000 }, { "epoch": 1.21, "learning_rate": 3.8662500000000005e-05, "loss": 2.2998, "step": 453500 }, { "epoch": 1.21, "learning_rate": 3.8650000000000004e-05, "loss": 2.321, "step": 454000 }, { "epoch": 1.21, "learning_rate": 3.8637500000000004e-05, "loss": 2.3256, "step": 454500 }, { "epoch": 1.21, "learning_rate": 3.8625e-05, "loss": 2.3175, "step": 455000 }, { "epoch": 1.21, "learning_rate": 3.8612500000000003e-05, "loss": 2.3294, "step": 455500 }, { "epoch": 1.21, "learning_rate": 3.86e-05, "loss": 2.312, "step": 456000 }, { "epoch": 1.22, "learning_rate": 3.85875e-05, "loss": 2.3223, "step": 456500 }, { "epoch": 1.22, "learning_rate": 3.8575e-05, "loss": 2.3175, "step": 457000 }, { "epoch": 1.22, "learning_rate": 3.85625e-05, "loss": 2.3335, "step": 457500 }, { "epoch": 1.22, "learning_rate": 3.855e-05, "loss": 2.3055, "step": 458000 }, { "epoch": 1.22, "learning_rate": 3.85375e-05, "loss": 2.3259, "step": 458500 }, { "epoch": 1.22, "learning_rate": 3.8525e-05, "loss": 2.3048, "step": 459000 }, { "epoch": 1.22, "learning_rate": 3.85125e-05, "loss": 2.3152, "step": 459500 }, { "epoch": 1.22, "learning_rate": 3.85e-05, "loss": 2.3041, "step": 460000 }, { "epoch": 1.23, "learning_rate": 3.848750000000001e-05, "loss": 2.3324, "step": 460500 }, { "epoch": 1.23, "learning_rate": 3.8475e-05, "loss": 2.3052, "step": 461000 }, { "epoch": 1.23, "learning_rate": 3.84625e-05, "loss": 2.3277, "step": 461500 }, { "epoch": 1.23, "learning_rate": 3.845e-05, "loss": 2.3254, "step": 462000 }, { "epoch": 1.23, "learning_rate": 3.8437500000000006e-05, "loss": 2.2967, "step": 462500 }, { "epoch": 1.23, "learning_rate": 3.8425e-05, "loss": 2.3023, "step": 463000 }, { "epoch": 1.23, "learning_rate": 3.84125e-05, "loss": 2.3093, "step": 463500 }, { "epoch": 1.24, "learning_rate": 3.8400000000000005e-05, "loss": 2.3062, "step": 464000 }, { "epoch": 1.24, "learning_rate": 3.8387500000000005e-05, "loss": 2.3149, "step": 464500 }, { "epoch": 1.24, "learning_rate": 3.8375e-05, "loss": 2.2942, "step": 465000 }, { "epoch": 1.24, "learning_rate": 3.8362500000000004e-05, "loss": 2.3148, "step": 465500 }, { "epoch": 1.24, "learning_rate": 3.8350000000000004e-05, "loss": 2.3028, "step": 466000 }, { "epoch": 1.24, "learning_rate": 3.8337500000000004e-05, "loss": 2.2973, "step": 466500 }, { "epoch": 1.24, "learning_rate": 3.8324999999999996e-05, "loss": 2.301, "step": 467000 }, { "epoch": 1.24, "learning_rate": 3.83125e-05, "loss": 2.3033, "step": 467500 }, { "epoch": 1.25, "learning_rate": 3.83e-05, "loss": 2.3037, "step": 468000 }, { "epoch": 1.25, "learning_rate": 3.82875e-05, "loss": 2.2902, "step": 468500 }, { "epoch": 1.25, "learning_rate": 3.8275e-05, "loss": 2.3058, "step": 469000 }, { "epoch": 1.25, "learning_rate": 3.82625e-05, "loss": 2.2883, "step": 469500 }, { "epoch": 1.25, "learning_rate": 3.825e-05, "loss": 2.2871, "step": 470000 }, { "epoch": 1.25, "learning_rate": 3.82375e-05, "loss": 2.3029, "step": 470500 }, { "epoch": 1.25, "learning_rate": 3.8225e-05, "loss": 2.3127, "step": 471000 }, { "epoch": 1.26, "learning_rate": 3.82125e-05, "loss": 2.2989, "step": 471500 }, { "epoch": 1.26, "learning_rate": 3.82e-05, "loss": 2.3081, "step": 472000 }, { "epoch": 1.26, "learning_rate": 3.818750000000001e-05, "loss": 2.2961, "step": 472500 }, { "epoch": 1.26, "learning_rate": 3.8175e-05, "loss": 2.3067, "step": 473000 }, { "epoch": 1.26, "learning_rate": 3.81625e-05, "loss": 2.2708, "step": 473500 }, { "epoch": 1.26, "learning_rate": 3.8150000000000006e-05, "loss": 2.2842, "step": 474000 }, { "epoch": 1.26, "learning_rate": 3.8137500000000005e-05, "loss": 2.2955, "step": 474500 }, { "epoch": 1.26, "learning_rate": 3.8125e-05, "loss": 2.2866, "step": 475000 }, { "epoch": 1.27, "learning_rate": 3.81125e-05, "loss": 2.3003, "step": 475500 }, { "epoch": 1.27, "learning_rate": 3.8100000000000005e-05, "loss": 2.2936, "step": 476000 }, { "epoch": 1.27, "learning_rate": 3.8087500000000004e-05, "loss": 2.2989, "step": 476500 }, { "epoch": 1.27, "learning_rate": 3.8075e-05, "loss": 2.277, "step": 477000 }, { "epoch": 1.27, "learning_rate": 3.8062500000000004e-05, "loss": 2.2942, "step": 477500 }, { "epoch": 1.27, "learning_rate": 3.805e-05, "loss": 2.276, "step": 478000 }, { "epoch": 1.27, "learning_rate": 3.80375e-05, "loss": 2.2937, "step": 478500 }, { "epoch": 1.28, "learning_rate": 3.8025e-05, "loss": 2.3004, "step": 479000 }, { "epoch": 1.28, "learning_rate": 3.80125e-05, "loss": 2.2932, "step": 479500 }, { "epoch": 1.28, "learning_rate": 3.8e-05, "loss": 2.2753, "step": 480000 }, { "epoch": 1.28, "learning_rate": 3.79875e-05, "loss": 2.3014, "step": 480500 }, { "epoch": 1.28, "learning_rate": 3.7975e-05, "loss": 2.2737, "step": 481000 }, { "epoch": 1.28, "learning_rate": 3.79625e-05, "loss": 2.2914, "step": 481500 }, { "epoch": 1.28, "learning_rate": 3.795e-05, "loss": 2.291, "step": 482000 }, { "epoch": 1.28, "learning_rate": 3.79375e-05, "loss": 2.2705, "step": 482500 }, { "epoch": 1.29, "learning_rate": 3.7925e-05, "loss": 2.2861, "step": 483000 }, { "epoch": 1.29, "learning_rate": 3.79125e-05, "loss": 2.2665, "step": 483500 }, { "epoch": 1.29, "learning_rate": 3.79e-05, "loss": 2.2691, "step": 484000 }, { "epoch": 1.29, "learning_rate": 3.7887500000000006e-05, "loss": 2.2693, "step": 484500 }, { "epoch": 1.29, "learning_rate": 3.7875e-05, "loss": 2.2661, "step": 485000 }, { "epoch": 1.29, "learning_rate": 3.78625e-05, "loss": 2.2716, "step": 485500 }, { "epoch": 1.29, "learning_rate": 3.7850000000000005e-05, "loss": 2.2804, "step": 486000 }, { "epoch": 1.3, "learning_rate": 3.7837500000000005e-05, "loss": 2.256, "step": 486500 }, { "epoch": 1.3, "learning_rate": 3.7825e-05, "loss": 2.2806, "step": 487000 }, { "epoch": 1.3, "learning_rate": 3.78125e-05, "loss": 2.2655, "step": 487500 }, { "epoch": 1.3, "learning_rate": 3.7800000000000004e-05, "loss": 2.2724, "step": 488000 }, { "epoch": 1.3, "learning_rate": 3.7787500000000004e-05, "loss": 2.2739, "step": 488500 }, { "epoch": 1.3, "learning_rate": 3.7775e-05, "loss": 2.2764, "step": 489000 }, { "epoch": 1.3, "learning_rate": 3.77625e-05, "loss": 2.2538, "step": 489500 }, { "epoch": 1.3, "learning_rate": 3.775e-05, "loss": 2.2756, "step": 490000 }, { "epoch": 1.31, "learning_rate": 3.77375e-05, "loss": 2.2756, "step": 490500 }, { "epoch": 1.31, "learning_rate": 3.7725e-05, "loss": 2.2613, "step": 491000 }, { "epoch": 1.31, "learning_rate": 3.77125e-05, "loss": 2.2623, "step": 491500 }, { "epoch": 1.31, "learning_rate": 3.77e-05, "loss": 2.277, "step": 492000 }, { "epoch": 1.31, "learning_rate": 3.76875e-05, "loss": 2.2668, "step": 492500 }, { "epoch": 1.31, "learning_rate": 3.7675e-05, "loss": 2.279, "step": 493000 }, { "epoch": 1.31, "learning_rate": 3.76625e-05, "loss": 2.2801, "step": 493500 }, { "epoch": 1.32, "learning_rate": 3.765e-05, "loss": 2.2629, "step": 494000 }, { "epoch": 1.32, "learning_rate": 3.76375e-05, "loss": 2.2639, "step": 494500 }, { "epoch": 1.32, "learning_rate": 3.7625e-05, "loss": 2.2504, "step": 495000 }, { "epoch": 1.32, "learning_rate": 3.76125e-05, "loss": 2.2511, "step": 495500 }, { "epoch": 1.32, "learning_rate": 3.76e-05, "loss": 2.2655, "step": 496000 }, { "epoch": 1.32, "learning_rate": 3.7587500000000006e-05, "loss": 2.2659, "step": 496500 }, { "epoch": 1.32, "learning_rate": 3.7575e-05, "loss": 2.2753, "step": 497000 }, { "epoch": 1.32, "learning_rate": 3.75625e-05, "loss": 2.256, "step": 497500 }, { "epoch": 1.33, "learning_rate": 3.7550000000000005e-05, "loss": 2.2709, "step": 498000 }, { "epoch": 1.33, "learning_rate": 3.7537500000000004e-05, "loss": 2.2572, "step": 498500 }, { "epoch": 1.33, "learning_rate": 3.7525e-05, "loss": 2.275, "step": 499000 }, { "epoch": 1.33, "learning_rate": 3.7512500000000004e-05, "loss": 2.2473, "step": 499500 }, { "epoch": 1.33, "learning_rate": 3.7500000000000003e-05, "loss": 2.2769, "step": 500000 }, { "epoch": 1.33, "learning_rate": 3.74875e-05, "loss": 2.2666, "step": 500500 }, { "epoch": 1.33, "learning_rate": 3.7475e-05, "loss": 2.2569, "step": 501000 }, { "epoch": 1.34, "learning_rate": 3.74625e-05, "loss": 2.2614, "step": 501500 }, { "epoch": 1.34, "learning_rate": 3.745e-05, "loss": 2.2503, "step": 502000 }, { "epoch": 1.34, "learning_rate": 3.74375e-05, "loss": 2.2467, "step": 502500 }, { "epoch": 1.34, "learning_rate": 3.7425e-05, "loss": 2.25, "step": 503000 }, { "epoch": 1.34, "learning_rate": 3.74125e-05, "loss": 2.2623, "step": 503500 }, { "epoch": 1.34, "learning_rate": 3.74e-05, "loss": 2.2599, "step": 504000 }, { "epoch": 1.34, "learning_rate": 3.73875e-05, "loss": 2.2491, "step": 504500 }, { "epoch": 1.34, "learning_rate": 3.737500000000001e-05, "loss": 2.2519, "step": 505000 }, { "epoch": 1.35, "learning_rate": 3.73625e-05, "loss": 2.2646, "step": 505500 }, { "epoch": 1.35, "learning_rate": 3.735e-05, "loss": 2.2473, "step": 506000 }, { "epoch": 1.35, "learning_rate": 3.7337500000000006e-05, "loss": 2.2273, "step": 506500 }, { "epoch": 1.35, "learning_rate": 3.7325000000000006e-05, "loss": 2.251, "step": 507000 }, { "epoch": 1.35, "learning_rate": 3.73125e-05, "loss": 2.2403, "step": 507500 }, { "epoch": 1.35, "learning_rate": 3.73e-05, "loss": 2.2504, "step": 508000 }, { "epoch": 1.35, "learning_rate": 3.7287500000000005e-05, "loss": 2.2488, "step": 508500 }, { "epoch": 1.36, "learning_rate": 3.7275000000000005e-05, "loss": 2.2359, "step": 509000 }, { "epoch": 1.36, "learning_rate": 3.72625e-05, "loss": 2.2409, "step": 509500 }, { "epoch": 1.36, "learning_rate": 3.7250000000000004e-05, "loss": 2.2434, "step": 510000 }, { "epoch": 1.36, "learning_rate": 3.7237500000000004e-05, "loss": 2.2483, "step": 510500 }, { "epoch": 1.36, "learning_rate": 3.7225000000000004e-05, "loss": 2.2477, "step": 511000 }, { "epoch": 1.36, "learning_rate": 3.72125e-05, "loss": 2.2507, "step": 511500 }, { "epoch": 1.36, "learning_rate": 3.72e-05, "loss": 2.2481, "step": 512000 }, { "epoch": 1.36, "learning_rate": 3.71875e-05, "loss": 2.2453, "step": 512500 }, { "epoch": 1.37, "learning_rate": 3.7175e-05, "loss": 2.2387, "step": 513000 }, { "epoch": 1.37, "learning_rate": 3.71625e-05, "loss": 2.2318, "step": 513500 }, { "epoch": 1.37, "learning_rate": 3.715e-05, "loss": 2.2275, "step": 514000 }, { "epoch": 1.37, "learning_rate": 3.71375e-05, "loss": 2.2427, "step": 514500 }, { "epoch": 1.37, "learning_rate": 3.7125e-05, "loss": 2.2348, "step": 515000 }, { "epoch": 1.37, "learning_rate": 3.71125e-05, "loss": 2.2328, "step": 515500 }, { "epoch": 1.37, "learning_rate": 3.71e-05, "loss": 2.2339, "step": 516000 }, { "epoch": 1.38, "learning_rate": 3.70875e-05, "loss": 2.2524, "step": 516500 }, { "epoch": 1.38, "learning_rate": 3.707500000000001e-05, "loss": 2.2403, "step": 517000 }, { "epoch": 1.38, "learning_rate": 3.70625e-05, "loss": 2.2302, "step": 517500 }, { "epoch": 1.38, "learning_rate": 3.705e-05, "loss": 2.2342, "step": 518000 }, { "epoch": 1.38, "learning_rate": 3.7037500000000006e-05, "loss": 2.2332, "step": 518500 }, { "epoch": 1.38, "learning_rate": 3.7025000000000005e-05, "loss": 2.2435, "step": 519000 }, { "epoch": 1.38, "learning_rate": 3.70125e-05, "loss": 2.221, "step": 519500 }, { "epoch": 1.38, "learning_rate": 3.7e-05, "loss": 2.2378, "step": 520000 }, { "epoch": 1.39, "learning_rate": 3.6987500000000005e-05, "loss": 2.2389, "step": 520500 }, { "epoch": 1.39, "learning_rate": 3.6975000000000004e-05, "loss": 2.23, "step": 521000 }, { "epoch": 1.39, "learning_rate": 3.69625e-05, "loss": 2.2336, "step": 521500 }, { "epoch": 1.39, "learning_rate": 3.6950000000000004e-05, "loss": 2.2194, "step": 522000 }, { "epoch": 1.39, "learning_rate": 3.69375e-05, "loss": 2.2369, "step": 522500 }, { "epoch": 1.39, "learning_rate": 3.6925e-05, "loss": 2.2327, "step": 523000 }, { "epoch": 1.39, "learning_rate": 3.69125e-05, "loss": 2.2289, "step": 523500 }, { "epoch": 1.4, "learning_rate": 3.69e-05, "loss": 2.2519, "step": 524000 }, { "epoch": 1.4, "learning_rate": 3.68875e-05, "loss": 2.2539, "step": 524500 }, { "epoch": 1.4, "learning_rate": 3.6875e-05, "loss": 2.232, "step": 525000 }, { "epoch": 1.4, "learning_rate": 3.68625e-05, "loss": 2.2532, "step": 525500 }, { "epoch": 1.4, "learning_rate": 3.685e-05, "loss": 2.2284, "step": 526000 }, { "epoch": 1.4, "learning_rate": 3.68375e-05, "loss": 2.2291, "step": 526500 }, { "epoch": 1.4, "learning_rate": 3.6825e-05, "loss": 2.2485, "step": 527000 }, { "epoch": 1.4, "learning_rate": 3.68125e-05, "loss": 2.2217, "step": 527500 }, { "epoch": 1.41, "learning_rate": 3.68e-05, "loss": 2.2425, "step": 528000 }, { "epoch": 1.41, "learning_rate": 3.67875e-05, "loss": 2.2302, "step": 528500 }, { "epoch": 1.41, "learning_rate": 3.6775000000000006e-05, "loss": 2.2088, "step": 529000 }, { "epoch": 1.41, "learning_rate": 3.67625e-05, "loss": 2.2247, "step": 529500 }, { "epoch": 1.41, "learning_rate": 3.675e-05, "loss": 2.2157, "step": 530000 }, { "epoch": 1.41, "learning_rate": 3.6737500000000005e-05, "loss": 2.2323, "step": 530500 }, { "epoch": 1.41, "learning_rate": 3.6725000000000005e-05, "loss": 2.2256, "step": 531000 }, { "epoch": 1.41, "learning_rate": 3.67125e-05, "loss": 2.2399, "step": 531500 }, { "epoch": 1.42, "learning_rate": 3.6700000000000004e-05, "loss": 2.231, "step": 532000 }, { "epoch": 1.42, "learning_rate": 3.6687500000000004e-05, "loss": 2.2116, "step": 532500 }, { "epoch": 1.42, "learning_rate": 3.6675000000000004e-05, "loss": 2.2279, "step": 533000 }, { "epoch": 1.42, "learning_rate": 3.66625e-05, "loss": 2.2164, "step": 533500 }, { "epoch": 1.42, "learning_rate": 3.665e-05, "loss": 2.2139, "step": 534000 }, { "epoch": 1.42, "learning_rate": 3.66375e-05, "loss": 2.231, "step": 534500 }, { "epoch": 1.42, "learning_rate": 3.6625e-05, "loss": 2.2095, "step": 535000 }, { "epoch": 1.43, "learning_rate": 3.66125e-05, "loss": 2.2042, "step": 535500 }, { "epoch": 1.43, "learning_rate": 3.66e-05, "loss": 2.2303, "step": 536000 }, { "epoch": 1.43, "learning_rate": 3.65875e-05, "loss": 2.214, "step": 536500 }, { "epoch": 1.43, "learning_rate": 3.6575e-05, "loss": 2.2152, "step": 537000 }, { "epoch": 1.43, "learning_rate": 3.65625e-05, "loss": 2.2337, "step": 537500 }, { "epoch": 1.43, "learning_rate": 3.655e-05, "loss": 2.2165, "step": 538000 }, { "epoch": 1.43, "learning_rate": 3.65375e-05, "loss": 2.2205, "step": 538500 }, { "epoch": 1.43, "learning_rate": 3.652500000000001e-05, "loss": 2.2139, "step": 539000 }, { "epoch": 1.44, "learning_rate": 3.65125e-05, "loss": 2.2061, "step": 539500 }, { "epoch": 1.44, "learning_rate": 3.65e-05, "loss": 2.203, "step": 540000 }, { "epoch": 1.44, "learning_rate": 3.64875e-05, "loss": 2.2196, "step": 540500 }, { "epoch": 1.44, "learning_rate": 3.6475000000000006e-05, "loss": 2.1992, "step": 541000 }, { "epoch": 1.44, "learning_rate": 3.64625e-05, "loss": 2.1924, "step": 541500 }, { "epoch": 1.44, "learning_rate": 3.645e-05, "loss": 2.214, "step": 542000 }, { "epoch": 1.44, "learning_rate": 3.6437500000000005e-05, "loss": 2.2072, "step": 542500 }, { "epoch": 1.45, "learning_rate": 3.6425000000000004e-05, "loss": 2.2166, "step": 543000 }, { "epoch": 1.45, "learning_rate": 3.64125e-05, "loss": 2.2121, "step": 543500 }, { "epoch": 1.45, "learning_rate": 3.6400000000000004e-05, "loss": 2.2101, "step": 544000 }, { "epoch": 1.45, "learning_rate": 3.6387500000000004e-05, "loss": 2.2105, "step": 544500 }, { "epoch": 1.45, "learning_rate": 3.6375e-05, "loss": 2.2088, "step": 545000 }, { "epoch": 1.45, "learning_rate": 3.6362499999999996e-05, "loss": 2.2065, "step": 545500 }, { "epoch": 1.45, "learning_rate": 3.635e-05, "loss": 2.2011, "step": 546000 }, { "epoch": 1.45, "learning_rate": 3.63375e-05, "loss": 2.2014, "step": 546500 }, { "epoch": 1.46, "learning_rate": 3.6325e-05, "loss": 2.214, "step": 547000 }, { "epoch": 1.46, "learning_rate": 3.63125e-05, "loss": 2.2287, "step": 547500 }, { "epoch": 1.46, "learning_rate": 3.63e-05, "loss": 2.2105, "step": 548000 }, { "epoch": 1.46, "learning_rate": 3.62875e-05, "loss": 2.2039, "step": 548500 }, { "epoch": 1.46, "learning_rate": 3.6275e-05, "loss": 2.2177, "step": 549000 }, { "epoch": 1.46, "learning_rate": 3.62625e-05, "loss": 2.1898, "step": 549500 }, { "epoch": 1.46, "learning_rate": 3.625e-05, "loss": 2.2003, "step": 550000 }, { "epoch": 1.47, "learning_rate": 3.62375e-05, "loss": 2.2133, "step": 550500 }, { "epoch": 1.47, "learning_rate": 3.6225000000000006e-05, "loss": 2.2056, "step": 551000 }, { "epoch": 1.47, "learning_rate": 3.62125e-05, "loss": 2.1962, "step": 551500 }, { "epoch": 1.47, "learning_rate": 3.62e-05, "loss": 2.1955, "step": 552000 }, { "epoch": 1.47, "learning_rate": 3.61875e-05, "loss": 2.2175, "step": 552500 }, { "epoch": 1.47, "learning_rate": 3.6175000000000005e-05, "loss": 2.2035, "step": 553000 }, { "epoch": 1.47, "learning_rate": 3.61625e-05, "loss": 2.2053, "step": 553500 }, { "epoch": 1.47, "learning_rate": 3.615e-05, "loss": 2.1863, "step": 554000 }, { "epoch": 1.48, "learning_rate": 3.6137500000000004e-05, "loss": 2.1967, "step": 554500 }, { "epoch": 1.48, "learning_rate": 3.6125000000000004e-05, "loss": 2.1976, "step": 555000 }, { "epoch": 1.48, "learning_rate": 3.61125e-05, "loss": 2.2182, "step": 555500 }, { "epoch": 1.48, "learning_rate": 3.61e-05, "loss": 2.1963, "step": 556000 }, { "epoch": 1.48, "learning_rate": 3.60875e-05, "loss": 2.2003, "step": 556500 }, { "epoch": 1.48, "learning_rate": 3.6075e-05, "loss": 2.1697, "step": 557000 }, { "epoch": 1.48, "learning_rate": 3.60625e-05, "loss": 2.2002, "step": 557500 }, { "epoch": 1.49, "learning_rate": 3.605e-05, "loss": 2.1983, "step": 558000 }, { "epoch": 1.49, "learning_rate": 3.60375e-05, "loss": 2.1948, "step": 558500 }, { "epoch": 1.49, "learning_rate": 3.6025e-05, "loss": 2.1925, "step": 559000 }, { "epoch": 1.49, "learning_rate": 3.60125e-05, "loss": 2.2183, "step": 559500 }, { "epoch": 1.49, "learning_rate": 3.6e-05, "loss": 2.1902, "step": 560000 }, { "epoch": 1.49, "learning_rate": 3.59875e-05, "loss": 2.1971, "step": 560500 }, { "epoch": 1.49, "learning_rate": 3.5975e-05, "loss": 2.2059, "step": 561000 }, { "epoch": 1.49, "learning_rate": 3.59625e-05, "loss": 2.2046, "step": 561500 }, { "epoch": 1.5, "learning_rate": 3.595e-05, "loss": 2.1816, "step": 562000 }, { "epoch": 1.5, "learning_rate": 3.59375e-05, "loss": 2.2105, "step": 562500 }, { "epoch": 1.5, "learning_rate": 3.5925000000000006e-05, "loss": 2.1952, "step": 563000 }, { "epoch": 1.5, "learning_rate": 3.5912500000000006e-05, "loss": 2.2042, "step": 563500 }, { "epoch": 1.5, "learning_rate": 3.59e-05, "loss": 2.1713, "step": 564000 }, { "epoch": 1.5, "learning_rate": 3.5887500000000005e-05, "loss": 2.198, "step": 564500 }, { "epoch": 1.5, "learning_rate": 3.5875000000000005e-05, "loss": 2.1963, "step": 565000 }, { "epoch": 1.51, "learning_rate": 3.5862500000000004e-05, "loss": 2.1863, "step": 565500 }, { "epoch": 1.51, "learning_rate": 3.585e-05, "loss": 2.1786, "step": 566000 }, { "epoch": 1.51, "learning_rate": 3.5837500000000004e-05, "loss": 2.1994, "step": 566500 }, { "epoch": 1.51, "learning_rate": 3.5825000000000003e-05, "loss": 2.1874, "step": 567000 }, { "epoch": 1.51, "learning_rate": 3.58125e-05, "loss": 2.1816, "step": 567500 }, { "epoch": 1.51, "learning_rate": 3.58e-05, "loss": 2.172, "step": 568000 }, { "epoch": 1.51, "learning_rate": 3.57875e-05, "loss": 2.1777, "step": 568500 }, { "epoch": 1.51, "learning_rate": 3.5775e-05, "loss": 2.1963, "step": 569000 }, { "epoch": 1.52, "learning_rate": 3.57625e-05, "loss": 2.1743, "step": 569500 }, { "epoch": 1.52, "learning_rate": 3.575e-05, "loss": 2.1963, "step": 570000 }, { "epoch": 1.52, "learning_rate": 3.57375e-05, "loss": 2.1764, "step": 570500 }, { "epoch": 1.52, "learning_rate": 3.5725e-05, "loss": 2.1796, "step": 571000 }, { "epoch": 1.52, "learning_rate": 3.571250000000001e-05, "loss": 2.1901, "step": 571500 }, { "epoch": 1.52, "learning_rate": 3.57e-05, "loss": 2.1948, "step": 572000 }, { "epoch": 1.52, "learning_rate": 3.56875e-05, "loss": 2.1999, "step": 572500 }, { "epoch": 1.53, "learning_rate": 3.5675e-05, "loss": 2.1858, "step": 573000 }, { "epoch": 1.53, "learning_rate": 3.5662500000000006e-05, "loss": 2.184, "step": 573500 }, { "epoch": 1.53, "learning_rate": 3.565e-05, "loss": 2.1787, "step": 574000 }, { "epoch": 1.53, "learning_rate": 3.56375e-05, "loss": 2.1779, "step": 574500 }, { "epoch": 1.53, "learning_rate": 3.5625000000000005e-05, "loss": 2.169, "step": 575000 }, { "epoch": 1.53, "learning_rate": 3.5612500000000005e-05, "loss": 2.1807, "step": 575500 }, { "epoch": 1.53, "learning_rate": 3.56e-05, "loss": 2.1784, "step": 576000 }, { "epoch": 1.53, "learning_rate": 3.5587500000000004e-05, "loss": 2.1722, "step": 576500 }, { "epoch": 1.54, "learning_rate": 3.5575000000000004e-05, "loss": 2.1907, "step": 577000 }, { "epoch": 1.54, "learning_rate": 3.5562500000000004e-05, "loss": 2.1685, "step": 577500 }, { "epoch": 1.54, "learning_rate": 3.555e-05, "loss": 2.1801, "step": 578000 }, { "epoch": 1.54, "learning_rate": 3.55375e-05, "loss": 2.174, "step": 578500 }, { "epoch": 1.54, "learning_rate": 3.5525e-05, "loss": 2.1785, "step": 579000 }, { "epoch": 1.54, "learning_rate": 3.55125e-05, "loss": 2.1781, "step": 579500 }, { "epoch": 1.54, "learning_rate": 3.55e-05, "loss": 2.1671, "step": 580000 }, { "epoch": 1.55, "learning_rate": 3.54875e-05, "loss": 2.1754, "step": 580500 }, { "epoch": 1.55, "learning_rate": 3.5475e-05, "loss": 2.1864, "step": 581000 }, { "epoch": 1.55, "learning_rate": 3.54625e-05, "loss": 2.1712, "step": 581500 }, { "epoch": 1.55, "learning_rate": 3.545e-05, "loss": 2.1612, "step": 582000 }, { "epoch": 1.55, "learning_rate": 3.54375e-05, "loss": 2.1619, "step": 582500 }, { "epoch": 1.55, "learning_rate": 3.5425e-05, "loss": 2.1771, "step": 583000 }, { "epoch": 1.55, "learning_rate": 3.541250000000001e-05, "loss": 2.178, "step": 583500 }, { "epoch": 1.55, "learning_rate": 3.54e-05, "loss": 2.1861, "step": 584000 }, { "epoch": 1.56, "learning_rate": 3.53875e-05, "loss": 2.174, "step": 584500 }, { "epoch": 1.56, "learning_rate": 3.5375e-05, "loss": 2.1839, "step": 585000 }, { "epoch": 1.56, "learning_rate": 3.5362500000000006e-05, "loss": 2.1706, "step": 585500 }, { "epoch": 1.56, "learning_rate": 3.535e-05, "loss": 2.1675, "step": 586000 }, { "epoch": 1.56, "learning_rate": 3.53375e-05, "loss": 2.1698, "step": 586500 }, { "epoch": 1.56, "learning_rate": 3.5325000000000005e-05, "loss": 2.1746, "step": 587000 }, { "epoch": 1.56, "learning_rate": 3.5312500000000005e-05, "loss": 2.1605, "step": 587500 }, { "epoch": 1.57, "learning_rate": 3.53e-05, "loss": 2.1654, "step": 588000 }, { "epoch": 1.57, "learning_rate": 3.5287500000000004e-05, "loss": 2.1591, "step": 588500 }, { "epoch": 1.57, "learning_rate": 3.5275000000000004e-05, "loss": 2.1569, "step": 589000 }, { "epoch": 1.57, "learning_rate": 3.52625e-05, "loss": 2.1557, "step": 589500 }, { "epoch": 1.57, "learning_rate": 3.525e-05, "loss": 2.1682, "step": 590000 }, { "epoch": 1.57, "learning_rate": 3.52375e-05, "loss": 2.1511, "step": 590500 }, { "epoch": 1.57, "learning_rate": 3.5225e-05, "loss": 2.1572, "step": 591000 }, { "epoch": 1.57, "learning_rate": 3.52125e-05, "loss": 2.1668, "step": 591500 }, { "epoch": 1.58, "learning_rate": 3.52e-05, "loss": 2.1754, "step": 592000 }, { "epoch": 1.58, "learning_rate": 3.51875e-05, "loss": 2.1613, "step": 592500 }, { "epoch": 1.58, "learning_rate": 3.5175e-05, "loss": 2.1675, "step": 593000 }, { "epoch": 1.58, "learning_rate": 3.51625e-05, "loss": 2.1699, "step": 593500 }, { "epoch": 1.58, "learning_rate": 3.515e-05, "loss": 2.1481, "step": 594000 }, { "epoch": 1.58, "learning_rate": 3.51375e-05, "loss": 2.1737, "step": 594500 }, { "epoch": 1.58, "learning_rate": 3.5125e-05, "loss": 2.1693, "step": 595000 }, { "epoch": 1.59, "learning_rate": 3.5112500000000006e-05, "loss": 2.169, "step": 595500 }, { "epoch": 1.59, "learning_rate": 3.51e-05, "loss": 2.1629, "step": 596000 }, { "epoch": 1.59, "learning_rate": 3.50875e-05, "loss": 2.1472, "step": 596500 }, { "epoch": 1.59, "learning_rate": 3.5075000000000006e-05, "loss": 2.1422, "step": 597000 }, { "epoch": 1.59, "learning_rate": 3.5062500000000005e-05, "loss": 2.1626, "step": 597500 }, { "epoch": 1.59, "learning_rate": 3.505e-05, "loss": 2.1624, "step": 598000 }, { "epoch": 1.59, "learning_rate": 3.50375e-05, "loss": 2.1497, "step": 598500 }, { "epoch": 1.59, "learning_rate": 3.5025000000000004e-05, "loss": 2.164, "step": 599000 }, { "epoch": 1.6, "learning_rate": 3.5012500000000004e-05, "loss": 2.161, "step": 599500 }, { "epoch": 1.6, "learning_rate": 3.5e-05, "loss": 2.1566, "step": 600000 }, { "epoch": 1.6, "learning_rate": 3.4987500000000003e-05, "loss": 2.1578, "step": 600500 }, { "epoch": 1.6, "learning_rate": 3.4975e-05, "loss": 2.1398, "step": 601000 }, { "epoch": 1.6, "learning_rate": 3.49625e-05, "loss": 2.1518, "step": 601500 }, { "epoch": 1.6, "learning_rate": 3.495e-05, "loss": 2.1647, "step": 602000 }, { "epoch": 1.6, "learning_rate": 3.49375e-05, "loss": 2.16, "step": 602500 }, { "epoch": 1.61, "learning_rate": 3.4925e-05, "loss": 2.1518, "step": 603000 }, { "epoch": 1.61, "learning_rate": 3.49125e-05, "loss": 2.167, "step": 603500 }, { "epoch": 1.61, "learning_rate": 3.49e-05, "loss": 2.1515, "step": 604000 }, { "epoch": 1.61, "learning_rate": 3.48875e-05, "loss": 2.1436, "step": 604500 }, { "epoch": 1.61, "learning_rate": 3.4875e-05, "loss": 2.1567, "step": 605000 }, { "epoch": 1.61, "learning_rate": 3.48625e-05, "loss": 2.1569, "step": 605500 }, { "epoch": 1.61, "learning_rate": 3.485e-05, "loss": 2.1668, "step": 606000 }, { "epoch": 1.61, "learning_rate": 3.48375e-05, "loss": 2.1551, "step": 606500 }, { "epoch": 1.62, "learning_rate": 3.4825e-05, "loss": 2.1378, "step": 607000 }, { "epoch": 1.62, "learning_rate": 3.4812500000000006e-05, "loss": 2.1508, "step": 607500 }, { "epoch": 1.62, "learning_rate": 3.48e-05, "loss": 2.1554, "step": 608000 }, { "epoch": 1.62, "learning_rate": 3.47875e-05, "loss": 2.1575, "step": 608500 }, { "epoch": 1.62, "learning_rate": 3.4775000000000005e-05, "loss": 2.1521, "step": 609000 }, { "epoch": 1.62, "learning_rate": 3.4762500000000005e-05, "loss": 2.1529, "step": 609500 }, { "epoch": 1.62, "learning_rate": 3.475e-05, "loss": 2.1537, "step": 610000 }, { "epoch": 1.63, "learning_rate": 3.47375e-05, "loss": 2.1436, "step": 610500 }, { "epoch": 1.63, "learning_rate": 3.4725000000000004e-05, "loss": 2.1524, "step": 611000 }, { "epoch": 1.63, "learning_rate": 3.4712500000000003e-05, "loss": 2.1649, "step": 611500 }, { "epoch": 1.63, "learning_rate": 3.4699999999999996e-05, "loss": 2.1409, "step": 612000 }, { "epoch": 1.63, "learning_rate": 3.46875e-05, "loss": 2.1666, "step": 612500 }, { "epoch": 1.63, "learning_rate": 3.4675e-05, "loss": 2.1359, "step": 613000 }, { "epoch": 1.63, "learning_rate": 3.46625e-05, "loss": 2.1382, "step": 613500 }, { "epoch": 1.63, "learning_rate": 3.465e-05, "loss": 2.1259, "step": 614000 }, { "epoch": 1.64, "learning_rate": 3.46375e-05, "loss": 2.1444, "step": 614500 }, { "epoch": 1.64, "learning_rate": 3.4625e-05, "loss": 2.1394, "step": 615000 }, { "epoch": 1.64, "learning_rate": 3.46125e-05, "loss": 2.1368, "step": 615500 }, { "epoch": 1.64, "learning_rate": 3.46e-05, "loss": 2.138, "step": 616000 }, { "epoch": 1.64, "learning_rate": 3.45875e-05, "loss": 2.1303, "step": 616500 }, { "epoch": 1.64, "learning_rate": 3.4575e-05, "loss": 2.1426, "step": 617000 }, { "epoch": 1.64, "learning_rate": 3.45625e-05, "loss": 2.1405, "step": 617500 }, { "epoch": 1.65, "learning_rate": 3.455e-05, "loss": 2.1398, "step": 618000 }, { "epoch": 1.65, "learning_rate": 3.45375e-05, "loss": 2.147, "step": 618500 }, { "epoch": 1.65, "learning_rate": 3.4525e-05, "loss": 2.1289, "step": 619000 }, { "epoch": 1.65, "learning_rate": 3.4512500000000005e-05, "loss": 2.1543, "step": 619500 }, { "epoch": 1.65, "learning_rate": 3.45e-05, "loss": 2.1408, "step": 620000 }, { "epoch": 1.65, "learning_rate": 3.44875e-05, "loss": 2.133, "step": 620500 }, { "epoch": 1.65, "learning_rate": 3.4475000000000005e-05, "loss": 2.1337, "step": 621000 }, { "epoch": 1.65, "learning_rate": 3.4462500000000004e-05, "loss": 2.1359, "step": 621500 }, { "epoch": 1.66, "learning_rate": 3.445e-05, "loss": 2.128, "step": 622000 }, { "epoch": 1.66, "learning_rate": 3.4437500000000004e-05, "loss": 2.1344, "step": 622500 }, { "epoch": 1.66, "learning_rate": 3.4425e-05, "loss": 2.1399, "step": 623000 }, { "epoch": 1.66, "learning_rate": 3.44125e-05, "loss": 2.1441, "step": 623500 }, { "epoch": 1.66, "learning_rate": 3.4399999999999996e-05, "loss": 2.1444, "step": 624000 }, { "epoch": 1.66, "learning_rate": 3.43875e-05, "loss": 2.1529, "step": 624500 }, { "epoch": 1.66, "learning_rate": 3.4375e-05, "loss": 2.1413, "step": 625000 }, { "epoch": 1.67, "learning_rate": 3.43625e-05, "loss": 2.143, "step": 625500 }, { "epoch": 1.67, "learning_rate": 3.435e-05, "loss": 2.1455, "step": 626000 }, { "epoch": 1.67, "learning_rate": 3.43375e-05, "loss": 2.1458, "step": 626500 }, { "epoch": 1.67, "learning_rate": 3.4325e-05, "loss": 2.1377, "step": 627000 }, { "epoch": 1.67, "learning_rate": 3.43125e-05, "loss": 2.1477, "step": 627500 }, { "epoch": 1.67, "learning_rate": 3.430000000000001e-05, "loss": 2.1252, "step": 628000 }, { "epoch": 1.67, "learning_rate": 3.42875e-05, "loss": 2.1364, "step": 628500 }, { "epoch": 1.67, "learning_rate": 3.4275e-05, "loss": 2.1293, "step": 629000 }, { "epoch": 1.68, "learning_rate": 3.4262500000000006e-05, "loss": 2.1181, "step": 629500 }, { "epoch": 1.68, "learning_rate": 3.4250000000000006e-05, "loss": 2.1193, "step": 630000 }, { "epoch": 1.68, "learning_rate": 3.42375e-05, "loss": 2.1443, "step": 630500 }, { "epoch": 1.68, "learning_rate": 3.4225e-05, "loss": 2.1431, "step": 631000 }, { "epoch": 1.68, "learning_rate": 3.4212500000000005e-05, "loss": 2.111, "step": 631500 }, { "epoch": 1.68, "learning_rate": 3.4200000000000005e-05, "loss": 2.1212, "step": 632000 }, { "epoch": 1.68, "learning_rate": 3.41875e-05, "loss": 2.1228, "step": 632500 }, { "epoch": 1.69, "learning_rate": 3.4175000000000004e-05, "loss": 2.1429, "step": 633000 }, { "epoch": 1.69, "learning_rate": 3.4162500000000004e-05, "loss": 2.1277, "step": 633500 }, { "epoch": 1.69, "learning_rate": 3.415e-05, "loss": 2.12, "step": 634000 }, { "epoch": 1.69, "learning_rate": 3.41375e-05, "loss": 2.1349, "step": 634500 }, { "epoch": 1.69, "learning_rate": 3.4125e-05, "loss": 2.1327, "step": 635000 }, { "epoch": 1.69, "learning_rate": 3.41125e-05, "loss": 2.1071, "step": 635500 }, { "epoch": 1.69, "learning_rate": 3.41e-05, "loss": 2.1254, "step": 636000 }, { "epoch": 1.69, "learning_rate": 3.40875e-05, "loss": 2.1229, "step": 636500 }, { "epoch": 1.7, "learning_rate": 3.4075e-05, "loss": 2.1194, "step": 637000 }, { "epoch": 1.7, "learning_rate": 3.40625e-05, "loss": 2.1273, "step": 637500 }, { "epoch": 1.7, "learning_rate": 3.405e-05, "loss": 2.1265, "step": 638000 }, { "epoch": 1.7, "learning_rate": 3.40375e-05, "loss": 2.1237, "step": 638500 }, { "epoch": 1.7, "learning_rate": 3.4025e-05, "loss": 2.1204, "step": 639000 }, { "epoch": 1.7, "learning_rate": 3.40125e-05, "loss": 2.1406, "step": 639500 }, { "epoch": 1.7, "learning_rate": 3.4000000000000007e-05, "loss": 2.1327, "step": 640000 }, { "epoch": 1.71, "learning_rate": 3.39875e-05, "loss": 2.1324, "step": 640500 }, { "epoch": 1.71, "learning_rate": 3.3975e-05, "loss": 2.1259, "step": 641000 }, { "epoch": 1.71, "learning_rate": 3.3962500000000006e-05, "loss": 2.1325, "step": 641500 }, { "epoch": 1.71, "learning_rate": 3.3950000000000005e-05, "loss": 2.128, "step": 642000 }, { "epoch": 1.71, "learning_rate": 3.39375e-05, "loss": 2.0937, "step": 642500 }, { "epoch": 1.71, "learning_rate": 3.3925e-05, "loss": 2.1193, "step": 643000 }, { "epoch": 1.71, "learning_rate": 3.3912500000000004e-05, "loss": 2.1217, "step": 643500 }, { "epoch": 1.71, "learning_rate": 3.3900000000000004e-05, "loss": 2.1171, "step": 644000 }, { "epoch": 1.72, "learning_rate": 3.38875e-05, "loss": 2.1337, "step": 644500 }, { "epoch": 1.72, "learning_rate": 3.3875000000000003e-05, "loss": 2.1133, "step": 645000 }, { "epoch": 1.72, "learning_rate": 3.38625e-05, "loss": 2.116, "step": 645500 }, { "epoch": 1.72, "learning_rate": 3.385e-05, "loss": 2.1114, "step": 646000 }, { "epoch": 1.72, "learning_rate": 3.38375e-05, "loss": 2.1271, "step": 646500 }, { "epoch": 1.72, "learning_rate": 3.3825e-05, "loss": 2.1092, "step": 647000 }, { "epoch": 1.72, "learning_rate": 3.38125e-05, "loss": 2.1231, "step": 647500 }, { "epoch": 1.73, "learning_rate": 3.38e-05, "loss": 2.1219, "step": 648000 }, { "epoch": 1.73, "learning_rate": 3.37875e-05, "loss": 2.1114, "step": 648500 }, { "epoch": 1.73, "learning_rate": 3.3775e-05, "loss": 2.1184, "step": 649000 }, { "epoch": 1.73, "learning_rate": 3.37625e-05, "loss": 2.1104, "step": 649500 }, { "epoch": 1.73, "learning_rate": 3.375000000000001e-05, "loss": 2.1087, "step": 650000 }, { "epoch": 1.73, "learning_rate": 3.37375e-05, "loss": 2.1204, "step": 650500 }, { "epoch": 1.73, "learning_rate": 3.3725e-05, "loss": 2.1137, "step": 651000 }, { "epoch": 1.73, "learning_rate": 3.37125e-05, "loss": 2.1131, "step": 651500 }, { "epoch": 1.74, "learning_rate": 3.3700000000000006e-05, "loss": 2.0957, "step": 652000 }, { "epoch": 1.74, "learning_rate": 3.36875e-05, "loss": 2.1098, "step": 652500 }, { "epoch": 1.74, "learning_rate": 3.3675e-05, "loss": 2.097, "step": 653000 }, { "epoch": 1.74, "learning_rate": 3.3662500000000005e-05, "loss": 2.1055, "step": 653500 }, { "epoch": 1.74, "learning_rate": 3.3650000000000005e-05, "loss": 2.1099, "step": 654000 }, { "epoch": 1.74, "learning_rate": 3.36375e-05, "loss": 2.1224, "step": 654500 }, { "epoch": 1.74, "learning_rate": 3.3625000000000004e-05, "loss": 2.1019, "step": 655000 }, { "epoch": 1.75, "learning_rate": 3.3612500000000004e-05, "loss": 2.1097, "step": 655500 }, { "epoch": 1.75, "learning_rate": 3.3600000000000004e-05, "loss": 2.0897, "step": 656000 }, { "epoch": 1.75, "learning_rate": 3.3587499999999996e-05, "loss": 2.1146, "step": 656500 }, { "epoch": 1.75, "learning_rate": 3.3575e-05, "loss": 2.0994, "step": 657000 }, { "epoch": 1.75, "learning_rate": 3.35625e-05, "loss": 2.1274, "step": 657500 }, { "epoch": 1.75, "learning_rate": 3.355e-05, "loss": 2.108, "step": 658000 }, { "epoch": 1.75, "learning_rate": 3.35375e-05, "loss": 2.1262, "step": 658500 }, { "epoch": 1.75, "learning_rate": 3.3525e-05, "loss": 2.1037, "step": 659000 }, { "epoch": 1.76, "learning_rate": 3.35125e-05, "loss": 2.1031, "step": 659500 }, { "epoch": 1.76, "learning_rate": 3.35e-05, "loss": 2.1039, "step": 660000 }, { "epoch": 1.76, "learning_rate": 3.34875e-05, "loss": 2.1041, "step": 660500 }, { "epoch": 1.76, "learning_rate": 3.3475e-05, "loss": 2.1007, "step": 661000 }, { "epoch": 1.76, "learning_rate": 3.34625e-05, "loss": 2.092, "step": 661500 }, { "epoch": 1.76, "learning_rate": 3.345000000000001e-05, "loss": 2.1143, "step": 662000 }, { "epoch": 1.76, "learning_rate": 3.34375e-05, "loss": 2.1024, "step": 662500 }, { "epoch": 1.77, "learning_rate": 3.3425e-05, "loss": 2.1124, "step": 663000 }, { "epoch": 1.77, "learning_rate": 3.34125e-05, "loss": 2.1115, "step": 663500 }, { "epoch": 1.77, "learning_rate": 3.3400000000000005e-05, "loss": 2.1081, "step": 664000 }, { "epoch": 1.77, "learning_rate": 3.33875e-05, "loss": 2.1077, "step": 664500 }, { "epoch": 1.77, "learning_rate": 3.3375e-05, "loss": 2.1053, "step": 665000 }, { "epoch": 1.77, "learning_rate": 3.3362500000000005e-05, "loss": 2.1037, "step": 665500 }, { "epoch": 1.77, "learning_rate": 3.3350000000000004e-05, "loss": 2.1044, "step": 666000 }, { "epoch": 1.77, "learning_rate": 3.33375e-05, "loss": 2.0955, "step": 666500 }, { "epoch": 1.78, "learning_rate": 3.3325000000000004e-05, "loss": 2.0971, "step": 667000 }, { "epoch": 1.78, "learning_rate": 3.33125e-05, "loss": 2.1108, "step": 667500 }, { "epoch": 1.78, "learning_rate": 3.33e-05, "loss": 2.1207, "step": 668000 }, { "epoch": 1.78, "learning_rate": 3.3287499999999996e-05, "loss": 2.091, "step": 668500 }, { "epoch": 1.78, "learning_rate": 3.3275e-05, "loss": 2.1003, "step": 669000 }, { "epoch": 1.78, "learning_rate": 3.32625e-05, "loss": 2.0997, "step": 669500 }, { "epoch": 1.78, "learning_rate": 3.325e-05, "loss": 2.1005, "step": 670000 }, { "epoch": 1.79, "learning_rate": 3.32375e-05, "loss": 2.104, "step": 670500 }, { "epoch": 1.79, "learning_rate": 3.3225e-05, "loss": 2.1054, "step": 671000 }, { "epoch": 1.79, "learning_rate": 3.32125e-05, "loss": 2.0908, "step": 671500 }, { "epoch": 1.79, "learning_rate": 3.32e-05, "loss": 2.0973, "step": 672000 }, { "epoch": 1.79, "learning_rate": 3.31875e-05, "loss": 2.0938, "step": 672500 }, { "epoch": 1.79, "learning_rate": 3.3175e-05, "loss": 2.0926, "step": 673000 }, { "epoch": 1.79, "learning_rate": 3.31625e-05, "loss": 2.1102, "step": 673500 }, { "epoch": 1.79, "learning_rate": 3.3150000000000006e-05, "loss": 2.0817, "step": 674000 }, { "epoch": 1.8, "learning_rate": 3.31375e-05, "loss": 2.1093, "step": 674500 }, { "epoch": 1.8, "learning_rate": 3.3125e-05, "loss": 2.0986, "step": 675000 }, { "epoch": 1.8, "learning_rate": 3.31125e-05, "loss": 2.1006, "step": 675500 }, { "epoch": 1.8, "learning_rate": 3.3100000000000005e-05, "loss": 2.0994, "step": 676000 }, { "epoch": 1.8, "learning_rate": 3.30875e-05, "loss": 2.0929, "step": 676500 }, { "epoch": 1.8, "learning_rate": 3.3075e-05, "loss": 2.0983, "step": 677000 }, { "epoch": 1.8, "learning_rate": 3.3062500000000004e-05, "loss": 2.088, "step": 677500 }, { "epoch": 1.81, "learning_rate": 3.3050000000000004e-05, "loss": 2.0951, "step": 678000 }, { "epoch": 1.81, "learning_rate": 3.30375e-05, "loss": 2.0919, "step": 678500 }, { "epoch": 1.81, "learning_rate": 3.3025e-05, "loss": 2.0916, "step": 679000 }, { "epoch": 1.81, "learning_rate": 3.30125e-05, "loss": 2.0849, "step": 679500 }, { "epoch": 1.81, "learning_rate": 3.3e-05, "loss": 2.1074, "step": 680000 }, { "epoch": 1.81, "learning_rate": 3.29875e-05, "loss": 2.1001, "step": 680500 }, { "epoch": 1.81, "learning_rate": 3.2975e-05, "loss": 2.0942, "step": 681000 }, { "epoch": 1.81, "learning_rate": 3.29625e-05, "loss": 2.104, "step": 681500 }, { "epoch": 1.82, "learning_rate": 3.295e-05, "loss": 2.0984, "step": 682000 }, { "epoch": 1.82, "learning_rate": 3.29375e-05, "loss": 2.1062, "step": 682500 }, { "epoch": 1.82, "learning_rate": 3.2925e-05, "loss": 2.0878, "step": 683000 }, { "epoch": 1.82, "learning_rate": 3.29125e-05, "loss": 2.0901, "step": 683500 }, { "epoch": 1.82, "learning_rate": 3.29e-05, "loss": 2.1096, "step": 684000 }, { "epoch": 1.82, "learning_rate": 3.28875e-05, "loss": 2.087, "step": 684500 }, { "epoch": 1.82, "learning_rate": 3.2875e-05, "loss": 2.0869, "step": 685000 }, { "epoch": 1.82, "learning_rate": 3.28625e-05, "loss": 2.0972, "step": 685500 }, { "epoch": 1.83, "learning_rate": 3.2850000000000006e-05, "loss": 2.0921, "step": 686000 }, { "epoch": 1.83, "learning_rate": 3.28375e-05, "loss": 2.0878, "step": 686500 }, { "epoch": 1.83, "learning_rate": 3.2825e-05, "loss": 2.0985, "step": 687000 }, { "epoch": 1.83, "learning_rate": 3.2812500000000005e-05, "loss": 2.1088, "step": 687500 }, { "epoch": 1.83, "learning_rate": 3.2800000000000004e-05, "loss": 2.0771, "step": 688000 }, { "epoch": 1.83, "learning_rate": 3.2787500000000004e-05, "loss": 2.0901, "step": 688500 }, { "epoch": 1.83, "learning_rate": 3.2775e-05, "loss": 2.0828, "step": 689000 }, { "epoch": 1.84, "learning_rate": 3.2762500000000004e-05, "loss": 2.0849, "step": 689500 }, { "epoch": 1.84, "learning_rate": 3.275e-05, "loss": 2.0826, "step": 690000 }, { "epoch": 1.84, "learning_rate": 3.27375e-05, "loss": 2.0907, "step": 690500 }, { "epoch": 1.84, "learning_rate": 3.2725e-05, "loss": 2.0861, "step": 691000 }, { "epoch": 1.84, "learning_rate": 3.27125e-05, "loss": 2.0918, "step": 691500 }, { "epoch": 1.84, "learning_rate": 3.27e-05, "loss": 2.0822, "step": 692000 }, { "epoch": 1.84, "learning_rate": 3.26875e-05, "loss": 2.088, "step": 692500 }, { "epoch": 1.84, "learning_rate": 3.2675e-05, "loss": 2.0879, "step": 693000 }, { "epoch": 1.85, "learning_rate": 3.26625e-05, "loss": 2.0747, "step": 693500 }, { "epoch": 1.85, "learning_rate": 3.265e-05, "loss": 2.1017, "step": 694000 }, { "epoch": 1.85, "learning_rate": 3.263750000000001e-05, "loss": 2.0813, "step": 694500 }, { "epoch": 1.85, "learning_rate": 3.2625e-05, "loss": 2.1043, "step": 695000 }, { "epoch": 1.85, "learning_rate": 3.26125e-05, "loss": 2.0853, "step": 695500 }, { "epoch": 1.85, "learning_rate": 3.26e-05, "loss": 2.0882, "step": 696000 }, { "epoch": 1.85, "learning_rate": 3.2587500000000006e-05, "loss": 2.0806, "step": 696500 }, { "epoch": 1.86, "learning_rate": 3.2575e-05, "loss": 2.0728, "step": 697000 }, { "epoch": 1.86, "learning_rate": 3.25625e-05, "loss": 2.0688, "step": 697500 }, { "epoch": 1.86, "learning_rate": 3.2550000000000005e-05, "loss": 2.0849, "step": 698000 }, { "epoch": 1.86, "learning_rate": 3.2537500000000005e-05, "loss": 2.0595, "step": 698500 }, { "epoch": 1.86, "learning_rate": 3.2525e-05, "loss": 2.0758, "step": 699000 }, { "epoch": 1.86, "learning_rate": 3.2512500000000004e-05, "loss": 2.0735, "step": 699500 }, { "epoch": 1.86, "learning_rate": 3.2500000000000004e-05, "loss": 2.0766, "step": 700000 }, { "epoch": 1.86, "learning_rate": 3.2487500000000004e-05, "loss": 2.0821, "step": 700500 }, { "epoch": 1.87, "learning_rate": 3.2474999999999997e-05, "loss": 2.0742, "step": 701000 }, { "epoch": 1.87, "learning_rate": 3.24625e-05, "loss": 2.0832, "step": 701500 }, { "epoch": 1.87, "learning_rate": 3.245e-05, "loss": 2.0901, "step": 702000 }, { "epoch": 1.87, "learning_rate": 3.24375e-05, "loss": 2.079, "step": 702500 }, { "epoch": 1.87, "learning_rate": 3.2425e-05, "loss": 2.0791, "step": 703000 }, { "epoch": 1.87, "learning_rate": 3.24125e-05, "loss": 2.0726, "step": 703500 }, { "epoch": 1.87, "learning_rate": 3.24e-05, "loss": 2.0779, "step": 704000 }, { "epoch": 1.88, "learning_rate": 3.23875e-05, "loss": 2.0689, "step": 704500 }, { "epoch": 1.88, "learning_rate": 3.2375e-05, "loss": 2.099, "step": 705000 }, { "epoch": 1.88, "learning_rate": 3.23625e-05, "loss": 2.0705, "step": 705500 }, { "epoch": 1.88, "learning_rate": 3.235e-05, "loss": 2.0833, "step": 706000 }, { "epoch": 1.88, "learning_rate": 3.233750000000001e-05, "loss": 2.081, "step": 706500 }, { "epoch": 1.88, "learning_rate": 3.2325e-05, "loss": 2.0753, "step": 707000 }, { "epoch": 1.88, "learning_rate": 3.23125e-05, "loss": 2.0812, "step": 707500 }, { "epoch": 1.88, "learning_rate": 3.2300000000000006e-05, "loss": 2.0754, "step": 708000 }, { "epoch": 1.89, "learning_rate": 3.2287500000000006e-05, "loss": 2.0644, "step": 708500 }, { "epoch": 1.89, "learning_rate": 3.2275e-05, "loss": 2.0633, "step": 709000 }, { "epoch": 1.89, "learning_rate": 3.22625e-05, "loss": 2.0704, "step": 709500 }, { "epoch": 1.89, "learning_rate": 3.2250000000000005e-05, "loss": 2.0693, "step": 710000 }, { "epoch": 1.89, "learning_rate": 3.2237500000000004e-05, "loss": 2.0621, "step": 710500 }, { "epoch": 1.89, "learning_rate": 3.2225e-05, "loss": 2.0613, "step": 711000 }, { "epoch": 1.89, "learning_rate": 3.2212500000000004e-05, "loss": 2.0829, "step": 711500 }, { "epoch": 1.9, "learning_rate": 3.2200000000000003e-05, "loss": 2.0776, "step": 712000 }, { "epoch": 1.9, "learning_rate": 3.21875e-05, "loss": 2.0666, "step": 712500 }, { "epoch": 1.9, "learning_rate": 3.2175e-05, "loss": 2.0718, "step": 713000 }, { "epoch": 1.9, "learning_rate": 3.21625e-05, "loss": 2.0557, "step": 713500 }, { "epoch": 1.9, "learning_rate": 3.215e-05, "loss": 2.0748, "step": 714000 }, { "epoch": 1.9, "learning_rate": 3.21375e-05, "loss": 2.0768, "step": 714500 }, { "epoch": 1.9, "learning_rate": 3.2125e-05, "loss": 2.0746, "step": 715000 }, { "epoch": 1.9, "learning_rate": 3.21125e-05, "loss": 2.0691, "step": 715500 }, { "epoch": 1.91, "learning_rate": 3.21e-05, "loss": 2.0981, "step": 716000 }, { "epoch": 1.91, "learning_rate": 3.20875e-05, "loss": 2.065, "step": 716500 }, { "epoch": 1.91, "learning_rate": 3.2075e-05, "loss": 2.0617, "step": 717000 }, { "epoch": 1.91, "learning_rate": 3.20625e-05, "loss": 2.0687, "step": 717500 }, { "epoch": 1.91, "learning_rate": 3.205e-05, "loss": 2.0796, "step": 718000 }, { "epoch": 1.91, "learning_rate": 3.2037500000000006e-05, "loss": 2.0771, "step": 718500 }, { "epoch": 1.91, "learning_rate": 3.2025e-05, "loss": 2.0565, "step": 719000 }, { "epoch": 1.92, "learning_rate": 3.20125e-05, "loss": 2.057, "step": 719500 }, { "epoch": 1.92, "learning_rate": 3.2000000000000005e-05, "loss": 2.0765, "step": 720000 }, { "epoch": 1.92, "learning_rate": 3.1987500000000005e-05, "loss": 2.0718, "step": 720500 }, { "epoch": 1.92, "learning_rate": 3.1975e-05, "loss": 2.0679, "step": 721000 }, { "epoch": 1.92, "learning_rate": 3.19625e-05, "loss": 2.0682, "step": 721500 }, { "epoch": 1.92, "learning_rate": 3.1950000000000004e-05, "loss": 2.0649, "step": 722000 }, { "epoch": 1.92, "learning_rate": 3.1937500000000004e-05, "loss": 2.0663, "step": 722500 }, { "epoch": 1.92, "learning_rate": 3.1925e-05, "loss": 2.0635, "step": 723000 }, { "epoch": 1.93, "learning_rate": 3.19125e-05, "loss": 2.0593, "step": 723500 }, { "epoch": 1.93, "learning_rate": 3.19e-05, "loss": 2.0553, "step": 724000 }, { "epoch": 1.93, "learning_rate": 3.18875e-05, "loss": 2.0704, "step": 724500 }, { "epoch": 1.93, "learning_rate": 3.1875e-05, "loss": 2.0626, "step": 725000 }, { "epoch": 1.93, "learning_rate": 3.18625e-05, "loss": 2.0684, "step": 725500 }, { "epoch": 1.93, "learning_rate": 3.185e-05, "loss": 2.0656, "step": 726000 }, { "epoch": 1.93, "learning_rate": 3.18375e-05, "loss": 2.063, "step": 726500 }, { "epoch": 1.94, "learning_rate": 3.1825e-05, "loss": 2.0791, "step": 727000 }, { "epoch": 1.94, "learning_rate": 3.18125e-05, "loss": 2.0736, "step": 727500 }, { "epoch": 1.94, "learning_rate": 3.18e-05, "loss": 2.0846, "step": 728000 }, { "epoch": 1.94, "learning_rate": 3.17875e-05, "loss": 2.0591, "step": 728500 }, { "epoch": 1.94, "learning_rate": 3.1775e-05, "loss": 2.0678, "step": 729000 }, { "epoch": 1.94, "learning_rate": 3.17625e-05, "loss": 2.0774, "step": 729500 }, { "epoch": 1.94, "learning_rate": 3.175e-05, "loss": 2.0609, "step": 730000 }, { "epoch": 1.94, "learning_rate": 3.1737500000000006e-05, "loss": 2.0616, "step": 730500 }, { "epoch": 1.95, "learning_rate": 3.1725e-05, "loss": 2.0559, "step": 731000 }, { "epoch": 1.95, "learning_rate": 3.17125e-05, "loss": 2.0679, "step": 731500 }, { "epoch": 1.95, "learning_rate": 3.1700000000000005e-05, "loss": 2.0627, "step": 732000 }, { "epoch": 1.95, "learning_rate": 3.1687500000000005e-05, "loss": 2.0718, "step": 732500 }, { "epoch": 1.95, "learning_rate": 3.1675e-05, "loss": 2.0618, "step": 733000 }, { "epoch": 1.95, "learning_rate": 3.16625e-05, "loss": 2.0481, "step": 733500 }, { "epoch": 1.95, "learning_rate": 3.1650000000000004e-05, "loss": 2.0564, "step": 734000 }, { "epoch": 1.96, "learning_rate": 3.16375e-05, "loss": 2.0481, "step": 734500 }, { "epoch": 1.96, "learning_rate": 3.1624999999999996e-05, "loss": 2.0705, "step": 735000 }, { "epoch": 1.96, "learning_rate": 3.16125e-05, "loss": 2.0598, "step": 735500 }, { "epoch": 1.96, "learning_rate": 3.16e-05, "loss": 2.06, "step": 736000 }, { "epoch": 1.96, "learning_rate": 3.15875e-05, "loss": 2.0592, "step": 736500 }, { "epoch": 1.96, "learning_rate": 3.1575e-05, "loss": 2.0589, "step": 737000 }, { "epoch": 1.96, "learning_rate": 3.15625e-05, "loss": 2.0496, "step": 737500 }, { "epoch": 1.96, "learning_rate": 3.155e-05, "loss": 2.0523, "step": 738000 }, { "epoch": 1.97, "learning_rate": 3.15375e-05, "loss": 2.048, "step": 738500 }, { "epoch": 1.97, "learning_rate": 3.1525e-05, "loss": 2.0522, "step": 739000 }, { "epoch": 1.97, "learning_rate": 3.15125e-05, "loss": 2.0568, "step": 739500 }, { "epoch": 1.97, "learning_rate": 3.15e-05, "loss": 2.0491, "step": 740000 }, { "epoch": 1.97, "learning_rate": 3.1487500000000006e-05, "loss": 2.0537, "step": 740500 }, { "epoch": 1.97, "learning_rate": 3.1475e-05, "loss": 2.0655, "step": 741000 }, { "epoch": 1.97, "learning_rate": 3.14625e-05, "loss": 2.0642, "step": 741500 }, { "epoch": 1.98, "learning_rate": 3.145e-05, "loss": 2.0672, "step": 742000 }, { "epoch": 1.98, "learning_rate": 3.1437500000000005e-05, "loss": 2.0545, "step": 742500 }, { "epoch": 1.98, "learning_rate": 3.1425e-05, "loss": 2.0377, "step": 743000 }, { "epoch": 1.98, "learning_rate": 3.14125e-05, "loss": 2.0605, "step": 743500 }, { "epoch": 1.98, "learning_rate": 3.1400000000000004e-05, "loss": 2.0607, "step": 744000 }, { "epoch": 1.98, "learning_rate": 3.1387500000000004e-05, "loss": 2.0574, "step": 744500 }, { "epoch": 1.98, "learning_rate": 3.1375e-05, "loss": 2.0641, "step": 745000 }, { "epoch": 1.98, "learning_rate": 3.13625e-05, "loss": 2.0424, "step": 745500 }, { "epoch": 1.99, "learning_rate": 3.135e-05, "loss": 2.04, "step": 746000 }, { "epoch": 1.99, "learning_rate": 3.13375e-05, "loss": 2.0515, "step": 746500 }, { "epoch": 1.99, "learning_rate": 3.1324999999999996e-05, "loss": 2.0395, "step": 747000 }, { "epoch": 1.99, "learning_rate": 3.13125e-05, "loss": 2.045, "step": 747500 }, { "epoch": 1.99, "learning_rate": 3.13e-05, "loss": 2.0568, "step": 748000 }, { "epoch": 1.99, "learning_rate": 3.12875e-05, "loss": 2.057, "step": 748500 }, { "epoch": 1.99, "learning_rate": 3.1275e-05, "loss": 2.0579, "step": 749000 }, { "epoch": 2.0, "learning_rate": 3.12625e-05, "loss": 2.05, "step": 749500 }, { "epoch": 2.0, "learning_rate": 3.125e-05, "loss": 2.0373, "step": 750000 }, { "epoch": 2.0, "learning_rate": 3.12375e-05, "loss": 2.0633, "step": 750500 }, { "epoch": 2.0, "learning_rate": 3.122500000000001e-05, "loss": 2.051, "step": 751000 }, { "epoch": 2.0, "learning_rate": 3.12125e-05, "loss": 2.0515, "step": 751500 }, { "epoch": 2.0, "learning_rate": 3.12e-05, "loss": 2.0439, "step": 752000 }, { "epoch": 2.0, "learning_rate": 3.1187500000000006e-05, "loss": 2.0632, "step": 752500 }, { "epoch": 2.0, "learning_rate": 3.1175000000000006e-05, "loss": 2.042, "step": 753000 }, { "epoch": 2.01, "learning_rate": 3.11625e-05, "loss": 2.0449, "step": 753500 }, { "epoch": 2.01, "learning_rate": 3.115e-05, "loss": 2.04, "step": 754000 }, { "epoch": 2.01, "learning_rate": 3.1137500000000005e-05, "loss": 2.0453, "step": 754500 }, { "epoch": 2.01, "learning_rate": 3.1125000000000004e-05, "loss": 2.0437, "step": 755000 }, { "epoch": 2.01, "learning_rate": 3.11125e-05, "loss": 2.0349, "step": 755500 }, { "epoch": 2.01, "learning_rate": 3.1100000000000004e-05, "loss": 2.043, "step": 756000 }, { "epoch": 2.01, "learning_rate": 3.1087500000000003e-05, "loss": 2.0314, "step": 756500 }, { "epoch": 2.02, "learning_rate": 3.1075e-05, "loss": 2.0399, "step": 757000 }, { "epoch": 2.02, "learning_rate": 3.10625e-05, "loss": 2.0407, "step": 757500 }, { "epoch": 2.02, "learning_rate": 3.105e-05, "loss": 2.0374, "step": 758000 }, { "epoch": 2.02, "learning_rate": 3.10375e-05, "loss": 2.0493, "step": 758500 }, { "epoch": 2.02, "learning_rate": 3.1025e-05, "loss": 2.0272, "step": 759000 }, { "epoch": 2.02, "learning_rate": 3.10125e-05, "loss": 2.0389, "step": 759500 }, { "epoch": 2.02, "learning_rate": 3.1e-05, "loss": 2.0403, "step": 760000 }, { "epoch": 2.02, "learning_rate": 3.09875e-05, "loss": 2.0514, "step": 760500 }, { "epoch": 2.03, "learning_rate": 3.0975e-05, "loss": 2.0416, "step": 761000 }, { "epoch": 2.03, "learning_rate": 3.09625e-05, "loss": 2.0277, "step": 761500 }, { "epoch": 2.03, "learning_rate": 3.095e-05, "loss": 2.0365, "step": 762000 }, { "epoch": 2.03, "learning_rate": 3.09375e-05, "loss": 2.0261, "step": 762500 }, { "epoch": 2.03, "learning_rate": 3.0925000000000006e-05, "loss": 2.0204, "step": 763000 }, { "epoch": 2.03, "learning_rate": 3.09125e-05, "loss": 2.0605, "step": 763500 }, { "epoch": 2.03, "learning_rate": 3.09e-05, "loss": 2.0272, "step": 764000 }, { "epoch": 2.04, "learning_rate": 3.0887500000000005e-05, "loss": 2.0436, "step": 764500 }, { "epoch": 2.04, "learning_rate": 3.0875000000000005e-05, "loss": 2.0482, "step": 765000 }, { "epoch": 2.04, "learning_rate": 3.08625e-05, "loss": 2.041, "step": 765500 }, { "epoch": 2.04, "learning_rate": 3.0850000000000004e-05, "loss": 2.0222, "step": 766000 }, { "epoch": 2.04, "learning_rate": 3.0837500000000004e-05, "loss": 2.0467, "step": 766500 }, { "epoch": 2.04, "learning_rate": 3.0825000000000004e-05, "loss": 2.0376, "step": 767000 }, { "epoch": 2.04, "learning_rate": 3.08125e-05, "loss": 2.0392, "step": 767500 }, { "epoch": 2.04, "learning_rate": 3.08e-05, "loss": 2.0393, "step": 768000 }, { "epoch": 2.05, "learning_rate": 3.07875e-05, "loss": 2.0262, "step": 768500 }, { "epoch": 2.05, "learning_rate": 3.0775e-05, "loss": 2.0396, "step": 769000 }, { "epoch": 2.05, "learning_rate": 3.07625e-05, "loss": 2.0376, "step": 769500 }, { "epoch": 2.05, "learning_rate": 3.075e-05, "loss": 2.0386, "step": 770000 }, { "epoch": 2.05, "learning_rate": 3.07375e-05, "loss": 2.0338, "step": 770500 }, { "epoch": 2.05, "learning_rate": 3.0725e-05, "loss": 2.0481, "step": 771000 }, { "epoch": 2.05, "learning_rate": 3.07125e-05, "loss": 2.0331, "step": 771500 }, { "epoch": 2.06, "learning_rate": 3.07e-05, "loss": 2.0302, "step": 772000 }, { "epoch": 2.06, "learning_rate": 3.06875e-05, "loss": 2.0375, "step": 772500 }, { "epoch": 2.06, "learning_rate": 3.067500000000001e-05, "loss": 2.0348, "step": 773000 }, { "epoch": 2.06, "learning_rate": 3.06625e-05, "loss": 2.0256, "step": 773500 }, { "epoch": 2.06, "learning_rate": 3.065e-05, "loss": 2.0322, "step": 774000 }, { "epoch": 2.06, "learning_rate": 3.06375e-05, "loss": 2.0451, "step": 774500 }, { "epoch": 2.06, "learning_rate": 3.0625000000000006e-05, "loss": 2.0283, "step": 775000 }, { "epoch": 2.06, "learning_rate": 3.06125e-05, "loss": 2.0338, "step": 775500 }, { "epoch": 2.07, "learning_rate": 3.06e-05, "loss": 2.0288, "step": 776000 }, { "epoch": 2.07, "learning_rate": 3.0587500000000005e-05, "loss": 2.0262, "step": 776500 }, { "epoch": 2.07, "learning_rate": 3.0575000000000005e-05, "loss": 2.0306, "step": 777000 }, { "epoch": 2.07, "learning_rate": 3.05625e-05, "loss": 2.04, "step": 777500 }, { "epoch": 2.07, "learning_rate": 3.0550000000000004e-05, "loss": 2.0251, "step": 778000 }, { "epoch": 2.07, "learning_rate": 3.0537500000000004e-05, "loss": 2.0251, "step": 778500 }, { "epoch": 2.07, "learning_rate": 3.0525e-05, "loss": 2.02, "step": 779000 }, { "epoch": 2.08, "learning_rate": 3.05125e-05, "loss": 2.0326, "step": 779500 }, { "epoch": 2.08, "learning_rate": 3.05e-05, "loss": 2.0229, "step": 780000 }, { "epoch": 2.08, "learning_rate": 3.0487500000000002e-05, "loss": 2.0287, "step": 780500 }, { "epoch": 2.08, "learning_rate": 3.0475000000000002e-05, "loss": 2.0291, "step": 781000 }, { "epoch": 2.08, "learning_rate": 3.04625e-05, "loss": 2.0176, "step": 781500 }, { "epoch": 2.08, "learning_rate": 3.045e-05, "loss": 2.0283, "step": 782000 }, { "epoch": 2.08, "learning_rate": 3.04375e-05, "loss": 2.0378, "step": 782500 }, { "epoch": 2.08, "learning_rate": 3.0425000000000004e-05, "loss": 2.0341, "step": 783000 }, { "epoch": 2.09, "learning_rate": 3.04125e-05, "loss": 2.0352, "step": 783500 }, { "epoch": 2.09, "learning_rate": 3.04e-05, "loss": 2.0223, "step": 784000 }, { "epoch": 2.09, "learning_rate": 3.0387500000000003e-05, "loss": 2.021, "step": 784500 }, { "epoch": 2.09, "learning_rate": 3.0375000000000003e-05, "loss": 2.0287, "step": 785000 }, { "epoch": 2.09, "learning_rate": 3.03625e-05, "loss": 2.0318, "step": 785500 }, { "epoch": 2.09, "learning_rate": 3.035e-05, "loss": 2.0242, "step": 786000 }, { "epoch": 2.09, "learning_rate": 3.0337500000000002e-05, "loss": 2.034, "step": 786500 }, { "epoch": 2.1, "learning_rate": 3.0325000000000002e-05, "loss": 2.0168, "step": 787000 }, { "epoch": 2.1, "learning_rate": 3.0312499999999998e-05, "loss": 2.0332, "step": 787500 }, { "epoch": 2.1, "learning_rate": 3.03e-05, "loss": 2.0148, "step": 788000 }, { "epoch": 2.1, "learning_rate": 3.02875e-05, "loss": 2.0194, "step": 788500 }, { "epoch": 2.1, "learning_rate": 3.0275000000000004e-05, "loss": 2.0164, "step": 789000 }, { "epoch": 2.1, "learning_rate": 3.02625e-05, "loss": 2.0192, "step": 789500 }, { "epoch": 2.1, "learning_rate": 3.025e-05, "loss": 2.0307, "step": 790000 }, { "epoch": 2.1, "learning_rate": 3.0237500000000003e-05, "loss": 2.032, "step": 790500 }, { "epoch": 2.11, "learning_rate": 3.0225000000000003e-05, "loss": 2.0252, "step": 791000 }, { "epoch": 2.11, "learning_rate": 3.02125e-05, "loss": 2.0326, "step": 791500 }, { "epoch": 2.11, "learning_rate": 3.02e-05, "loss": 2.025, "step": 792000 }, { "epoch": 2.11, "learning_rate": 3.0187500000000002e-05, "loss": 2.0161, "step": 792500 }, { "epoch": 2.11, "learning_rate": 3.0175e-05, "loss": 2.0236, "step": 793000 }, { "epoch": 2.11, "learning_rate": 3.0162499999999998e-05, "loss": 2.0369, "step": 793500 }, { "epoch": 2.11, "learning_rate": 3.015e-05, "loss": 2.0184, "step": 794000 }, { "epoch": 2.12, "learning_rate": 3.01375e-05, "loss": 2.0274, "step": 794500 }, { "epoch": 2.12, "learning_rate": 3.0125000000000004e-05, "loss": 2.0171, "step": 795000 }, { "epoch": 2.12, "learning_rate": 3.01125e-05, "loss": 2.0274, "step": 795500 }, { "epoch": 2.12, "learning_rate": 3.01e-05, "loss": 2.0272, "step": 796000 }, { "epoch": 2.12, "learning_rate": 3.0087500000000003e-05, "loss": 2.0089, "step": 796500 }, { "epoch": 2.12, "learning_rate": 3.0075000000000003e-05, "loss": 2.0187, "step": 797000 }, { "epoch": 2.12, "learning_rate": 3.00625e-05, "loss": 2.0127, "step": 797500 }, { "epoch": 2.12, "learning_rate": 3.0050000000000002e-05, "loss": 2.0112, "step": 798000 }, { "epoch": 2.13, "learning_rate": 3.00375e-05, "loss": 2.0369, "step": 798500 }, { "epoch": 2.13, "learning_rate": 3.0025000000000005e-05, "loss": 2.0094, "step": 799000 }, { "epoch": 2.13, "learning_rate": 3.0012499999999998e-05, "loss": 2.028, "step": 799500 }, { "epoch": 2.13, "learning_rate": 3e-05, "loss": 1.9975, "step": 800000 }, { "epoch": 2.13, "learning_rate": 2.99875e-05, "loss": 2.0067, "step": 800500 }, { "epoch": 2.13, "learning_rate": 2.9975000000000004e-05, "loss": 2.0244, "step": 801000 }, { "epoch": 2.13, "learning_rate": 2.99625e-05, "loss": 2.0172, "step": 801500 }, { "epoch": 2.14, "learning_rate": 2.995e-05, "loss": 2.0065, "step": 802000 }, { "epoch": 2.14, "learning_rate": 2.9937500000000003e-05, "loss": 2.0219, "step": 802500 }, { "epoch": 2.14, "learning_rate": 2.9925000000000002e-05, "loss": 2.0079, "step": 803000 }, { "epoch": 2.14, "learning_rate": 2.99125e-05, "loss": 2.0004, "step": 803500 }, { "epoch": 2.14, "learning_rate": 2.9900000000000002e-05, "loss": 2.0283, "step": 804000 }, { "epoch": 2.14, "learning_rate": 2.98875e-05, "loss": 2.0314, "step": 804500 }, { "epoch": 2.14, "learning_rate": 2.9875000000000004e-05, "loss": 2.0361, "step": 805000 }, { "epoch": 2.14, "learning_rate": 2.9862499999999997e-05, "loss": 2.0109, "step": 805500 }, { "epoch": 2.15, "learning_rate": 2.985e-05, "loss": 2.0033, "step": 806000 }, { "epoch": 2.15, "learning_rate": 2.98375e-05, "loss": 2.0203, "step": 806500 }, { "epoch": 2.15, "learning_rate": 2.9825000000000003e-05, "loss": 2.0174, "step": 807000 }, { "epoch": 2.15, "learning_rate": 2.98125e-05, "loss": 2.0379, "step": 807500 }, { "epoch": 2.15, "learning_rate": 2.98e-05, "loss": 2.0096, "step": 808000 }, { "epoch": 2.15, "learning_rate": 2.9787500000000002e-05, "loss": 1.987, "step": 808500 }, { "epoch": 2.15, "learning_rate": 2.9775000000000002e-05, "loss": 2.0102, "step": 809000 }, { "epoch": 2.16, "learning_rate": 2.97625e-05, "loss": 2.0037, "step": 809500 }, { "epoch": 2.16, "learning_rate": 2.975e-05, "loss": 2.0052, "step": 810000 }, { "epoch": 2.16, "learning_rate": 2.97375e-05, "loss": 2.0073, "step": 810500 }, { "epoch": 2.16, "learning_rate": 2.9725000000000004e-05, "loss": 1.9947, "step": 811000 }, { "epoch": 2.16, "learning_rate": 2.9712499999999997e-05, "loss": 2.0309, "step": 811500 }, { "epoch": 2.16, "learning_rate": 2.97e-05, "loss": 2.0188, "step": 812000 }, { "epoch": 2.16, "learning_rate": 2.96875e-05, "loss": 2.0266, "step": 812500 }, { "epoch": 2.16, "learning_rate": 2.9675000000000003e-05, "loss": 2.0019, "step": 813000 }, { "epoch": 2.17, "learning_rate": 2.9662500000000003e-05, "loss": 2.0142, "step": 813500 }, { "epoch": 2.17, "learning_rate": 2.965e-05, "loss": 1.9963, "step": 814000 }, { "epoch": 2.17, "learning_rate": 2.9637500000000002e-05, "loss": 2.0204, "step": 814500 }, { "epoch": 2.17, "learning_rate": 2.9625000000000002e-05, "loss": 2.0103, "step": 815000 }, { "epoch": 2.17, "learning_rate": 2.9612500000000005e-05, "loss": 2.0074, "step": 815500 }, { "epoch": 2.17, "learning_rate": 2.96e-05, "loss": 2.0169, "step": 816000 }, { "epoch": 2.17, "learning_rate": 2.95875e-05, "loss": 2.0138, "step": 816500 }, { "epoch": 2.18, "learning_rate": 2.9575000000000004e-05, "loss": 2.0138, "step": 817000 }, { "epoch": 2.18, "learning_rate": 2.9562500000000004e-05, "loss": 2.0041, "step": 817500 }, { "epoch": 2.18, "learning_rate": 2.955e-05, "loss": 2.0042, "step": 818000 }, { "epoch": 2.18, "learning_rate": 2.95375e-05, "loss": 2.0068, "step": 818500 }, { "epoch": 2.18, "learning_rate": 2.9525000000000003e-05, "loss": 2.0202, "step": 819000 }, { "epoch": 2.18, "learning_rate": 2.9512500000000002e-05, "loss": 2.0035, "step": 819500 }, { "epoch": 2.18, "learning_rate": 2.95e-05, "loss": 2.0067, "step": 820000 }, { "epoch": 2.18, "learning_rate": 2.9487500000000002e-05, "loss": 2.006, "step": 820500 }, { "epoch": 2.19, "learning_rate": 2.9475e-05, "loss": 2.0035, "step": 821000 }, { "epoch": 2.19, "learning_rate": 2.9462500000000005e-05, "loss": 2.0069, "step": 821500 }, { "epoch": 2.19, "learning_rate": 2.945e-05, "loss": 1.9899, "step": 822000 }, { "epoch": 2.19, "learning_rate": 2.94375e-05, "loss": 2.0201, "step": 822500 }, { "epoch": 2.19, "learning_rate": 2.9425000000000004e-05, "loss": 2.0077, "step": 823000 }, { "epoch": 2.19, "learning_rate": 2.9412500000000003e-05, "loss": 2.0136, "step": 823500 }, { "epoch": 2.19, "learning_rate": 2.94e-05, "loss": 2.0134, "step": 824000 }, { "epoch": 2.2, "learning_rate": 2.9387500000000003e-05, "loss": 2.02, "step": 824500 }, { "epoch": 2.2, "learning_rate": 2.9375000000000003e-05, "loss": 2.011, "step": 825000 }, { "epoch": 2.2, "learning_rate": 2.9362500000000002e-05, "loss": 2.0038, "step": 825500 }, { "epoch": 2.2, "learning_rate": 2.935e-05, "loss": 2.0082, "step": 826000 }, { "epoch": 2.2, "learning_rate": 2.93375e-05, "loss": 2.0052, "step": 826500 }, { "epoch": 2.2, "learning_rate": 2.9325e-05, "loss": 2.0097, "step": 827000 }, { "epoch": 2.2, "learning_rate": 2.9312500000000004e-05, "loss": 1.9888, "step": 827500 }, { "epoch": 2.2, "learning_rate": 2.93e-05, "loss": 2.0067, "step": 828000 }, { "epoch": 2.21, "learning_rate": 2.92875e-05, "loss": 1.993, "step": 828500 }, { "epoch": 2.21, "learning_rate": 2.9275000000000003e-05, "loss": 2.01, "step": 829000 }, { "epoch": 2.21, "learning_rate": 2.9262500000000003e-05, "loss": 2.0034, "step": 829500 }, { "epoch": 2.21, "learning_rate": 2.925e-05, "loss": 1.9942, "step": 830000 }, { "epoch": 2.21, "learning_rate": 2.9237500000000003e-05, "loss": 2.0122, "step": 830500 }, { "epoch": 2.21, "learning_rate": 2.9225000000000002e-05, "loss": 2.0039, "step": 831000 }, { "epoch": 2.21, "learning_rate": 2.9212500000000005e-05, "loss": 2.0017, "step": 831500 }, { "epoch": 2.21, "learning_rate": 2.9199999999999998e-05, "loss": 2.012, "step": 832000 }, { "epoch": 2.22, "learning_rate": 2.91875e-05, "loss": 2.014, "step": 832500 }, { "epoch": 2.22, "learning_rate": 2.9175e-05, "loss": 1.9917, "step": 833000 }, { "epoch": 2.22, "learning_rate": 2.9162500000000004e-05, "loss": 2.0039, "step": 833500 }, { "epoch": 2.22, "learning_rate": 2.915e-05, "loss": 1.9957, "step": 834000 }, { "epoch": 2.22, "learning_rate": 2.91375e-05, "loss": 1.9899, "step": 834500 }, { "epoch": 2.22, "learning_rate": 2.9125000000000003e-05, "loss": 2.0133, "step": 835000 }, { "epoch": 2.22, "learning_rate": 2.9112500000000003e-05, "loss": 2.0037, "step": 835500 }, { "epoch": 2.23, "learning_rate": 2.91e-05, "loss": 1.9949, "step": 836000 }, { "epoch": 2.23, "learning_rate": 2.9087500000000002e-05, "loss": 2.0, "step": 836500 }, { "epoch": 2.23, "learning_rate": 2.9075000000000002e-05, "loss": 2.0079, "step": 837000 }, { "epoch": 2.23, "learning_rate": 2.9062500000000005e-05, "loss": 1.9791, "step": 837500 }, { "epoch": 2.23, "learning_rate": 2.9049999999999998e-05, "loss": 2.0051, "step": 838000 }, { "epoch": 2.23, "learning_rate": 2.90375e-05, "loss": 1.9925, "step": 838500 }, { "epoch": 2.23, "learning_rate": 2.9025e-05, "loss": 2.0001, "step": 839000 }, { "epoch": 2.23, "learning_rate": 2.9012500000000004e-05, "loss": 1.9851, "step": 839500 }, { "epoch": 2.24, "learning_rate": 2.9e-05, "loss": 1.9886, "step": 840000 }, { "epoch": 2.24, "learning_rate": 2.89875e-05, "loss": 1.9976, "step": 840500 }, { "epoch": 2.24, "learning_rate": 2.8975000000000003e-05, "loss": 1.9844, "step": 841000 }, { "epoch": 2.24, "learning_rate": 2.8962500000000003e-05, "loss": 1.9894, "step": 841500 }, { "epoch": 2.24, "learning_rate": 2.895e-05, "loss": 1.9974, "step": 842000 }, { "epoch": 2.24, "learning_rate": 2.8937500000000002e-05, "loss": 1.9755, "step": 842500 }, { "epoch": 2.24, "learning_rate": 2.8925000000000002e-05, "loss": 2.0146, "step": 843000 }, { "epoch": 2.25, "learning_rate": 2.8912500000000005e-05, "loss": 1.9939, "step": 843500 }, { "epoch": 2.25, "learning_rate": 2.8899999999999998e-05, "loss": 1.9954, "step": 844000 }, { "epoch": 2.25, "learning_rate": 2.88875e-05, "loss": 1.9926, "step": 844500 }, { "epoch": 2.25, "learning_rate": 2.8875e-05, "loss": 1.9923, "step": 845000 }, { "epoch": 2.25, "learning_rate": 2.8862500000000004e-05, "loss": 1.9989, "step": 845500 }, { "epoch": 2.25, "learning_rate": 2.885e-05, "loss": 1.991, "step": 846000 }, { "epoch": 2.25, "learning_rate": 2.88375e-05, "loss": 1.9918, "step": 846500 }, { "epoch": 2.25, "learning_rate": 2.8825000000000003e-05, "loss": 1.9912, "step": 847000 }, { "epoch": 2.26, "learning_rate": 2.8812500000000002e-05, "loss": 1.9914, "step": 847500 }, { "epoch": 2.26, "learning_rate": 2.88e-05, "loss": 1.9874, "step": 848000 }, { "epoch": 2.26, "learning_rate": 2.8787500000000002e-05, "loss": 2.0126, "step": 848500 }, { "epoch": 2.26, "learning_rate": 2.8775e-05, "loss": 1.9757, "step": 849000 }, { "epoch": 2.26, "learning_rate": 2.8762500000000005e-05, "loss": 1.9932, "step": 849500 }, { "epoch": 2.26, "learning_rate": 2.8749999999999997e-05, "loss": 1.9934, "step": 850000 }, { "epoch": 2.26, "learning_rate": 2.87375e-05, "loss": 1.9828, "step": 850500 }, { "epoch": 2.27, "learning_rate": 2.8725e-05, "loss": 1.9969, "step": 851000 }, { "epoch": 2.27, "learning_rate": 2.8712500000000003e-05, "loss": 1.9896, "step": 851500 }, { "epoch": 2.27, "learning_rate": 2.87e-05, "loss": 1.9766, "step": 852000 }, { "epoch": 2.27, "learning_rate": 2.86875e-05, "loss": 1.9979, "step": 852500 }, { "epoch": 2.27, "learning_rate": 2.8675000000000002e-05, "loss": 1.9831, "step": 853000 }, { "epoch": 2.27, "learning_rate": 2.8662500000000002e-05, "loss": 2.0093, "step": 853500 }, { "epoch": 2.27, "learning_rate": 2.865e-05, "loss": 1.9822, "step": 854000 }, { "epoch": 2.27, "learning_rate": 2.86375e-05, "loss": 1.9853, "step": 854500 }, { "epoch": 2.28, "learning_rate": 2.8625e-05, "loss": 1.9908, "step": 855000 }, { "epoch": 2.28, "learning_rate": 2.8612500000000004e-05, "loss": 1.9851, "step": 855500 }, { "epoch": 2.28, "learning_rate": 2.86e-05, "loss": 2.0079, "step": 856000 }, { "epoch": 2.28, "learning_rate": 2.85875e-05, "loss": 1.9747, "step": 856500 }, { "epoch": 2.28, "learning_rate": 2.8575000000000003e-05, "loss": 1.9918, "step": 857000 }, { "epoch": 2.28, "learning_rate": 2.8562500000000003e-05, "loss": 1.9855, "step": 857500 }, { "epoch": 2.28, "learning_rate": 2.855e-05, "loss": 2.0001, "step": 858000 }, { "epoch": 2.29, "learning_rate": 2.85375e-05, "loss": 2.0026, "step": 858500 }, { "epoch": 2.29, "learning_rate": 2.8525000000000002e-05, "loss": 1.9926, "step": 859000 }, { "epoch": 2.29, "learning_rate": 2.8512500000000002e-05, "loss": 1.9789, "step": 859500 }, { "epoch": 2.29, "learning_rate": 2.8499999999999998e-05, "loss": 1.9996, "step": 860000 }, { "epoch": 2.29, "learning_rate": 2.84875e-05, "loss": 1.9909, "step": 860500 }, { "epoch": 2.29, "learning_rate": 2.8475e-05, "loss": 1.9654, "step": 861000 }, { "epoch": 2.29, "learning_rate": 2.8462500000000004e-05, "loss": 1.9739, "step": 861500 }, { "epoch": 2.29, "learning_rate": 2.845e-05, "loss": 1.9801, "step": 862000 }, { "epoch": 2.3, "learning_rate": 2.84375e-05, "loss": 1.9701, "step": 862500 }, { "epoch": 2.3, "learning_rate": 2.8425000000000003e-05, "loss": 1.9752, "step": 863000 }, { "epoch": 2.3, "learning_rate": 2.8412500000000003e-05, "loss": 1.9665, "step": 863500 }, { "epoch": 2.3, "learning_rate": 2.84e-05, "loss": 1.9754, "step": 864000 }, { "epoch": 2.3, "learning_rate": 2.83875e-05, "loss": 1.9954, "step": 864500 }, { "epoch": 2.3, "learning_rate": 2.8375000000000002e-05, "loss": 1.9743, "step": 865000 }, { "epoch": 2.3, "learning_rate": 2.83625e-05, "loss": 1.9793, "step": 865500 }, { "epoch": 2.31, "learning_rate": 2.8349999999999998e-05, "loss": 1.9773, "step": 866000 }, { "epoch": 2.31, "learning_rate": 2.83375e-05, "loss": 1.9857, "step": 866500 }, { "epoch": 2.31, "learning_rate": 2.8325e-05, "loss": 1.9851, "step": 867000 }, { "epoch": 2.31, "learning_rate": 2.8312500000000004e-05, "loss": 1.9839, "step": 867500 }, { "epoch": 2.31, "learning_rate": 2.83e-05, "loss": 1.9799, "step": 868000 }, { "epoch": 2.31, "learning_rate": 2.82875e-05, "loss": 1.9774, "step": 868500 }, { "epoch": 2.31, "learning_rate": 2.8275000000000003e-05, "loss": 1.9994, "step": 869000 }, { "epoch": 2.31, "learning_rate": 2.8262500000000003e-05, "loss": 1.9642, "step": 869500 }, { "epoch": 2.32, "learning_rate": 2.825e-05, "loss": 1.9983, "step": 870000 }, { "epoch": 2.32, "learning_rate": 2.82375e-05, "loss": 1.9737, "step": 870500 }, { "epoch": 2.32, "learning_rate": 2.8225e-05, "loss": 1.9822, "step": 871000 }, { "epoch": 2.32, "learning_rate": 2.82125e-05, "loss": 1.9806, "step": 871500 }, { "epoch": 2.32, "learning_rate": 2.8199999999999998e-05, "loss": 1.9808, "step": 872000 }, { "epoch": 2.32, "learning_rate": 2.81875e-05, "loss": 1.9813, "step": 872500 }, { "epoch": 2.32, "learning_rate": 2.8175e-05, "loss": 1.9649, "step": 873000 }, { "epoch": 2.33, "learning_rate": 2.8162500000000004e-05, "loss": 1.9752, "step": 873500 }, { "epoch": 2.33, "learning_rate": 2.815e-05, "loss": 1.9793, "step": 874000 }, { "epoch": 2.33, "learning_rate": 2.81375e-05, "loss": 1.9818, "step": 874500 }, { "epoch": 2.33, "learning_rate": 2.8125000000000003e-05, "loss": 1.985, "step": 875000 }, { "epoch": 2.33, "learning_rate": 2.8112500000000002e-05, "loss": 1.9871, "step": 875500 }, { "epoch": 2.33, "learning_rate": 2.8100000000000005e-05, "loss": 1.9852, "step": 876000 }, { "epoch": 2.33, "learning_rate": 2.80875e-05, "loss": 1.9801, "step": 876500 }, { "epoch": 2.33, "learning_rate": 2.8075e-05, "loss": 1.9791, "step": 877000 }, { "epoch": 2.34, "learning_rate": 2.80625e-05, "loss": 1.9788, "step": 877500 }, { "epoch": 2.34, "learning_rate": 2.8050000000000004e-05, "loss": 1.9885, "step": 878000 }, { "epoch": 2.34, "learning_rate": 2.80375e-05, "loss": 1.9588, "step": 878500 }, { "epoch": 2.34, "learning_rate": 2.8025e-05, "loss": 1.9693, "step": 879000 }, { "epoch": 2.34, "learning_rate": 2.8012500000000003e-05, "loss": 1.9718, "step": 879500 }, { "epoch": 2.34, "learning_rate": 2.8000000000000003e-05, "loss": 1.9896, "step": 880000 }, { "epoch": 2.34, "learning_rate": 2.79875e-05, "loss": 1.98, "step": 880500 }, { "epoch": 2.35, "learning_rate": 2.7975000000000002e-05, "loss": 1.9685, "step": 881000 }, { "epoch": 2.35, "learning_rate": 2.7962500000000002e-05, "loss": 1.9795, "step": 881500 }, { "epoch": 2.35, "learning_rate": 2.7950000000000005e-05, "loss": 1.9739, "step": 882000 }, { "epoch": 2.35, "learning_rate": 2.79375e-05, "loss": 1.9574, "step": 882500 }, { "epoch": 2.35, "learning_rate": 2.7925e-05, "loss": 1.9598, "step": 883000 }, { "epoch": 2.35, "learning_rate": 2.7912500000000004e-05, "loss": 1.9896, "step": 883500 }, { "epoch": 2.35, "learning_rate": 2.7900000000000004e-05, "loss": 1.9819, "step": 884000 }, { "epoch": 2.35, "learning_rate": 2.78875e-05, "loss": 1.9795, "step": 884500 }, { "epoch": 2.36, "learning_rate": 2.7875e-05, "loss": 1.9823, "step": 885000 }, { "epoch": 2.36, "learning_rate": 2.7862500000000003e-05, "loss": 1.9673, "step": 885500 }, { "epoch": 2.36, "learning_rate": 2.7850000000000003e-05, "loss": 1.9751, "step": 886000 }, { "epoch": 2.36, "learning_rate": 2.78375e-05, "loss": 1.9869, "step": 886500 }, { "epoch": 2.36, "learning_rate": 2.7825000000000002e-05, "loss": 1.9643, "step": 887000 }, { "epoch": 2.36, "learning_rate": 2.7812500000000002e-05, "loss": 1.9687, "step": 887500 }, { "epoch": 2.36, "learning_rate": 2.7800000000000005e-05, "loss": 1.973, "step": 888000 }, { "epoch": 2.37, "learning_rate": 2.77875e-05, "loss": 1.9725, "step": 888500 }, { "epoch": 2.37, "learning_rate": 2.7775e-05, "loss": 1.9773, "step": 889000 }, { "epoch": 2.37, "learning_rate": 2.7762500000000004e-05, "loss": 1.9643, "step": 889500 }, { "epoch": 2.37, "learning_rate": 2.7750000000000004e-05, "loss": 1.9746, "step": 890000 }, { "epoch": 2.37, "learning_rate": 2.77375e-05, "loss": 1.983, "step": 890500 }, { "epoch": 2.37, "learning_rate": 2.7725e-05, "loss": 1.9574, "step": 891000 }, { "epoch": 2.37, "learning_rate": 2.7712500000000003e-05, "loss": 1.9667, "step": 891500 }, { "epoch": 2.37, "learning_rate": 2.7700000000000002e-05, "loss": 1.9758, "step": 892000 }, { "epoch": 2.38, "learning_rate": 2.76875e-05, "loss": 1.96, "step": 892500 }, { "epoch": 2.38, "learning_rate": 2.7675000000000002e-05, "loss": 1.9862, "step": 893000 }, { "epoch": 2.38, "learning_rate": 2.76625e-05, "loss": 1.9678, "step": 893500 }, { "epoch": 2.38, "learning_rate": 2.7650000000000005e-05, "loss": 1.9768, "step": 894000 }, { "epoch": 2.38, "learning_rate": 2.76375e-05, "loss": 1.973, "step": 894500 }, { "epoch": 2.38, "learning_rate": 2.7625e-05, "loss": 1.9788, "step": 895000 }, { "epoch": 2.38, "learning_rate": 2.7612500000000004e-05, "loss": 1.9743, "step": 895500 }, { "epoch": 2.39, "learning_rate": 2.7600000000000003e-05, "loss": 1.9693, "step": 896000 }, { "epoch": 2.39, "learning_rate": 2.75875e-05, "loss": 1.9636, "step": 896500 }, { "epoch": 2.39, "learning_rate": 2.7575e-05, "loss": 1.9736, "step": 897000 }, { "epoch": 2.39, "learning_rate": 2.7562500000000002e-05, "loss": 1.9862, "step": 897500 }, { "epoch": 2.39, "learning_rate": 2.7550000000000002e-05, "loss": 1.9648, "step": 898000 }, { "epoch": 2.39, "learning_rate": 2.75375e-05, "loss": 1.9792, "step": 898500 }, { "epoch": 2.39, "learning_rate": 2.7525e-05, "loss": 1.9758, "step": 899000 }, { "epoch": 2.39, "learning_rate": 2.75125e-05, "loss": 1.9626, "step": 899500 }, { "epoch": 2.4, "learning_rate": 2.7500000000000004e-05, "loss": 1.9728, "step": 900000 }, { "epoch": 2.4, "learning_rate": 2.74875e-05, "loss": 1.974, "step": 900500 }, { "epoch": 2.4, "learning_rate": 2.7475e-05, "loss": 1.972, "step": 901000 }, { "epoch": 2.4, "learning_rate": 2.7462500000000003e-05, "loss": 1.9666, "step": 901500 }, { "epoch": 2.4, "learning_rate": 2.7450000000000003e-05, "loss": 1.9657, "step": 902000 }, { "epoch": 2.4, "learning_rate": 2.74375e-05, "loss": 1.9773, "step": 902500 }, { "epoch": 2.4, "learning_rate": 2.7425e-05, "loss": 1.9741, "step": 903000 }, { "epoch": 2.41, "learning_rate": 2.7412500000000002e-05, "loss": 1.9732, "step": 903500 }, { "epoch": 2.41, "learning_rate": 2.7400000000000002e-05, "loss": 1.9725, "step": 904000 }, { "epoch": 2.41, "learning_rate": 2.7387499999999998e-05, "loss": 1.9635, "step": 904500 }, { "epoch": 2.41, "learning_rate": 2.7375e-05, "loss": 1.9718, "step": 905000 }, { "epoch": 2.41, "learning_rate": 2.73625e-05, "loss": 1.9734, "step": 905500 }, { "epoch": 2.41, "learning_rate": 2.7350000000000004e-05, "loss": 1.9806, "step": 906000 }, { "epoch": 2.41, "learning_rate": 2.73375e-05, "loss": 1.9718, "step": 906500 }, { "epoch": 2.41, "learning_rate": 2.7325e-05, "loss": 1.9695, "step": 907000 }, { "epoch": 2.42, "learning_rate": 2.7312500000000003e-05, "loss": 1.9684, "step": 907500 }, { "epoch": 2.42, "learning_rate": 2.7300000000000003e-05, "loss": 1.963, "step": 908000 }, { "epoch": 2.42, "learning_rate": 2.72875e-05, "loss": 1.9645, "step": 908500 }, { "epoch": 2.42, "learning_rate": 2.7275e-05, "loss": 1.9482, "step": 909000 }, { "epoch": 2.42, "learning_rate": 2.7262500000000002e-05, "loss": 1.955, "step": 909500 }, { "epoch": 2.42, "learning_rate": 2.725e-05, "loss": 1.9588, "step": 910000 }, { "epoch": 2.42, "learning_rate": 2.7237499999999998e-05, "loss": 1.9575, "step": 910500 }, { "epoch": 2.43, "learning_rate": 2.7225e-05, "loss": 1.9575, "step": 911000 }, { "epoch": 2.43, "learning_rate": 2.72125e-05, "loss": 1.958, "step": 911500 }, { "epoch": 2.43, "learning_rate": 2.7200000000000004e-05, "loss": 1.9621, "step": 912000 }, { "epoch": 2.43, "learning_rate": 2.71875e-05, "loss": 1.9665, "step": 912500 }, { "epoch": 2.43, "learning_rate": 2.7175e-05, "loss": 1.9651, "step": 913000 }, { "epoch": 2.43, "learning_rate": 2.7162500000000003e-05, "loss": 1.9685, "step": 913500 }, { "epoch": 2.43, "learning_rate": 2.7150000000000003e-05, "loss": 1.9586, "step": 914000 }, { "epoch": 2.43, "learning_rate": 2.71375e-05, "loss": 1.976, "step": 914500 }, { "epoch": 2.44, "learning_rate": 2.7125000000000002e-05, "loss": 1.9631, "step": 915000 }, { "epoch": 2.44, "learning_rate": 2.7112500000000002e-05, "loss": 1.97, "step": 915500 }, { "epoch": 2.44, "learning_rate": 2.7100000000000005e-05, "loss": 1.9551, "step": 916000 }, { "epoch": 2.44, "learning_rate": 2.7087499999999998e-05, "loss": 1.9627, "step": 916500 }, { "epoch": 2.44, "learning_rate": 2.7075e-05, "loss": 1.9478, "step": 917000 }, { "epoch": 2.44, "learning_rate": 2.70625e-05, "loss": 1.9671, "step": 917500 }, { "epoch": 2.44, "learning_rate": 2.7050000000000004e-05, "loss": 1.9559, "step": 918000 }, { "epoch": 2.45, "learning_rate": 2.70375e-05, "loss": 1.9628, "step": 918500 }, { "epoch": 2.45, "learning_rate": 2.7025e-05, "loss": 1.9529, "step": 919000 }, { "epoch": 2.45, "learning_rate": 2.7012500000000003e-05, "loss": 1.9446, "step": 919500 }, { "epoch": 2.45, "learning_rate": 2.7000000000000002e-05, "loss": 1.9533, "step": 920000 }, { "epoch": 2.45, "learning_rate": 2.69875e-05, "loss": 1.961, "step": 920500 }, { "epoch": 2.45, "learning_rate": 2.6975000000000002e-05, "loss": 1.9515, "step": 921000 }, { "epoch": 2.45, "learning_rate": 2.69625e-05, "loss": 1.9652, "step": 921500 }, { "epoch": 2.45, "learning_rate": 2.6950000000000005e-05, "loss": 1.9686, "step": 922000 }, { "epoch": 2.46, "learning_rate": 2.6937499999999997e-05, "loss": 1.9636, "step": 922500 }, { "epoch": 2.46, "learning_rate": 2.6925e-05, "loss": 1.9433, "step": 923000 }, { "epoch": 2.46, "learning_rate": 2.69125e-05, "loss": 1.9508, "step": 923500 }, { "epoch": 2.46, "learning_rate": 2.6900000000000003e-05, "loss": 1.9651, "step": 924000 }, { "epoch": 2.46, "learning_rate": 2.68875e-05, "loss": 1.9667, "step": 924500 }, { "epoch": 2.46, "learning_rate": 2.6875e-05, "loss": 1.9643, "step": 925000 }, { "epoch": 2.46, "learning_rate": 2.6862500000000002e-05, "loss": 1.9501, "step": 925500 }, { "epoch": 2.47, "learning_rate": 2.6850000000000002e-05, "loss": 1.9739, "step": 926000 }, { "epoch": 2.47, "learning_rate": 2.68375e-05, "loss": 1.9451, "step": 926500 }, { "epoch": 2.47, "learning_rate": 2.6825e-05, "loss": 1.9573, "step": 927000 }, { "epoch": 2.47, "learning_rate": 2.68125e-05, "loss": 1.963, "step": 927500 }, { "epoch": 2.47, "learning_rate": 2.6800000000000004e-05, "loss": 1.9453, "step": 928000 }, { "epoch": 2.47, "learning_rate": 2.6787499999999997e-05, "loss": 1.9418, "step": 928500 }, { "epoch": 2.47, "learning_rate": 2.6775e-05, "loss": 1.9541, "step": 929000 }, { "epoch": 2.47, "learning_rate": 2.67625e-05, "loss": 1.9497, "step": 929500 }, { "epoch": 2.48, "learning_rate": 2.6750000000000003e-05, "loss": 1.9723, "step": 930000 }, { "epoch": 2.48, "learning_rate": 2.67375e-05, "loss": 1.9536, "step": 930500 }, { "epoch": 2.48, "learning_rate": 2.6725e-05, "loss": 1.955, "step": 931000 }, { "epoch": 2.48, "learning_rate": 2.6712500000000002e-05, "loss": 1.9461, "step": 931500 }, { "epoch": 2.48, "learning_rate": 2.6700000000000002e-05, "loss": 1.9491, "step": 932000 }, { "epoch": 2.48, "learning_rate": 2.6687499999999998e-05, "loss": 1.9497, "step": 932500 }, { "epoch": 2.48, "learning_rate": 2.6675e-05, "loss": 1.9586, "step": 933000 }, { "epoch": 2.49, "learning_rate": 2.66625e-05, "loss": 1.9692, "step": 933500 }, { "epoch": 2.49, "learning_rate": 2.6650000000000004e-05, "loss": 1.9614, "step": 934000 }, { "epoch": 2.49, "learning_rate": 2.6637499999999997e-05, "loss": 1.9657, "step": 934500 }, { "epoch": 2.49, "learning_rate": 2.6625e-05, "loss": 1.937, "step": 935000 }, { "epoch": 2.49, "learning_rate": 2.66125e-05, "loss": 1.9544, "step": 935500 }, { "epoch": 2.49, "learning_rate": 2.6600000000000003e-05, "loss": 1.9653, "step": 936000 }, { "epoch": 2.49, "learning_rate": 2.65875e-05, "loss": 1.9489, "step": 936500 }, { "epoch": 2.49, "learning_rate": 2.6575e-05, "loss": 1.9456, "step": 937000 }, { "epoch": 2.5, "learning_rate": 2.6562500000000002e-05, "loss": 1.9587, "step": 937500 }, { "epoch": 2.5, "learning_rate": 2.655e-05, "loss": 1.9507, "step": 938000 }, { "epoch": 2.5, "learning_rate": 2.6537500000000005e-05, "loss": 1.9427, "step": 938500 }, { "epoch": 2.5, "learning_rate": 2.6525e-05, "loss": 1.9411, "step": 939000 }, { "epoch": 2.5, "learning_rate": 2.65125e-05, "loss": 1.9547, "step": 939500 }, { "epoch": 2.5, "learning_rate": 2.6500000000000004e-05, "loss": 1.9456, "step": 940000 }, { "epoch": 2.5, "learning_rate": 2.6487500000000003e-05, "loss": 1.9585, "step": 940500 }, { "epoch": 2.51, "learning_rate": 2.6475e-05, "loss": 1.9341, "step": 941000 }, { "epoch": 2.51, "learning_rate": 2.6462500000000003e-05, "loss": 1.9379, "step": 941500 }, { "epoch": 2.51, "learning_rate": 2.6450000000000003e-05, "loss": 1.9584, "step": 942000 }, { "epoch": 2.51, "learning_rate": 2.6437500000000002e-05, "loss": 1.9395, "step": 942500 }, { "epoch": 2.51, "learning_rate": 2.6425e-05, "loss": 1.9489, "step": 943000 }, { "epoch": 2.51, "learning_rate": 2.64125e-05, "loss": 1.9449, "step": 943500 }, { "epoch": 2.51, "learning_rate": 2.64e-05, "loss": 1.9605, "step": 944000 }, { "epoch": 2.51, "learning_rate": 2.6387500000000004e-05, "loss": 1.9447, "step": 944500 }, { "epoch": 2.52, "learning_rate": 2.6375e-05, "loss": 1.9541, "step": 945000 }, { "epoch": 2.52, "learning_rate": 2.63625e-05, "loss": 1.9293, "step": 945500 }, { "epoch": 2.52, "learning_rate": 2.6350000000000004e-05, "loss": 1.9389, "step": 946000 }, { "epoch": 2.52, "learning_rate": 2.6337500000000003e-05, "loss": 1.9539, "step": 946500 }, { "epoch": 2.52, "learning_rate": 2.6325e-05, "loss": 1.9496, "step": 947000 }, { "epoch": 2.52, "learning_rate": 2.6312500000000003e-05, "loss": 1.9532, "step": 947500 }, { "epoch": 2.52, "learning_rate": 2.6300000000000002e-05, "loss": 1.9367, "step": 948000 }, { "epoch": 2.53, "learning_rate": 2.6287500000000005e-05, "loss": 1.933, "step": 948500 }, { "epoch": 2.53, "learning_rate": 2.6275e-05, "loss": 1.9556, "step": 949000 }, { "epoch": 2.53, "learning_rate": 2.62625e-05, "loss": 1.964, "step": 949500 }, { "epoch": 2.53, "learning_rate": 2.625e-05, "loss": 1.9493, "step": 950000 }, { "epoch": 2.53, "learning_rate": 2.6237500000000004e-05, "loss": 1.9413, "step": 950500 }, { "epoch": 2.53, "learning_rate": 2.6225e-05, "loss": 1.9471, "step": 951000 }, { "epoch": 2.53, "learning_rate": 2.62125e-05, "loss": 1.9354, "step": 951500 }, { "epoch": 2.53, "learning_rate": 2.6200000000000003e-05, "loss": 1.9373, "step": 952000 }, { "epoch": 2.54, "learning_rate": 2.6187500000000003e-05, "loss": 1.9338, "step": 952500 }, { "epoch": 2.54, "learning_rate": 2.6175e-05, "loss": 1.9402, "step": 953000 }, { "epoch": 2.54, "learning_rate": 2.6162500000000002e-05, "loss": 1.944, "step": 953500 }, { "epoch": 2.54, "learning_rate": 2.6150000000000002e-05, "loss": 1.9404, "step": 954000 }, { "epoch": 2.54, "learning_rate": 2.6137500000000005e-05, "loss": 1.9359, "step": 954500 }, { "epoch": 2.54, "learning_rate": 2.6124999999999998e-05, "loss": 1.9456, "step": 955000 }, { "epoch": 2.54, "learning_rate": 2.61125e-05, "loss": 1.9237, "step": 955500 }, { "epoch": 2.55, "learning_rate": 2.61e-05, "loss": 1.9473, "step": 956000 }, { "epoch": 2.55, "learning_rate": 2.6087500000000004e-05, "loss": 1.9399, "step": 956500 }, { "epoch": 2.55, "learning_rate": 2.6075e-05, "loss": 1.9506, "step": 957000 }, { "epoch": 2.55, "learning_rate": 2.60625e-05, "loss": 1.9335, "step": 957500 }, { "epoch": 2.55, "learning_rate": 2.6050000000000003e-05, "loss": 1.9412, "step": 958000 }, { "epoch": 2.55, "learning_rate": 2.6037500000000003e-05, "loss": 1.9526, "step": 958500 }, { "epoch": 2.55, "learning_rate": 2.6025e-05, "loss": 1.9446, "step": 959000 }, { "epoch": 2.55, "learning_rate": 2.6012500000000002e-05, "loss": 1.9496, "step": 959500 }, { "epoch": 2.56, "learning_rate": 2.6000000000000002e-05, "loss": 1.9269, "step": 960000 }, { "epoch": 2.56, "learning_rate": 2.5987500000000005e-05, "loss": 1.9418, "step": 960500 }, { "epoch": 2.56, "learning_rate": 2.5974999999999998e-05, "loss": 1.9535, "step": 961000 }, { "epoch": 2.56, "learning_rate": 2.59625e-05, "loss": 1.9535, "step": 961500 }, { "epoch": 2.56, "learning_rate": 2.595e-05, "loss": 1.9416, "step": 962000 }, { "epoch": 2.56, "learning_rate": 2.5937500000000004e-05, "loss": 1.9317, "step": 962500 }, { "epoch": 2.56, "learning_rate": 2.5925e-05, "loss": 1.9312, "step": 963000 }, { "epoch": 2.57, "learning_rate": 2.59125e-05, "loss": 1.9278, "step": 963500 }, { "epoch": 2.57, "learning_rate": 2.5900000000000003e-05, "loss": 1.9408, "step": 964000 }, { "epoch": 2.57, "learning_rate": 2.5887500000000002e-05, "loss": 1.9356, "step": 964500 }, { "epoch": 2.57, "learning_rate": 2.5875e-05, "loss": 1.9404, "step": 965000 }, { "epoch": 2.57, "learning_rate": 2.5862500000000002e-05, "loss": 1.9379, "step": 965500 }, { "epoch": 2.57, "learning_rate": 2.585e-05, "loss": 1.9447, "step": 966000 }, { "epoch": 2.57, "learning_rate": 2.5837500000000005e-05, "loss": 1.9338, "step": 966500 }, { "epoch": 2.57, "learning_rate": 2.5824999999999998e-05, "loss": 1.921, "step": 967000 }, { "epoch": 2.58, "learning_rate": 2.58125e-05, "loss": 1.9374, "step": 967500 }, { "epoch": 2.58, "learning_rate": 2.58e-05, "loss": 1.9359, "step": 968000 }, { "epoch": 2.58, "learning_rate": 2.5787500000000003e-05, "loss": 1.925, "step": 968500 }, { "epoch": 2.58, "learning_rate": 2.5775e-05, "loss": 1.9335, "step": 969000 }, { "epoch": 2.58, "learning_rate": 2.57625e-05, "loss": 1.9419, "step": 969500 }, { "epoch": 2.58, "learning_rate": 2.5750000000000002e-05, "loss": 1.9505, "step": 970000 }, { "epoch": 2.58, "learning_rate": 2.5737500000000002e-05, "loss": 1.9336, "step": 970500 }, { "epoch": 2.59, "learning_rate": 2.5725e-05, "loss": 1.9366, "step": 971000 }, { "epoch": 2.59, "learning_rate": 2.57125e-05, "loss": 1.9371, "step": 971500 }, { "epoch": 2.59, "learning_rate": 2.57e-05, "loss": 1.9311, "step": 972000 }, { "epoch": 2.59, "learning_rate": 2.5687500000000004e-05, "loss": 1.9456, "step": 972500 }, { "epoch": 2.59, "learning_rate": 2.5675e-05, "loss": 1.9207, "step": 973000 }, { "epoch": 2.59, "learning_rate": 2.56625e-05, "loss": 1.9307, "step": 973500 }, { "epoch": 2.59, "learning_rate": 2.5650000000000003e-05, "loss": 1.9541, "step": 974000 }, { "epoch": 2.59, "learning_rate": 2.5637500000000003e-05, "loss": 1.9232, "step": 974500 }, { "epoch": 2.6, "learning_rate": 2.5625e-05, "loss": 1.9332, "step": 975000 }, { "epoch": 2.6, "learning_rate": 2.56125e-05, "loss": 1.9365, "step": 975500 }, { "epoch": 2.6, "learning_rate": 2.5600000000000002e-05, "loss": 1.9282, "step": 976000 }, { "epoch": 2.6, "learning_rate": 2.5587500000000002e-05, "loss": 1.9381, "step": 976500 }, { "epoch": 2.6, "learning_rate": 2.5574999999999998e-05, "loss": 1.933, "step": 977000 }, { "epoch": 2.6, "learning_rate": 2.55625e-05, "loss": 1.921, "step": 977500 }, { "epoch": 2.6, "learning_rate": 2.555e-05, "loss": 1.9415, "step": 978000 }, { "epoch": 2.61, "learning_rate": 2.5537500000000004e-05, "loss": 1.9431, "step": 978500 }, { "epoch": 2.61, "learning_rate": 2.5525e-05, "loss": 1.9428, "step": 979000 }, { "epoch": 2.61, "learning_rate": 2.55125e-05, "loss": 1.9368, "step": 979500 }, { "epoch": 2.61, "learning_rate": 2.5500000000000003e-05, "loss": 1.928, "step": 980000 }, { "epoch": 2.61, "learning_rate": 2.5487500000000003e-05, "loss": 1.9368, "step": 980500 }, { "epoch": 2.61, "learning_rate": 2.5475e-05, "loss": 1.9389, "step": 981000 }, { "epoch": 2.61, "learning_rate": 2.54625e-05, "loss": 1.9316, "step": 981500 }, { "epoch": 2.61, "learning_rate": 2.5450000000000002e-05, "loss": 1.9403, "step": 982000 }, { "epoch": 2.62, "learning_rate": 2.54375e-05, "loss": 1.9338, "step": 982500 }, { "epoch": 2.62, "learning_rate": 2.5424999999999998e-05, "loss": 1.9464, "step": 983000 }, { "epoch": 2.62, "learning_rate": 2.54125e-05, "loss": 1.9373, "step": 983500 }, { "epoch": 2.62, "learning_rate": 2.54e-05, "loss": 1.9304, "step": 984000 }, { "epoch": 2.62, "learning_rate": 2.5387500000000004e-05, "loss": 1.9335, "step": 984500 }, { "epoch": 2.62, "learning_rate": 2.5375e-05, "loss": 1.9353, "step": 985000 }, { "epoch": 2.62, "learning_rate": 2.53625e-05, "loss": 1.9429, "step": 985500 }, { "epoch": 2.62, "learning_rate": 2.5350000000000003e-05, "loss": 1.9258, "step": 986000 }, { "epoch": 2.63, "learning_rate": 2.5337500000000003e-05, "loss": 1.9201, "step": 986500 }, { "epoch": 2.63, "learning_rate": 2.5325e-05, "loss": 1.9309, "step": 987000 }, { "epoch": 2.63, "learning_rate": 2.53125e-05, "loss": 1.9173, "step": 987500 }, { "epoch": 2.63, "learning_rate": 2.5300000000000002e-05, "loss": 1.9226, "step": 988000 }, { "epoch": 2.63, "learning_rate": 2.52875e-05, "loss": 1.9395, "step": 988500 }, { "epoch": 2.63, "learning_rate": 2.5274999999999998e-05, "loss": 1.9383, "step": 989000 }, { "epoch": 2.63, "learning_rate": 2.52625e-05, "loss": 1.9229, "step": 989500 }, { "epoch": 2.64, "learning_rate": 2.525e-05, "loss": 1.9185, "step": 990000 }, { "epoch": 2.64, "learning_rate": 2.5237500000000004e-05, "loss": 1.9286, "step": 990500 }, { "epoch": 2.64, "learning_rate": 2.5225e-05, "loss": 1.9384, "step": 991000 }, { "epoch": 2.64, "learning_rate": 2.52125e-05, "loss": 1.9153, "step": 991500 }, { "epoch": 2.64, "learning_rate": 2.5200000000000003e-05, "loss": 1.9267, "step": 992000 }, { "epoch": 2.64, "learning_rate": 2.5187500000000002e-05, "loss": 1.9232, "step": 992500 }, { "epoch": 2.64, "learning_rate": 2.5175e-05, "loss": 1.9378, "step": 993000 }, { "epoch": 2.64, "learning_rate": 2.51625e-05, "loss": 1.9348, "step": 993500 }, { "epoch": 2.65, "learning_rate": 2.515e-05, "loss": 1.9364, "step": 994000 }, { "epoch": 2.65, "learning_rate": 2.51375e-05, "loss": 1.9218, "step": 994500 }, { "epoch": 2.65, "learning_rate": 2.5124999999999997e-05, "loss": 1.9303, "step": 995000 }, { "epoch": 2.65, "learning_rate": 2.51125e-05, "loss": 1.9125, "step": 995500 }, { "epoch": 2.65, "learning_rate": 2.51e-05, "loss": 1.9196, "step": 996000 }, { "epoch": 2.65, "learning_rate": 2.5087500000000003e-05, "loss": 1.9221, "step": 996500 }, { "epoch": 2.65, "learning_rate": 2.5075e-05, "loss": 1.9291, "step": 997000 }, { "epoch": 2.66, "learning_rate": 2.50625e-05, "loss": 1.9345, "step": 997500 }, { "epoch": 2.66, "learning_rate": 2.5050000000000002e-05, "loss": 1.9168, "step": 998000 }, { "epoch": 2.66, "learning_rate": 2.5037500000000002e-05, "loss": 1.9151, "step": 998500 }, { "epoch": 2.66, "learning_rate": 2.5025e-05, "loss": 1.9174, "step": 999000 }, { "epoch": 2.66, "learning_rate": 2.50125e-05, "loss": 1.925, "step": 999500 }, { "epoch": 2.66, "learning_rate": 2.5e-05, "loss": 1.9265, "step": 1000000 }, { "epoch": 2.66, "learning_rate": 2.49875e-05, "loss": 1.9373, "step": 1000500 }, { "epoch": 2.66, "learning_rate": 2.4975e-05, "loss": 1.9233, "step": 1001000 }, { "epoch": 2.67, "learning_rate": 2.4962500000000004e-05, "loss": 1.9327, "step": 1001500 }, { "epoch": 2.67, "learning_rate": 2.495e-05, "loss": 1.9402, "step": 1002000 }, { "epoch": 2.67, "learning_rate": 2.4937500000000003e-05, "loss": 1.916, "step": 1002500 }, { "epoch": 2.67, "learning_rate": 2.4925000000000003e-05, "loss": 1.9159, "step": 1003000 }, { "epoch": 2.67, "learning_rate": 2.4912500000000002e-05, "loss": 1.9213, "step": 1003500 }, { "epoch": 2.67, "learning_rate": 2.4900000000000002e-05, "loss": 1.9265, "step": 1004000 }, { "epoch": 2.67, "learning_rate": 2.4887500000000002e-05, "loss": 1.9324, "step": 1004500 }, { "epoch": 2.68, "learning_rate": 2.4875e-05, "loss": 1.9218, "step": 1005000 }, { "epoch": 2.68, "learning_rate": 2.48625e-05, "loss": 1.9137, "step": 1005500 }, { "epoch": 2.68, "learning_rate": 2.485e-05, "loss": 1.9139, "step": 1006000 }, { "epoch": 2.68, "learning_rate": 2.4837500000000004e-05, "loss": 1.9323, "step": 1006500 }, { "epoch": 2.68, "learning_rate": 2.4825e-05, "loss": 1.9128, "step": 1007000 }, { "epoch": 2.68, "learning_rate": 2.4812500000000003e-05, "loss": 1.9374, "step": 1007500 }, { "epoch": 2.68, "learning_rate": 2.48e-05, "loss": 1.9355, "step": 1008000 }, { "epoch": 2.68, "learning_rate": 2.4787500000000003e-05, "loss": 1.9166, "step": 1008500 }, { "epoch": 2.69, "learning_rate": 2.4775000000000003e-05, "loss": 1.9133, "step": 1009000 }, { "epoch": 2.69, "learning_rate": 2.4762500000000002e-05, "loss": 1.9175, "step": 1009500 }, { "epoch": 2.69, "learning_rate": 2.4750000000000002e-05, "loss": 1.9209, "step": 1010000 }, { "epoch": 2.69, "learning_rate": 2.47375e-05, "loss": 1.9262, "step": 1010500 }, { "epoch": 2.69, "learning_rate": 2.4725e-05, "loss": 1.9189, "step": 1011000 }, { "epoch": 2.69, "learning_rate": 2.47125e-05, "loss": 1.9129, "step": 1011500 }, { "epoch": 2.69, "learning_rate": 2.47e-05, "loss": 1.9324, "step": 1012000 }, { "epoch": 2.7, "learning_rate": 2.4687500000000004e-05, "loss": 1.9134, "step": 1012500 }, { "epoch": 2.7, "learning_rate": 2.4675e-05, "loss": 1.92, "step": 1013000 }, { "epoch": 2.7, "learning_rate": 2.4662500000000003e-05, "loss": 1.9108, "step": 1013500 }, { "epoch": 2.7, "learning_rate": 2.465e-05, "loss": 1.9152, "step": 1014000 }, { "epoch": 2.7, "learning_rate": 2.4637500000000003e-05, "loss": 1.902, "step": 1014500 }, { "epoch": 2.7, "learning_rate": 2.4625000000000002e-05, "loss": 1.9175, "step": 1015000 }, { "epoch": 2.7, "learning_rate": 2.4612500000000002e-05, "loss": 1.9163, "step": 1015500 }, { "epoch": 2.7, "learning_rate": 2.46e-05, "loss": 1.9246, "step": 1016000 }, { "epoch": 2.71, "learning_rate": 2.45875e-05, "loss": 1.9213, "step": 1016500 }, { "epoch": 2.71, "learning_rate": 2.4575e-05, "loss": 1.9234, "step": 1017000 }, { "epoch": 2.71, "learning_rate": 2.45625e-05, "loss": 1.9167, "step": 1017500 }, { "epoch": 2.71, "learning_rate": 2.455e-05, "loss": 1.9103, "step": 1018000 }, { "epoch": 2.71, "learning_rate": 2.4537500000000004e-05, "loss": 1.9203, "step": 1018500 }, { "epoch": 2.71, "learning_rate": 2.4525e-05, "loss": 1.8977, "step": 1019000 }, { "epoch": 2.71, "learning_rate": 2.4512500000000003e-05, "loss": 1.9212, "step": 1019500 }, { "epoch": 2.72, "learning_rate": 2.45e-05, "loss": 1.9329, "step": 1020000 }, { "epoch": 2.72, "learning_rate": 2.4487500000000002e-05, "loss": 1.9209, "step": 1020500 }, { "epoch": 2.72, "learning_rate": 2.4475000000000002e-05, "loss": 1.9155, "step": 1021000 }, { "epoch": 2.72, "learning_rate": 2.44625e-05, "loss": 1.9294, "step": 1021500 }, { "epoch": 2.72, "learning_rate": 2.445e-05, "loss": 1.93, "step": 1022000 }, { "epoch": 2.72, "learning_rate": 2.44375e-05, "loss": 1.9323, "step": 1022500 }, { "epoch": 2.72, "learning_rate": 2.4425e-05, "loss": 1.9137, "step": 1023000 }, { "epoch": 2.72, "learning_rate": 2.44125e-05, "loss": 1.9108, "step": 1023500 }, { "epoch": 2.73, "learning_rate": 2.44e-05, "loss": 1.9097, "step": 1024000 }, { "epoch": 2.73, "learning_rate": 2.4387500000000003e-05, "loss": 1.9132, "step": 1024500 }, { "epoch": 2.73, "learning_rate": 2.4375e-05, "loss": 1.9087, "step": 1025000 }, { "epoch": 2.73, "learning_rate": 2.4362500000000003e-05, "loss": 1.9213, "step": 1025500 }, { "epoch": 2.73, "learning_rate": 2.435e-05, "loss": 1.9189, "step": 1026000 }, { "epoch": 2.73, "learning_rate": 2.4337500000000002e-05, "loss": 1.9086, "step": 1026500 }, { "epoch": 2.73, "learning_rate": 2.4325000000000002e-05, "loss": 1.9094, "step": 1027000 }, { "epoch": 2.74, "learning_rate": 2.43125e-05, "loss": 1.9161, "step": 1027500 }, { "epoch": 2.74, "learning_rate": 2.43e-05, "loss": 1.8975, "step": 1028000 }, { "epoch": 2.74, "learning_rate": 2.42875e-05, "loss": 1.9108, "step": 1028500 }, { "epoch": 2.74, "learning_rate": 2.4275e-05, "loss": 1.9094, "step": 1029000 }, { "epoch": 2.74, "learning_rate": 2.42625e-05, "loss": 1.9126, "step": 1029500 }, { "epoch": 2.74, "learning_rate": 2.425e-05, "loss": 1.9027, "step": 1030000 }, { "epoch": 2.74, "learning_rate": 2.4237500000000003e-05, "loss": 1.9028, "step": 1030500 }, { "epoch": 2.74, "learning_rate": 2.4225e-05, "loss": 1.9223, "step": 1031000 }, { "epoch": 2.75, "learning_rate": 2.4212500000000002e-05, "loss": 1.9152, "step": 1031500 }, { "epoch": 2.75, "learning_rate": 2.4200000000000002e-05, "loss": 1.8914, "step": 1032000 }, { "epoch": 2.75, "learning_rate": 2.4187500000000002e-05, "loss": 1.9336, "step": 1032500 }, { "epoch": 2.75, "learning_rate": 2.4175e-05, "loss": 1.9023, "step": 1033000 }, { "epoch": 2.75, "learning_rate": 2.41625e-05, "loss": 1.9035, "step": 1033500 }, { "epoch": 2.75, "learning_rate": 2.415e-05, "loss": 1.8964, "step": 1034000 }, { "epoch": 2.75, "learning_rate": 2.41375e-05, "loss": 1.903, "step": 1034500 }, { "epoch": 2.76, "learning_rate": 2.4125e-05, "loss": 1.8989, "step": 1035000 }, { "epoch": 2.76, "learning_rate": 2.4112500000000003e-05, "loss": 1.9157, "step": 1035500 }, { "epoch": 2.76, "learning_rate": 2.41e-05, "loss": 1.9077, "step": 1036000 }, { "epoch": 2.76, "learning_rate": 2.4087500000000003e-05, "loss": 1.9043, "step": 1036500 }, { "epoch": 2.76, "learning_rate": 2.4075e-05, "loss": 1.9, "step": 1037000 }, { "epoch": 2.76, "learning_rate": 2.4062500000000002e-05, "loss": 1.919, "step": 1037500 }, { "epoch": 2.76, "learning_rate": 2.4050000000000002e-05, "loss": 1.9007, "step": 1038000 }, { "epoch": 2.76, "learning_rate": 2.40375e-05, "loss": 1.907, "step": 1038500 }, { "epoch": 2.77, "learning_rate": 2.4025e-05, "loss": 1.9105, "step": 1039000 }, { "epoch": 2.77, "learning_rate": 2.40125e-05, "loss": 1.9117, "step": 1039500 }, { "epoch": 2.77, "learning_rate": 2.4e-05, "loss": 1.9162, "step": 1040000 }, { "epoch": 2.77, "learning_rate": 2.39875e-05, "loss": 1.9195, "step": 1040500 }, { "epoch": 2.77, "learning_rate": 2.3975e-05, "loss": 1.9028, "step": 1041000 }, { "epoch": 2.77, "learning_rate": 2.3962500000000003e-05, "loss": 1.9155, "step": 1041500 }, { "epoch": 2.77, "learning_rate": 2.395e-05, "loss": 1.9009, "step": 1042000 }, { "epoch": 2.78, "learning_rate": 2.3937500000000002e-05, "loss": 1.8987, "step": 1042500 }, { "epoch": 2.78, "learning_rate": 2.3925e-05, "loss": 1.9136, "step": 1043000 }, { "epoch": 2.78, "learning_rate": 2.3912500000000002e-05, "loss": 1.9127, "step": 1043500 }, { "epoch": 2.78, "learning_rate": 2.39e-05, "loss": 1.9128, "step": 1044000 }, { "epoch": 2.78, "learning_rate": 2.38875e-05, "loss": 1.9133, "step": 1044500 }, { "epoch": 2.78, "learning_rate": 2.3875e-05, "loss": 1.9132, "step": 1045000 }, { "epoch": 2.78, "learning_rate": 2.38625e-05, "loss": 1.9178, "step": 1045500 }, { "epoch": 2.78, "learning_rate": 2.385e-05, "loss": 1.91, "step": 1046000 }, { "epoch": 2.79, "learning_rate": 2.38375e-05, "loss": 1.9076, "step": 1046500 }, { "epoch": 2.79, "learning_rate": 2.3825e-05, "loss": 1.9175, "step": 1047000 }, { "epoch": 2.79, "learning_rate": 2.3812500000000003e-05, "loss": 1.8945, "step": 1047500 }, { "epoch": 2.79, "learning_rate": 2.38e-05, "loss": 1.9071, "step": 1048000 }, { "epoch": 2.79, "learning_rate": 2.3787500000000002e-05, "loss": 1.9025, "step": 1048500 }, { "epoch": 2.79, "learning_rate": 2.3775e-05, "loss": 1.909, "step": 1049000 }, { "epoch": 2.79, "learning_rate": 2.37625e-05, "loss": 1.9125, "step": 1049500 }, { "epoch": 2.8, "learning_rate": 2.375e-05, "loss": 1.9126, "step": 1050000 }, { "epoch": 2.8, "learning_rate": 2.37375e-05, "loss": 1.8892, "step": 1050500 }, { "epoch": 2.8, "learning_rate": 2.3725e-05, "loss": 1.9134, "step": 1051000 }, { "epoch": 2.8, "learning_rate": 2.37125e-05, "loss": 1.9049, "step": 1051500 }, { "epoch": 2.8, "learning_rate": 2.37e-05, "loss": 1.9131, "step": 1052000 }, { "epoch": 2.8, "learning_rate": 2.36875e-05, "loss": 1.9035, "step": 1052500 }, { "epoch": 2.8, "learning_rate": 2.3675e-05, "loss": 1.8969, "step": 1053000 }, { "epoch": 2.8, "learning_rate": 2.3662500000000003e-05, "loss": 1.9089, "step": 1053500 }, { "epoch": 2.81, "learning_rate": 2.365e-05, "loss": 1.8875, "step": 1054000 }, { "epoch": 2.81, "learning_rate": 2.3637500000000002e-05, "loss": 1.9175, "step": 1054500 }, { "epoch": 2.81, "learning_rate": 2.3624999999999998e-05, "loss": 1.8957, "step": 1055000 }, { "epoch": 2.81, "learning_rate": 2.36125e-05, "loss": 1.917, "step": 1055500 }, { "epoch": 2.81, "learning_rate": 2.36e-05, "loss": 1.9144, "step": 1056000 }, { "epoch": 2.81, "learning_rate": 2.35875e-05, "loss": 1.8875, "step": 1056500 }, { "epoch": 2.81, "learning_rate": 2.3575e-05, "loss": 1.9239, "step": 1057000 }, { "epoch": 2.82, "learning_rate": 2.35625e-05, "loss": 1.9149, "step": 1057500 }, { "epoch": 2.82, "learning_rate": 2.355e-05, "loss": 1.9146, "step": 1058000 }, { "epoch": 2.82, "learning_rate": 2.35375e-05, "loss": 1.8884, "step": 1058500 }, { "epoch": 2.82, "learning_rate": 2.3525e-05, "loss": 1.8961, "step": 1059000 }, { "epoch": 2.82, "learning_rate": 2.3512500000000002e-05, "loss": 1.8946, "step": 1059500 }, { "epoch": 2.82, "learning_rate": 2.35e-05, "loss": 1.8962, "step": 1060000 }, { "epoch": 2.82, "learning_rate": 2.3487500000000002e-05, "loss": 1.8941, "step": 1060500 }, { "epoch": 2.82, "learning_rate": 2.3475e-05, "loss": 1.909, "step": 1061000 }, { "epoch": 2.83, "learning_rate": 2.34625e-05, "loss": 1.9036, "step": 1061500 }, { "epoch": 2.83, "learning_rate": 2.345e-05, "loss": 1.8984, "step": 1062000 }, { "epoch": 2.83, "learning_rate": 2.34375e-05, "loss": 1.9166, "step": 1062500 }, { "epoch": 2.83, "learning_rate": 2.3425000000000004e-05, "loss": 1.8974, "step": 1063000 }, { "epoch": 2.83, "learning_rate": 2.34125e-05, "loss": 1.9083, "step": 1063500 }, { "epoch": 2.83, "learning_rate": 2.3400000000000003e-05, "loss": 1.8869, "step": 1064000 }, { "epoch": 2.83, "learning_rate": 2.3387500000000003e-05, "loss": 1.8933, "step": 1064500 }, { "epoch": 2.84, "learning_rate": 2.3375000000000002e-05, "loss": 1.9023, "step": 1065000 }, { "epoch": 2.84, "learning_rate": 2.3362500000000002e-05, "loss": 1.8944, "step": 1065500 }, { "epoch": 2.84, "learning_rate": 2.3350000000000002e-05, "loss": 1.8861, "step": 1066000 }, { "epoch": 2.84, "learning_rate": 2.33375e-05, "loss": 1.8877, "step": 1066500 }, { "epoch": 2.84, "learning_rate": 2.3325e-05, "loss": 1.8976, "step": 1067000 }, { "epoch": 2.84, "learning_rate": 2.33125e-05, "loss": 1.8935, "step": 1067500 }, { "epoch": 2.84, "learning_rate": 2.3300000000000004e-05, "loss": 1.8937, "step": 1068000 }, { "epoch": 2.84, "learning_rate": 2.32875e-05, "loss": 1.9043, "step": 1068500 }, { "epoch": 2.85, "learning_rate": 2.3275000000000003e-05, "loss": 1.8836, "step": 1069000 }, { "epoch": 2.85, "learning_rate": 2.32625e-05, "loss": 1.8981, "step": 1069500 }, { "epoch": 2.85, "learning_rate": 2.3250000000000003e-05, "loss": 1.8854, "step": 1070000 }, { "epoch": 2.85, "learning_rate": 2.3237500000000002e-05, "loss": 1.8828, "step": 1070500 }, { "epoch": 2.85, "learning_rate": 2.3225000000000002e-05, "loss": 1.9001, "step": 1071000 }, { "epoch": 2.85, "learning_rate": 2.3212500000000002e-05, "loss": 1.9018, "step": 1071500 }, { "epoch": 2.85, "learning_rate": 2.32e-05, "loss": 1.8983, "step": 1072000 }, { "epoch": 2.86, "learning_rate": 2.31875e-05, "loss": 1.9087, "step": 1072500 }, { "epoch": 2.86, "learning_rate": 2.3175e-05, "loss": 1.8991, "step": 1073000 }, { "epoch": 2.86, "learning_rate": 2.31625e-05, "loss": 1.9123, "step": 1073500 }, { "epoch": 2.86, "learning_rate": 2.3150000000000004e-05, "loss": 1.8935, "step": 1074000 }, { "epoch": 2.86, "learning_rate": 2.31375e-05, "loss": 1.8894, "step": 1074500 }, { "epoch": 2.86, "learning_rate": 2.3125000000000003e-05, "loss": 1.8941, "step": 1075000 }, { "epoch": 2.86, "learning_rate": 2.31125e-05, "loss": 1.9141, "step": 1075500 }, { "epoch": 2.86, "learning_rate": 2.3100000000000002e-05, "loss": 1.8939, "step": 1076000 }, { "epoch": 2.87, "learning_rate": 2.3087500000000002e-05, "loss": 1.8853, "step": 1076500 }, { "epoch": 2.87, "learning_rate": 2.3075000000000002e-05, "loss": 1.9009, "step": 1077000 }, { "epoch": 2.87, "learning_rate": 2.30625e-05, "loss": 1.8875, "step": 1077500 }, { "epoch": 2.87, "learning_rate": 2.305e-05, "loss": 1.895, "step": 1078000 }, { "epoch": 2.87, "learning_rate": 2.30375e-05, "loss": 1.8897, "step": 1078500 }, { "epoch": 2.87, "learning_rate": 2.3025e-05, "loss": 1.902, "step": 1079000 }, { "epoch": 2.87, "learning_rate": 2.30125e-05, "loss": 1.893, "step": 1079500 }, { "epoch": 2.88, "learning_rate": 2.3000000000000003e-05, "loss": 1.893, "step": 1080000 }, { "epoch": 2.88, "learning_rate": 2.29875e-05, "loss": 1.8841, "step": 1080500 }, { "epoch": 2.88, "learning_rate": 2.2975000000000003e-05, "loss": 1.9174, "step": 1081000 }, { "epoch": 2.88, "learning_rate": 2.29625e-05, "loss": 1.8956, "step": 1081500 }, { "epoch": 2.88, "learning_rate": 2.2950000000000002e-05, "loss": 1.8788, "step": 1082000 }, { "epoch": 2.88, "learning_rate": 2.2937500000000002e-05, "loss": 1.8844, "step": 1082500 }, { "epoch": 2.88, "learning_rate": 2.2925e-05, "loss": 1.8921, "step": 1083000 }, { "epoch": 2.88, "learning_rate": 2.29125e-05, "loss": 1.886, "step": 1083500 }, { "epoch": 2.89, "learning_rate": 2.29e-05, "loss": 1.8925, "step": 1084000 }, { "epoch": 2.89, "learning_rate": 2.28875e-05, "loss": 1.8881, "step": 1084500 }, { "epoch": 2.89, "learning_rate": 2.2875e-05, "loss": 1.8766, "step": 1085000 }, { "epoch": 2.89, "learning_rate": 2.28625e-05, "loss": 1.883, "step": 1085500 }, { "epoch": 2.89, "learning_rate": 2.2850000000000003e-05, "loss": 1.909, "step": 1086000 }, { "epoch": 2.89, "learning_rate": 2.28375e-05, "loss": 1.882, "step": 1086500 }, { "epoch": 2.89, "learning_rate": 2.2825000000000003e-05, "loss": 1.887, "step": 1087000 }, { "epoch": 2.9, "learning_rate": 2.28125e-05, "loss": 1.878, "step": 1087500 }, { "epoch": 2.9, "learning_rate": 2.2800000000000002e-05, "loss": 1.9032, "step": 1088000 }, { "epoch": 2.9, "learning_rate": 2.27875e-05, "loss": 1.9103, "step": 1088500 }, { "epoch": 2.9, "learning_rate": 2.2775e-05, "loss": 1.9134, "step": 1089000 }, { "epoch": 2.9, "learning_rate": 2.27625e-05, "loss": 1.8785, "step": 1089500 }, { "epoch": 2.9, "learning_rate": 2.275e-05, "loss": 1.8804, "step": 1090000 }, { "epoch": 2.9, "learning_rate": 2.27375e-05, "loss": 1.9018, "step": 1090500 }, { "epoch": 2.9, "learning_rate": 2.2725000000000003e-05, "loss": 1.89, "step": 1091000 }, { "epoch": 2.91, "learning_rate": 2.27125e-05, "loss": 1.8847, "step": 1091500 }, { "epoch": 2.91, "learning_rate": 2.2700000000000003e-05, "loss": 1.8863, "step": 1092000 }, { "epoch": 2.91, "learning_rate": 2.26875e-05, "loss": 1.8868, "step": 1092500 }, { "epoch": 2.91, "learning_rate": 2.2675000000000002e-05, "loss": 1.8903, "step": 1093000 }, { "epoch": 2.91, "learning_rate": 2.2662500000000002e-05, "loss": 1.8783, "step": 1093500 }, { "epoch": 2.91, "learning_rate": 2.265e-05, "loss": 1.8755, "step": 1094000 }, { "epoch": 2.91, "learning_rate": 2.26375e-05, "loss": 1.884, "step": 1094500 }, { "epoch": 2.92, "learning_rate": 2.2625e-05, "loss": 1.8901, "step": 1095000 }, { "epoch": 2.92, "learning_rate": 2.26125e-05, "loss": 1.8738, "step": 1095500 }, { "epoch": 2.92, "learning_rate": 2.26e-05, "loss": 1.8956, "step": 1096000 }, { "epoch": 2.92, "learning_rate": 2.25875e-05, "loss": 1.884, "step": 1096500 }, { "epoch": 2.92, "learning_rate": 2.2575000000000003e-05, "loss": 1.8858, "step": 1097000 }, { "epoch": 2.92, "learning_rate": 2.25625e-05, "loss": 1.8872, "step": 1097500 }, { "epoch": 2.92, "learning_rate": 2.2550000000000003e-05, "loss": 1.8923, "step": 1098000 }, { "epoch": 2.92, "learning_rate": 2.25375e-05, "loss": 1.8743, "step": 1098500 }, { "epoch": 2.93, "learning_rate": 2.2525000000000002e-05, "loss": 1.8775, "step": 1099000 }, { "epoch": 2.93, "learning_rate": 2.2512500000000002e-05, "loss": 1.8949, "step": 1099500 }, { "epoch": 2.93, "learning_rate": 2.25e-05, "loss": 1.8788, "step": 1100000 }, { "epoch": 2.93, "learning_rate": 2.24875e-05, "loss": 1.8952, "step": 1100500 }, { "epoch": 2.93, "learning_rate": 2.2475e-05, "loss": 1.8924, "step": 1101000 }, { "epoch": 2.93, "learning_rate": 2.24625e-05, "loss": 1.8681, "step": 1101500 }, { "epoch": 2.93, "learning_rate": 2.245e-05, "loss": 1.8911, "step": 1102000 }, { "epoch": 2.94, "learning_rate": 2.24375e-05, "loss": 1.8917, "step": 1102500 }, { "epoch": 2.94, "learning_rate": 2.2425000000000003e-05, "loss": 1.9006, "step": 1103000 }, { "epoch": 2.94, "learning_rate": 2.24125e-05, "loss": 1.9014, "step": 1103500 }, { "epoch": 2.94, "learning_rate": 2.2400000000000002e-05, "loss": 1.8753, "step": 1104000 }, { "epoch": 2.94, "learning_rate": 2.23875e-05, "loss": 1.8878, "step": 1104500 }, { "epoch": 2.94, "learning_rate": 2.2375000000000002e-05, "loss": 1.8847, "step": 1105000 }, { "epoch": 2.94, "learning_rate": 2.23625e-05, "loss": 1.8851, "step": 1105500 }, { "epoch": 2.94, "learning_rate": 2.235e-05, "loss": 1.8727, "step": 1106000 }, { "epoch": 2.95, "learning_rate": 2.23375e-05, "loss": 1.8835, "step": 1106500 }, { "epoch": 2.95, "learning_rate": 2.2325e-05, "loss": 1.8883, "step": 1107000 }, { "epoch": 2.95, "learning_rate": 2.23125e-05, "loss": 1.8835, "step": 1107500 }, { "epoch": 2.95, "learning_rate": 2.23e-05, "loss": 1.8818, "step": 1108000 }, { "epoch": 2.95, "learning_rate": 2.22875e-05, "loss": 1.8765, "step": 1108500 }, { "epoch": 2.95, "learning_rate": 2.2275000000000003e-05, "loss": 1.8665, "step": 1109000 }, { "epoch": 2.95, "learning_rate": 2.22625e-05, "loss": 1.8862, "step": 1109500 }, { "epoch": 2.96, "learning_rate": 2.2250000000000002e-05, "loss": 1.8909, "step": 1110000 }, { "epoch": 2.96, "learning_rate": 2.22375e-05, "loss": 1.9014, "step": 1110500 }, { "epoch": 2.96, "learning_rate": 2.2225e-05, "loss": 1.8824, "step": 1111000 }, { "epoch": 2.96, "learning_rate": 2.22125e-05, "loss": 1.9019, "step": 1111500 }, { "epoch": 2.96, "learning_rate": 2.22e-05, "loss": 1.8718, "step": 1112000 }, { "epoch": 2.96, "learning_rate": 2.21875e-05, "loss": 1.8741, "step": 1112500 }, { "epoch": 2.96, "learning_rate": 2.2175e-05, "loss": 1.8684, "step": 1113000 }, { "epoch": 2.96, "learning_rate": 2.21625e-05, "loss": 1.8965, "step": 1113500 }, { "epoch": 2.97, "learning_rate": 2.215e-05, "loss": 1.8911, "step": 1114000 }, { "epoch": 2.97, "learning_rate": 2.21375e-05, "loss": 1.8889, "step": 1114500 }, { "epoch": 2.97, "learning_rate": 2.2125000000000002e-05, "loss": 1.8681, "step": 1115000 }, { "epoch": 2.97, "learning_rate": 2.21125e-05, "loss": 1.878, "step": 1115500 }, { "epoch": 2.97, "learning_rate": 2.2100000000000002e-05, "loss": 1.8773, "step": 1116000 }, { "epoch": 2.97, "learning_rate": 2.2087499999999998e-05, "loss": 1.8789, "step": 1116500 }, { "epoch": 2.97, "learning_rate": 2.2075e-05, "loss": 1.8884, "step": 1117000 }, { "epoch": 2.98, "learning_rate": 2.20625e-05, "loss": 1.8874, "step": 1117500 }, { "epoch": 2.98, "learning_rate": 2.205e-05, "loss": 1.8717, "step": 1118000 }, { "epoch": 2.98, "learning_rate": 2.20375e-05, "loss": 1.8747, "step": 1118500 }, { "epoch": 2.98, "learning_rate": 2.2025e-05, "loss": 1.8741, "step": 1119000 }, { "epoch": 2.98, "learning_rate": 2.20125e-05, "loss": 1.8771, "step": 1119500 }, { "epoch": 2.98, "learning_rate": 2.2000000000000003e-05, "loss": 1.857, "step": 1120000 }, { "epoch": 2.98, "learning_rate": 2.19875e-05, "loss": 1.8869, "step": 1120500 }, { "epoch": 2.98, "learning_rate": 2.1975000000000002e-05, "loss": 1.8718, "step": 1121000 }, { "epoch": 2.99, "learning_rate": 2.19625e-05, "loss": 1.8799, "step": 1121500 }, { "epoch": 2.99, "learning_rate": 2.195e-05, "loss": 1.8922, "step": 1122000 }, { "epoch": 2.99, "learning_rate": 2.19375e-05, "loss": 1.8728, "step": 1122500 }, { "epoch": 2.99, "learning_rate": 2.1925e-05, "loss": 1.8722, "step": 1123000 }, { "epoch": 2.99, "learning_rate": 2.19125e-05, "loss": 1.8972, "step": 1123500 }, { "epoch": 2.99, "learning_rate": 2.19e-05, "loss": 1.8873, "step": 1124000 }, { "epoch": 2.99, "learning_rate": 2.18875e-05, "loss": 1.8629, "step": 1124500 }, { "epoch": 3.0, "learning_rate": 2.1875e-05, "loss": 1.8715, "step": 1125000 }, { "epoch": 3.0, "learning_rate": 2.1862500000000003e-05, "loss": 1.8894, "step": 1125500 }, { "epoch": 3.0, "learning_rate": 2.1850000000000003e-05, "loss": 1.884, "step": 1126000 }, { "epoch": 3.0, "learning_rate": 2.1837500000000002e-05, "loss": 1.885, "step": 1126500 }, { "epoch": 3.0, "learning_rate": 2.1825000000000002e-05, "loss": 1.8778, "step": 1127000 }, { "epoch": 3.0, "learning_rate": 2.18125e-05, "loss": 1.8808, "step": 1127500 }, { "epoch": 3.0, "learning_rate": 2.18e-05, "loss": 1.8546, "step": 1128000 }, { "epoch": 3.0, "learning_rate": 2.17875e-05, "loss": 1.8648, "step": 1128500 }, { "epoch": 3.01, "learning_rate": 2.1775e-05, "loss": 1.8687, "step": 1129000 }, { "epoch": 3.01, "learning_rate": 2.1762500000000004e-05, "loss": 1.8725, "step": 1129500 }, { "epoch": 3.01, "learning_rate": 2.175e-05, "loss": 1.8812, "step": 1130000 }, { "epoch": 3.01, "learning_rate": 2.1737500000000003e-05, "loss": 1.8651, "step": 1130500 }, { "epoch": 3.01, "learning_rate": 2.1725e-05, "loss": 1.8873, "step": 1131000 }, { "epoch": 3.01, "learning_rate": 2.1712500000000003e-05, "loss": 1.8641, "step": 1131500 }, { "epoch": 3.01, "learning_rate": 2.1700000000000002e-05, "loss": 1.8641, "step": 1132000 }, { "epoch": 3.02, "learning_rate": 2.1687500000000002e-05, "loss": 1.8675, "step": 1132500 }, { "epoch": 3.02, "learning_rate": 2.1675e-05, "loss": 1.8651, "step": 1133000 }, { "epoch": 3.02, "learning_rate": 2.16625e-05, "loss": 1.8757, "step": 1133500 }, { "epoch": 3.02, "learning_rate": 2.165e-05, "loss": 1.872, "step": 1134000 }, { "epoch": 3.02, "learning_rate": 2.16375e-05, "loss": 1.851, "step": 1134500 }, { "epoch": 3.02, "learning_rate": 2.1625e-05, "loss": 1.8743, "step": 1135000 }, { "epoch": 3.02, "learning_rate": 2.1612500000000004e-05, "loss": 1.8842, "step": 1135500 }, { "epoch": 3.02, "learning_rate": 2.16e-05, "loss": 1.8718, "step": 1136000 }, { "epoch": 3.03, "learning_rate": 2.1587500000000003e-05, "loss": 1.8727, "step": 1136500 }, { "epoch": 3.03, "learning_rate": 2.1575e-05, "loss": 1.8709, "step": 1137000 }, { "epoch": 3.03, "learning_rate": 2.1562500000000002e-05, "loss": 1.8525, "step": 1137500 }, { "epoch": 3.03, "learning_rate": 2.1550000000000002e-05, "loss": 1.8559, "step": 1138000 }, { "epoch": 3.03, "learning_rate": 2.1537500000000002e-05, "loss": 1.8498, "step": 1138500 }, { "epoch": 3.03, "learning_rate": 2.1525e-05, "loss": 1.8711, "step": 1139000 }, { "epoch": 3.03, "learning_rate": 2.15125e-05, "loss": 1.8674, "step": 1139500 }, { "epoch": 3.03, "learning_rate": 2.15e-05, "loss": 1.8557, "step": 1140000 }, { "epoch": 3.04, "learning_rate": 2.14875e-05, "loss": 1.8808, "step": 1140500 }, { "epoch": 3.04, "learning_rate": 2.1475e-05, "loss": 1.8692, "step": 1141000 }, { "epoch": 3.04, "learning_rate": 2.1462500000000003e-05, "loss": 1.8541, "step": 1141500 }, { "epoch": 3.04, "learning_rate": 2.145e-05, "loss": 1.8715, "step": 1142000 }, { "epoch": 3.04, "learning_rate": 2.1437500000000003e-05, "loss": 1.8606, "step": 1142500 }, { "epoch": 3.04, "learning_rate": 2.1425e-05, "loss": 1.8725, "step": 1143000 }, { "epoch": 3.04, "learning_rate": 2.1412500000000002e-05, "loss": 1.8665, "step": 1143500 }, { "epoch": 3.05, "learning_rate": 2.1400000000000002e-05, "loss": 1.8597, "step": 1144000 }, { "epoch": 3.05, "learning_rate": 2.13875e-05, "loss": 1.8692, "step": 1144500 }, { "epoch": 3.05, "learning_rate": 2.1375e-05, "loss": 1.8613, "step": 1145000 }, { "epoch": 3.05, "learning_rate": 2.13625e-05, "loss": 1.865, "step": 1145500 }, { "epoch": 3.05, "learning_rate": 2.135e-05, "loss": 1.8661, "step": 1146000 }, { "epoch": 3.05, "learning_rate": 2.13375e-05, "loss": 1.8649, "step": 1146500 }, { "epoch": 3.05, "learning_rate": 2.1325e-05, "loss": 1.8623, "step": 1147000 }, { "epoch": 3.05, "learning_rate": 2.1312500000000003e-05, "loss": 1.8773, "step": 1147500 }, { "epoch": 3.06, "learning_rate": 2.13e-05, "loss": 1.8729, "step": 1148000 }, { "epoch": 3.06, "learning_rate": 2.1287500000000002e-05, "loss": 1.8537, "step": 1148500 }, { "epoch": 3.06, "learning_rate": 2.1275000000000002e-05, "loss": 1.8741, "step": 1149000 }, { "epoch": 3.06, "learning_rate": 2.1262500000000002e-05, "loss": 1.8554, "step": 1149500 }, { "epoch": 3.06, "learning_rate": 2.125e-05, "loss": 1.883, "step": 1150000 }, { "epoch": 3.06, "learning_rate": 2.12375e-05, "loss": 1.8697, "step": 1150500 }, { "epoch": 3.06, "learning_rate": 2.1225e-05, "loss": 1.8683, "step": 1151000 }, { "epoch": 3.07, "learning_rate": 2.12125e-05, "loss": 1.889, "step": 1151500 }, { "epoch": 3.07, "learning_rate": 2.12e-05, "loss": 1.8685, "step": 1152000 }, { "epoch": 3.07, "learning_rate": 2.1187500000000003e-05, "loss": 1.8778, "step": 1152500 }, { "epoch": 3.07, "learning_rate": 2.1175e-05, "loss": 1.8571, "step": 1153000 }, { "epoch": 3.07, "learning_rate": 2.1162500000000003e-05, "loss": 1.8672, "step": 1153500 }, { "epoch": 3.07, "learning_rate": 2.115e-05, "loss": 1.8724, "step": 1154000 }, { "epoch": 3.07, "learning_rate": 2.1137500000000002e-05, "loss": 1.8551, "step": 1154500 }, { "epoch": 3.07, "learning_rate": 2.1125000000000002e-05, "loss": 1.871, "step": 1155000 }, { "epoch": 3.08, "learning_rate": 2.11125e-05, "loss": 1.8589, "step": 1155500 }, { "epoch": 3.08, "learning_rate": 2.11e-05, "loss": 1.8536, "step": 1156000 }, { "epoch": 3.08, "learning_rate": 2.10875e-05, "loss": 1.874, "step": 1156500 }, { "epoch": 3.08, "learning_rate": 2.1075e-05, "loss": 1.8507, "step": 1157000 }, { "epoch": 3.08, "learning_rate": 2.10625e-05, "loss": 1.8518, "step": 1157500 }, { "epoch": 3.08, "learning_rate": 2.105e-05, "loss": 1.8624, "step": 1158000 }, { "epoch": 3.08, "learning_rate": 2.1037500000000003e-05, "loss": 1.8908, "step": 1158500 }, { "epoch": 3.09, "learning_rate": 2.1025e-05, "loss": 1.849, "step": 1159000 }, { "epoch": 3.09, "learning_rate": 2.1012500000000003e-05, "loss": 1.8569, "step": 1159500 }, { "epoch": 3.09, "learning_rate": 2.1e-05, "loss": 1.8613, "step": 1160000 }, { "epoch": 3.09, "learning_rate": 2.0987500000000002e-05, "loss": 1.8452, "step": 1160500 }, { "epoch": 3.09, "learning_rate": 2.0975e-05, "loss": 1.8547, "step": 1161000 }, { "epoch": 3.09, "learning_rate": 2.09625e-05, "loss": 1.8764, "step": 1161500 }, { "epoch": 3.09, "learning_rate": 2.095e-05, "loss": 1.8654, "step": 1162000 }, { "epoch": 3.09, "learning_rate": 2.09375e-05, "loss": 1.8542, "step": 1162500 }, { "epoch": 3.1, "learning_rate": 2.0925e-05, "loss": 1.8619, "step": 1163000 }, { "epoch": 3.1, "learning_rate": 2.09125e-05, "loss": 1.8595, "step": 1163500 }, { "epoch": 3.1, "learning_rate": 2.09e-05, "loss": 1.8637, "step": 1164000 }, { "epoch": 3.1, "learning_rate": 2.0887500000000003e-05, "loss": 1.8571, "step": 1164500 }, { "epoch": 3.1, "learning_rate": 2.0875e-05, "loss": 1.8712, "step": 1165000 }, { "epoch": 3.1, "learning_rate": 2.0862500000000002e-05, "loss": 1.8567, "step": 1165500 }, { "epoch": 3.1, "learning_rate": 2.085e-05, "loss": 1.8464, "step": 1166000 }, { "epoch": 3.11, "learning_rate": 2.08375e-05, "loss": 1.8553, "step": 1166500 }, { "epoch": 3.11, "learning_rate": 2.0825e-05, "loss": 1.8591, "step": 1167000 }, { "epoch": 3.11, "learning_rate": 2.08125e-05, "loss": 1.8699, "step": 1167500 }, { "epoch": 3.11, "learning_rate": 2.08e-05, "loss": 1.8469, "step": 1168000 }, { "epoch": 3.11, "learning_rate": 2.07875e-05, "loss": 1.859, "step": 1168500 }, { "epoch": 3.11, "learning_rate": 2.0775e-05, "loss": 1.8479, "step": 1169000 }, { "epoch": 3.11, "learning_rate": 2.07625e-05, "loss": 1.8484, "step": 1169500 }, { "epoch": 3.11, "learning_rate": 2.075e-05, "loss": 1.8547, "step": 1170000 }, { "epoch": 3.12, "learning_rate": 2.0737500000000003e-05, "loss": 1.86, "step": 1170500 }, { "epoch": 3.12, "learning_rate": 2.0725e-05, "loss": 1.8547, "step": 1171000 }, { "epoch": 3.12, "learning_rate": 2.0712500000000002e-05, "loss": 1.8557, "step": 1171500 }, { "epoch": 3.12, "learning_rate": 2.07e-05, "loss": 1.8476, "step": 1172000 }, { "epoch": 3.12, "learning_rate": 2.06875e-05, "loss": 1.8594, "step": 1172500 }, { "epoch": 3.12, "learning_rate": 2.0675e-05, "loss": 1.8446, "step": 1173000 }, { "epoch": 3.12, "learning_rate": 2.06625e-05, "loss": 1.8622, "step": 1173500 }, { "epoch": 3.13, "learning_rate": 2.065e-05, "loss": 1.8549, "step": 1174000 }, { "epoch": 3.13, "learning_rate": 2.06375e-05, "loss": 1.8671, "step": 1174500 }, { "epoch": 3.13, "learning_rate": 2.0625e-05, "loss": 1.842, "step": 1175000 }, { "epoch": 3.13, "learning_rate": 2.06125e-05, "loss": 1.8534, "step": 1175500 }, { "epoch": 3.13, "learning_rate": 2.06e-05, "loss": 1.8549, "step": 1176000 }, { "epoch": 3.13, "learning_rate": 2.0587500000000002e-05, "loss": 1.8621, "step": 1176500 }, { "epoch": 3.13, "learning_rate": 2.0575e-05, "loss": 1.8454, "step": 1177000 }, { "epoch": 3.13, "learning_rate": 2.0562500000000002e-05, "loss": 1.8619, "step": 1177500 }, { "epoch": 3.14, "learning_rate": 2.055e-05, "loss": 1.8449, "step": 1178000 }, { "epoch": 3.14, "learning_rate": 2.05375e-05, "loss": 1.8411, "step": 1178500 }, { "epoch": 3.14, "learning_rate": 2.0525e-05, "loss": 1.8457, "step": 1179000 }, { "epoch": 3.14, "learning_rate": 2.05125e-05, "loss": 1.854, "step": 1179500 }, { "epoch": 3.14, "learning_rate": 2.05e-05, "loss": 1.861, "step": 1180000 }, { "epoch": 3.14, "learning_rate": 2.04875e-05, "loss": 1.8447, "step": 1180500 }, { "epoch": 3.14, "learning_rate": 2.0475e-05, "loss": 1.8444, "step": 1181000 }, { "epoch": 3.15, "learning_rate": 2.0462500000000003e-05, "loss": 1.8477, "step": 1181500 }, { "epoch": 3.15, "learning_rate": 2.045e-05, "loss": 1.8693, "step": 1182000 }, { "epoch": 3.15, "learning_rate": 2.0437500000000002e-05, "loss": 1.8592, "step": 1182500 }, { "epoch": 3.15, "learning_rate": 2.0425e-05, "loss": 1.8557, "step": 1183000 }, { "epoch": 3.15, "learning_rate": 2.04125e-05, "loss": 1.8569, "step": 1183500 }, { "epoch": 3.15, "learning_rate": 2.04e-05, "loss": 1.8559, "step": 1184000 }, { "epoch": 3.15, "learning_rate": 2.03875e-05, "loss": 1.8448, "step": 1184500 }, { "epoch": 3.15, "learning_rate": 2.0375e-05, "loss": 1.8572, "step": 1185000 }, { "epoch": 3.16, "learning_rate": 2.03625e-05, "loss": 1.871, "step": 1185500 }, { "epoch": 3.16, "learning_rate": 2.035e-05, "loss": 1.8401, "step": 1186000 }, { "epoch": 3.16, "learning_rate": 2.03375e-05, "loss": 1.8543, "step": 1186500 }, { "epoch": 3.16, "learning_rate": 2.0325e-05, "loss": 1.8489, "step": 1187000 }, { "epoch": 3.16, "learning_rate": 2.0312500000000002e-05, "loss": 1.8535, "step": 1187500 }, { "epoch": 3.16, "learning_rate": 2.0300000000000002e-05, "loss": 1.8505, "step": 1188000 }, { "epoch": 3.16, "learning_rate": 2.0287500000000002e-05, "loss": 1.8457, "step": 1188500 }, { "epoch": 3.17, "learning_rate": 2.0275e-05, "loss": 1.8555, "step": 1189000 }, { "epoch": 3.17, "learning_rate": 2.02625e-05, "loss": 1.853, "step": 1189500 }, { "epoch": 3.17, "learning_rate": 2.025e-05, "loss": 1.8677, "step": 1190000 }, { "epoch": 3.17, "learning_rate": 2.02375e-05, "loss": 1.8524, "step": 1190500 }, { "epoch": 3.17, "learning_rate": 2.0225000000000004e-05, "loss": 1.84, "step": 1191000 }, { "epoch": 3.17, "learning_rate": 2.02125e-05, "loss": 1.8625, "step": 1191500 }, { "epoch": 3.17, "learning_rate": 2.0200000000000003e-05, "loss": 1.8475, "step": 1192000 }, { "epoch": 3.17, "learning_rate": 2.01875e-05, "loss": 1.8563, "step": 1192500 }, { "epoch": 3.18, "learning_rate": 2.0175000000000003e-05, "loss": 1.8421, "step": 1193000 }, { "epoch": 3.18, "learning_rate": 2.0162500000000002e-05, "loss": 1.8586, "step": 1193500 }, { "epoch": 3.18, "learning_rate": 2.0150000000000002e-05, "loss": 1.8506, "step": 1194000 }, { "epoch": 3.18, "learning_rate": 2.01375e-05, "loss": 1.867, "step": 1194500 }, { "epoch": 3.18, "learning_rate": 2.0125e-05, "loss": 1.8645, "step": 1195000 }, { "epoch": 3.18, "learning_rate": 2.01125e-05, "loss": 1.8558, "step": 1195500 }, { "epoch": 3.18, "learning_rate": 2.01e-05, "loss": 1.8564, "step": 1196000 }, { "epoch": 3.19, "learning_rate": 2.00875e-05, "loss": 1.8604, "step": 1196500 }, { "epoch": 3.19, "learning_rate": 2.0075000000000003e-05, "loss": 1.8371, "step": 1197000 }, { "epoch": 3.19, "learning_rate": 2.00625e-05, "loss": 1.8481, "step": 1197500 }, { "epoch": 3.19, "learning_rate": 2.0050000000000003e-05, "loss": 1.8443, "step": 1198000 }, { "epoch": 3.19, "learning_rate": 2.00375e-05, "loss": 1.8455, "step": 1198500 }, { "epoch": 3.19, "learning_rate": 2.0025000000000002e-05, "loss": 1.8515, "step": 1199000 }, { "epoch": 3.19, "learning_rate": 2.0012500000000002e-05, "loss": 1.8571, "step": 1199500 }, { "epoch": 3.19, "learning_rate": 2e-05, "loss": 1.8463, "step": 1200000 }, { "epoch": 3.2, "learning_rate": 1.99875e-05, "loss": 1.8432, "step": 1200500 }, { "epoch": 3.2, "learning_rate": 1.9975e-05, "loss": 1.8453, "step": 1201000 }, { "epoch": 3.2, "learning_rate": 1.99625e-05, "loss": 1.8572, "step": 1201500 }, { "epoch": 3.2, "learning_rate": 1.995e-05, "loss": 1.8701, "step": 1202000 }, { "epoch": 3.2, "learning_rate": 1.99375e-05, "loss": 1.8473, "step": 1202500 }, { "epoch": 3.2, "learning_rate": 1.9925000000000003e-05, "loss": 1.8542, "step": 1203000 }, { "epoch": 3.2, "learning_rate": 1.99125e-05, "loss": 1.8418, "step": 1203500 }, { "epoch": 3.21, "learning_rate": 1.9900000000000003e-05, "loss": 1.8444, "step": 1204000 }, { "epoch": 3.21, "learning_rate": 1.98875e-05, "loss": 1.8316, "step": 1204500 }, { "epoch": 3.21, "learning_rate": 1.9875000000000002e-05, "loss": 1.8436, "step": 1205000 }, { "epoch": 3.21, "learning_rate": 1.98625e-05, "loss": 1.8569, "step": 1205500 }, { "epoch": 3.21, "learning_rate": 1.985e-05, "loss": 1.8643, "step": 1206000 }, { "epoch": 3.21, "learning_rate": 1.98375e-05, "loss": 1.8585, "step": 1206500 }, { "epoch": 3.21, "learning_rate": 1.9825e-05, "loss": 1.8429, "step": 1207000 }, { "epoch": 3.21, "learning_rate": 1.98125e-05, "loss": 1.8384, "step": 1207500 }, { "epoch": 3.22, "learning_rate": 1.9800000000000004e-05, "loss": 1.8362, "step": 1208000 }, { "epoch": 3.22, "learning_rate": 1.97875e-05, "loss": 1.8562, "step": 1208500 }, { "epoch": 3.22, "learning_rate": 1.9775000000000003e-05, "loss": 1.8405, "step": 1209000 }, { "epoch": 3.22, "learning_rate": 1.97625e-05, "loss": 1.8326, "step": 1209500 }, { "epoch": 3.22, "learning_rate": 1.9750000000000002e-05, "loss": 1.843, "step": 1210000 }, { "epoch": 3.22, "learning_rate": 1.9737500000000002e-05, "loss": 1.847, "step": 1210500 }, { "epoch": 3.22, "learning_rate": 1.9725000000000002e-05, "loss": 1.8643, "step": 1211000 }, { "epoch": 3.23, "learning_rate": 1.97125e-05, "loss": 1.8568, "step": 1211500 }, { "epoch": 3.23, "learning_rate": 1.97e-05, "loss": 1.8291, "step": 1212000 }, { "epoch": 3.23, "learning_rate": 1.96875e-05, "loss": 1.8452, "step": 1212500 }, { "epoch": 3.23, "learning_rate": 1.9675e-05, "loss": 1.8464, "step": 1213000 }, { "epoch": 3.23, "learning_rate": 1.96625e-05, "loss": 1.844, "step": 1213500 }, { "epoch": 3.23, "learning_rate": 1.9650000000000003e-05, "loss": 1.8355, "step": 1214000 }, { "epoch": 3.23, "learning_rate": 1.96375e-05, "loss": 1.8436, "step": 1214500 }, { "epoch": 3.23, "learning_rate": 1.9625000000000003e-05, "loss": 1.8408, "step": 1215000 }, { "epoch": 3.24, "learning_rate": 1.96125e-05, "loss": 1.8323, "step": 1215500 }, { "epoch": 3.24, "learning_rate": 1.9600000000000002e-05, "loss": 1.8265, "step": 1216000 }, { "epoch": 3.24, "learning_rate": 1.9587500000000002e-05, "loss": 1.8405, "step": 1216500 }, { "epoch": 3.24, "learning_rate": 1.9575e-05, "loss": 1.8372, "step": 1217000 }, { "epoch": 3.24, "learning_rate": 1.95625e-05, "loss": 1.8422, "step": 1217500 }, { "epoch": 3.24, "learning_rate": 1.955e-05, "loss": 1.8403, "step": 1218000 }, { "epoch": 3.24, "learning_rate": 1.95375e-05, "loss": 1.8435, "step": 1218500 }, { "epoch": 3.25, "learning_rate": 1.9525e-05, "loss": 1.8397, "step": 1219000 }, { "epoch": 3.25, "learning_rate": 1.95125e-05, "loss": 1.8524, "step": 1219500 }, { "epoch": 3.25, "learning_rate": 1.9500000000000003e-05, "loss": 1.8347, "step": 1220000 }, { "epoch": 3.25, "learning_rate": 1.94875e-05, "loss": 1.8482, "step": 1220500 }, { "epoch": 3.25, "learning_rate": 1.9475000000000002e-05, "loss": 1.8465, "step": 1221000 }, { "epoch": 3.25, "learning_rate": 1.94625e-05, "loss": 1.8334, "step": 1221500 }, { "epoch": 3.25, "learning_rate": 1.9450000000000002e-05, "loss": 1.8422, "step": 1222000 }, { "epoch": 3.25, "learning_rate": 1.94375e-05, "loss": 1.8296, "step": 1222500 }, { "epoch": 3.26, "learning_rate": 1.9425e-05, "loss": 1.8471, "step": 1223000 }, { "epoch": 3.26, "learning_rate": 1.94125e-05, "loss": 1.8373, "step": 1223500 }, { "epoch": 3.26, "learning_rate": 1.94e-05, "loss": 1.8439, "step": 1224000 }, { "epoch": 3.26, "learning_rate": 1.93875e-05, "loss": 1.8411, "step": 1224500 }, { "epoch": 3.26, "learning_rate": 1.9375e-05, "loss": 1.8499, "step": 1225000 }, { "epoch": 3.26, "learning_rate": 1.93625e-05, "loss": 1.8493, "step": 1225500 }, { "epoch": 3.26, "learning_rate": 1.9350000000000003e-05, "loss": 1.8343, "step": 1226000 }, { "epoch": 3.27, "learning_rate": 1.93375e-05, "loss": 1.8341, "step": 1226500 }, { "epoch": 3.27, "learning_rate": 1.9325000000000002e-05, "loss": 1.8383, "step": 1227000 }, { "epoch": 3.27, "learning_rate": 1.93125e-05, "loss": 1.8486, "step": 1227500 }, { "epoch": 3.27, "learning_rate": 1.93e-05, "loss": 1.826, "step": 1228000 }, { "epoch": 3.27, "learning_rate": 1.92875e-05, "loss": 1.8428, "step": 1228500 }, { "epoch": 3.27, "learning_rate": 1.9275e-05, "loss": 1.8369, "step": 1229000 }, { "epoch": 3.27, "learning_rate": 1.92625e-05, "loss": 1.8444, "step": 1229500 }, { "epoch": 3.27, "learning_rate": 1.925e-05, "loss": 1.8299, "step": 1230000 }, { "epoch": 3.28, "learning_rate": 1.92375e-05, "loss": 1.8301, "step": 1230500 }, { "epoch": 3.28, "learning_rate": 1.9225e-05, "loss": 1.8233, "step": 1231000 }, { "epoch": 3.28, "learning_rate": 1.92125e-05, "loss": 1.8394, "step": 1231500 }, { "epoch": 3.28, "learning_rate": 1.9200000000000003e-05, "loss": 1.8431, "step": 1232000 }, { "epoch": 3.28, "learning_rate": 1.91875e-05, "loss": 1.8498, "step": 1232500 }, { "epoch": 3.28, "learning_rate": 1.9175000000000002e-05, "loss": 1.8402, "step": 1233000 }, { "epoch": 3.28, "learning_rate": 1.9162499999999998e-05, "loss": 1.8439, "step": 1233500 }, { "epoch": 3.29, "learning_rate": 1.915e-05, "loss": 1.8359, "step": 1234000 }, { "epoch": 3.29, "learning_rate": 1.91375e-05, "loss": 1.8359, "step": 1234500 }, { "epoch": 3.29, "learning_rate": 1.9125e-05, "loss": 1.8232, "step": 1235000 }, { "epoch": 3.29, "learning_rate": 1.91125e-05, "loss": 1.8337, "step": 1235500 }, { "epoch": 3.29, "learning_rate": 1.91e-05, "loss": 1.8509, "step": 1236000 }, { "epoch": 3.29, "learning_rate": 1.90875e-05, "loss": 1.8446, "step": 1236500 }, { "epoch": 3.29, "learning_rate": 1.9075000000000003e-05, "loss": 1.8417, "step": 1237000 }, { "epoch": 3.29, "learning_rate": 1.90625e-05, "loss": 1.8543, "step": 1237500 }, { "epoch": 3.3, "learning_rate": 1.9050000000000002e-05, "loss": 1.843, "step": 1238000 }, { "epoch": 3.3, "learning_rate": 1.90375e-05, "loss": 1.8444, "step": 1238500 }, { "epoch": 3.3, "learning_rate": 1.9025e-05, "loss": 1.8367, "step": 1239000 }, { "epoch": 3.3, "learning_rate": 1.90125e-05, "loss": 1.8314, "step": 1239500 }, { "epoch": 3.3, "learning_rate": 1.9e-05, "loss": 1.842, "step": 1240000 }, { "epoch": 3.3, "learning_rate": 1.89875e-05, "loss": 1.8431, "step": 1240500 }, { "epoch": 3.3, "learning_rate": 1.8975e-05, "loss": 1.8278, "step": 1241000 }, { "epoch": 3.31, "learning_rate": 1.89625e-05, "loss": 1.8401, "step": 1241500 }, { "epoch": 3.31, "learning_rate": 1.895e-05, "loss": 1.8397, "step": 1242000 }, { "epoch": 3.31, "learning_rate": 1.89375e-05, "loss": 1.8272, "step": 1242500 }, { "epoch": 3.31, "learning_rate": 1.8925000000000003e-05, "loss": 1.8199, "step": 1243000 }, { "epoch": 3.31, "learning_rate": 1.89125e-05, "loss": 1.8074, "step": 1243500 }, { "epoch": 3.31, "learning_rate": 1.8900000000000002e-05, "loss": 1.8322, "step": 1244000 }, { "epoch": 3.31, "learning_rate": 1.88875e-05, "loss": 1.8286, "step": 1244500 }, { "epoch": 3.31, "learning_rate": 1.8875e-05, "loss": 1.822, "step": 1245000 }, { "epoch": 3.32, "learning_rate": 1.88625e-05, "loss": 1.8351, "step": 1245500 }, { "epoch": 3.32, "learning_rate": 1.885e-05, "loss": 1.8138, "step": 1246000 }, { "epoch": 3.32, "learning_rate": 1.88375e-05, "loss": 1.8208, "step": 1246500 }, { "epoch": 3.32, "learning_rate": 1.8825e-05, "loss": 1.8225, "step": 1247000 }, { "epoch": 3.32, "learning_rate": 1.88125e-05, "loss": 1.8212, "step": 1247500 }, { "epoch": 3.32, "learning_rate": 1.88e-05, "loss": 1.8254, "step": 1248000 }, { "epoch": 3.32, "learning_rate": 1.87875e-05, "loss": 1.8383, "step": 1248500 }, { "epoch": 3.33, "learning_rate": 1.8775000000000002e-05, "loss": 1.8171, "step": 1249000 }, { "epoch": 3.33, "learning_rate": 1.87625e-05, "loss": 1.8298, "step": 1249500 }, { "epoch": 3.33, "learning_rate": 1.8750000000000002e-05, "loss": 1.8361, "step": 1250000 }, { "epoch": 3.33, "learning_rate": 1.87375e-05, "loss": 1.8318, "step": 1250500 }, { "epoch": 3.33, "learning_rate": 1.8725e-05, "loss": 1.85, "step": 1251000 }, { "epoch": 3.33, "learning_rate": 1.87125e-05, "loss": 1.8348, "step": 1251500 }, { "epoch": 3.33, "learning_rate": 1.87e-05, "loss": 1.8347, "step": 1252000 }, { "epoch": 3.33, "learning_rate": 1.8687500000000004e-05, "loss": 1.82, "step": 1252500 }, { "epoch": 3.34, "learning_rate": 1.8675e-05, "loss": 1.8276, "step": 1253000 }, { "epoch": 3.34, "learning_rate": 1.8662500000000003e-05, "loss": 1.8373, "step": 1253500 }, { "epoch": 3.34, "learning_rate": 1.865e-05, "loss": 1.8468, "step": 1254000 }, { "epoch": 3.34, "learning_rate": 1.8637500000000002e-05, "loss": 1.8341, "step": 1254500 }, { "epoch": 3.34, "learning_rate": 1.8625000000000002e-05, "loss": 1.8218, "step": 1255000 }, { "epoch": 3.34, "learning_rate": 1.8612500000000002e-05, "loss": 1.8414, "step": 1255500 }, { "epoch": 3.34, "learning_rate": 1.86e-05, "loss": 1.8386, "step": 1256000 }, { "epoch": 3.35, "learning_rate": 1.85875e-05, "loss": 1.8259, "step": 1256500 }, { "epoch": 3.35, "learning_rate": 1.8575e-05, "loss": 1.8226, "step": 1257000 }, { "epoch": 3.35, "learning_rate": 1.85625e-05, "loss": 1.8425, "step": 1257500 }, { "epoch": 3.35, "learning_rate": 1.855e-05, "loss": 1.8192, "step": 1258000 }, { "epoch": 3.35, "learning_rate": 1.8537500000000003e-05, "loss": 1.8291, "step": 1258500 }, { "epoch": 3.35, "learning_rate": 1.8525e-05, "loss": 1.8385, "step": 1259000 }, { "epoch": 3.35, "learning_rate": 1.8512500000000003e-05, "loss": 1.8242, "step": 1259500 }, { "epoch": 3.35, "learning_rate": 1.85e-05, "loss": 1.8473, "step": 1260000 }, { "epoch": 3.36, "learning_rate": 1.8487500000000002e-05, "loss": 1.8263, "step": 1260500 }, { "epoch": 3.36, "learning_rate": 1.8475000000000002e-05, "loss": 1.8409, "step": 1261000 }, { "epoch": 3.36, "learning_rate": 1.84625e-05, "loss": 1.8317, "step": 1261500 }, { "epoch": 3.36, "learning_rate": 1.845e-05, "loss": 1.8296, "step": 1262000 }, { "epoch": 3.36, "learning_rate": 1.84375e-05, "loss": 1.8409, "step": 1262500 }, { "epoch": 3.36, "learning_rate": 1.8425e-05, "loss": 1.8301, "step": 1263000 }, { "epoch": 3.36, "learning_rate": 1.84125e-05, "loss": 1.8294, "step": 1263500 }, { "epoch": 3.37, "learning_rate": 1.84e-05, "loss": 1.8193, "step": 1264000 }, { "epoch": 3.37, "learning_rate": 1.8387500000000003e-05, "loss": 1.8216, "step": 1264500 }, { "epoch": 3.37, "learning_rate": 1.8375e-05, "loss": 1.8359, "step": 1265000 }, { "epoch": 3.37, "learning_rate": 1.8362500000000002e-05, "loss": 1.8211, "step": 1265500 }, { "epoch": 3.37, "learning_rate": 1.8350000000000002e-05, "loss": 1.8248, "step": 1266000 }, { "epoch": 3.37, "learning_rate": 1.8337500000000002e-05, "loss": 1.8173, "step": 1266500 }, { "epoch": 3.37, "learning_rate": 1.8325e-05, "loss": 1.8286, "step": 1267000 }, { "epoch": 3.37, "learning_rate": 1.83125e-05, "loss": 1.8378, "step": 1267500 }, { "epoch": 3.38, "learning_rate": 1.83e-05, "loss": 1.8266, "step": 1268000 }, { "epoch": 3.38, "learning_rate": 1.82875e-05, "loss": 1.829, "step": 1268500 }, { "epoch": 3.38, "learning_rate": 1.8275e-05, "loss": 1.8209, "step": 1269000 }, { "epoch": 3.38, "learning_rate": 1.8262500000000003e-05, "loss": 1.8338, "step": 1269500 }, { "epoch": 3.38, "learning_rate": 1.825e-05, "loss": 1.8326, "step": 1270000 }, { "epoch": 3.38, "learning_rate": 1.8237500000000003e-05, "loss": 1.8289, "step": 1270500 }, { "epoch": 3.38, "learning_rate": 1.8225e-05, "loss": 1.829, "step": 1271000 }, { "epoch": 3.39, "learning_rate": 1.8212500000000002e-05, "loss": 1.8186, "step": 1271500 }, { "epoch": 3.39, "learning_rate": 1.8200000000000002e-05, "loss": 1.8341, "step": 1272000 }, { "epoch": 3.39, "learning_rate": 1.81875e-05, "loss": 1.8307, "step": 1272500 }, { "epoch": 3.39, "learning_rate": 1.8175e-05, "loss": 1.8165, "step": 1273000 }, { "epoch": 3.39, "learning_rate": 1.81625e-05, "loss": 1.8187, "step": 1273500 }, { "epoch": 3.39, "learning_rate": 1.815e-05, "loss": 1.8313, "step": 1274000 }, { "epoch": 3.39, "learning_rate": 1.81375e-05, "loss": 1.8223, "step": 1274500 }, { "epoch": 3.39, "learning_rate": 1.8125e-05, "loss": 1.8203, "step": 1275000 }, { "epoch": 3.4, "learning_rate": 1.8112500000000003e-05, "loss": 1.8291, "step": 1275500 }, { "epoch": 3.4, "learning_rate": 1.81e-05, "loss": 1.8236, "step": 1276000 }, { "epoch": 3.4, "learning_rate": 1.8087500000000003e-05, "loss": 1.8357, "step": 1276500 }, { "epoch": 3.4, "learning_rate": 1.8075e-05, "loss": 1.8311, "step": 1277000 }, { "epoch": 3.4, "learning_rate": 1.8062500000000002e-05, "loss": 1.818, "step": 1277500 }, { "epoch": 3.4, "learning_rate": 1.805e-05, "loss": 1.8359, "step": 1278000 }, { "epoch": 3.4, "learning_rate": 1.80375e-05, "loss": 1.8308, "step": 1278500 }, { "epoch": 3.41, "learning_rate": 1.8025e-05, "loss": 1.8172, "step": 1279000 }, { "epoch": 3.41, "learning_rate": 1.80125e-05, "loss": 1.8184, "step": 1279500 }, { "epoch": 3.41, "learning_rate": 1.8e-05, "loss": 1.8271, "step": 1280000 }, { "epoch": 3.41, "learning_rate": 1.79875e-05, "loss": 1.8248, "step": 1280500 }, { "epoch": 3.41, "learning_rate": 1.7975e-05, "loss": 1.8279, "step": 1281000 }, { "epoch": 3.41, "learning_rate": 1.7962500000000003e-05, "loss": 1.8319, "step": 1281500 }, { "epoch": 3.41, "learning_rate": 1.795e-05, "loss": 1.8361, "step": 1282000 }, { "epoch": 3.41, "learning_rate": 1.7937500000000002e-05, "loss": 1.8037, "step": 1282500 }, { "epoch": 3.42, "learning_rate": 1.7925e-05, "loss": 1.808, "step": 1283000 }, { "epoch": 3.42, "learning_rate": 1.7912500000000002e-05, "loss": 1.8186, "step": 1283500 }, { "epoch": 3.42, "learning_rate": 1.79e-05, "loss": 1.8119, "step": 1284000 }, { "epoch": 3.42, "learning_rate": 1.78875e-05, "loss": 1.8135, "step": 1284500 }, { "epoch": 3.42, "learning_rate": 1.7875e-05, "loss": 1.8139, "step": 1285000 }, { "epoch": 3.42, "learning_rate": 1.78625e-05, "loss": 1.8226, "step": 1285500 }, { "epoch": 3.42, "learning_rate": 1.785e-05, "loss": 1.8186, "step": 1286000 }, { "epoch": 3.42, "learning_rate": 1.78375e-05, "loss": 1.8102, "step": 1286500 }, { "epoch": 3.43, "learning_rate": 1.7825e-05, "loss": 1.8324, "step": 1287000 }, { "epoch": 3.43, "learning_rate": 1.7812500000000003e-05, "loss": 1.8232, "step": 1287500 }, { "epoch": 3.43, "learning_rate": 1.78e-05, "loss": 1.8237, "step": 1288000 }, { "epoch": 3.43, "learning_rate": 1.7787500000000002e-05, "loss": 1.8204, "step": 1288500 }, { "epoch": 3.43, "learning_rate": 1.7775e-05, "loss": 1.8092, "step": 1289000 }, { "epoch": 3.43, "learning_rate": 1.77625e-05, "loss": 1.8373, "step": 1289500 }, { "epoch": 3.43, "learning_rate": 1.775e-05, "loss": 1.8241, "step": 1290000 }, { "epoch": 3.44, "learning_rate": 1.77375e-05, "loss": 1.8198, "step": 1290500 }, { "epoch": 3.44, "learning_rate": 1.7725e-05, "loss": 1.8133, "step": 1291000 }, { "epoch": 3.44, "learning_rate": 1.77125e-05, "loss": 1.8326, "step": 1291500 }, { "epoch": 3.44, "learning_rate": 1.77e-05, "loss": 1.8219, "step": 1292000 }, { "epoch": 3.44, "learning_rate": 1.76875e-05, "loss": 1.8117, "step": 1292500 }, { "epoch": 3.44, "learning_rate": 1.7675e-05, "loss": 1.8243, "step": 1293000 }, { "epoch": 3.44, "learning_rate": 1.7662500000000002e-05, "loss": 1.8243, "step": 1293500 }, { "epoch": 3.44, "learning_rate": 1.765e-05, "loss": 1.8133, "step": 1294000 }, { "epoch": 3.45, "learning_rate": 1.7637500000000002e-05, "loss": 1.8198, "step": 1294500 }, { "epoch": 3.45, "learning_rate": 1.7625e-05, "loss": 1.8125, "step": 1295000 }, { "epoch": 3.45, "learning_rate": 1.76125e-05, "loss": 1.8164, "step": 1295500 }, { "epoch": 3.45, "learning_rate": 1.76e-05, "loss": 1.8419, "step": 1296000 }, { "epoch": 3.45, "learning_rate": 1.75875e-05, "loss": 1.8323, "step": 1296500 }, { "epoch": 3.45, "learning_rate": 1.7575e-05, "loss": 1.8149, "step": 1297000 }, { "epoch": 3.45, "learning_rate": 1.75625e-05, "loss": 1.8248, "step": 1297500 }, { "epoch": 3.46, "learning_rate": 1.755e-05, "loss": 1.838, "step": 1298000 }, { "epoch": 3.46, "learning_rate": 1.7537500000000003e-05, "loss": 1.811, "step": 1298500 }, { "epoch": 3.46, "learning_rate": 1.7525e-05, "loss": 1.815, "step": 1299000 }, { "epoch": 3.46, "learning_rate": 1.7512500000000002e-05, "loss": 1.8136, "step": 1299500 }, { "epoch": 3.46, "learning_rate": 1.75e-05, "loss": 1.8176, "step": 1300000 }, { "epoch": 3.46, "learning_rate": 1.74875e-05, "loss": 1.8215, "step": 1300500 }, { "epoch": 6.0, "learning_rate": 4.070714285714286e-05, "loss": 1.8306, "step": 1301000 }, { "epoch": 6.01, "learning_rate": 4.070357142857143e-05, "loss": 1.8357, "step": 1301500 }, { "epoch": 6.01, "learning_rate": 4.07e-05, "loss": 1.8459, "step": 1302000 }, { "epoch": 6.01, "learning_rate": 4.0696428571428574e-05, "loss": 1.8436, "step": 1302500 }, { "epoch": 6.01, "learning_rate": 4.069285714285714e-05, "loss": 1.8441, "step": 1303000 }, { "epoch": 6.02, "learning_rate": 4.0689285714285716e-05, "loss": 1.8631, "step": 1303500 }, { "epoch": 6.02, "learning_rate": 4.068571428571429e-05, "loss": 1.8582, "step": 1304000 }, { "epoch": 6.02, "learning_rate": 4.068214285714286e-05, "loss": 1.8581, "step": 1304500 }, { "epoch": 6.02, "learning_rate": 4.067857142857143e-05, "loss": 1.851, "step": 1305000 }, { "epoch": 6.03, "learning_rate": 4.0675e-05, "loss": 1.8631, "step": 1305500 }, { "epoch": 6.03, "learning_rate": 4.0671428571428574e-05, "loss": 1.8628, "step": 1306000 }, { "epoch": 6.03, "learning_rate": 4.066785714285715e-05, "loss": 1.8619, "step": 1306500 }, { "epoch": 6.03, "learning_rate": 4.0664285714285715e-05, "loss": 1.8604, "step": 1307000 }, { "epoch": 6.04, "learning_rate": 4.066071428571428e-05, "loss": 1.8589, "step": 1307500 }, { "epoch": 6.04, "learning_rate": 4.065714285714286e-05, "loss": 1.8683, "step": 1308000 }, { "epoch": 6.04, "learning_rate": 4.065357142857143e-05, "loss": 1.8537, "step": 1308500 }, { "epoch": 6.05, "learning_rate": 4.065e-05, "loss": 1.8636, "step": 1309000 }, { "epoch": 6.05, "learning_rate": 4.064642857142857e-05, "loss": 1.8584, "step": 1309500 }, { "epoch": 6.05, "learning_rate": 4.064285714285714e-05, "loss": 1.8557, "step": 1310000 }, { "epoch": 6.05, "learning_rate": 4.063928571428572e-05, "loss": 1.8603, "step": 1310500 }, { "epoch": 6.06, "learning_rate": 4.063571428571429e-05, "loss": 1.8525, "step": 1311000 }, { "epoch": 6.06, "learning_rate": 4.0632142857142856e-05, "loss": 1.8598, "step": 1311500 }, { "epoch": 6.06, "learning_rate": 4.062857142857143e-05, "loss": 1.8562, "step": 1312000 }, { "epoch": 6.06, "learning_rate": 4.0625000000000005e-05, "loss": 1.8557, "step": 1312500 }, { "epoch": 6.07, "learning_rate": 4.062142857142857e-05, "loss": 1.853, "step": 1313000 }, { "epoch": 6.07, "learning_rate": 4.061785714285715e-05, "loss": 1.8658, "step": 1313500 }, { "epoch": 6.07, "learning_rate": 4.0614285714285714e-05, "loss": 1.8616, "step": 1314000 }, { "epoch": 6.07, "learning_rate": 4.061071428571429e-05, "loss": 1.8545, "step": 1314500 }, { "epoch": 6.08, "learning_rate": 4.060714285714286e-05, "loss": 1.8446, "step": 1315000 }, { "epoch": 6.08, "learning_rate": 4.060357142857143e-05, "loss": 1.858, "step": 1315500 }, { "epoch": 6.08, "learning_rate": 4.0600000000000004e-05, "loss": 1.8581, "step": 1316000 }, { "epoch": 6.09, "learning_rate": 4.059642857142857e-05, "loss": 1.8535, "step": 1316500 }, { "epoch": 6.09, "learning_rate": 4.0592857142857146e-05, "loss": 1.859, "step": 1317000 }, { "epoch": 6.09, "learning_rate": 4.058928571428572e-05, "loss": 1.8643, "step": 1317500 }, { "epoch": 6.09, "learning_rate": 4.058571428571429e-05, "loss": 1.8528, "step": 1318000 }, { "epoch": 6.1, "learning_rate": 4.0582142857142855e-05, "loss": 1.8532, "step": 1318500 }, { "epoch": 6.1, "learning_rate": 4.057857142857143e-05, "loss": 1.8502, "step": 1319000 }, { "epoch": 6.1, "learning_rate": 4.0575000000000004e-05, "loss": 1.8637, "step": 1319500 }, { "epoch": 6.1, "learning_rate": 4.057142857142857e-05, "loss": 1.8588, "step": 1320000 }, { "epoch": 6.11, "learning_rate": 4.0567857142857145e-05, "loss": 1.8624, "step": 1320500 }, { "epoch": 6.11, "learning_rate": 4.056428571428571e-05, "loss": 1.8542, "step": 1321000 }, { "epoch": 6.11, "learning_rate": 4.056071428571429e-05, "loss": 1.8542, "step": 1321500 }, { "epoch": 6.11, "learning_rate": 4.055714285714286e-05, "loss": 1.8637, "step": 1322000 }, { "epoch": 6.12, "learning_rate": 4.055357142857143e-05, "loss": 1.8554, "step": 1322500 }, { "epoch": 6.12, "learning_rate": 4.055e-05, "loss": 1.8523, "step": 1323000 }, { "epoch": 6.12, "learning_rate": 4.054642857142858e-05, "loss": 1.8722, "step": 1323500 }, { "epoch": 6.13, "learning_rate": 4.0542857142857145e-05, "loss": 1.8517, "step": 1324000 }, { "epoch": 6.13, "learning_rate": 4.053928571428572e-05, "loss": 1.8593, "step": 1324500 }, { "epoch": 6.13, "learning_rate": 4.0535714285714287e-05, "loss": 1.8531, "step": 1325000 }, { "epoch": 6.13, "learning_rate": 4.053214285714286e-05, "loss": 1.8538, "step": 1325500 }, { "epoch": 6.14, "learning_rate": 4.0528571428571435e-05, "loss": 1.851, "step": 1326000 }, { "epoch": 6.14, "learning_rate": 4.0525e-05, "loss": 1.8587, "step": 1326500 }, { "epoch": 6.14, "learning_rate": 4.052142857142857e-05, "loss": 1.862, "step": 1327000 }, { "epoch": 6.14, "learning_rate": 4.0517857142857144e-05, "loss": 1.8573, "step": 1327500 }, { "epoch": 6.15, "learning_rate": 4.051428571428572e-05, "loss": 1.8518, "step": 1328000 }, { "epoch": 6.15, "learning_rate": 4.0510714285714286e-05, "loss": 1.8563, "step": 1328500 }, { "epoch": 6.15, "learning_rate": 4.050714285714286e-05, "loss": 1.8653, "step": 1329000 }, { "epoch": 6.15, "learning_rate": 4.050357142857143e-05, "loss": 1.8637, "step": 1329500 }, { "epoch": 6.16, "learning_rate": 4.05e-05, "loss": 1.8577, "step": 1330000 }, { "epoch": 6.16, "learning_rate": 4.0496428571428576e-05, "loss": 1.8503, "step": 1330500 }, { "epoch": 6.16, "learning_rate": 4.0492857142857144e-05, "loss": 1.8516, "step": 1331000 }, { "epoch": 6.17, "learning_rate": 4.048928571428572e-05, "loss": 1.8543, "step": 1331500 }, { "epoch": 6.17, "learning_rate": 4.0485714285714285e-05, "loss": 1.8432, "step": 1332000 }, { "epoch": 6.17, "learning_rate": 4.048214285714286e-05, "loss": 1.8613, "step": 1332500 }, { "epoch": 6.17, "learning_rate": 4.0478571428571434e-05, "loss": 1.841, "step": 1333000 }, { "epoch": 6.18, "learning_rate": 4.0475e-05, "loss": 1.8459, "step": 1333500 }, { "epoch": 6.18, "learning_rate": 4.047142857142857e-05, "loss": 1.8485, "step": 1334000 }, { "epoch": 6.18, "learning_rate": 4.046785714285714e-05, "loss": 1.8536, "step": 1334500 }, { "epoch": 6.18, "learning_rate": 4.046428571428572e-05, "loss": 1.8484, "step": 1335000 }, { "epoch": 6.19, "learning_rate": 4.0460714285714285e-05, "loss": 1.8514, "step": 1335500 }, { "epoch": 6.19, "learning_rate": 4.045714285714286e-05, "loss": 1.8492, "step": 1336000 }, { "epoch": 6.19, "learning_rate": 4.0453571428571426e-05, "loss": 1.8482, "step": 1336500 }, { "epoch": 6.19, "learning_rate": 4.045000000000001e-05, "loss": 1.8532, "step": 1337000 }, { "epoch": 6.2, "learning_rate": 4.0446428571428575e-05, "loss": 1.8516, "step": 1337500 }, { "epoch": 6.2, "learning_rate": 4.044285714285714e-05, "loss": 1.8496, "step": 1338000 }, { "epoch": 6.2, "learning_rate": 4.043928571428572e-05, "loss": 1.8537, "step": 1338500 }, { "epoch": 6.2, "learning_rate": 4.043571428571429e-05, "loss": 1.8648, "step": 1339000 }, { "epoch": 6.21, "learning_rate": 4.043214285714286e-05, "loss": 1.8323, "step": 1339500 }, { "epoch": 6.21, "learning_rate": 4.042857142857143e-05, "loss": 1.8498, "step": 1340000 }, { "epoch": 6.21, "learning_rate": 4.0425e-05, "loss": 1.8538, "step": 1340500 }, { "epoch": 6.22, "learning_rate": 4.0421428571428574e-05, "loss": 1.841, "step": 1341000 }, { "epoch": 6.22, "learning_rate": 4.041785714285715e-05, "loss": 1.8296, "step": 1341500 }, { "epoch": 6.22, "learning_rate": 4.0414285714285716e-05, "loss": 1.842, "step": 1342000 }, { "epoch": 6.22, "learning_rate": 4.0410714285714283e-05, "loss": 1.8481, "step": 1342500 }, { "epoch": 6.23, "learning_rate": 4.040714285714286e-05, "loss": 1.8397, "step": 1343000 }, { "epoch": 6.23, "learning_rate": 4.040357142857143e-05, "loss": 1.8504, "step": 1343500 }, { "epoch": 6.23, "learning_rate": 4.0400000000000006e-05, "loss": 1.8444, "step": 1344000 }, { "epoch": 6.23, "learning_rate": 4.0396428571428574e-05, "loss": 1.8348, "step": 1344500 }, { "epoch": 6.24, "learning_rate": 4.039285714285714e-05, "loss": 1.8408, "step": 1345000 }, { "epoch": 6.24, "learning_rate": 4.0389285714285715e-05, "loss": 1.8347, "step": 1345500 }, { "epoch": 6.24, "learning_rate": 4.038571428571429e-05, "loss": 1.8448, "step": 1346000 }, { "epoch": 6.24, "learning_rate": 4.038214285714286e-05, "loss": 1.8476, "step": 1346500 }, { "epoch": 6.25, "learning_rate": 4.037857142857143e-05, "loss": 1.844, "step": 1347000 }, { "epoch": 6.25, "learning_rate": 4.0375e-05, "loss": 1.8317, "step": 1347500 }, { "epoch": 6.25, "learning_rate": 4.037142857142857e-05, "loss": 1.8457, "step": 1348000 }, { "epoch": 6.26, "learning_rate": 4.036785714285715e-05, "loss": 1.8505, "step": 1348500 }, { "epoch": 6.26, "learning_rate": 4.0364285714285715e-05, "loss": 1.8464, "step": 1349000 }, { "epoch": 6.26, "learning_rate": 4.036071428571428e-05, "loss": 1.8488, "step": 1349500 }, { "epoch": 6.26, "learning_rate": 4.035714285714286e-05, "loss": 1.8453, "step": 1350000 }, { "epoch": 6.27, "learning_rate": 4.035357142857143e-05, "loss": 1.8466, "step": 1350500 }, { "epoch": 6.27, "learning_rate": 4.0350000000000005e-05, "loss": 1.8381, "step": 1351000 }, { "epoch": 6.27, "learning_rate": 4.034642857142857e-05, "loss": 1.8523, "step": 1351500 }, { "epoch": 6.27, "learning_rate": 4.034285714285715e-05, "loss": 1.8438, "step": 1352000 }, { "epoch": 6.28, "learning_rate": 4.033928571428572e-05, "loss": 1.837, "step": 1352500 }, { "epoch": 6.28, "learning_rate": 4.033571428571429e-05, "loss": 1.8418, "step": 1353000 }, { "epoch": 6.28, "learning_rate": 4.0332142857142856e-05, "loss": 1.8316, "step": 1353500 }, { "epoch": 6.28, "learning_rate": 4.032857142857143e-05, "loss": 1.852, "step": 1354000 }, { "epoch": 6.29, "learning_rate": 4.0325000000000004e-05, "loss": 1.8387, "step": 1354500 }, { "epoch": 6.29, "learning_rate": 4.032142857142857e-05, "loss": 1.8369, "step": 1355000 }, { "epoch": 6.29, "learning_rate": 4.0317857142857146e-05, "loss": 1.8439, "step": 1355500 }, { "epoch": 6.3, "learning_rate": 4.0314285714285714e-05, "loss": 1.8397, "step": 1356000 }, { "epoch": 6.3, "learning_rate": 4.031071428571429e-05, "loss": 1.8357, "step": 1356500 }, { "epoch": 6.3, "learning_rate": 4.030714285714286e-05, "loss": 1.8395, "step": 1357000 }, { "epoch": 6.3, "learning_rate": 4.030357142857143e-05, "loss": 1.8415, "step": 1357500 }, { "epoch": 6.31, "learning_rate": 4.0300000000000004e-05, "loss": 1.8384, "step": 1358000 }, { "epoch": 6.31, "learning_rate": 4.029642857142857e-05, "loss": 1.8309, "step": 1358500 }, { "epoch": 6.31, "learning_rate": 4.0292857142857146e-05, "loss": 1.8398, "step": 1359000 }, { "epoch": 6.31, "learning_rate": 4.028928571428572e-05, "loss": 1.8406, "step": 1359500 }, { "epoch": 6.32, "learning_rate": 4.028571428571429e-05, "loss": 1.8546, "step": 1360000 }, { "epoch": 6.32, "learning_rate": 4.0282142857142855e-05, "loss": 1.8394, "step": 1360500 }, { "epoch": 6.32, "learning_rate": 4.027857142857143e-05, "loss": 1.8319, "step": 1361000 }, { "epoch": 6.32, "learning_rate": 4.0275e-05, "loss": 1.8491, "step": 1361500 }, { "epoch": 6.33, "learning_rate": 4.027142857142857e-05, "loss": 1.8375, "step": 1362000 }, { "epoch": 6.33, "learning_rate": 4.0267857142857145e-05, "loss": 1.8408, "step": 1362500 }, { "epoch": 6.33, "learning_rate": 4.026428571428571e-05, "loss": 1.8394, "step": 1363000 }, { "epoch": 6.34, "learning_rate": 4.0260714285714287e-05, "loss": 1.8418, "step": 1363500 }, { "epoch": 6.34, "learning_rate": 4.025714285714286e-05, "loss": 1.8363, "step": 1364000 }, { "epoch": 6.34, "learning_rate": 4.025357142857143e-05, "loss": 1.8409, "step": 1364500 }, { "epoch": 6.34, "learning_rate": 4.025e-05, "loss": 1.8467, "step": 1365000 }, { "epoch": 6.35, "learning_rate": 4.024642857142858e-05, "loss": 1.8496, "step": 1365500 }, { "epoch": 6.35, "learning_rate": 4.0242857142857144e-05, "loss": 1.829, "step": 1366000 }, { "epoch": 6.35, "learning_rate": 4.023928571428572e-05, "loss": 1.8462, "step": 1366500 }, { "epoch": 6.35, "learning_rate": 4.0235714285714286e-05, "loss": 1.8364, "step": 1367000 }, { "epoch": 6.36, "learning_rate": 4.023214285714286e-05, "loss": 1.8443, "step": 1367500 }, { "epoch": 6.36, "learning_rate": 4.0228571428571434e-05, "loss": 1.8457, "step": 1368000 }, { "epoch": 6.36, "learning_rate": 4.0225e-05, "loss": 1.8438, "step": 1368500 }, { "epoch": 6.36, "learning_rate": 4.022142857142857e-05, "loss": 1.8376, "step": 1369000 }, { "epoch": 6.37, "learning_rate": 4.0217857142857144e-05, "loss": 1.8234, "step": 1369500 }, { "epoch": 6.37, "learning_rate": 4.021428571428572e-05, "loss": 1.8266, "step": 1370000 }, { "epoch": 6.37, "learning_rate": 4.0210714285714285e-05, "loss": 1.844, "step": 1370500 }, { "epoch": 6.38, "learning_rate": 4.020714285714286e-05, "loss": 1.8258, "step": 1371000 }, { "epoch": 6.38, "learning_rate": 4.020357142857143e-05, "loss": 1.8321, "step": 1371500 }, { "epoch": 6.38, "learning_rate": 4.02e-05, "loss": 1.8444, "step": 1372000 }, { "epoch": 6.38, "learning_rate": 4.0196428571428576e-05, "loss": 1.8337, "step": 1372500 }, { "epoch": 6.39, "learning_rate": 4.019285714285714e-05, "loss": 1.8369, "step": 1373000 }, { "epoch": 6.39, "learning_rate": 4.018928571428572e-05, "loss": 1.8204, "step": 1373500 }, { "epoch": 6.39, "learning_rate": 4.0185714285714285e-05, "loss": 1.8378, "step": 1374000 }, { "epoch": 6.39, "learning_rate": 4.018214285714286e-05, "loss": 1.8317, "step": 1374500 }, { "epoch": 6.4, "learning_rate": 4.017857142857143e-05, "loss": 1.8331, "step": 1375000 }, { "epoch": 6.4, "learning_rate": 4.0175e-05, "loss": 1.8353, "step": 1375500 }, { "epoch": 6.4, "learning_rate": 4.017142857142857e-05, "loss": 1.8251, "step": 1376000 }, { "epoch": 6.4, "learning_rate": 4.016785714285715e-05, "loss": 1.837, "step": 1376500 }, { "epoch": 6.41, "learning_rate": 4.016428571428572e-05, "loss": 1.8328, "step": 1377000 }, { "epoch": 6.41, "learning_rate": 4.016071428571429e-05, "loss": 1.8248, "step": 1377500 }, { "epoch": 6.41, "learning_rate": 4.015714285714286e-05, "loss": 1.8366, "step": 1378000 }, { "epoch": 6.42, "learning_rate": 4.0153571428571426e-05, "loss": 1.8185, "step": 1378500 }, { "epoch": 6.42, "learning_rate": 4.015000000000001e-05, "loss": 1.8224, "step": 1379000 }, { "epoch": 6.42, "learning_rate": 4.0146428571428574e-05, "loss": 1.8356, "step": 1379500 }, { "epoch": 6.42, "learning_rate": 4.014285714285714e-05, "loss": 1.8375, "step": 1380000 }, { "epoch": 6.43, "learning_rate": 4.0139285714285716e-05, "loss": 1.8232, "step": 1380500 }, { "epoch": 6.43, "learning_rate": 4.013571428571429e-05, "loss": 1.8208, "step": 1381000 }, { "epoch": 6.43, "learning_rate": 4.013214285714286e-05, "loss": 1.8311, "step": 1381500 }, { "epoch": 6.43, "learning_rate": 4.012857142857143e-05, "loss": 1.8394, "step": 1382000 }, { "epoch": 6.44, "learning_rate": 4.0125e-05, "loss": 1.8296, "step": 1382500 }, { "epoch": 6.44, "learning_rate": 4.0121428571428574e-05, "loss": 1.8329, "step": 1383000 }, { "epoch": 6.44, "learning_rate": 4.011785714285715e-05, "loss": 1.8298, "step": 1383500 }, { "epoch": 6.44, "learning_rate": 4.0114285714285715e-05, "loss": 1.8387, "step": 1384000 }, { "epoch": 6.45, "learning_rate": 4.011071428571429e-05, "loss": 1.8401, "step": 1384500 }, { "epoch": 6.45, "learning_rate": 4.010714285714286e-05, "loss": 1.8285, "step": 1385000 }, { "epoch": 6.45, "learning_rate": 4.010357142857143e-05, "loss": 1.8339, "step": 1385500 }, { "epoch": 6.46, "learning_rate": 4.0100000000000006e-05, "loss": 1.8351, "step": 1386000 }, { "epoch": 6.46, "learning_rate": 4.009642857142857e-05, "loss": 1.825, "step": 1386500 }, { "epoch": 6.46, "learning_rate": 4.009285714285714e-05, "loss": 1.8343, "step": 1387000 }, { "epoch": 6.46, "learning_rate": 4.0089285714285715e-05, "loss": 1.8281, "step": 1387500 }, { "epoch": 6.47, "learning_rate": 4.008571428571429e-05, "loss": 1.826, "step": 1388000 }, { "epoch": 6.47, "learning_rate": 4.0082142857142857e-05, "loss": 1.813, "step": 1388500 }, { "epoch": 6.47, "learning_rate": 4.007857142857143e-05, "loss": 1.8334, "step": 1389000 }, { "epoch": 6.47, "learning_rate": 4.0075e-05, "loss": 1.8344, "step": 1389500 }, { "epoch": 6.48, "learning_rate": 4.007142857142857e-05, "loss": 1.8205, "step": 1390000 }, { "epoch": 6.48, "learning_rate": 4.006785714285715e-05, "loss": 1.8275, "step": 1390500 }, { "epoch": 6.48, "learning_rate": 4.0064285714285714e-05, "loss": 1.8282, "step": 1391000 }, { "epoch": 6.48, "learning_rate": 4.006071428571429e-05, "loss": 1.8187, "step": 1391500 }, { "epoch": 6.49, "learning_rate": 4.005714285714286e-05, "loss": 1.8286, "step": 1392000 }, { "epoch": 6.49, "learning_rate": 4.005357142857143e-05, "loss": 1.821, "step": 1392500 }, { "epoch": 6.49, "learning_rate": 4.0050000000000004e-05, "loss": 1.8248, "step": 1393000 }, { "epoch": 6.5, "learning_rate": 4.004642857142857e-05, "loss": 1.8207, "step": 1393500 }, { "epoch": 6.5, "learning_rate": 4.0042857142857146e-05, "loss": 1.8226, "step": 1394000 }, { "epoch": 6.5, "learning_rate": 4.003928571428572e-05, "loss": 1.8266, "step": 1394500 }, { "epoch": 6.5, "learning_rate": 4.003571428571429e-05, "loss": 1.8222, "step": 1395000 }, { "epoch": 6.51, "learning_rate": 4.0032142857142855e-05, "loss": 1.8336, "step": 1395500 }, { "epoch": 6.51, "learning_rate": 4.002857142857143e-05, "loss": 1.8205, "step": 1396000 }, { "epoch": 6.51, "learning_rate": 4.0025000000000004e-05, "loss": 1.8313, "step": 1396500 }, { "epoch": 6.51, "learning_rate": 4.002142857142857e-05, "loss": 1.8288, "step": 1397000 }, { "epoch": 6.52, "learning_rate": 4.0017857142857146e-05, "loss": 1.8227, "step": 1397500 }, { "epoch": 6.52, "learning_rate": 4.001428571428571e-05, "loss": 1.8202, "step": 1398000 }, { "epoch": 6.52, "learning_rate": 4.001071428571429e-05, "loss": 1.8255, "step": 1398500 }, { "epoch": 6.52, "learning_rate": 4.000714285714286e-05, "loss": 1.8201, "step": 1399000 }, { "epoch": 6.53, "learning_rate": 4.000357142857143e-05, "loss": 1.8234, "step": 1399500 }, { "epoch": 6.53, "learning_rate": 4e-05, "loss": 1.8195, "step": 1400000 }, { "epoch": 6.53, "learning_rate": 3.999642857142857e-05, "loss": 1.8199, "step": 1400500 }, { "epoch": 6.54, "learning_rate": 3.9992857142857145e-05, "loss": 1.8267, "step": 1401000 }, { "epoch": 6.54, "learning_rate": 3.998928571428572e-05, "loss": 1.824, "step": 1401500 }, { "epoch": 6.54, "learning_rate": 3.998571428571429e-05, "loss": 1.8174, "step": 1402000 }, { "epoch": 6.54, "learning_rate": 3.9982142857142854e-05, "loss": 1.8234, "step": 1402500 }, { "epoch": 6.55, "learning_rate": 3.9978571428571435e-05, "loss": 1.8265, "step": 1403000 }, { "epoch": 6.55, "learning_rate": 3.9975e-05, "loss": 1.8248, "step": 1403500 }, { "epoch": 6.55, "learning_rate": 3.997142857142857e-05, "loss": 1.822, "step": 1404000 }, { "epoch": 6.55, "learning_rate": 3.9967857142857144e-05, "loss": 1.8226, "step": 1404500 }, { "epoch": 6.56, "learning_rate": 3.996428571428571e-05, "loss": 1.8261, "step": 1405000 }, { "epoch": 6.56, "learning_rate": 3.996071428571429e-05, "loss": 1.8187, "step": 1405500 }, { "epoch": 6.56, "learning_rate": 3.995714285714286e-05, "loss": 1.8282, "step": 1406000 }, { "epoch": 6.56, "learning_rate": 3.995357142857143e-05, "loss": 1.8178, "step": 1406500 }, { "epoch": 6.57, "learning_rate": 3.995e-05, "loss": 1.8285, "step": 1407000 }, { "epoch": 6.57, "learning_rate": 3.9946428571428576e-05, "loss": 1.8218, "step": 1407500 }, { "epoch": 6.57, "learning_rate": 3.9942857142857144e-05, "loss": 1.8203, "step": 1408000 }, { "epoch": 6.58, "learning_rate": 3.993928571428572e-05, "loss": 1.8236, "step": 1408500 }, { "epoch": 6.58, "learning_rate": 3.9935714285714285e-05, "loss": 1.8214, "step": 1409000 }, { "epoch": 6.58, "learning_rate": 3.993214285714286e-05, "loss": 1.8273, "step": 1409500 }, { "epoch": 6.58, "learning_rate": 3.9928571428571434e-05, "loss": 1.8164, "step": 1410000 }, { "epoch": 6.59, "learning_rate": 3.9925e-05, "loss": 1.8155, "step": 1410500 }, { "epoch": 6.59, "learning_rate": 3.992142857142857e-05, "loss": 1.8217, "step": 1411000 }, { "epoch": 6.59, "learning_rate": 3.991785714285714e-05, "loss": 1.8112, "step": 1411500 }, { "epoch": 6.59, "learning_rate": 3.991428571428572e-05, "loss": 1.8146, "step": 1412000 }, { "epoch": 6.6, "learning_rate": 3.991071428571429e-05, "loss": 1.8079, "step": 1412500 }, { "epoch": 6.6, "learning_rate": 3.990714285714286e-05, "loss": 1.8123, "step": 1413000 }, { "epoch": 6.6, "learning_rate": 3.9903571428571427e-05, "loss": 1.8086, "step": 1413500 }, { "epoch": 6.6, "learning_rate": 3.99e-05, "loss": 1.8158, "step": 1414000 }, { "epoch": 6.61, "learning_rate": 3.9896428571428575e-05, "loss": 1.8242, "step": 1414500 }, { "epoch": 6.61, "learning_rate": 3.989285714285714e-05, "loss": 1.8109, "step": 1415000 }, { "epoch": 6.61, "learning_rate": 3.988928571428572e-05, "loss": 1.8153, "step": 1415500 }, { "epoch": 6.61, "learning_rate": 3.9885714285714284e-05, "loss": 1.8318, "step": 1416000 }, { "epoch": 6.62, "learning_rate": 3.988214285714286e-05, "loss": 1.8216, "step": 1416500 }, { "epoch": 6.62, "learning_rate": 3.987857142857143e-05, "loss": 1.8115, "step": 1417000 }, { "epoch": 6.62, "learning_rate": 3.9875e-05, "loss": 1.8129, "step": 1417500 }, { "epoch": 6.63, "learning_rate": 3.987142857142857e-05, "loss": 1.8102, "step": 1418000 }, { "epoch": 6.63, "learning_rate": 3.986785714285715e-05, "loss": 1.8085, "step": 1418500 }, { "epoch": 6.63, "learning_rate": 3.9864285714285716e-05, "loss": 1.8095, "step": 1419000 }, { "epoch": 6.63, "learning_rate": 3.986071428571429e-05, "loss": 1.822, "step": 1419500 }, { "epoch": 6.64, "learning_rate": 3.985714285714286e-05, "loss": 1.8174, "step": 1420000 }, { "epoch": 6.64, "learning_rate": 3.985357142857143e-05, "loss": 1.8242, "step": 1420500 }, { "epoch": 6.64, "learning_rate": 3.9850000000000006e-05, "loss": 1.8096, "step": 1421000 }, { "epoch": 6.64, "learning_rate": 3.9846428571428574e-05, "loss": 1.8158, "step": 1421500 }, { "epoch": 6.65, "learning_rate": 3.984285714285714e-05, "loss": 1.8078, "step": 1422000 }, { "epoch": 6.65, "learning_rate": 3.9839285714285716e-05, "loss": 1.8018, "step": 1422500 }, { "epoch": 6.65, "learning_rate": 3.983571428571429e-05, "loss": 1.8256, "step": 1423000 }, { "epoch": 6.65, "learning_rate": 3.983214285714286e-05, "loss": 1.8132, "step": 1423500 }, { "epoch": 6.66, "learning_rate": 3.982857142857143e-05, "loss": 1.8149, "step": 1424000 }, { "epoch": 6.66, "learning_rate": 3.9825e-05, "loss": 1.8057, "step": 1424500 }, { "epoch": 6.66, "learning_rate": 3.982142857142857e-05, "loss": 1.8155, "step": 1425000 }, { "epoch": 6.67, "learning_rate": 3.981785714285715e-05, "loss": 1.8123, "step": 1425500 }, { "epoch": 6.67, "learning_rate": 3.9814285714285715e-05, "loss": 1.8173, "step": 1426000 }, { "epoch": 6.67, "learning_rate": 3.981071428571429e-05, "loss": 1.7982, "step": 1426500 }, { "epoch": 6.67, "learning_rate": 3.980714285714286e-05, "loss": 1.8216, "step": 1427000 }, { "epoch": 6.68, "learning_rate": 3.980357142857143e-05, "loss": 1.819, "step": 1427500 }, { "epoch": 6.68, "learning_rate": 3.9800000000000005e-05, "loss": 1.8154, "step": 1428000 }, { "epoch": 6.68, "learning_rate": 3.979642857142857e-05, "loss": 1.8136, "step": 1428500 }, { "epoch": 6.68, "learning_rate": 3.979285714285714e-05, "loss": 1.8141, "step": 1429000 }, { "epoch": 6.69, "learning_rate": 3.978928571428572e-05, "loss": 1.8126, "step": 1429500 }, { "epoch": 6.69, "learning_rate": 3.978571428571429e-05, "loss": 1.7951, "step": 1430000 }, { "epoch": 6.69, "learning_rate": 3.9782142857142856e-05, "loss": 1.803, "step": 1430500 }, { "epoch": 6.69, "learning_rate": 3.977857142857143e-05, "loss": 1.8245, "step": 1431000 }, { "epoch": 6.7, "learning_rate": 3.9775e-05, "loss": 1.8084, "step": 1431500 }, { "epoch": 6.7, "learning_rate": 3.977142857142857e-05, "loss": 1.8118, "step": 1432000 }, { "epoch": 6.7, "learning_rate": 3.9767857142857146e-05, "loss": 1.8125, "step": 1432500 }, { "epoch": 6.71, "learning_rate": 3.9764285714285714e-05, "loss": 1.8174, "step": 1433000 }, { "epoch": 6.71, "learning_rate": 3.976071428571429e-05, "loss": 1.8092, "step": 1433500 }, { "epoch": 6.71, "learning_rate": 3.975714285714286e-05, "loss": 1.8131, "step": 1434000 }, { "epoch": 6.71, "learning_rate": 3.975357142857143e-05, "loss": 1.827, "step": 1434500 }, { "epoch": 6.72, "learning_rate": 3.9750000000000004e-05, "loss": 1.8029, "step": 1435000 }, { "epoch": 6.72, "learning_rate": 3.974642857142857e-05, "loss": 1.8128, "step": 1435500 }, { "epoch": 6.72, "learning_rate": 3.9742857142857146e-05, "loss": 1.8108, "step": 1436000 }, { "epoch": 6.72, "learning_rate": 3.973928571428572e-05, "loss": 1.8059, "step": 1436500 }, { "epoch": 6.73, "learning_rate": 3.973571428571429e-05, "loss": 1.7995, "step": 1437000 }, { "epoch": 6.73, "learning_rate": 3.9732142857142855e-05, "loss": 1.8073, "step": 1437500 }, { "epoch": 6.73, "learning_rate": 3.972857142857143e-05, "loss": 1.8156, "step": 1438000 }, { "epoch": 6.73, "learning_rate": 3.9725e-05, "loss": 1.7997, "step": 1438500 }, { "epoch": 6.74, "learning_rate": 3.972142857142858e-05, "loss": 1.8106, "step": 1439000 }, { "epoch": 6.74, "learning_rate": 3.9717857142857145e-05, "loss": 1.8031, "step": 1439500 }, { "epoch": 6.74, "learning_rate": 3.971428571428571e-05, "loss": 1.7966, "step": 1440000 }, { "epoch": 6.75, "learning_rate": 3.971071428571429e-05, "loss": 1.8052, "step": 1440500 }, { "epoch": 6.75, "learning_rate": 3.970714285714286e-05, "loss": 1.8147, "step": 1441000 }, { "epoch": 6.75, "learning_rate": 3.970357142857143e-05, "loss": 1.8039, "step": 1441500 }, { "epoch": 6.75, "learning_rate": 3.97e-05, "loss": 1.81, "step": 1442000 }, { "epoch": 6.76, "learning_rate": 3.969642857142857e-05, "loss": 1.8037, "step": 1442500 }, { "epoch": 6.76, "learning_rate": 3.9692857142857144e-05, "loss": 1.8022, "step": 1443000 }, { "epoch": 6.76, "learning_rate": 3.968928571428572e-05, "loss": 1.8104, "step": 1443500 }, { "epoch": 6.76, "learning_rate": 3.9685714285714286e-05, "loss": 1.8198, "step": 1444000 }, { "epoch": 6.77, "learning_rate": 3.9682142857142854e-05, "loss": 1.7931, "step": 1444500 }, { "epoch": 6.77, "learning_rate": 3.9678571428571435e-05, "loss": 1.8102, "step": 1445000 }, { "epoch": 6.77, "learning_rate": 3.9675e-05, "loss": 1.811, "step": 1445500 }, { "epoch": 6.77, "learning_rate": 3.9671428571428576e-05, "loss": 1.8102, "step": 1446000 }, { "epoch": 6.78, "learning_rate": 3.9667857142857144e-05, "loss": 1.8048, "step": 1446500 }, { "epoch": 6.78, "learning_rate": 3.966428571428572e-05, "loss": 1.8068, "step": 1447000 }, { "epoch": 6.78, "learning_rate": 3.966071428571429e-05, "loss": 1.8014, "step": 1447500 }, { "epoch": 6.79, "learning_rate": 3.965714285714286e-05, "loss": 1.8075, "step": 1448000 }, { "epoch": 6.79, "learning_rate": 3.965357142857143e-05, "loss": 1.8052, "step": 1448500 }, { "epoch": 6.79, "learning_rate": 3.965e-05, "loss": 1.8112, "step": 1449000 }, { "epoch": 6.79, "learning_rate": 3.9646428571428576e-05, "loss": 1.8026, "step": 1449500 }, { "epoch": 6.8, "learning_rate": 3.964285714285714e-05, "loss": 1.7933, "step": 1450000 }, { "epoch": 6.8, "learning_rate": 3.963928571428572e-05, "loss": 1.8172, "step": 1450500 }, { "epoch": 6.8, "learning_rate": 3.9635714285714285e-05, "loss": 1.7904, "step": 1451000 }, { "epoch": 6.8, "learning_rate": 3.963214285714286e-05, "loss": 1.8062, "step": 1451500 }, { "epoch": 6.81, "learning_rate": 3.9628571428571433e-05, "loss": 1.788, "step": 1452000 }, { "epoch": 6.81, "learning_rate": 3.9625e-05, "loss": 1.8007, "step": 1452500 }, { "epoch": 6.81, "learning_rate": 3.9621428571428575e-05, "loss": 1.8045, "step": 1453000 }, { "epoch": 6.81, "learning_rate": 3.961785714285714e-05, "loss": 1.7986, "step": 1453500 }, { "epoch": 6.82, "learning_rate": 3.961428571428572e-05, "loss": 1.8029, "step": 1454000 }, { "epoch": 6.82, "learning_rate": 3.961071428571429e-05, "loss": 1.8012, "step": 1454500 }, { "epoch": 6.82, "learning_rate": 3.960714285714286e-05, "loss": 1.8079, "step": 1455000 }, { "epoch": 6.83, "learning_rate": 3.9603571428571426e-05, "loss": 1.798, "step": 1455500 }, { "epoch": 6.83, "learning_rate": 3.960000000000001e-05, "loss": 1.7982, "step": 1456000 }, { "epoch": 6.83, "learning_rate": 3.9596428571428575e-05, "loss": 1.8069, "step": 1456500 }, { "epoch": 6.83, "learning_rate": 3.959285714285714e-05, "loss": 1.8057, "step": 1457000 }, { "epoch": 6.84, "learning_rate": 3.9589285714285716e-05, "loss": 1.8077, "step": 1457500 }, { "epoch": 6.84, "learning_rate": 3.9585714285714284e-05, "loss": 1.7941, "step": 1458000 }, { "epoch": 6.84, "learning_rate": 3.958214285714286e-05, "loss": 1.8098, "step": 1458500 }, { "epoch": 6.84, "learning_rate": 3.957857142857143e-05, "loss": 1.8068, "step": 1459000 }, { "epoch": 6.85, "learning_rate": 3.9575e-05, "loss": 1.8062, "step": 1459500 }, { "epoch": 6.85, "learning_rate": 3.9571428571428574e-05, "loss": 1.812, "step": 1460000 }, { "epoch": 6.85, "learning_rate": 3.956785714285715e-05, "loss": 1.8037, "step": 1460500 }, { "epoch": 6.85, "learning_rate": 3.9564285714285716e-05, "loss": 1.8041, "step": 1461000 }, { "epoch": 6.86, "learning_rate": 3.956071428571429e-05, "loss": 1.8043, "step": 1461500 }, { "epoch": 6.86, "learning_rate": 3.955714285714286e-05, "loss": 1.8066, "step": 1462000 }, { "epoch": 6.86, "learning_rate": 3.955357142857143e-05, "loss": 1.8037, "step": 1462500 }, { "epoch": 6.87, "learning_rate": 3.9550000000000006e-05, "loss": 1.7861, "step": 1463000 }, { "epoch": 6.87, "learning_rate": 3.954642857142857e-05, "loss": 1.8015, "step": 1463500 }, { "epoch": 6.87, "learning_rate": 3.954285714285714e-05, "loss": 1.8048, "step": 1464000 }, { "epoch": 6.87, "learning_rate": 3.9539285714285715e-05, "loss": 1.7866, "step": 1464500 }, { "epoch": 6.88, "learning_rate": 3.953571428571429e-05, "loss": 1.7997, "step": 1465000 }, { "epoch": 6.88, "learning_rate": 3.953214285714286e-05, "loss": 1.7968, "step": 1465500 }, { "epoch": 6.88, "learning_rate": 3.952857142857143e-05, "loss": 1.8047, "step": 1466000 }, { "epoch": 6.88, "learning_rate": 3.9525e-05, "loss": 1.8004, "step": 1466500 }, { "epoch": 6.89, "learning_rate": 3.952142857142857e-05, "loss": 1.8012, "step": 1467000 }, { "epoch": 6.89, "learning_rate": 3.951785714285715e-05, "loss": 1.7978, "step": 1467500 }, { "epoch": 6.89, "learning_rate": 3.9514285714285714e-05, "loss": 1.8087, "step": 1468000 }, { "epoch": 6.89, "learning_rate": 3.951071428571429e-05, "loss": 1.7879, "step": 1468500 }, { "epoch": 6.9, "learning_rate": 3.9507142857142856e-05, "loss": 1.8079, "step": 1469000 }, { "epoch": 6.9, "learning_rate": 3.950357142857143e-05, "loss": 1.7922, "step": 1469500 }, { "epoch": 6.9, "learning_rate": 3.9500000000000005e-05, "loss": 1.7949, "step": 1470000 }, { "epoch": 6.91, "learning_rate": 3.949642857142857e-05, "loss": 1.789, "step": 1470500 }, { "epoch": 6.91, "learning_rate": 3.949285714285714e-05, "loss": 1.7961, "step": 1471000 }, { "epoch": 6.91, "learning_rate": 3.948928571428572e-05, "loss": 1.7912, "step": 1471500 }, { "epoch": 6.91, "learning_rate": 3.948571428571429e-05, "loss": 1.8009, "step": 1472000 }, { "epoch": 6.92, "learning_rate": 3.9482142857142856e-05, "loss": 1.796, "step": 1472500 }, { "epoch": 6.92, "learning_rate": 3.947857142857143e-05, "loss": 1.7983, "step": 1473000 }, { "epoch": 6.92, "learning_rate": 3.9475000000000004e-05, "loss": 1.7952, "step": 1473500 }, { "epoch": 6.92, "learning_rate": 3.947142857142858e-05, "loss": 1.7887, "step": 1474000 }, { "epoch": 6.93, "learning_rate": 3.9467857142857146e-05, "loss": 1.796, "step": 1474500 }, { "epoch": 6.93, "learning_rate": 3.946428571428571e-05, "loss": 1.8096, "step": 1475000 }, { "epoch": 6.93, "learning_rate": 3.946071428571429e-05, "loss": 1.8001, "step": 1475500 }, { "epoch": 6.93, "learning_rate": 3.945714285714286e-05, "loss": 1.7936, "step": 1476000 }, { "epoch": 6.94, "learning_rate": 3.945357142857143e-05, "loss": 1.7902, "step": 1476500 }, { "epoch": 6.94, "learning_rate": 3.9450000000000003e-05, "loss": 1.8063, "step": 1477000 }, { "epoch": 6.94, "learning_rate": 3.944642857142857e-05, "loss": 1.8011, "step": 1477500 }, { "epoch": 6.95, "learning_rate": 3.9442857142857145e-05, "loss": 1.7895, "step": 1478000 }, { "epoch": 6.95, "learning_rate": 3.943928571428572e-05, "loss": 1.7911, "step": 1478500 }, { "epoch": 6.95, "learning_rate": 3.943571428571429e-05, "loss": 1.8058, "step": 1479000 }, { "epoch": 6.95, "learning_rate": 3.9432142857142854e-05, "loss": 1.7918, "step": 1479500 }, { "epoch": 6.96, "learning_rate": 3.942857142857143e-05, "loss": 1.7944, "step": 1480000 }, { "epoch": 6.96, "learning_rate": 3.9425e-05, "loss": 1.7957, "step": 1480500 }, { "epoch": 6.96, "learning_rate": 3.942142857142858e-05, "loss": 1.7969, "step": 1481000 }, { "epoch": 6.96, "learning_rate": 3.9417857142857145e-05, "loss": 1.7965, "step": 1481500 }, { "epoch": 6.97, "learning_rate": 3.941428571428571e-05, "loss": 1.7976, "step": 1482000 }, { "epoch": 6.97, "learning_rate": 3.941071428571429e-05, "loss": 1.7886, "step": 1482500 }, { "epoch": 6.97, "learning_rate": 3.940714285714286e-05, "loss": 1.796, "step": 1483000 }, { "epoch": 6.97, "learning_rate": 3.940357142857143e-05, "loss": 1.7873, "step": 1483500 }, { "epoch": 6.98, "learning_rate": 3.94e-05, "loss": 1.7871, "step": 1484000 }, { "epoch": 6.98, "learning_rate": 3.939642857142857e-05, "loss": 1.7938, "step": 1484500 }, { "epoch": 6.98, "learning_rate": 3.9392857142857144e-05, "loss": 1.7955, "step": 1485000 }, { "epoch": 6.99, "learning_rate": 3.938928571428572e-05, "loss": 1.793, "step": 1485500 }, { "epoch": 6.99, "learning_rate": 3.9385714285714286e-05, "loss": 1.7997, "step": 1486000 }, { "epoch": 6.99, "learning_rate": 3.938214285714285e-05, "loss": 1.7901, "step": 1486500 }, { "epoch": 6.99, "learning_rate": 3.9378571428571434e-05, "loss": 1.7856, "step": 1487000 }, { "epoch": 7.0, "learning_rate": 3.9375e-05, "loss": 1.7839, "step": 1487500 }, { "epoch": 7.0, "learning_rate": 3.9371428571428576e-05, "loss": 1.7836, "step": 1488000 }, { "epoch": 7.0, "learning_rate": 3.936785714285714e-05, "loss": 1.7939, "step": 1488500 }, { "epoch": 7.0, "learning_rate": 3.936428571428572e-05, "loss": 1.7945, "step": 1489000 }, { "epoch": 7.01, "learning_rate": 3.936071428571429e-05, "loss": 1.7835, "step": 1489500 }, { "epoch": 7.01, "learning_rate": 3.935714285714286e-05, "loss": 1.7998, "step": 1490000 }, { "epoch": 7.01, "learning_rate": 3.935357142857143e-05, "loss": 1.7829, "step": 1490500 }, { "epoch": 7.01, "learning_rate": 3.935e-05, "loss": 1.7817, "step": 1491000 }, { "epoch": 7.02, "learning_rate": 3.9346428571428575e-05, "loss": 1.7908, "step": 1491500 }, { "epoch": 7.02, "learning_rate": 3.934285714285714e-05, "loss": 1.773, "step": 1492000 }, { "epoch": 7.02, "learning_rate": 3.933928571428572e-05, "loss": 1.7784, "step": 1492500 }, { "epoch": 7.02, "learning_rate": 3.9335714285714284e-05, "loss": 1.7909, "step": 1493000 }, { "epoch": 7.03, "learning_rate": 3.933214285714286e-05, "loss": 1.7774, "step": 1493500 }, { "epoch": 7.03, "learning_rate": 3.932857142857143e-05, "loss": 1.7863, "step": 1494000 }, { "epoch": 7.03, "learning_rate": 3.9325e-05, "loss": 1.7954, "step": 1494500 }, { "epoch": 7.04, "learning_rate": 3.9321428571428575e-05, "loss": 1.7906, "step": 1495000 }, { "epoch": 7.04, "learning_rate": 3.931785714285714e-05, "loss": 1.7849, "step": 1495500 }, { "epoch": 7.04, "learning_rate": 3.9314285714285716e-05, "loss": 1.7787, "step": 1496000 }, { "epoch": 7.04, "learning_rate": 3.931071428571429e-05, "loss": 1.7726, "step": 1496500 }, { "epoch": 7.05, "learning_rate": 3.930714285714286e-05, "loss": 1.7819, "step": 1497000 }, { "epoch": 7.05, "learning_rate": 3.9303571428571426e-05, "loss": 1.7864, "step": 1497500 }, { "epoch": 7.05, "learning_rate": 3.9300000000000007e-05, "loss": 1.7806, "step": 1498000 }, { "epoch": 7.05, "learning_rate": 3.9296428571428574e-05, "loss": 1.7805, "step": 1498500 }, { "epoch": 7.06, "learning_rate": 3.929285714285714e-05, "loss": 1.7916, "step": 1499000 }, { "epoch": 7.06, "learning_rate": 3.9289285714285716e-05, "loss": 1.7794, "step": 1499500 }, { "epoch": 7.06, "learning_rate": 3.928571428571429e-05, "loss": 1.7861, "step": 1500000 }, { "epoch": 7.06, "learning_rate": 3.9282142857142864e-05, "loss": 1.7852, "step": 1500500 }, { "epoch": 7.07, "learning_rate": 3.927857142857143e-05, "loss": 1.7811, "step": 1501000 }, { "epoch": 7.07, "learning_rate": 3.9275e-05, "loss": 1.7866, "step": 1501500 }, { "epoch": 7.07, "learning_rate": 3.9271428571428573e-05, "loss": 1.7834, "step": 1502000 }, { "epoch": 7.08, "learning_rate": 3.926785714285715e-05, "loss": 1.791, "step": 1502500 }, { "epoch": 7.08, "learning_rate": 3.9264285714285715e-05, "loss": 1.7833, "step": 1503000 }, { "epoch": 7.08, "learning_rate": 3.926071428571429e-05, "loss": 1.7866, "step": 1503500 }, { "epoch": 7.08, "learning_rate": 3.925714285714286e-05, "loss": 1.7847, "step": 1504000 }, { "epoch": 7.09, "learning_rate": 3.925357142857143e-05, "loss": 1.782, "step": 1504500 }, { "epoch": 7.09, "learning_rate": 3.9250000000000005e-05, "loss": 1.7804, "step": 1505000 }, { "epoch": 7.09, "learning_rate": 3.924642857142857e-05, "loss": 1.7735, "step": 1505500 }, { "epoch": 7.09, "learning_rate": 3.924285714285714e-05, "loss": 1.7719, "step": 1506000 }, { "epoch": 7.1, "learning_rate": 3.9239285714285715e-05, "loss": 1.777, "step": 1506500 }, { "epoch": 7.1, "learning_rate": 3.923571428571429e-05, "loss": 1.7755, "step": 1507000 }, { "epoch": 7.1, "learning_rate": 3.923214285714286e-05, "loss": 1.7836, "step": 1507500 }, { "epoch": 7.1, "learning_rate": 3.922857142857143e-05, "loss": 1.7633, "step": 1508000 }, { "epoch": 7.11, "learning_rate": 3.9225e-05, "loss": 1.7814, "step": 1508500 }, { "epoch": 7.11, "learning_rate": 3.922142857142858e-05, "loss": 1.7817, "step": 1509000 }, { "epoch": 7.11, "learning_rate": 3.9217857142857146e-05, "loss": 1.7786, "step": 1509500 }, { "epoch": 7.12, "learning_rate": 3.9214285714285714e-05, "loss": 1.786, "step": 1510000 }, { "epoch": 7.12, "learning_rate": 3.921071428571429e-05, "loss": 1.7806, "step": 1510500 }, { "epoch": 7.12, "learning_rate": 3.9207142857142856e-05, "loss": 1.775, "step": 1511000 }, { "epoch": 7.12, "learning_rate": 3.920357142857143e-05, "loss": 1.7773, "step": 1511500 }, { "epoch": 7.13, "learning_rate": 3.9200000000000004e-05, "loss": 1.7812, "step": 1512000 }, { "epoch": 7.13, "learning_rate": 3.919642857142857e-05, "loss": 1.7849, "step": 1512500 }, { "epoch": 7.13, "learning_rate": 3.919285714285714e-05, "loss": 1.7715, "step": 1513000 }, { "epoch": 7.13, "learning_rate": 3.918928571428572e-05, "loss": 1.7782, "step": 1513500 }, { "epoch": 7.14, "learning_rate": 3.918571428571429e-05, "loss": 1.7717, "step": 1514000 }, { "epoch": 7.14, "learning_rate": 3.918214285714286e-05, "loss": 1.7773, "step": 1514500 }, { "epoch": 7.14, "learning_rate": 3.917857142857143e-05, "loss": 1.7798, "step": 1515000 }, { "epoch": 7.14, "learning_rate": 3.9175000000000004e-05, "loss": 1.7901, "step": 1515500 }, { "epoch": 7.15, "learning_rate": 3.917142857142858e-05, "loss": 1.7795, "step": 1516000 }, { "epoch": 7.15, "learning_rate": 3.9167857142857145e-05, "loss": 1.7845, "step": 1516500 }, { "epoch": 7.15, "learning_rate": 3.916428571428571e-05, "loss": 1.7811, "step": 1517000 }, { "epoch": 7.16, "learning_rate": 3.916071428571429e-05, "loss": 1.7871, "step": 1517500 }, { "epoch": 7.16, "learning_rate": 3.915714285714286e-05, "loss": 1.7834, "step": 1518000 }, { "epoch": 7.16, "learning_rate": 3.915357142857143e-05, "loss": 1.7739, "step": 1518500 }, { "epoch": 7.16, "learning_rate": 3.915e-05, "loss": 1.7734, "step": 1519000 }, { "epoch": 7.17, "learning_rate": 3.914642857142857e-05, "loss": 1.78, "step": 1519500 }, { "epoch": 7.17, "learning_rate": 3.9142857142857145e-05, "loss": 1.7822, "step": 1520000 }, { "epoch": 7.17, "learning_rate": 3.913928571428572e-05, "loss": 1.7765, "step": 1520500 }, { "epoch": 7.17, "learning_rate": 3.9135714285714286e-05, "loss": 1.788, "step": 1521000 }, { "epoch": 7.18, "learning_rate": 3.913214285714286e-05, "loss": 1.7842, "step": 1521500 }, { "epoch": 7.18, "learning_rate": 3.912857142857143e-05, "loss": 1.7742, "step": 1522000 }, { "epoch": 7.18, "learning_rate": 3.9125e-05, "loss": 1.774, "step": 1522500 }, { "epoch": 7.18, "learning_rate": 3.9121428571428577e-05, "loss": 1.7803, "step": 1523000 }, { "epoch": 7.19, "learning_rate": 3.9117857142857144e-05, "loss": 1.7739, "step": 1523500 }, { "epoch": 7.19, "learning_rate": 3.911428571428571e-05, "loss": 1.7791, "step": 1524000 }, { "epoch": 7.19, "learning_rate": 3.911071428571429e-05, "loss": 1.79, "step": 1524500 }, { "epoch": 7.2, "learning_rate": 3.910714285714286e-05, "loss": 1.7748, "step": 1525000 }, { "epoch": 7.2, "learning_rate": 3.910357142857143e-05, "loss": 1.7794, "step": 1525500 }, { "epoch": 7.2, "learning_rate": 3.91e-05, "loss": 1.782, "step": 1526000 }, { "epoch": 7.2, "learning_rate": 3.9096428571428576e-05, "loss": 1.7778, "step": 1526500 }, { "epoch": 7.21, "learning_rate": 3.9092857142857143e-05, "loss": 1.7737, "step": 1527000 }, { "epoch": 7.21, "learning_rate": 3.908928571428572e-05, "loss": 1.7626, "step": 1527500 }, { "epoch": 7.21, "learning_rate": 3.9085714285714285e-05, "loss": 1.7837, "step": 1528000 }, { "epoch": 7.21, "learning_rate": 3.908214285714286e-05, "loss": 1.7702, "step": 1528500 }, { "epoch": 7.22, "learning_rate": 3.9078571428571434e-05, "loss": 1.7738, "step": 1529000 }, { "epoch": 7.22, "learning_rate": 3.9075e-05, "loss": 1.7787, "step": 1529500 }, { "epoch": 7.22, "learning_rate": 3.9071428571428575e-05, "loss": 1.7646, "step": 1530000 }, { "epoch": 7.22, "learning_rate": 3.906785714285714e-05, "loss": 1.7695, "step": 1530500 }, { "epoch": 7.23, "learning_rate": 3.906428571428572e-05, "loss": 1.7665, "step": 1531000 }, { "epoch": 7.23, "learning_rate": 3.906071428571429e-05, "loss": 1.7681, "step": 1531500 }, { "epoch": 7.23, "learning_rate": 3.905714285714286e-05, "loss": 1.7811, "step": 1532000 }, { "epoch": 7.24, "learning_rate": 3.9053571428571426e-05, "loss": 1.7826, "step": 1532500 }, { "epoch": 7.24, "learning_rate": 3.905e-05, "loss": 1.7725, "step": 1533000 }, { "epoch": 7.24, "learning_rate": 3.9046428571428575e-05, "loss": 1.7716, "step": 1533500 }, { "epoch": 7.24, "learning_rate": 3.904285714285714e-05, "loss": 1.772, "step": 1534000 }, { "epoch": 7.25, "learning_rate": 3.9039285714285716e-05, "loss": 1.7732, "step": 1534500 }, { "epoch": 7.25, "learning_rate": 3.9035714285714284e-05, "loss": 1.7684, "step": 1535000 }, { "epoch": 7.25, "learning_rate": 3.903214285714286e-05, "loss": 1.7667, "step": 1535500 }, { "epoch": 7.25, "learning_rate": 3.902857142857143e-05, "loss": 1.7801, "step": 1536000 }, { "epoch": 7.26, "learning_rate": 3.9025e-05, "loss": 1.7696, "step": 1536500 }, { "epoch": 7.26, "learning_rate": 3.9021428571428574e-05, "loss": 1.7677, "step": 1537000 }, { "epoch": 7.26, "learning_rate": 3.901785714285714e-05, "loss": 1.7636, "step": 1537500 }, { "epoch": 7.26, "learning_rate": 3.9014285714285716e-05, "loss": 1.7858, "step": 1538000 }, { "epoch": 7.27, "learning_rate": 3.901071428571429e-05, "loss": 1.777, "step": 1538500 }, { "epoch": 7.27, "learning_rate": 3.900714285714286e-05, "loss": 1.7735, "step": 1539000 }, { "epoch": 7.27, "learning_rate": 3.9003571428571425e-05, "loss": 1.7721, "step": 1539500 }, { "epoch": 7.28, "learning_rate": 3.9000000000000006e-05, "loss": 1.7777, "step": 1540000 }, { "epoch": 7.28, "learning_rate": 3.8996428571428574e-05, "loss": 1.762, "step": 1540500 }, { "epoch": 7.28, "learning_rate": 3.899285714285714e-05, "loss": 1.7734, "step": 1541000 }, { "epoch": 7.28, "learning_rate": 3.8989285714285715e-05, "loss": 1.7747, "step": 1541500 }, { "epoch": 7.29, "learning_rate": 3.898571428571429e-05, "loss": 1.7702, "step": 1542000 }, { "epoch": 7.29, "learning_rate": 3.8982142857142864e-05, "loss": 1.7573, "step": 1542500 }, { "epoch": 7.29, "learning_rate": 3.897857142857143e-05, "loss": 1.7611, "step": 1543000 }, { "epoch": 7.29, "learning_rate": 3.8975e-05, "loss": 1.774, "step": 1543500 }, { "epoch": 7.3, "learning_rate": 3.897142857142857e-05, "loss": 1.7611, "step": 1544000 }, { "epoch": 7.3, "learning_rate": 3.896785714285715e-05, "loss": 1.7647, "step": 1544500 }, { "epoch": 7.3, "learning_rate": 3.8964285714285715e-05, "loss": 1.7742, "step": 1545000 }, { "epoch": 7.3, "learning_rate": 3.896071428571429e-05, "loss": 1.7649, "step": 1545500 }, { "epoch": 7.31, "learning_rate": 3.8957142857142856e-05, "loss": 1.7745, "step": 1546000 }, { "epoch": 7.31, "learning_rate": 3.895357142857143e-05, "loss": 1.7693, "step": 1546500 }, { "epoch": 7.31, "learning_rate": 3.8950000000000005e-05, "loss": 1.7665, "step": 1547000 }, { "epoch": 7.32, "learning_rate": 3.894642857142857e-05, "loss": 1.7672, "step": 1547500 }, { "epoch": 7.32, "learning_rate": 3.894285714285714e-05, "loss": 1.7717, "step": 1548000 }, { "epoch": 7.32, "learning_rate": 3.8939285714285714e-05, "loss": 1.7646, "step": 1548500 }, { "epoch": 7.32, "learning_rate": 3.893571428571429e-05, "loss": 1.7685, "step": 1549000 }, { "epoch": 7.33, "learning_rate": 3.893214285714286e-05, "loss": 1.7651, "step": 1549500 }, { "epoch": 7.33, "learning_rate": 3.892857142857143e-05, "loss": 1.7652, "step": 1550000 }, { "epoch": 7.33, "learning_rate": 3.8925e-05, "loss": 1.7704, "step": 1550500 }, { "epoch": 7.33, "learning_rate": 3.892142857142858e-05, "loss": 1.7705, "step": 1551000 }, { "epoch": 7.34, "learning_rate": 3.8917857142857146e-05, "loss": 1.7746, "step": 1551500 }, { "epoch": 7.34, "learning_rate": 3.8914285714285713e-05, "loss": 1.767, "step": 1552000 }, { "epoch": 7.34, "learning_rate": 3.891071428571429e-05, "loss": 1.7771, "step": 1552500 }, { "epoch": 7.34, "learning_rate": 3.890714285714286e-05, "loss": 1.764, "step": 1553000 }, { "epoch": 7.35, "learning_rate": 3.890357142857143e-05, "loss": 1.7664, "step": 1553500 }, { "epoch": 7.35, "learning_rate": 3.8900000000000004e-05, "loss": 1.7733, "step": 1554000 }, { "epoch": 7.35, "learning_rate": 3.889642857142857e-05, "loss": 1.7761, "step": 1554500 }, { "epoch": 7.36, "learning_rate": 3.8892857142857145e-05, "loss": 1.7629, "step": 1555000 }, { "epoch": 7.36, "learning_rate": 3.888928571428572e-05, "loss": 1.761, "step": 1555500 }, { "epoch": 7.36, "learning_rate": 3.888571428571429e-05, "loss": 1.7736, "step": 1556000 }, { "epoch": 7.36, "learning_rate": 3.888214285714286e-05, "loss": 1.7634, "step": 1556500 }, { "epoch": 7.37, "learning_rate": 3.887857142857143e-05, "loss": 1.7768, "step": 1557000 }, { "epoch": 7.37, "learning_rate": 3.8875e-05, "loss": 1.7513, "step": 1557500 }, { "epoch": 7.37, "learning_rate": 3.887142857142858e-05, "loss": 1.7642, "step": 1558000 }, { "epoch": 7.37, "learning_rate": 3.8867857142857145e-05, "loss": 1.7605, "step": 1558500 }, { "epoch": 7.38, "learning_rate": 3.886428571428571e-05, "loss": 1.7601, "step": 1559000 }, { "epoch": 7.38, "learning_rate": 3.8860714285714286e-05, "loss": 1.7771, "step": 1559500 }, { "epoch": 7.38, "learning_rate": 3.885714285714286e-05, "loss": 1.7708, "step": 1560000 }, { "epoch": 7.38, "learning_rate": 3.885357142857143e-05, "loss": 1.7626, "step": 1560500 }, { "epoch": 7.39, "learning_rate": 3.885e-05, "loss": 1.7679, "step": 1561000 }, { "epoch": 7.39, "learning_rate": 3.884642857142857e-05, "loss": 1.7551, "step": 1561500 }, { "epoch": 7.39, "learning_rate": 3.8842857142857144e-05, "loss": 1.765, "step": 1562000 }, { "epoch": 7.4, "learning_rate": 3.883928571428572e-05, "loss": 1.7615, "step": 1562500 }, { "epoch": 7.4, "learning_rate": 3.8835714285714286e-05, "loss": 1.7677, "step": 1563000 }, { "epoch": 7.4, "learning_rate": 3.883214285714286e-05, "loss": 1.7674, "step": 1563500 }, { "epoch": 7.4, "learning_rate": 3.882857142857143e-05, "loss": 1.776, "step": 1564000 }, { "epoch": 7.41, "learning_rate": 3.8825e-05, "loss": 1.7585, "step": 1564500 }, { "epoch": 7.41, "learning_rate": 3.8821428571428576e-05, "loss": 1.7545, "step": 1565000 }, { "epoch": 7.41, "learning_rate": 3.8817857142857144e-05, "loss": 1.7654, "step": 1565500 }, { "epoch": 7.41, "learning_rate": 3.881428571428571e-05, "loss": 1.7636, "step": 1566000 }, { "epoch": 7.42, "learning_rate": 3.881071428571429e-05, "loss": 1.7589, "step": 1566500 }, { "epoch": 7.42, "learning_rate": 3.880714285714286e-05, "loss": 1.7713, "step": 1567000 }, { "epoch": 7.42, "learning_rate": 3.880357142857143e-05, "loss": 1.7626, "step": 1567500 }, { "epoch": 7.42, "learning_rate": 3.88e-05, "loss": 1.7535, "step": 1568000 }, { "epoch": 7.43, "learning_rate": 3.8796428571428575e-05, "loss": 1.7577, "step": 1568500 }, { "epoch": 7.43, "learning_rate": 3.879285714285715e-05, "loss": 1.7579, "step": 1569000 }, { "epoch": 7.43, "learning_rate": 3.878928571428572e-05, "loss": 1.7738, "step": 1569500 }, { "epoch": 7.43, "learning_rate": 3.8785714285714285e-05, "loss": 1.7703, "step": 1570000 }, { "epoch": 7.44, "learning_rate": 3.878214285714286e-05, "loss": 1.7641, "step": 1570500 }, { "epoch": 7.44, "learning_rate": 3.877857142857143e-05, "loss": 1.7666, "step": 1571000 }, { "epoch": 7.44, "learning_rate": 3.8775e-05, "loss": 1.7656, "step": 1571500 }, { "epoch": 7.45, "learning_rate": 3.8771428571428575e-05, "loss": 1.7555, "step": 1572000 }, { "epoch": 7.45, "learning_rate": 3.876785714285714e-05, "loss": 1.7589, "step": 1572500 }, { "epoch": 7.45, "learning_rate": 3.8764285714285717e-05, "loss": 1.769, "step": 1573000 }, { "epoch": 7.45, "learning_rate": 3.876071428571429e-05, "loss": 1.7446, "step": 1573500 }, { "epoch": 7.46, "learning_rate": 3.875714285714286e-05, "loss": 1.7597, "step": 1574000 }, { "epoch": 7.46, "learning_rate": 3.8753571428571426e-05, "loss": 1.7635, "step": 1574500 }, { "epoch": 7.46, "learning_rate": 3.875e-05, "loss": 1.7643, "step": 1575000 }, { "epoch": 7.46, "learning_rate": 3.8746428571428574e-05, "loss": 1.7511, "step": 1575500 }, { "epoch": 7.47, "learning_rate": 3.874285714285715e-05, "loss": 1.7584, "step": 1576000 }, { "epoch": 7.47, "learning_rate": 3.8739285714285716e-05, "loss": 1.7618, "step": 1576500 }, { "epoch": 7.47, "learning_rate": 3.8735714285714283e-05, "loss": 1.7691, "step": 1577000 }, { "epoch": 7.47, "learning_rate": 3.8732142857142864e-05, "loss": 1.7603, "step": 1577500 }, { "epoch": 7.48, "learning_rate": 3.872857142857143e-05, "loss": 1.7652, "step": 1578000 }, { "epoch": 7.48, "learning_rate": 3.8725e-05, "loss": 1.7703, "step": 1578500 }, { "epoch": 7.48, "learning_rate": 3.8721428571428574e-05, "loss": 1.7565, "step": 1579000 }, { "epoch": 7.49, "learning_rate": 3.871785714285715e-05, "loss": 1.7611, "step": 1579500 }, { "epoch": 7.49, "learning_rate": 3.8714285714285715e-05, "loss": 1.756, "step": 1580000 }, { "epoch": 7.49, "learning_rate": 3.871071428571429e-05, "loss": 1.7681, "step": 1580500 }, { "epoch": 7.49, "learning_rate": 3.870714285714286e-05, "loss": 1.758, "step": 1581000 }, { "epoch": 7.5, "learning_rate": 3.870357142857143e-05, "loss": 1.7594, "step": 1581500 }, { "epoch": 7.5, "learning_rate": 3.8700000000000006e-05, "loss": 1.7577, "step": 1582000 }, { "epoch": 7.5, "learning_rate": 3.869642857142857e-05, "loss": 1.753, "step": 1582500 }, { "epoch": 7.5, "learning_rate": 3.869285714285715e-05, "loss": 1.7665, "step": 1583000 }, { "epoch": 7.51, "learning_rate": 3.8689285714285715e-05, "loss": 1.756, "step": 1583500 }, { "epoch": 7.51, "learning_rate": 3.868571428571429e-05, "loss": 1.7688, "step": 1584000 }, { "epoch": 7.51, "learning_rate": 3.868214285714286e-05, "loss": 1.7597, "step": 1584500 }, { "epoch": 7.51, "learning_rate": 3.867857142857143e-05, "loss": 1.7542, "step": 1585000 }, { "epoch": 7.52, "learning_rate": 3.8675e-05, "loss": 1.7611, "step": 1585500 }, { "epoch": 7.52, "learning_rate": 3.867142857142857e-05, "loss": 1.7579, "step": 1586000 }, { "epoch": 7.52, "learning_rate": 3.866785714285715e-05, "loss": 1.7623, "step": 1586500 }, { "epoch": 7.53, "learning_rate": 3.8664285714285714e-05, "loss": 1.7649, "step": 1587000 }, { "epoch": 7.53, "learning_rate": 3.866071428571429e-05, "loss": 1.7717, "step": 1587500 }, { "epoch": 7.53, "learning_rate": 3.8657142857142856e-05, "loss": 1.7509, "step": 1588000 }, { "epoch": 7.53, "learning_rate": 3.865357142857143e-05, "loss": 1.7529, "step": 1588500 }, { "epoch": 7.54, "learning_rate": 3.8650000000000004e-05, "loss": 1.7543, "step": 1589000 }, { "epoch": 7.54, "learning_rate": 3.864642857142857e-05, "loss": 1.7562, "step": 1589500 }, { "epoch": 7.54, "learning_rate": 3.8642857142857146e-05, "loss": 1.7674, "step": 1590000 }, { "epoch": 7.54, "learning_rate": 3.8639285714285714e-05, "loss": 1.7586, "step": 1590500 }, { "epoch": 7.55, "learning_rate": 3.863571428571429e-05, "loss": 1.7546, "step": 1591000 }, { "epoch": 7.55, "learning_rate": 3.863214285714286e-05, "loss": 1.7573, "step": 1591500 }, { "epoch": 7.55, "learning_rate": 3.862857142857143e-05, "loss": 1.7568, "step": 1592000 }, { "epoch": 7.55, "learning_rate": 3.8625e-05, "loss": 1.7542, "step": 1592500 }, { "epoch": 7.56, "learning_rate": 3.862142857142858e-05, "loss": 1.7596, "step": 1593000 }, { "epoch": 7.56, "learning_rate": 3.8617857142857145e-05, "loss": 1.7605, "step": 1593500 }, { "epoch": 7.56, "learning_rate": 3.861428571428571e-05, "loss": 1.7499, "step": 1594000 }, { "epoch": 7.57, "learning_rate": 3.861071428571429e-05, "loss": 1.7555, "step": 1594500 }, { "epoch": 7.57, "learning_rate": 3.860714285714286e-05, "loss": 1.766, "step": 1595000 }, { "epoch": 7.57, "learning_rate": 3.860357142857143e-05, "loss": 1.75, "step": 1595500 }, { "epoch": 7.57, "learning_rate": 3.86e-05, "loss": 1.7504, "step": 1596000 }, { "epoch": 7.58, "learning_rate": 3.859642857142857e-05, "loss": 1.7632, "step": 1596500 }, { "epoch": 7.58, "learning_rate": 3.8592857142857145e-05, "loss": 1.7593, "step": 1597000 }, { "epoch": 7.58, "learning_rate": 3.858928571428572e-05, "loss": 1.7556, "step": 1597500 }, { "epoch": 7.58, "learning_rate": 3.8585714285714287e-05, "loss": 1.7594, "step": 1598000 }, { "epoch": 7.59, "learning_rate": 3.858214285714286e-05, "loss": 1.7585, "step": 1598500 }, { "epoch": 7.59, "learning_rate": 3.857857142857143e-05, "loss": 1.7523, "step": 1599000 }, { "epoch": 7.59, "learning_rate": 3.8575e-05, "loss": 1.7539, "step": 1599500 }, { "epoch": 7.59, "learning_rate": 3.857142857142858e-05, "loss": 1.7536, "step": 1600000 }, { "epoch": 7.6, "learning_rate": 3.8567857142857144e-05, "loss": 1.7442, "step": 1600500 }, { "epoch": 7.6, "learning_rate": 3.856428571428571e-05, "loss": 1.7564, "step": 1601000 }, { "epoch": 7.6, "learning_rate": 3.8560714285714286e-05, "loss": 1.755, "step": 1601500 }, { "epoch": 7.61, "learning_rate": 3.855714285714286e-05, "loss": 1.7494, "step": 1602000 }, { "epoch": 7.61, "learning_rate": 3.855357142857143e-05, "loss": 1.7559, "step": 1602500 }, { "epoch": 7.61, "learning_rate": 3.855e-05, "loss": 1.7589, "step": 1603000 }, { "epoch": 7.61, "learning_rate": 3.854642857142857e-05, "loss": 1.7504, "step": 1603500 }, { "epoch": 7.62, "learning_rate": 3.854285714285715e-05, "loss": 1.7464, "step": 1604000 }, { "epoch": 7.62, "learning_rate": 3.853928571428572e-05, "loss": 1.7519, "step": 1604500 }, { "epoch": 7.62, "learning_rate": 3.8535714285714285e-05, "loss": 1.7565, "step": 1605000 }, { "epoch": 7.62, "learning_rate": 3.853214285714286e-05, "loss": 1.7532, "step": 1605500 }, { "epoch": 7.63, "learning_rate": 3.8528571428571434e-05, "loss": 1.752, "step": 1606000 }, { "epoch": 7.63, "learning_rate": 3.8525e-05, "loss": 1.7599, "step": 1606500 }, { "epoch": 7.63, "learning_rate": 3.8521428571428576e-05, "loss": 1.7603, "step": 1607000 }, { "epoch": 7.63, "learning_rate": 3.851785714285714e-05, "loss": 1.7542, "step": 1607500 }, { "epoch": 7.64, "learning_rate": 3.851428571428571e-05, "loss": 1.747, "step": 1608000 }, { "epoch": 7.64, "learning_rate": 3.851071428571429e-05, "loss": 1.7433, "step": 1608500 }, { "epoch": 7.64, "learning_rate": 3.850714285714286e-05, "loss": 1.7486, "step": 1609000 }, { "epoch": 7.65, "learning_rate": 3.8503571428571426e-05, "loss": 1.7583, "step": 1609500 }, { "epoch": 7.65, "learning_rate": 3.85e-05, "loss": 1.7503, "step": 1610000 }, { "epoch": 7.65, "learning_rate": 3.8496428571428575e-05, "loss": 1.7402, "step": 1610500 }, { "epoch": 7.65, "learning_rate": 3.849285714285715e-05, "loss": 1.7434, "step": 1611000 }, { "epoch": 7.66, "learning_rate": 3.848928571428572e-05, "loss": 1.7513, "step": 1611500 }, { "epoch": 7.66, "learning_rate": 3.8485714285714284e-05, "loss": 1.7552, "step": 1612000 }, { "epoch": 7.66, "learning_rate": 3.848214285714286e-05, "loss": 1.7511, "step": 1612500 }, { "epoch": 7.66, "learning_rate": 3.847857142857143e-05, "loss": 1.751, "step": 1613000 }, { "epoch": 7.67, "learning_rate": 3.8475e-05, "loss": 1.7448, "step": 1613500 }, { "epoch": 7.67, "learning_rate": 3.8471428571428574e-05, "loss": 1.7569, "step": 1614000 }, { "epoch": 7.67, "learning_rate": 3.846785714285714e-05, "loss": 1.7521, "step": 1614500 }, { "epoch": 7.67, "learning_rate": 3.8464285714285716e-05, "loss": 1.7551, "step": 1615000 }, { "epoch": 7.68, "learning_rate": 3.846071428571429e-05, "loss": 1.7418, "step": 1615500 }, { "epoch": 7.68, "learning_rate": 3.845714285714286e-05, "loss": 1.7534, "step": 1616000 }, { "epoch": 7.68, "learning_rate": 3.8453571428571425e-05, "loss": 1.7429, "step": 1616500 }, { "epoch": 7.69, "learning_rate": 3.845e-05, "loss": 1.7427, "step": 1617000 }, { "epoch": 7.69, "learning_rate": 3.8446428571428574e-05, "loss": 1.75, "step": 1617500 }, { "epoch": 7.69, "learning_rate": 3.844285714285715e-05, "loss": 1.7629, "step": 1618000 }, { "epoch": 7.69, "learning_rate": 3.8439285714285715e-05, "loss": 1.7504, "step": 1618500 }, { "epoch": 7.7, "learning_rate": 3.843571428571428e-05, "loss": 1.7449, "step": 1619000 }, { "epoch": 7.7, "learning_rate": 3.8432142857142864e-05, "loss": 1.747, "step": 1619500 }, { "epoch": 7.7, "learning_rate": 3.842857142857143e-05, "loss": 1.7611, "step": 1620000 }, { "epoch": 7.7, "learning_rate": 3.8425e-05, "loss": 1.748, "step": 1620500 }, { "epoch": 7.71, "learning_rate": 3.842142857142857e-05, "loss": 1.7478, "step": 1621000 }, { "epoch": 7.71, "learning_rate": 3.841785714285715e-05, "loss": 1.7559, "step": 1621500 }, { "epoch": 7.71, "learning_rate": 3.8414285714285715e-05, "loss": 1.7385, "step": 1622000 }, { "epoch": 7.71, "learning_rate": 3.841071428571429e-05, "loss": 1.7487, "step": 1622500 }, { "epoch": 7.72, "learning_rate": 3.8407142857142857e-05, "loss": 1.7501, "step": 1623000 }, { "epoch": 7.72, "learning_rate": 3.840357142857143e-05, "loss": 1.7387, "step": 1623500 }, { "epoch": 7.72, "learning_rate": 3.8400000000000005e-05, "loss": 1.7526, "step": 1624000 }, { "epoch": 7.73, "learning_rate": 3.839642857142857e-05, "loss": 1.7563, "step": 1624500 }, { "epoch": 7.73, "learning_rate": 3.839285714285715e-05, "loss": 1.746, "step": 1625000 }, { "epoch": 7.73, "learning_rate": 3.8389285714285714e-05, "loss": 1.7473, "step": 1625500 }, { "epoch": 7.73, "learning_rate": 3.838571428571429e-05, "loss": 1.7576, "step": 1626000 }, { "epoch": 7.74, "learning_rate": 3.838214285714286e-05, "loss": 1.7509, "step": 1626500 }, { "epoch": 7.74, "learning_rate": 3.837857142857143e-05, "loss": 1.7552, "step": 1627000 }, { "epoch": 7.74, "learning_rate": 3.8375e-05, "loss": 1.7444, "step": 1627500 }, { "epoch": 7.74, "learning_rate": 3.837142857142857e-05, "loss": 1.7529, "step": 1628000 }, { "epoch": 7.75, "learning_rate": 3.8367857142857146e-05, "loss": 1.7418, "step": 1628500 }, { "epoch": 7.75, "learning_rate": 3.8364285714285714e-05, "loss": 1.7461, "step": 1629000 }, { "epoch": 7.75, "learning_rate": 3.836071428571429e-05, "loss": 1.7445, "step": 1629500 }, { "epoch": 7.75, "learning_rate": 3.8357142857142855e-05, "loss": 1.7571, "step": 1630000 }, { "epoch": 7.76, "learning_rate": 3.8353571428571436e-05, "loss": 1.7464, "step": 1630500 }, { "epoch": 7.76, "learning_rate": 3.8350000000000004e-05, "loss": 1.7459, "step": 1631000 }, { "epoch": 7.76, "learning_rate": 3.834642857142857e-05, "loss": 1.7489, "step": 1631500 }, { "epoch": 7.77, "learning_rate": 3.8342857142857146e-05, "loss": 1.7427, "step": 1632000 }, { "epoch": 7.77, "learning_rate": 3.833928571428572e-05, "loss": 1.7524, "step": 1632500 }, { "epoch": 7.77, "learning_rate": 3.833571428571429e-05, "loss": 1.7576, "step": 1633000 }, { "epoch": 7.77, "learning_rate": 3.833214285714286e-05, "loss": 1.7405, "step": 1633500 }, { "epoch": 7.78, "learning_rate": 3.832857142857143e-05, "loss": 1.7445, "step": 1634000 }, { "epoch": 7.78, "learning_rate": 3.8324999999999996e-05, "loss": 1.7416, "step": 1634500 }, { "epoch": 7.78, "learning_rate": 3.832142857142858e-05, "loss": 1.755, "step": 1635000 }, { "epoch": 7.78, "learning_rate": 3.8317857142857145e-05, "loss": 1.7458, "step": 1635500 }, { "epoch": 7.79, "learning_rate": 3.831428571428571e-05, "loss": 1.7531, "step": 1636000 }, { "epoch": 7.79, "learning_rate": 3.831071428571429e-05, "loss": 1.7431, "step": 1636500 }, { "epoch": 7.79, "learning_rate": 3.830714285714286e-05, "loss": 1.7435, "step": 1637000 }, { "epoch": 7.79, "learning_rate": 3.8303571428571435e-05, "loss": 1.7483, "step": 1637500 }, { "epoch": 7.8, "learning_rate": 3.83e-05, "loss": 1.7492, "step": 1638000 }, { "epoch": 7.8, "learning_rate": 3.829642857142857e-05, "loss": 1.7437, "step": 1638500 }, { "epoch": 7.8, "learning_rate": 3.8292857142857144e-05, "loss": 1.746, "step": 1639000 }, { "epoch": 7.81, "learning_rate": 3.828928571428572e-05, "loss": 1.7304, "step": 1639500 }, { "epoch": 7.81, "learning_rate": 3.8285714285714286e-05, "loss": 1.7373, "step": 1640000 }, { "epoch": 7.81, "learning_rate": 3.828214285714286e-05, "loss": 1.7467, "step": 1640500 }, { "epoch": 7.81, "learning_rate": 3.827857142857143e-05, "loss": 1.7406, "step": 1641000 }, { "epoch": 7.82, "learning_rate": 3.8275e-05, "loss": 1.7473, "step": 1641500 }, { "epoch": 7.82, "learning_rate": 3.8271428571428576e-05, "loss": 1.7369, "step": 1642000 }, { "epoch": 7.82, "learning_rate": 3.8267857142857144e-05, "loss": 1.7494, "step": 1642500 }, { "epoch": 7.82, "learning_rate": 3.826428571428571e-05, "loss": 1.7521, "step": 1643000 }, { "epoch": 7.83, "learning_rate": 3.8260714285714285e-05, "loss": 1.7407, "step": 1643500 }, { "epoch": 7.83, "learning_rate": 3.825714285714286e-05, "loss": 1.7403, "step": 1644000 }, { "epoch": 7.83, "learning_rate": 3.8253571428571434e-05, "loss": 1.7411, "step": 1644500 }, { "epoch": 7.83, "learning_rate": 3.825e-05, "loss": 1.7466, "step": 1645000 }, { "epoch": 7.84, "learning_rate": 3.824642857142857e-05, "loss": 1.7495, "step": 1645500 }, { "epoch": 7.84, "learning_rate": 3.824285714285715e-05, "loss": 1.7397, "step": 1646000 }, { "epoch": 7.84, "learning_rate": 3.823928571428572e-05, "loss": 1.7432, "step": 1646500 }, { "epoch": 7.84, "learning_rate": 3.8235714285714285e-05, "loss": 1.7431, "step": 1647000 }, { "epoch": 7.85, "learning_rate": 3.823214285714286e-05, "loss": 1.7505, "step": 1647500 }, { "epoch": 7.85, "learning_rate": 3.822857142857143e-05, "loss": 1.756, "step": 1648000 }, { "epoch": 7.85, "learning_rate": 3.8225e-05, "loss": 1.7472, "step": 1648500 }, { "epoch": 7.86, "learning_rate": 3.8221428571428575e-05, "loss": 1.7389, "step": 1649000 }, { "epoch": 7.86, "learning_rate": 3.821785714285714e-05, "loss": 1.7461, "step": 1649500 }, { "epoch": 7.86, "learning_rate": 3.821428571428572e-05, "loss": 1.7235, "step": 1650000 }, { "epoch": 7.86, "learning_rate": 3.821071428571429e-05, "loss": 1.74, "step": 1650500 }, { "epoch": 7.87, "learning_rate": 3.820714285714286e-05, "loss": 1.7399, "step": 1651000 }, { "epoch": 7.87, "learning_rate": 3.820357142857143e-05, "loss": 1.7386, "step": 1651500 }, { "epoch": 7.87, "learning_rate": 3.82e-05, "loss": 1.7433, "step": 1652000 }, { "epoch": 7.87, "learning_rate": 3.8196428571428574e-05, "loss": 1.7427, "step": 1652500 }, { "epoch": 7.88, "learning_rate": 3.819285714285715e-05, "loss": 1.7448, "step": 1653000 }, { "epoch": 7.88, "learning_rate": 3.8189285714285716e-05, "loss": 1.7367, "step": 1653500 }, { "epoch": 7.88, "learning_rate": 3.8185714285714284e-05, "loss": 1.7391, "step": 1654000 }, { "epoch": 7.88, "learning_rate": 3.818214285714286e-05, "loss": 1.7432, "step": 1654500 }, { "epoch": 7.89, "learning_rate": 3.817857142857143e-05, "loss": 1.7234, "step": 1655000 }, { "epoch": 7.89, "learning_rate": 3.8175e-05, "loss": 1.7286, "step": 1655500 }, { "epoch": 7.89, "learning_rate": 3.8171428571428574e-05, "loss": 1.7369, "step": 1656000 }, { "epoch": 7.9, "learning_rate": 3.816785714285714e-05, "loss": 1.7518, "step": 1656500 }, { "epoch": 7.9, "learning_rate": 3.8164285714285716e-05, "loss": 1.7302, "step": 1657000 }, { "epoch": 7.9, "learning_rate": 3.816071428571429e-05, "loss": 1.7327, "step": 1657500 }, { "epoch": 7.9, "learning_rate": 3.815714285714286e-05, "loss": 1.7369, "step": 1658000 }, { "epoch": 7.91, "learning_rate": 3.815357142857143e-05, "loss": 1.7368, "step": 1658500 }, { "epoch": 7.91, "learning_rate": 3.8150000000000006e-05, "loss": 1.7321, "step": 1659000 }, { "epoch": 7.91, "learning_rate": 3.814642857142857e-05, "loss": 1.7383, "step": 1659500 }, { "epoch": 7.91, "learning_rate": 3.814285714285715e-05, "loss": 1.7439, "step": 1660000 }, { "epoch": 7.92, "learning_rate": 3.8139285714285715e-05, "loss": 1.7308, "step": 1660500 }, { "epoch": 7.92, "learning_rate": 3.813571428571428e-05, "loss": 1.7478, "step": 1661000 }, { "epoch": 7.92, "learning_rate": 3.8132142857142863e-05, "loss": 1.7369, "step": 1661500 }, { "epoch": 7.92, "learning_rate": 3.812857142857143e-05, "loss": 1.7317, "step": 1662000 }, { "epoch": 7.93, "learning_rate": 3.8125e-05, "loss": 1.738, "step": 1662500 }, { "epoch": 7.93, "learning_rate": 3.812142857142857e-05, "loss": 1.7525, "step": 1663000 }, { "epoch": 7.93, "learning_rate": 3.811785714285715e-05, "loss": 1.7496, "step": 1663500 }, { "epoch": 7.94, "learning_rate": 3.8114285714285714e-05, "loss": 1.7478, "step": 1664000 }, { "epoch": 7.94, "learning_rate": 3.811071428571429e-05, "loss": 1.7509, "step": 1664500 }, { "epoch": 7.94, "learning_rate": 3.8107142857142856e-05, "loss": 1.7322, "step": 1665000 }, { "epoch": 7.94, "learning_rate": 3.810357142857143e-05, "loss": 1.733, "step": 1665500 }, { "epoch": 7.95, "learning_rate": 3.8100000000000005e-05, "loss": 1.7289, "step": 1666000 }, { "epoch": 7.95, "learning_rate": 3.809642857142857e-05, "loss": 1.7432, "step": 1666500 }, { "epoch": 7.95, "learning_rate": 3.8092857142857146e-05, "loss": 1.7354, "step": 1667000 }, { "epoch": 7.95, "learning_rate": 3.8089285714285714e-05, "loss": 1.7319, "step": 1667500 }, { "epoch": 7.96, "learning_rate": 3.808571428571429e-05, "loss": 1.7341, "step": 1668000 }, { "epoch": 7.96, "learning_rate": 3.808214285714286e-05, "loss": 1.7276, "step": 1668500 }, { "epoch": 7.96, "learning_rate": 3.807857142857143e-05, "loss": 1.731, "step": 1669000 }, { "epoch": 7.96, "learning_rate": 3.8075e-05, "loss": 1.7343, "step": 1669500 }, { "epoch": 7.97, "learning_rate": 3.807142857142857e-05, "loss": 1.731, "step": 1670000 }, { "epoch": 7.97, "learning_rate": 3.8067857142857146e-05, "loss": 1.7296, "step": 1670500 }, { "epoch": 7.97, "learning_rate": 3.806428571428571e-05, "loss": 1.7323, "step": 1671000 }, { "epoch": 7.98, "learning_rate": 3.806071428571429e-05, "loss": 1.7342, "step": 1671500 }, { "epoch": 7.98, "learning_rate": 3.8057142857142855e-05, "loss": 1.7351, "step": 1672000 }, { "epoch": 7.98, "learning_rate": 3.8053571428571436e-05, "loss": 1.7359, "step": 1672500 }, { "epoch": 7.98, "learning_rate": 3.805e-05, "loss": 1.7378, "step": 1673000 }, { "epoch": 7.99, "learning_rate": 3.804642857142857e-05, "loss": 1.7373, "step": 1673500 }, { "epoch": 7.99, "learning_rate": 3.8042857142857145e-05, "loss": 1.7347, "step": 1674000 }, { "epoch": 7.99, "learning_rate": 3.803928571428572e-05, "loss": 1.7367, "step": 1674500 }, { "epoch": 7.99, "learning_rate": 3.803571428571429e-05, "loss": 1.7404, "step": 1675000 }, { "epoch": 8.0, "learning_rate": 3.803214285714286e-05, "loss": 1.7433, "step": 1675500 }, { "epoch": 8.0, "learning_rate": 3.802857142857143e-05, "loss": 1.742, "step": 1676000 }, { "epoch": 8.0, "learning_rate": 3.8025e-05, "loss": 1.7432, "step": 1676500 }, { "epoch": 8.0, "learning_rate": 3.802142857142858e-05, "loss": 1.7332, "step": 1677000 }, { "epoch": 8.01, "learning_rate": 3.8017857142857144e-05, "loss": 1.7251, "step": 1677500 }, { "epoch": 8.01, "learning_rate": 3.801428571428571e-05, "loss": 1.7262, "step": 1678000 }, { "epoch": 8.01, "learning_rate": 3.8010714285714286e-05, "loss": 1.7255, "step": 1678500 }, { "epoch": 8.02, "learning_rate": 3.800714285714286e-05, "loss": 1.7319, "step": 1679000 }, { "epoch": 8.02, "learning_rate": 3.8003571428571435e-05, "loss": 1.7356, "step": 1679500 }, { "epoch": 8.02, "learning_rate": 3.8e-05, "loss": 1.7294, "step": 1680000 }, { "epoch": 8.02, "learning_rate": 3.799642857142857e-05, "loss": 1.7315, "step": 1680500 }, { "epoch": 8.03, "learning_rate": 3.7992857142857144e-05, "loss": 1.7326, "step": 1681000 }, { "epoch": 8.03, "learning_rate": 3.798928571428572e-05, "loss": 1.72, "step": 1681500 }, { "epoch": 8.03, "learning_rate": 3.7985714285714286e-05, "loss": 1.7261, "step": 1682000 }, { "epoch": 8.03, "learning_rate": 3.798214285714286e-05, "loss": 1.7279, "step": 1682500 }, { "epoch": 8.04, "learning_rate": 3.797857142857143e-05, "loss": 1.7396, "step": 1683000 }, { "epoch": 8.04, "learning_rate": 3.7975e-05, "loss": 1.7204, "step": 1683500 }, { "epoch": 8.04, "learning_rate": 3.7971428571428576e-05, "loss": 1.7254, "step": 1684000 }, { "epoch": 8.04, "learning_rate": 3.796785714285714e-05, "loss": 1.7233, "step": 1684500 }, { "epoch": 8.05, "learning_rate": 3.796428571428571e-05, "loss": 1.7158, "step": 1685000 }, { "epoch": 8.05, "learning_rate": 3.796071428571429e-05, "loss": 1.7315, "step": 1685500 }, { "epoch": 8.05, "learning_rate": 3.795714285714286e-05, "loss": 1.7349, "step": 1686000 }, { "epoch": 8.06, "learning_rate": 3.7953571428571433e-05, "loss": 1.7415, "step": 1686500 }, { "epoch": 8.06, "learning_rate": 3.795e-05, "loss": 1.7343, "step": 1687000 }, { "epoch": 8.06, "learning_rate": 3.794642857142857e-05, "loss": 1.7362, "step": 1687500 }, { "epoch": 8.06, "learning_rate": 3.794285714285715e-05, "loss": 1.7288, "step": 1688000 }, { "epoch": 8.07, "learning_rate": 3.793928571428572e-05, "loss": 1.7234, "step": 1688500 }, { "epoch": 8.07, "learning_rate": 3.7935714285714284e-05, "loss": 1.7234, "step": 1689000 }, { "epoch": 8.07, "learning_rate": 3.793214285714286e-05, "loss": 1.7307, "step": 1689500 }, { "epoch": 8.07, "learning_rate": 3.792857142857143e-05, "loss": 1.727, "step": 1690000 }, { "epoch": 8.08, "learning_rate": 3.7925e-05, "loss": 1.7261, "step": 1690500 }, { "epoch": 8.08, "learning_rate": 3.7921428571428575e-05, "loss": 1.7157, "step": 1691000 }, { "epoch": 8.08, "learning_rate": 3.791785714285714e-05, "loss": 1.7207, "step": 1691500 }, { "epoch": 8.08, "learning_rate": 3.7914285714285716e-05, "loss": 1.7265, "step": 1692000 }, { "epoch": 8.09, "learning_rate": 3.791071428571429e-05, "loss": 1.7224, "step": 1692500 }, { "epoch": 8.09, "learning_rate": 3.790714285714286e-05, "loss": 1.7286, "step": 1693000 }, { "epoch": 8.09, "learning_rate": 3.790357142857143e-05, "loss": 1.7204, "step": 1693500 }, { "epoch": 8.1, "learning_rate": 3.79e-05, "loss": 1.7211, "step": 1694000 }, { "epoch": 8.1, "learning_rate": 3.7896428571428574e-05, "loss": 1.729, "step": 1694500 }, { "epoch": 8.1, "learning_rate": 3.789285714285715e-05, "loss": 1.723, "step": 1695000 }, { "epoch": 8.1, "learning_rate": 3.7889285714285716e-05, "loss": 1.7379, "step": 1695500 }, { "epoch": 8.11, "learning_rate": 3.788571428571428e-05, "loss": 1.7176, "step": 1696000 }, { "epoch": 8.11, "learning_rate": 3.788214285714286e-05, "loss": 1.7246, "step": 1696500 }, { "epoch": 8.11, "learning_rate": 3.787857142857143e-05, "loss": 1.7257, "step": 1697000 }, { "epoch": 8.11, "learning_rate": 3.7875e-05, "loss": 1.7161, "step": 1697500 }, { "epoch": 8.12, "learning_rate": 3.787142857142857e-05, "loss": 1.7263, "step": 1698000 }, { "epoch": 8.12, "learning_rate": 3.786785714285714e-05, "loss": 1.7364, "step": 1698500 }, { "epoch": 8.12, "learning_rate": 3.786428571428572e-05, "loss": 1.724, "step": 1699000 }, { "epoch": 8.12, "learning_rate": 3.786071428571429e-05, "loss": 1.7249, "step": 1699500 }, { "epoch": 8.13, "learning_rate": 3.785714285714286e-05, "loss": 1.7414, "step": 1700000 }, { "epoch": 8.13, "learning_rate": 3.785357142857143e-05, "loss": 1.715, "step": 1700500 }, { "epoch": 8.13, "learning_rate": 3.7850000000000005e-05, "loss": 1.7197, "step": 1701000 }, { "epoch": 8.14, "learning_rate": 3.784642857142857e-05, "loss": 1.7172, "step": 1701500 }, { "epoch": 8.14, "learning_rate": 3.784285714285715e-05, "loss": 1.7203, "step": 1702000 }, { "epoch": 8.14, "learning_rate": 3.7839285714285714e-05, "loss": 1.7366, "step": 1702500 }, { "epoch": 8.14, "learning_rate": 3.783571428571429e-05, "loss": 1.7198, "step": 1703000 }, { "epoch": 8.15, "learning_rate": 3.783214285714286e-05, "loss": 1.7203, "step": 1703500 }, { "epoch": 8.15, "learning_rate": 3.782857142857143e-05, "loss": 1.7261, "step": 1704000 }, { "epoch": 8.15, "learning_rate": 3.7825e-05, "loss": 1.7246, "step": 1704500 }, { "epoch": 8.15, "learning_rate": 3.782142857142857e-05, "loss": 1.7206, "step": 1705000 }, { "epoch": 8.16, "learning_rate": 3.7817857142857146e-05, "loss": 1.7288, "step": 1705500 }, { "epoch": 8.16, "learning_rate": 3.781428571428572e-05, "loss": 1.7211, "step": 1706000 }, { "epoch": 8.16, "learning_rate": 3.781071428571429e-05, "loss": 1.7379, "step": 1706500 }, { "epoch": 8.16, "learning_rate": 3.7807142857142856e-05, "loss": 1.7322, "step": 1707000 }, { "epoch": 8.17, "learning_rate": 3.780357142857143e-05, "loss": 1.7241, "step": 1707500 }, { "epoch": 8.17, "learning_rate": 3.7800000000000004e-05, "loss": 1.7303, "step": 1708000 }, { "epoch": 8.17, "learning_rate": 3.779642857142857e-05, "loss": 1.7251, "step": 1708500 }, { "epoch": 8.18, "learning_rate": 3.7792857142857146e-05, "loss": 1.7275, "step": 1709000 }, { "epoch": 8.18, "learning_rate": 3.778928571428571e-05, "loss": 1.7261, "step": 1709500 }, { "epoch": 8.18, "learning_rate": 3.778571428571429e-05, "loss": 1.7244, "step": 1710000 }, { "epoch": 8.18, "learning_rate": 3.778214285714286e-05, "loss": 1.7283, "step": 1710500 }, { "epoch": 8.19, "learning_rate": 3.777857142857143e-05, "loss": 1.7217, "step": 1711000 }, { "epoch": 8.19, "learning_rate": 3.7775e-05, "loss": 1.7207, "step": 1711500 }, { "epoch": 8.19, "learning_rate": 3.777142857142858e-05, "loss": 1.7244, "step": 1712000 }, { "epoch": 8.19, "learning_rate": 3.7767857142857145e-05, "loss": 1.7212, "step": 1712500 }, { "epoch": 8.2, "learning_rate": 3.776428571428572e-05, "loss": 1.7355, "step": 1713000 }, { "epoch": 8.2, "learning_rate": 3.776071428571429e-05, "loss": 1.7306, "step": 1713500 }, { "epoch": 8.2, "learning_rate": 3.7757142857142854e-05, "loss": 1.7295, "step": 1714000 }, { "epoch": 8.2, "learning_rate": 3.7753571428571435e-05, "loss": 1.7146, "step": 1714500 }, { "epoch": 8.21, "learning_rate": 3.775e-05, "loss": 1.7154, "step": 1715000 }, { "epoch": 8.21, "learning_rate": 3.774642857142857e-05, "loss": 1.7217, "step": 1715500 }, { "epoch": 8.21, "learning_rate": 3.7742857142857145e-05, "loss": 1.7226, "step": 1716000 }, { "epoch": 8.21, "learning_rate": 3.773928571428572e-05, "loss": 1.7386, "step": 1716500 }, { "epoch": 8.22, "learning_rate": 3.7735714285714286e-05, "loss": 1.7148, "step": 1717000 }, { "epoch": 8.22, "learning_rate": 3.773214285714286e-05, "loss": 1.7178, "step": 1717500 }, { "epoch": 8.22, "learning_rate": 3.772857142857143e-05, "loss": 1.7369, "step": 1718000 }, { "epoch": 8.23, "learning_rate": 3.7725e-05, "loss": 1.7153, "step": 1718500 }, { "epoch": 8.23, "learning_rate": 3.7721428571428576e-05, "loss": 1.7244, "step": 1719000 }, { "epoch": 8.23, "learning_rate": 3.7717857142857144e-05, "loss": 1.7149, "step": 1719500 }, { "epoch": 8.23, "learning_rate": 3.771428571428572e-05, "loss": 1.7226, "step": 1720000 }, { "epoch": 8.24, "learning_rate": 3.7710714285714286e-05, "loss": 1.7141, "step": 1720500 }, { "epoch": 8.24, "learning_rate": 3.770714285714286e-05, "loss": 1.7236, "step": 1721000 }, { "epoch": 8.24, "learning_rate": 3.7703571428571434e-05, "loss": 1.7259, "step": 1721500 }, { "epoch": 8.24, "learning_rate": 3.77e-05, "loss": 1.7226, "step": 1722000 }, { "epoch": 8.25, "learning_rate": 3.769642857142857e-05, "loss": 1.7252, "step": 1722500 }, { "epoch": 8.25, "learning_rate": 3.769285714285714e-05, "loss": 1.7101, "step": 1723000 }, { "epoch": 8.25, "learning_rate": 3.768928571428572e-05, "loss": 1.7233, "step": 1723500 }, { "epoch": 8.25, "learning_rate": 3.7685714285714285e-05, "loss": 1.7285, "step": 1724000 }, { "epoch": 8.26, "learning_rate": 3.768214285714286e-05, "loss": 1.7199, "step": 1724500 }, { "epoch": 8.26, "learning_rate": 3.767857142857143e-05, "loss": 1.7142, "step": 1725000 }, { "epoch": 8.26, "learning_rate": 3.7675e-05, "loss": 1.7245, "step": 1725500 }, { "epoch": 8.27, "learning_rate": 3.7671428571428575e-05, "loss": 1.716, "step": 1726000 }, { "epoch": 8.27, "learning_rate": 3.766785714285714e-05, "loss": 1.7252, "step": 1726500 }, { "epoch": 8.27, "learning_rate": 3.766428571428572e-05, "loss": 1.7254, "step": 1727000 }, { "epoch": 8.27, "learning_rate": 3.766071428571429e-05, "loss": 1.7131, "step": 1727500 }, { "epoch": 8.28, "learning_rate": 3.765714285714286e-05, "loss": 1.7128, "step": 1728000 }, { "epoch": 8.28, "learning_rate": 3.765357142857143e-05, "loss": 1.7341, "step": 1728500 }, { "epoch": 8.28, "learning_rate": 3.765e-05, "loss": 1.7156, "step": 1729000 }, { "epoch": 8.28, "learning_rate": 3.7646428571428575e-05, "loss": 1.7312, "step": 1729500 }, { "epoch": 8.29, "learning_rate": 3.764285714285715e-05, "loss": 1.7182, "step": 1730000 }, { "epoch": 8.29, "learning_rate": 3.7639285714285716e-05, "loss": 1.7277, "step": 1730500 }, { "epoch": 8.29, "learning_rate": 3.7635714285714284e-05, "loss": 1.7219, "step": 1731000 }, { "epoch": 8.29, "learning_rate": 3.763214285714286e-05, "loss": 1.7167, "step": 1731500 }, { "epoch": 8.3, "learning_rate": 3.762857142857143e-05, "loss": 1.7229, "step": 1732000 }, { "epoch": 8.3, "learning_rate": 3.7625e-05, "loss": 1.72, "step": 1732500 }, { "epoch": 8.3, "learning_rate": 3.7621428571428574e-05, "loss": 1.7312, "step": 1733000 }, { "epoch": 8.31, "learning_rate": 3.761785714285714e-05, "loss": 1.7149, "step": 1733500 }, { "epoch": 8.31, "learning_rate": 3.7614285714285716e-05, "loss": 1.72, "step": 1734000 }, { "epoch": 8.31, "learning_rate": 3.761071428571429e-05, "loss": 1.7215, "step": 1734500 }, { "epoch": 8.31, "learning_rate": 3.760714285714286e-05, "loss": 1.7158, "step": 1735000 }, { "epoch": 8.32, "learning_rate": 3.760357142857143e-05, "loss": 1.7242, "step": 1735500 }, { "epoch": 8.32, "learning_rate": 3.76e-05, "loss": 1.7131, "step": 1736000 }, { "epoch": 8.32, "learning_rate": 3.7596428571428573e-05, "loss": 1.7183, "step": 1736500 }, { "epoch": 8.32, "learning_rate": 3.759285714285715e-05, "loss": 1.7242, "step": 1737000 }, { "epoch": 8.33, "learning_rate": 3.7589285714285715e-05, "loss": 1.7145, "step": 1737500 }, { "epoch": 8.33, "learning_rate": 3.758571428571428e-05, "loss": 1.7153, "step": 1738000 }, { "epoch": 8.33, "learning_rate": 3.758214285714286e-05, "loss": 1.7266, "step": 1738500 }, { "epoch": 8.33, "learning_rate": 3.757857142857143e-05, "loss": 1.7243, "step": 1739000 }, { "epoch": 8.34, "learning_rate": 3.7575e-05, "loss": 1.7177, "step": 1739500 }, { "epoch": 8.34, "learning_rate": 3.757142857142857e-05, "loss": 1.7184, "step": 1740000 }, { "epoch": 8.34, "learning_rate": 3.756785714285714e-05, "loss": 1.7196, "step": 1740500 }, { "epoch": 8.35, "learning_rate": 3.756428571428572e-05, "loss": 1.7231, "step": 1741000 }, { "epoch": 8.35, "learning_rate": 3.756071428571429e-05, "loss": 1.7166, "step": 1741500 }, { "epoch": 8.35, "learning_rate": 3.7557142857142856e-05, "loss": 1.7138, "step": 1742000 }, { "epoch": 8.35, "learning_rate": 3.755357142857143e-05, "loss": 1.7222, "step": 1742500 }, { "epoch": 8.36, "learning_rate": 3.7550000000000005e-05, "loss": 1.7281, "step": 1743000 }, { "epoch": 8.36, "learning_rate": 3.754642857142857e-05, "loss": 1.7157, "step": 1743500 }, { "epoch": 8.36, "learning_rate": 3.7542857142857146e-05, "loss": 1.719, "step": 1744000 }, { "epoch": 8.36, "learning_rate": 3.7539285714285714e-05, "loss": 1.7104, "step": 1744500 }, { "epoch": 8.37, "learning_rate": 3.753571428571429e-05, "loss": 1.7146, "step": 1745000 }, { "epoch": 8.37, "learning_rate": 3.753214285714286e-05, "loss": 1.7135, "step": 1745500 }, { "epoch": 8.37, "learning_rate": 3.752857142857143e-05, "loss": 1.7147, "step": 1746000 }, { "epoch": 8.37, "learning_rate": 3.7525e-05, "loss": 1.7196, "step": 1746500 }, { "epoch": 8.38, "learning_rate": 3.752142857142857e-05, "loss": 1.7157, "step": 1747000 }, { "epoch": 8.38, "learning_rate": 3.7517857142857146e-05, "loss": 1.7199, "step": 1747500 }, { "epoch": 8.38, "learning_rate": 3.751428571428572e-05, "loss": 1.7183, "step": 1748000 }, { "epoch": 8.39, "learning_rate": 3.751071428571429e-05, "loss": 1.7213, "step": 1748500 }, { "epoch": 8.39, "learning_rate": 3.7507142857142855e-05, "loss": 1.7117, "step": 1749000 }, { "epoch": 8.39, "learning_rate": 3.750357142857143e-05, "loss": 1.7126, "step": 1749500 }, { "epoch": 8.39, "learning_rate": 3.7500000000000003e-05, "loss": 1.703, "step": 1750000 }, { "epoch": 8.4, "learning_rate": 3.749642857142857e-05, "loss": 1.7079, "step": 1750500 }, { "epoch": 8.4, "learning_rate": 3.7492857142857145e-05, "loss": 1.7215, "step": 1751000 }, { "epoch": 8.4, "learning_rate": 3.748928571428571e-05, "loss": 1.7094, "step": 1751500 }, { "epoch": 8.4, "learning_rate": 3.748571428571429e-05, "loss": 1.7136, "step": 1752000 }, { "epoch": 8.41, "learning_rate": 3.748214285714286e-05, "loss": 1.7121, "step": 1752500 }, { "epoch": 8.41, "learning_rate": 3.747857142857143e-05, "loss": 1.7169, "step": 1753000 }, { "epoch": 8.41, "learning_rate": 3.7475e-05, "loss": 1.7249, "step": 1753500 }, { "epoch": 8.41, "learning_rate": 3.747142857142858e-05, "loss": 1.7207, "step": 1754000 }, { "epoch": 8.42, "learning_rate": 3.7467857142857145e-05, "loss": 1.7165, "step": 1754500 }, { "epoch": 8.42, "learning_rate": 3.746428571428572e-05, "loss": 1.7147, "step": 1755000 }, { "epoch": 8.42, "learning_rate": 3.7460714285714286e-05, "loss": 1.7147, "step": 1755500 }, { "epoch": 8.43, "learning_rate": 3.745714285714286e-05, "loss": 1.7068, "step": 1756000 }, { "epoch": 8.43, "learning_rate": 3.7453571428571435e-05, "loss": 1.7071, "step": 1756500 }, { "epoch": 8.43, "learning_rate": 3.745e-05, "loss": 1.7159, "step": 1757000 }, { "epoch": 8.43, "learning_rate": 3.744642857142857e-05, "loss": 1.7218, "step": 1757500 }, { "epoch": 8.44, "learning_rate": 3.7442857142857144e-05, "loss": 1.7163, "step": 1758000 }, { "epoch": 8.44, "learning_rate": 3.743928571428572e-05, "loss": 1.7192, "step": 1758500 }, { "epoch": 8.44, "learning_rate": 3.7435714285714286e-05, "loss": 1.711, "step": 1759000 }, { "epoch": 8.44, "learning_rate": 3.743214285714286e-05, "loss": 1.7152, "step": 1759500 }, { "epoch": 8.45, "learning_rate": 3.742857142857143e-05, "loss": 1.7126, "step": 1760000 }, { "epoch": 8.45, "learning_rate": 3.7425e-05, "loss": 1.7173, "step": 1760500 }, { "epoch": 8.45, "learning_rate": 3.7421428571428576e-05, "loss": 1.7148, "step": 1761000 }, { "epoch": 8.45, "learning_rate": 3.741785714285714e-05, "loss": 1.7089, "step": 1761500 }, { "epoch": 8.46, "learning_rate": 3.741428571428572e-05, "loss": 1.7169, "step": 1762000 }, { "epoch": 8.46, "learning_rate": 3.7410714285714285e-05, "loss": 1.712, "step": 1762500 }, { "epoch": 8.46, "learning_rate": 3.740714285714286e-05, "loss": 1.7171, "step": 1763000 }, { "epoch": 8.47, "learning_rate": 3.7403571428571434e-05, "loss": 1.7113, "step": 1763500 }, { "epoch": 8.47, "learning_rate": 3.74e-05, "loss": 1.7066, "step": 1764000 }, { "epoch": 8.47, "learning_rate": 3.739642857142857e-05, "loss": 1.7208, "step": 1764500 }, { "epoch": 8.47, "learning_rate": 3.739285714285714e-05, "loss": 1.7182, "step": 1765000 }, { "epoch": 8.48, "learning_rate": 3.738928571428572e-05, "loss": 1.7143, "step": 1765500 }, { "epoch": 8.48, "learning_rate": 3.7385714285714284e-05, "loss": 1.7121, "step": 1766000 }, { "epoch": 8.48, "learning_rate": 3.738214285714286e-05, "loss": 1.7103, "step": 1766500 }, { "epoch": 8.48, "learning_rate": 3.7378571428571426e-05, "loss": 1.7179, "step": 1767000 }, { "epoch": 8.49, "learning_rate": 3.737500000000001e-05, "loss": 1.7202, "step": 1767500 }, { "epoch": 8.49, "learning_rate": 3.7371428571428575e-05, "loss": 1.7139, "step": 1768000 }, { "epoch": 8.49, "learning_rate": 3.736785714285714e-05, "loss": 1.715, "step": 1768500 }, { "epoch": 8.49, "learning_rate": 3.7364285714285716e-05, "loss": 1.708, "step": 1769000 }, { "epoch": 8.5, "learning_rate": 3.736071428571429e-05, "loss": 1.7079, "step": 1769500 }, { "epoch": 8.5, "learning_rate": 3.735714285714286e-05, "loss": 1.7176, "step": 1770000 }, { "epoch": 8.5, "learning_rate": 3.735357142857143e-05, "loss": 1.7071, "step": 1770500 }, { "epoch": 8.51, "learning_rate": 3.735e-05, "loss": 1.7136, "step": 1771000 }, { "epoch": 8.51, "learning_rate": 3.7346428571428574e-05, "loss": 1.7111, "step": 1771500 }, { "epoch": 8.51, "learning_rate": 3.734285714285715e-05, "loss": 1.7194, "step": 1772000 }, { "epoch": 8.51, "learning_rate": 3.7339285714285716e-05, "loss": 1.7236, "step": 1772500 }, { "epoch": 8.52, "learning_rate": 3.733571428571428e-05, "loss": 1.718, "step": 1773000 }, { "epoch": 8.52, "learning_rate": 3.733214285714286e-05, "loss": 1.7, "step": 1773500 }, { "epoch": 8.52, "learning_rate": 3.732857142857143e-05, "loss": 1.7091, "step": 1774000 }, { "epoch": 8.52, "learning_rate": 3.7325000000000006e-05, "loss": 1.7083, "step": 1774500 }, { "epoch": 8.53, "learning_rate": 3.7321428571428573e-05, "loss": 1.7036, "step": 1775000 }, { "epoch": 8.53, "learning_rate": 3.731785714285714e-05, "loss": 1.7099, "step": 1775500 }, { "epoch": 8.53, "learning_rate": 3.7314285714285715e-05, "loss": 1.7051, "step": 1776000 }, { "epoch": 8.53, "learning_rate": 3.731071428571429e-05, "loss": 1.7091, "step": 1776500 }, { "epoch": 8.54, "learning_rate": 3.730714285714286e-05, "loss": 1.7194, "step": 1777000 }, { "epoch": 8.54, "learning_rate": 3.730357142857143e-05, "loss": 1.7337, "step": 1777500 }, { "epoch": 8.54, "learning_rate": 3.73e-05, "loss": 1.7133, "step": 1778000 }, { "epoch": 8.55, "learning_rate": 3.729642857142857e-05, "loss": 1.7114, "step": 1778500 }, { "epoch": 8.55, "learning_rate": 3.729285714285715e-05, "loss": 1.7141, "step": 1779000 }, { "epoch": 8.55, "learning_rate": 3.7289285714285715e-05, "loss": 1.6977, "step": 1779500 }, { "epoch": 8.55, "learning_rate": 3.728571428571428e-05, "loss": 1.7031, "step": 1780000 }, { "epoch": 8.56, "learning_rate": 3.728214285714286e-05, "loss": 1.7034, "step": 1780500 }, { "epoch": 8.56, "learning_rate": 3.727857142857143e-05, "loss": 1.7189, "step": 1781000 }, { "epoch": 8.56, "learning_rate": 3.7275000000000005e-05, "loss": 1.7076, "step": 1781500 }, { "epoch": 8.56, "learning_rate": 3.727142857142857e-05, "loss": 1.7093, "step": 1782000 }, { "epoch": 8.57, "learning_rate": 3.7267857142857147e-05, "loss": 1.7037, "step": 1782500 }, { "epoch": 8.57, "learning_rate": 3.726428571428572e-05, "loss": 1.7156, "step": 1783000 }, { "epoch": 8.57, "learning_rate": 3.726071428571429e-05, "loss": 1.7039, "step": 1783500 }, { "epoch": 8.57, "learning_rate": 3.7257142857142856e-05, "loss": 1.7018, "step": 1784000 }, { "epoch": 8.58, "learning_rate": 3.725357142857143e-05, "loss": 1.7012, "step": 1784500 }, { "epoch": 8.58, "learning_rate": 3.7250000000000004e-05, "loss": 1.6979, "step": 1785000 }, { "epoch": 8.58, "learning_rate": 3.724642857142857e-05, "loss": 1.6971, "step": 1785500 }, { "epoch": 8.59, "learning_rate": 3.7242857142857146e-05, "loss": 1.7147, "step": 1786000 }, { "epoch": 8.59, "learning_rate": 3.723928571428571e-05, "loss": 1.6971, "step": 1786500 }, { "epoch": 8.59, "learning_rate": 3.723571428571429e-05, "loss": 1.6995, "step": 1787000 }, { "epoch": 8.59, "learning_rate": 3.723214285714286e-05, "loss": 1.6982, "step": 1787500 }, { "epoch": 8.6, "learning_rate": 3.722857142857143e-05, "loss": 1.7084, "step": 1788000 }, { "epoch": 8.6, "learning_rate": 3.7225000000000004e-05, "loss": 1.7137, "step": 1788500 }, { "epoch": 8.6, "learning_rate": 3.722142857142857e-05, "loss": 1.7038, "step": 1789000 }, { "epoch": 8.6, "learning_rate": 3.7217857142857145e-05, "loss": 1.7015, "step": 1789500 }, { "epoch": 8.61, "learning_rate": 3.721428571428572e-05, "loss": 1.714, "step": 1790000 }, { "epoch": 8.61, "learning_rate": 3.721071428571429e-05, "loss": 1.7071, "step": 1790500 }, { "epoch": 8.61, "learning_rate": 3.7207142857142854e-05, "loss": 1.7096, "step": 1791000 }, { "epoch": 8.61, "learning_rate": 3.720357142857143e-05, "loss": 1.7128, "step": 1791500 }, { "epoch": 8.62, "learning_rate": 3.72e-05, "loss": 1.7124, "step": 1792000 }, { "epoch": 8.62, "learning_rate": 3.719642857142857e-05, "loss": 1.7007, "step": 1792500 }, { "epoch": 8.62, "learning_rate": 3.7192857142857145e-05, "loss": 1.6993, "step": 1793000 }, { "epoch": 8.62, "learning_rate": 3.718928571428571e-05, "loss": 1.7116, "step": 1793500 }, { "epoch": 8.63, "learning_rate": 3.7185714285714286e-05, "loss": 1.7011, "step": 1794000 }, { "epoch": 8.63, "learning_rate": 3.718214285714286e-05, "loss": 1.7054, "step": 1794500 }, { "epoch": 8.63, "learning_rate": 3.717857142857143e-05, "loss": 1.7019, "step": 1795000 }, { "epoch": 8.64, "learning_rate": 3.7175e-05, "loss": 1.7075, "step": 1795500 }, { "epoch": 8.64, "learning_rate": 3.717142857142858e-05, "loss": 1.7097, "step": 1796000 }, { "epoch": 8.64, "learning_rate": 3.7167857142857144e-05, "loss": 1.7082, "step": 1796500 }, { "epoch": 8.64, "learning_rate": 3.716428571428572e-05, "loss": 1.698, "step": 1797000 }, { "epoch": 8.65, "learning_rate": 3.7160714285714286e-05, "loss": 1.7091, "step": 1797500 }, { "epoch": 8.65, "learning_rate": 3.715714285714286e-05, "loss": 1.7004, "step": 1798000 }, { "epoch": 8.65, "learning_rate": 3.7153571428571434e-05, "loss": 1.7042, "step": 1798500 }, { "epoch": 8.65, "learning_rate": 3.715e-05, "loss": 1.7005, "step": 1799000 }, { "epoch": 8.66, "learning_rate": 3.714642857142857e-05, "loss": 1.7089, "step": 1799500 }, { "epoch": 8.66, "learning_rate": 3.7142857142857143e-05, "loss": 1.7101, "step": 1800000 }, { "epoch": 8.66, "learning_rate": 3.713928571428572e-05, "loss": 1.7101, "step": 1800500 }, { "epoch": 8.66, "learning_rate": 3.7135714285714285e-05, "loss": 1.712, "step": 1801000 }, { "epoch": 8.67, "learning_rate": 3.713214285714286e-05, "loss": 1.7176, "step": 1801500 }, { "epoch": 8.67, "learning_rate": 3.712857142857143e-05, "loss": 1.7137, "step": 1802000 }, { "epoch": 8.67, "learning_rate": 3.7125e-05, "loss": 1.7053, "step": 1802500 }, { "epoch": 8.68, "learning_rate": 3.7121428571428575e-05, "loss": 1.6959, "step": 1803000 }, { "epoch": 8.68, "learning_rate": 3.711785714285714e-05, "loss": 1.7057, "step": 1803500 }, { "epoch": 8.68, "learning_rate": 3.711428571428572e-05, "loss": 1.6988, "step": 1804000 }, { "epoch": 8.68, "learning_rate": 3.7110714285714285e-05, "loss": 1.7021, "step": 1804500 }, { "epoch": 8.69, "learning_rate": 3.710714285714286e-05, "loss": 1.7005, "step": 1805000 }, { "epoch": 8.69, "learning_rate": 3.710357142857143e-05, "loss": 1.7025, "step": 1805500 }, { "epoch": 8.69, "learning_rate": 3.71e-05, "loss": 1.7061, "step": 1806000 }, { "epoch": 8.69, "learning_rate": 3.709642857142857e-05, "loss": 1.7005, "step": 1806500 }, { "epoch": 8.7, "learning_rate": 3.709285714285715e-05, "loss": 1.7083, "step": 1807000 }, { "epoch": 8.7, "learning_rate": 3.7089285714285717e-05, "loss": 1.704, "step": 1807500 }, { "epoch": 8.7, "learning_rate": 3.7085714285714284e-05, "loss": 1.7084, "step": 1808000 }, { "epoch": 8.7, "learning_rate": 3.708214285714286e-05, "loss": 1.703, "step": 1808500 }, { "epoch": 8.71, "learning_rate": 3.707857142857143e-05, "loss": 1.7051, "step": 1809000 }, { "epoch": 8.71, "learning_rate": 3.707500000000001e-05, "loss": 1.7037, "step": 1809500 }, { "epoch": 8.71, "learning_rate": 3.7071428571428574e-05, "loss": 1.6963, "step": 1810000 }, { "epoch": 8.72, "learning_rate": 3.706785714285714e-05, "loss": 1.694, "step": 1810500 }, { "epoch": 8.72, "learning_rate": 3.7064285714285716e-05, "loss": 1.6994, "step": 1811000 }, { "epoch": 8.72, "learning_rate": 3.706071428571429e-05, "loss": 1.7045, "step": 1811500 }, { "epoch": 8.72, "learning_rate": 3.705714285714286e-05, "loss": 1.7012, "step": 1812000 }, { "epoch": 8.73, "learning_rate": 3.705357142857143e-05, "loss": 1.7075, "step": 1812500 }, { "epoch": 8.73, "learning_rate": 3.705e-05, "loss": 1.7059, "step": 1813000 }, { "epoch": 8.73, "learning_rate": 3.7046428571428574e-05, "loss": 1.7083, "step": 1813500 }, { "epoch": 8.73, "learning_rate": 3.704285714285715e-05, "loss": 1.7059, "step": 1814000 }, { "epoch": 8.74, "learning_rate": 3.7039285714285715e-05, "loss": 1.6967, "step": 1814500 }, { "epoch": 8.74, "learning_rate": 3.703571428571429e-05, "loss": 1.7109, "step": 1815000 }, { "epoch": 8.74, "learning_rate": 3.703214285714286e-05, "loss": 1.7038, "step": 1815500 }, { "epoch": 8.74, "learning_rate": 3.702857142857143e-05, "loss": 1.7069, "step": 1816000 }, { "epoch": 8.75, "learning_rate": 3.7025000000000005e-05, "loss": 1.7011, "step": 1816500 }, { "epoch": 8.75, "learning_rate": 3.702142857142857e-05, "loss": 1.6961, "step": 1817000 }, { "epoch": 8.75, "learning_rate": 3.701785714285714e-05, "loss": 1.702, "step": 1817500 }, { "epoch": 8.76, "learning_rate": 3.7014285714285715e-05, "loss": 1.702, "step": 1818000 }, { "epoch": 8.76, "learning_rate": 3.701071428571429e-05, "loss": 1.6999, "step": 1818500 }, { "epoch": 8.76, "learning_rate": 3.7007142857142856e-05, "loss": 1.7026, "step": 1819000 }, { "epoch": 8.76, "learning_rate": 3.700357142857143e-05, "loss": 1.709, "step": 1819500 }, { "epoch": 8.77, "learning_rate": 3.7e-05, "loss": 1.7029, "step": 1820000 }, { "epoch": 8.77, "learning_rate": 3.699642857142857e-05, "loss": 1.6988, "step": 1820500 }, { "epoch": 8.77, "learning_rate": 3.6992857142857147e-05, "loss": 1.7031, "step": 1821000 }, { "epoch": 8.77, "learning_rate": 3.6989285714285714e-05, "loss": 1.7057, "step": 1821500 }, { "epoch": 8.78, "learning_rate": 3.698571428571429e-05, "loss": 1.6907, "step": 1822000 }, { "epoch": 8.78, "learning_rate": 3.698214285714286e-05, "loss": 1.7021, "step": 1822500 }, { "epoch": 8.78, "learning_rate": 3.697857142857143e-05, "loss": 1.6995, "step": 1823000 }, { "epoch": 8.78, "learning_rate": 3.6975000000000004e-05, "loss": 1.6978, "step": 1823500 }, { "epoch": 8.79, "learning_rate": 3.697142857142857e-05, "loss": 1.6973, "step": 1824000 }, { "epoch": 8.79, "learning_rate": 3.6967857142857146e-05, "loss": 1.7012, "step": 1824500 }, { "epoch": 8.79, "learning_rate": 3.696428571428572e-05, "loss": 1.6883, "step": 1825000 }, { "epoch": 8.8, "learning_rate": 3.696071428571429e-05, "loss": 1.7061, "step": 1825500 }, { "epoch": 8.8, "learning_rate": 3.6957142857142855e-05, "loss": 1.6854, "step": 1826000 }, { "epoch": 8.8, "learning_rate": 3.695357142857143e-05, "loss": 1.6995, "step": 1826500 }, { "epoch": 8.8, "learning_rate": 3.6950000000000004e-05, "loss": 1.7004, "step": 1827000 }, { "epoch": 8.81, "learning_rate": 3.694642857142857e-05, "loss": 1.6962, "step": 1827500 }, { "epoch": 8.81, "learning_rate": 3.6942857142857145e-05, "loss": 1.7109, "step": 1828000 }, { "epoch": 8.81, "learning_rate": 3.693928571428571e-05, "loss": 1.6952, "step": 1828500 }, { "epoch": 8.81, "learning_rate": 3.693571428571429e-05, "loss": 1.6992, "step": 1829000 }, { "epoch": 8.82, "learning_rate": 3.693214285714286e-05, "loss": 1.7032, "step": 1829500 }, { "epoch": 8.82, "learning_rate": 3.692857142857143e-05, "loss": 1.6946, "step": 1830000 }, { "epoch": 8.82, "learning_rate": 3.6925e-05, "loss": 1.7022, "step": 1830500 }, { "epoch": 8.82, "learning_rate": 3.692142857142857e-05, "loss": 1.6965, "step": 1831000 }, { "epoch": 8.83, "learning_rate": 3.6917857142857145e-05, "loss": 1.7016, "step": 1831500 }, { "epoch": 8.83, "learning_rate": 3.691428571428572e-05, "loss": 1.7002, "step": 1832000 }, { "epoch": 8.83, "learning_rate": 3.6910714285714286e-05, "loss": 1.6884, "step": 1832500 }, { "epoch": 8.84, "learning_rate": 3.6907142857142854e-05, "loss": 1.6966, "step": 1833000 }, { "epoch": 8.84, "learning_rate": 3.6903571428571435e-05, "loss": 1.6977, "step": 1833500 }, { "epoch": 8.84, "learning_rate": 3.69e-05, "loss": 1.6966, "step": 1834000 }, { "epoch": 8.84, "learning_rate": 3.689642857142857e-05, "loss": 1.6975, "step": 1834500 }, { "epoch": 8.85, "learning_rate": 3.6892857142857144e-05, "loss": 1.6857, "step": 1835000 }, { "epoch": 8.85, "learning_rate": 3.688928571428572e-05, "loss": 1.6831, "step": 1835500 }, { "epoch": 8.85, "learning_rate": 3.688571428571429e-05, "loss": 1.6966, "step": 1836000 }, { "epoch": 8.85, "learning_rate": 3.688214285714286e-05, "loss": 1.697, "step": 1836500 }, { "epoch": 8.86, "learning_rate": 3.687857142857143e-05, "loss": 1.6909, "step": 1837000 }, { "epoch": 8.86, "learning_rate": 3.6875e-05, "loss": 1.7067, "step": 1837500 }, { "epoch": 8.86, "learning_rate": 3.6871428571428576e-05, "loss": 1.698, "step": 1838000 }, { "epoch": 8.86, "learning_rate": 3.6867857142857144e-05, "loss": 1.7039, "step": 1838500 }, { "epoch": 8.87, "learning_rate": 3.686428571428572e-05, "loss": 1.6824, "step": 1839000 }, { "epoch": 8.87, "learning_rate": 3.6860714285714285e-05, "loss": 1.6994, "step": 1839500 }, { "epoch": 8.87, "learning_rate": 3.685714285714286e-05, "loss": 1.697, "step": 1840000 }, { "epoch": 8.88, "learning_rate": 3.6853571428571434e-05, "loss": 1.7024, "step": 1840500 }, { "epoch": 8.88, "learning_rate": 3.685e-05, "loss": 1.7106, "step": 1841000 }, { "epoch": 8.88, "learning_rate": 3.684642857142857e-05, "loss": 1.692, "step": 1841500 }, { "epoch": 8.88, "learning_rate": 3.684285714285714e-05, "loss": 1.7023, "step": 1842000 }, { "epoch": 8.89, "learning_rate": 3.683928571428572e-05, "loss": 1.6928, "step": 1842500 }, { "epoch": 8.89, "learning_rate": 3.683571428571429e-05, "loss": 1.6985, "step": 1843000 }, { "epoch": 8.89, "learning_rate": 3.683214285714286e-05, "loss": 1.7034, "step": 1843500 }, { "epoch": 8.89, "learning_rate": 3.6828571428571426e-05, "loss": 1.7001, "step": 1844000 }, { "epoch": 8.9, "learning_rate": 3.6825e-05, "loss": 1.6951, "step": 1844500 }, { "epoch": 8.9, "learning_rate": 3.6821428571428575e-05, "loss": 1.6955, "step": 1845000 }, { "epoch": 8.9, "learning_rate": 3.681785714285714e-05, "loss": 1.6914, "step": 1845500 }, { "epoch": 8.9, "learning_rate": 3.6814285714285717e-05, "loss": 1.6912, "step": 1846000 }, { "epoch": 8.91, "learning_rate": 3.6810714285714284e-05, "loss": 1.6916, "step": 1846500 }, { "epoch": 8.91, "learning_rate": 3.680714285714286e-05, "loss": 1.6994, "step": 1847000 }, { "epoch": 8.91, "learning_rate": 3.680357142857143e-05, "loss": 1.6939, "step": 1847500 }, { "epoch": 8.92, "learning_rate": 3.68e-05, "loss": 1.6908, "step": 1848000 }, { "epoch": 8.92, "learning_rate": 3.679642857142857e-05, "loss": 1.6938, "step": 1848500 }, { "epoch": 8.92, "learning_rate": 3.679285714285715e-05, "loss": 1.6978, "step": 1849000 }, { "epoch": 8.92, "learning_rate": 3.6789285714285716e-05, "loss": 1.6943, "step": 1849500 }, { "epoch": 8.93, "learning_rate": 3.678571428571429e-05, "loss": 1.691, "step": 1850000 }, { "epoch": 8.93, "learning_rate": 3.678214285714286e-05, "loss": 1.6855, "step": 1850500 }, { "epoch": 8.93, "learning_rate": 3.677857142857143e-05, "loss": 1.6952, "step": 1851000 }, { "epoch": 8.93, "learning_rate": 3.6775000000000006e-05, "loss": 1.7001, "step": 1851500 }, { "epoch": 8.94, "learning_rate": 3.6771428571428574e-05, "loss": 1.6972, "step": 1852000 }, { "epoch": 8.94, "learning_rate": 3.676785714285714e-05, "loss": 1.7035, "step": 1852500 }, { "epoch": 8.94, "learning_rate": 3.6764285714285715e-05, "loss": 1.6947, "step": 1853000 }, { "epoch": 8.94, "learning_rate": 3.676071428571429e-05, "loss": 1.6924, "step": 1853500 }, { "epoch": 8.95, "learning_rate": 3.675714285714286e-05, "loss": 1.6865, "step": 1854000 }, { "epoch": 8.95, "learning_rate": 3.675357142857143e-05, "loss": 1.698, "step": 1854500 }, { "epoch": 8.95, "learning_rate": 3.675e-05, "loss": 1.6957, "step": 1855000 }, { "epoch": 8.96, "learning_rate": 3.674642857142857e-05, "loss": 1.6946, "step": 1855500 }, { "epoch": 8.96, "learning_rate": 3.674285714285715e-05, "loss": 1.6956, "step": 1856000 }, { "epoch": 8.96, "learning_rate": 3.6739285714285715e-05, "loss": 1.7033, "step": 1856500 }, { "epoch": 8.96, "learning_rate": 3.673571428571429e-05, "loss": 1.6978, "step": 1857000 }, { "epoch": 8.97, "learning_rate": 3.6732142857142856e-05, "loss": 1.7006, "step": 1857500 }, { "epoch": 8.97, "learning_rate": 3.672857142857143e-05, "loss": 1.6977, "step": 1858000 }, { "epoch": 8.97, "learning_rate": 3.6725000000000005e-05, "loss": 1.7022, "step": 1858500 }, { "epoch": 8.97, "learning_rate": 3.672142857142857e-05, "loss": 1.6876, "step": 1859000 }, { "epoch": 8.98, "learning_rate": 3.671785714285714e-05, "loss": 1.6981, "step": 1859500 }, { "epoch": 8.98, "learning_rate": 3.671428571428572e-05, "loss": 1.7029, "step": 1860000 }, { "epoch": 8.98, "learning_rate": 3.671071428571429e-05, "loss": 1.6913, "step": 1860500 }, { "epoch": 8.98, "learning_rate": 3.6707142857142856e-05, "loss": 1.7064, "step": 1861000 }, { "epoch": 8.99, "learning_rate": 3.670357142857143e-05, "loss": 1.6936, "step": 1861500 }, { "epoch": 8.99, "learning_rate": 3.6700000000000004e-05, "loss": 1.6964, "step": 1862000 }, { "epoch": 8.99, "learning_rate": 3.669642857142857e-05, "loss": 1.6912, "step": 1862500 }, { "epoch": 9.0, "learning_rate": 3.6692857142857146e-05, "loss": 1.6885, "step": 1863000 }, { "epoch": 9.0, "learning_rate": 3.6689285714285714e-05, "loss": 1.6995, "step": 1863500 }, { "epoch": 9.0, "learning_rate": 3.668571428571429e-05, "loss": 1.6938, "step": 1864000 }, { "epoch": 9.0, "learning_rate": 3.668214285714286e-05, "loss": 1.6884, "step": 1864500 }, { "epoch": 9.01, "learning_rate": 3.667857142857143e-05, "loss": 1.689, "step": 1865000 }, { "epoch": 9.01, "learning_rate": 3.6675000000000004e-05, "loss": 1.6923, "step": 1865500 }, { "epoch": 9.01, "learning_rate": 3.667142857142857e-05, "loss": 1.6889, "step": 1866000 }, { "epoch": 9.01, "learning_rate": 3.6667857142857145e-05, "loss": 1.6907, "step": 1866500 }, { "epoch": 9.02, "learning_rate": 3.666428571428572e-05, "loss": 1.6935, "step": 1867000 }, { "epoch": 9.02, "learning_rate": 3.666071428571429e-05, "loss": 1.6924, "step": 1867500 }, { "epoch": 9.02, "learning_rate": 3.6657142857142855e-05, "loss": 1.6901, "step": 1868000 }, { "epoch": 9.02, "learning_rate": 3.665357142857143e-05, "loss": 1.6835, "step": 1868500 }, { "epoch": 9.03, "learning_rate": 3.665e-05, "loss": 1.6934, "step": 1869000 }, { "epoch": 9.03, "learning_rate": 3.664642857142857e-05, "loss": 1.6862, "step": 1869500 }, { "epoch": 9.03, "learning_rate": 3.6642857142857145e-05, "loss": 1.6712, "step": 1870000 }, { "epoch": 9.03, "learning_rate": 3.663928571428571e-05, "loss": 1.6828, "step": 1870500 }, { "epoch": 9.04, "learning_rate": 3.6635714285714287e-05, "loss": 1.689, "step": 1871000 }, { "epoch": 9.04, "learning_rate": 3.663214285714286e-05, "loss": 1.681, "step": 1871500 }, { "epoch": 9.04, "learning_rate": 3.662857142857143e-05, "loss": 1.685, "step": 1872000 }, { "epoch": 9.05, "learning_rate": 3.6625e-05, "loss": 1.6944, "step": 1872500 }, { "epoch": 9.05, "learning_rate": 3.662142857142857e-05, "loss": 1.6814, "step": 1873000 }, { "epoch": 9.05, "learning_rate": 3.6617857142857144e-05, "loss": 1.6864, "step": 1873500 }, { "epoch": 9.05, "learning_rate": 3.661428571428572e-05, "loss": 1.6805, "step": 1874000 }, { "epoch": 9.06, "learning_rate": 3.6610714285714286e-05, "loss": 1.6794, "step": 1874500 }, { "epoch": 9.06, "learning_rate": 3.6607142857142853e-05, "loss": 1.6773, "step": 1875000 }, { "epoch": 9.06, "learning_rate": 3.6603571428571434e-05, "loss": 1.6966, "step": 1875500 }, { "epoch": 9.06, "learning_rate": 3.66e-05, "loss": 1.6812, "step": 1876000 }, { "epoch": 9.07, "learning_rate": 3.6596428571428576e-05, "loss": 1.6885, "step": 1876500 }, { "epoch": 9.07, "learning_rate": 3.6592857142857144e-05, "loss": 1.6772, "step": 1877000 }, { "epoch": 9.07, "learning_rate": 3.658928571428572e-05, "loss": 1.6943, "step": 1877500 }, { "epoch": 9.07, "learning_rate": 3.658571428571429e-05, "loss": 1.6862, "step": 1878000 }, { "epoch": 9.08, "learning_rate": 3.658214285714286e-05, "loss": 1.689, "step": 1878500 }, { "epoch": 9.08, "learning_rate": 3.657857142857143e-05, "loss": 1.6764, "step": 1879000 }, { "epoch": 9.08, "learning_rate": 3.6575e-05, "loss": 1.6828, "step": 1879500 }, { "epoch": 9.09, "learning_rate": 3.6571428571428576e-05, "loss": 1.696, "step": 1880000 }, { "epoch": 9.09, "learning_rate": 3.656785714285714e-05, "loss": 1.6943, "step": 1880500 }, { "epoch": 9.09, "learning_rate": 3.656428571428572e-05, "loss": 1.6836, "step": 1881000 }, { "epoch": 9.09, "learning_rate": 3.6560714285714285e-05, "loss": 1.6807, "step": 1881500 }, { "epoch": 9.1, "learning_rate": 3.655714285714286e-05, "loss": 1.6696, "step": 1882000 }, { "epoch": 9.1, "learning_rate": 3.655357142857143e-05, "loss": 1.6932, "step": 1882500 }, { "epoch": 9.1, "learning_rate": 3.655e-05, "loss": 1.6908, "step": 1883000 }, { "epoch": 9.1, "learning_rate": 3.6546428571428575e-05, "loss": 1.676, "step": 1883500 }, { "epoch": 9.11, "learning_rate": 3.654285714285714e-05, "loss": 1.6829, "step": 1884000 }, { "epoch": 9.11, "learning_rate": 3.653928571428572e-05, "loss": 1.6868, "step": 1884500 }, { "epoch": 9.11, "learning_rate": 3.653571428571429e-05, "loss": 1.6809, "step": 1885000 }, { "epoch": 9.11, "learning_rate": 3.653214285714286e-05, "loss": 1.6822, "step": 1885500 }, { "epoch": 9.12, "learning_rate": 3.6528571428571426e-05, "loss": 1.6814, "step": 1886000 }, { "epoch": 9.12, "learning_rate": 3.652500000000001e-05, "loss": 1.6832, "step": 1886500 }, { "epoch": 9.12, "learning_rate": 3.6521428571428574e-05, "loss": 1.6786, "step": 1887000 }, { "epoch": 9.13, "learning_rate": 3.651785714285714e-05, "loss": 1.6968, "step": 1887500 }, { "epoch": 9.13, "learning_rate": 3.6514285714285716e-05, "loss": 1.6829, "step": 1888000 }, { "epoch": 9.13, "learning_rate": 3.651071428571429e-05, "loss": 1.6769, "step": 1888500 }, { "epoch": 9.13, "learning_rate": 3.650714285714286e-05, "loss": 1.6839, "step": 1889000 }, { "epoch": 9.14, "learning_rate": 3.650357142857143e-05, "loss": 1.6831, "step": 1889500 }, { "epoch": 9.14, "learning_rate": 3.65e-05, "loss": 1.6851, "step": 1890000 }, { "epoch": 9.14, "learning_rate": 3.6496428571428574e-05, "loss": 1.6728, "step": 1890500 }, { "epoch": 9.14, "learning_rate": 3.649285714285715e-05, "loss": 1.6777, "step": 1891000 }, { "epoch": 9.15, "learning_rate": 3.6489285714285715e-05, "loss": 1.688, "step": 1891500 }, { "epoch": 9.15, "learning_rate": 3.648571428571429e-05, "loss": 1.6843, "step": 1892000 }, { "epoch": 9.15, "learning_rate": 3.648214285714286e-05, "loss": 1.6775, "step": 1892500 }, { "epoch": 9.15, "learning_rate": 3.647857142857143e-05, "loss": 1.6902, "step": 1893000 }, { "epoch": 9.16, "learning_rate": 3.6475000000000006e-05, "loss": 1.6891, "step": 1893500 }, { "epoch": 9.16, "learning_rate": 3.647142857142857e-05, "loss": 1.6839, "step": 1894000 }, { "epoch": 9.16, "learning_rate": 3.646785714285714e-05, "loss": 1.6823, "step": 1894500 }, { "epoch": 9.17, "learning_rate": 3.6464285714285715e-05, "loss": 1.6791, "step": 1895000 }, { "epoch": 9.17, "learning_rate": 3.646071428571429e-05, "loss": 1.6667, "step": 1895500 }, { "epoch": 9.17, "learning_rate": 3.6457142857142857e-05, "loss": 1.6833, "step": 1896000 }, { "epoch": 9.17, "learning_rate": 3.645357142857143e-05, "loss": 1.6796, "step": 1896500 }, { "epoch": 9.18, "learning_rate": 3.645e-05, "loss": 1.68, "step": 1897000 }, { "epoch": 9.18, "learning_rate": 3.644642857142857e-05, "loss": 1.6824, "step": 1897500 }, { "epoch": 9.18, "learning_rate": 3.644285714285715e-05, "loss": 1.6899, "step": 1898000 }, { "epoch": 9.18, "learning_rate": 3.6439285714285714e-05, "loss": 1.6796, "step": 1898500 }, { "epoch": 9.19, "learning_rate": 3.643571428571429e-05, "loss": 1.6837, "step": 1899000 }, { "epoch": 9.19, "learning_rate": 3.6432142857142856e-05, "loss": 1.6855, "step": 1899500 }, { "epoch": 9.19, "learning_rate": 3.642857142857143e-05, "loss": 1.679, "step": 1900000 }, { "epoch": 9.19, "learning_rate": 3.6425000000000004e-05, "loss": 1.6817, "step": 1900500 }, { "epoch": 9.2, "learning_rate": 3.642142857142857e-05, "loss": 1.6849, "step": 1901000 }, { "epoch": 9.2, "learning_rate": 3.641785714285714e-05, "loss": 1.6898, "step": 1901500 }, { "epoch": 9.2, "learning_rate": 3.641428571428572e-05, "loss": 1.6804, "step": 1902000 }, { "epoch": 9.21, "learning_rate": 3.641071428571429e-05, "loss": 1.6784, "step": 1902500 }, { "epoch": 9.21, "learning_rate": 3.6407142857142855e-05, "loss": 1.6833, "step": 1903000 }, { "epoch": 9.21, "learning_rate": 3.640357142857143e-05, "loss": 1.6866, "step": 1903500 }, { "epoch": 9.21, "learning_rate": 3.6400000000000004e-05, "loss": 1.6882, "step": 1904000 }, { "epoch": 9.22, "learning_rate": 3.639642857142858e-05, "loss": 1.6776, "step": 1904500 }, { "epoch": 9.22, "learning_rate": 3.6392857142857146e-05, "loss": 1.6843, "step": 1905000 }, { "epoch": 9.22, "learning_rate": 3.638928571428571e-05, "loss": 1.6775, "step": 1905500 }, { "epoch": 9.22, "learning_rate": 3.638571428571429e-05, "loss": 1.6806, "step": 1906000 }, { "epoch": 9.23, "learning_rate": 3.638214285714286e-05, "loss": 1.678, "step": 1906500 }, { "epoch": 9.23, "learning_rate": 3.637857142857143e-05, "loss": 1.6851, "step": 1907000 }, { "epoch": 9.23, "learning_rate": 3.6375e-05, "loss": 1.6868, "step": 1907500 }, { "epoch": 9.23, "learning_rate": 3.637142857142857e-05, "loss": 1.6724, "step": 1908000 }, { "epoch": 9.24, "learning_rate": 3.6367857142857145e-05, "loss": 1.6781, "step": 1908500 }, { "epoch": 9.24, "learning_rate": 3.636428571428572e-05, "loss": 1.674, "step": 1909000 }, { "epoch": 9.24, "learning_rate": 3.636071428571429e-05, "loss": 1.6917, "step": 1909500 }, { "epoch": 9.25, "learning_rate": 3.6357142857142854e-05, "loss": 1.6801, "step": 1910000 }, { "epoch": 9.25, "learning_rate": 3.635357142857143e-05, "loss": 1.6828, "step": 1910500 }, { "epoch": 9.25, "learning_rate": 3.635e-05, "loss": 1.6835, "step": 1911000 }, { "epoch": 9.25, "learning_rate": 3.634642857142858e-05, "loss": 1.6822, "step": 1911500 }, { "epoch": 9.26, "learning_rate": 3.6342857142857144e-05, "loss": 1.6848, "step": 1912000 }, { "epoch": 9.26, "learning_rate": 3.633928571428571e-05, "loss": 1.6715, "step": 1912500 }, { "epoch": 9.26, "learning_rate": 3.633571428571429e-05, "loss": 1.6739, "step": 1913000 }, { "epoch": 9.26, "learning_rate": 3.633214285714286e-05, "loss": 1.6766, "step": 1913500 }, { "epoch": 9.27, "learning_rate": 3.632857142857143e-05, "loss": 1.6801, "step": 1914000 }, { "epoch": 9.27, "learning_rate": 3.6325e-05, "loss": 1.6911, "step": 1914500 }, { "epoch": 9.27, "learning_rate": 3.6321428571428576e-05, "loss": 1.6747, "step": 1915000 }, { "epoch": 9.27, "learning_rate": 3.6317857142857144e-05, "loss": 1.6692, "step": 1915500 }, { "epoch": 9.28, "learning_rate": 3.631428571428572e-05, "loss": 1.6649, "step": 1916000 }, { "epoch": 9.28, "learning_rate": 3.6310714285714285e-05, "loss": 1.6902, "step": 1916500 }, { "epoch": 9.28, "learning_rate": 3.630714285714285e-05, "loss": 1.6841, "step": 1917000 }, { "epoch": 9.29, "learning_rate": 3.6303571428571434e-05, "loss": 1.6868, "step": 1917500 }, { "epoch": 9.29, "learning_rate": 3.63e-05, "loss": 1.6793, "step": 1918000 }, { "epoch": 9.29, "learning_rate": 3.6296428571428576e-05, "loss": 1.6807, "step": 1918500 }, { "epoch": 9.29, "learning_rate": 3.629285714285714e-05, "loss": 1.6835, "step": 1919000 }, { "epoch": 9.3, "learning_rate": 3.628928571428572e-05, "loss": 1.6828, "step": 1919500 }, { "epoch": 9.3, "learning_rate": 3.628571428571429e-05, "loss": 1.6722, "step": 1920000 }, { "epoch": 9.3, "learning_rate": 3.628214285714286e-05, "loss": 1.6834, "step": 1920500 }, { "epoch": 9.3, "learning_rate": 3.6278571428571427e-05, "loss": 1.6727, "step": 1921000 }, { "epoch": 9.31, "learning_rate": 3.6275e-05, "loss": 1.6911, "step": 1921500 }, { "epoch": 9.31, "learning_rate": 3.6271428571428575e-05, "loss": 1.6684, "step": 1922000 }, { "epoch": 9.31, "learning_rate": 3.626785714285714e-05, "loss": 1.6793, "step": 1922500 }, { "epoch": 9.31, "learning_rate": 3.626428571428572e-05, "loss": 1.6812, "step": 1923000 }, { "epoch": 9.32, "learning_rate": 3.6260714285714284e-05, "loss": 1.673, "step": 1923500 }, { "epoch": 9.32, "learning_rate": 3.625714285714286e-05, "loss": 1.6824, "step": 1924000 }, { "epoch": 9.32, "learning_rate": 3.625357142857143e-05, "loss": 1.6821, "step": 1924500 }, { "epoch": 9.33, "learning_rate": 3.625e-05, "loss": 1.673, "step": 1925000 }, { "epoch": 9.33, "learning_rate": 3.6246428571428574e-05, "loss": 1.6688, "step": 1925500 }, { "epoch": 9.33, "learning_rate": 3.624285714285714e-05, "loss": 1.6866, "step": 1926000 }, { "epoch": 9.33, "learning_rate": 3.6239285714285716e-05, "loss": 1.6772, "step": 1926500 }, { "epoch": 9.34, "learning_rate": 3.623571428571429e-05, "loss": 1.6749, "step": 1927000 }, { "epoch": 9.34, "learning_rate": 3.623214285714286e-05, "loss": 1.6858, "step": 1927500 }, { "epoch": 9.34, "learning_rate": 3.6228571428571425e-05, "loss": 1.6832, "step": 1928000 }, { "epoch": 9.34, "learning_rate": 3.6225000000000006e-05, "loss": 1.6933, "step": 1928500 }, { "epoch": 9.35, "learning_rate": 3.6221428571428574e-05, "loss": 1.6617, "step": 1929000 }, { "epoch": 9.35, "learning_rate": 3.621785714285714e-05, "loss": 1.6745, "step": 1929500 }, { "epoch": 9.35, "learning_rate": 3.6214285714285716e-05, "loss": 1.6788, "step": 1930000 }, { "epoch": 9.35, "learning_rate": 3.621071428571429e-05, "loss": 1.6757, "step": 1930500 }, { "epoch": 9.36, "learning_rate": 3.620714285714286e-05, "loss": 1.6776, "step": 1931000 }, { "epoch": 9.36, "learning_rate": 3.620357142857143e-05, "loss": 1.6702, "step": 1931500 }, { "epoch": 9.36, "learning_rate": 3.62e-05, "loss": 1.6737, "step": 1932000 }, { "epoch": 9.37, "learning_rate": 3.619642857142857e-05, "loss": 1.6746, "step": 1932500 }, { "epoch": 9.37, "learning_rate": 3.619285714285715e-05, "loss": 1.6828, "step": 1933000 }, { "epoch": 9.37, "learning_rate": 3.6189285714285715e-05, "loss": 1.6646, "step": 1933500 }, { "epoch": 9.37, "learning_rate": 3.618571428571429e-05, "loss": 1.6699, "step": 1934000 }, { "epoch": 9.38, "learning_rate": 3.618214285714286e-05, "loss": 1.6815, "step": 1934500 }, { "epoch": 9.38, "learning_rate": 3.617857142857143e-05, "loss": 1.6797, "step": 1935000 }, { "epoch": 9.38, "learning_rate": 3.6175000000000005e-05, "loss": 1.6779, "step": 1935500 }, { "epoch": 9.38, "learning_rate": 3.617142857142857e-05, "loss": 1.6706, "step": 1936000 }, { "epoch": 9.39, "learning_rate": 3.616785714285714e-05, "loss": 1.6821, "step": 1936500 }, { "epoch": 9.39, "learning_rate": 3.6164285714285714e-05, "loss": 1.683, "step": 1937000 }, { "epoch": 9.39, "learning_rate": 3.616071428571429e-05, "loss": 1.6728, "step": 1937500 }, { "epoch": 9.39, "learning_rate": 3.615714285714286e-05, "loss": 1.6765, "step": 1938000 }, { "epoch": 9.4, "learning_rate": 3.615357142857143e-05, "loss": 1.6784, "step": 1938500 }, { "epoch": 9.4, "learning_rate": 3.615e-05, "loss": 1.6813, "step": 1939000 }, { "epoch": 9.4, "learning_rate": 3.614642857142858e-05, "loss": 1.6633, "step": 1939500 }, { "epoch": 9.41, "learning_rate": 3.6142857142857146e-05, "loss": 1.6728, "step": 1940000 }, { "epoch": 9.41, "learning_rate": 3.6139285714285714e-05, "loss": 1.6781, "step": 1940500 }, { "epoch": 9.41, "learning_rate": 3.613571428571429e-05, "loss": 1.6762, "step": 1941000 }, { "epoch": 9.41, "learning_rate": 3.613214285714286e-05, "loss": 1.678, "step": 1941500 }, { "epoch": 9.42, "learning_rate": 3.612857142857143e-05, "loss": 1.6615, "step": 1942000 }, { "epoch": 9.42, "learning_rate": 3.6125000000000004e-05, "loss": 1.6811, "step": 1942500 }, { "epoch": 9.42, "learning_rate": 3.612142857142857e-05, "loss": 1.6733, "step": 1943000 }, { "epoch": 9.42, "learning_rate": 3.611785714285714e-05, "loss": 1.6702, "step": 1943500 }, { "epoch": 9.43, "learning_rate": 3.611428571428572e-05, "loss": 1.6864, "step": 1944000 }, { "epoch": 9.43, "learning_rate": 3.611071428571429e-05, "loss": 1.6812, "step": 1944500 }, { "epoch": 9.43, "learning_rate": 3.610714285714286e-05, "loss": 1.6696, "step": 1945000 }, { "epoch": 9.43, "learning_rate": 3.610357142857143e-05, "loss": 1.6811, "step": 1945500 }, { "epoch": 9.44, "learning_rate": 3.61e-05, "loss": 1.6685, "step": 1946000 }, { "epoch": 9.44, "learning_rate": 3.609642857142858e-05, "loss": 1.6762, "step": 1946500 }, { "epoch": 9.44, "learning_rate": 3.6092857142857145e-05, "loss": 1.6744, "step": 1947000 }, { "epoch": 9.44, "learning_rate": 3.608928571428571e-05, "loss": 1.6814, "step": 1947500 }, { "epoch": 9.45, "learning_rate": 3.608571428571429e-05, "loss": 1.6797, "step": 1948000 }, { "epoch": 9.45, "learning_rate": 3.608214285714286e-05, "loss": 1.6691, "step": 1948500 }, { "epoch": 9.45, "learning_rate": 3.607857142857143e-05, "loss": 1.6912, "step": 1949000 }, { "epoch": 9.46, "learning_rate": 3.6075e-05, "loss": 1.6739, "step": 1949500 }, { "epoch": 9.46, "learning_rate": 3.607142857142857e-05, "loss": 1.6707, "step": 1950000 }, { "epoch": 9.46, "learning_rate": 3.6067857142857144e-05, "loss": 1.6746, "step": 1950500 }, { "epoch": 9.46, "learning_rate": 3.606428571428572e-05, "loss": 1.6777, "step": 1951000 }, { "epoch": 9.47, "learning_rate": 3.6060714285714286e-05, "loss": 1.671, "step": 1951500 }, { "epoch": 9.47, "learning_rate": 3.605714285714286e-05, "loss": 1.6731, "step": 1952000 }, { "epoch": 9.47, "learning_rate": 3.605357142857143e-05, "loss": 1.6798, "step": 1952500 }, { "epoch": 9.47, "learning_rate": 3.605e-05, "loss": 1.6795, "step": 1953000 }, { "epoch": 9.48, "learning_rate": 3.6046428571428576e-05, "loss": 1.6721, "step": 1953500 }, { "epoch": 9.48, "learning_rate": 3.6042857142857144e-05, "loss": 1.6731, "step": 1954000 }, { "epoch": 9.48, "learning_rate": 3.603928571428571e-05, "loss": 1.6815, "step": 1954500 }, { "epoch": 9.48, "learning_rate": 3.603571428571429e-05, "loss": 1.6694, "step": 1955000 }, { "epoch": 9.49, "learning_rate": 3.603214285714286e-05, "loss": 1.6821, "step": 1955500 }, { "epoch": 9.49, "learning_rate": 3.602857142857143e-05, "loss": 1.6784, "step": 1956000 }, { "epoch": 9.49, "learning_rate": 3.6025e-05, "loss": 1.6874, "step": 1956500 }, { "epoch": 9.5, "learning_rate": 3.6021428571428576e-05, "loss": 1.6834, "step": 1957000 }, { "epoch": 9.5, "learning_rate": 3.601785714285714e-05, "loss": 1.6802, "step": 1957500 }, { "epoch": 9.5, "learning_rate": 3.601428571428572e-05, "loss": 1.6698, "step": 1958000 }, { "epoch": 9.5, "learning_rate": 3.6010714285714285e-05, "loss": 1.6732, "step": 1958500 }, { "epoch": 9.51, "learning_rate": 3.600714285714286e-05, "loss": 1.6753, "step": 1959000 }, { "epoch": 9.51, "learning_rate": 3.6003571428571433e-05, "loss": 1.6685, "step": 1959500 }, { "epoch": 9.51, "learning_rate": 3.6e-05, "loss": 1.6682, "step": 1960000 }, { "epoch": 9.51, "learning_rate": 3.5996428571428575e-05, "loss": 1.6617, "step": 1960500 }, { "epoch": 9.52, "learning_rate": 3.599285714285714e-05, "loss": 1.6724, "step": 1961000 }, { "epoch": 9.52, "learning_rate": 3.598928571428572e-05, "loss": 1.6598, "step": 1961500 }, { "epoch": 9.52, "learning_rate": 3.598571428571429e-05, "loss": 1.6648, "step": 1962000 }, { "epoch": 9.52, "learning_rate": 3.598214285714286e-05, "loss": 1.6696, "step": 1962500 }, { "epoch": 9.53, "learning_rate": 3.5978571428571426e-05, "loss": 1.6786, "step": 1963000 }, { "epoch": 9.53, "learning_rate": 3.5975e-05, "loss": 1.6748, "step": 1963500 }, { "epoch": 9.53, "learning_rate": 3.5971428571428575e-05, "loss": 1.6756, "step": 1964000 }, { "epoch": 9.54, "learning_rate": 3.596785714285714e-05, "loss": 1.667, "step": 1964500 }, { "epoch": 9.54, "learning_rate": 3.5964285714285716e-05, "loss": 1.6683, "step": 1965000 }, { "epoch": 9.54, "learning_rate": 3.5960714285714284e-05, "loss": 1.68, "step": 1965500 }, { "epoch": 9.54, "learning_rate": 3.5957142857142865e-05, "loss": 1.6764, "step": 1966000 }, { "epoch": 9.55, "learning_rate": 3.595357142857143e-05, "loss": 1.664, "step": 1966500 }, { "epoch": 9.55, "learning_rate": 3.595e-05, "loss": 1.6652, "step": 1967000 }, { "epoch": 9.55, "learning_rate": 3.5946428571428574e-05, "loss": 1.6696, "step": 1967500 }, { "epoch": 9.55, "learning_rate": 3.594285714285714e-05, "loss": 1.6793, "step": 1968000 }, { "epoch": 9.56, "learning_rate": 3.5939285714285716e-05, "loss": 1.6687, "step": 1968500 }, { "epoch": 9.56, "learning_rate": 3.593571428571429e-05, "loss": 1.662, "step": 1969000 }, { "epoch": 9.56, "learning_rate": 3.593214285714286e-05, "loss": 1.679, "step": 1969500 }, { "epoch": 9.56, "learning_rate": 3.5928571428571425e-05, "loss": 1.6655, "step": 1970000 }, { "epoch": 9.57, "learning_rate": 3.5925000000000006e-05, "loss": 1.6792, "step": 1970500 }, { "epoch": 9.57, "learning_rate": 3.592142857142857e-05, "loss": 1.6879, "step": 1971000 }, { "epoch": 9.57, "learning_rate": 3.591785714285714e-05, "loss": 1.6614, "step": 1971500 }, { "epoch": 9.58, "learning_rate": 3.5914285714285715e-05, "loss": 1.6726, "step": 1972000 }, { "epoch": 9.58, "learning_rate": 3.591071428571429e-05, "loss": 1.6637, "step": 1972500 }, { "epoch": 9.58, "learning_rate": 3.5907142857142864e-05, "loss": 1.6586, "step": 1973000 }, { "epoch": 9.58, "learning_rate": 3.590357142857143e-05, "loss": 1.6826, "step": 1973500 }, { "epoch": 9.59, "learning_rate": 3.59e-05, "loss": 1.6683, "step": 1974000 }, { "epoch": 9.59, "learning_rate": 3.589642857142857e-05, "loss": 1.6711, "step": 1974500 }, { "epoch": 9.59, "learning_rate": 3.589285714285715e-05, "loss": 1.6693, "step": 1975000 }, { "epoch": 9.59, "learning_rate": 3.5889285714285714e-05, "loss": 1.6747, "step": 1975500 }, { "epoch": 9.6, "learning_rate": 3.588571428571429e-05, "loss": 1.6649, "step": 1976000 }, { "epoch": 9.6, "learning_rate": 3.5882142857142856e-05, "loss": 1.6737, "step": 1976500 }, { "epoch": 9.6, "learning_rate": 3.587857142857143e-05, "loss": 1.6685, "step": 1977000 }, { "epoch": 9.6, "learning_rate": 3.5875000000000005e-05, "loss": 1.6683, "step": 1977500 }, { "epoch": 9.61, "learning_rate": 3.587142857142857e-05, "loss": 1.6769, "step": 1978000 }, { "epoch": 9.61, "learning_rate": 3.586785714285714e-05, "loss": 1.669, "step": 1978500 }, { "epoch": 9.61, "learning_rate": 3.5864285714285714e-05, "loss": 1.6758, "step": 1979000 }, { "epoch": 9.62, "learning_rate": 3.586071428571429e-05, "loss": 1.6683, "step": 1979500 }, { "epoch": 9.62, "learning_rate": 3.585714285714286e-05, "loss": 1.6751, "step": 1980000 }, { "epoch": 9.62, "learning_rate": 3.585357142857143e-05, "loss": 1.6756, "step": 1980500 }, { "epoch": 9.62, "learning_rate": 3.585e-05, "loss": 1.6757, "step": 1981000 }, { "epoch": 9.63, "learning_rate": 3.584642857142858e-05, "loss": 1.6698, "step": 1981500 }, { "epoch": 9.63, "learning_rate": 3.5842857142857146e-05, "loss": 1.6624, "step": 1982000 }, { "epoch": 9.63, "learning_rate": 3.583928571428571e-05, "loss": 1.6629, "step": 1982500 }, { "epoch": 9.63, "learning_rate": 3.583571428571429e-05, "loss": 1.6756, "step": 1983000 }, { "epoch": 9.64, "learning_rate": 3.583214285714286e-05, "loss": 1.6784, "step": 1983500 }, { "epoch": 9.64, "learning_rate": 3.582857142857143e-05, "loss": 1.6666, "step": 1984000 }, { "epoch": 9.64, "learning_rate": 3.5825000000000003e-05, "loss": 1.6781, "step": 1984500 }, { "epoch": 9.64, "learning_rate": 3.582142857142857e-05, "loss": 1.6643, "step": 1985000 }, { "epoch": 9.65, "learning_rate": 3.5817857142857145e-05, "loss": 1.6618, "step": 1985500 }, { "epoch": 9.65, "learning_rate": 3.581428571428572e-05, "loss": 1.6745, "step": 1986000 }, { "epoch": 9.65, "learning_rate": 3.581071428571429e-05, "loss": 1.675, "step": 1986500 }, { "epoch": 9.66, "learning_rate": 3.580714285714286e-05, "loss": 1.6703, "step": 1987000 }, { "epoch": 9.66, "learning_rate": 3.580357142857143e-05, "loss": 1.668, "step": 1987500 }, { "epoch": 9.66, "learning_rate": 3.58e-05, "loss": 1.6724, "step": 1988000 }, { "epoch": 9.66, "learning_rate": 3.579642857142858e-05, "loss": 1.6664, "step": 1988500 }, { "epoch": 9.67, "learning_rate": 3.5792857142857145e-05, "loss": 1.6635, "step": 1989000 }, { "epoch": 9.67, "learning_rate": 3.578928571428571e-05, "loss": 1.6789, "step": 1989500 }, { "epoch": 9.67, "learning_rate": 3.5785714285714286e-05, "loss": 1.6749, "step": 1990000 }, { "epoch": 9.67, "learning_rate": 3.578214285714286e-05, "loss": 1.6716, "step": 1990500 }, { "epoch": 9.68, "learning_rate": 3.577857142857143e-05, "loss": 1.6762, "step": 1991000 }, { "epoch": 9.68, "learning_rate": 3.5775e-05, "loss": 1.6724, "step": 1991500 }, { "epoch": 9.68, "learning_rate": 3.577142857142857e-05, "loss": 1.6804, "step": 1992000 }, { "epoch": 9.68, "learning_rate": 3.5767857142857144e-05, "loss": 1.6764, "step": 1992500 }, { "epoch": 9.69, "learning_rate": 3.576428571428572e-05, "loss": 1.6727, "step": 1993000 }, { "epoch": 9.69, "learning_rate": 3.5760714285714286e-05, "loss": 1.6726, "step": 1993500 }, { "epoch": 9.69, "learning_rate": 3.575714285714286e-05, "loss": 1.6748, "step": 1994000 }, { "epoch": 9.7, "learning_rate": 3.575357142857143e-05, "loss": 1.671, "step": 1994500 }, { "epoch": 9.7, "learning_rate": 3.575e-05, "loss": 1.6597, "step": 1995000 }, { "epoch": 9.7, "learning_rate": 3.5746428571428576e-05, "loss": 1.6788, "step": 1995500 }, { "epoch": 9.7, "learning_rate": 3.574285714285714e-05, "loss": 1.669, "step": 1996000 }, { "epoch": 9.71, "learning_rate": 3.573928571428571e-05, "loss": 1.6687, "step": 1996500 }, { "epoch": 9.71, "learning_rate": 3.573571428571429e-05, "loss": 1.6764, "step": 1997000 }, { "epoch": 9.71, "learning_rate": 3.573214285714286e-05, "loss": 1.6642, "step": 1997500 }, { "epoch": 9.71, "learning_rate": 3.572857142857143e-05, "loss": 1.6746, "step": 1998000 }, { "epoch": 9.72, "learning_rate": 3.5725e-05, "loss": 1.6733, "step": 1998500 }, { "epoch": 9.72, "learning_rate": 3.5721428571428575e-05, "loss": 1.6706, "step": 1999000 }, { "epoch": 9.72, "learning_rate": 3.571785714285714e-05, "loss": 1.6631, "step": 1999500 }, { "epoch": 9.72, "learning_rate": 3.571428571428572e-05, "loss": 1.6587, "step": 2000000 }, { "epoch": 9.73, "learning_rate": 3.5710714285714284e-05, "loss": 1.6664, "step": 2000500 }, { "epoch": 9.73, "learning_rate": 3.570714285714286e-05, "loss": 1.6636, "step": 2001000 }, { "epoch": 9.73, "learning_rate": 3.570357142857143e-05, "loss": 1.6691, "step": 2001500 }, { "epoch": 9.74, "learning_rate": 3.57e-05, "loss": 1.6749, "step": 2002000 }, { "epoch": 9.74, "learning_rate": 3.5696428571428575e-05, "loss": 1.6722, "step": 2002500 }, { "epoch": 9.74, "learning_rate": 3.569285714285714e-05, "loss": 1.6719, "step": 2003000 }, { "epoch": 9.74, "learning_rate": 3.5689285714285716e-05, "loss": 1.657, "step": 2003500 }, { "epoch": 9.75, "learning_rate": 3.568571428571429e-05, "loss": 1.664, "step": 2004000 }, { "epoch": 9.75, "learning_rate": 3.568214285714286e-05, "loss": 1.6723, "step": 2004500 }, { "epoch": 9.75, "learning_rate": 3.5678571428571426e-05, "loss": 1.6672, "step": 2005000 }, { "epoch": 9.75, "learning_rate": 3.5675e-05, "loss": 1.6632, "step": 2005500 }, { "epoch": 9.76, "learning_rate": 3.5671428571428574e-05, "loss": 1.6607, "step": 2006000 }, { "epoch": 9.76, "learning_rate": 3.566785714285715e-05, "loss": 1.6681, "step": 2006500 }, { "epoch": 9.76, "learning_rate": 3.5664285714285716e-05, "loss": 1.6769, "step": 2007000 }, { "epoch": 9.76, "learning_rate": 3.566071428571428e-05, "loss": 1.6654, "step": 2007500 }, { "epoch": 9.77, "learning_rate": 3.5657142857142864e-05, "loss": 1.6624, "step": 2008000 }, { "epoch": 9.77, "learning_rate": 3.565357142857143e-05, "loss": 1.6561, "step": 2008500 }, { "epoch": 9.77, "learning_rate": 3.565e-05, "loss": 1.6579, "step": 2009000 }, { "epoch": 9.78, "learning_rate": 3.5646428571428573e-05, "loss": 1.6763, "step": 2009500 }, { "epoch": 9.78, "learning_rate": 3.564285714285715e-05, "loss": 1.6715, "step": 2010000 }, { "epoch": 9.78, "learning_rate": 3.5639285714285715e-05, "loss": 1.6614, "step": 2010500 }, { "epoch": 9.78, "learning_rate": 3.563571428571429e-05, "loss": 1.6678, "step": 2011000 }, { "epoch": 9.79, "learning_rate": 3.563214285714286e-05, "loss": 1.6667, "step": 2011500 }, { "epoch": 9.79, "learning_rate": 3.562857142857143e-05, "loss": 1.6711, "step": 2012000 }, { "epoch": 9.79, "learning_rate": 3.5625000000000005e-05, "loss": 1.6526, "step": 2012500 }, { "epoch": 9.79, "learning_rate": 3.562142857142857e-05, "loss": 1.6723, "step": 2013000 }, { "epoch": 9.8, "learning_rate": 3.561785714285715e-05, "loss": 1.6638, "step": 2013500 }, { "epoch": 9.8, "learning_rate": 3.5614285714285715e-05, "loss": 1.6734, "step": 2014000 }, { "epoch": 9.8, "learning_rate": 3.561071428571429e-05, "loss": 1.6662, "step": 2014500 }, { "epoch": 9.8, "learning_rate": 3.560714285714286e-05, "loss": 1.6723, "step": 2015000 }, { "epoch": 9.81, "learning_rate": 3.560357142857143e-05, "loss": 1.6635, "step": 2015500 }, { "epoch": 9.81, "learning_rate": 3.56e-05, "loss": 1.6631, "step": 2016000 }, { "epoch": 9.81, "learning_rate": 3.559642857142857e-05, "loss": 1.6626, "step": 2016500 }, { "epoch": 9.82, "learning_rate": 3.5592857142857146e-05, "loss": 1.67, "step": 2017000 }, { "epoch": 9.82, "learning_rate": 3.5589285714285714e-05, "loss": 1.6581, "step": 2017500 }, { "epoch": 9.82, "learning_rate": 3.558571428571429e-05, "loss": 1.6506, "step": 2018000 }, { "epoch": 9.82, "learning_rate": 3.5582142857142856e-05, "loss": 1.6595, "step": 2018500 }, { "epoch": 9.83, "learning_rate": 3.557857142857143e-05, "loss": 1.6685, "step": 2019000 }, { "epoch": 9.83, "learning_rate": 3.5575000000000004e-05, "loss": 1.6537, "step": 2019500 }, { "epoch": 9.83, "learning_rate": 3.557142857142857e-05, "loss": 1.6629, "step": 2020000 }, { "epoch": 9.83, "learning_rate": 3.5567857142857146e-05, "loss": 1.6694, "step": 2020500 }, { "epoch": 9.84, "learning_rate": 3.556428571428571e-05, "loss": 1.6586, "step": 2021000 }, { "epoch": 9.84, "learning_rate": 3.556071428571429e-05, "loss": 1.6545, "step": 2021500 }, { "epoch": 9.84, "learning_rate": 3.555714285714286e-05, "loss": 1.6643, "step": 2022000 }, { "epoch": 9.84, "learning_rate": 3.555357142857143e-05, "loss": 1.6619, "step": 2022500 }, { "epoch": 9.85, "learning_rate": 3.555e-05, "loss": 1.6688, "step": 2023000 }, { "epoch": 9.85, "learning_rate": 3.554642857142858e-05, "loss": 1.6665, "step": 2023500 }, { "epoch": 9.85, "learning_rate": 3.5542857142857145e-05, "loss": 1.661, "step": 2024000 }, { "epoch": 9.85, "learning_rate": 3.553928571428571e-05, "loss": 1.6533, "step": 2024500 }, { "epoch": 9.86, "learning_rate": 3.553571428571429e-05, "loss": 1.6607, "step": 2025000 }, { "epoch": 9.86, "learning_rate": 3.553214285714286e-05, "loss": 1.6671, "step": 2025500 }, { "epoch": 9.86, "learning_rate": 3.552857142857143e-05, "loss": 1.6498, "step": 2026000 }, { "epoch": 9.87, "learning_rate": 3.5525e-05, "loss": 1.6597, "step": 2026500 }, { "epoch": 9.87, "learning_rate": 3.552142857142857e-05, "loss": 1.6602, "step": 2027000 }, { "epoch": 9.87, "learning_rate": 3.5517857142857145e-05, "loss": 1.6622, "step": 2027500 }, { "epoch": 9.87, "learning_rate": 3.551428571428572e-05, "loss": 1.6518, "step": 2028000 }, { "epoch": 9.88, "learning_rate": 3.5510714285714286e-05, "loss": 1.6669, "step": 2028500 }, { "epoch": 9.88, "learning_rate": 3.550714285714286e-05, "loss": 1.6586, "step": 2029000 }, { "epoch": 9.88, "learning_rate": 3.550357142857143e-05, "loss": 1.6692, "step": 2029500 }, { "epoch": 9.88, "learning_rate": 3.55e-05, "loss": 1.6546, "step": 2030000 }, { "epoch": 9.89, "learning_rate": 3.5496428571428577e-05, "loss": 1.6558, "step": 2030500 }, { "epoch": 9.89, "learning_rate": 3.5492857142857144e-05, "loss": 1.6519, "step": 2031000 }, { "epoch": 9.89, "learning_rate": 3.548928571428571e-05, "loss": 1.6639, "step": 2031500 }, { "epoch": 9.89, "learning_rate": 3.5485714285714286e-05, "loss": 1.6653, "step": 2032000 }, { "epoch": 9.9, "learning_rate": 3.548214285714286e-05, "loss": 1.6677, "step": 2032500 }, { "epoch": 9.9, "learning_rate": 3.547857142857143e-05, "loss": 1.6688, "step": 2033000 }, { "epoch": 9.9, "learning_rate": 3.5475e-05, "loss": 1.6608, "step": 2033500 }, { "epoch": 9.91, "learning_rate": 3.547142857142857e-05, "loss": 1.6606, "step": 2034000 }, { "epoch": 9.91, "learning_rate": 3.546785714285715e-05, "loss": 1.6624, "step": 2034500 }, { "epoch": 9.91, "learning_rate": 3.546428571428572e-05, "loss": 1.6534, "step": 2035000 }, { "epoch": 9.91, "learning_rate": 3.5460714285714285e-05, "loss": 1.6631, "step": 2035500 }, { "epoch": 9.92, "learning_rate": 3.545714285714286e-05, "loss": 1.6623, "step": 2036000 }, { "epoch": 9.92, "learning_rate": 3.5453571428571434e-05, "loss": 1.6591, "step": 2036500 }, { "epoch": 9.92, "learning_rate": 3.545e-05, "loss": 1.6639, "step": 2037000 }, { "epoch": 9.92, "learning_rate": 3.5446428571428575e-05, "loss": 1.6562, "step": 2037500 }, { "epoch": 9.93, "learning_rate": 3.544285714285714e-05, "loss": 1.6643, "step": 2038000 }, { "epoch": 9.93, "learning_rate": 3.543928571428572e-05, "loss": 1.6603, "step": 2038500 }, { "epoch": 9.93, "learning_rate": 3.543571428571429e-05, "loss": 1.6701, "step": 2039000 }, { "epoch": 9.93, "learning_rate": 3.543214285714286e-05, "loss": 1.6556, "step": 2039500 }, { "epoch": 9.94, "learning_rate": 3.5428571428571426e-05, "loss": 1.6613, "step": 2040000 }, { "epoch": 9.94, "learning_rate": 3.5425e-05, "loss": 1.6567, "step": 2040500 }, { "epoch": 9.94, "learning_rate": 3.5421428571428575e-05, "loss": 1.6632, "step": 2041000 }, { "epoch": 9.95, "learning_rate": 3.541785714285715e-05, "loss": 1.6573, "step": 2041500 }, { "epoch": 9.95, "learning_rate": 3.5414285714285716e-05, "loss": 1.681, "step": 2042000 }, { "epoch": 9.95, "learning_rate": 3.5410714285714284e-05, "loss": 1.6705, "step": 2042500 }, { "epoch": 9.95, "learning_rate": 3.540714285714286e-05, "loss": 1.6532, "step": 2043000 }, { "epoch": 9.96, "learning_rate": 3.540357142857143e-05, "loss": 1.6687, "step": 2043500 }, { "epoch": 9.96, "learning_rate": 3.54e-05, "loss": 1.6633, "step": 2044000 }, { "epoch": 9.96, "learning_rate": 3.5396428571428574e-05, "loss": 1.6503, "step": 2044500 }, { "epoch": 9.96, "learning_rate": 3.539285714285714e-05, "loss": 1.665, "step": 2045000 }, { "epoch": 9.97, "learning_rate": 3.5389285714285716e-05, "loss": 1.6669, "step": 2045500 }, { "epoch": 9.97, "learning_rate": 3.538571428571429e-05, "loss": 1.6699, "step": 2046000 }, { "epoch": 9.97, "learning_rate": 3.538214285714286e-05, "loss": 1.6593, "step": 2046500 }, { "epoch": 9.97, "learning_rate": 3.5378571428571425e-05, "loss": 1.6666, "step": 2047000 }, { "epoch": 9.98, "learning_rate": 3.5375e-05, "loss": 1.6577, "step": 2047500 }, { "epoch": 9.98, "learning_rate": 3.5371428571428574e-05, "loss": 1.6607, "step": 2048000 }, { "epoch": 9.98, "learning_rate": 3.536785714285715e-05, "loss": 1.6584, "step": 2048500 }, { "epoch": 9.99, "learning_rate": 3.5364285714285715e-05, "loss": 1.6556, "step": 2049000 }, { "epoch": 9.99, "learning_rate": 3.536071428571428e-05, "loss": 1.651, "step": 2049500 }, { "epoch": 9.99, "learning_rate": 3.5357142857142864e-05, "loss": 1.6457, "step": 2050000 }, { "epoch": 9.99, "learning_rate": 3.535357142857143e-05, "loss": 1.6538, "step": 2050500 }, { "epoch": 10.0, "learning_rate": 3.535e-05, "loss": 1.6575, "step": 2051000 }, { "epoch": 10.0, "learning_rate": 3.534642857142857e-05, "loss": 1.6626, "step": 2051500 }, { "epoch": 10.0, "learning_rate": 3.534285714285715e-05, "loss": 1.6575, "step": 2052000 }, { "epoch": 10.0, "learning_rate": 3.5339285714285715e-05, "loss": 1.6524, "step": 2052500 }, { "epoch": 10.01, "learning_rate": 3.533571428571429e-05, "loss": 1.6568, "step": 2053000 }, { "epoch": 10.01, "learning_rate": 3.5332142857142856e-05, "loss": 1.6543, "step": 2053500 }, { "epoch": 10.01, "learning_rate": 3.532857142857143e-05, "loss": 1.6458, "step": 2054000 }, { "epoch": 10.01, "learning_rate": 3.5325000000000005e-05, "loss": 1.6546, "step": 2054500 }, { "epoch": 10.02, "learning_rate": 3.532142857142857e-05, "loss": 1.6479, "step": 2055000 }, { "epoch": 10.02, "learning_rate": 3.5317857142857147e-05, "loss": 1.6548, "step": 2055500 }, { "epoch": 10.02, "learning_rate": 3.5314285714285714e-05, "loss": 1.6454, "step": 2056000 }, { "epoch": 10.03, "learning_rate": 3.531071428571429e-05, "loss": 1.6644, "step": 2056500 }, { "epoch": 10.03, "learning_rate": 3.530714285714286e-05, "loss": 1.6502, "step": 2057000 }, { "epoch": 10.03, "learning_rate": 3.530357142857143e-05, "loss": 1.6664, "step": 2057500 }, { "epoch": 10.03, "learning_rate": 3.53e-05, "loss": 1.6563, "step": 2058000 }, { "epoch": 10.04, "learning_rate": 3.529642857142857e-05, "loss": 1.6487, "step": 2058500 }, { "epoch": 10.04, "learning_rate": 3.5292857142857146e-05, "loss": 1.6441, "step": 2059000 }, { "epoch": 10.04, "learning_rate": 3.5289285714285713e-05, "loss": 1.6543, "step": 2059500 }, { "epoch": 10.04, "learning_rate": 3.528571428571429e-05, "loss": 1.665, "step": 2060000 }, { "epoch": 10.05, "learning_rate": 3.5282142857142855e-05, "loss": 1.645, "step": 2060500 }, { "epoch": 10.05, "learning_rate": 3.527857142857143e-05, "loss": 1.6446, "step": 2061000 }, { "epoch": 10.05, "learning_rate": 3.5275000000000004e-05, "loss": 1.6462, "step": 2061500 }, { "epoch": 10.05, "learning_rate": 3.527142857142857e-05, "loss": 1.6567, "step": 2062000 }, { "epoch": 10.06, "learning_rate": 3.5267857142857145e-05, "loss": 1.6535, "step": 2062500 }, { "epoch": 10.06, "learning_rate": 3.526428571428572e-05, "loss": 1.6485, "step": 2063000 }, { "epoch": 10.06, "learning_rate": 3.526071428571429e-05, "loss": 1.6517, "step": 2063500 }, { "epoch": 10.07, "learning_rate": 3.525714285714286e-05, "loss": 1.6623, "step": 2064000 }, { "epoch": 10.07, "learning_rate": 3.525357142857143e-05, "loss": 1.6529, "step": 2064500 }, { "epoch": 10.07, "learning_rate": 3.525e-05, "loss": 1.6537, "step": 2065000 }, { "epoch": 10.07, "learning_rate": 3.524642857142858e-05, "loss": 1.6519, "step": 2065500 }, { "epoch": 10.08, "learning_rate": 3.5242857142857145e-05, "loss": 1.6498, "step": 2066000 }, { "epoch": 10.08, "learning_rate": 3.523928571428571e-05, "loss": 1.6526, "step": 2066500 }, { "epoch": 10.08, "learning_rate": 3.5235714285714286e-05, "loss": 1.6493, "step": 2067000 }, { "epoch": 10.08, "learning_rate": 3.523214285714286e-05, "loss": 1.653, "step": 2067500 }, { "epoch": 10.09, "learning_rate": 3.5228571428571435e-05, "loss": 1.6641, "step": 2068000 }, { "epoch": 10.09, "learning_rate": 3.5225e-05, "loss": 1.6498, "step": 2068500 }, { "epoch": 10.09, "learning_rate": 3.522142857142857e-05, "loss": 1.642, "step": 2069000 }, { "epoch": 10.09, "learning_rate": 3.5217857142857144e-05, "loss": 1.6621, "step": 2069500 }, { "epoch": 10.1, "learning_rate": 3.521428571428572e-05, "loss": 1.6536, "step": 2070000 }, { "epoch": 10.1, "learning_rate": 3.5210714285714286e-05, "loss": 1.6488, "step": 2070500 }, { "epoch": 10.1, "learning_rate": 3.520714285714286e-05, "loss": 1.6548, "step": 2071000 }, { "epoch": 10.11, "learning_rate": 3.520357142857143e-05, "loss": 1.6525, "step": 2071500 }, { "epoch": 10.11, "learning_rate": 3.52e-05, "loss": 1.6523, "step": 2072000 }, { "epoch": 10.11, "learning_rate": 3.5196428571428576e-05, "loss": 1.6363, "step": 2072500 }, { "epoch": 10.11, "learning_rate": 3.5192857142857144e-05, "loss": 1.6502, "step": 2073000 }, { "epoch": 10.12, "learning_rate": 3.518928571428571e-05, "loss": 1.651, "step": 2073500 }, { "epoch": 10.12, "learning_rate": 3.5185714285714285e-05, "loss": 1.651, "step": 2074000 }, { "epoch": 10.12, "learning_rate": 3.518214285714286e-05, "loss": 1.6435, "step": 2074500 }, { "epoch": 10.12, "learning_rate": 3.5178571428571434e-05, "loss": 1.6553, "step": 2075000 }, { "epoch": 10.13, "learning_rate": 3.5175e-05, "loss": 1.6576, "step": 2075500 }, { "epoch": 10.13, "learning_rate": 3.517142857142857e-05, "loss": 1.6473, "step": 2076000 }, { "epoch": 10.13, "learning_rate": 3.516785714285715e-05, "loss": 1.6554, "step": 2076500 }, { "epoch": 10.13, "learning_rate": 3.516428571428572e-05, "loss": 1.6565, "step": 2077000 }, { "epoch": 10.14, "learning_rate": 3.5160714285714285e-05, "loss": 1.6484, "step": 2077500 }, { "epoch": 10.14, "learning_rate": 3.515714285714286e-05, "loss": 1.6505, "step": 2078000 }, { "epoch": 10.14, "learning_rate": 3.515357142857143e-05, "loss": 1.6431, "step": 2078500 }, { "epoch": 10.15, "learning_rate": 3.515e-05, "loss": 1.6613, "step": 2079000 }, { "epoch": 10.15, "learning_rate": 3.5146428571428575e-05, "loss": 1.6558, "step": 2079500 }, { "epoch": 10.15, "learning_rate": 3.514285714285714e-05, "loss": 1.6523, "step": 2080000 }, { "epoch": 10.15, "learning_rate": 3.5139285714285717e-05, "loss": 1.6476, "step": 2080500 }, { "epoch": 10.16, "learning_rate": 3.513571428571429e-05, "loss": 1.6539, "step": 2081000 }, { "epoch": 10.16, "learning_rate": 3.513214285714286e-05, "loss": 1.6477, "step": 2081500 }, { "epoch": 10.16, "learning_rate": 3.512857142857143e-05, "loss": 1.6465, "step": 2082000 }, { "epoch": 10.16, "learning_rate": 3.5125e-05, "loss": 1.6471, "step": 2082500 }, { "epoch": 10.17, "learning_rate": 3.5121428571428574e-05, "loss": 1.6542, "step": 2083000 }, { "epoch": 10.17, "learning_rate": 3.511785714285715e-05, "loss": 1.6605, "step": 2083500 }, { "epoch": 10.17, "learning_rate": 3.5114285714285716e-05, "loss": 1.6527, "step": 2084000 }, { "epoch": 10.17, "learning_rate": 3.5110714285714283e-05, "loss": 1.6513, "step": 2084500 }, { "epoch": 10.18, "learning_rate": 3.510714285714286e-05, "loss": 1.6488, "step": 2085000 }, { "epoch": 10.18, "learning_rate": 3.510357142857143e-05, "loss": 1.649, "step": 2085500 }, { "epoch": 10.18, "learning_rate": 3.51e-05, "loss": 1.6466, "step": 2086000 }, { "epoch": 10.19, "learning_rate": 3.5096428571428574e-05, "loss": 1.6386, "step": 2086500 }, { "epoch": 10.19, "learning_rate": 3.509285714285714e-05, "loss": 1.6509, "step": 2087000 }, { "epoch": 10.19, "learning_rate": 3.5089285714285715e-05, "loss": 1.6487, "step": 2087500 }, { "epoch": 10.19, "learning_rate": 3.508571428571429e-05, "loss": 1.6571, "step": 2088000 }, { "epoch": 10.2, "learning_rate": 3.508214285714286e-05, "loss": 1.6468, "step": 2088500 }, { "epoch": 10.2, "learning_rate": 3.507857142857143e-05, "loss": 1.6422, "step": 2089000 }, { "epoch": 10.2, "learning_rate": 3.5075000000000006e-05, "loss": 1.6498, "step": 2089500 }, { "epoch": 10.2, "learning_rate": 3.507142857142857e-05, "loss": 1.6545, "step": 2090000 }, { "epoch": 10.21, "learning_rate": 3.506785714285715e-05, "loss": 1.6462, "step": 2090500 }, { "epoch": 10.21, "learning_rate": 3.5064285714285715e-05, "loss": 1.6521, "step": 2091000 }, { "epoch": 10.21, "learning_rate": 3.506071428571429e-05, "loss": 1.6564, "step": 2091500 }, { "epoch": 10.21, "learning_rate": 3.505714285714286e-05, "loss": 1.6548, "step": 2092000 }, { "epoch": 10.22, "learning_rate": 3.505357142857143e-05, "loss": 1.6485, "step": 2092500 }, { "epoch": 10.22, "learning_rate": 3.505e-05, "loss": 1.6502, "step": 2093000 }, { "epoch": 10.22, "learning_rate": 3.504642857142857e-05, "loss": 1.6575, "step": 2093500 }, { "epoch": 10.22, "learning_rate": 3.504285714285715e-05, "loss": 1.6533, "step": 2094000 }, { "epoch": 10.23, "learning_rate": 3.5039285714285714e-05, "loss": 1.6571, "step": 2094500 }, { "epoch": 10.23, "learning_rate": 3.503571428571429e-05, "loss": 1.6563, "step": 2095000 }, { "epoch": 10.23, "learning_rate": 3.5032142857142856e-05, "loss": 1.6538, "step": 2095500 }, { "epoch": 10.24, "learning_rate": 3.502857142857143e-05, "loss": 1.6501, "step": 2096000 }, { "epoch": 10.24, "learning_rate": 3.5025000000000004e-05, "loss": 1.6467, "step": 2096500 }, { "epoch": 10.24, "learning_rate": 3.502142857142857e-05, "loss": 1.6422, "step": 2097000 }, { "epoch": 10.24, "learning_rate": 3.5017857142857146e-05, "loss": 1.6481, "step": 2097500 }, { "epoch": 10.25, "learning_rate": 3.5014285714285714e-05, "loss": 1.6519, "step": 2098000 }, { "epoch": 10.25, "learning_rate": 3.501071428571429e-05, "loss": 1.6433, "step": 2098500 }, { "epoch": 10.25, "learning_rate": 3.500714285714286e-05, "loss": 1.6356, "step": 2099000 }, { "epoch": 10.25, "learning_rate": 3.500357142857143e-05, "loss": 1.6505, "step": 2099500 }, { "epoch": 10.26, "learning_rate": 3.5e-05, "loss": 1.6447, "step": 2100000 }, { "epoch": 10.26, "learning_rate": 3.499642857142857e-05, "loss": 1.6436, "step": 2100500 }, { "epoch": 10.26, "learning_rate": 3.4992857142857145e-05, "loss": 1.6527, "step": 2101000 }, { "epoch": 10.26, "learning_rate": 3.498928571428571e-05, "loss": 1.6344, "step": 2101500 }, { "epoch": 10.27, "learning_rate": 3.498571428571429e-05, "loss": 1.653, "step": 2102000 }, { "epoch": 10.27, "learning_rate": 3.4982142857142855e-05, "loss": 1.6466, "step": 2102500 }, { "epoch": 10.27, "learning_rate": 3.4978571428571436e-05, "loss": 1.6398, "step": 2103000 }, { "epoch": 10.28, "learning_rate": 3.4975e-05, "loss": 1.6489, "step": 2103500 }, { "epoch": 10.28, "learning_rate": 3.497142857142857e-05, "loss": 1.6452, "step": 2104000 }, { "epoch": 10.28, "learning_rate": 3.4967857142857145e-05, "loss": 1.6528, "step": 2104500 }, { "epoch": 10.28, "learning_rate": 3.496428571428572e-05, "loss": 1.6517, "step": 2105000 }, { "epoch": 10.29, "learning_rate": 3.4960714285714287e-05, "loss": 1.6426, "step": 2105500 }, { "epoch": 10.29, "learning_rate": 3.495714285714286e-05, "loss": 1.642, "step": 2106000 }, { "epoch": 10.29, "learning_rate": 3.495357142857143e-05, "loss": 1.6511, "step": 2106500 }, { "epoch": 10.29, "learning_rate": 3.495e-05, "loss": 1.6583, "step": 2107000 }, { "epoch": 10.3, "learning_rate": 3.494642857142858e-05, "loss": 1.6396, "step": 2107500 }, { "epoch": 10.3, "learning_rate": 3.4942857142857144e-05, "loss": 1.6449, "step": 2108000 }, { "epoch": 10.3, "learning_rate": 3.493928571428571e-05, "loss": 1.6528, "step": 2108500 }, { "epoch": 10.3, "learning_rate": 3.4935714285714286e-05, "loss": 1.6432, "step": 2109000 }, { "epoch": 10.31, "learning_rate": 3.493214285714286e-05, "loss": 1.6473, "step": 2109500 }, { "epoch": 10.31, "learning_rate": 3.4928571428571434e-05, "loss": 1.6402, "step": 2110000 }, { "epoch": 10.31, "learning_rate": 3.4925e-05, "loss": 1.6471, "step": 2110500 }, { "epoch": 10.32, "learning_rate": 3.492142857142857e-05, "loss": 1.6448, "step": 2111000 }, { "epoch": 10.32, "learning_rate": 3.4917857142857144e-05, "loss": 1.6419, "step": 2111500 }, { "epoch": 10.32, "learning_rate": 3.491428571428572e-05, "loss": 1.6506, "step": 2112000 }, { "epoch": 10.32, "learning_rate": 3.4910714285714285e-05, "loss": 1.6428, "step": 2112500 }, { "epoch": 10.33, "learning_rate": 3.490714285714286e-05, "loss": 1.6571, "step": 2113000 }, { "epoch": 10.33, "learning_rate": 3.490357142857143e-05, "loss": 1.6551, "step": 2113500 }, { "epoch": 10.33, "learning_rate": 3.49e-05, "loss": 1.6514, "step": 2114000 }, { "epoch": 10.33, "learning_rate": 3.4896428571428576e-05, "loss": 1.647, "step": 2114500 }, { "epoch": 10.34, "learning_rate": 3.489285714285714e-05, "loss": 1.6373, "step": 2115000 }, { "epoch": 10.34, "learning_rate": 3.488928571428571e-05, "loss": 1.6485, "step": 2115500 }, { "epoch": 10.34, "learning_rate": 3.488571428571429e-05, "loss": 1.6469, "step": 2116000 }, { "epoch": 10.34, "learning_rate": 3.488214285714286e-05, "loss": 1.6427, "step": 2116500 }, { "epoch": 10.35, "learning_rate": 3.487857142857143e-05, "loss": 1.6486, "step": 2117000 }, { "epoch": 10.35, "learning_rate": 3.4875e-05, "loss": 1.6436, "step": 2117500 }, { "epoch": 10.35, "learning_rate": 3.4871428571428575e-05, "loss": 1.6428, "step": 2118000 }, { "epoch": 10.36, "learning_rate": 3.486785714285715e-05, "loss": 1.6432, "step": 2118500 }, { "epoch": 10.36, "learning_rate": 3.486428571428572e-05, "loss": 1.6471, "step": 2119000 }, { "epoch": 10.36, "learning_rate": 3.4860714285714284e-05, "loss": 1.6432, "step": 2119500 }, { "epoch": 10.36, "learning_rate": 3.485714285714286e-05, "loss": 1.6419, "step": 2120000 }, { "epoch": 10.37, "learning_rate": 3.485357142857143e-05, "loss": 1.6448, "step": 2120500 }, { "epoch": 10.37, "learning_rate": 3.485e-05, "loss": 1.6431, "step": 2121000 }, { "epoch": 10.37, "learning_rate": 3.4846428571428574e-05, "loss": 1.6381, "step": 2121500 }, { "epoch": 10.37, "learning_rate": 3.484285714285714e-05, "loss": 1.6414, "step": 2122000 }, { "epoch": 10.38, "learning_rate": 3.4839285714285716e-05, "loss": 1.6362, "step": 2122500 }, { "epoch": 10.38, "learning_rate": 3.483571428571429e-05, "loss": 1.6576, "step": 2123000 }, { "epoch": 10.38, "learning_rate": 3.483214285714286e-05, "loss": 1.6475, "step": 2123500 }, { "epoch": 10.38, "learning_rate": 3.482857142857143e-05, "loss": 1.6449, "step": 2124000 }, { "epoch": 10.39, "learning_rate": 3.4825e-05, "loss": 1.6462, "step": 2124500 }, { "epoch": 10.39, "learning_rate": 3.4821428571428574e-05, "loss": 1.6423, "step": 2125000 }, { "epoch": 10.39, "learning_rate": 3.481785714285715e-05, "loss": 1.6404, "step": 2125500 }, { "epoch": 10.4, "learning_rate": 3.4814285714285715e-05, "loss": 1.6501, "step": 2126000 }, { "epoch": 10.4, "learning_rate": 3.481071428571428e-05, "loss": 1.64, "step": 2126500 }, { "epoch": 10.4, "learning_rate": 3.480714285714286e-05, "loss": 1.6402, "step": 2127000 }, { "epoch": 10.4, "learning_rate": 3.480357142857143e-05, "loss": 1.6477, "step": 2127500 }, { "epoch": 10.41, "learning_rate": 3.48e-05, "loss": 1.6354, "step": 2128000 }, { "epoch": 10.41, "learning_rate": 3.479642857142857e-05, "loss": 1.6454, "step": 2128500 }, { "epoch": 10.41, "learning_rate": 3.479285714285714e-05, "loss": 1.6416, "step": 2129000 }, { "epoch": 10.41, "learning_rate": 3.478928571428572e-05, "loss": 1.6412, "step": 2129500 }, { "epoch": 10.42, "learning_rate": 3.478571428571429e-05, "loss": 1.65, "step": 2130000 }, { "epoch": 10.42, "learning_rate": 3.4782142857142857e-05, "loss": 1.6455, "step": 2130500 }, { "epoch": 10.42, "learning_rate": 3.477857142857143e-05, "loss": 1.648, "step": 2131000 }, { "epoch": 10.42, "learning_rate": 3.4775000000000005e-05, "loss": 1.645, "step": 2131500 }, { "epoch": 10.43, "learning_rate": 3.477142857142857e-05, "loss": 1.6465, "step": 2132000 }, { "epoch": 10.43, "learning_rate": 3.476785714285715e-05, "loss": 1.6422, "step": 2132500 }, { "epoch": 10.43, "learning_rate": 3.4764285714285714e-05, "loss": 1.6354, "step": 2133000 }, { "epoch": 10.44, "learning_rate": 3.476071428571429e-05, "loss": 1.6469, "step": 2133500 }, { "epoch": 10.44, "learning_rate": 3.475714285714286e-05, "loss": 1.6443, "step": 2134000 }, { "epoch": 10.44, "learning_rate": 3.475357142857143e-05, "loss": 1.6409, "step": 2134500 }, { "epoch": 10.44, "learning_rate": 3.475e-05, "loss": 1.6339, "step": 2135000 }, { "epoch": 10.45, "learning_rate": 3.474642857142857e-05, "loss": 1.6372, "step": 2135500 }, { "epoch": 10.45, "learning_rate": 3.4742857142857146e-05, "loss": 1.6369, "step": 2136000 }, { "epoch": 10.45, "learning_rate": 3.473928571428572e-05, "loss": 1.6468, "step": 2136500 }, { "epoch": 10.45, "learning_rate": 3.473571428571429e-05, "loss": 1.6424, "step": 2137000 }, { "epoch": 10.46, "learning_rate": 3.4732142857142855e-05, "loss": 1.6355, "step": 2137500 }, { "epoch": 10.46, "learning_rate": 3.472857142857143e-05, "loss": 1.6367, "step": 2138000 }, { "epoch": 10.46, "learning_rate": 3.4725000000000004e-05, "loss": 1.654, "step": 2138500 }, { "epoch": 10.46, "learning_rate": 3.472142857142857e-05, "loss": 1.653, "step": 2139000 }, { "epoch": 10.47, "learning_rate": 3.4717857142857146e-05, "loss": 1.6411, "step": 2139500 }, { "epoch": 10.47, "learning_rate": 3.471428571428571e-05, "loss": 1.6527, "step": 2140000 }, { "epoch": 10.47, "learning_rate": 3.471071428571429e-05, "loss": 1.6449, "step": 2140500 }, { "epoch": 10.48, "learning_rate": 3.470714285714286e-05, "loss": 1.6487, "step": 2141000 }, { "epoch": 10.48, "learning_rate": 3.470357142857143e-05, "loss": 1.6415, "step": 2141500 }, { "epoch": 10.48, "learning_rate": 3.4699999999999996e-05, "loss": 1.6458, "step": 2142000 }, { "epoch": 10.48, "learning_rate": 3.469642857142858e-05, "loss": 1.6393, "step": 2142500 }, { "epoch": 10.49, "learning_rate": 3.4692857142857145e-05, "loss": 1.6365, "step": 2143000 }, { "epoch": 10.49, "learning_rate": 3.468928571428572e-05, "loss": 1.6491, "step": 2143500 }, { "epoch": 10.49, "learning_rate": 3.468571428571429e-05, "loss": 1.639, "step": 2144000 }, { "epoch": 10.49, "learning_rate": 3.468214285714286e-05, "loss": 1.6426, "step": 2144500 }, { "epoch": 10.5, "learning_rate": 3.4678571428571435e-05, "loss": 1.6452, "step": 2145000 }, { "epoch": 10.5, "learning_rate": 3.4675e-05, "loss": 1.636, "step": 2145500 }, { "epoch": 10.5, "learning_rate": 3.467142857142857e-05, "loss": 1.6313, "step": 2146000 }, { "epoch": 10.5, "learning_rate": 3.4667857142857144e-05, "loss": 1.6431, "step": 2146500 }, { "epoch": 10.51, "learning_rate": 3.466428571428572e-05, "loss": 1.647, "step": 2147000 }, { "epoch": 10.51, "learning_rate": 3.4660714285714286e-05, "loss": 1.6426, "step": 2147500 }, { "epoch": 10.51, "learning_rate": 3.465714285714286e-05, "loss": 1.6467, "step": 2148000 }, { "epoch": 10.52, "learning_rate": 3.465357142857143e-05, "loss": 1.6484, "step": 2148500 }, { "epoch": 10.52, "learning_rate": 3.465e-05, "loss": 1.6428, "step": 2149000 }, { "epoch": 10.52, "learning_rate": 3.4646428571428576e-05, "loss": 1.6388, "step": 2149500 }, { "epoch": 10.52, "learning_rate": 3.4642857142857144e-05, "loss": 1.6513, "step": 2150000 }, { "epoch": 10.53, "learning_rate": 3.463928571428572e-05, "loss": 1.6402, "step": 2150500 }, { "epoch": 10.53, "learning_rate": 3.4635714285714285e-05, "loss": 1.6292, "step": 2151000 }, { "epoch": 10.53, "learning_rate": 3.463214285714286e-05, "loss": 1.6496, "step": 2151500 }, { "epoch": 10.53, "learning_rate": 3.4628571428571434e-05, "loss": 1.6398, "step": 2152000 }, { "epoch": 10.54, "learning_rate": 3.4625e-05, "loss": 1.6373, "step": 2152500 }, { "epoch": 10.54, "learning_rate": 3.462142857142857e-05, "loss": 1.6379, "step": 2153000 }, { "epoch": 10.54, "learning_rate": 3.461785714285714e-05, "loss": 1.6392, "step": 2153500 }, { "epoch": 10.54, "learning_rate": 3.461428571428572e-05, "loss": 1.6415, "step": 2154000 }, { "epoch": 10.55, "learning_rate": 3.4610714285714285e-05, "loss": 1.6434, "step": 2154500 }, { "epoch": 10.55, "learning_rate": 3.460714285714286e-05, "loss": 1.6439, "step": 2155000 }, { "epoch": 10.55, "learning_rate": 3.4603571428571427e-05, "loss": 1.6435, "step": 2155500 }, { "epoch": 10.56, "learning_rate": 3.46e-05, "loss": 1.6414, "step": 2156000 }, { "epoch": 10.56, "learning_rate": 3.4596428571428575e-05, "loss": 1.6536, "step": 2156500 }, { "epoch": 10.56, "learning_rate": 3.459285714285714e-05, "loss": 1.632, "step": 2157000 }, { "epoch": 10.56, "learning_rate": 3.458928571428572e-05, "loss": 1.6522, "step": 2157500 }, { "epoch": 10.57, "learning_rate": 3.458571428571429e-05, "loss": 1.6421, "step": 2158000 }, { "epoch": 10.57, "learning_rate": 3.458214285714286e-05, "loss": 1.636, "step": 2158500 }, { "epoch": 10.57, "learning_rate": 3.457857142857143e-05, "loss": 1.6418, "step": 2159000 }, { "epoch": 10.57, "learning_rate": 3.4575e-05, "loss": 1.6443, "step": 2159500 }, { "epoch": 10.58, "learning_rate": 3.4571428571428574e-05, "loss": 1.6432, "step": 2160000 }, { "epoch": 10.58, "learning_rate": 3.456785714285715e-05, "loss": 1.6611, "step": 2160500 }, { "epoch": 10.58, "learning_rate": 3.4564285714285716e-05, "loss": 1.6415, "step": 2161000 }, { "epoch": 10.58, "learning_rate": 3.4560714285714284e-05, "loss": 1.636, "step": 2161500 }, { "epoch": 10.59, "learning_rate": 3.455714285714286e-05, "loss": 1.6375, "step": 2162000 }, { "epoch": 10.59, "learning_rate": 3.455357142857143e-05, "loss": 1.6429, "step": 2162500 }, { "epoch": 10.59, "learning_rate": 3.455e-05, "loss": 1.6452, "step": 2163000 }, { "epoch": 10.6, "learning_rate": 3.4546428571428574e-05, "loss": 1.6383, "step": 2163500 }, { "epoch": 10.6, "learning_rate": 3.454285714285714e-05, "loss": 1.6478, "step": 2164000 }, { "epoch": 10.6, "learning_rate": 3.4539285714285716e-05, "loss": 1.6397, "step": 2164500 }, { "epoch": 10.6, "learning_rate": 3.453571428571429e-05, "loss": 1.6304, "step": 2165000 }, { "epoch": 10.61, "learning_rate": 3.453214285714286e-05, "loss": 1.6466, "step": 2165500 }, { "epoch": 10.61, "learning_rate": 3.452857142857143e-05, "loss": 1.639, "step": 2166000 }, { "epoch": 10.61, "learning_rate": 3.4525e-05, "loss": 1.643, "step": 2166500 }, { "epoch": 10.61, "learning_rate": 3.452142857142857e-05, "loss": 1.6345, "step": 2167000 }, { "epoch": 10.62, "learning_rate": 3.451785714285715e-05, "loss": 1.633, "step": 2167500 }, { "epoch": 10.62, "learning_rate": 3.4514285714285715e-05, "loss": 1.6335, "step": 2168000 }, { "epoch": 10.62, "learning_rate": 3.451071428571428e-05, "loss": 1.6337, "step": 2168500 }, { "epoch": 10.62, "learning_rate": 3.4507142857142863e-05, "loss": 1.6358, "step": 2169000 }, { "epoch": 10.63, "learning_rate": 3.450357142857143e-05, "loss": 1.6354, "step": 2169500 }, { "epoch": 10.63, "learning_rate": 3.45e-05, "loss": 1.6393, "step": 2170000 }, { "epoch": 10.63, "learning_rate": 3.449642857142857e-05, "loss": 1.6391, "step": 2170500 }, { "epoch": 10.63, "learning_rate": 3.449285714285714e-05, "loss": 1.6463, "step": 2171000 }, { "epoch": 10.64, "learning_rate": 3.448928571428572e-05, "loss": 1.6348, "step": 2171500 }, { "epoch": 10.64, "learning_rate": 3.448571428571429e-05, "loss": 1.6358, "step": 2172000 }, { "epoch": 10.64, "learning_rate": 3.4482142857142856e-05, "loss": 1.6313, "step": 2172500 }, { "epoch": 10.65, "learning_rate": 3.447857142857143e-05, "loss": 1.631, "step": 2173000 }, { "epoch": 10.65, "learning_rate": 3.4475000000000005e-05, "loss": 1.6394, "step": 2173500 }, { "epoch": 10.65, "learning_rate": 3.447142857142857e-05, "loss": 1.6305, "step": 2174000 }, { "epoch": 10.65, "learning_rate": 3.4467857142857146e-05, "loss": 1.6457, "step": 2174500 }, { "epoch": 10.66, "learning_rate": 3.4464285714285714e-05, "loss": 1.635, "step": 2175000 }, { "epoch": 10.66, "learning_rate": 3.446071428571429e-05, "loss": 1.6391, "step": 2175500 }, { "epoch": 10.66, "learning_rate": 3.445714285714286e-05, "loss": 1.6406, "step": 2176000 }, { "epoch": 10.66, "learning_rate": 3.445357142857143e-05, "loss": 1.6435, "step": 2176500 }, { "epoch": 10.67, "learning_rate": 3.445e-05, "loss": 1.6375, "step": 2177000 }, { "epoch": 10.67, "learning_rate": 3.444642857142857e-05, "loss": 1.6386, "step": 2177500 }, { "epoch": 10.67, "learning_rate": 3.4442857142857146e-05, "loss": 1.6389, "step": 2178000 }, { "epoch": 10.67, "learning_rate": 3.443928571428572e-05, "loss": 1.6327, "step": 2178500 }, { "epoch": 10.68, "learning_rate": 3.443571428571429e-05, "loss": 1.6418, "step": 2179000 }, { "epoch": 10.68, "learning_rate": 3.4432142857142855e-05, "loss": 1.6502, "step": 2179500 }, { "epoch": 10.68, "learning_rate": 3.442857142857143e-05, "loss": 1.6383, "step": 2180000 }, { "epoch": 10.69, "learning_rate": 3.4425e-05, "loss": 1.6402, "step": 2180500 }, { "epoch": 10.69, "learning_rate": 3.442142857142857e-05, "loss": 1.6435, "step": 2181000 }, { "epoch": 10.69, "learning_rate": 3.4417857142857145e-05, "loss": 1.6401, "step": 2181500 }, { "epoch": 10.69, "learning_rate": 3.441428571428571e-05, "loss": 1.6311, "step": 2182000 }, { "epoch": 10.7, "learning_rate": 3.441071428571429e-05, "loss": 1.6439, "step": 2182500 }, { "epoch": 10.7, "learning_rate": 3.440714285714286e-05, "loss": 1.6473, "step": 2183000 }, { "epoch": 10.7, "learning_rate": 3.440357142857143e-05, "loss": 1.6412, "step": 2183500 }, { "epoch": 10.7, "learning_rate": 3.4399999999999996e-05, "loss": 1.6352, "step": 2184000 }, { "epoch": 10.71, "learning_rate": 3.439642857142858e-05, "loss": 1.6342, "step": 2184500 }, { "epoch": 10.71, "learning_rate": 3.4392857142857144e-05, "loss": 1.6344, "step": 2185000 }, { "epoch": 10.71, "learning_rate": 3.438928571428572e-05, "loss": 1.6441, "step": 2185500 }, { "epoch": 10.71, "learning_rate": 3.4385714285714286e-05, "loss": 1.6386, "step": 2186000 }, { "epoch": 10.72, "learning_rate": 3.438214285714286e-05, "loss": 1.6439, "step": 2186500 }, { "epoch": 10.72, "learning_rate": 3.4378571428571435e-05, "loss": 1.6341, "step": 2187000 }, { "epoch": 10.72, "learning_rate": 3.4375e-05, "loss": 1.6359, "step": 2187500 }, { "epoch": 10.73, "learning_rate": 3.437142857142857e-05, "loss": 1.6343, "step": 2188000 }, { "epoch": 10.73, "learning_rate": 3.4367857142857144e-05, "loss": 1.6417, "step": 2188500 }, { "epoch": 10.73, "learning_rate": 3.436428571428572e-05, "loss": 1.6337, "step": 2189000 }, { "epoch": 10.73, "learning_rate": 3.4360714285714286e-05, "loss": 1.6414, "step": 2189500 }, { "epoch": 10.74, "learning_rate": 3.435714285714286e-05, "loss": 1.6398, "step": 2190000 }, { "epoch": 10.74, "learning_rate": 3.435357142857143e-05, "loss": 1.6465, "step": 2190500 }, { "epoch": 10.74, "learning_rate": 3.435e-05, "loss": 1.632, "step": 2191000 }, { "epoch": 10.74, "learning_rate": 3.4346428571428576e-05, "loss": 1.6362, "step": 2191500 }, { "epoch": 10.75, "learning_rate": 3.434285714285714e-05, "loss": 1.6365, "step": 2192000 }, { "epoch": 10.75, "learning_rate": 3.433928571428572e-05, "loss": 1.6403, "step": 2192500 }, { "epoch": 10.75, "learning_rate": 3.4335714285714285e-05, "loss": 1.6344, "step": 2193000 }, { "epoch": 10.75, "learning_rate": 3.433214285714286e-05, "loss": 1.6439, "step": 2193500 }, { "epoch": 10.76, "learning_rate": 3.432857142857143e-05, "loss": 1.6441, "step": 2194000 }, { "epoch": 10.76, "learning_rate": 3.4325e-05, "loss": 1.641, "step": 2194500 }, { "epoch": 10.76, "learning_rate": 3.432142857142857e-05, "loss": 1.6376, "step": 2195000 }, { "epoch": 10.77, "learning_rate": 3.431785714285715e-05, "loss": 1.6375, "step": 2195500 }, { "epoch": 10.77, "learning_rate": 3.431428571428572e-05, "loss": 1.6305, "step": 2196000 }, { "epoch": 10.77, "learning_rate": 3.4310714285714284e-05, "loss": 1.6402, "step": 2196500 }, { "epoch": 10.77, "learning_rate": 3.430714285714286e-05, "loss": 1.6272, "step": 2197000 }, { "epoch": 10.78, "learning_rate": 3.4303571428571426e-05, "loss": 1.6352, "step": 2197500 }, { "epoch": 10.78, "learning_rate": 3.430000000000001e-05, "loss": 1.6245, "step": 2198000 }, { "epoch": 10.78, "learning_rate": 3.4296428571428574e-05, "loss": 1.6436, "step": 2198500 }, { "epoch": 10.78, "learning_rate": 3.429285714285714e-05, "loss": 1.6442, "step": 2199000 }, { "epoch": 10.79, "learning_rate": 3.4289285714285716e-05, "loss": 1.6301, "step": 2199500 }, { "epoch": 10.79, "learning_rate": 3.428571428571429e-05, "loss": 1.6307, "step": 2200000 }, { "epoch": 10.79, "learning_rate": 3.428214285714286e-05, "loss": 1.646, "step": 2200500 }, { "epoch": 10.79, "learning_rate": 3.427857142857143e-05, "loss": 1.6311, "step": 2201000 }, { "epoch": 10.8, "learning_rate": 3.4275e-05, "loss": 1.6394, "step": 2201500 }, { "epoch": 10.8, "learning_rate": 3.4271428571428574e-05, "loss": 1.6414, "step": 2202000 }, { "epoch": 10.8, "learning_rate": 3.426785714285715e-05, "loss": 1.6378, "step": 2202500 }, { "epoch": 10.81, "learning_rate": 3.4264285714285716e-05, "loss": 1.6324, "step": 2203000 }, { "epoch": 10.81, "learning_rate": 3.426071428571428e-05, "loss": 1.6277, "step": 2203500 }, { "epoch": 10.81, "learning_rate": 3.425714285714286e-05, "loss": 1.6382, "step": 2204000 }, { "epoch": 10.81, "learning_rate": 3.425357142857143e-05, "loss": 1.6316, "step": 2204500 }, { "epoch": 10.82, "learning_rate": 3.4250000000000006e-05, "loss": 1.6453, "step": 2205000 }, { "epoch": 10.82, "learning_rate": 3.424642857142857e-05, "loss": 1.6419, "step": 2205500 }, { "epoch": 10.82, "learning_rate": 3.424285714285714e-05, "loss": 1.6382, "step": 2206000 }, { "epoch": 10.82, "learning_rate": 3.4239285714285715e-05, "loss": 1.6402, "step": 2206500 }, { "epoch": 10.83, "learning_rate": 3.423571428571429e-05, "loss": 1.636, "step": 2207000 }, { "epoch": 10.83, "learning_rate": 3.423214285714286e-05, "loss": 1.6316, "step": 2207500 }, { "epoch": 10.83, "learning_rate": 3.422857142857143e-05, "loss": 1.6328, "step": 2208000 }, { "epoch": 10.83, "learning_rate": 3.4225e-05, "loss": 1.623, "step": 2208500 }, { "epoch": 10.84, "learning_rate": 3.422142857142857e-05, "loss": 1.6313, "step": 2209000 }, { "epoch": 10.84, "learning_rate": 3.421785714285715e-05, "loss": 1.6398, "step": 2209500 }, { "epoch": 10.84, "learning_rate": 3.4214285714285714e-05, "loss": 1.6465, "step": 2210000 }, { "epoch": 10.85, "learning_rate": 3.421071428571428e-05, "loss": 1.6314, "step": 2210500 }, { "epoch": 10.85, "learning_rate": 3.420714285714286e-05, "loss": 1.6292, "step": 2211000 }, { "epoch": 10.85, "learning_rate": 3.420357142857143e-05, "loss": 1.6466, "step": 2211500 }, { "epoch": 10.85, "learning_rate": 3.4200000000000005e-05, "loss": 1.6387, "step": 2212000 }, { "epoch": 10.86, "learning_rate": 3.419642857142857e-05, "loss": 1.6265, "step": 2212500 }, { "epoch": 10.86, "learning_rate": 3.4192857142857146e-05, "loss": 1.6246, "step": 2213000 }, { "epoch": 10.86, "learning_rate": 3.418928571428572e-05, "loss": 1.6397, "step": 2213500 }, { "epoch": 10.86, "learning_rate": 3.418571428571429e-05, "loss": 1.6386, "step": 2214000 }, { "epoch": 10.87, "learning_rate": 3.4182142857142855e-05, "loss": 1.6328, "step": 2214500 }, { "epoch": 10.87, "learning_rate": 3.417857142857143e-05, "loss": 1.6497, "step": 2215000 }, { "epoch": 10.87, "learning_rate": 3.4175000000000004e-05, "loss": 1.6396, "step": 2215500 }, { "epoch": 10.87, "learning_rate": 3.417142857142857e-05, "loss": 1.6331, "step": 2216000 }, { "epoch": 10.88, "learning_rate": 3.4167857142857146e-05, "loss": 1.6395, "step": 2216500 }, { "epoch": 10.88, "learning_rate": 3.416428571428571e-05, "loss": 1.6296, "step": 2217000 }, { "epoch": 10.88, "learning_rate": 3.416071428571429e-05, "loss": 1.6427, "step": 2217500 }, { "epoch": 10.89, "learning_rate": 3.415714285714286e-05, "loss": 1.6289, "step": 2218000 }, { "epoch": 10.89, "learning_rate": 3.415357142857143e-05, "loss": 1.6287, "step": 2218500 }, { "epoch": 10.89, "learning_rate": 3.415e-05, "loss": 1.635, "step": 2219000 }, { "epoch": 10.89, "learning_rate": 3.414642857142857e-05, "loss": 1.6317, "step": 2219500 }, { "epoch": 10.9, "learning_rate": 3.4142857142857145e-05, "loss": 1.6328, "step": 2220000 }, { "epoch": 10.9, "learning_rate": 3.413928571428572e-05, "loss": 1.6305, "step": 2220500 }, { "epoch": 10.9, "learning_rate": 3.413571428571429e-05, "loss": 1.6364, "step": 2221000 }, { "epoch": 10.9, "learning_rate": 3.4132142857142854e-05, "loss": 1.6384, "step": 2221500 }, { "epoch": 10.91, "learning_rate": 3.4128571428571435e-05, "loss": 1.6348, "step": 2222000 }, { "epoch": 10.91, "learning_rate": 3.4125e-05, "loss": 1.642, "step": 2222500 }, { "epoch": 10.91, "learning_rate": 3.412142857142857e-05, "loss": 1.6392, "step": 2223000 }, { "epoch": 10.91, "learning_rate": 3.4117857142857144e-05, "loss": 1.6307, "step": 2223500 }, { "epoch": 10.92, "learning_rate": 3.411428571428571e-05, "loss": 1.6205, "step": 2224000 }, { "epoch": 10.92, "learning_rate": 3.4110714285714286e-05, "loss": 1.6384, "step": 2224500 }, { "epoch": 10.92, "learning_rate": 3.410714285714286e-05, "loss": 1.6185, "step": 2225000 }, { "epoch": 10.93, "learning_rate": 3.410357142857143e-05, "loss": 1.6417, "step": 2225500 }, { "epoch": 10.93, "learning_rate": 3.41e-05, "loss": 1.6252, "step": 2226000 }, { "epoch": 10.93, "learning_rate": 3.4096428571428576e-05, "loss": 1.6344, "step": 2226500 }, { "epoch": 10.93, "learning_rate": 3.4092857142857144e-05, "loss": 1.6333, "step": 2227000 }, { "epoch": 10.94, "learning_rate": 3.408928571428572e-05, "loss": 1.6367, "step": 2227500 }, { "epoch": 10.94, "learning_rate": 3.4085714285714286e-05, "loss": 1.6402, "step": 2228000 }, { "epoch": 10.94, "learning_rate": 3.408214285714286e-05, "loss": 1.6345, "step": 2228500 }, { "epoch": 10.94, "learning_rate": 3.4078571428571434e-05, "loss": 1.6334, "step": 2229000 }, { "epoch": 10.95, "learning_rate": 3.4075e-05, "loss": 1.6258, "step": 2229500 }, { "epoch": 10.95, "learning_rate": 3.407142857142857e-05, "loss": 1.6377, "step": 2230000 }, { "epoch": 10.95, "learning_rate": 3.406785714285714e-05, "loss": 1.6304, "step": 2230500 }, { "epoch": 10.95, "learning_rate": 3.406428571428572e-05, "loss": 1.6369, "step": 2231000 }, { "epoch": 10.96, "learning_rate": 3.4060714285714285e-05, "loss": 1.6326, "step": 2231500 }, { "epoch": 10.96, "learning_rate": 3.405714285714286e-05, "loss": 1.6321, "step": 2232000 }, { "epoch": 10.96, "learning_rate": 3.405357142857143e-05, "loss": 1.6321, "step": 2232500 }, { "epoch": 10.97, "learning_rate": 3.405e-05, "loss": 1.6329, "step": 2233000 }, { "epoch": 10.97, "learning_rate": 3.4046428571428575e-05, "loss": 1.6353, "step": 2233500 }, { "epoch": 10.97, "learning_rate": 3.404285714285714e-05, "loss": 1.6283, "step": 2234000 }, { "epoch": 10.97, "learning_rate": 3.403928571428572e-05, "loss": 1.6328, "step": 2234500 }, { "epoch": 10.98, "learning_rate": 3.4035714285714284e-05, "loss": 1.6398, "step": 2235000 }, { "epoch": 10.98, "learning_rate": 3.403214285714286e-05, "loss": 1.6241, "step": 2235500 }, { "epoch": 10.98, "learning_rate": 3.402857142857143e-05, "loss": 1.6335, "step": 2236000 }, { "epoch": 10.98, "learning_rate": 3.4025e-05, "loss": 1.6286, "step": 2236500 }, { "epoch": 10.99, "learning_rate": 3.402142857142857e-05, "loss": 1.6211, "step": 2237000 }, { "epoch": 10.99, "learning_rate": 3.401785714285715e-05, "loss": 1.6319, "step": 2237500 }, { "epoch": 10.99, "learning_rate": 3.4014285714285716e-05, "loss": 1.62, "step": 2238000 }, { "epoch": 10.99, "learning_rate": 3.4010714285714284e-05, "loss": 1.622, "step": 2238500 }, { "epoch": 11.0, "learning_rate": 3.400714285714286e-05, "loss": 1.6303, "step": 2239000 }, { "epoch": 11.0, "learning_rate": 3.400357142857143e-05, "loss": 1.6302, "step": 2239500 }, { "epoch": 11.0, "learning_rate": 3.4000000000000007e-05, "loss": 1.6199, "step": 2240000 }, { "epoch": 11.01, "learning_rate": 3.3996428571428574e-05, "loss": 1.6317, "step": 2240500 }, { "epoch": 11.01, "learning_rate": 3.399285714285714e-05, "loss": 1.6338, "step": 2241000 }, { "epoch": 11.01, "learning_rate": 3.3989285714285716e-05, "loss": 1.6258, "step": 2241500 }, { "epoch": 11.01, "learning_rate": 3.398571428571429e-05, "loss": 1.6234, "step": 2242000 }, { "epoch": 11.02, "learning_rate": 3.398214285714286e-05, "loss": 1.6225, "step": 2242500 }, { "epoch": 11.02, "learning_rate": 3.397857142857143e-05, "loss": 1.6075, "step": 2243000 }, { "epoch": 11.02, "learning_rate": 3.3975e-05, "loss": 1.6201, "step": 2243500 }, { "epoch": 11.02, "learning_rate": 3.397142857142857e-05, "loss": 1.6299, "step": 2244000 }, { "epoch": 11.03, "learning_rate": 3.396785714285715e-05, "loss": 1.6363, "step": 2244500 }, { "epoch": 11.03, "learning_rate": 3.3964285714285715e-05, "loss": 1.638, "step": 2245000 }, { "epoch": 11.03, "learning_rate": 3.396071428571428e-05, "loss": 1.6196, "step": 2245500 }, { "epoch": 11.03, "learning_rate": 3.395714285714286e-05, "loss": 1.6214, "step": 2246000 }, { "epoch": 11.04, "learning_rate": 3.395357142857143e-05, "loss": 1.6283, "step": 2246500 }, { "epoch": 11.04, "learning_rate": 3.3950000000000005e-05, "loss": 1.6207, "step": 2247000 }, { "epoch": 11.04, "learning_rate": 3.394642857142857e-05, "loss": 1.6228, "step": 2247500 }, { "epoch": 11.04, "learning_rate": 3.394285714285714e-05, "loss": 1.6229, "step": 2248000 }, { "epoch": 11.05, "learning_rate": 3.393928571428572e-05, "loss": 1.6215, "step": 2248500 }, { "epoch": 11.05, "learning_rate": 3.393571428571429e-05, "loss": 1.6263, "step": 2249000 }, { "epoch": 11.05, "learning_rate": 3.3932142857142856e-05, "loss": 1.6195, "step": 2249500 }, { "epoch": 11.06, "learning_rate": 3.392857142857143e-05, "loss": 1.6215, "step": 2250000 }, { "epoch": 11.06, "learning_rate": 3.3925e-05, "loss": 1.6244, "step": 2250500 }, { "epoch": 11.06, "learning_rate": 3.392142857142857e-05, "loss": 1.6194, "step": 2251000 }, { "epoch": 11.06, "learning_rate": 3.3917857142857146e-05, "loss": 1.6189, "step": 2251500 }, { "epoch": 11.07, "learning_rate": 3.3914285714285714e-05, "loss": 1.6348, "step": 2252000 }, { "epoch": 11.07, "learning_rate": 3.391071428571429e-05, "loss": 1.6193, "step": 2252500 }, { "epoch": 11.07, "learning_rate": 3.390714285714286e-05, "loss": 1.6215, "step": 2253000 }, { "epoch": 11.07, "learning_rate": 3.390357142857143e-05, "loss": 1.6222, "step": 2253500 }, { "epoch": 11.08, "learning_rate": 3.3900000000000004e-05, "loss": 1.6279, "step": 2254000 }, { "epoch": 11.08, "learning_rate": 3.389642857142857e-05, "loss": 1.6221, "step": 2254500 }, { "epoch": 11.08, "learning_rate": 3.3892857142857146e-05, "loss": 1.635, "step": 2255000 }, { "epoch": 11.08, "learning_rate": 3.388928571428572e-05, "loss": 1.619, "step": 2255500 }, { "epoch": 11.09, "learning_rate": 3.388571428571429e-05, "loss": 1.633, "step": 2256000 }, { "epoch": 11.09, "learning_rate": 3.3882142857142855e-05, "loss": 1.6293, "step": 2256500 }, { "epoch": 11.09, "learning_rate": 3.387857142857143e-05, "loss": 1.6209, "step": 2257000 }, { "epoch": 11.1, "learning_rate": 3.3875000000000003e-05, "loss": 1.6183, "step": 2257500 }, { "epoch": 11.1, "learning_rate": 3.387142857142857e-05, "loss": 1.6314, "step": 2258000 }, { "epoch": 11.1, "learning_rate": 3.3867857142857145e-05, "loss": 1.6224, "step": 2258500 }, { "epoch": 11.1, "learning_rate": 3.386428571428571e-05, "loss": 1.625, "step": 2259000 }, { "epoch": 11.11, "learning_rate": 3.386071428571429e-05, "loss": 1.6298, "step": 2259500 }, { "epoch": 11.11, "learning_rate": 3.385714285714286e-05, "loss": 1.6311, "step": 2260000 }, { "epoch": 11.11, "learning_rate": 3.385357142857143e-05, "loss": 1.6263, "step": 2260500 }, { "epoch": 11.11, "learning_rate": 3.385e-05, "loss": 1.6205, "step": 2261000 }, { "epoch": 11.12, "learning_rate": 3.384642857142857e-05, "loss": 1.62, "step": 2261500 }, { "epoch": 11.12, "learning_rate": 3.3842857142857145e-05, "loss": 1.6287, "step": 2262000 }, { "epoch": 11.12, "learning_rate": 3.383928571428572e-05, "loss": 1.6226, "step": 2262500 }, { "epoch": 11.12, "learning_rate": 3.3835714285714286e-05, "loss": 1.6234, "step": 2263000 }, { "epoch": 11.13, "learning_rate": 3.3832142857142854e-05, "loss": 1.6273, "step": 2263500 }, { "epoch": 11.13, "learning_rate": 3.3828571428571435e-05, "loss": 1.6227, "step": 2264000 }, { "epoch": 11.13, "learning_rate": 3.3825e-05, "loss": 1.6215, "step": 2264500 }, { "epoch": 11.14, "learning_rate": 3.382142857142857e-05, "loss": 1.6141, "step": 2265000 }, { "epoch": 11.14, "learning_rate": 3.3817857142857144e-05, "loss": 1.627, "step": 2265500 }, { "epoch": 11.14, "learning_rate": 3.381428571428572e-05, "loss": 1.6298, "step": 2266000 }, { "epoch": 11.14, "learning_rate": 3.381071428571429e-05, "loss": 1.6241, "step": 2266500 }, { "epoch": 11.15, "learning_rate": 3.380714285714286e-05, "loss": 1.6173, "step": 2267000 }, { "epoch": 11.15, "learning_rate": 3.380357142857143e-05, "loss": 1.6343, "step": 2267500 }, { "epoch": 11.15, "learning_rate": 3.38e-05, "loss": 1.6172, "step": 2268000 }, { "epoch": 11.15, "learning_rate": 3.3796428571428576e-05, "loss": 1.6295, "step": 2268500 }, { "epoch": 11.16, "learning_rate": 3.379285714285714e-05, "loss": 1.6152, "step": 2269000 }, { "epoch": 11.16, "learning_rate": 3.378928571428572e-05, "loss": 1.6228, "step": 2269500 }, { "epoch": 11.16, "learning_rate": 3.3785714285714285e-05, "loss": 1.6194, "step": 2270000 }, { "epoch": 11.16, "learning_rate": 3.378214285714286e-05, "loss": 1.6262, "step": 2270500 }, { "epoch": 11.17, "learning_rate": 3.3778571428571434e-05, "loss": 1.6227, "step": 2271000 }, { "epoch": 11.17, "learning_rate": 3.3775e-05, "loss": 1.6341, "step": 2271500 }, { "epoch": 11.17, "learning_rate": 3.377142857142857e-05, "loss": 1.6214, "step": 2272000 }, { "epoch": 11.18, "learning_rate": 3.376785714285714e-05, "loss": 1.626, "step": 2272500 }, { "epoch": 11.18, "learning_rate": 3.376428571428572e-05, "loss": 1.6253, "step": 2273000 }, { "epoch": 11.18, "learning_rate": 3.376071428571429e-05, "loss": 1.6207, "step": 2273500 }, { "epoch": 11.18, "learning_rate": 3.375714285714286e-05, "loss": 1.6254, "step": 2274000 }, { "epoch": 11.19, "learning_rate": 3.3753571428571426e-05, "loss": 1.6298, "step": 2274500 }, { "epoch": 11.19, "learning_rate": 3.375000000000001e-05, "loss": 1.6164, "step": 2275000 }, { "epoch": 11.19, "learning_rate": 3.3746428571428575e-05, "loss": 1.6282, "step": 2275500 }, { "epoch": 11.19, "learning_rate": 3.374285714285714e-05, "loss": 1.6258, "step": 2276000 }, { "epoch": 11.2, "learning_rate": 3.3739285714285716e-05, "loss": 1.6334, "step": 2276500 }, { "epoch": 11.2, "learning_rate": 3.3735714285714284e-05, "loss": 1.6177, "step": 2277000 }, { "epoch": 11.2, "learning_rate": 3.373214285714286e-05, "loss": 1.6207, "step": 2277500 }, { "epoch": 11.2, "learning_rate": 3.372857142857143e-05, "loss": 1.6162, "step": 2278000 }, { "epoch": 11.21, "learning_rate": 3.3725e-05, "loss": 1.6238, "step": 2278500 }, { "epoch": 11.21, "learning_rate": 3.372142857142857e-05, "loss": 1.6215, "step": 2279000 }, { "epoch": 11.21, "learning_rate": 3.371785714285715e-05, "loss": 1.6294, "step": 2279500 }, { "epoch": 11.22, "learning_rate": 3.3714285714285716e-05, "loss": 1.6206, "step": 2280000 }, { "epoch": 11.22, "learning_rate": 3.371071428571429e-05, "loss": 1.6282, "step": 2280500 }, { "epoch": 11.22, "learning_rate": 3.370714285714286e-05, "loss": 1.6096, "step": 2281000 }, { "epoch": 11.22, "learning_rate": 3.370357142857143e-05, "loss": 1.619, "step": 2281500 }, { "epoch": 11.23, "learning_rate": 3.3700000000000006e-05, "loss": 1.6082, "step": 2282000 }, { "epoch": 11.23, "learning_rate": 3.3696428571428573e-05, "loss": 1.618, "step": 2282500 }, { "epoch": 11.23, "learning_rate": 3.369285714285714e-05, "loss": 1.6239, "step": 2283000 }, { "epoch": 11.23, "learning_rate": 3.3689285714285715e-05, "loss": 1.6203, "step": 2283500 }, { "epoch": 11.24, "learning_rate": 3.368571428571429e-05, "loss": 1.6209, "step": 2284000 }, { "epoch": 11.24, "learning_rate": 3.368214285714286e-05, "loss": 1.6157, "step": 2284500 }, { "epoch": 11.24, "learning_rate": 3.367857142857143e-05, "loss": 1.6179, "step": 2285000 }, { "epoch": 11.24, "learning_rate": 3.3675e-05, "loss": 1.6255, "step": 2285500 }, { "epoch": 11.25, "learning_rate": 3.367142857142857e-05, "loss": 1.6152, "step": 2286000 }, { "epoch": 11.25, "learning_rate": 3.366785714285715e-05, "loss": 1.6147, "step": 2286500 }, { "epoch": 11.25, "learning_rate": 3.3664285714285715e-05, "loss": 1.6283, "step": 2287000 }, { "epoch": 11.26, "learning_rate": 3.366071428571429e-05, "loss": 1.6168, "step": 2287500 }, { "epoch": 11.26, "learning_rate": 3.3657142857142856e-05, "loss": 1.6188, "step": 2288000 }, { "epoch": 11.26, "learning_rate": 3.365357142857143e-05, "loss": 1.6144, "step": 2288500 }, { "epoch": 11.26, "learning_rate": 3.3650000000000005e-05, "loss": 1.6195, "step": 2289000 }, { "epoch": 11.27, "learning_rate": 3.364642857142857e-05, "loss": 1.6331, "step": 2289500 }, { "epoch": 11.27, "learning_rate": 3.364285714285714e-05, "loss": 1.6205, "step": 2290000 }, { "epoch": 11.27, "learning_rate": 3.363928571428572e-05, "loss": 1.6244, "step": 2290500 }, { "epoch": 11.27, "learning_rate": 3.363571428571429e-05, "loss": 1.6312, "step": 2291000 }, { "epoch": 11.28, "learning_rate": 3.3632142857142856e-05, "loss": 1.6197, "step": 2291500 }, { "epoch": 11.28, "learning_rate": 3.362857142857143e-05, "loss": 1.6313, "step": 2292000 }, { "epoch": 11.28, "learning_rate": 3.3625000000000004e-05, "loss": 1.6282, "step": 2292500 }, { "epoch": 11.28, "learning_rate": 3.362142857142857e-05, "loss": 1.6332, "step": 2293000 }, { "epoch": 11.29, "learning_rate": 3.3617857142857146e-05, "loss": 1.6265, "step": 2293500 }, { "epoch": 11.29, "learning_rate": 3.361428571428571e-05, "loss": 1.6226, "step": 2294000 }, { "epoch": 11.29, "learning_rate": 3.361071428571429e-05, "loss": 1.6193, "step": 2294500 }, { "epoch": 11.3, "learning_rate": 3.360714285714286e-05, "loss": 1.6204, "step": 2295000 }, { "epoch": 11.3, "learning_rate": 3.360357142857143e-05, "loss": 1.6213, "step": 2295500 }, { "epoch": 11.3, "learning_rate": 3.3600000000000004e-05, "loss": 1.6189, "step": 2296000 }, { "epoch": 11.3, "learning_rate": 3.359642857142857e-05, "loss": 1.6262, "step": 2296500 }, { "epoch": 11.31, "learning_rate": 3.3592857142857145e-05, "loss": 1.6166, "step": 2297000 }, { "epoch": 11.31, "learning_rate": 3.358928571428572e-05, "loss": 1.612, "step": 2297500 }, { "epoch": 11.31, "learning_rate": 3.358571428571429e-05, "loss": 1.608, "step": 2298000 }, { "epoch": 11.31, "learning_rate": 3.3582142857142854e-05, "loss": 1.6284, "step": 2298500 }, { "epoch": 11.32, "learning_rate": 3.357857142857143e-05, "loss": 1.6199, "step": 2299000 }, { "epoch": 11.32, "learning_rate": 3.3575e-05, "loss": 1.6281, "step": 2299500 }, { "epoch": 11.32, "learning_rate": 3.357142857142857e-05, "loss": 1.6145, "step": 2300000 }, { "epoch": 11.32, "learning_rate": 3.3567857142857145e-05, "loss": 1.6135, "step": 2300500 }, { "epoch": 11.33, "learning_rate": 3.356428571428571e-05, "loss": 1.6169, "step": 2301000 }, { "epoch": 11.33, "learning_rate": 3.3560714285714286e-05, "loss": 1.6255, "step": 2301500 }, { "epoch": 11.33, "learning_rate": 3.355714285714286e-05, "loss": 1.6228, "step": 2302000 }, { "epoch": 11.34, "learning_rate": 3.355357142857143e-05, "loss": 1.6194, "step": 2302500 }, { "epoch": 11.34, "learning_rate": 3.355e-05, "loss": 1.6179, "step": 2303000 }, { "epoch": 11.34, "learning_rate": 3.354642857142857e-05, "loss": 1.6256, "step": 2303500 }, { "epoch": 11.34, "learning_rate": 3.3542857142857144e-05, "loss": 1.6192, "step": 2304000 }, { "epoch": 11.35, "learning_rate": 3.353928571428572e-05, "loss": 1.6228, "step": 2304500 }, { "epoch": 11.35, "learning_rate": 3.3535714285714286e-05, "loss": 1.6153, "step": 2305000 }, { "epoch": 11.35, "learning_rate": 3.353214285714285e-05, "loss": 1.6149, "step": 2305500 }, { "epoch": 11.35, "learning_rate": 3.3528571428571434e-05, "loss": 1.6158, "step": 2306000 }, { "epoch": 11.36, "learning_rate": 3.3525e-05, "loss": 1.6155, "step": 2306500 }, { "epoch": 11.36, "learning_rate": 3.352142857142857e-05, "loss": 1.6277, "step": 2307000 }, { "epoch": 11.36, "learning_rate": 3.3517857142857143e-05, "loss": 1.6231, "step": 2307500 }, { "epoch": 11.36, "learning_rate": 3.351428571428572e-05, "loss": 1.6256, "step": 2308000 }, { "epoch": 11.37, "learning_rate": 3.351071428571429e-05, "loss": 1.6185, "step": 2308500 }, { "epoch": 11.37, "learning_rate": 3.350714285714286e-05, "loss": 1.6217, "step": 2309000 }, { "epoch": 11.37, "learning_rate": 3.350357142857143e-05, "loss": 1.6142, "step": 2309500 }, { "epoch": 11.38, "learning_rate": 3.35e-05, "loss": 1.6175, "step": 2310000 }, { "epoch": 11.38, "learning_rate": 3.3496428571428575e-05, "loss": 1.6204, "step": 2310500 }, { "epoch": 11.38, "learning_rate": 3.349285714285714e-05, "loss": 1.6186, "step": 2311000 }, { "epoch": 11.38, "learning_rate": 3.348928571428572e-05, "loss": 1.614, "step": 2311500 }, { "epoch": 11.39, "learning_rate": 3.3485714285714285e-05, "loss": 1.6191, "step": 2312000 }, { "epoch": 11.39, "learning_rate": 3.348214285714286e-05, "loss": 1.6094, "step": 2312500 }, { "epoch": 11.39, "learning_rate": 3.347857142857143e-05, "loss": 1.6241, "step": 2313000 }, { "epoch": 11.39, "learning_rate": 3.3475e-05, "loss": 1.6235, "step": 2313500 }, { "epoch": 11.4, "learning_rate": 3.3471428571428575e-05, "loss": 1.614, "step": 2314000 }, { "epoch": 11.4, "learning_rate": 3.346785714285714e-05, "loss": 1.622, "step": 2314500 }, { "epoch": 11.4, "learning_rate": 3.3464285714285716e-05, "loss": 1.6183, "step": 2315000 }, { "epoch": 11.4, "learning_rate": 3.346071428571429e-05, "loss": 1.6233, "step": 2315500 }, { "epoch": 11.41, "learning_rate": 3.345714285714286e-05, "loss": 1.614, "step": 2316000 }, { "epoch": 11.41, "learning_rate": 3.3453571428571426e-05, "loss": 1.6236, "step": 2316500 }, { "epoch": 11.41, "learning_rate": 3.345000000000001e-05, "loss": 1.6268, "step": 2317000 }, { "epoch": 11.42, "learning_rate": 3.3446428571428574e-05, "loss": 1.6195, "step": 2317500 }, { "epoch": 11.42, "learning_rate": 3.344285714285714e-05, "loss": 1.6078, "step": 2318000 }, { "epoch": 11.42, "learning_rate": 3.3439285714285716e-05, "loss": 1.6197, "step": 2318500 }, { "epoch": 11.42, "learning_rate": 3.343571428571429e-05, "loss": 1.6048, "step": 2319000 }, { "epoch": 11.43, "learning_rate": 3.343214285714286e-05, "loss": 1.6158, "step": 2319500 }, { "epoch": 11.43, "learning_rate": 3.342857142857143e-05, "loss": 1.6146, "step": 2320000 }, { "epoch": 11.43, "learning_rate": 3.3425e-05, "loss": 1.6119, "step": 2320500 }, { "epoch": 11.43, "learning_rate": 3.3421428571428574e-05, "loss": 1.6175, "step": 2321000 }, { "epoch": 11.44, "learning_rate": 3.341785714285715e-05, "loss": 1.6115, "step": 2321500 }, { "epoch": 11.44, "learning_rate": 3.3414285714285715e-05, "loss": 1.6153, "step": 2322000 }, { "epoch": 11.44, "learning_rate": 3.341071428571429e-05, "loss": 1.6138, "step": 2322500 }, { "epoch": 11.44, "learning_rate": 3.340714285714286e-05, "loss": 1.6302, "step": 2323000 }, { "epoch": 11.45, "learning_rate": 3.340357142857143e-05, "loss": 1.6216, "step": 2323500 }, { "epoch": 11.45, "learning_rate": 3.3400000000000005e-05, "loss": 1.6133, "step": 2324000 }, { "epoch": 11.45, "learning_rate": 3.339642857142857e-05, "loss": 1.6151, "step": 2324500 }, { "epoch": 11.45, "learning_rate": 3.339285714285714e-05, "loss": 1.6178, "step": 2325000 }, { "epoch": 11.46, "learning_rate": 3.3389285714285715e-05, "loss": 1.6133, "step": 2325500 }, { "epoch": 11.46, "learning_rate": 3.338571428571429e-05, "loss": 1.6187, "step": 2326000 }, { "epoch": 11.46, "learning_rate": 3.3382142857142856e-05, "loss": 1.6144, "step": 2326500 }, { "epoch": 11.47, "learning_rate": 3.337857142857143e-05, "loss": 1.6175, "step": 2327000 }, { "epoch": 11.47, "learning_rate": 3.3375e-05, "loss": 1.6224, "step": 2327500 }, { "epoch": 11.47, "learning_rate": 3.337142857142857e-05, "loss": 1.6133, "step": 2328000 }, { "epoch": 11.47, "learning_rate": 3.3367857142857147e-05, "loss": 1.6148, "step": 2328500 }, { "epoch": 11.48, "learning_rate": 3.3364285714285714e-05, "loss": 1.6104, "step": 2329000 }, { "epoch": 11.48, "learning_rate": 3.336071428571429e-05, "loss": 1.6223, "step": 2329500 }, { "epoch": 11.48, "learning_rate": 3.3357142857142856e-05, "loss": 1.6218, "step": 2330000 }, { "epoch": 11.48, "learning_rate": 3.335357142857143e-05, "loss": 1.6197, "step": 2330500 }, { "epoch": 11.49, "learning_rate": 3.3350000000000004e-05, "loss": 1.6193, "step": 2331000 }, { "epoch": 11.49, "learning_rate": 3.334642857142857e-05, "loss": 1.6192, "step": 2331500 }, { "epoch": 11.49, "learning_rate": 3.334285714285714e-05, "loss": 1.6156, "step": 2332000 }, { "epoch": 11.49, "learning_rate": 3.333928571428572e-05, "loss": 1.6015, "step": 2332500 }, { "epoch": 11.5, "learning_rate": 3.333571428571429e-05, "loss": 1.6123, "step": 2333000 }, { "epoch": 11.5, "learning_rate": 3.3332142857142855e-05, "loss": 1.6219, "step": 2333500 }, { "epoch": 11.5, "learning_rate": 3.332857142857143e-05, "loss": 1.6167, "step": 2334000 }, { "epoch": 11.51, "learning_rate": 3.3325000000000004e-05, "loss": 1.6164, "step": 2334500 }, { "epoch": 11.51, "learning_rate": 3.332142857142858e-05, "loss": 1.6089, "step": 2335000 }, { "epoch": 11.51, "learning_rate": 3.3317857142857145e-05, "loss": 1.6151, "step": 2335500 }, { "epoch": 11.51, "learning_rate": 3.331428571428571e-05, "loss": 1.6176, "step": 2336000 }, { "epoch": 11.52, "learning_rate": 3.331071428571429e-05, "loss": 1.6207, "step": 2336500 }, { "epoch": 11.52, "learning_rate": 3.330714285714286e-05, "loss": 1.6173, "step": 2337000 }, { "epoch": 11.52, "learning_rate": 3.330357142857143e-05, "loss": 1.6072, "step": 2337500 }, { "epoch": 11.52, "learning_rate": 3.33e-05, "loss": 1.6135, "step": 2338000 }, { "epoch": 11.53, "learning_rate": 3.329642857142857e-05, "loss": 1.6181, "step": 2338500 }, { "epoch": 11.53, "learning_rate": 3.3292857142857145e-05, "loss": 1.6062, "step": 2339000 }, { "epoch": 11.53, "learning_rate": 3.328928571428572e-05, "loss": 1.6106, "step": 2339500 }, { "epoch": 11.53, "learning_rate": 3.3285714285714286e-05, "loss": 1.6202, "step": 2340000 }, { "epoch": 11.54, "learning_rate": 3.3282142857142854e-05, "loss": 1.6136, "step": 2340500 }, { "epoch": 11.54, "learning_rate": 3.327857142857143e-05, "loss": 1.6185, "step": 2341000 }, { "epoch": 11.54, "learning_rate": 3.3275e-05, "loss": 1.6167, "step": 2341500 }, { "epoch": 11.55, "learning_rate": 3.327142857142858e-05, "loss": 1.6132, "step": 2342000 }, { "epoch": 11.55, "learning_rate": 3.3267857142857144e-05, "loss": 1.6178, "step": 2342500 }, { "epoch": 11.55, "learning_rate": 3.326428571428571e-05, "loss": 1.6325, "step": 2343000 }, { "epoch": 11.55, "learning_rate": 3.326071428571429e-05, "loss": 1.6126, "step": 2343500 }, { "epoch": 11.56, "learning_rate": 3.325714285714286e-05, "loss": 1.6233, "step": 2344000 }, { "epoch": 11.56, "learning_rate": 3.325357142857143e-05, "loss": 1.618, "step": 2344500 }, { "epoch": 11.56, "learning_rate": 3.325e-05, "loss": 1.6061, "step": 2345000 }, { "epoch": 11.56, "learning_rate": 3.3246428571428576e-05, "loss": 1.618, "step": 2345500 }, { "epoch": 11.57, "learning_rate": 3.3242857142857144e-05, "loss": 1.617, "step": 2346000 }, { "epoch": 11.57, "learning_rate": 3.323928571428572e-05, "loss": 1.6188, "step": 2346500 }, { "epoch": 11.57, "learning_rate": 3.3235714285714285e-05, "loss": 1.6213, "step": 2347000 }, { "epoch": 11.57, "learning_rate": 3.323214285714286e-05, "loss": 1.6314, "step": 2347500 }, { "epoch": 11.58, "learning_rate": 3.3228571428571434e-05, "loss": 1.6167, "step": 2348000 }, { "epoch": 11.58, "learning_rate": 3.3225e-05, "loss": 1.6196, "step": 2348500 }, { "epoch": 11.58, "learning_rate": 3.3221428571428575e-05, "loss": 1.6069, "step": 2349000 }, { "epoch": 11.59, "learning_rate": 3.321785714285714e-05, "loss": 1.6116, "step": 2349500 }, { "epoch": 11.59, "learning_rate": 3.321428571428572e-05, "loss": 1.6161, "step": 2350000 }, { "epoch": 11.59, "learning_rate": 3.321071428571429e-05, "loss": 1.6151, "step": 2350500 }, { "epoch": 11.59, "learning_rate": 3.320714285714286e-05, "loss": 1.6062, "step": 2351000 }, { "epoch": 11.6, "learning_rate": 3.3203571428571426e-05, "loss": 1.6144, "step": 2351500 }, { "epoch": 11.6, "learning_rate": 3.32e-05, "loss": 1.6108, "step": 2352000 }, { "epoch": 11.6, "learning_rate": 3.3196428571428575e-05, "loss": 1.6107, "step": 2352500 }, { "epoch": 11.6, "learning_rate": 3.319285714285714e-05, "loss": 1.6146, "step": 2353000 }, { "epoch": 11.61, "learning_rate": 3.3189285714285717e-05, "loss": 1.6213, "step": 2353500 }, { "epoch": 11.61, "learning_rate": 3.3185714285714284e-05, "loss": 1.6242, "step": 2354000 }, { "epoch": 11.61, "learning_rate": 3.318214285714286e-05, "loss": 1.6262, "step": 2354500 }, { "epoch": 11.61, "learning_rate": 3.317857142857143e-05, "loss": 1.6222, "step": 2355000 }, { "epoch": 11.62, "learning_rate": 3.3175e-05, "loss": 1.6217, "step": 2355500 }, { "epoch": 11.62, "learning_rate": 3.3171428571428574e-05, "loss": 1.6152, "step": 2356000 }, { "epoch": 11.62, "learning_rate": 3.316785714285714e-05, "loss": 1.6084, "step": 2356500 }, { "epoch": 11.63, "learning_rate": 3.3164285714285716e-05, "loss": 1.6249, "step": 2357000 }, { "epoch": 11.63, "learning_rate": 3.316071428571429e-05, "loss": 1.6132, "step": 2357500 }, { "epoch": 11.63, "learning_rate": 3.315714285714286e-05, "loss": 1.6166, "step": 2358000 }, { "epoch": 11.63, "learning_rate": 3.3153571428571425e-05, "loss": 1.618, "step": 2358500 }, { "epoch": 11.64, "learning_rate": 3.3150000000000006e-05, "loss": 1.6076, "step": 2359000 }, { "epoch": 11.64, "learning_rate": 3.3146428571428574e-05, "loss": 1.6194, "step": 2359500 }, { "epoch": 11.64, "learning_rate": 3.314285714285714e-05, "loss": 1.6107, "step": 2360000 }, { "epoch": 11.64, "learning_rate": 3.3139285714285715e-05, "loss": 1.611, "step": 2360500 }, { "epoch": 11.65, "learning_rate": 3.313571428571429e-05, "loss": 1.6252, "step": 2361000 }, { "epoch": 11.65, "learning_rate": 3.313214285714286e-05, "loss": 1.616, "step": 2361500 }, { "epoch": 11.65, "learning_rate": 3.312857142857143e-05, "loss": 1.6216, "step": 2362000 }, { "epoch": 11.65, "learning_rate": 3.3125e-05, "loss": 1.6094, "step": 2362500 }, { "epoch": 11.66, "learning_rate": 3.312142857142857e-05, "loss": 1.6172, "step": 2363000 }, { "epoch": 11.66, "learning_rate": 3.311785714285715e-05, "loss": 1.6195, "step": 2363500 }, { "epoch": 11.66, "learning_rate": 3.3114285714285715e-05, "loss": 1.6149, "step": 2364000 }, { "epoch": 11.67, "learning_rate": 3.311071428571429e-05, "loss": 1.6087, "step": 2364500 }, { "epoch": 11.67, "learning_rate": 3.3107142857142856e-05, "loss": 1.6166, "step": 2365000 }, { "epoch": 11.67, "learning_rate": 3.310357142857143e-05, "loss": 1.6088, "step": 2365500 }, { "epoch": 11.67, "learning_rate": 3.3100000000000005e-05, "loss": 1.6159, "step": 2366000 }, { "epoch": 11.68, "learning_rate": 3.309642857142857e-05, "loss": 1.6171, "step": 2366500 }, { "epoch": 11.68, "learning_rate": 3.309285714285714e-05, "loss": 1.6075, "step": 2367000 }, { "epoch": 11.68, "learning_rate": 3.3089285714285714e-05, "loss": 1.6099, "step": 2367500 }, { "epoch": 11.68, "learning_rate": 3.308571428571429e-05, "loss": 1.6163, "step": 2368000 }, { "epoch": 11.69, "learning_rate": 3.3082142857142856e-05, "loss": 1.6193, "step": 2368500 }, { "epoch": 11.69, "learning_rate": 3.307857142857143e-05, "loss": 1.6106, "step": 2369000 }, { "epoch": 11.69, "learning_rate": 3.3075e-05, "loss": 1.6198, "step": 2369500 }, { "epoch": 11.69, "learning_rate": 3.307142857142858e-05, "loss": 1.6055, "step": 2370000 }, { "epoch": 11.7, "learning_rate": 3.3067857142857146e-05, "loss": 1.6148, "step": 2370500 }, { "epoch": 11.7, "learning_rate": 3.3064285714285714e-05, "loss": 1.6046, "step": 2371000 }, { "epoch": 11.7, "learning_rate": 3.306071428571429e-05, "loss": 1.6137, "step": 2371500 }, { "epoch": 11.71, "learning_rate": 3.305714285714286e-05, "loss": 1.6139, "step": 2372000 }, { "epoch": 11.71, "learning_rate": 3.305357142857143e-05, "loss": 1.6118, "step": 2372500 }, { "epoch": 11.71, "learning_rate": 3.3050000000000004e-05, "loss": 1.6105, "step": 2373000 }, { "epoch": 11.71, "learning_rate": 3.304642857142857e-05, "loss": 1.614, "step": 2373500 }, { "epoch": 11.72, "learning_rate": 3.304285714285714e-05, "loss": 1.6158, "step": 2374000 }, { "epoch": 11.72, "learning_rate": 3.303928571428572e-05, "loss": 1.6244, "step": 2374500 }, { "epoch": 11.72, "learning_rate": 3.303571428571429e-05, "loss": 1.6077, "step": 2375000 }, { "epoch": 11.72, "learning_rate": 3.303214285714286e-05, "loss": 1.621, "step": 2375500 }, { "epoch": 11.73, "learning_rate": 3.302857142857143e-05, "loss": 1.6241, "step": 2376000 }, { "epoch": 11.73, "learning_rate": 3.3025e-05, "loss": 1.6201, "step": 2376500 }, { "epoch": 11.73, "learning_rate": 3.302142857142858e-05, "loss": 1.6119, "step": 2377000 }, { "epoch": 11.73, "learning_rate": 3.3017857142857145e-05, "loss": 1.6088, "step": 2377500 }, { "epoch": 11.74, "learning_rate": 3.301428571428571e-05, "loss": 1.6041, "step": 2378000 }, { "epoch": 11.74, "learning_rate": 3.3010714285714287e-05, "loss": 1.6163, "step": 2378500 }, { "epoch": 11.74, "learning_rate": 3.300714285714286e-05, "loss": 1.6169, "step": 2379000 }, { "epoch": 11.75, "learning_rate": 3.300357142857143e-05, "loss": 1.6248, "step": 2379500 }, { "epoch": 11.75, "learning_rate": 3.3e-05, "loss": 1.6208, "step": 2380000 }, { "epoch": 11.75, "learning_rate": 3.299642857142857e-05, "loss": 1.6185, "step": 2380500 }, { "epoch": 11.75, "learning_rate": 3.2992857142857144e-05, "loss": 1.609, "step": 2381000 }, { "epoch": 11.76, "learning_rate": 3.298928571428572e-05, "loss": 1.6023, "step": 2381500 }, { "epoch": 11.76, "learning_rate": 3.2985714285714286e-05, "loss": 1.6129, "step": 2382000 }, { "epoch": 11.76, "learning_rate": 3.298214285714286e-05, "loss": 1.6012, "step": 2382500 }, { "epoch": 11.76, "learning_rate": 3.297857142857143e-05, "loss": 1.6091, "step": 2383000 }, { "epoch": 11.77, "learning_rate": 3.2975e-05, "loss": 1.6123, "step": 2383500 }, { "epoch": 11.77, "learning_rate": 3.2971428571428576e-05, "loss": 1.6101, "step": 2384000 }, { "epoch": 11.77, "learning_rate": 3.2967857142857144e-05, "loss": 1.6071, "step": 2384500 }, { "epoch": 11.77, "learning_rate": 3.296428571428571e-05, "loss": 1.6043, "step": 2385000 }, { "epoch": 11.78, "learning_rate": 3.296071428571429e-05, "loss": 1.6154, "step": 2385500 }, { "epoch": 11.78, "learning_rate": 3.295714285714286e-05, "loss": 1.6209, "step": 2386000 }, { "epoch": 11.78, "learning_rate": 3.295357142857143e-05, "loss": 1.6, "step": 2386500 }, { "epoch": 11.79, "learning_rate": 3.295e-05, "loss": 1.601, "step": 2387000 }, { "epoch": 11.79, "learning_rate": 3.2946428571428576e-05, "loss": 1.6229, "step": 2387500 }, { "epoch": 11.79, "learning_rate": 3.294285714285714e-05, "loss": 1.6139, "step": 2388000 }, { "epoch": 11.79, "learning_rate": 3.293928571428572e-05, "loss": 1.6071, "step": 2388500 }, { "epoch": 11.8, "learning_rate": 3.2935714285714285e-05, "loss": 1.6037, "step": 2389000 }, { "epoch": 11.8, "learning_rate": 3.293214285714286e-05, "loss": 1.6102, "step": 2389500 }, { "epoch": 11.8, "learning_rate": 3.292857142857143e-05, "loss": 1.6195, "step": 2390000 }, { "epoch": 11.8, "learning_rate": 3.2925e-05, "loss": 1.6112, "step": 2390500 }, { "epoch": 11.81, "learning_rate": 3.2921428571428575e-05, "loss": 1.6032, "step": 2391000 }, { "epoch": 11.81, "learning_rate": 3.291785714285714e-05, "loss": 1.6141, "step": 2391500 }, { "epoch": 11.81, "learning_rate": 3.291428571428572e-05, "loss": 1.6102, "step": 2392000 }, { "epoch": 11.81, "learning_rate": 3.291071428571429e-05, "loss": 1.6001, "step": 2392500 }, { "epoch": 11.82, "learning_rate": 3.290714285714286e-05, "loss": 1.6148, "step": 2393000 }, { "epoch": 11.82, "learning_rate": 3.2903571428571426e-05, "loss": 1.606, "step": 2393500 }, { "epoch": 11.82, "learning_rate": 3.29e-05, "loss": 1.6176, "step": 2394000 }, { "epoch": 11.83, "learning_rate": 3.2896428571428574e-05, "loss": 1.6065, "step": 2394500 }, { "epoch": 11.83, "learning_rate": 3.289285714285714e-05, "loss": 1.6065, "step": 2395000 }, { "epoch": 11.83, "learning_rate": 3.2889285714285716e-05, "loss": 1.6108, "step": 2395500 }, { "epoch": 11.83, "learning_rate": 3.2885714285714284e-05, "loss": 1.6113, "step": 2396000 }, { "epoch": 11.84, "learning_rate": 3.2882142857142865e-05, "loss": 1.6111, "step": 2396500 }, { "epoch": 11.84, "learning_rate": 3.287857142857143e-05, "loss": 1.6005, "step": 2397000 }, { "epoch": 11.84, "learning_rate": 3.2875e-05, "loss": 1.6076, "step": 2397500 }, { "epoch": 11.84, "learning_rate": 3.2871428571428574e-05, "loss": 1.6058, "step": 2398000 }, { "epoch": 11.85, "learning_rate": 3.286785714285715e-05, "loss": 1.6052, "step": 2398500 }, { "epoch": 11.85, "learning_rate": 3.2864285714285715e-05, "loss": 1.6071, "step": 2399000 }, { "epoch": 11.85, "learning_rate": 3.286071428571429e-05, "loss": 1.6108, "step": 2399500 }, { "epoch": 11.85, "learning_rate": 3.285714285714286e-05, "loss": 1.6117, "step": 2400000 }, { "epoch": 11.86, "learning_rate": 3.2853571428571425e-05, "loss": 1.6172, "step": 2400500 }, { "epoch": 11.86, "learning_rate": 3.2850000000000006e-05, "loss": 1.6101, "step": 2401000 }, { "epoch": 11.86, "learning_rate": 3.284642857142857e-05, "loss": 1.6133, "step": 2401500 }, { "epoch": 11.86, "learning_rate": 3.284285714285714e-05, "loss": 1.6051, "step": 2402000 }, { "epoch": 11.87, "learning_rate": 3.2839285714285715e-05, "loss": 1.6057, "step": 2402500 }, { "epoch": 11.87, "learning_rate": 3.283571428571429e-05, "loss": 1.604, "step": 2403000 }, { "epoch": 11.87, "learning_rate": 3.283214285714286e-05, "loss": 1.6022, "step": 2403500 }, { "epoch": 11.88, "learning_rate": 3.282857142857143e-05, "loss": 1.6101, "step": 2404000 }, { "epoch": 11.88, "learning_rate": 3.2825e-05, "loss": 1.6074, "step": 2404500 }, { "epoch": 11.88, "learning_rate": 3.282142857142857e-05, "loss": 1.604, "step": 2405000 }, { "epoch": 11.88, "learning_rate": 3.281785714285715e-05, "loss": 1.6112, "step": 2405500 }, { "epoch": 11.89, "learning_rate": 3.2814285714285714e-05, "loss": 1.6086, "step": 2406000 }, { "epoch": 11.89, "learning_rate": 3.281071428571429e-05, "loss": 1.6027, "step": 2406500 }, { "epoch": 11.89, "learning_rate": 3.2807142857142856e-05, "loss": 1.6079, "step": 2407000 }, { "epoch": 11.89, "learning_rate": 3.280357142857143e-05, "loss": 1.6169, "step": 2407500 }, { "epoch": 11.9, "learning_rate": 3.2800000000000004e-05, "loss": 1.5967, "step": 2408000 }, { "epoch": 11.9, "learning_rate": 3.279642857142857e-05, "loss": 1.6136, "step": 2408500 }, { "epoch": 11.9, "learning_rate": 3.279285714285714e-05, "loss": 1.6062, "step": 2409000 }, { "epoch": 11.9, "learning_rate": 3.2789285714285714e-05, "loss": 1.612, "step": 2409500 }, { "epoch": 11.91, "learning_rate": 3.278571428571429e-05, "loss": 1.6102, "step": 2410000 }, { "epoch": 11.91, "learning_rate": 3.278214285714286e-05, "loss": 1.6071, "step": 2410500 }, { "epoch": 11.91, "learning_rate": 3.277857142857143e-05, "loss": 1.6072, "step": 2411000 }, { "epoch": 11.92, "learning_rate": 3.2775e-05, "loss": 1.6101, "step": 2411500 }, { "epoch": 11.92, "learning_rate": 3.277142857142858e-05, "loss": 1.6125, "step": 2412000 }, { "epoch": 11.92, "learning_rate": 3.2767857142857146e-05, "loss": 1.6071, "step": 2412500 }, { "epoch": 11.92, "learning_rate": 3.276428571428571e-05, "loss": 1.5977, "step": 2413000 }, { "epoch": 11.93, "learning_rate": 3.276071428571429e-05, "loss": 1.6013, "step": 2413500 }, { "epoch": 11.93, "learning_rate": 3.275714285714286e-05, "loss": 1.6171, "step": 2414000 }, { "epoch": 11.93, "learning_rate": 3.275357142857143e-05, "loss": 1.6068, "step": 2414500 }, { "epoch": 11.93, "learning_rate": 3.275e-05, "loss": 1.6045, "step": 2415000 }, { "epoch": 11.94, "learning_rate": 3.274642857142857e-05, "loss": 1.5998, "step": 2415500 }, { "epoch": 11.94, "learning_rate": 3.2742857142857145e-05, "loss": 1.6062, "step": 2416000 }, { "epoch": 11.94, "learning_rate": 3.273928571428572e-05, "loss": 1.6067, "step": 2416500 }, { "epoch": 11.94, "learning_rate": 3.273571428571429e-05, "loss": 1.6119, "step": 2417000 }, { "epoch": 11.95, "learning_rate": 3.273214285714286e-05, "loss": 1.6234, "step": 2417500 }, { "epoch": 11.95, "learning_rate": 3.272857142857143e-05, "loss": 1.6137, "step": 2418000 }, { "epoch": 11.95, "learning_rate": 3.2725e-05, "loss": 1.6037, "step": 2418500 }, { "epoch": 11.96, "learning_rate": 3.272142857142858e-05, "loss": 1.6123, "step": 2419000 }, { "epoch": 11.96, "learning_rate": 3.2717857142857144e-05, "loss": 1.6075, "step": 2419500 }, { "epoch": 11.96, "learning_rate": 3.271428571428571e-05, "loss": 1.6146, "step": 2420000 }, { "epoch": 11.96, "learning_rate": 3.2710714285714286e-05, "loss": 1.6166, "step": 2420500 }, { "epoch": 11.97, "learning_rate": 3.270714285714286e-05, "loss": 1.6116, "step": 2421000 }, { "epoch": 11.97, "learning_rate": 3.270357142857143e-05, "loss": 1.6111, "step": 2421500 }, { "epoch": 11.97, "learning_rate": 3.27e-05, "loss": 1.6072, "step": 2422000 }, { "epoch": 11.97, "learning_rate": 3.269642857142857e-05, "loss": 1.6102, "step": 2422500 }, { "epoch": 11.98, "learning_rate": 3.2692857142857144e-05, "loss": 1.6062, "step": 2423000 }, { "epoch": 11.98, "learning_rate": 3.268928571428572e-05, "loss": 1.6078, "step": 2423500 }, { "epoch": 11.98, "learning_rate": 3.2685714285714285e-05, "loss": 1.6039, "step": 2424000 }, { "epoch": 11.98, "learning_rate": 3.268214285714286e-05, "loss": 1.6115, "step": 2424500 }, { "epoch": 11.99, "learning_rate": 3.2678571428571434e-05, "loss": 1.6065, "step": 2425000 }, { "epoch": 11.99, "learning_rate": 3.2675e-05, "loss": 1.611, "step": 2425500 }, { "epoch": 11.99, "learning_rate": 3.2671428571428576e-05, "loss": 1.6148, "step": 2426000 }, { "epoch": 12.0, "learning_rate": 3.266785714285714e-05, "loss": 1.6071, "step": 2426500 }, { "epoch": 12.0, "learning_rate": 3.266428571428571e-05, "loss": 1.599, "step": 2427000 }, { "epoch": 12.0, "learning_rate": 3.266071428571429e-05, "loss": 1.6071, "step": 2427500 }, { "epoch": 12.0, "learning_rate": 3.265714285714286e-05, "loss": 1.5964, "step": 2428000 }, { "epoch": 12.01, "learning_rate": 3.2653571428571427e-05, "loss": 1.6065, "step": 2428500 }, { "epoch": 12.01, "learning_rate": 3.265e-05, "loss": 1.595, "step": 2429000 }, { "epoch": 12.01, "learning_rate": 3.2646428571428575e-05, "loss": 1.6076, "step": 2429500 }, { "epoch": 12.01, "learning_rate": 3.264285714285714e-05, "loss": 1.5934, "step": 2430000 }, { "epoch": 12.02, "learning_rate": 3.263928571428572e-05, "loss": 1.6016, "step": 2430500 }, { "epoch": 12.02, "learning_rate": 3.2635714285714284e-05, "loss": 1.6107, "step": 2431000 }, { "epoch": 12.02, "learning_rate": 3.263214285714286e-05, "loss": 1.6062, "step": 2431500 }, { "epoch": 12.02, "learning_rate": 3.262857142857143e-05, "loss": 1.6051, "step": 2432000 }, { "epoch": 12.03, "learning_rate": 3.2625e-05, "loss": 1.5965, "step": 2432500 }, { "epoch": 12.03, "learning_rate": 3.2621428571428574e-05, "loss": 1.5915, "step": 2433000 }, { "epoch": 12.03, "learning_rate": 3.261785714285714e-05, "loss": 1.6019, "step": 2433500 }, { "epoch": 12.04, "learning_rate": 3.2614285714285716e-05, "loss": 1.5855, "step": 2434000 }, { "epoch": 12.04, "learning_rate": 3.261071428571429e-05, "loss": 1.5979, "step": 2434500 }, { "epoch": 12.04, "learning_rate": 3.260714285714286e-05, "loss": 1.5951, "step": 2435000 }, { "epoch": 12.04, "learning_rate": 3.2603571428571425e-05, "loss": 1.5994, "step": 2435500 }, { "epoch": 12.05, "learning_rate": 3.26e-05, "loss": 1.6058, "step": 2436000 }, { "epoch": 12.05, "learning_rate": 3.2596428571428574e-05, "loss": 1.6029, "step": 2436500 }, { "epoch": 12.05, "learning_rate": 3.259285714285714e-05, "loss": 1.6117, "step": 2437000 }, { "epoch": 12.05, "learning_rate": 3.2589285714285716e-05, "loss": 1.5873, "step": 2437500 }, { "epoch": 12.06, "learning_rate": 3.258571428571428e-05, "loss": 1.6037, "step": 2438000 }, { "epoch": 12.06, "learning_rate": 3.2582142857142864e-05, "loss": 1.6074, "step": 2438500 }, { "epoch": 12.06, "learning_rate": 3.257857142857143e-05, "loss": 1.5924, "step": 2439000 }, { "epoch": 12.06, "learning_rate": 3.2575e-05, "loss": 1.6005, "step": 2439500 }, { "epoch": 12.07, "learning_rate": 3.257142857142857e-05, "loss": 1.6042, "step": 2440000 }, { "epoch": 12.07, "learning_rate": 3.256785714285715e-05, "loss": 1.6082, "step": 2440500 }, { "epoch": 12.07, "learning_rate": 3.2564285714285715e-05, "loss": 1.5986, "step": 2441000 }, { "epoch": 12.08, "learning_rate": 3.256071428571429e-05, "loss": 1.606, "step": 2441500 }, { "epoch": 12.08, "learning_rate": 3.255714285714286e-05, "loss": 1.6033, "step": 2442000 }, { "epoch": 12.08, "learning_rate": 3.255357142857143e-05, "loss": 1.5982, "step": 2442500 }, { "epoch": 12.08, "learning_rate": 3.2550000000000005e-05, "loss": 1.5915, "step": 2443000 }, { "epoch": 12.09, "learning_rate": 3.254642857142857e-05, "loss": 1.5917, "step": 2443500 }, { "epoch": 12.09, "learning_rate": 3.254285714285715e-05, "loss": 1.6092, "step": 2444000 }, { "epoch": 12.09, "learning_rate": 3.2539285714285714e-05, "loss": 1.6021, "step": 2444500 }, { "epoch": 12.09, "learning_rate": 3.253571428571429e-05, "loss": 1.5974, "step": 2445000 }, { "epoch": 12.1, "learning_rate": 3.253214285714286e-05, "loss": 1.6037, "step": 2445500 }, { "epoch": 12.1, "learning_rate": 3.252857142857143e-05, "loss": 1.5915, "step": 2446000 }, { "epoch": 12.1, "learning_rate": 3.2525e-05, "loss": 1.6111, "step": 2446500 }, { "epoch": 12.1, "learning_rate": 3.252142857142857e-05, "loss": 1.5975, "step": 2447000 }, { "epoch": 12.11, "learning_rate": 3.2517857142857146e-05, "loss": 1.5931, "step": 2447500 }, { "epoch": 12.11, "learning_rate": 3.2514285714285714e-05, "loss": 1.5918, "step": 2448000 }, { "epoch": 12.11, "learning_rate": 3.251071428571429e-05, "loss": 1.594, "step": 2448500 }, { "epoch": 12.12, "learning_rate": 3.2507142857142855e-05, "loss": 1.6027, "step": 2449000 }, { "epoch": 12.12, "learning_rate": 3.250357142857143e-05, "loss": 1.6028, "step": 2449500 }, { "epoch": 12.12, "learning_rate": 3.2500000000000004e-05, "loss": 1.6089, "step": 2450000 }, { "epoch": 12.12, "learning_rate": 3.249642857142857e-05, "loss": 1.6046, "step": 2450500 }, { "epoch": 12.13, "learning_rate": 3.2492857142857146e-05, "loss": 1.592, "step": 2451000 }, { "epoch": 12.13, "learning_rate": 3.248928571428572e-05, "loss": 1.5991, "step": 2451500 }, { "epoch": 12.13, "learning_rate": 3.248571428571429e-05, "loss": 1.6058, "step": 2452000 }, { "epoch": 12.13, "learning_rate": 3.248214285714286e-05, "loss": 1.6018, "step": 2452500 }, { "epoch": 12.14, "learning_rate": 3.247857142857143e-05, "loss": 1.5987, "step": 2453000 }, { "epoch": 12.14, "learning_rate": 3.2474999999999997e-05, "loss": 1.5933, "step": 2453500 }, { "epoch": 12.14, "learning_rate": 3.247142857142858e-05, "loss": 1.6045, "step": 2454000 }, { "epoch": 12.14, "learning_rate": 3.2467857142857145e-05, "loss": 1.6086, "step": 2454500 }, { "epoch": 12.15, "learning_rate": 3.246428571428571e-05, "loss": 1.6053, "step": 2455000 }, { "epoch": 12.15, "learning_rate": 3.246071428571429e-05, "loss": 1.5976, "step": 2455500 }, { "epoch": 12.15, "learning_rate": 3.245714285714286e-05, "loss": 1.5941, "step": 2456000 }, { "epoch": 12.16, "learning_rate": 3.245357142857143e-05, "loss": 1.6089, "step": 2456500 }, { "epoch": 12.16, "learning_rate": 3.245e-05, "loss": 1.603, "step": 2457000 }, { "epoch": 12.16, "learning_rate": 3.244642857142857e-05, "loss": 1.6038, "step": 2457500 }, { "epoch": 12.16, "learning_rate": 3.2442857142857144e-05, "loss": 1.595, "step": 2458000 }, { "epoch": 12.17, "learning_rate": 3.243928571428572e-05, "loss": 1.6121, "step": 2458500 }, { "epoch": 12.17, "learning_rate": 3.2435714285714286e-05, "loss": 1.5976, "step": 2459000 }, { "epoch": 12.17, "learning_rate": 3.243214285714286e-05, "loss": 1.6037, "step": 2459500 }, { "epoch": 12.17, "learning_rate": 3.242857142857143e-05, "loss": 1.5973, "step": 2460000 }, { "epoch": 12.18, "learning_rate": 3.2425e-05, "loss": 1.5998, "step": 2460500 }, { "epoch": 12.18, "learning_rate": 3.2421428571428576e-05, "loss": 1.5925, "step": 2461000 }, { "epoch": 12.18, "learning_rate": 3.2417857142857144e-05, "loss": 1.5916, "step": 2461500 }, { "epoch": 12.18, "learning_rate": 3.241428571428571e-05, "loss": 1.5973, "step": 2462000 }, { "epoch": 12.19, "learning_rate": 3.2410714285714286e-05, "loss": 1.6023, "step": 2462500 }, { "epoch": 12.19, "learning_rate": 3.240714285714286e-05, "loss": 1.6006, "step": 2463000 }, { "epoch": 12.19, "learning_rate": 3.240357142857143e-05, "loss": 1.6116, "step": 2463500 }, { "epoch": 12.2, "learning_rate": 3.24e-05, "loss": 1.5983, "step": 2464000 }, { "epoch": 12.2, "learning_rate": 3.239642857142857e-05, "loss": 1.6024, "step": 2464500 }, { "epoch": 12.2, "learning_rate": 3.239285714285715e-05, "loss": 1.6002, "step": 2465000 }, { "epoch": 12.2, "learning_rate": 3.238928571428572e-05, "loss": 1.6064, "step": 2465500 }, { "epoch": 12.21, "learning_rate": 3.2385714285714285e-05, "loss": 1.5966, "step": 2466000 }, { "epoch": 12.21, "learning_rate": 3.238214285714286e-05, "loss": 1.6041, "step": 2466500 }, { "epoch": 12.21, "learning_rate": 3.2378571428571433e-05, "loss": 1.6042, "step": 2467000 }, { "epoch": 12.21, "learning_rate": 3.2375e-05, "loss": 1.601, "step": 2467500 }, { "epoch": 12.22, "learning_rate": 3.2371428571428575e-05, "loss": 1.5979, "step": 2468000 }, { "epoch": 12.22, "learning_rate": 3.236785714285714e-05, "loss": 1.5928, "step": 2468500 }, { "epoch": 12.22, "learning_rate": 3.236428571428572e-05, "loss": 1.6019, "step": 2469000 }, { "epoch": 12.22, "learning_rate": 3.236071428571429e-05, "loss": 1.5918, "step": 2469500 }, { "epoch": 12.23, "learning_rate": 3.235714285714286e-05, "loss": 1.6062, "step": 2470000 }, { "epoch": 12.23, "learning_rate": 3.2353571428571426e-05, "loss": 1.5928, "step": 2470500 }, { "epoch": 12.23, "learning_rate": 3.235e-05, "loss": 1.588, "step": 2471000 }, { "epoch": 12.23, "learning_rate": 3.2346428571428575e-05, "loss": 1.5899, "step": 2471500 }, { "epoch": 12.24, "learning_rate": 3.234285714285715e-05, "loss": 1.6014, "step": 2472000 }, { "epoch": 12.24, "learning_rate": 3.2339285714285716e-05, "loss": 1.5974, "step": 2472500 }, { "epoch": 12.24, "learning_rate": 3.2335714285714284e-05, "loss": 1.6022, "step": 2473000 }, { "epoch": 12.25, "learning_rate": 3.233214285714286e-05, "loss": 1.6024, "step": 2473500 }, { "epoch": 12.25, "learning_rate": 3.232857142857143e-05, "loss": 1.5946, "step": 2474000 }, { "epoch": 12.25, "learning_rate": 3.2325e-05, "loss": 1.5958, "step": 2474500 }, { "epoch": 12.25, "learning_rate": 3.2321428571428574e-05, "loss": 1.5986, "step": 2475000 }, { "epoch": 12.26, "learning_rate": 3.231785714285714e-05, "loss": 1.602, "step": 2475500 }, { "epoch": 12.26, "learning_rate": 3.2314285714285716e-05, "loss": 1.5814, "step": 2476000 }, { "epoch": 12.26, "learning_rate": 3.231071428571429e-05, "loss": 1.5855, "step": 2476500 }, { "epoch": 12.26, "learning_rate": 3.230714285714286e-05, "loss": 1.5952, "step": 2477000 }, { "epoch": 12.27, "learning_rate": 3.2303571428571425e-05, "loss": 1.5941, "step": 2477500 }, { "epoch": 12.27, "learning_rate": 3.2300000000000006e-05, "loss": 1.5982, "step": 2478000 }, { "epoch": 12.27, "learning_rate": 3.229642857142857e-05, "loss": 1.5987, "step": 2478500 }, { "epoch": 12.27, "learning_rate": 3.229285714285715e-05, "loss": 1.6044, "step": 2479000 }, { "epoch": 12.28, "learning_rate": 3.2289285714285715e-05, "loss": 1.5902, "step": 2479500 }, { "epoch": 12.28, "learning_rate": 3.228571428571428e-05, "loss": 1.6004, "step": 2480000 }, { "epoch": 12.28, "learning_rate": 3.2282142857142864e-05, "loss": 1.6062, "step": 2480500 }, { "epoch": 12.29, "learning_rate": 3.227857142857143e-05, "loss": 1.6081, "step": 2481000 }, { "epoch": 12.29, "learning_rate": 3.2275e-05, "loss": 1.5977, "step": 2481500 }, { "epoch": 12.29, "learning_rate": 3.227142857142857e-05, "loss": 1.6029, "step": 2482000 }, { "epoch": 12.29, "learning_rate": 3.226785714285715e-05, "loss": 1.5965, "step": 2482500 }, { "epoch": 12.3, "learning_rate": 3.2264285714285714e-05, "loss": 1.5993, "step": 2483000 }, { "epoch": 12.3, "learning_rate": 3.226071428571429e-05, "loss": 1.5945, "step": 2483500 }, { "epoch": 12.3, "learning_rate": 3.2257142857142856e-05, "loss": 1.5896, "step": 2484000 }, { "epoch": 12.3, "learning_rate": 3.225357142857143e-05, "loss": 1.5969, "step": 2484500 }, { "epoch": 12.31, "learning_rate": 3.2250000000000005e-05, "loss": 1.5987, "step": 2485000 }, { "epoch": 12.31, "learning_rate": 3.224642857142857e-05, "loss": 1.5754, "step": 2485500 }, { "epoch": 12.31, "learning_rate": 3.2242857142857146e-05, "loss": 1.5992, "step": 2486000 }, { "epoch": 12.31, "learning_rate": 3.2239285714285714e-05, "loss": 1.6008, "step": 2486500 }, { "epoch": 12.32, "learning_rate": 3.223571428571429e-05, "loss": 1.5986, "step": 2487000 }, { "epoch": 12.32, "learning_rate": 3.223214285714286e-05, "loss": 1.5961, "step": 2487500 }, { "epoch": 12.32, "learning_rate": 3.222857142857143e-05, "loss": 1.6081, "step": 2488000 }, { "epoch": 12.33, "learning_rate": 3.2225e-05, "loss": 1.5988, "step": 2488500 }, { "epoch": 12.33, "learning_rate": 3.222142857142857e-05, "loss": 1.5963, "step": 2489000 }, { "epoch": 12.33, "learning_rate": 3.2217857142857146e-05, "loss": 1.5896, "step": 2489500 }, { "epoch": 12.33, "learning_rate": 3.221428571428571e-05, "loss": 1.5969, "step": 2490000 }, { "epoch": 12.34, "learning_rate": 3.221071428571429e-05, "loss": 1.5966, "step": 2490500 }, { "epoch": 12.34, "learning_rate": 3.2207142857142855e-05, "loss": 1.6085, "step": 2491000 }, { "epoch": 12.34, "learning_rate": 3.220357142857143e-05, "loss": 1.6054, "step": 2491500 }, { "epoch": 12.34, "learning_rate": 3.2200000000000003e-05, "loss": 1.6022, "step": 2492000 }, { "epoch": 12.35, "learning_rate": 3.219642857142857e-05, "loss": 1.5984, "step": 2492500 }, { "epoch": 12.35, "learning_rate": 3.2192857142857145e-05, "loss": 1.5995, "step": 2493000 }, { "epoch": 12.35, "learning_rate": 3.218928571428572e-05, "loss": 1.5935, "step": 2493500 }, { "epoch": 12.35, "learning_rate": 3.218571428571429e-05, "loss": 1.5934, "step": 2494000 }, { "epoch": 12.36, "learning_rate": 3.218214285714286e-05, "loss": 1.5925, "step": 2494500 }, { "epoch": 12.36, "learning_rate": 3.217857142857143e-05, "loss": 1.5951, "step": 2495000 }, { "epoch": 12.36, "learning_rate": 3.2175e-05, "loss": 1.592, "step": 2495500 }, { "epoch": 12.37, "learning_rate": 3.217142857142858e-05, "loss": 1.5922, "step": 2496000 }, { "epoch": 12.37, "learning_rate": 3.2167857142857145e-05, "loss": 1.5927, "step": 2496500 }, { "epoch": 12.37, "learning_rate": 3.216428571428571e-05, "loss": 1.5973, "step": 2497000 }, { "epoch": 12.37, "learning_rate": 3.2160714285714286e-05, "loss": 1.6056, "step": 2497500 }, { "epoch": 12.38, "learning_rate": 3.215714285714286e-05, "loss": 1.5936, "step": 2498000 }, { "epoch": 12.38, "learning_rate": 3.215357142857143e-05, "loss": 1.5962, "step": 2498500 }, { "epoch": 12.38, "learning_rate": 3.215e-05, "loss": 1.5987, "step": 2499000 }, { "epoch": 12.38, "learning_rate": 3.214642857142857e-05, "loss": 1.5967, "step": 2499500 }, { "epoch": 12.39, "learning_rate": 3.2142857142857144e-05, "loss": 1.6007, "step": 2500000 }, { "epoch": 12.39, "learning_rate": 3.213928571428572e-05, "loss": 1.5931, "step": 2500500 }, { "epoch": 12.39, "learning_rate": 3.2135714285714286e-05, "loss": 1.6077, "step": 2501000 }, { "epoch": 12.39, "learning_rate": 3.213214285714286e-05, "loss": 1.5942, "step": 2501500 }, { "epoch": 12.4, "learning_rate": 3.212857142857143e-05, "loss": 1.5933, "step": 2502000 }, { "epoch": 12.4, "learning_rate": 3.2125e-05, "loss": 1.5955, "step": 2502500 }, { "epoch": 12.4, "learning_rate": 3.2121428571428576e-05, "loss": 1.6037, "step": 2503000 }, { "epoch": 12.41, "learning_rate": 3.211785714285714e-05, "loss": 1.5965, "step": 2503500 }, { "epoch": 12.41, "learning_rate": 3.211428571428571e-05, "loss": 1.5925, "step": 2504000 }, { "epoch": 12.41, "learning_rate": 3.211071428571429e-05, "loss": 1.5943, "step": 2504500 }, { "epoch": 12.41, "learning_rate": 3.210714285714286e-05, "loss": 1.592, "step": 2505000 }, { "epoch": 12.42, "learning_rate": 3.2103571428571434e-05, "loss": 1.585, "step": 2505500 }, { "epoch": 12.42, "learning_rate": 3.21e-05, "loss": 1.5927, "step": 2506000 }, { "epoch": 12.42, "learning_rate": 3.209642857142857e-05, "loss": 1.5881, "step": 2506500 }, { "epoch": 12.42, "learning_rate": 3.209285714285715e-05, "loss": 1.5874, "step": 2507000 }, { "epoch": 12.43, "learning_rate": 3.208928571428572e-05, "loss": 1.5963, "step": 2507500 }, { "epoch": 12.43, "learning_rate": 3.2085714285714284e-05, "loss": 1.5925, "step": 2508000 }, { "epoch": 12.43, "learning_rate": 3.208214285714286e-05, "loss": 1.5968, "step": 2508500 }, { "epoch": 12.43, "learning_rate": 3.207857142857143e-05, "loss": 1.5827, "step": 2509000 }, { "epoch": 12.44, "learning_rate": 3.2075e-05, "loss": 1.6056, "step": 2509500 }, { "epoch": 12.44, "learning_rate": 3.2071428571428575e-05, "loss": 1.5961, "step": 2510000 }, { "epoch": 12.44, "learning_rate": 3.206785714285714e-05, "loss": 1.5899, "step": 2510500 }, { "epoch": 12.45, "learning_rate": 3.2064285714285716e-05, "loss": 1.5899, "step": 2511000 }, { "epoch": 12.45, "learning_rate": 3.206071428571429e-05, "loss": 1.5892, "step": 2511500 }, { "epoch": 12.45, "learning_rate": 3.205714285714286e-05, "loss": 1.6096, "step": 2512000 }, { "epoch": 12.45, "learning_rate": 3.205357142857143e-05, "loss": 1.6003, "step": 2512500 }, { "epoch": 12.46, "learning_rate": 3.205e-05, "loss": 1.5996, "step": 2513000 }, { "epoch": 12.46, "learning_rate": 3.2046428571428574e-05, "loss": 1.5977, "step": 2513500 }, { "epoch": 12.46, "learning_rate": 3.204285714285715e-05, "loss": 1.589, "step": 2514000 }, { "epoch": 12.46, "learning_rate": 3.2039285714285716e-05, "loss": 1.5926, "step": 2514500 }, { "epoch": 12.47, "learning_rate": 3.203571428571428e-05, "loss": 1.6081, "step": 2515000 }, { "epoch": 12.47, "learning_rate": 3.203214285714286e-05, "loss": 1.5999, "step": 2515500 }, { "epoch": 12.47, "learning_rate": 3.202857142857143e-05, "loss": 1.5962, "step": 2516000 }, { "epoch": 12.47, "learning_rate": 3.2025e-05, "loss": 1.5913, "step": 2516500 }, { "epoch": 12.48, "learning_rate": 3.2021428571428573e-05, "loss": 1.5835, "step": 2517000 }, { "epoch": 12.48, "learning_rate": 3.201785714285714e-05, "loss": 1.5933, "step": 2517500 }, { "epoch": 12.48, "learning_rate": 3.2014285714285715e-05, "loss": 1.5916, "step": 2518000 }, { "epoch": 12.49, "learning_rate": 3.201071428571429e-05, "loss": 1.5965, "step": 2518500 }, { "epoch": 12.49, "learning_rate": 3.200714285714286e-05, "loss": 1.6067, "step": 2519000 }, { "epoch": 12.49, "learning_rate": 3.200357142857143e-05, "loss": 1.5969, "step": 2519500 }, { "epoch": 12.49, "learning_rate": 3.2000000000000005e-05, "loss": 1.5937, "step": 2520000 }, { "epoch": 12.5, "learning_rate": 3.199642857142857e-05, "loss": 1.5906, "step": 2520500 }, { "epoch": 12.5, "learning_rate": 3.199285714285715e-05, "loss": 1.5862, "step": 2521000 }, { "epoch": 12.5, "learning_rate": 3.1989285714285715e-05, "loss": 1.5954, "step": 2521500 }, { "epoch": 12.5, "learning_rate": 3.198571428571429e-05, "loss": 1.6028, "step": 2522000 }, { "epoch": 12.51, "learning_rate": 3.198214285714286e-05, "loss": 1.5895, "step": 2522500 }, { "epoch": 12.51, "learning_rate": 3.197857142857143e-05, "loss": 1.5942, "step": 2523000 }, { "epoch": 12.51, "learning_rate": 3.1975e-05, "loss": 1.6025, "step": 2523500 }, { "epoch": 12.51, "learning_rate": 3.197142857142857e-05, "loss": 1.5939, "step": 2524000 }, { "epoch": 12.52, "learning_rate": 3.1967857142857146e-05, "loss": 1.5982, "step": 2524500 }, { "epoch": 12.52, "learning_rate": 3.1964285714285714e-05, "loss": 1.5894, "step": 2525000 }, { "epoch": 12.52, "learning_rate": 3.196071428571429e-05, "loss": 1.59, "step": 2525500 }, { "epoch": 12.53, "learning_rate": 3.1957142857142856e-05, "loss": 1.596, "step": 2526000 }, { "epoch": 12.53, "learning_rate": 3.195357142857143e-05, "loss": 1.5941, "step": 2526500 }, { "epoch": 12.53, "learning_rate": 3.1950000000000004e-05, "loss": 1.5975, "step": 2527000 }, { "epoch": 12.53, "learning_rate": 3.194642857142857e-05, "loss": 1.5995, "step": 2527500 }, { "epoch": 12.54, "learning_rate": 3.1942857142857146e-05, "loss": 1.5798, "step": 2528000 }, { "epoch": 12.54, "learning_rate": 3.193928571428571e-05, "loss": 1.5941, "step": 2528500 }, { "epoch": 12.54, "learning_rate": 3.193571428571429e-05, "loss": 1.5829, "step": 2529000 }, { "epoch": 12.54, "learning_rate": 3.193214285714286e-05, "loss": 1.6025, "step": 2529500 }, { "epoch": 12.55, "learning_rate": 3.192857142857143e-05, "loss": 1.5995, "step": 2530000 }, { "epoch": 12.55, "learning_rate": 3.1925e-05, "loss": 1.5922, "step": 2530500 }, { "epoch": 12.55, "learning_rate": 3.192142857142857e-05, "loss": 1.5841, "step": 2531000 }, { "epoch": 12.55, "learning_rate": 3.1917857142857145e-05, "loss": 1.5981, "step": 2531500 }, { "epoch": 12.56, "learning_rate": 3.191428571428571e-05, "loss": 1.5941, "step": 2532000 }, { "epoch": 12.56, "learning_rate": 3.191071428571429e-05, "loss": 1.583, "step": 2532500 }, { "epoch": 12.56, "learning_rate": 3.1907142857142854e-05, "loss": 1.595, "step": 2533000 }, { "epoch": 12.57, "learning_rate": 3.1903571428571435e-05, "loss": 1.5884, "step": 2533500 }, { "epoch": 12.57, "learning_rate": 3.19e-05, "loss": 1.5866, "step": 2534000 }, { "epoch": 12.57, "learning_rate": 3.189642857142857e-05, "loss": 1.5962, "step": 2534500 }, { "epoch": 12.57, "learning_rate": 3.1892857142857145e-05, "loss": 1.5908, "step": 2535000 }, { "epoch": 12.58, "learning_rate": 3.188928571428572e-05, "loss": 1.5843, "step": 2535500 }, { "epoch": 12.58, "learning_rate": 3.1885714285714286e-05, "loss": 1.5939, "step": 2536000 }, { "epoch": 12.58, "learning_rate": 3.188214285714286e-05, "loss": 1.5962, "step": 2536500 }, { "epoch": 12.58, "learning_rate": 3.187857142857143e-05, "loss": 1.5919, "step": 2537000 }, { "epoch": 12.59, "learning_rate": 3.1875e-05, "loss": 1.5965, "step": 2537500 }, { "epoch": 12.59, "learning_rate": 3.1871428571428577e-05, "loss": 1.5953, "step": 2538000 }, { "epoch": 12.59, "learning_rate": 3.1867857142857144e-05, "loss": 1.6023, "step": 2538500 }, { "epoch": 12.59, "learning_rate": 3.186428571428571e-05, "loss": 1.5851, "step": 2539000 }, { "epoch": 12.6, "learning_rate": 3.1860714285714286e-05, "loss": 1.5912, "step": 2539500 }, { "epoch": 12.6, "learning_rate": 3.185714285714286e-05, "loss": 1.5888, "step": 2540000 }, { "epoch": 12.6, "learning_rate": 3.1853571428571434e-05, "loss": 1.5945, "step": 2540500 }, { "epoch": 12.61, "learning_rate": 3.185e-05, "loss": 1.5934, "step": 2541000 }, { "epoch": 12.61, "learning_rate": 3.184642857142857e-05, "loss": 1.5977, "step": 2541500 }, { "epoch": 12.61, "learning_rate": 3.1842857142857143e-05, "loss": 1.5815, "step": 2542000 }, { "epoch": 12.61, "learning_rate": 3.183928571428572e-05, "loss": 1.5971, "step": 2542500 }, { "epoch": 12.62, "learning_rate": 3.1835714285714285e-05, "loss": 1.5815, "step": 2543000 }, { "epoch": 12.62, "learning_rate": 3.183214285714286e-05, "loss": 1.594, "step": 2543500 }, { "epoch": 12.62, "learning_rate": 3.182857142857143e-05, "loss": 1.5873, "step": 2544000 }, { "epoch": 12.62, "learning_rate": 3.1825e-05, "loss": 1.5952, "step": 2544500 }, { "epoch": 12.63, "learning_rate": 3.1821428571428575e-05, "loss": 1.5971, "step": 2545000 }, { "epoch": 12.63, "learning_rate": 3.181785714285714e-05, "loss": 1.5974, "step": 2545500 }, { "epoch": 12.63, "learning_rate": 3.181428571428571e-05, "loss": 1.5931, "step": 2546000 }, { "epoch": 12.63, "learning_rate": 3.181071428571429e-05, "loss": 1.5796, "step": 2546500 }, { "epoch": 12.64, "learning_rate": 3.180714285714286e-05, "loss": 1.6038, "step": 2547000 }, { "epoch": 12.64, "learning_rate": 3.180357142857143e-05, "loss": 1.5901, "step": 2547500 }, { "epoch": 12.64, "learning_rate": 3.18e-05, "loss": 1.5862, "step": 2548000 }, { "epoch": 12.64, "learning_rate": 3.1796428571428575e-05, "loss": 1.5932, "step": 2548500 }, { "epoch": 12.65, "learning_rate": 3.179285714285715e-05, "loss": 1.5934, "step": 2549000 }, { "epoch": 12.65, "learning_rate": 3.1789285714285716e-05, "loss": 1.595, "step": 2549500 }, { "epoch": 12.65, "learning_rate": 3.1785714285714284e-05, "loss": 1.5953, "step": 2550000 }, { "epoch": 12.66, "learning_rate": 3.178214285714286e-05, "loss": 1.5847, "step": 2550500 }, { "epoch": 12.66, "learning_rate": 3.177857142857143e-05, "loss": 1.5885, "step": 2551000 }, { "epoch": 12.66, "learning_rate": 3.1775e-05, "loss": 1.5805, "step": 2551500 }, { "epoch": 12.66, "learning_rate": 3.1771428571428574e-05, "loss": 1.5983, "step": 2552000 }, { "epoch": 12.67, "learning_rate": 3.176785714285714e-05, "loss": 1.5967, "step": 2552500 }, { "epoch": 12.67, "learning_rate": 3.1764285714285716e-05, "loss": 1.5953, "step": 2553000 }, { "epoch": 12.67, "learning_rate": 3.176071428571429e-05, "loss": 1.5932, "step": 2553500 }, { "epoch": 12.67, "learning_rate": 3.175714285714286e-05, "loss": 1.5817, "step": 2554000 }, { "epoch": 12.68, "learning_rate": 3.175357142857143e-05, "loss": 1.5878, "step": 2554500 }, { "epoch": 12.68, "learning_rate": 3.175e-05, "loss": 1.591, "step": 2555000 }, { "epoch": 12.68, "learning_rate": 3.1746428571428573e-05, "loss": 1.5942, "step": 2555500 }, { "epoch": 12.68, "learning_rate": 3.174285714285715e-05, "loss": 1.5844, "step": 2556000 }, { "epoch": 12.69, "learning_rate": 3.1739285714285715e-05, "loss": 1.5977, "step": 2556500 }, { "epoch": 12.69, "learning_rate": 3.173571428571428e-05, "loss": 1.5876, "step": 2557000 }, { "epoch": 12.69, "learning_rate": 3.173214285714286e-05, "loss": 1.5915, "step": 2557500 }, { "epoch": 12.7, "learning_rate": 3.172857142857143e-05, "loss": 1.5909, "step": 2558000 }, { "epoch": 12.7, "learning_rate": 3.1725e-05, "loss": 1.5977, "step": 2558500 }, { "epoch": 12.7, "learning_rate": 3.172142857142857e-05, "loss": 1.6001, "step": 2559000 }, { "epoch": 12.7, "learning_rate": 3.171785714285714e-05, "loss": 1.593, "step": 2559500 }, { "epoch": 12.71, "learning_rate": 3.1714285714285715e-05, "loss": 1.5863, "step": 2560000 }, { "epoch": 12.71, "learning_rate": 3.171071428571429e-05, "loss": 1.5805, "step": 2560500 }, { "epoch": 12.71, "learning_rate": 3.1707142857142856e-05, "loss": 1.5897, "step": 2561000 }, { "epoch": 12.71, "learning_rate": 3.170357142857143e-05, "loss": 1.5973, "step": 2561500 }, { "epoch": 12.72, "learning_rate": 3.1700000000000005e-05, "loss": 1.5899, "step": 2562000 }, { "epoch": 12.72, "learning_rate": 3.169642857142857e-05, "loss": 1.5942, "step": 2562500 }, { "epoch": 12.72, "learning_rate": 3.1692857142857147e-05, "loss": 1.5857, "step": 2563000 }, { "epoch": 12.72, "learning_rate": 3.1689285714285714e-05, "loss": 1.5927, "step": 2563500 }, { "epoch": 12.73, "learning_rate": 3.168571428571429e-05, "loss": 1.5885, "step": 2564000 }, { "epoch": 12.73, "learning_rate": 3.168214285714286e-05, "loss": 1.6004, "step": 2564500 }, { "epoch": 12.73, "learning_rate": 3.167857142857143e-05, "loss": 1.5929, "step": 2565000 }, { "epoch": 12.74, "learning_rate": 3.1675e-05, "loss": 1.5782, "step": 2565500 }, { "epoch": 12.74, "learning_rate": 3.167142857142857e-05, "loss": 1.5925, "step": 2566000 }, { "epoch": 12.74, "learning_rate": 3.1667857142857146e-05, "loss": 1.5985, "step": 2566500 }, { "epoch": 12.74, "learning_rate": 3.166428571428572e-05, "loss": 1.5921, "step": 2567000 }, { "epoch": 12.75, "learning_rate": 3.166071428571429e-05, "loss": 1.5895, "step": 2567500 }, { "epoch": 12.75, "learning_rate": 3.1657142857142855e-05, "loss": 1.5893, "step": 2568000 }, { "epoch": 12.75, "learning_rate": 3.165357142857143e-05, "loss": 1.5916, "step": 2568500 }, { "epoch": 12.75, "learning_rate": 3.1650000000000004e-05, "loss": 1.5858, "step": 2569000 }, { "epoch": 12.76, "learning_rate": 3.164642857142857e-05, "loss": 1.5948, "step": 2569500 }, { "epoch": 12.76, "learning_rate": 3.1642857142857145e-05, "loss": 1.5938, "step": 2570000 }, { "epoch": 12.76, "learning_rate": 3.163928571428571e-05, "loss": 1.5762, "step": 2570500 }, { "epoch": 12.76, "learning_rate": 3.163571428571429e-05, "loss": 1.6054, "step": 2571000 }, { "epoch": 12.77, "learning_rate": 3.163214285714286e-05, "loss": 1.5853, "step": 2571500 }, { "epoch": 12.77, "learning_rate": 3.162857142857143e-05, "loss": 1.5947, "step": 2572000 }, { "epoch": 12.77, "learning_rate": 3.1624999999999996e-05, "loss": 1.5956, "step": 2572500 }, { "epoch": 12.78, "learning_rate": 3.162142857142858e-05, "loss": 1.5924, "step": 2573000 }, { "epoch": 12.78, "learning_rate": 3.1617857142857145e-05, "loss": 1.5835, "step": 2573500 }, { "epoch": 12.78, "learning_rate": 3.161428571428572e-05, "loss": 1.5948, "step": 2574000 }, { "epoch": 12.78, "learning_rate": 3.1610714285714286e-05, "loss": 1.5931, "step": 2574500 }, { "epoch": 12.79, "learning_rate": 3.160714285714286e-05, "loss": 1.5997, "step": 2575000 }, { "epoch": 12.79, "learning_rate": 3.1603571428571435e-05, "loss": 1.5899, "step": 2575500 }, { "epoch": 12.79, "learning_rate": 3.16e-05, "loss": 1.5842, "step": 2576000 }, { "epoch": 12.79, "learning_rate": 3.159642857142857e-05, "loss": 1.596, "step": 2576500 }, { "epoch": 12.8, "learning_rate": 3.1592857142857144e-05, "loss": 1.5891, "step": 2577000 }, { "epoch": 12.8, "learning_rate": 3.158928571428572e-05, "loss": 1.5865, "step": 2577500 }, { "epoch": 12.8, "learning_rate": 3.1585714285714286e-05, "loss": 1.5896, "step": 2578000 }, { "epoch": 12.8, "learning_rate": 3.158214285714286e-05, "loss": 1.5883, "step": 2578500 }, { "epoch": 12.81, "learning_rate": 3.157857142857143e-05, "loss": 1.585, "step": 2579000 }, { "epoch": 12.81, "learning_rate": 3.1575e-05, "loss": 1.5773, "step": 2579500 }, { "epoch": 12.81, "learning_rate": 3.1571428571428576e-05, "loss": 1.5822, "step": 2580000 }, { "epoch": 12.82, "learning_rate": 3.1567857142857143e-05, "loss": 1.5915, "step": 2580500 }, { "epoch": 12.82, "learning_rate": 3.156428571428572e-05, "loss": 1.5834, "step": 2581000 }, { "epoch": 12.82, "learning_rate": 3.1560714285714285e-05, "loss": 1.5873, "step": 2581500 }, { "epoch": 12.82, "learning_rate": 3.155714285714286e-05, "loss": 1.5925, "step": 2582000 }, { "epoch": 12.83, "learning_rate": 3.1553571428571434e-05, "loss": 1.584, "step": 2582500 }, { "epoch": 12.83, "learning_rate": 3.155e-05, "loss": 1.5915, "step": 2583000 }, { "epoch": 12.83, "learning_rate": 3.154642857142857e-05, "loss": 1.5915, "step": 2583500 }, { "epoch": 12.83, "learning_rate": 3.154285714285714e-05, "loss": 1.5959, "step": 2584000 }, { "epoch": 12.84, "learning_rate": 3.153928571428572e-05, "loss": 1.5857, "step": 2584500 }, { "epoch": 12.84, "learning_rate": 3.1535714285714285e-05, "loss": 1.5876, "step": 2585000 }, { "epoch": 12.84, "learning_rate": 3.153214285714286e-05, "loss": 1.5808, "step": 2585500 }, { "epoch": 12.84, "learning_rate": 3.1528571428571426e-05, "loss": 1.5873, "step": 2586000 }, { "epoch": 12.85, "learning_rate": 3.1525e-05, "loss": 1.5928, "step": 2586500 }, { "epoch": 12.85, "learning_rate": 3.1521428571428575e-05, "loss": 1.5945, "step": 2587000 }, { "epoch": 12.85, "learning_rate": 3.151785714285714e-05, "loss": 1.5946, "step": 2587500 }, { "epoch": 12.86, "learning_rate": 3.1514285714285717e-05, "loss": 1.5842, "step": 2588000 }, { "epoch": 12.86, "learning_rate": 3.151071428571429e-05, "loss": 1.5991, "step": 2588500 }, { "epoch": 12.86, "learning_rate": 3.150714285714286e-05, "loss": 1.5969, "step": 2589000 }, { "epoch": 12.86, "learning_rate": 3.150357142857143e-05, "loss": 1.5915, "step": 2589500 }, { "epoch": 12.87, "learning_rate": 3.15e-05, "loss": 1.6, "step": 2590000 }, { "epoch": 12.87, "learning_rate": 3.1496428571428574e-05, "loss": 1.5997, "step": 2590500 }, { "epoch": 12.87, "learning_rate": 3.149285714285715e-05, "loss": 1.5977, "step": 2591000 }, { "epoch": 12.87, "learning_rate": 3.1489285714285716e-05, "loss": 1.5803, "step": 2591500 }, { "epoch": 12.88, "learning_rate": 3.148571428571428e-05, "loss": 1.5935, "step": 2592000 }, { "epoch": 12.88, "learning_rate": 3.148214285714286e-05, "loss": 1.5921, "step": 2592500 }, { "epoch": 12.88, "learning_rate": 3.147857142857143e-05, "loss": 1.5836, "step": 2593000 }, { "epoch": 12.88, "learning_rate": 3.1475e-05, "loss": 1.5922, "step": 2593500 }, { "epoch": 12.89, "learning_rate": 3.1471428571428574e-05, "loss": 1.5981, "step": 2594000 }, { "epoch": 12.89, "learning_rate": 3.146785714285714e-05, "loss": 1.5931, "step": 2594500 }, { "epoch": 12.89, "learning_rate": 3.1464285714285715e-05, "loss": 1.5924, "step": 2595000 }, { "epoch": 12.9, "learning_rate": 3.146071428571429e-05, "loss": 1.59, "step": 2595500 }, { "epoch": 12.9, "learning_rate": 3.145714285714286e-05, "loss": 1.5818, "step": 2596000 }, { "epoch": 12.9, "learning_rate": 3.145357142857143e-05, "loss": 1.5819, "step": 2596500 }, { "epoch": 12.9, "learning_rate": 3.145e-05, "loss": 1.5828, "step": 2597000 }, { "epoch": 12.91, "learning_rate": 3.144642857142857e-05, "loss": 1.589, "step": 2597500 }, { "epoch": 12.91, "learning_rate": 3.144285714285715e-05, "loss": 1.5863, "step": 2598000 }, { "epoch": 12.91, "learning_rate": 3.1439285714285715e-05, "loss": 1.5796, "step": 2598500 }, { "epoch": 12.91, "learning_rate": 3.143571428571428e-05, "loss": 1.5922, "step": 2599000 }, { "epoch": 12.92, "learning_rate": 3.143214285714286e-05, "loss": 1.5819, "step": 2599500 }, { "epoch": 12.92, "learning_rate": 3.142857142857143e-05, "loss": 1.5816, "step": 2600000 }, { "epoch": 12.92, "learning_rate": 3.1425e-05, "loss": 1.5836, "step": 2600500 }, { "epoch": 12.92, "learning_rate": 3.142142857142857e-05, "loss": 1.5851, "step": 2601000 }, { "epoch": 12.93, "learning_rate": 3.141785714285715e-05, "loss": 1.5814, "step": 2601500 }, { "epoch": 12.93, "learning_rate": 3.141428571428572e-05, "loss": 1.582, "step": 2602000 }, { "epoch": 12.93, "learning_rate": 3.141071428571429e-05, "loss": 1.5992, "step": 2602500 }, { "epoch": 12.94, "learning_rate": 3.1407142857142856e-05, "loss": 1.5843, "step": 2603000 }, { "epoch": 12.94, "learning_rate": 3.140357142857143e-05, "loss": 1.5798, "step": 2603500 }, { "epoch": 12.94, "learning_rate": 3.1400000000000004e-05, "loss": 1.5777, "step": 2604000 }, { "epoch": 12.94, "learning_rate": 3.139642857142857e-05, "loss": 1.5803, "step": 2604500 }, { "epoch": 12.95, "learning_rate": 3.1392857142857146e-05, "loss": 1.5926, "step": 2605000 }, { "epoch": 12.95, "learning_rate": 3.1389285714285713e-05, "loss": 1.605, "step": 2605500 }, { "epoch": 12.95, "learning_rate": 3.138571428571429e-05, "loss": 1.5832, "step": 2606000 }, { "epoch": 12.95, "learning_rate": 3.138214285714286e-05, "loss": 1.5848, "step": 2606500 }, { "epoch": 12.96, "learning_rate": 3.137857142857143e-05, "loss": 1.5907, "step": 2607000 }, { "epoch": 12.96, "learning_rate": 3.1375e-05, "loss": 1.5811, "step": 2607500 }, { "epoch": 12.96, "learning_rate": 3.137142857142857e-05, "loss": 1.595, "step": 2608000 }, { "epoch": 12.96, "learning_rate": 3.1367857142857145e-05, "loss": 1.5893, "step": 2608500 }, { "epoch": 12.97, "learning_rate": 3.136428571428572e-05, "loss": 1.5925, "step": 2609000 }, { "epoch": 12.97, "learning_rate": 3.136071428571429e-05, "loss": 1.5915, "step": 2609500 }, { "epoch": 12.97, "learning_rate": 3.1357142857142855e-05, "loss": 1.5887, "step": 2610000 }, { "epoch": 12.98, "learning_rate": 3.135357142857143e-05, "loss": 1.5909, "step": 2610500 }, { "epoch": 12.98, "learning_rate": 3.135e-05, "loss": 1.5811, "step": 2611000 }, { "epoch": 12.98, "learning_rate": 3.134642857142857e-05, "loss": 1.5913, "step": 2611500 }, { "epoch": 12.98, "learning_rate": 3.1342857142857145e-05, "loss": 1.591, "step": 2612000 }, { "epoch": 12.99, "learning_rate": 3.133928571428571e-05, "loss": 1.5847, "step": 2612500 }, { "epoch": 12.99, "learning_rate": 3.1335714285714287e-05, "loss": 1.5913, "step": 2613000 }, { "epoch": 12.99, "learning_rate": 3.133214285714286e-05, "loss": 1.5968, "step": 2613500 }, { "epoch": 12.99, "learning_rate": 3.132857142857143e-05, "loss": 1.5888, "step": 2614000 }, { "epoch": 13.0, "learning_rate": 3.1324999999999996e-05, "loss": 1.5864, "step": 2614500 }, { "epoch": 13.0, "learning_rate": 3.132142857142858e-05, "loss": 1.586, "step": 2615000 }, { "epoch": 13.0, "learning_rate": 3.1317857142857144e-05, "loss": 1.5805, "step": 2615500 }, { "epoch": 13.0, "learning_rate": 3.131428571428572e-05, "loss": 1.5758, "step": 2616000 }, { "epoch": 13.01, "learning_rate": 3.1310714285714286e-05, "loss": 1.5773, "step": 2616500 }, { "epoch": 13.01, "learning_rate": 3.130714285714286e-05, "loss": 1.581, "step": 2617000 }, { "epoch": 13.01, "learning_rate": 3.1303571428571434e-05, "loss": 1.5764, "step": 2617500 }, { "epoch": 13.02, "learning_rate": 3.13e-05, "loss": 1.5769, "step": 2618000 }, { "epoch": 13.02, "learning_rate": 3.129642857142857e-05, "loss": 1.5703, "step": 2618500 }, { "epoch": 13.02, "learning_rate": 3.1292857142857144e-05, "loss": 1.5886, "step": 2619000 }, { "epoch": 13.02, "learning_rate": 3.128928571428572e-05, "loss": 1.5852, "step": 2619500 }, { "epoch": 13.03, "learning_rate": 3.1285714285714285e-05, "loss": 1.5867, "step": 2620000 }, { "epoch": 13.03, "learning_rate": 3.128214285714286e-05, "loss": 1.581, "step": 2620500 }, { "epoch": 13.03, "learning_rate": 3.127857142857143e-05, "loss": 1.5779, "step": 2621000 }, { "epoch": 13.03, "learning_rate": 3.1275e-05, "loss": 1.574, "step": 2621500 }, { "epoch": 13.04, "learning_rate": 3.1271428571428576e-05, "loss": 1.581, "step": 2622000 }, { "epoch": 13.04, "learning_rate": 3.126785714285714e-05, "loss": 1.6035, "step": 2622500 }, { "epoch": 13.04, "learning_rate": 3.126428571428572e-05, "loss": 1.5869, "step": 2623000 }, { "epoch": 13.04, "learning_rate": 3.1260714285714285e-05, "loss": 1.5859, "step": 2623500 }, { "epoch": 13.05, "learning_rate": 3.125714285714286e-05, "loss": 1.5814, "step": 2624000 }, { "epoch": 13.05, "learning_rate": 3.125357142857143e-05, "loss": 1.5803, "step": 2624500 }, { "epoch": 13.05, "learning_rate": 3.125e-05, "loss": 1.5768, "step": 2625000 }, { "epoch": 13.05, "learning_rate": 3.124642857142857e-05, "loss": 1.5697, "step": 2625500 }, { "epoch": 13.06, "learning_rate": 3.124285714285715e-05, "loss": 1.5775, "step": 2626000 }, { "epoch": 13.06, "learning_rate": 3.123928571428572e-05, "loss": 1.5903, "step": 2626500 }, { "epoch": 13.06, "learning_rate": 3.1235714285714284e-05, "loss": 1.5749, "step": 2627000 }, { "epoch": 13.07, "learning_rate": 3.123214285714286e-05, "loss": 1.5872, "step": 2627500 }, { "epoch": 13.07, "learning_rate": 3.122857142857143e-05, "loss": 1.5754, "step": 2628000 }, { "epoch": 13.07, "learning_rate": 3.122500000000001e-05, "loss": 1.5816, "step": 2628500 }, { "epoch": 13.07, "learning_rate": 3.1221428571428574e-05, "loss": 1.5791, "step": 2629000 }, { "epoch": 13.08, "learning_rate": 3.121785714285714e-05, "loss": 1.5835, "step": 2629500 }, { "epoch": 13.08, "learning_rate": 3.1214285714285716e-05, "loss": 1.5752, "step": 2630000 }, { "epoch": 13.08, "learning_rate": 3.121071428571429e-05, "loss": 1.5632, "step": 2630500 }, { "epoch": 13.08, "learning_rate": 3.120714285714286e-05, "loss": 1.5783, "step": 2631000 }, { "epoch": 13.09, "learning_rate": 3.120357142857143e-05, "loss": 1.5831, "step": 2631500 }, { "epoch": 13.09, "learning_rate": 3.12e-05, "loss": 1.5836, "step": 2632000 }, { "epoch": 13.09, "learning_rate": 3.1196428571428574e-05, "loss": 1.5839, "step": 2632500 }, { "epoch": 13.09, "learning_rate": 3.119285714285715e-05, "loss": 1.5789, "step": 2633000 }, { "epoch": 13.1, "learning_rate": 3.1189285714285715e-05, "loss": 1.5787, "step": 2633500 }, { "epoch": 13.1, "learning_rate": 3.118571428571428e-05, "loss": 1.5793, "step": 2634000 }, { "epoch": 13.1, "learning_rate": 3.118214285714286e-05, "loss": 1.5823, "step": 2634500 }, { "epoch": 13.11, "learning_rate": 3.117857142857143e-05, "loss": 1.5805, "step": 2635000 }, { "epoch": 13.11, "learning_rate": 3.1175000000000006e-05, "loss": 1.5815, "step": 2635500 }, { "epoch": 13.11, "learning_rate": 3.117142857142857e-05, "loss": 1.5775, "step": 2636000 }, { "epoch": 13.11, "learning_rate": 3.116785714285714e-05, "loss": 1.5653, "step": 2636500 }, { "epoch": 13.12, "learning_rate": 3.1164285714285715e-05, "loss": 1.5838, "step": 2637000 }, { "epoch": 13.12, "learning_rate": 3.116071428571429e-05, "loss": 1.5943, "step": 2637500 }, { "epoch": 13.12, "learning_rate": 3.1157142857142857e-05, "loss": 1.5891, "step": 2638000 }, { "epoch": 13.12, "learning_rate": 3.115357142857143e-05, "loss": 1.5839, "step": 2638500 }, { "epoch": 13.13, "learning_rate": 3.115e-05, "loss": 1.589, "step": 2639000 }, { "epoch": 13.13, "learning_rate": 3.114642857142857e-05, "loss": 1.5783, "step": 2639500 }, { "epoch": 13.13, "learning_rate": 3.114285714285715e-05, "loss": 1.5757, "step": 2640000 }, { "epoch": 13.13, "learning_rate": 3.1139285714285714e-05, "loss": 1.5706, "step": 2640500 }, { "epoch": 13.14, "learning_rate": 3.113571428571428e-05, "loss": 1.5887, "step": 2641000 }, { "epoch": 13.14, "learning_rate": 3.113214285714286e-05, "loss": 1.5863, "step": 2641500 }, { "epoch": 13.14, "learning_rate": 3.112857142857143e-05, "loss": 1.5861, "step": 2642000 }, { "epoch": 13.15, "learning_rate": 3.1125000000000004e-05, "loss": 1.5772, "step": 2642500 }, { "epoch": 13.15, "learning_rate": 3.112142857142857e-05, "loss": 1.5818, "step": 2643000 }, { "epoch": 13.15, "learning_rate": 3.1117857142857146e-05, "loss": 1.5894, "step": 2643500 }, { "epoch": 13.15, "learning_rate": 3.111428571428572e-05, "loss": 1.5803, "step": 2644000 }, { "epoch": 13.16, "learning_rate": 3.111071428571429e-05, "loss": 1.5808, "step": 2644500 }, { "epoch": 13.16, "learning_rate": 3.1107142857142855e-05, "loss": 1.5818, "step": 2645000 }, { "epoch": 13.16, "learning_rate": 3.110357142857143e-05, "loss": 1.5674, "step": 2645500 }, { "epoch": 13.16, "learning_rate": 3.1100000000000004e-05, "loss": 1.5782, "step": 2646000 }, { "epoch": 13.17, "learning_rate": 3.109642857142857e-05, "loss": 1.5754, "step": 2646500 }, { "epoch": 13.17, "learning_rate": 3.1092857142857145e-05, "loss": 1.5822, "step": 2647000 }, { "epoch": 13.17, "learning_rate": 3.108928571428571e-05, "loss": 1.5774, "step": 2647500 }, { "epoch": 13.17, "learning_rate": 3.108571428571429e-05, "loss": 1.5709, "step": 2648000 }, { "epoch": 13.18, "learning_rate": 3.108214285714286e-05, "loss": 1.5784, "step": 2648500 }, { "epoch": 13.18, "learning_rate": 3.107857142857143e-05, "loss": 1.5798, "step": 2649000 }, { "epoch": 13.18, "learning_rate": 3.1075e-05, "loss": 1.5843, "step": 2649500 }, { "epoch": 13.19, "learning_rate": 3.107142857142857e-05, "loss": 1.5688, "step": 2650000 }, { "epoch": 13.19, "learning_rate": 3.1067857142857145e-05, "loss": 1.5876, "step": 2650500 }, { "epoch": 13.19, "learning_rate": 3.106428571428572e-05, "loss": 1.5825, "step": 2651000 }, { "epoch": 13.19, "learning_rate": 3.1060714285714287e-05, "loss": 1.5823, "step": 2651500 }, { "epoch": 13.2, "learning_rate": 3.1057142857142854e-05, "loss": 1.5818, "step": 2652000 }, { "epoch": 13.2, "learning_rate": 3.1053571428571435e-05, "loss": 1.5753, "step": 2652500 }, { "epoch": 13.2, "learning_rate": 3.105e-05, "loss": 1.5734, "step": 2653000 }, { "epoch": 13.2, "learning_rate": 3.104642857142857e-05, "loss": 1.5836, "step": 2653500 }, { "epoch": 13.21, "learning_rate": 3.1042857142857144e-05, "loss": 1.5593, "step": 2654000 }, { "epoch": 13.21, "learning_rate": 3.103928571428572e-05, "loss": 1.5886, "step": 2654500 }, { "epoch": 13.21, "learning_rate": 3.1035714285714286e-05, "loss": 1.5761, "step": 2655000 }, { "epoch": 13.21, "learning_rate": 3.103214285714286e-05, "loss": 1.5756, "step": 2655500 }, { "epoch": 13.22, "learning_rate": 3.102857142857143e-05, "loss": 1.5807, "step": 2656000 }, { "epoch": 13.22, "learning_rate": 3.1025e-05, "loss": 1.5836, "step": 2656500 }, { "epoch": 13.22, "learning_rate": 3.1021428571428576e-05, "loss": 1.5662, "step": 2657000 }, { "epoch": 13.23, "learning_rate": 3.1017857142857144e-05, "loss": 1.5773, "step": 2657500 }, { "epoch": 13.23, "learning_rate": 3.101428571428572e-05, "loss": 1.5754, "step": 2658000 }, { "epoch": 13.23, "learning_rate": 3.1010714285714285e-05, "loss": 1.5686, "step": 2658500 }, { "epoch": 13.23, "learning_rate": 3.100714285714286e-05, "loss": 1.5741, "step": 2659000 }, { "epoch": 13.24, "learning_rate": 3.1003571428571434e-05, "loss": 1.5839, "step": 2659500 }, { "epoch": 13.24, "learning_rate": 3.1e-05, "loss": 1.5861, "step": 2660000 }, { "epoch": 13.24, "learning_rate": 3.099642857142857e-05, "loss": 1.5699, "step": 2660500 }, { "epoch": 13.24, "learning_rate": 3.099285714285714e-05, "loss": 1.5714, "step": 2661000 }, { "epoch": 13.25, "learning_rate": 3.098928571428572e-05, "loss": 1.5836, "step": 2661500 }, { "epoch": 13.25, "learning_rate": 3.0985714285714285e-05, "loss": 1.5739, "step": 2662000 }, { "epoch": 13.25, "learning_rate": 3.098214285714286e-05, "loss": 1.5734, "step": 2662500 }, { "epoch": 13.25, "learning_rate": 3.0978571428571426e-05, "loss": 1.574, "step": 2663000 }, { "epoch": 13.26, "learning_rate": 3.0975e-05, "loss": 1.582, "step": 2663500 }, { "epoch": 13.26, "learning_rate": 3.0971428571428575e-05, "loss": 1.5793, "step": 2664000 }, { "epoch": 13.26, "learning_rate": 3.096785714285714e-05, "loss": 1.5632, "step": 2664500 }, { "epoch": 13.27, "learning_rate": 3.096428571428572e-05, "loss": 1.571, "step": 2665000 }, { "epoch": 13.27, "learning_rate": 3.0960714285714284e-05, "loss": 1.5801, "step": 2665500 }, { "epoch": 13.27, "learning_rate": 3.095714285714286e-05, "loss": 1.5804, "step": 2666000 }, { "epoch": 13.27, "learning_rate": 3.095357142857143e-05, "loss": 1.5777, "step": 2666500 }, { "epoch": 13.28, "learning_rate": 3.095e-05, "loss": 1.5818, "step": 2667000 }, { "epoch": 13.28, "learning_rate": 3.094642857142857e-05, "loss": 1.58, "step": 2667500 }, { "epoch": 13.28, "learning_rate": 3.094285714285715e-05, "loss": 1.5809, "step": 2668000 }, { "epoch": 13.28, "learning_rate": 3.0939285714285716e-05, "loss": 1.5816, "step": 2668500 }, { "epoch": 13.29, "learning_rate": 3.0935714285714284e-05, "loss": 1.576, "step": 2669000 }, { "epoch": 13.29, "learning_rate": 3.093214285714286e-05, "loss": 1.5695, "step": 2669500 }, { "epoch": 13.29, "learning_rate": 3.092857142857143e-05, "loss": 1.5769, "step": 2670000 }, { "epoch": 13.29, "learning_rate": 3.0925000000000006e-05, "loss": 1.5727, "step": 2670500 }, { "epoch": 13.3, "learning_rate": 3.0921428571428574e-05, "loss": 1.5753, "step": 2671000 }, { "epoch": 13.3, "learning_rate": 3.091785714285714e-05, "loss": 1.5632, "step": 2671500 }, { "epoch": 13.3, "learning_rate": 3.0914285714285715e-05, "loss": 1.5795, "step": 2672000 }, { "epoch": 13.31, "learning_rate": 3.091071428571429e-05, "loss": 1.5783, "step": 2672500 }, { "epoch": 13.31, "learning_rate": 3.090714285714286e-05, "loss": 1.5714, "step": 2673000 }, { "epoch": 13.31, "learning_rate": 3.090357142857143e-05, "loss": 1.5766, "step": 2673500 }, { "epoch": 13.31, "learning_rate": 3.09e-05, "loss": 1.5691, "step": 2674000 }, { "epoch": 13.32, "learning_rate": 3.089642857142857e-05, "loss": 1.588, "step": 2674500 }, { "epoch": 13.32, "learning_rate": 3.089285714285715e-05, "loss": 1.5685, "step": 2675000 }, { "epoch": 13.32, "learning_rate": 3.0889285714285715e-05, "loss": 1.5765, "step": 2675500 }, { "epoch": 13.32, "learning_rate": 3.088571428571428e-05, "loss": 1.5834, "step": 2676000 }, { "epoch": 13.33, "learning_rate": 3.0882142857142857e-05, "loss": 1.5753, "step": 2676500 }, { "epoch": 13.33, "learning_rate": 3.087857142857143e-05, "loss": 1.5833, "step": 2677000 }, { "epoch": 13.33, "learning_rate": 3.0875000000000005e-05, "loss": 1.585, "step": 2677500 }, { "epoch": 13.33, "learning_rate": 3.087142857142857e-05, "loss": 1.5741, "step": 2678000 }, { "epoch": 13.34, "learning_rate": 3.086785714285714e-05, "loss": 1.5781, "step": 2678500 }, { "epoch": 13.34, "learning_rate": 3.086428571428572e-05, "loss": 1.5871, "step": 2679000 }, { "epoch": 13.34, "learning_rate": 3.086071428571429e-05, "loss": 1.5836, "step": 2679500 }, { "epoch": 13.35, "learning_rate": 3.0857142857142856e-05, "loss": 1.579, "step": 2680000 }, { "epoch": 13.35, "learning_rate": 3.085357142857143e-05, "loss": 1.5782, "step": 2680500 }, { "epoch": 13.35, "learning_rate": 3.0850000000000004e-05, "loss": 1.5822, "step": 2681000 }, { "epoch": 13.35, "learning_rate": 3.084642857142857e-05, "loss": 1.5801, "step": 2681500 }, { "epoch": 13.36, "learning_rate": 3.0842857142857146e-05, "loss": 1.5854, "step": 2682000 }, { "epoch": 13.36, "learning_rate": 3.0839285714285714e-05, "loss": 1.5736, "step": 2682500 }, { "epoch": 13.36, "learning_rate": 3.083571428571428e-05, "loss": 1.5769, "step": 2683000 }, { "epoch": 13.36, "learning_rate": 3.083214285714286e-05, "loss": 1.584, "step": 2683500 }, { "epoch": 13.37, "learning_rate": 3.082857142857143e-05, "loss": 1.5666, "step": 2684000 }, { "epoch": 13.37, "learning_rate": 3.0825000000000004e-05, "loss": 1.5758, "step": 2684500 }, { "epoch": 13.37, "learning_rate": 3.082142857142857e-05, "loss": 1.5772, "step": 2685000 }, { "epoch": 13.37, "learning_rate": 3.0817857142857146e-05, "loss": 1.5861, "step": 2685500 }, { "epoch": 13.38, "learning_rate": 3.081428571428572e-05, "loss": 1.5811, "step": 2686000 }, { "epoch": 13.38, "learning_rate": 3.081071428571429e-05, "loss": 1.581, "step": 2686500 }, { "epoch": 13.38, "learning_rate": 3.0807142857142855e-05, "loss": 1.5799, "step": 2687000 }, { "epoch": 13.39, "learning_rate": 3.080357142857143e-05, "loss": 1.5757, "step": 2687500 }, { "epoch": 13.39, "learning_rate": 3.08e-05, "loss": 1.5788, "step": 2688000 }, { "epoch": 13.39, "learning_rate": 3.079642857142857e-05, "loss": 1.5789, "step": 2688500 }, { "epoch": 13.39, "learning_rate": 3.0792857142857145e-05, "loss": 1.5785, "step": 2689000 }, { "epoch": 13.4, "learning_rate": 3.078928571428571e-05, "loss": 1.5812, "step": 2689500 }, { "epoch": 13.4, "learning_rate": 3.078571428571429e-05, "loss": 1.5739, "step": 2690000 }, { "epoch": 13.4, "learning_rate": 3.078214285714286e-05, "loss": 1.5729, "step": 2690500 }, { "epoch": 13.4, "learning_rate": 3.077857142857143e-05, "loss": 1.5743, "step": 2691000 }, { "epoch": 13.41, "learning_rate": 3.0775e-05, "loss": 1.5715, "step": 2691500 }, { "epoch": 13.41, "learning_rate": 3.077142857142857e-05, "loss": 1.575, "step": 2692000 }, { "epoch": 13.41, "learning_rate": 3.0767857142857144e-05, "loss": 1.5794, "step": 2692500 }, { "epoch": 13.41, "learning_rate": 3.076428571428572e-05, "loss": 1.5786, "step": 2693000 }, { "epoch": 13.42, "learning_rate": 3.0760714285714286e-05, "loss": 1.5706, "step": 2693500 }, { "epoch": 13.42, "learning_rate": 3.0757142857142854e-05, "loss": 1.5811, "step": 2694000 }, { "epoch": 13.42, "learning_rate": 3.0753571428571435e-05, "loss": 1.5693, "step": 2694500 }, { "epoch": 13.43, "learning_rate": 3.075e-05, "loss": 1.5731, "step": 2695000 }, { "epoch": 13.43, "learning_rate": 3.074642857142857e-05, "loss": 1.5787, "step": 2695500 }, { "epoch": 13.43, "learning_rate": 3.0742857142857144e-05, "loss": 1.5813, "step": 2696000 }, { "epoch": 13.43, "learning_rate": 3.073928571428572e-05, "loss": 1.5638, "step": 2696500 }, { "epoch": 13.44, "learning_rate": 3.073571428571429e-05, "loss": 1.5702, "step": 2697000 }, { "epoch": 13.44, "learning_rate": 3.073214285714286e-05, "loss": 1.5796, "step": 2697500 }, { "epoch": 13.44, "learning_rate": 3.072857142857143e-05, "loss": 1.5838, "step": 2698000 }, { "epoch": 13.44, "learning_rate": 3.0725e-05, "loss": 1.5743, "step": 2698500 }, { "epoch": 13.45, "learning_rate": 3.0721428571428576e-05, "loss": 1.575, "step": 2699000 }, { "epoch": 13.45, "learning_rate": 3.071785714285714e-05, "loss": 1.5695, "step": 2699500 }, { "epoch": 13.45, "learning_rate": 3.071428571428572e-05, "loss": 1.5785, "step": 2700000 }, { "epoch": 13.45, "learning_rate": 3.0710714285714285e-05, "loss": 1.5786, "step": 2700500 }, { "epoch": 13.46, "learning_rate": 3.070714285714286e-05, "loss": 1.5813, "step": 2701000 }, { "epoch": 13.46, "learning_rate": 3.070357142857143e-05, "loss": 1.5759, "step": 2701500 }, { "epoch": 13.46, "learning_rate": 3.07e-05, "loss": 1.5751, "step": 2702000 }, { "epoch": 13.46, "learning_rate": 3.069642857142857e-05, "loss": 1.572, "step": 2702500 }, { "epoch": 13.47, "learning_rate": 3.069285714285714e-05, "loss": 1.5719, "step": 2703000 }, { "epoch": 13.47, "learning_rate": 3.068928571428572e-05, "loss": 1.5774, "step": 2703500 }, { "epoch": 13.47, "learning_rate": 3.068571428571429e-05, "loss": 1.5661, "step": 2704000 }, { "epoch": 13.48, "learning_rate": 3.068214285714286e-05, "loss": 1.5781, "step": 2704500 }, { "epoch": 13.48, "learning_rate": 3.0678571428571426e-05, "loss": 1.5715, "step": 2705000 }, { "epoch": 13.48, "learning_rate": 3.067500000000001e-05, "loss": 1.5704, "step": 2705500 }, { "epoch": 13.48, "learning_rate": 3.0671428571428574e-05, "loss": 1.5747, "step": 2706000 }, { "epoch": 13.49, "learning_rate": 3.066785714285714e-05, "loss": 1.5798, "step": 2706500 }, { "epoch": 13.49, "learning_rate": 3.0664285714285716e-05, "loss": 1.5805, "step": 2707000 }, { "epoch": 13.49, "learning_rate": 3.066071428571429e-05, "loss": 1.5798, "step": 2707500 }, { "epoch": 13.49, "learning_rate": 3.065714285714286e-05, "loss": 1.5567, "step": 2708000 }, { "epoch": 13.5, "learning_rate": 3.065357142857143e-05, "loss": 1.5645, "step": 2708500 }, { "epoch": 13.5, "learning_rate": 3.065e-05, "loss": 1.5721, "step": 2709000 }, { "epoch": 13.5, "learning_rate": 3.064642857142857e-05, "loss": 1.5752, "step": 2709500 }, { "epoch": 13.5, "learning_rate": 3.064285714285715e-05, "loss": 1.57, "step": 2710000 }, { "epoch": 13.51, "learning_rate": 3.0639285714285716e-05, "loss": 1.5672, "step": 2710500 }, { "epoch": 13.51, "learning_rate": 3.063571428571429e-05, "loss": 1.577, "step": 2711000 }, { "epoch": 13.51, "learning_rate": 3.063214285714286e-05, "loss": 1.5756, "step": 2711500 }, { "epoch": 13.52, "learning_rate": 3.062857142857143e-05, "loss": 1.565, "step": 2712000 }, { "epoch": 13.52, "learning_rate": 3.0625000000000006e-05, "loss": 1.5704, "step": 2712500 }, { "epoch": 13.52, "learning_rate": 3.062142857142857e-05, "loss": 1.5736, "step": 2713000 }, { "epoch": 13.52, "learning_rate": 3.061785714285714e-05, "loss": 1.5707, "step": 2713500 }, { "epoch": 13.53, "learning_rate": 3.0614285714285715e-05, "loss": 1.5681, "step": 2714000 }, { "epoch": 13.53, "learning_rate": 3.061071428571429e-05, "loss": 1.564, "step": 2714500 }, { "epoch": 13.53, "learning_rate": 3.060714285714286e-05, "loss": 1.5846, "step": 2715000 }, { "epoch": 13.53, "learning_rate": 3.060357142857143e-05, "loss": 1.5787, "step": 2715500 }, { "epoch": 13.54, "learning_rate": 3.06e-05, "loss": 1.567, "step": 2716000 }, { "epoch": 13.54, "learning_rate": 3.059642857142857e-05, "loss": 1.5667, "step": 2716500 }, { "epoch": 13.54, "learning_rate": 3.059285714285715e-05, "loss": 1.5795, "step": 2717000 }, { "epoch": 13.54, "learning_rate": 3.0589285714285714e-05, "loss": 1.5796, "step": 2717500 }, { "epoch": 13.55, "learning_rate": 3.058571428571429e-05, "loss": 1.5673, "step": 2718000 }, { "epoch": 13.55, "learning_rate": 3.0582142857142856e-05, "loss": 1.5645, "step": 2718500 }, { "epoch": 13.55, "learning_rate": 3.057857142857143e-05, "loss": 1.576, "step": 2719000 }, { "epoch": 13.56, "learning_rate": 3.0575000000000005e-05, "loss": 1.5778, "step": 2719500 }, { "epoch": 13.56, "learning_rate": 3.057142857142857e-05, "loss": 1.5722, "step": 2720000 }, { "epoch": 13.56, "learning_rate": 3.056785714285714e-05, "loss": 1.5629, "step": 2720500 }, { "epoch": 13.56, "learning_rate": 3.056428571428572e-05, "loss": 1.5636, "step": 2721000 }, { "epoch": 13.57, "learning_rate": 3.056071428571429e-05, "loss": 1.5657, "step": 2721500 }, { "epoch": 13.57, "learning_rate": 3.0557142857142855e-05, "loss": 1.5766, "step": 2722000 }, { "epoch": 13.57, "learning_rate": 3.055357142857143e-05, "loss": 1.5791, "step": 2722500 }, { "epoch": 13.57, "learning_rate": 3.0550000000000004e-05, "loss": 1.5763, "step": 2723000 }, { "epoch": 13.58, "learning_rate": 3.054642857142857e-05, "loss": 1.5813, "step": 2723500 }, { "epoch": 13.58, "learning_rate": 3.0542857142857146e-05, "loss": 1.5717, "step": 2724000 }, { "epoch": 13.58, "learning_rate": 3.053928571428571e-05, "loss": 1.5678, "step": 2724500 }, { "epoch": 13.58, "learning_rate": 3.053571428571429e-05, "loss": 1.5729, "step": 2725000 }, { "epoch": 13.59, "learning_rate": 3.053214285714286e-05, "loss": 1.571, "step": 2725500 }, { "epoch": 13.59, "learning_rate": 3.052857142857143e-05, "loss": 1.5755, "step": 2726000 }, { "epoch": 13.59, "learning_rate": 3.0525e-05, "loss": 1.5752, "step": 2726500 }, { "epoch": 13.6, "learning_rate": 3.052142857142857e-05, "loss": 1.5791, "step": 2727000 }, { "epoch": 13.6, "learning_rate": 3.0517857142857145e-05, "loss": 1.5694, "step": 2727500 }, { "epoch": 13.6, "learning_rate": 3.0514285714285716e-05, "loss": 1.5685, "step": 2728000 }, { "epoch": 13.6, "learning_rate": 3.0510714285714287e-05, "loss": 1.5927, "step": 2728500 }, { "epoch": 13.61, "learning_rate": 3.0507142857142858e-05, "loss": 1.5768, "step": 2729000 }, { "epoch": 13.61, "learning_rate": 3.0503571428571432e-05, "loss": 1.5716, "step": 2729500 }, { "epoch": 13.61, "learning_rate": 3.05e-05, "loss": 1.562, "step": 2730000 }, { "epoch": 13.61, "learning_rate": 3.049642857142857e-05, "loss": 1.5729, "step": 2730500 }, { "epoch": 13.62, "learning_rate": 3.0492857142857144e-05, "loss": 1.574, "step": 2731000 }, { "epoch": 13.62, "learning_rate": 3.0489285714285715e-05, "loss": 1.5639, "step": 2731500 }, { "epoch": 13.62, "learning_rate": 3.048571428571429e-05, "loss": 1.5747, "step": 2732000 }, { "epoch": 13.62, "learning_rate": 3.048214285714286e-05, "loss": 1.5674, "step": 2732500 }, { "epoch": 13.63, "learning_rate": 3.0478571428571428e-05, "loss": 1.5747, "step": 2733000 }, { "epoch": 13.63, "learning_rate": 3.0475000000000002e-05, "loss": 1.5649, "step": 2733500 }, { "epoch": 13.63, "learning_rate": 3.0471428571428573e-05, "loss": 1.5727, "step": 2734000 }, { "epoch": 13.64, "learning_rate": 3.0467857142857144e-05, "loss": 1.574, "step": 2734500 }, { "epoch": 13.64, "learning_rate": 3.0464285714285718e-05, "loss": 1.578, "step": 2735000 }, { "epoch": 13.64, "learning_rate": 3.0460714285714286e-05, "loss": 1.5757, "step": 2735500 }, { "epoch": 13.64, "learning_rate": 3.0457142857142856e-05, "loss": 1.5679, "step": 2736000 }, { "epoch": 13.65, "learning_rate": 3.045357142857143e-05, "loss": 1.57, "step": 2736500 }, { "epoch": 13.65, "learning_rate": 3.045e-05, "loss": 1.5611, "step": 2737000 }, { "epoch": 13.65, "learning_rate": 3.044642857142857e-05, "loss": 1.5809, "step": 2737500 }, { "epoch": 13.65, "learning_rate": 3.0442857142857147e-05, "loss": 1.5748, "step": 2738000 }, { "epoch": 13.66, "learning_rate": 3.0439285714285714e-05, "loss": 1.5717, "step": 2738500 }, { "epoch": 13.66, "learning_rate": 3.043571428571429e-05, "loss": 1.5618, "step": 2739000 }, { "epoch": 13.66, "learning_rate": 3.043214285714286e-05, "loss": 1.5698, "step": 2739500 }, { "epoch": 13.66, "learning_rate": 3.042857142857143e-05, "loss": 1.5674, "step": 2740000 }, { "epoch": 13.67, "learning_rate": 3.0425000000000004e-05, "loss": 1.5733, "step": 2740500 }, { "epoch": 13.67, "learning_rate": 3.0421428571428572e-05, "loss": 1.5691, "step": 2741000 }, { "epoch": 13.67, "learning_rate": 3.0417857142857143e-05, "loss": 1.5706, "step": 2741500 }, { "epoch": 13.68, "learning_rate": 3.0414285714285717e-05, "loss": 1.5732, "step": 2742000 }, { "epoch": 13.68, "learning_rate": 3.0410714285714288e-05, "loss": 1.5695, "step": 2742500 }, { "epoch": 13.68, "learning_rate": 3.0407142857142855e-05, "loss": 1.5691, "step": 2743000 }, { "epoch": 13.68, "learning_rate": 3.0403571428571433e-05, "loss": 1.5812, "step": 2743500 }, { "epoch": 13.69, "learning_rate": 3.04e-05, "loss": 1.5721, "step": 2744000 }, { "epoch": 13.69, "learning_rate": 3.039642857142857e-05, "loss": 1.5649, "step": 2744500 }, { "epoch": 13.69, "learning_rate": 3.0392857142857145e-05, "loss": 1.5661, "step": 2745000 }, { "epoch": 13.69, "learning_rate": 3.0389285714285713e-05, "loss": 1.5669, "step": 2745500 }, { "epoch": 13.7, "learning_rate": 3.038571428571429e-05, "loss": 1.5595, "step": 2746000 }, { "epoch": 13.7, "learning_rate": 3.0382142857142858e-05, "loss": 1.5564, "step": 2746500 }, { "epoch": 13.7, "learning_rate": 3.037857142857143e-05, "loss": 1.5742, "step": 2747000 }, { "epoch": 13.7, "learning_rate": 3.0375000000000003e-05, "loss": 1.5826, "step": 2747500 }, { "epoch": 13.71, "learning_rate": 3.0371428571428574e-05, "loss": 1.5678, "step": 2748000 }, { "epoch": 13.71, "learning_rate": 3.036785714285714e-05, "loss": 1.57, "step": 2748500 }, { "epoch": 13.71, "learning_rate": 3.0364285714285716e-05, "loss": 1.5529, "step": 2749000 }, { "epoch": 13.72, "learning_rate": 3.0360714285714287e-05, "loss": 1.5526, "step": 2749500 }, { "epoch": 13.72, "learning_rate": 3.0357142857142857e-05, "loss": 1.5758, "step": 2750000 }, { "epoch": 13.72, "learning_rate": 3.035357142857143e-05, "loss": 1.5631, "step": 2750500 }, { "epoch": 13.72, "learning_rate": 3.035e-05, "loss": 1.5625, "step": 2751000 }, { "epoch": 13.73, "learning_rate": 3.0346428571428577e-05, "loss": 1.5706, "step": 2751500 }, { "epoch": 13.73, "learning_rate": 3.0342857142857144e-05, "loss": 1.5719, "step": 2752000 }, { "epoch": 13.73, "learning_rate": 3.0339285714285715e-05, "loss": 1.57, "step": 2752500 }, { "epoch": 13.73, "learning_rate": 3.033571428571429e-05, "loss": 1.5593, "step": 2753000 }, { "epoch": 13.74, "learning_rate": 3.033214285714286e-05, "loss": 1.5749, "step": 2753500 }, { "epoch": 13.74, "learning_rate": 3.0328571428571428e-05, "loss": 1.5688, "step": 2754000 }, { "epoch": 13.74, "learning_rate": 3.0325000000000002e-05, "loss": 1.5842, "step": 2754500 }, { "epoch": 13.74, "learning_rate": 3.0321428571428573e-05, "loss": 1.5846, "step": 2755000 }, { "epoch": 13.75, "learning_rate": 3.0317857142857144e-05, "loss": 1.5703, "step": 2755500 }, { "epoch": 13.75, "learning_rate": 3.0314285714285718e-05, "loss": 1.5724, "step": 2756000 }, { "epoch": 13.75, "learning_rate": 3.0310714285714285e-05, "loss": 1.5697, "step": 2756500 }, { "epoch": 13.76, "learning_rate": 3.0307142857142856e-05, "loss": 1.5794, "step": 2757000 }, { "epoch": 13.76, "learning_rate": 3.030357142857143e-05, "loss": 1.5672, "step": 2757500 }, { "epoch": 13.76, "learning_rate": 3.03e-05, "loss": 1.5745, "step": 2758000 }, { "epoch": 13.76, "learning_rate": 3.0296428571428576e-05, "loss": 1.565, "step": 2758500 }, { "epoch": 13.77, "learning_rate": 3.0292857142857146e-05, "loss": 1.5642, "step": 2759000 }, { "epoch": 13.77, "learning_rate": 3.0289285714285714e-05, "loss": 1.5781, "step": 2759500 }, { "epoch": 13.77, "learning_rate": 3.0285714285714288e-05, "loss": 1.578, "step": 2760000 }, { "epoch": 13.77, "learning_rate": 3.028214285714286e-05, "loss": 1.5816, "step": 2760500 }, { "epoch": 13.78, "learning_rate": 3.027857142857143e-05, "loss": 1.5657, "step": 2761000 }, { "epoch": 13.78, "learning_rate": 3.0275000000000004e-05, "loss": 1.577, "step": 2761500 }, { "epoch": 13.78, "learning_rate": 3.027142857142857e-05, "loss": 1.5771, "step": 2762000 }, { "epoch": 13.78, "learning_rate": 3.0267857142857142e-05, "loss": 1.5742, "step": 2762500 }, { "epoch": 13.79, "learning_rate": 3.0264285714285717e-05, "loss": 1.5633, "step": 2763000 }, { "epoch": 13.79, "learning_rate": 3.0260714285714287e-05, "loss": 1.578, "step": 2763500 }, { "epoch": 13.79, "learning_rate": 3.0257142857142855e-05, "loss": 1.5657, "step": 2764000 }, { "epoch": 13.8, "learning_rate": 3.0253571428571433e-05, "loss": 1.5737, "step": 2764500 }, { "epoch": 13.8, "learning_rate": 3.025e-05, "loss": 1.5696, "step": 2765000 }, { "epoch": 13.8, "learning_rate": 3.0246428571428574e-05, "loss": 1.5698, "step": 2765500 }, { "epoch": 13.8, "learning_rate": 3.0242857142857145e-05, "loss": 1.575, "step": 2766000 }, { "epoch": 13.81, "learning_rate": 3.0239285714285716e-05, "loss": 1.5732, "step": 2766500 }, { "epoch": 13.81, "learning_rate": 3.023571428571429e-05, "loss": 1.5589, "step": 2767000 }, { "epoch": 13.81, "learning_rate": 3.0232142857142858e-05, "loss": 1.5763, "step": 2767500 }, { "epoch": 13.81, "learning_rate": 3.022857142857143e-05, "loss": 1.5664, "step": 2768000 }, { "epoch": 13.82, "learning_rate": 3.0225000000000003e-05, "loss": 1.5617, "step": 2768500 }, { "epoch": 13.82, "learning_rate": 3.0221428571428574e-05, "loss": 1.5721, "step": 2769000 }, { "epoch": 13.82, "learning_rate": 3.021785714285714e-05, "loss": 1.5708, "step": 2769500 }, { "epoch": 13.82, "learning_rate": 3.021428571428572e-05, "loss": 1.5654, "step": 2770000 }, { "epoch": 13.83, "learning_rate": 3.0210714285714286e-05, "loss": 1.5708, "step": 2770500 }, { "epoch": 13.83, "learning_rate": 3.0207142857142857e-05, "loss": 1.5731, "step": 2771000 }, { "epoch": 13.83, "learning_rate": 3.020357142857143e-05, "loss": 1.5775, "step": 2771500 }, { "epoch": 13.84, "learning_rate": 3.02e-05, "loss": 1.5638, "step": 2772000 }, { "epoch": 13.84, "learning_rate": 3.0196428571428576e-05, "loss": 1.5624, "step": 2772500 }, { "epoch": 13.84, "learning_rate": 3.0192857142857144e-05, "loss": 1.5759, "step": 2773000 }, { "epoch": 13.84, "learning_rate": 3.0189285714285715e-05, "loss": 1.5731, "step": 2773500 }, { "epoch": 13.85, "learning_rate": 3.018571428571429e-05, "loss": 1.5732, "step": 2774000 }, { "epoch": 13.85, "learning_rate": 3.018214285714286e-05, "loss": 1.5615, "step": 2774500 }, { "epoch": 13.85, "learning_rate": 3.0178571428571427e-05, "loss": 1.5703, "step": 2775000 }, { "epoch": 13.85, "learning_rate": 3.0175e-05, "loss": 1.571, "step": 2775500 }, { "epoch": 13.86, "learning_rate": 3.0171428571428572e-05, "loss": 1.5671, "step": 2776000 }, { "epoch": 13.86, "learning_rate": 3.0167857142857143e-05, "loss": 1.5712, "step": 2776500 }, { "epoch": 13.86, "learning_rate": 3.0164285714285718e-05, "loss": 1.5628, "step": 2777000 }, { "epoch": 13.86, "learning_rate": 3.0160714285714285e-05, "loss": 1.5703, "step": 2777500 }, { "epoch": 13.87, "learning_rate": 3.0157142857142856e-05, "loss": 1.5772, "step": 2778000 }, { "epoch": 13.87, "learning_rate": 3.015357142857143e-05, "loss": 1.5732, "step": 2778500 }, { "epoch": 13.87, "learning_rate": 3.015e-05, "loss": 1.5593, "step": 2779000 }, { "epoch": 13.87, "learning_rate": 3.0146428571428575e-05, "loss": 1.5707, "step": 2779500 }, { "epoch": 13.88, "learning_rate": 3.0142857142857146e-05, "loss": 1.5609, "step": 2780000 }, { "epoch": 13.88, "learning_rate": 3.0139285714285714e-05, "loss": 1.564, "step": 2780500 }, { "epoch": 13.88, "learning_rate": 3.0135714285714288e-05, "loss": 1.5748, "step": 2781000 }, { "epoch": 13.89, "learning_rate": 3.013214285714286e-05, "loss": 1.5801, "step": 2781500 }, { "epoch": 13.89, "learning_rate": 3.012857142857143e-05, "loss": 1.5638, "step": 2782000 }, { "epoch": 13.89, "learning_rate": 3.0125000000000004e-05, "loss": 1.5731, "step": 2782500 }, { "epoch": 13.89, "learning_rate": 3.012142857142857e-05, "loss": 1.5761, "step": 2783000 }, { "epoch": 13.9, "learning_rate": 3.0117857142857142e-05, "loss": 1.5737, "step": 2783500 }, { "epoch": 13.9, "learning_rate": 3.0114285714285716e-05, "loss": 1.5687, "step": 2784000 }, { "epoch": 13.9, "learning_rate": 3.0110714285714287e-05, "loss": 1.5745, "step": 2784500 }, { "epoch": 13.9, "learning_rate": 3.0107142857142855e-05, "loss": 1.5713, "step": 2785000 }, { "epoch": 13.91, "learning_rate": 3.0103571428571432e-05, "loss": 1.5715, "step": 2785500 }, { "epoch": 13.91, "learning_rate": 3.01e-05, "loss": 1.5785, "step": 2786000 }, { "epoch": 13.91, "learning_rate": 3.0096428571428574e-05, "loss": 1.5662, "step": 2786500 }, { "epoch": 13.91, "learning_rate": 3.0092857142857145e-05, "loss": 1.5566, "step": 2787000 }, { "epoch": 13.92, "learning_rate": 3.0089285714285716e-05, "loss": 1.5606, "step": 2787500 }, { "epoch": 13.92, "learning_rate": 3.008571428571429e-05, "loss": 1.5752, "step": 2788000 }, { "epoch": 13.92, "learning_rate": 3.0082142857142857e-05, "loss": 1.5676, "step": 2788500 }, { "epoch": 13.93, "learning_rate": 3.007857142857143e-05, "loss": 1.5561, "step": 2789000 }, { "epoch": 13.93, "learning_rate": 3.0075000000000003e-05, "loss": 1.5615, "step": 2789500 }, { "epoch": 13.93, "learning_rate": 3.0071428571428573e-05, "loss": 1.5743, "step": 2790000 }, { "epoch": 13.93, "learning_rate": 3.006785714285714e-05, "loss": 1.5636, "step": 2790500 }, { "epoch": 13.94, "learning_rate": 3.006428571428572e-05, "loss": 1.5696, "step": 2791000 }, { "epoch": 13.94, "learning_rate": 3.0060714285714286e-05, "loss": 1.5665, "step": 2791500 }, { "epoch": 13.94, "learning_rate": 3.0057142857142857e-05, "loss": 1.5703, "step": 2792000 }, { "epoch": 13.94, "learning_rate": 3.005357142857143e-05, "loss": 1.5711, "step": 2792500 }, { "epoch": 13.95, "learning_rate": 3.0050000000000002e-05, "loss": 1.5706, "step": 2793000 }, { "epoch": 13.95, "learning_rate": 3.0046428571428576e-05, "loss": 1.568, "step": 2793500 }, { "epoch": 13.95, "learning_rate": 3.0042857142857144e-05, "loss": 1.5656, "step": 2794000 }, { "epoch": 13.95, "learning_rate": 3.0039285714285715e-05, "loss": 1.5704, "step": 2794500 }, { "epoch": 13.96, "learning_rate": 3.003571428571429e-05, "loss": 1.5793, "step": 2795000 }, { "epoch": 13.96, "learning_rate": 3.003214285714286e-05, "loss": 1.5613, "step": 2795500 }, { "epoch": 13.96, "learning_rate": 3.0028571428571427e-05, "loss": 1.57, "step": 2796000 }, { "epoch": 13.97, "learning_rate": 3.0025000000000005e-05, "loss": 1.5637, "step": 2796500 }, { "epoch": 13.97, "learning_rate": 3.0021428571428572e-05, "loss": 1.5664, "step": 2797000 }, { "epoch": 13.97, "learning_rate": 3.0017857142857143e-05, "loss": 1.5704, "step": 2797500 }, { "epoch": 13.97, "learning_rate": 3.0014285714285717e-05, "loss": 1.5665, "step": 2798000 }, { "epoch": 13.98, "learning_rate": 3.0010714285714285e-05, "loss": 1.5563, "step": 2798500 }, { "epoch": 13.98, "learning_rate": 3.0007142857142856e-05, "loss": 1.5726, "step": 2799000 }, { "epoch": 13.98, "learning_rate": 3.000357142857143e-05, "loss": 1.5661, "step": 2799500 }, { "epoch": 13.98, "learning_rate": 3e-05, "loss": 1.5637, "step": 2800000 }, { "epoch": 13.99, "learning_rate": 2.9996428571428575e-05, "loss": 1.5676, "step": 2800500 }, { "epoch": 13.99, "learning_rate": 2.9992857142857146e-05, "loss": 1.5651, "step": 2801000 }, { "epoch": 13.99, "learning_rate": 2.9989285714285713e-05, "loss": 1.5668, "step": 2801500 }, { "epoch": 13.99, "learning_rate": 2.9985714285714288e-05, "loss": 1.5688, "step": 2802000 }, { "epoch": 14.0, "learning_rate": 2.998214285714286e-05, "loss": 1.5594, "step": 2802500 }, { "epoch": 14.0, "learning_rate": 2.997857142857143e-05, "loss": 1.5644, "step": 2803000 }, { "epoch": 14.0, "learning_rate": 2.9975000000000004e-05, "loss": 1.5507, "step": 2803500 }, { "epoch": 14.01, "learning_rate": 2.997142857142857e-05, "loss": 1.5625, "step": 2804000 }, { "epoch": 14.01, "learning_rate": 2.9967857142857142e-05, "loss": 1.556, "step": 2804500 }, { "epoch": 14.01, "learning_rate": 2.9964285714285716e-05, "loss": 1.5573, "step": 2805000 }, { "epoch": 14.01, "learning_rate": 2.9960714285714287e-05, "loss": 1.5605, "step": 2805500 }, { "epoch": 14.02, "learning_rate": 2.9957142857142854e-05, "loss": 1.5599, "step": 2806000 }, { "epoch": 14.02, "learning_rate": 2.9953571428571432e-05, "loss": 1.5692, "step": 2806500 }, { "epoch": 14.02, "learning_rate": 2.995e-05, "loss": 1.5619, "step": 2807000 }, { "epoch": 14.02, "learning_rate": 2.9946428571428574e-05, "loss": 1.5678, "step": 2807500 }, { "epoch": 14.03, "learning_rate": 2.9942857142857145e-05, "loss": 1.5621, "step": 2808000 }, { "epoch": 14.03, "learning_rate": 2.9939285714285716e-05, "loss": 1.5554, "step": 2808500 }, { "epoch": 14.03, "learning_rate": 2.993571428571429e-05, "loss": 1.5649, "step": 2809000 }, { "epoch": 14.03, "learning_rate": 2.9932142857142857e-05, "loss": 1.5567, "step": 2809500 }, { "epoch": 14.04, "learning_rate": 2.9928571428571428e-05, "loss": 1.5686, "step": 2810000 }, { "epoch": 14.04, "learning_rate": 2.9925000000000002e-05, "loss": 1.5652, "step": 2810500 }, { "epoch": 14.04, "learning_rate": 2.9921428571428573e-05, "loss": 1.5473, "step": 2811000 }, { "epoch": 14.05, "learning_rate": 2.991785714285714e-05, "loss": 1.5575, "step": 2811500 }, { "epoch": 14.05, "learning_rate": 2.9914285714285718e-05, "loss": 1.548, "step": 2812000 }, { "epoch": 14.05, "learning_rate": 2.9910714285714286e-05, "loss": 1.5544, "step": 2812500 }, { "epoch": 14.05, "learning_rate": 2.990714285714286e-05, "loss": 1.5584, "step": 2813000 }, { "epoch": 14.06, "learning_rate": 2.990357142857143e-05, "loss": 1.5574, "step": 2813500 }, { "epoch": 14.06, "learning_rate": 2.9900000000000002e-05, "loss": 1.5466, "step": 2814000 }, { "epoch": 14.06, "learning_rate": 2.9896428571428576e-05, "loss": 1.5573, "step": 2814500 }, { "epoch": 14.06, "learning_rate": 2.9892857142857143e-05, "loss": 1.5549, "step": 2815000 }, { "epoch": 14.07, "learning_rate": 2.9889285714285714e-05, "loss": 1.5633, "step": 2815500 }, { "epoch": 14.07, "learning_rate": 2.988571428571429e-05, "loss": 1.5719, "step": 2816000 }, { "epoch": 14.07, "learning_rate": 2.988214285714286e-05, "loss": 1.5606, "step": 2816500 }, { "epoch": 14.07, "learning_rate": 2.9878571428571427e-05, "loss": 1.5637, "step": 2817000 }, { "epoch": 14.08, "learning_rate": 2.9875000000000004e-05, "loss": 1.5717, "step": 2817500 }, { "epoch": 14.08, "learning_rate": 2.9871428571428572e-05, "loss": 1.5586, "step": 2818000 }, { "epoch": 14.08, "learning_rate": 2.9867857142857143e-05, "loss": 1.5593, "step": 2818500 }, { "epoch": 14.09, "learning_rate": 2.9864285714285717e-05, "loss": 1.5611, "step": 2819000 }, { "epoch": 14.09, "learning_rate": 2.9860714285714288e-05, "loss": 1.5569, "step": 2819500 }, { "epoch": 14.09, "learning_rate": 2.9857142857142862e-05, "loss": 1.5632, "step": 2820000 }, { "epoch": 14.09, "learning_rate": 2.985357142857143e-05, "loss": 1.5623, "step": 2820500 }, { "epoch": 14.1, "learning_rate": 2.985e-05, "loss": 1.5544, "step": 2821000 }, { "epoch": 14.1, "learning_rate": 2.9846428571428575e-05, "loss": 1.5571, "step": 2821500 }, { "epoch": 14.1, "learning_rate": 2.9842857142857146e-05, "loss": 1.5539, "step": 2822000 }, { "epoch": 14.1, "learning_rate": 2.9839285714285713e-05, "loss": 1.5521, "step": 2822500 }, { "epoch": 14.11, "learning_rate": 2.983571428571429e-05, "loss": 1.5651, "step": 2823000 }, { "epoch": 14.11, "learning_rate": 2.9832142857142858e-05, "loss": 1.5562, "step": 2823500 }, { "epoch": 14.11, "learning_rate": 2.982857142857143e-05, "loss": 1.569, "step": 2824000 }, { "epoch": 14.11, "learning_rate": 2.9825000000000003e-05, "loss": 1.5552, "step": 2824500 }, { "epoch": 14.12, "learning_rate": 2.982142857142857e-05, "loss": 1.553, "step": 2825000 }, { "epoch": 14.12, "learning_rate": 2.981785714285714e-05, "loss": 1.5697, "step": 2825500 }, { "epoch": 14.12, "learning_rate": 2.9814285714285716e-05, "loss": 1.5625, "step": 2826000 }, { "epoch": 14.13, "learning_rate": 2.9810714285714287e-05, "loss": 1.5574, "step": 2826500 }, { "epoch": 14.13, "learning_rate": 2.980714285714286e-05, "loss": 1.5591, "step": 2827000 }, { "epoch": 14.13, "learning_rate": 2.9803571428571432e-05, "loss": 1.5576, "step": 2827500 }, { "epoch": 14.13, "learning_rate": 2.98e-05, "loss": 1.5631, "step": 2828000 }, { "epoch": 14.14, "learning_rate": 2.9796428571428574e-05, "loss": 1.5602, "step": 2828500 }, { "epoch": 14.14, "learning_rate": 2.9792857142857144e-05, "loss": 1.5603, "step": 2829000 }, { "epoch": 14.14, "learning_rate": 2.9789285714285715e-05, "loss": 1.5667, "step": 2829500 }, { "epoch": 14.14, "learning_rate": 2.978571428571429e-05, "loss": 1.5531, "step": 2830000 }, { "epoch": 14.15, "learning_rate": 2.9782142857142857e-05, "loss": 1.5546, "step": 2830500 }, { "epoch": 14.15, "learning_rate": 2.9778571428571428e-05, "loss": 1.5634, "step": 2831000 }, { "epoch": 14.15, "learning_rate": 2.9775000000000002e-05, "loss": 1.5583, "step": 2831500 }, { "epoch": 14.15, "learning_rate": 2.9771428571428573e-05, "loss": 1.557, "step": 2832000 }, { "epoch": 14.16, "learning_rate": 2.976785714285714e-05, "loss": 1.5603, "step": 2832500 }, { "epoch": 14.16, "learning_rate": 2.9764285714285718e-05, "loss": 1.5758, "step": 2833000 }, { "epoch": 14.16, "learning_rate": 2.9760714285714285e-05, "loss": 1.5611, "step": 2833500 }, { "epoch": 14.17, "learning_rate": 2.975714285714286e-05, "loss": 1.5645, "step": 2834000 }, { "epoch": 14.17, "learning_rate": 2.975357142857143e-05, "loss": 1.5624, "step": 2834500 }, { "epoch": 14.17, "learning_rate": 2.975e-05, "loss": 1.5627, "step": 2835000 }, { "epoch": 14.17, "learning_rate": 2.9746428571428576e-05, "loss": 1.5542, "step": 2835500 }, { "epoch": 14.18, "learning_rate": 2.9742857142857143e-05, "loss": 1.5519, "step": 2836000 }, { "epoch": 14.18, "learning_rate": 2.9739285714285714e-05, "loss": 1.5585, "step": 2836500 }, { "epoch": 14.18, "learning_rate": 2.9735714285714288e-05, "loss": 1.556, "step": 2837000 }, { "epoch": 14.18, "learning_rate": 2.973214285714286e-05, "loss": 1.5537, "step": 2837500 }, { "epoch": 14.19, "learning_rate": 2.9728571428571427e-05, "loss": 1.5629, "step": 2838000 }, { "epoch": 14.19, "learning_rate": 2.9725000000000004e-05, "loss": 1.5529, "step": 2838500 }, { "epoch": 14.19, "learning_rate": 2.9721428571428572e-05, "loss": 1.562, "step": 2839000 }, { "epoch": 14.19, "learning_rate": 2.9717857142857143e-05, "loss": 1.5641, "step": 2839500 }, { "epoch": 14.2, "learning_rate": 2.9714285714285717e-05, "loss": 1.5608, "step": 2840000 }, { "epoch": 14.2, "learning_rate": 2.9710714285714288e-05, "loss": 1.5593, "step": 2840500 }, { "epoch": 14.2, "learning_rate": 2.9707142857142862e-05, "loss": 1.564, "step": 2841000 }, { "epoch": 14.21, "learning_rate": 2.970357142857143e-05, "loss": 1.5534, "step": 2841500 }, { "epoch": 14.21, "learning_rate": 2.97e-05, "loss": 1.5565, "step": 2842000 }, { "epoch": 14.21, "learning_rate": 2.9696428571428574e-05, "loss": 1.5602, "step": 2842500 }, { "epoch": 14.21, "learning_rate": 2.9692857142857145e-05, "loss": 1.5521, "step": 2843000 }, { "epoch": 14.22, "learning_rate": 2.9689285714285713e-05, "loss": 1.5545, "step": 2843500 }, { "epoch": 14.22, "learning_rate": 2.968571428571429e-05, "loss": 1.5553, "step": 2844000 }, { "epoch": 14.22, "learning_rate": 2.9682142857142858e-05, "loss": 1.5563, "step": 2844500 }, { "epoch": 14.22, "learning_rate": 2.967857142857143e-05, "loss": 1.5485, "step": 2845000 }, { "epoch": 14.23, "learning_rate": 2.9675000000000003e-05, "loss": 1.5627, "step": 2845500 }, { "epoch": 14.23, "learning_rate": 2.9671428571428574e-05, "loss": 1.5596, "step": 2846000 }, { "epoch": 14.23, "learning_rate": 2.966785714285714e-05, "loss": 1.5633, "step": 2846500 }, { "epoch": 14.23, "learning_rate": 2.9664285714285716e-05, "loss": 1.5564, "step": 2847000 }, { "epoch": 14.24, "learning_rate": 2.9660714285714286e-05, "loss": 1.5593, "step": 2847500 }, { "epoch": 14.24, "learning_rate": 2.965714285714286e-05, "loss": 1.5522, "step": 2848000 }, { "epoch": 14.24, "learning_rate": 2.965357142857143e-05, "loss": 1.5482, "step": 2848500 }, { "epoch": 14.24, "learning_rate": 2.965e-05, "loss": 1.5617, "step": 2849000 }, { "epoch": 14.25, "learning_rate": 2.9646428571428577e-05, "loss": 1.5732, "step": 2849500 }, { "epoch": 14.25, "learning_rate": 2.9642857142857144e-05, "loss": 1.5489, "step": 2850000 }, { "epoch": 14.25, "learning_rate": 2.9639285714285715e-05, "loss": 1.5486, "step": 2850500 }, { "epoch": 14.26, "learning_rate": 2.963571428571429e-05, "loss": 1.5552, "step": 2851000 }, { "epoch": 14.26, "learning_rate": 2.9632142857142857e-05, "loss": 1.576, "step": 2851500 }, { "epoch": 14.26, "learning_rate": 2.9628571428571428e-05, "loss": 1.5654, "step": 2852000 }, { "epoch": 14.26, "learning_rate": 2.9625000000000002e-05, "loss": 1.5521, "step": 2852500 }, { "epoch": 14.27, "learning_rate": 2.9621428571428573e-05, "loss": 1.5687, "step": 2853000 }, { "epoch": 14.27, "learning_rate": 2.961785714285714e-05, "loss": 1.5635, "step": 2853500 }, { "epoch": 14.27, "learning_rate": 2.9614285714285718e-05, "loss": 1.5613, "step": 2854000 }, { "epoch": 14.27, "learning_rate": 2.9610714285714285e-05, "loss": 1.5591, "step": 2854500 }, { "epoch": 14.28, "learning_rate": 2.960714285714286e-05, "loss": 1.5526, "step": 2855000 }, { "epoch": 14.28, "learning_rate": 2.960357142857143e-05, "loss": 1.5551, "step": 2855500 }, { "epoch": 14.28, "learning_rate": 2.96e-05, "loss": 1.5562, "step": 2856000 }, { "epoch": 14.28, "learning_rate": 2.9596428571428575e-05, "loss": 1.5532, "step": 2856500 }, { "epoch": 14.29, "learning_rate": 2.9592857142857143e-05, "loss": 1.5674, "step": 2857000 }, { "epoch": 14.29, "learning_rate": 2.9589285714285714e-05, "loss": 1.56, "step": 2857500 }, { "epoch": 14.29, "learning_rate": 2.9585714285714288e-05, "loss": 1.5538, "step": 2858000 }, { "epoch": 14.3, "learning_rate": 2.958214285714286e-05, "loss": 1.5518, "step": 2858500 }, { "epoch": 14.3, "learning_rate": 2.9578571428571426e-05, "loss": 1.5655, "step": 2859000 }, { "epoch": 14.3, "learning_rate": 2.9575000000000004e-05, "loss": 1.5584, "step": 2859500 }, { "epoch": 14.3, "learning_rate": 2.957142857142857e-05, "loss": 1.5673, "step": 2860000 }, { "epoch": 14.31, "learning_rate": 2.9567857142857142e-05, "loss": 1.5551, "step": 2860500 }, { "epoch": 14.31, "learning_rate": 2.9564285714285717e-05, "loss": 1.5589, "step": 2861000 }, { "epoch": 14.31, "learning_rate": 2.9560714285714287e-05, "loss": 1.5566, "step": 2861500 }, { "epoch": 14.31, "learning_rate": 2.955714285714286e-05, "loss": 1.5578, "step": 2862000 }, { "epoch": 14.32, "learning_rate": 2.955357142857143e-05, "loss": 1.5618, "step": 2862500 }, { "epoch": 14.32, "learning_rate": 2.955e-05, "loss": 1.5629, "step": 2863000 }, { "epoch": 14.32, "learning_rate": 2.9546428571428574e-05, "loss": 1.5524, "step": 2863500 }, { "epoch": 14.32, "learning_rate": 2.9542857142857145e-05, "loss": 1.5603, "step": 2864000 }, { "epoch": 14.33, "learning_rate": 2.9539285714285713e-05, "loss": 1.5526, "step": 2864500 }, { "epoch": 14.33, "learning_rate": 2.953571428571429e-05, "loss": 1.551, "step": 2865000 }, { "epoch": 14.33, "learning_rate": 2.9532142857142858e-05, "loss": 1.5666, "step": 2865500 }, { "epoch": 14.34, "learning_rate": 2.952857142857143e-05, "loss": 1.5699, "step": 2866000 }, { "epoch": 14.34, "learning_rate": 2.9525000000000003e-05, "loss": 1.5596, "step": 2866500 }, { "epoch": 14.34, "learning_rate": 2.9521428571428574e-05, "loss": 1.5602, "step": 2867000 }, { "epoch": 14.34, "learning_rate": 2.951785714285714e-05, "loss": 1.5519, "step": 2867500 }, { "epoch": 14.35, "learning_rate": 2.9514285714285715e-05, "loss": 1.5621, "step": 2868000 }, { "epoch": 14.35, "learning_rate": 2.9510714285714286e-05, "loss": 1.5542, "step": 2868500 }, { "epoch": 14.35, "learning_rate": 2.950714285714286e-05, "loss": 1.5603, "step": 2869000 }, { "epoch": 14.35, "learning_rate": 2.950357142857143e-05, "loss": 1.5569, "step": 2869500 }, { "epoch": 14.36, "learning_rate": 2.95e-05, "loss": 1.5585, "step": 2870000 }, { "epoch": 14.36, "learning_rate": 2.9496428571428576e-05, "loss": 1.557, "step": 2870500 }, { "epoch": 14.36, "learning_rate": 2.9492857142857144e-05, "loss": 1.5447, "step": 2871000 }, { "epoch": 14.36, "learning_rate": 2.9489285714285715e-05, "loss": 1.5637, "step": 2871500 }, { "epoch": 14.37, "learning_rate": 2.948571428571429e-05, "loss": 1.5539, "step": 2872000 }, { "epoch": 14.37, "learning_rate": 2.948214285714286e-05, "loss": 1.5669, "step": 2872500 }, { "epoch": 14.37, "learning_rate": 2.9478571428571427e-05, "loss": 1.5555, "step": 2873000 }, { "epoch": 14.38, "learning_rate": 2.9475e-05, "loss": 1.5525, "step": 2873500 }, { "epoch": 14.38, "learning_rate": 2.9471428571428572e-05, "loss": 1.5607, "step": 2874000 }, { "epoch": 14.38, "learning_rate": 2.946785714285714e-05, "loss": 1.5551, "step": 2874500 }, { "epoch": 14.38, "learning_rate": 2.9464285714285718e-05, "loss": 1.5494, "step": 2875000 }, { "epoch": 14.39, "learning_rate": 2.9460714285714285e-05, "loss": 1.5591, "step": 2875500 }, { "epoch": 14.39, "learning_rate": 2.9457142857142863e-05, "loss": 1.5615, "step": 2876000 }, { "epoch": 14.39, "learning_rate": 2.945357142857143e-05, "loss": 1.5571, "step": 2876500 }, { "epoch": 14.39, "learning_rate": 2.945e-05, "loss": 1.5602, "step": 2877000 }, { "epoch": 14.4, "learning_rate": 2.9446428571428575e-05, "loss": 1.5546, "step": 2877500 }, { "epoch": 14.4, "learning_rate": 2.9442857142857143e-05, "loss": 1.5624, "step": 2878000 }, { "epoch": 14.4, "learning_rate": 2.9439285714285714e-05, "loss": 1.5526, "step": 2878500 }, { "epoch": 14.4, "learning_rate": 2.9435714285714288e-05, "loss": 1.5615, "step": 2879000 }, { "epoch": 14.41, "learning_rate": 2.943214285714286e-05, "loss": 1.5608, "step": 2879500 }, { "epoch": 14.41, "learning_rate": 2.9428571428571426e-05, "loss": 1.5661, "step": 2880000 }, { "epoch": 14.41, "learning_rate": 2.9425000000000004e-05, "loss": 1.5523, "step": 2880500 }, { "epoch": 14.42, "learning_rate": 2.942142857142857e-05, "loss": 1.5615, "step": 2881000 }, { "epoch": 14.42, "learning_rate": 2.9417857142857145e-05, "loss": 1.551, "step": 2881500 }, { "epoch": 14.42, "learning_rate": 2.9414285714285716e-05, "loss": 1.5533, "step": 2882000 }, { "epoch": 14.42, "learning_rate": 2.9410714285714287e-05, "loss": 1.5586, "step": 2882500 }, { "epoch": 14.43, "learning_rate": 2.940714285714286e-05, "loss": 1.5609, "step": 2883000 }, { "epoch": 14.43, "learning_rate": 2.940357142857143e-05, "loss": 1.5603, "step": 2883500 }, { "epoch": 14.43, "learning_rate": 2.94e-05, "loss": 1.5506, "step": 2884000 }, { "epoch": 14.43, "learning_rate": 2.9396428571428574e-05, "loss": 1.5561, "step": 2884500 }, { "epoch": 14.44, "learning_rate": 2.9392857142857145e-05, "loss": 1.5662, "step": 2885000 }, { "epoch": 14.44, "learning_rate": 2.9389285714285712e-05, "loss": 1.5568, "step": 2885500 }, { "epoch": 14.44, "learning_rate": 2.938571428571429e-05, "loss": 1.5531, "step": 2886000 }, { "epoch": 14.44, "learning_rate": 2.9382142857142857e-05, "loss": 1.5452, "step": 2886500 }, { "epoch": 14.45, "learning_rate": 2.9378571428571428e-05, "loss": 1.5444, "step": 2887000 }, { "epoch": 14.45, "learning_rate": 2.9375000000000003e-05, "loss": 1.5591, "step": 2887500 }, { "epoch": 14.45, "learning_rate": 2.9371428571428573e-05, "loss": 1.553, "step": 2888000 }, { "epoch": 14.46, "learning_rate": 2.9367857142857148e-05, "loss": 1.5569, "step": 2888500 }, { "epoch": 14.46, "learning_rate": 2.9364285714285715e-05, "loss": 1.5514, "step": 2889000 }, { "epoch": 14.46, "learning_rate": 2.9360714285714286e-05, "loss": 1.5516, "step": 2889500 }, { "epoch": 14.46, "learning_rate": 2.935714285714286e-05, "loss": 1.5632, "step": 2890000 }, { "epoch": 14.47, "learning_rate": 2.935357142857143e-05, "loss": 1.549, "step": 2890500 }, { "epoch": 14.47, "learning_rate": 2.935e-05, "loss": 1.5527, "step": 2891000 }, { "epoch": 14.47, "learning_rate": 2.9346428571428576e-05, "loss": 1.5478, "step": 2891500 }, { "epoch": 14.47, "learning_rate": 2.9342857142857144e-05, "loss": 1.5519, "step": 2892000 }, { "epoch": 14.48, "learning_rate": 2.9339285714285714e-05, "loss": 1.5628, "step": 2892500 }, { "epoch": 14.48, "learning_rate": 2.933571428571429e-05, "loss": 1.5634, "step": 2893000 }, { "epoch": 14.48, "learning_rate": 2.933214285714286e-05, "loss": 1.5491, "step": 2893500 }, { "epoch": 14.48, "learning_rate": 2.9328571428571427e-05, "loss": 1.5505, "step": 2894000 }, { "epoch": 14.49, "learning_rate": 2.9325e-05, "loss": 1.5551, "step": 2894500 }, { "epoch": 14.49, "learning_rate": 2.9321428571428572e-05, "loss": 1.5497, "step": 2895000 }, { "epoch": 14.49, "learning_rate": 2.9317857142857146e-05, "loss": 1.5592, "step": 2895500 }, { "epoch": 14.5, "learning_rate": 2.9314285714285717e-05, "loss": 1.545, "step": 2896000 }, { "epoch": 14.5, "learning_rate": 2.9310714285714285e-05, "loss": 1.5593, "step": 2896500 }, { "epoch": 14.5, "learning_rate": 2.9307142857142862e-05, "loss": 1.5495, "step": 2897000 }, { "epoch": 14.5, "learning_rate": 2.930357142857143e-05, "loss": 1.5446, "step": 2897500 }, { "epoch": 14.51, "learning_rate": 2.93e-05, "loss": 1.5537, "step": 2898000 }, { "epoch": 14.51, "learning_rate": 2.9296428571428575e-05, "loss": 1.559, "step": 2898500 }, { "epoch": 14.51, "learning_rate": 2.9292857142857146e-05, "loss": 1.5573, "step": 2899000 }, { "epoch": 14.51, "learning_rate": 2.9289285714285713e-05, "loss": 1.5495, "step": 2899500 }, { "epoch": 14.52, "learning_rate": 2.9285714285714288e-05, "loss": 1.5588, "step": 2900000 }, { "epoch": 14.52, "learning_rate": 2.928214285714286e-05, "loss": 1.5583, "step": 2900500 }, { "epoch": 14.52, "learning_rate": 2.9278571428571426e-05, "loss": 1.5656, "step": 2901000 }, { "epoch": 14.52, "learning_rate": 2.9275000000000003e-05, "loss": 1.5654, "step": 2901500 }, { "epoch": 14.53, "learning_rate": 2.927142857142857e-05, "loss": 1.5586, "step": 2902000 }, { "epoch": 14.53, "learning_rate": 2.9267857142857145e-05, "loss": 1.5529, "step": 2902500 }, { "epoch": 14.53, "learning_rate": 2.9264285714285716e-05, "loss": 1.5564, "step": 2903000 }, { "epoch": 14.54, "learning_rate": 2.9260714285714287e-05, "loss": 1.5522, "step": 2903500 }, { "epoch": 14.54, "learning_rate": 2.925714285714286e-05, "loss": 1.5517, "step": 2904000 }, { "epoch": 14.54, "learning_rate": 2.925357142857143e-05, "loss": 1.5492, "step": 2904500 }, { "epoch": 14.54, "learning_rate": 2.925e-05, "loss": 1.5556, "step": 2905000 }, { "epoch": 14.55, "learning_rate": 2.9246428571428574e-05, "loss": 1.5548, "step": 2905500 }, { "epoch": 14.55, "learning_rate": 2.9242857142857145e-05, "loss": 1.5483, "step": 2906000 }, { "epoch": 14.55, "learning_rate": 2.9239285714285712e-05, "loss": 1.5474, "step": 2906500 }, { "epoch": 14.55, "learning_rate": 2.923571428571429e-05, "loss": 1.5528, "step": 2907000 }, { "epoch": 14.56, "learning_rate": 2.9232142857142857e-05, "loss": 1.5628, "step": 2907500 }, { "epoch": 14.56, "learning_rate": 2.9228571428571428e-05, "loss": 1.5636, "step": 2908000 }, { "epoch": 14.56, "learning_rate": 2.9225000000000002e-05, "loss": 1.5487, "step": 2908500 }, { "epoch": 14.56, "learning_rate": 2.9221428571428573e-05, "loss": 1.5652, "step": 2909000 }, { "epoch": 14.57, "learning_rate": 2.9217857142857147e-05, "loss": 1.5655, "step": 2909500 }, { "epoch": 14.57, "learning_rate": 2.9214285714285715e-05, "loss": 1.5549, "step": 2910000 }, { "epoch": 14.57, "learning_rate": 2.9210714285714286e-05, "loss": 1.5552, "step": 2910500 }, { "epoch": 14.58, "learning_rate": 2.920714285714286e-05, "loss": 1.555, "step": 2911000 }, { "epoch": 14.58, "learning_rate": 2.920357142857143e-05, "loss": 1.5508, "step": 2911500 }, { "epoch": 14.58, "learning_rate": 2.9199999999999998e-05, "loss": 1.5533, "step": 2912000 }, { "epoch": 14.58, "learning_rate": 2.9196428571428576e-05, "loss": 1.5589, "step": 2912500 }, { "epoch": 14.59, "learning_rate": 2.9192857142857143e-05, "loss": 1.5377, "step": 2913000 }, { "epoch": 14.59, "learning_rate": 2.9189285714285714e-05, "loss": 1.5542, "step": 2913500 }, { "epoch": 14.59, "learning_rate": 2.918571428571429e-05, "loss": 1.5504, "step": 2914000 }, { "epoch": 14.59, "learning_rate": 2.918214285714286e-05, "loss": 1.553, "step": 2914500 }, { "epoch": 14.6, "learning_rate": 2.9178571428571427e-05, "loss": 1.5603, "step": 2915000 }, { "epoch": 14.6, "learning_rate": 2.9175e-05, "loss": 1.5547, "step": 2915500 }, { "epoch": 14.6, "learning_rate": 2.9171428571428572e-05, "loss": 1.5392, "step": 2916000 }, { "epoch": 14.6, "learning_rate": 2.9167857142857146e-05, "loss": 1.5462, "step": 2916500 }, { "epoch": 14.61, "learning_rate": 2.9164285714285717e-05, "loss": 1.5602, "step": 2917000 }, { "epoch": 14.61, "learning_rate": 2.9160714285714284e-05, "loss": 1.5609, "step": 2917500 }, { "epoch": 14.61, "learning_rate": 2.9157142857142862e-05, "loss": 1.5621, "step": 2918000 }, { "epoch": 14.62, "learning_rate": 2.915357142857143e-05, "loss": 1.5555, "step": 2918500 }, { "epoch": 14.62, "learning_rate": 2.915e-05, "loss": 1.5513, "step": 2919000 }, { "epoch": 14.62, "learning_rate": 2.9146428571428575e-05, "loss": 1.5523, "step": 2919500 }, { "epoch": 14.62, "learning_rate": 2.9142857142857146e-05, "loss": 1.5644, "step": 2920000 }, { "epoch": 14.63, "learning_rate": 2.9139285714285713e-05, "loss": 1.5499, "step": 2920500 }, { "epoch": 14.63, "learning_rate": 2.9135714285714287e-05, "loss": 1.5566, "step": 2921000 }, { "epoch": 14.63, "learning_rate": 2.9132142857142858e-05, "loss": 1.5574, "step": 2921500 }, { "epoch": 14.63, "learning_rate": 2.912857142857143e-05, "loss": 1.555, "step": 2922000 }, { "epoch": 14.64, "learning_rate": 2.9125000000000003e-05, "loss": 1.5508, "step": 2922500 }, { "epoch": 14.64, "learning_rate": 2.912142857142857e-05, "loss": 1.5627, "step": 2923000 }, { "epoch": 14.64, "learning_rate": 2.911785714285715e-05, "loss": 1.5475, "step": 2923500 }, { "epoch": 14.64, "learning_rate": 2.9114285714285716e-05, "loss": 1.5497, "step": 2924000 }, { "epoch": 14.65, "learning_rate": 2.9110714285714287e-05, "loss": 1.5427, "step": 2924500 }, { "epoch": 14.65, "learning_rate": 2.910714285714286e-05, "loss": 1.5369, "step": 2925000 }, { "epoch": 14.65, "learning_rate": 2.910357142857143e-05, "loss": 1.5493, "step": 2925500 }, { "epoch": 14.65, "learning_rate": 2.91e-05, "loss": 1.5536, "step": 2926000 }, { "epoch": 14.66, "learning_rate": 2.9096428571428573e-05, "loss": 1.5492, "step": 2926500 }, { "epoch": 14.66, "learning_rate": 2.9092857142857144e-05, "loss": 1.5521, "step": 2927000 }, { "epoch": 14.66, "learning_rate": 2.9089285714285712e-05, "loss": 1.5508, "step": 2927500 }, { "epoch": 14.67, "learning_rate": 2.908571428571429e-05, "loss": 1.5522, "step": 2928000 }, { "epoch": 14.67, "learning_rate": 2.9082142857142857e-05, "loss": 1.5488, "step": 2928500 }, { "epoch": 14.67, "learning_rate": 2.9078571428571428e-05, "loss": 1.5551, "step": 2929000 }, { "epoch": 14.67, "learning_rate": 2.9075000000000002e-05, "loss": 1.5547, "step": 2929500 }, { "epoch": 14.68, "learning_rate": 2.9071428571428573e-05, "loss": 1.5552, "step": 2930000 }, { "epoch": 14.68, "learning_rate": 2.9067857142857147e-05, "loss": 1.5626, "step": 2930500 }, { "epoch": 14.68, "learning_rate": 2.9064285714285715e-05, "loss": 1.5459, "step": 2931000 }, { "epoch": 14.68, "learning_rate": 2.9060714285714285e-05, "loss": 1.5569, "step": 2931500 }, { "epoch": 14.69, "learning_rate": 2.905714285714286e-05, "loss": 1.5507, "step": 2932000 }, { "epoch": 14.69, "learning_rate": 2.905357142857143e-05, "loss": 1.5437, "step": 2932500 }, { "epoch": 14.69, "learning_rate": 2.9049999999999998e-05, "loss": 1.5539, "step": 2933000 }, { "epoch": 14.69, "learning_rate": 2.9046428571428576e-05, "loss": 1.5569, "step": 2933500 }, { "epoch": 14.7, "learning_rate": 2.9042857142857143e-05, "loss": 1.5531, "step": 2934000 }, { "epoch": 14.7, "learning_rate": 2.9039285714285714e-05, "loss": 1.5586, "step": 2934500 }, { "epoch": 14.7, "learning_rate": 2.9035714285714288e-05, "loss": 1.5656, "step": 2935000 }, { "epoch": 14.71, "learning_rate": 2.903214285714286e-05, "loss": 1.5479, "step": 2935500 }, { "epoch": 14.71, "learning_rate": 2.9028571428571427e-05, "loss": 1.554, "step": 2936000 }, { "epoch": 14.71, "learning_rate": 2.9025e-05, "loss": 1.5619, "step": 2936500 }, { "epoch": 14.71, "learning_rate": 2.902142857142857e-05, "loss": 1.5565, "step": 2937000 }, { "epoch": 14.72, "learning_rate": 2.9017857142857146e-05, "loss": 1.5634, "step": 2937500 }, { "epoch": 14.72, "learning_rate": 2.9014285714285717e-05, "loss": 1.5451, "step": 2938000 }, { "epoch": 14.72, "learning_rate": 2.9010714285714284e-05, "loss": 1.5531, "step": 2938500 }, { "epoch": 14.72, "learning_rate": 2.9007142857142862e-05, "loss": 1.5526, "step": 2939000 }, { "epoch": 14.73, "learning_rate": 2.900357142857143e-05, "loss": 1.5584, "step": 2939500 }, { "epoch": 14.73, "learning_rate": 2.9e-05, "loss": 1.5559, "step": 2940000 }, { "epoch": 14.73, "learning_rate": 2.8996428571428574e-05, "loss": 1.5493, "step": 2940500 }, { "epoch": 14.73, "learning_rate": 2.8992857142857145e-05, "loss": 1.557, "step": 2941000 }, { "epoch": 14.74, "learning_rate": 2.8989285714285713e-05, "loss": 1.551, "step": 2941500 }, { "epoch": 14.74, "learning_rate": 2.8985714285714287e-05, "loss": 1.5602, "step": 2942000 }, { "epoch": 14.74, "learning_rate": 2.8982142857142858e-05, "loss": 1.5509, "step": 2942500 }, { "epoch": 14.75, "learning_rate": 2.8978571428571432e-05, "loss": 1.5588, "step": 2943000 }, { "epoch": 14.75, "learning_rate": 2.8975000000000003e-05, "loss": 1.5545, "step": 2943500 }, { "epoch": 14.75, "learning_rate": 2.897142857142857e-05, "loss": 1.5456, "step": 2944000 }, { "epoch": 14.75, "learning_rate": 2.8967857142857148e-05, "loss": 1.5487, "step": 2944500 }, { "epoch": 14.76, "learning_rate": 2.8964285714285716e-05, "loss": 1.5517, "step": 2945000 }, { "epoch": 14.76, "learning_rate": 2.8960714285714286e-05, "loss": 1.5626, "step": 2945500 }, { "epoch": 14.76, "learning_rate": 2.895714285714286e-05, "loss": 1.5494, "step": 2946000 }, { "epoch": 14.76, "learning_rate": 2.895357142857143e-05, "loss": 1.5414, "step": 2946500 }, { "epoch": 14.77, "learning_rate": 2.895e-05, "loss": 1.5603, "step": 2947000 }, { "epoch": 14.77, "learning_rate": 2.8946428571428573e-05, "loss": 1.5474, "step": 2947500 }, { "epoch": 14.77, "learning_rate": 2.8942857142857144e-05, "loss": 1.548, "step": 2948000 }, { "epoch": 14.77, "learning_rate": 2.8939285714285715e-05, "loss": 1.5475, "step": 2948500 }, { "epoch": 14.78, "learning_rate": 2.893571428571429e-05, "loss": 1.5456, "step": 2949000 }, { "epoch": 14.78, "learning_rate": 2.8932142857142857e-05, "loss": 1.5457, "step": 2949500 }, { "epoch": 14.78, "learning_rate": 2.8928571428571434e-05, "loss": 1.5512, "step": 2950000 }, { "epoch": 14.79, "learning_rate": 2.8925000000000002e-05, "loss": 1.5509, "step": 2950500 }, { "epoch": 14.79, "learning_rate": 2.8921428571428573e-05, "loss": 1.5678, "step": 2951000 }, { "epoch": 14.79, "learning_rate": 2.8917857142857147e-05, "loss": 1.5441, "step": 2951500 }, { "epoch": 14.79, "learning_rate": 2.8914285714285714e-05, "loss": 1.5543, "step": 2952000 }, { "epoch": 14.8, "learning_rate": 2.8910714285714285e-05, "loss": 1.5435, "step": 2952500 }, { "epoch": 14.8, "learning_rate": 2.890714285714286e-05, "loss": 1.5408, "step": 2953000 }, { "epoch": 14.8, "learning_rate": 2.890357142857143e-05, "loss": 1.567, "step": 2953500 }, { "epoch": 14.8, "learning_rate": 2.8899999999999998e-05, "loss": 1.5523, "step": 2954000 }, { "epoch": 14.81, "learning_rate": 2.8896428571428575e-05, "loss": 1.5515, "step": 2954500 }, { "epoch": 14.81, "learning_rate": 2.8892857142857143e-05, "loss": 1.5586, "step": 2955000 }, { "epoch": 14.81, "learning_rate": 2.8889285714285714e-05, "loss": 1.5536, "step": 2955500 }, { "epoch": 14.81, "learning_rate": 2.8885714285714288e-05, "loss": 1.5473, "step": 2956000 }, { "epoch": 14.82, "learning_rate": 2.888214285714286e-05, "loss": 1.5454, "step": 2956500 }, { "epoch": 14.82, "learning_rate": 2.8878571428571433e-05, "loss": 1.5517, "step": 2957000 }, { "epoch": 14.82, "learning_rate": 2.8875e-05, "loss": 1.553, "step": 2957500 }, { "epoch": 14.83, "learning_rate": 2.887142857142857e-05, "loss": 1.5457, "step": 2958000 }, { "epoch": 14.83, "learning_rate": 2.8867857142857146e-05, "loss": 1.5554, "step": 2958500 }, { "epoch": 14.83, "learning_rate": 2.8864285714285716e-05, "loss": 1.5502, "step": 2959000 }, { "epoch": 14.83, "learning_rate": 2.8860714285714284e-05, "loss": 1.5539, "step": 2959500 }, { "epoch": 14.84, "learning_rate": 2.885714285714286e-05, "loss": 1.5478, "step": 2960000 }, { "epoch": 14.84, "learning_rate": 2.885357142857143e-05, "loss": 1.5585, "step": 2960500 }, { "epoch": 14.84, "learning_rate": 2.885e-05, "loss": 1.5484, "step": 2961000 }, { "epoch": 14.84, "learning_rate": 2.8846428571428574e-05, "loss": 1.5465, "step": 2961500 }, { "epoch": 14.85, "learning_rate": 2.8842857142857145e-05, "loss": 1.5506, "step": 2962000 }, { "epoch": 14.85, "learning_rate": 2.8839285714285712e-05, "loss": 1.5495, "step": 2962500 }, { "epoch": 14.85, "learning_rate": 2.8835714285714287e-05, "loss": 1.5487, "step": 2963000 }, { "epoch": 14.85, "learning_rate": 2.8832142857142858e-05, "loss": 1.5652, "step": 2963500 }, { "epoch": 14.86, "learning_rate": 2.8828571428571432e-05, "loss": 1.5491, "step": 2964000 }, { "epoch": 14.86, "learning_rate": 2.8825000000000003e-05, "loss": 1.5508, "step": 2964500 }, { "epoch": 14.86, "learning_rate": 2.882142857142857e-05, "loss": 1.5396, "step": 2965000 }, { "epoch": 14.87, "learning_rate": 2.8817857142857148e-05, "loss": 1.5477, "step": 2965500 }, { "epoch": 14.87, "learning_rate": 2.8814285714285715e-05, "loss": 1.5479, "step": 2966000 }, { "epoch": 14.87, "learning_rate": 2.8810714285714286e-05, "loss": 1.5493, "step": 2966500 }, { "epoch": 14.87, "learning_rate": 2.880714285714286e-05, "loss": 1.5408, "step": 2967000 }, { "epoch": 14.88, "learning_rate": 2.880357142857143e-05, "loss": 1.5537, "step": 2967500 }, { "epoch": 14.88, "learning_rate": 2.88e-05, "loss": 1.5567, "step": 2968000 }, { "epoch": 14.88, "learning_rate": 2.8796428571428573e-05, "loss": 1.5502, "step": 2968500 }, { "epoch": 14.88, "learning_rate": 2.8792857142857144e-05, "loss": 1.5514, "step": 2969000 }, { "epoch": 14.89, "learning_rate": 2.8789285714285715e-05, "loss": 1.5451, "step": 2969500 }, { "epoch": 14.89, "learning_rate": 2.878571428571429e-05, "loss": 1.5495, "step": 2970000 }, { "epoch": 14.89, "learning_rate": 2.8782142857142856e-05, "loss": 1.5458, "step": 2970500 }, { "epoch": 14.89, "learning_rate": 2.8778571428571434e-05, "loss": 1.5557, "step": 2971000 }, { "epoch": 14.9, "learning_rate": 2.8775e-05, "loss": 1.5481, "step": 2971500 }, { "epoch": 14.9, "learning_rate": 2.8771428571428572e-05, "loss": 1.534, "step": 2972000 }, { "epoch": 14.9, "learning_rate": 2.8767857142857147e-05, "loss": 1.5526, "step": 2972500 }, { "epoch": 14.91, "learning_rate": 2.8764285714285717e-05, "loss": 1.5528, "step": 2973000 }, { "epoch": 14.91, "learning_rate": 2.8760714285714285e-05, "loss": 1.5411, "step": 2973500 }, { "epoch": 14.91, "learning_rate": 2.875714285714286e-05, "loss": 1.5514, "step": 2974000 }, { "epoch": 14.91, "learning_rate": 2.875357142857143e-05, "loss": 1.5534, "step": 2974500 }, { "epoch": 14.92, "learning_rate": 2.8749999999999997e-05, "loss": 1.5528, "step": 2975000 }, { "epoch": 14.92, "learning_rate": 2.8746428571428575e-05, "loss": 1.5497, "step": 2975500 }, { "epoch": 14.92, "learning_rate": 2.8742857142857143e-05, "loss": 1.54, "step": 2976000 }, { "epoch": 14.92, "learning_rate": 2.8739285714285713e-05, "loss": 1.5463, "step": 2976500 }, { "epoch": 14.93, "learning_rate": 2.8735714285714288e-05, "loss": 1.5514, "step": 2977000 }, { "epoch": 14.93, "learning_rate": 2.873214285714286e-05, "loss": 1.5617, "step": 2977500 }, { "epoch": 14.93, "learning_rate": 2.8728571428571433e-05, "loss": 1.549, "step": 2978000 }, { "epoch": 14.93, "learning_rate": 2.8725e-05, "loss": 1.555, "step": 2978500 }, { "epoch": 14.94, "learning_rate": 2.872142857142857e-05, "loss": 1.5518, "step": 2979000 }, { "epoch": 14.94, "learning_rate": 2.8717857142857145e-05, "loss": 1.55, "step": 2979500 }, { "epoch": 14.94, "learning_rate": 2.8714285714285716e-05, "loss": 1.5467, "step": 2980000 }, { "epoch": 14.95, "learning_rate": 2.8710714285714284e-05, "loss": 1.5496, "step": 2980500 }, { "epoch": 14.95, "learning_rate": 2.870714285714286e-05, "loss": 1.5545, "step": 2981000 }, { "epoch": 14.95, "learning_rate": 2.870357142857143e-05, "loss": 1.5448, "step": 2981500 }, { "epoch": 14.95, "learning_rate": 2.87e-05, "loss": 1.5552, "step": 2982000 }, { "epoch": 14.96, "learning_rate": 2.8696428571428574e-05, "loss": 1.5438, "step": 2982500 }, { "epoch": 14.96, "learning_rate": 2.8692857142857145e-05, "loss": 1.5531, "step": 2983000 }, { "epoch": 14.96, "learning_rate": 2.8689285714285712e-05, "loss": 1.5458, "step": 2983500 }, { "epoch": 14.96, "learning_rate": 2.8685714285714286e-05, "loss": 1.5387, "step": 2984000 }, { "epoch": 14.97, "learning_rate": 2.8682142857142857e-05, "loss": 1.55, "step": 2984500 }, { "epoch": 14.97, "learning_rate": 2.867857142857143e-05, "loss": 1.555, "step": 2985000 }, { "epoch": 14.97, "learning_rate": 2.8675000000000002e-05, "loss": 1.55, "step": 2985500 }, { "epoch": 14.97, "learning_rate": 2.867142857142857e-05, "loss": 1.5437, "step": 2986000 }, { "epoch": 14.98, "learning_rate": 2.8667857142857148e-05, "loss": 1.5513, "step": 2986500 }, { "epoch": 14.98, "learning_rate": 2.8664285714285715e-05, "loss": 1.554, "step": 2987000 }, { "epoch": 14.98, "learning_rate": 2.8660714285714286e-05, "loss": 1.5595, "step": 2987500 }, { "epoch": 14.99, "learning_rate": 2.865714285714286e-05, "loss": 1.5368, "step": 2988000 }, { "epoch": 14.99, "learning_rate": 2.865357142857143e-05, "loss": 1.5477, "step": 2988500 }, { "epoch": 14.99, "learning_rate": 2.865e-05, "loss": 1.5483, "step": 2989000 }, { "epoch": 14.99, "learning_rate": 2.8646428571428573e-05, "loss": 1.5499, "step": 2989500 }, { "epoch": 15.0, "learning_rate": 2.8642857142857144e-05, "loss": 1.5449, "step": 2990000 }, { "epoch": 15.0, "learning_rate": 2.8639285714285714e-05, "loss": 1.5481, "step": 2990500 }, { "epoch": 15.0, "learning_rate": 2.863571428571429e-05, "loss": 1.5498, "step": 2991000 }, { "epoch": 15.0, "learning_rate": 2.8632142857142856e-05, "loss": 1.5494, "step": 2991500 }, { "epoch": 15.01, "learning_rate": 2.8628571428571434e-05, "loss": 1.5564, "step": 2992000 }, { "epoch": 15.01, "learning_rate": 2.8625e-05, "loss": 1.5453, "step": 2992500 }, { "epoch": 15.01, "learning_rate": 2.8621428571428572e-05, "loss": 1.548, "step": 2993000 }, { "epoch": 15.01, "learning_rate": 2.8617857142857146e-05, "loss": 1.5372, "step": 2993500 }, { "epoch": 15.02, "learning_rate": 2.8614285714285717e-05, "loss": 1.5561, "step": 2994000 }, { "epoch": 15.02, "learning_rate": 2.8610714285714285e-05, "loss": 1.5466, "step": 2994500 }, { "epoch": 15.02, "learning_rate": 2.860714285714286e-05, "loss": 1.5468, "step": 2995000 }, { "epoch": 15.03, "learning_rate": 2.860357142857143e-05, "loss": 1.538, "step": 2995500 }, { "epoch": 15.03, "learning_rate": 2.86e-05, "loss": 1.5427, "step": 2996000 }, { "epoch": 15.03, "learning_rate": 2.8596428571428575e-05, "loss": 1.5417, "step": 2996500 }, { "epoch": 15.03, "learning_rate": 2.8592857142857142e-05, "loss": 1.5518, "step": 2997000 }, { "epoch": 15.04, "learning_rate": 2.8589285714285713e-05, "loss": 1.5425, "step": 2997500 }, { "epoch": 15.04, "learning_rate": 2.8585714285714287e-05, "loss": 1.5534, "step": 2998000 }, { "epoch": 15.04, "learning_rate": 2.8582142857142858e-05, "loss": 1.5409, "step": 2998500 }, { "epoch": 15.04, "learning_rate": 2.8578571428571433e-05, "loss": 1.5278, "step": 2999000 }, { "epoch": 15.05, "learning_rate": 2.8575000000000003e-05, "loss": 1.5455, "step": 2999500 }, { "epoch": 15.05, "learning_rate": 2.857142857142857e-05, "loss": 1.5464, "step": 3000000 }, { "epoch": 15.05, "learning_rate": 2.8567857142857145e-05, "loss": 1.5473, "step": 3000500 }, { "epoch": 15.05, "learning_rate": 2.8564285714285716e-05, "loss": 1.5448, "step": 3001000 }, { "epoch": 15.06, "learning_rate": 2.8560714285714283e-05, "loss": 1.5374, "step": 3001500 }, { "epoch": 15.06, "learning_rate": 2.855714285714286e-05, "loss": 1.5369, "step": 3002000 }, { "epoch": 15.06, "learning_rate": 2.855357142857143e-05, "loss": 1.5499, "step": 3002500 }, { "epoch": 15.06, "learning_rate": 2.855e-05, "loss": 1.5508, "step": 3003000 }, { "epoch": 15.07, "learning_rate": 2.8546428571428574e-05, "loss": 1.5437, "step": 3003500 }, { "epoch": 15.07, "learning_rate": 2.8542857142857144e-05, "loss": 1.5359, "step": 3004000 }, { "epoch": 15.07, "learning_rate": 2.853928571428572e-05, "loss": 1.5471, "step": 3004500 }, { "epoch": 15.08, "learning_rate": 2.8535714285714286e-05, "loss": 1.5432, "step": 3005000 }, { "epoch": 15.08, "learning_rate": 2.8532142857142857e-05, "loss": 1.537, "step": 3005500 }, { "epoch": 15.08, "learning_rate": 2.852857142857143e-05, "loss": 1.5424, "step": 3006000 }, { "epoch": 15.08, "learning_rate": 2.8525000000000002e-05, "loss": 1.5414, "step": 3006500 }, { "epoch": 15.09, "learning_rate": 2.852142857142857e-05, "loss": 1.5459, "step": 3007000 }, { "epoch": 15.09, "learning_rate": 2.8517857142857147e-05, "loss": 1.5333, "step": 3007500 }, { "epoch": 15.09, "learning_rate": 2.8514285714285715e-05, "loss": 1.5397, "step": 3008000 }, { "epoch": 15.09, "learning_rate": 2.8510714285714286e-05, "loss": 1.5411, "step": 3008500 }, { "epoch": 15.1, "learning_rate": 2.850714285714286e-05, "loss": 1.5463, "step": 3009000 }, { "epoch": 15.1, "learning_rate": 2.850357142857143e-05, "loss": 1.5542, "step": 3009500 }, { "epoch": 15.1, "learning_rate": 2.8499999999999998e-05, "loss": 1.5458, "step": 3010000 }, { "epoch": 15.1, "learning_rate": 2.8496428571428572e-05, "loss": 1.5398, "step": 3010500 }, { "epoch": 15.11, "learning_rate": 2.8492857142857143e-05, "loss": 1.5418, "step": 3011000 }, { "epoch": 15.11, "learning_rate": 2.8489285714285718e-05, "loss": 1.5475, "step": 3011500 }, { "epoch": 15.11, "learning_rate": 2.848571428571429e-05, "loss": 1.5406, "step": 3012000 }, { "epoch": 15.12, "learning_rate": 2.8482142857142856e-05, "loss": 1.544, "step": 3012500 }, { "epoch": 15.12, "learning_rate": 2.8478571428571433e-05, "loss": 1.5548, "step": 3013000 }, { "epoch": 15.12, "learning_rate": 2.8475e-05, "loss": 1.5487, "step": 3013500 }, { "epoch": 15.12, "learning_rate": 2.8471428571428572e-05, "loss": 1.55, "step": 3014000 }, { "epoch": 15.13, "learning_rate": 2.8467857142857146e-05, "loss": 1.5395, "step": 3014500 }, { "epoch": 15.13, "learning_rate": 2.8464285714285717e-05, "loss": 1.5384, "step": 3015000 }, { "epoch": 15.13, "learning_rate": 2.8460714285714284e-05, "loss": 1.5459, "step": 3015500 }, { "epoch": 15.13, "learning_rate": 2.845714285714286e-05, "loss": 1.5416, "step": 3016000 }, { "epoch": 15.14, "learning_rate": 2.845357142857143e-05, "loss": 1.5345, "step": 3016500 }, { "epoch": 15.14, "learning_rate": 2.845e-05, "loss": 1.5454, "step": 3017000 }, { "epoch": 15.14, "learning_rate": 2.8446428571428575e-05, "loss": 1.5446, "step": 3017500 }, { "epoch": 15.14, "learning_rate": 2.8442857142857142e-05, "loss": 1.5438, "step": 3018000 }, { "epoch": 15.15, "learning_rate": 2.843928571428572e-05, "loss": 1.5401, "step": 3018500 }, { "epoch": 15.15, "learning_rate": 2.8435714285714287e-05, "loss": 1.5407, "step": 3019000 }, { "epoch": 15.15, "learning_rate": 2.8432142857142858e-05, "loss": 1.5459, "step": 3019500 }, { "epoch": 15.16, "learning_rate": 2.8428571428571432e-05, "loss": 1.5378, "step": 3020000 }, { "epoch": 15.16, "learning_rate": 2.8425000000000003e-05, "loss": 1.5442, "step": 3020500 }, { "epoch": 15.16, "learning_rate": 2.842142857142857e-05, "loss": 1.5437, "step": 3021000 }, { "epoch": 15.16, "learning_rate": 2.8417857142857145e-05, "loss": 1.5356, "step": 3021500 }, { "epoch": 15.17, "learning_rate": 2.8414285714285716e-05, "loss": 1.5387, "step": 3022000 }, { "epoch": 15.17, "learning_rate": 2.8410714285714287e-05, "loss": 1.545, "step": 3022500 }, { "epoch": 15.17, "learning_rate": 2.840714285714286e-05, "loss": 1.5371, "step": 3023000 }, { "epoch": 15.17, "learning_rate": 2.8403571428571428e-05, "loss": 1.5477, "step": 3023500 }, { "epoch": 15.18, "learning_rate": 2.84e-05, "loss": 1.5446, "step": 3024000 }, { "epoch": 15.18, "learning_rate": 2.8396428571428573e-05, "loss": 1.5354, "step": 3024500 }, { "epoch": 15.18, "learning_rate": 2.8392857142857144e-05, "loss": 1.5441, "step": 3025000 }, { "epoch": 15.18, "learning_rate": 2.838928571428572e-05, "loss": 1.5414, "step": 3025500 }, { "epoch": 15.19, "learning_rate": 2.838571428571429e-05, "loss": 1.5341, "step": 3026000 }, { "epoch": 15.19, "learning_rate": 2.8382142857142857e-05, "loss": 1.5488, "step": 3026500 }, { "epoch": 15.19, "learning_rate": 2.837857142857143e-05, "loss": 1.5419, "step": 3027000 }, { "epoch": 15.2, "learning_rate": 2.8375000000000002e-05, "loss": 1.5345, "step": 3027500 }, { "epoch": 15.2, "learning_rate": 2.837142857142857e-05, "loss": 1.5468, "step": 3028000 }, { "epoch": 15.2, "learning_rate": 2.8367857142857147e-05, "loss": 1.549, "step": 3028500 }, { "epoch": 15.2, "learning_rate": 2.8364285714285714e-05, "loss": 1.5411, "step": 3029000 }, { "epoch": 15.21, "learning_rate": 2.8360714285714285e-05, "loss": 1.5451, "step": 3029500 }, { "epoch": 15.21, "learning_rate": 2.835714285714286e-05, "loss": 1.5398, "step": 3030000 }, { "epoch": 15.21, "learning_rate": 2.835357142857143e-05, "loss": 1.5315, "step": 3030500 }, { "epoch": 15.21, "learning_rate": 2.8349999999999998e-05, "loss": 1.5405, "step": 3031000 }, { "epoch": 15.22, "learning_rate": 2.8346428571428572e-05, "loss": 1.5504, "step": 3031500 }, { "epoch": 15.22, "learning_rate": 2.8342857142857143e-05, "loss": 1.5469, "step": 3032000 }, { "epoch": 15.22, "learning_rate": 2.8339285714285717e-05, "loss": 1.5362, "step": 3032500 }, { "epoch": 15.22, "learning_rate": 2.8335714285714288e-05, "loss": 1.5388, "step": 3033000 }, { "epoch": 15.23, "learning_rate": 2.8332142857142856e-05, "loss": 1.5317, "step": 3033500 }, { "epoch": 15.23, "learning_rate": 2.8328571428571433e-05, "loss": 1.5552, "step": 3034000 }, { "epoch": 15.23, "learning_rate": 2.8325e-05, "loss": 1.5409, "step": 3034500 }, { "epoch": 15.24, "learning_rate": 2.832142857142857e-05, "loss": 1.5405, "step": 3035000 }, { "epoch": 15.24, "learning_rate": 2.8317857142857146e-05, "loss": 1.5432, "step": 3035500 }, { "epoch": 15.24, "learning_rate": 2.8314285714285717e-05, "loss": 1.5559, "step": 3036000 }, { "epoch": 15.24, "learning_rate": 2.8310714285714284e-05, "loss": 1.5397, "step": 3036500 }, { "epoch": 15.25, "learning_rate": 2.830714285714286e-05, "loss": 1.5486, "step": 3037000 }, { "epoch": 15.25, "learning_rate": 2.830357142857143e-05, "loss": 1.5481, "step": 3037500 }, { "epoch": 15.25, "learning_rate": 2.83e-05, "loss": 1.537, "step": 3038000 }, { "epoch": 15.25, "learning_rate": 2.8296428571428574e-05, "loss": 1.5478, "step": 3038500 }, { "epoch": 15.26, "learning_rate": 2.8292857142857142e-05, "loss": 1.544, "step": 3039000 }, { "epoch": 15.26, "learning_rate": 2.828928571428572e-05, "loss": 1.5445, "step": 3039500 }, { "epoch": 15.26, "learning_rate": 2.8285714285714287e-05, "loss": 1.5474, "step": 3040000 }, { "epoch": 15.26, "learning_rate": 2.8282142857142858e-05, "loss": 1.5467, "step": 3040500 }, { "epoch": 15.27, "learning_rate": 2.8278571428571432e-05, "loss": 1.5351, "step": 3041000 }, { "epoch": 15.27, "learning_rate": 2.8275000000000003e-05, "loss": 1.5341, "step": 3041500 }, { "epoch": 15.27, "learning_rate": 2.827142857142857e-05, "loss": 1.5491, "step": 3042000 }, { "epoch": 15.28, "learning_rate": 2.8267857142857145e-05, "loss": 1.5481, "step": 3042500 }, { "epoch": 15.28, "learning_rate": 2.8264285714285715e-05, "loss": 1.5404, "step": 3043000 }, { "epoch": 15.28, "learning_rate": 2.8260714285714286e-05, "loss": 1.5501, "step": 3043500 }, { "epoch": 15.28, "learning_rate": 2.825714285714286e-05, "loss": 1.5555, "step": 3044000 }, { "epoch": 15.29, "learning_rate": 2.8253571428571428e-05, "loss": 1.5462, "step": 3044500 }, { "epoch": 15.29, "learning_rate": 2.825e-05, "loss": 1.5379, "step": 3045000 }, { "epoch": 15.29, "learning_rate": 2.8246428571428573e-05, "loss": 1.5534, "step": 3045500 }, { "epoch": 15.29, "learning_rate": 2.8242857142857144e-05, "loss": 1.5525, "step": 3046000 }, { "epoch": 15.3, "learning_rate": 2.8239285714285718e-05, "loss": 1.5426, "step": 3046500 }, { "epoch": 15.3, "learning_rate": 2.823571428571429e-05, "loss": 1.5424, "step": 3047000 }, { "epoch": 15.3, "learning_rate": 2.8232142857142857e-05, "loss": 1.5411, "step": 3047500 }, { "epoch": 15.3, "learning_rate": 2.822857142857143e-05, "loss": 1.5408, "step": 3048000 }, { "epoch": 15.31, "learning_rate": 2.8225e-05, "loss": 1.5442, "step": 3048500 }, { "epoch": 15.31, "learning_rate": 2.8221428571428573e-05, "loss": 1.5386, "step": 3049000 }, { "epoch": 15.31, "learning_rate": 2.8217857142857147e-05, "loss": 1.5342, "step": 3049500 }, { "epoch": 15.32, "learning_rate": 2.8214285714285714e-05, "loss": 1.5505, "step": 3050000 }, { "epoch": 15.32, "learning_rate": 2.8210714285714285e-05, "loss": 1.5427, "step": 3050500 }, { "epoch": 15.32, "learning_rate": 2.820714285714286e-05, "loss": 1.5454, "step": 3051000 }, { "epoch": 15.32, "learning_rate": 2.820357142857143e-05, "loss": 1.5438, "step": 3051500 }, { "epoch": 15.33, "learning_rate": 2.8199999999999998e-05, "loss": 1.5388, "step": 3052000 }, { "epoch": 15.33, "learning_rate": 2.8196428571428575e-05, "loss": 1.5438, "step": 3052500 }, { "epoch": 15.33, "learning_rate": 2.8192857142857143e-05, "loss": 1.5437, "step": 3053000 }, { "epoch": 15.33, "learning_rate": 2.8189285714285717e-05, "loss": 1.5408, "step": 3053500 }, { "epoch": 15.34, "learning_rate": 2.8185714285714288e-05, "loss": 1.5341, "step": 3054000 }, { "epoch": 15.34, "learning_rate": 2.8182142857142855e-05, "loss": 1.541, "step": 3054500 }, { "epoch": 15.34, "learning_rate": 2.8178571428571433e-05, "loss": 1.5439, "step": 3055000 }, { "epoch": 15.34, "learning_rate": 2.8175e-05, "loss": 1.5427, "step": 3055500 }, { "epoch": 15.35, "learning_rate": 2.817142857142857e-05, "loss": 1.5386, "step": 3056000 }, { "epoch": 15.35, "learning_rate": 2.8167857142857146e-05, "loss": 1.5374, "step": 3056500 }, { "epoch": 15.35, "learning_rate": 2.8164285714285716e-05, "loss": 1.5381, "step": 3057000 }, { "epoch": 15.36, "learning_rate": 2.8160714285714284e-05, "loss": 1.5454, "step": 3057500 }, { "epoch": 15.36, "learning_rate": 2.8157142857142858e-05, "loss": 1.5379, "step": 3058000 }, { "epoch": 15.36, "learning_rate": 2.815357142857143e-05, "loss": 1.5478, "step": 3058500 }, { "epoch": 15.36, "learning_rate": 2.815e-05, "loss": 1.5512, "step": 3059000 }, { "epoch": 15.37, "learning_rate": 2.8146428571428574e-05, "loss": 1.5407, "step": 3059500 }, { "epoch": 15.37, "learning_rate": 2.814285714285714e-05, "loss": 1.5346, "step": 3060000 }, { "epoch": 15.37, "learning_rate": 2.813928571428572e-05, "loss": 1.5409, "step": 3060500 }, { "epoch": 15.37, "learning_rate": 2.8135714285714287e-05, "loss": 1.5223, "step": 3061000 }, { "epoch": 15.38, "learning_rate": 2.8132142857142858e-05, "loss": 1.5407, "step": 3061500 }, { "epoch": 15.38, "learning_rate": 2.8128571428571432e-05, "loss": 1.5431, "step": 3062000 }, { "epoch": 15.38, "learning_rate": 2.8125000000000003e-05, "loss": 1.5518, "step": 3062500 }, { "epoch": 15.38, "learning_rate": 2.812142857142857e-05, "loss": 1.5392, "step": 3063000 }, { "epoch": 15.39, "learning_rate": 2.8117857142857144e-05, "loss": 1.5404, "step": 3063500 }, { "epoch": 15.39, "learning_rate": 2.8114285714285715e-05, "loss": 1.5419, "step": 3064000 }, { "epoch": 15.39, "learning_rate": 2.8110714285714286e-05, "loss": 1.5425, "step": 3064500 }, { "epoch": 15.4, "learning_rate": 2.810714285714286e-05, "loss": 1.5463, "step": 3065000 }, { "epoch": 15.4, "learning_rate": 2.8103571428571428e-05, "loss": 1.5426, "step": 3065500 }, { "epoch": 15.4, "learning_rate": 2.8100000000000005e-05, "loss": 1.5357, "step": 3066000 }, { "epoch": 15.4, "learning_rate": 2.8096428571428573e-05, "loss": 1.54, "step": 3066500 }, { "epoch": 15.41, "learning_rate": 2.8092857142857144e-05, "loss": 1.5418, "step": 3067000 }, { "epoch": 15.41, "learning_rate": 2.8089285714285718e-05, "loss": 1.5354, "step": 3067500 }, { "epoch": 15.41, "learning_rate": 2.808571428571429e-05, "loss": 1.5371, "step": 3068000 }, { "epoch": 15.41, "learning_rate": 2.8082142857142856e-05, "loss": 1.5381, "step": 3068500 }, { "epoch": 15.42, "learning_rate": 2.807857142857143e-05, "loss": 1.5382, "step": 3069000 }, { "epoch": 15.42, "learning_rate": 2.8075e-05, "loss": 1.536, "step": 3069500 }, { "epoch": 15.42, "learning_rate": 2.8071428571428572e-05, "loss": 1.5267, "step": 3070000 }, { "epoch": 15.42, "learning_rate": 2.8067857142857147e-05, "loss": 1.5377, "step": 3070500 }, { "epoch": 15.43, "learning_rate": 2.8064285714285714e-05, "loss": 1.5346, "step": 3071000 }, { "epoch": 15.43, "learning_rate": 2.8060714285714285e-05, "loss": 1.5333, "step": 3071500 }, { "epoch": 15.43, "learning_rate": 2.805714285714286e-05, "loss": 1.5368, "step": 3072000 }, { "epoch": 15.44, "learning_rate": 2.805357142857143e-05, "loss": 1.532, "step": 3072500 }, { "epoch": 15.44, "learning_rate": 2.8050000000000004e-05, "loss": 1.5431, "step": 3073000 }, { "epoch": 15.44, "learning_rate": 2.8046428571428575e-05, "loss": 1.5434, "step": 3073500 }, { "epoch": 15.44, "learning_rate": 2.8042857142857143e-05, "loss": 1.5316, "step": 3074000 }, { "epoch": 15.45, "learning_rate": 2.8039285714285717e-05, "loss": 1.5455, "step": 3074500 }, { "epoch": 15.45, "learning_rate": 2.8035714285714288e-05, "loss": 1.5357, "step": 3075000 }, { "epoch": 15.45, "learning_rate": 2.803214285714286e-05, "loss": 1.5426, "step": 3075500 }, { "epoch": 15.45, "learning_rate": 2.8028571428571433e-05, "loss": 1.5275, "step": 3076000 }, { "epoch": 15.46, "learning_rate": 2.8025e-05, "loss": 1.5393, "step": 3076500 }, { "epoch": 15.46, "learning_rate": 2.802142857142857e-05, "loss": 1.54, "step": 3077000 }, { "epoch": 15.46, "learning_rate": 2.8017857142857145e-05, "loss": 1.5468, "step": 3077500 }, { "epoch": 15.46, "learning_rate": 2.8014285714285716e-05, "loss": 1.545, "step": 3078000 }, { "epoch": 15.47, "learning_rate": 2.8010714285714284e-05, "loss": 1.5324, "step": 3078500 }, { "epoch": 15.47, "learning_rate": 2.800714285714286e-05, "loss": 1.5333, "step": 3079000 }, { "epoch": 15.47, "learning_rate": 2.800357142857143e-05, "loss": 1.5429, "step": 3079500 }, { "epoch": 15.47, "learning_rate": 2.8000000000000003e-05, "loss": 1.5425, "step": 3080000 }, { "epoch": 15.48, "learning_rate": 2.7996428571428574e-05, "loss": 1.5428, "step": 3080500 }, { "epoch": 15.48, "learning_rate": 2.799285714285714e-05, "loss": 1.5388, "step": 3081000 }, { "epoch": 15.48, "learning_rate": 2.798928571428572e-05, "loss": 1.5454, "step": 3081500 }, { "epoch": 15.49, "learning_rate": 2.7985714285714286e-05, "loss": 1.5437, "step": 3082000 }, { "epoch": 15.49, "learning_rate": 2.7982142857142857e-05, "loss": 1.5369, "step": 3082500 }, { "epoch": 15.49, "learning_rate": 2.797857142857143e-05, "loss": 1.5362, "step": 3083000 }, { "epoch": 15.49, "learning_rate": 2.7975000000000002e-05, "loss": 1.5406, "step": 3083500 }, { "epoch": 15.5, "learning_rate": 2.797142857142857e-05, "loss": 1.5427, "step": 3084000 }, { "epoch": 15.5, "learning_rate": 2.7967857142857144e-05, "loss": 1.5304, "step": 3084500 }, { "epoch": 15.5, "learning_rate": 2.7964285714285715e-05, "loss": 1.5276, "step": 3085000 }, { "epoch": 15.5, "learning_rate": 2.7960714285714286e-05, "loss": 1.5454, "step": 3085500 }, { "epoch": 15.51, "learning_rate": 2.795714285714286e-05, "loss": 1.5412, "step": 3086000 }, { "epoch": 15.51, "learning_rate": 2.7953571428571428e-05, "loss": 1.5381, "step": 3086500 }, { "epoch": 15.51, "learning_rate": 2.7950000000000005e-05, "loss": 1.5403, "step": 3087000 }, { "epoch": 15.51, "learning_rate": 2.7946428571428573e-05, "loss": 1.5381, "step": 3087500 }, { "epoch": 15.52, "learning_rate": 2.7942857142857143e-05, "loss": 1.5512, "step": 3088000 }, { "epoch": 15.52, "learning_rate": 2.7939285714285718e-05, "loss": 1.5393, "step": 3088500 }, { "epoch": 15.52, "learning_rate": 2.793571428571429e-05, "loss": 1.5365, "step": 3089000 }, { "epoch": 15.53, "learning_rate": 2.7932142857142856e-05, "loss": 1.5484, "step": 3089500 }, { "epoch": 15.53, "learning_rate": 2.792857142857143e-05, "loss": 1.5352, "step": 3090000 }, { "epoch": 15.53, "learning_rate": 2.7925e-05, "loss": 1.5472, "step": 3090500 }, { "epoch": 15.53, "learning_rate": 2.7921428571428572e-05, "loss": 1.5414, "step": 3091000 }, { "epoch": 15.54, "learning_rate": 2.7917857142857146e-05, "loss": 1.5386, "step": 3091500 }, { "epoch": 15.54, "learning_rate": 2.7914285714285714e-05, "loss": 1.5512, "step": 3092000 }, { "epoch": 15.54, "learning_rate": 2.7910714285714285e-05, "loss": 1.5413, "step": 3092500 }, { "epoch": 15.54, "learning_rate": 2.790714285714286e-05, "loss": 1.5356, "step": 3093000 }, { "epoch": 15.55, "learning_rate": 2.790357142857143e-05, "loss": 1.544, "step": 3093500 }, { "epoch": 15.55, "learning_rate": 2.7900000000000004e-05, "loss": 1.5339, "step": 3094000 }, { "epoch": 15.55, "learning_rate": 2.7896428571428575e-05, "loss": 1.5457, "step": 3094500 }, { "epoch": 15.55, "learning_rate": 2.7892857142857142e-05, "loss": 1.5293, "step": 3095000 }, { "epoch": 15.56, "learning_rate": 2.7889285714285716e-05, "loss": 1.5396, "step": 3095500 }, { "epoch": 15.56, "learning_rate": 2.7885714285714287e-05, "loss": 1.532, "step": 3096000 }, { "epoch": 15.56, "learning_rate": 2.7882142857142858e-05, "loss": 1.5359, "step": 3096500 }, { "epoch": 15.57, "learning_rate": 2.7878571428571432e-05, "loss": 1.5345, "step": 3097000 }, { "epoch": 15.57, "learning_rate": 2.7875e-05, "loss": 1.5294, "step": 3097500 }, { "epoch": 15.57, "learning_rate": 2.787142857142857e-05, "loss": 1.5436, "step": 3098000 }, { "epoch": 15.57, "learning_rate": 2.7867857142857145e-05, "loss": 1.5435, "step": 3098500 }, { "epoch": 15.58, "learning_rate": 2.7864285714285716e-05, "loss": 1.5418, "step": 3099000 }, { "epoch": 15.58, "learning_rate": 2.7860714285714283e-05, "loss": 1.5359, "step": 3099500 }, { "epoch": 15.58, "learning_rate": 2.785714285714286e-05, "loss": 1.5342, "step": 3100000 }, { "epoch": 15.58, "learning_rate": 2.785357142857143e-05, "loss": 1.536, "step": 3100500 }, { "epoch": 15.59, "learning_rate": 2.7850000000000003e-05, "loss": 1.5381, "step": 3101000 }, { "epoch": 15.59, "learning_rate": 2.7846428571428574e-05, "loss": 1.5392, "step": 3101500 }, { "epoch": 15.59, "learning_rate": 2.7842857142857144e-05, "loss": 1.5354, "step": 3102000 }, { "epoch": 15.59, "learning_rate": 2.783928571428572e-05, "loss": 1.5399, "step": 3102500 }, { "epoch": 15.6, "learning_rate": 2.7835714285714286e-05, "loss": 1.5421, "step": 3103000 }, { "epoch": 15.6, "learning_rate": 2.7832142857142857e-05, "loss": 1.5414, "step": 3103500 }, { "epoch": 15.6, "learning_rate": 2.782857142857143e-05, "loss": 1.5419, "step": 3104000 }, { "epoch": 15.61, "learning_rate": 2.7825000000000002e-05, "loss": 1.5379, "step": 3104500 }, { "epoch": 15.61, "learning_rate": 2.782142857142857e-05, "loss": 1.5315, "step": 3105000 }, { "epoch": 15.61, "learning_rate": 2.7817857142857144e-05, "loss": 1.5281, "step": 3105500 }, { "epoch": 15.61, "learning_rate": 2.7814285714285715e-05, "loss": 1.5371, "step": 3106000 }, { "epoch": 15.62, "learning_rate": 2.7810714285714286e-05, "loss": 1.5345, "step": 3106500 }, { "epoch": 15.62, "learning_rate": 2.780714285714286e-05, "loss": 1.5362, "step": 3107000 }, { "epoch": 15.62, "learning_rate": 2.7803571428571427e-05, "loss": 1.5448, "step": 3107500 }, { "epoch": 15.62, "learning_rate": 2.7800000000000005e-05, "loss": 1.5412, "step": 3108000 }, { "epoch": 15.63, "learning_rate": 2.7796428571428572e-05, "loss": 1.5328, "step": 3108500 }, { "epoch": 15.63, "learning_rate": 2.7792857142857143e-05, "loss": 1.5354, "step": 3109000 }, { "epoch": 15.63, "learning_rate": 2.7789285714285717e-05, "loss": 1.5301, "step": 3109500 }, { "epoch": 15.63, "learning_rate": 2.778571428571429e-05, "loss": 1.5307, "step": 3110000 }, { "epoch": 15.64, "learning_rate": 2.7782142857142856e-05, "loss": 1.5323, "step": 3110500 }, { "epoch": 15.64, "learning_rate": 2.777857142857143e-05, "loss": 1.5317, "step": 3111000 }, { "epoch": 15.64, "learning_rate": 2.7775e-05, "loss": 1.5371, "step": 3111500 }, { "epoch": 15.65, "learning_rate": 2.7771428571428572e-05, "loss": 1.5303, "step": 3112000 }, { "epoch": 15.65, "learning_rate": 2.7767857142857146e-05, "loss": 1.5431, "step": 3112500 }, { "epoch": 15.65, "learning_rate": 2.7764285714285713e-05, "loss": 1.5426, "step": 3113000 }, { "epoch": 15.65, "learning_rate": 2.7760714285714284e-05, "loss": 1.537, "step": 3113500 }, { "epoch": 15.66, "learning_rate": 2.775714285714286e-05, "loss": 1.5337, "step": 3114000 }, { "epoch": 15.66, "learning_rate": 2.775357142857143e-05, "loss": 1.5387, "step": 3114500 }, { "epoch": 15.66, "learning_rate": 2.7750000000000004e-05, "loss": 1.526, "step": 3115000 }, { "epoch": 15.66, "learning_rate": 2.7746428571428575e-05, "loss": 1.5455, "step": 3115500 }, { "epoch": 15.67, "learning_rate": 2.7742857142857142e-05, "loss": 1.5384, "step": 3116000 }, { "epoch": 15.67, "learning_rate": 2.7739285714285716e-05, "loss": 1.5394, "step": 3116500 }, { "epoch": 15.67, "learning_rate": 2.7735714285714287e-05, "loss": 1.5395, "step": 3117000 }, { "epoch": 15.67, "learning_rate": 2.7732142857142858e-05, "loss": 1.5405, "step": 3117500 }, { "epoch": 15.68, "learning_rate": 2.7728571428571432e-05, "loss": 1.5359, "step": 3118000 }, { "epoch": 15.68, "learning_rate": 2.7725e-05, "loss": 1.5312, "step": 3118500 }, { "epoch": 15.68, "learning_rate": 2.772142857142857e-05, "loss": 1.5293, "step": 3119000 }, { "epoch": 15.69, "learning_rate": 2.7717857142857145e-05, "loss": 1.5369, "step": 3119500 }, { "epoch": 15.69, "learning_rate": 2.7714285714285716e-05, "loss": 1.5386, "step": 3120000 }, { "epoch": 15.69, "learning_rate": 2.7710714285714283e-05, "loss": 1.5345, "step": 3120500 }, { "epoch": 15.69, "learning_rate": 2.770714285714286e-05, "loss": 1.5429, "step": 3121000 }, { "epoch": 15.7, "learning_rate": 2.7703571428571428e-05, "loss": 1.5331, "step": 3121500 }, { "epoch": 15.7, "learning_rate": 2.7700000000000002e-05, "loss": 1.5345, "step": 3122000 }, { "epoch": 15.7, "learning_rate": 2.7696428571428573e-05, "loss": 1.5331, "step": 3122500 }, { "epoch": 15.7, "learning_rate": 2.7692857142857144e-05, "loss": 1.5364, "step": 3123000 }, { "epoch": 15.71, "learning_rate": 2.768928571428572e-05, "loss": 1.5381, "step": 3123500 }, { "epoch": 15.71, "learning_rate": 2.7685714285714286e-05, "loss": 1.5374, "step": 3124000 }, { "epoch": 15.71, "learning_rate": 2.7682142857142857e-05, "loss": 1.5416, "step": 3124500 }, { "epoch": 15.71, "learning_rate": 2.767857142857143e-05, "loss": 1.5456, "step": 3125000 }, { "epoch": 15.72, "learning_rate": 2.7675000000000002e-05, "loss": 1.5354, "step": 3125500 }, { "epoch": 15.72, "learning_rate": 2.767142857142857e-05, "loss": 1.5343, "step": 3126000 }, { "epoch": 15.72, "learning_rate": 2.7667857142857147e-05, "loss": 1.5474, "step": 3126500 }, { "epoch": 15.73, "learning_rate": 2.7664285714285714e-05, "loss": 1.5448, "step": 3127000 }, { "epoch": 15.73, "learning_rate": 2.766071428571429e-05, "loss": 1.5312, "step": 3127500 }, { "epoch": 15.73, "learning_rate": 2.765714285714286e-05, "loss": 1.5389, "step": 3128000 }, { "epoch": 15.73, "learning_rate": 2.7653571428571427e-05, "loss": 1.5443, "step": 3128500 }, { "epoch": 15.74, "learning_rate": 2.7650000000000005e-05, "loss": 1.5284, "step": 3129000 }, { "epoch": 15.74, "learning_rate": 2.7646428571428572e-05, "loss": 1.5352, "step": 3129500 }, { "epoch": 15.74, "learning_rate": 2.7642857142857143e-05, "loss": 1.5398, "step": 3130000 }, { "epoch": 15.74, "learning_rate": 2.7639285714285717e-05, "loss": 1.5344, "step": 3130500 }, { "epoch": 15.75, "learning_rate": 2.7635714285714288e-05, "loss": 1.5351, "step": 3131000 }, { "epoch": 15.75, "learning_rate": 2.7632142857142856e-05, "loss": 1.5302, "step": 3131500 }, { "epoch": 15.75, "learning_rate": 2.762857142857143e-05, "loss": 1.5348, "step": 3132000 }, { "epoch": 15.75, "learning_rate": 2.7625e-05, "loss": 1.5394, "step": 3132500 }, { "epoch": 15.76, "learning_rate": 2.762142857142857e-05, "loss": 1.5344, "step": 3133000 }, { "epoch": 15.76, "learning_rate": 2.7617857142857146e-05, "loss": 1.5378, "step": 3133500 }, { "epoch": 15.76, "learning_rate": 2.7614285714285713e-05, "loss": 1.5377, "step": 3134000 }, { "epoch": 15.77, "learning_rate": 2.761071428571429e-05, "loss": 1.5334, "step": 3134500 }, { "epoch": 15.77, "learning_rate": 2.760714285714286e-05, "loss": 1.5412, "step": 3135000 }, { "epoch": 15.77, "learning_rate": 2.760357142857143e-05, "loss": 1.5358, "step": 3135500 }, { "epoch": 15.77, "learning_rate": 2.7600000000000003e-05, "loss": 1.5465, "step": 3136000 }, { "epoch": 15.78, "learning_rate": 2.7596428571428574e-05, "loss": 1.5291, "step": 3136500 }, { "epoch": 15.78, "learning_rate": 2.7592857142857142e-05, "loss": 1.538, "step": 3137000 }, { "epoch": 15.78, "learning_rate": 2.7589285714285716e-05, "loss": 1.5432, "step": 3137500 }, { "epoch": 15.78, "learning_rate": 2.7585714285714287e-05, "loss": 1.5526, "step": 3138000 }, { "epoch": 15.79, "learning_rate": 2.7582142857142858e-05, "loss": 1.5223, "step": 3138500 }, { "epoch": 15.79, "learning_rate": 2.7578571428571432e-05, "loss": 1.535, "step": 3139000 }, { "epoch": 15.79, "learning_rate": 2.7575e-05, "loss": 1.5419, "step": 3139500 }, { "epoch": 15.79, "learning_rate": 2.757142857142857e-05, "loss": 1.5327, "step": 3140000 }, { "epoch": 15.8, "learning_rate": 2.7567857142857145e-05, "loss": 1.5401, "step": 3140500 }, { "epoch": 15.8, "learning_rate": 2.7564285714285715e-05, "loss": 1.5255, "step": 3141000 }, { "epoch": 15.8, "learning_rate": 2.756071428571429e-05, "loss": 1.5317, "step": 3141500 }, { "epoch": 15.81, "learning_rate": 2.755714285714286e-05, "loss": 1.5392, "step": 3142000 }, { "epoch": 15.81, "learning_rate": 2.7553571428571428e-05, "loss": 1.5355, "step": 3142500 }, { "epoch": 15.81, "learning_rate": 2.7550000000000002e-05, "loss": 1.5243, "step": 3143000 }, { "epoch": 15.81, "learning_rate": 2.7546428571428573e-05, "loss": 1.548, "step": 3143500 }, { "epoch": 15.82, "learning_rate": 2.7542857142857144e-05, "loss": 1.5383, "step": 3144000 }, { "epoch": 15.82, "learning_rate": 2.7539285714285718e-05, "loss": 1.5395, "step": 3144500 }, { "epoch": 15.82, "learning_rate": 2.7535714285714286e-05, "loss": 1.5278, "step": 3145000 }, { "epoch": 15.82, "learning_rate": 2.7532142857142856e-05, "loss": 1.5262, "step": 3145500 }, { "epoch": 15.83, "learning_rate": 2.752857142857143e-05, "loss": 1.5311, "step": 3146000 }, { "epoch": 15.83, "learning_rate": 2.7525e-05, "loss": 1.5449, "step": 3146500 }, { "epoch": 15.83, "learning_rate": 2.752142857142857e-05, "loss": 1.536, "step": 3147000 }, { "epoch": 15.83, "learning_rate": 2.7517857142857147e-05, "loss": 1.5337, "step": 3147500 }, { "epoch": 15.84, "learning_rate": 2.7514285714285714e-05, "loss": 1.5426, "step": 3148000 }, { "epoch": 15.84, "learning_rate": 2.751071428571429e-05, "loss": 1.5435, "step": 3148500 }, { "epoch": 15.84, "learning_rate": 2.750714285714286e-05, "loss": 1.5415, "step": 3149000 }, { "epoch": 15.85, "learning_rate": 2.750357142857143e-05, "loss": 1.5243, "step": 3149500 }, { "epoch": 15.85, "learning_rate": 2.7500000000000004e-05, "loss": 1.5285, "step": 3150000 }, { "epoch": 15.85, "learning_rate": 2.7496428571428572e-05, "loss": 1.539, "step": 3150500 }, { "epoch": 15.85, "learning_rate": 2.7492857142857143e-05, "loss": 1.54, "step": 3151000 }, { "epoch": 15.86, "learning_rate": 2.7489285714285717e-05, "loss": 1.5442, "step": 3151500 }, { "epoch": 15.86, "learning_rate": 2.7485714285714288e-05, "loss": 1.5313, "step": 3152000 }, { "epoch": 15.86, "learning_rate": 2.7482142857142855e-05, "loss": 1.5387, "step": 3152500 }, { "epoch": 15.86, "learning_rate": 2.7478571428571433e-05, "loss": 1.5276, "step": 3153000 }, { "epoch": 15.87, "learning_rate": 2.7475e-05, "loss": 1.53, "step": 3153500 }, { "epoch": 15.87, "learning_rate": 2.747142857142857e-05, "loss": 1.5315, "step": 3154000 }, { "epoch": 15.87, "learning_rate": 2.7467857142857145e-05, "loss": 1.5366, "step": 3154500 }, { "epoch": 15.87, "learning_rate": 2.7464285714285713e-05, "loss": 1.541, "step": 3155000 }, { "epoch": 15.88, "learning_rate": 2.746071428571429e-05, "loss": 1.5407, "step": 3155500 }, { "epoch": 15.88, "learning_rate": 2.7457142857142858e-05, "loss": 1.5369, "step": 3156000 }, { "epoch": 15.88, "learning_rate": 2.745357142857143e-05, "loss": 1.5282, "step": 3156500 }, { "epoch": 15.88, "learning_rate": 2.7450000000000003e-05, "loss": 1.5335, "step": 3157000 }, { "epoch": 15.89, "learning_rate": 2.7446428571428574e-05, "loss": 1.5363, "step": 3157500 }, { "epoch": 15.89, "learning_rate": 2.744285714285714e-05, "loss": 1.5359, "step": 3158000 }, { "epoch": 15.89, "learning_rate": 2.7439285714285716e-05, "loss": 1.536, "step": 3158500 }, { "epoch": 15.9, "learning_rate": 2.7435714285714287e-05, "loss": 1.5386, "step": 3159000 }, { "epoch": 15.9, "learning_rate": 2.7432142857142857e-05, "loss": 1.5262, "step": 3159500 }, { "epoch": 15.9, "learning_rate": 2.742857142857143e-05, "loss": 1.537, "step": 3160000 }, { "epoch": 15.9, "learning_rate": 2.7425e-05, "loss": 1.547, "step": 3160500 }, { "epoch": 15.91, "learning_rate": 2.742142857142857e-05, "loss": 1.54, "step": 3161000 }, { "epoch": 15.91, "learning_rate": 2.7417857142857144e-05, "loss": 1.5287, "step": 3161500 }, { "epoch": 15.91, "learning_rate": 2.7414285714285715e-05, "loss": 1.5346, "step": 3162000 }, { "epoch": 15.91, "learning_rate": 2.741071428571429e-05, "loss": 1.5247, "step": 3162500 }, { "epoch": 15.92, "learning_rate": 2.740714285714286e-05, "loss": 1.5404, "step": 3163000 }, { "epoch": 15.92, "learning_rate": 2.7403571428571428e-05, "loss": 1.5331, "step": 3163500 }, { "epoch": 15.92, "learning_rate": 2.7400000000000002e-05, "loss": 1.5381, "step": 3164000 }, { "epoch": 15.92, "learning_rate": 2.7396428571428573e-05, "loss": 1.5328, "step": 3164500 }, { "epoch": 15.93, "learning_rate": 2.7392857142857144e-05, "loss": 1.5373, "step": 3165000 }, { "epoch": 15.93, "learning_rate": 2.7389285714285718e-05, "loss": 1.532, "step": 3165500 }, { "epoch": 15.93, "learning_rate": 2.7385714285714285e-05, "loss": 1.5288, "step": 3166000 }, { "epoch": 15.94, "learning_rate": 2.7382142857142856e-05, "loss": 1.5303, "step": 3166500 }, { "epoch": 15.94, "learning_rate": 2.737857142857143e-05, "loss": 1.5417, "step": 3167000 }, { "epoch": 15.94, "learning_rate": 2.7375e-05, "loss": 1.5302, "step": 3167500 }, { "epoch": 15.94, "learning_rate": 2.737142857142857e-05, "loss": 1.5361, "step": 3168000 }, { "epoch": 15.95, "learning_rate": 2.7367857142857146e-05, "loss": 1.5355, "step": 3168500 }, { "epoch": 15.95, "learning_rate": 2.7364285714285714e-05, "loss": 1.5292, "step": 3169000 }, { "epoch": 15.95, "learning_rate": 2.7360714285714288e-05, "loss": 1.5322, "step": 3169500 }, { "epoch": 15.95, "learning_rate": 2.735714285714286e-05, "loss": 1.5251, "step": 3170000 }, { "epoch": 15.96, "learning_rate": 2.735357142857143e-05, "loss": 1.5291, "step": 3170500 }, { "epoch": 15.96, "learning_rate": 2.7350000000000004e-05, "loss": 1.5371, "step": 3171000 }, { "epoch": 15.96, "learning_rate": 2.734642857142857e-05, "loss": 1.5319, "step": 3171500 }, { "epoch": 15.96, "learning_rate": 2.7342857142857142e-05, "loss": 1.5297, "step": 3172000 }, { "epoch": 15.97, "learning_rate": 2.7339285714285717e-05, "loss": 1.53, "step": 3172500 }, { "epoch": 15.97, "learning_rate": 2.7335714285714288e-05, "loss": 1.5368, "step": 3173000 }, { "epoch": 15.97, "learning_rate": 2.7332142857142855e-05, "loss": 1.5358, "step": 3173500 }, { "epoch": 15.98, "learning_rate": 2.7328571428571433e-05, "loss": 1.5369, "step": 3174000 }, { "epoch": 15.98, "learning_rate": 2.7325e-05, "loss": 1.5295, "step": 3174500 }, { "epoch": 15.98, "learning_rate": 2.732142857142857e-05, "loss": 1.5401, "step": 3175000 }, { "epoch": 15.98, "learning_rate": 2.7317857142857145e-05, "loss": 1.5282, "step": 3175500 }, { "epoch": 15.99, "learning_rate": 2.7314285714285716e-05, "loss": 1.5341, "step": 3176000 }, { "epoch": 15.99, "learning_rate": 2.731071428571429e-05, "loss": 1.5359, "step": 3176500 }, { "epoch": 15.99, "learning_rate": 2.7307142857142858e-05, "loss": 1.5326, "step": 3177000 }, { "epoch": 15.99, "learning_rate": 2.730357142857143e-05, "loss": 1.5369, "step": 3177500 }, { "epoch": 16.0, "learning_rate": 2.7300000000000003e-05, "loss": 1.535, "step": 3178000 }, { "epoch": 16.0, "learning_rate": 2.7296428571428574e-05, "loss": 1.5319, "step": 3178500 }, { "epoch": 16.0, "learning_rate": 2.729285714285714e-05, "loss": 1.5235, "step": 3179000 }, { "epoch": 16.0, "learning_rate": 2.728928571428572e-05, "loss": 1.525, "step": 3179500 }, { "epoch": 16.01, "learning_rate": 2.7285714285714286e-05, "loss": 1.5184, "step": 3180000 }, { "epoch": 16.01, "learning_rate": 2.7282142857142857e-05, "loss": 1.5213, "step": 3180500 }, { "epoch": 16.01, "learning_rate": 2.727857142857143e-05, "loss": 1.5187, "step": 3181000 }, { "epoch": 16.02, "learning_rate": 2.7275e-05, "loss": 1.5214, "step": 3181500 }, { "epoch": 16.02, "learning_rate": 2.727142857142857e-05, "loss": 1.514, "step": 3182000 }, { "epoch": 16.02, "learning_rate": 2.7267857142857144e-05, "loss": 1.5312, "step": 3182500 }, { "epoch": 16.02, "learning_rate": 2.7264285714285715e-05, "loss": 1.5299, "step": 3183000 }, { "epoch": 16.03, "learning_rate": 2.726071428571429e-05, "loss": 1.5321, "step": 3183500 }, { "epoch": 16.03, "learning_rate": 2.725714285714286e-05, "loss": 1.5358, "step": 3184000 }, { "epoch": 16.03, "learning_rate": 2.7253571428571427e-05, "loss": 1.5283, "step": 3184500 }, { "epoch": 16.03, "learning_rate": 2.725e-05, "loss": 1.5272, "step": 3185000 }, { "epoch": 16.04, "learning_rate": 2.7246428571428573e-05, "loss": 1.538, "step": 3185500 }, { "epoch": 16.04, "learning_rate": 2.7242857142857143e-05, "loss": 1.5307, "step": 3186000 }, { "epoch": 16.04, "learning_rate": 2.7239285714285718e-05, "loss": 1.5266, "step": 3186500 }, { "epoch": 16.04, "learning_rate": 2.7235714285714285e-05, "loss": 1.524, "step": 3187000 }, { "epoch": 16.05, "learning_rate": 2.7232142857142856e-05, "loss": 1.5243, "step": 3187500 }, { "epoch": 16.05, "learning_rate": 2.722857142857143e-05, "loss": 1.528, "step": 3188000 }, { "epoch": 16.05, "learning_rate": 2.7225e-05, "loss": 1.535, "step": 3188500 }, { "epoch": 16.06, "learning_rate": 2.7221428571428575e-05, "loss": 1.5267, "step": 3189000 }, { "epoch": 16.06, "learning_rate": 2.7217857142857146e-05, "loss": 1.5262, "step": 3189500 }, { "epoch": 16.06, "learning_rate": 2.7214285714285714e-05, "loss": 1.5322, "step": 3190000 }, { "epoch": 16.06, "learning_rate": 2.7210714285714288e-05, "loss": 1.5282, "step": 3190500 }, { "epoch": 16.07, "learning_rate": 2.720714285714286e-05, "loss": 1.5318, "step": 3191000 }, { "epoch": 16.07, "learning_rate": 2.720357142857143e-05, "loss": 1.5268, "step": 3191500 }, { "epoch": 16.07, "learning_rate": 2.7200000000000004e-05, "loss": 1.5274, "step": 3192000 }, { "epoch": 16.07, "learning_rate": 2.719642857142857e-05, "loss": 1.528, "step": 3192500 }, { "epoch": 16.08, "learning_rate": 2.7192857142857142e-05, "loss": 1.5317, "step": 3193000 }, { "epoch": 16.08, "learning_rate": 2.7189285714285716e-05, "loss": 1.5502, "step": 3193500 }, { "epoch": 16.08, "learning_rate": 2.7185714285714287e-05, "loss": 1.5286, "step": 3194000 }, { "epoch": 16.08, "learning_rate": 2.7182142857142855e-05, "loss": 1.5333, "step": 3194500 }, { "epoch": 16.09, "learning_rate": 2.7178571428571432e-05, "loss": 1.526, "step": 3195000 }, { "epoch": 16.09, "learning_rate": 2.7175e-05, "loss": 1.5259, "step": 3195500 }, { "epoch": 16.09, "learning_rate": 2.7171428571428574e-05, "loss": 1.5318, "step": 3196000 }, { "epoch": 16.1, "learning_rate": 2.7167857142857145e-05, "loss": 1.5232, "step": 3196500 }, { "epoch": 16.1, "learning_rate": 2.7164285714285716e-05, "loss": 1.5248, "step": 3197000 }, { "epoch": 16.1, "learning_rate": 2.716071428571429e-05, "loss": 1.535, "step": 3197500 }, { "epoch": 16.1, "learning_rate": 2.7157142857142858e-05, "loss": 1.535, "step": 3198000 }, { "epoch": 16.11, "learning_rate": 2.715357142857143e-05, "loss": 1.531, "step": 3198500 }, { "epoch": 16.11, "learning_rate": 2.7150000000000003e-05, "loss": 1.5332, "step": 3199000 }, { "epoch": 16.11, "learning_rate": 2.7146428571428573e-05, "loss": 1.5243, "step": 3199500 }, { "epoch": 16.11, "learning_rate": 2.714285714285714e-05, "loss": 1.5289, "step": 3200000 }, { "epoch": 16.12, "learning_rate": 2.713928571428572e-05, "loss": 1.5208, "step": 3200500 }, { "epoch": 16.12, "learning_rate": 2.7135714285714286e-05, "loss": 1.5251, "step": 3201000 }, { "epoch": 16.12, "learning_rate": 2.7132142857142857e-05, "loss": 1.5237, "step": 3201500 }, { "epoch": 16.12, "learning_rate": 2.712857142857143e-05, "loss": 1.5307, "step": 3202000 }, { "epoch": 16.13, "learning_rate": 2.7125000000000002e-05, "loss": 1.5243, "step": 3202500 }, { "epoch": 16.13, "learning_rate": 2.7121428571428576e-05, "loss": 1.5271, "step": 3203000 }, { "epoch": 16.13, "learning_rate": 2.7117857142857144e-05, "loss": 1.5245, "step": 3203500 }, { "epoch": 16.14, "learning_rate": 2.7114285714285715e-05, "loss": 1.5274, "step": 3204000 }, { "epoch": 16.14, "learning_rate": 2.711071428571429e-05, "loss": 1.5269, "step": 3204500 }, { "epoch": 16.14, "learning_rate": 2.710714285714286e-05, "loss": 1.5288, "step": 3205000 }, { "epoch": 16.14, "learning_rate": 2.7103571428571427e-05, "loss": 1.5231, "step": 3205500 }, { "epoch": 16.15, "learning_rate": 2.7100000000000005e-05, "loss": 1.528, "step": 3206000 }, { "epoch": 16.15, "learning_rate": 2.7096428571428572e-05, "loss": 1.5253, "step": 3206500 }, { "epoch": 16.15, "learning_rate": 2.7092857142857143e-05, "loss": 1.5294, "step": 3207000 }, { "epoch": 16.15, "learning_rate": 2.7089285714285717e-05, "loss": 1.527, "step": 3207500 }, { "epoch": 16.16, "learning_rate": 2.7085714285714285e-05, "loss": 1.5267, "step": 3208000 }, { "epoch": 16.16, "learning_rate": 2.7082142857142856e-05, "loss": 1.5252, "step": 3208500 }, { "epoch": 16.16, "learning_rate": 2.707857142857143e-05, "loss": 1.5212, "step": 3209000 }, { "epoch": 16.16, "learning_rate": 2.7075e-05, "loss": 1.5208, "step": 3209500 }, { "epoch": 16.17, "learning_rate": 2.7071428571428575e-05, "loss": 1.5231, "step": 3210000 }, { "epoch": 16.17, "learning_rate": 2.7067857142857146e-05, "loss": 1.5206, "step": 3210500 }, { "epoch": 16.17, "learning_rate": 2.7064285714285713e-05, "loss": 1.5223, "step": 3211000 }, { "epoch": 16.18, "learning_rate": 2.7060714285714288e-05, "loss": 1.5199, "step": 3211500 }, { "epoch": 16.18, "learning_rate": 2.705714285714286e-05, "loss": 1.5221, "step": 3212000 }, { "epoch": 16.18, "learning_rate": 2.705357142857143e-05, "loss": 1.5167, "step": 3212500 }, { "epoch": 16.18, "learning_rate": 2.7050000000000004e-05, "loss": 1.5297, "step": 3213000 }, { "epoch": 16.19, "learning_rate": 2.704642857142857e-05, "loss": 1.5222, "step": 3213500 }, { "epoch": 16.19, "learning_rate": 2.7042857142857142e-05, "loss": 1.5345, "step": 3214000 }, { "epoch": 16.19, "learning_rate": 2.7039285714285716e-05, "loss": 1.5313, "step": 3214500 }, { "epoch": 16.19, "learning_rate": 2.7035714285714287e-05, "loss": 1.5311, "step": 3215000 }, { "epoch": 16.2, "learning_rate": 2.7032142857142854e-05, "loss": 1.5273, "step": 3215500 }, { "epoch": 16.2, "learning_rate": 2.7028571428571432e-05, "loss": 1.5243, "step": 3216000 }, { "epoch": 16.2, "learning_rate": 2.7025e-05, "loss": 1.5196, "step": 3216500 }, { "epoch": 16.2, "learning_rate": 2.7021428571428574e-05, "loss": 1.5272, "step": 3217000 }, { "epoch": 16.21, "learning_rate": 2.7017857142857145e-05, "loss": 1.5232, "step": 3217500 }, { "epoch": 16.21, "learning_rate": 2.7014285714285716e-05, "loss": 1.5245, "step": 3218000 }, { "epoch": 16.21, "learning_rate": 2.701071428571429e-05, "loss": 1.5266, "step": 3218500 }, { "epoch": 16.22, "learning_rate": 2.7007142857142857e-05, "loss": 1.5237, "step": 3219000 }, { "epoch": 16.22, "learning_rate": 2.7003571428571428e-05, "loss": 1.5231, "step": 3219500 }, { "epoch": 16.22, "learning_rate": 2.7000000000000002e-05, "loss": 1.526, "step": 3220000 }, { "epoch": 16.22, "learning_rate": 2.6996428571428573e-05, "loss": 1.5347, "step": 3220500 }, { "epoch": 16.23, "learning_rate": 2.699285714285714e-05, "loss": 1.5264, "step": 3221000 }, { "epoch": 16.23, "learning_rate": 2.698928571428572e-05, "loss": 1.5199, "step": 3221500 }, { "epoch": 16.23, "learning_rate": 2.6985714285714286e-05, "loss": 1.5312, "step": 3222000 }, { "epoch": 16.23, "learning_rate": 2.6982142857142857e-05, "loss": 1.529, "step": 3222500 }, { "epoch": 16.24, "learning_rate": 2.697857142857143e-05, "loss": 1.5372, "step": 3223000 }, { "epoch": 16.24, "learning_rate": 2.6975000000000002e-05, "loss": 1.5071, "step": 3223500 }, { "epoch": 16.24, "learning_rate": 2.6971428571428576e-05, "loss": 1.521, "step": 3224000 }, { "epoch": 16.24, "learning_rate": 2.6967857142857143e-05, "loss": 1.5308, "step": 3224500 }, { "epoch": 16.25, "learning_rate": 2.6964285714285714e-05, "loss": 1.5253, "step": 3225000 }, { "epoch": 16.25, "learning_rate": 2.696071428571429e-05, "loss": 1.5226, "step": 3225500 }, { "epoch": 16.25, "learning_rate": 2.695714285714286e-05, "loss": 1.5233, "step": 3226000 }, { "epoch": 16.25, "learning_rate": 2.6953571428571427e-05, "loss": 1.5295, "step": 3226500 }, { "epoch": 16.26, "learning_rate": 2.6950000000000005e-05, "loss": 1.5236, "step": 3227000 }, { "epoch": 16.26, "learning_rate": 2.6946428571428572e-05, "loss": 1.5282, "step": 3227500 }, { "epoch": 16.26, "learning_rate": 2.6942857142857143e-05, "loss": 1.5295, "step": 3228000 }, { "epoch": 16.27, "learning_rate": 2.6939285714285717e-05, "loss": 1.5208, "step": 3228500 }, { "epoch": 16.27, "learning_rate": 2.6935714285714288e-05, "loss": 1.5192, "step": 3229000 }, { "epoch": 16.27, "learning_rate": 2.6932142857142855e-05, "loss": 1.5293, "step": 3229500 }, { "epoch": 16.27, "learning_rate": 2.692857142857143e-05, "loss": 1.5286, "step": 3230000 }, { "epoch": 16.28, "learning_rate": 2.6925e-05, "loss": 1.528, "step": 3230500 }, { "epoch": 16.28, "learning_rate": 2.6921428571428575e-05, "loss": 1.5259, "step": 3231000 }, { "epoch": 16.28, "learning_rate": 2.6917857142857146e-05, "loss": 1.5149, "step": 3231500 }, { "epoch": 16.28, "learning_rate": 2.6914285714285713e-05, "loss": 1.5289, "step": 3232000 }, { "epoch": 16.29, "learning_rate": 2.691071428571429e-05, "loss": 1.5228, "step": 3232500 }, { "epoch": 16.29, "learning_rate": 2.6907142857142858e-05, "loss": 1.5295, "step": 3233000 }, { "epoch": 16.29, "learning_rate": 2.690357142857143e-05, "loss": 1.534, "step": 3233500 }, { "epoch": 16.29, "learning_rate": 2.6900000000000003e-05, "loss": 1.5382, "step": 3234000 }, { "epoch": 16.3, "learning_rate": 2.689642857142857e-05, "loss": 1.5319, "step": 3234500 }, { "epoch": 16.3, "learning_rate": 2.689285714285714e-05, "loss": 1.5142, "step": 3235000 }, { "epoch": 16.3, "learning_rate": 2.6889285714285716e-05, "loss": 1.532, "step": 3235500 }, { "epoch": 16.31, "learning_rate": 2.6885714285714287e-05, "loss": 1.5194, "step": 3236000 }, { "epoch": 16.31, "learning_rate": 2.6882142857142854e-05, "loss": 1.5237, "step": 3236500 }, { "epoch": 16.31, "learning_rate": 2.6878571428571432e-05, "loss": 1.5263, "step": 3237000 }, { "epoch": 16.31, "learning_rate": 2.6875e-05, "loss": 1.5335, "step": 3237500 }, { "epoch": 16.32, "learning_rate": 2.6871428571428574e-05, "loss": 1.5212, "step": 3238000 }, { "epoch": 16.32, "learning_rate": 2.6867857142857144e-05, "loss": 1.5226, "step": 3238500 }, { "epoch": 16.32, "learning_rate": 2.6864285714285715e-05, "loss": 1.5163, "step": 3239000 }, { "epoch": 16.32, "learning_rate": 2.686071428571429e-05, "loss": 1.531, "step": 3239500 }, { "epoch": 16.33, "learning_rate": 2.6857142857142857e-05, "loss": 1.5251, "step": 3240000 }, { "epoch": 16.33, "learning_rate": 2.6853571428571428e-05, "loss": 1.5267, "step": 3240500 }, { "epoch": 16.33, "learning_rate": 2.6850000000000002e-05, "loss": 1.5207, "step": 3241000 }, { "epoch": 16.33, "learning_rate": 2.6846428571428573e-05, "loss": 1.5335, "step": 3241500 }, { "epoch": 16.34, "learning_rate": 2.684285714285714e-05, "loss": 1.5304, "step": 3242000 }, { "epoch": 16.34, "learning_rate": 2.6839285714285718e-05, "loss": 1.5256, "step": 3242500 }, { "epoch": 16.34, "learning_rate": 2.6835714285714286e-05, "loss": 1.5292, "step": 3243000 }, { "epoch": 16.35, "learning_rate": 2.6832142857142856e-05, "loss": 1.5283, "step": 3243500 }, { "epoch": 16.35, "learning_rate": 2.682857142857143e-05, "loss": 1.5173, "step": 3244000 }, { "epoch": 16.35, "learning_rate": 2.6825e-05, "loss": 1.5198, "step": 3244500 }, { "epoch": 16.35, "learning_rate": 2.6821428571428576e-05, "loss": 1.5222, "step": 3245000 }, { "epoch": 16.36, "learning_rate": 2.6817857142857143e-05, "loss": 1.5255, "step": 3245500 }, { "epoch": 16.36, "learning_rate": 2.6814285714285714e-05, "loss": 1.5245, "step": 3246000 }, { "epoch": 16.36, "learning_rate": 2.681071428571429e-05, "loss": 1.5247, "step": 3246500 }, { "epoch": 16.36, "learning_rate": 2.680714285714286e-05, "loss": 1.5242, "step": 3247000 }, { "epoch": 16.37, "learning_rate": 2.6803571428571427e-05, "loss": 1.5285, "step": 3247500 }, { "epoch": 16.37, "learning_rate": 2.6800000000000004e-05, "loss": 1.5309, "step": 3248000 }, { "epoch": 16.37, "learning_rate": 2.6796428571428572e-05, "loss": 1.5313, "step": 3248500 }, { "epoch": 16.37, "learning_rate": 2.6792857142857143e-05, "loss": 1.5456, "step": 3249000 }, { "epoch": 16.38, "learning_rate": 2.6789285714285717e-05, "loss": 1.519, "step": 3249500 }, { "epoch": 16.38, "learning_rate": 2.6785714285714288e-05, "loss": 1.5273, "step": 3250000 }, { "epoch": 16.38, "learning_rate": 2.6782142857142862e-05, "loss": 1.514, "step": 3250500 }, { "epoch": 16.39, "learning_rate": 2.677857142857143e-05, "loss": 1.5245, "step": 3251000 }, { "epoch": 16.39, "learning_rate": 2.6775e-05, "loss": 1.5279, "step": 3251500 }, { "epoch": 16.39, "learning_rate": 2.6771428571428575e-05, "loss": 1.5251, "step": 3252000 }, { "epoch": 16.39, "learning_rate": 2.6767857142857145e-05, "loss": 1.5258, "step": 3252500 }, { "epoch": 16.4, "learning_rate": 2.6764285714285713e-05, "loss": 1.5228, "step": 3253000 }, { "epoch": 16.4, "learning_rate": 2.676071428571429e-05, "loss": 1.5232, "step": 3253500 }, { "epoch": 16.4, "learning_rate": 2.6757142857142858e-05, "loss": 1.5188, "step": 3254000 }, { "epoch": 16.4, "learning_rate": 2.675357142857143e-05, "loss": 1.533, "step": 3254500 }, { "epoch": 16.41, "learning_rate": 2.6750000000000003e-05, "loss": 1.5285, "step": 3255000 }, { "epoch": 16.41, "learning_rate": 2.6746428571428574e-05, "loss": 1.5271, "step": 3255500 }, { "epoch": 16.41, "learning_rate": 2.674285714285714e-05, "loss": 1.5185, "step": 3256000 }, { "epoch": 16.41, "learning_rate": 2.6739285714285716e-05, "loss": 1.5302, "step": 3256500 }, { "epoch": 16.42, "learning_rate": 2.6735714285714287e-05, "loss": 1.5311, "step": 3257000 }, { "epoch": 16.42, "learning_rate": 2.673214285714286e-05, "loss": 1.5226, "step": 3257500 }, { "epoch": 16.42, "learning_rate": 2.672857142857143e-05, "loss": 1.5302, "step": 3258000 }, { "epoch": 16.43, "learning_rate": 2.6725e-05, "loss": 1.5227, "step": 3258500 }, { "epoch": 16.43, "learning_rate": 2.6721428571428577e-05, "loss": 1.5225, "step": 3259000 }, { "epoch": 16.43, "learning_rate": 2.6717857142857144e-05, "loss": 1.5361, "step": 3259500 }, { "epoch": 16.43, "learning_rate": 2.6714285714285715e-05, "loss": 1.5335, "step": 3260000 }, { "epoch": 16.44, "learning_rate": 2.671071428571429e-05, "loss": 1.5261, "step": 3260500 }, { "epoch": 16.44, "learning_rate": 2.6707142857142857e-05, "loss": 1.5123, "step": 3261000 }, { "epoch": 16.44, "learning_rate": 2.6703571428571428e-05, "loss": 1.5188, "step": 3261500 }, { "epoch": 16.44, "learning_rate": 2.6700000000000002e-05, "loss": 1.531, "step": 3262000 }, { "epoch": 16.45, "learning_rate": 2.6696428571428573e-05, "loss": 1.5288, "step": 3262500 }, { "epoch": 16.45, "learning_rate": 2.669285714285714e-05, "loss": 1.5223, "step": 3263000 }, { "epoch": 16.45, "learning_rate": 2.6689285714285718e-05, "loss": 1.5343, "step": 3263500 }, { "epoch": 16.45, "learning_rate": 2.6685714285714285e-05, "loss": 1.5224, "step": 3264000 }, { "epoch": 16.46, "learning_rate": 2.668214285714286e-05, "loss": 1.5358, "step": 3264500 }, { "epoch": 16.46, "learning_rate": 2.667857142857143e-05, "loss": 1.5257, "step": 3265000 }, { "epoch": 16.46, "learning_rate": 2.6675e-05, "loss": 1.5245, "step": 3265500 }, { "epoch": 16.47, "learning_rate": 2.6671428571428576e-05, "loss": 1.5278, "step": 3266000 }, { "epoch": 16.47, "learning_rate": 2.6667857142857143e-05, "loss": 1.5178, "step": 3266500 }, { "epoch": 16.47, "learning_rate": 2.6664285714285714e-05, "loss": 1.5255, "step": 3267000 }, { "epoch": 16.47, "learning_rate": 2.6660714285714288e-05, "loss": 1.5255, "step": 3267500 }, { "epoch": 16.48, "learning_rate": 2.665714285714286e-05, "loss": 1.5285, "step": 3268000 }, { "epoch": 16.48, "learning_rate": 2.6653571428571426e-05, "loss": 1.5293, "step": 3268500 }, { "epoch": 16.48, "learning_rate": 2.6650000000000004e-05, "loss": 1.528, "step": 3269000 }, { "epoch": 16.48, "learning_rate": 2.664642857142857e-05, "loss": 1.5168, "step": 3269500 }, { "epoch": 16.49, "learning_rate": 2.6642857142857142e-05, "loss": 1.5082, "step": 3270000 }, { "epoch": 16.49, "learning_rate": 2.6639285714285717e-05, "loss": 1.5196, "step": 3270500 }, { "epoch": 16.49, "learning_rate": 2.6635714285714287e-05, "loss": 1.5184, "step": 3271000 }, { "epoch": 16.49, "learning_rate": 2.6632142857142862e-05, "loss": 1.5296, "step": 3271500 }, { "epoch": 16.5, "learning_rate": 2.662857142857143e-05, "loss": 1.5169, "step": 3272000 }, { "epoch": 16.5, "learning_rate": 2.6625e-05, "loss": 1.5327, "step": 3272500 }, { "epoch": 16.5, "learning_rate": 2.6621428571428574e-05, "loss": 1.5219, "step": 3273000 }, { "epoch": 16.51, "learning_rate": 2.6617857142857145e-05, "loss": 1.5249, "step": 3273500 }, { "epoch": 16.51, "learning_rate": 2.6614285714285713e-05, "loss": 1.5189, "step": 3274000 }, { "epoch": 16.51, "learning_rate": 2.661071428571429e-05, "loss": 1.5183, "step": 3274500 }, { "epoch": 16.51, "learning_rate": 2.6607142857142858e-05, "loss": 1.5178, "step": 3275000 }, { "epoch": 16.52, "learning_rate": 2.660357142857143e-05, "loss": 1.5157, "step": 3275500 }, { "epoch": 16.52, "learning_rate": 2.6600000000000003e-05, "loss": 1.5212, "step": 3276000 }, { "epoch": 16.52, "learning_rate": 2.6596428571428574e-05, "loss": 1.5232, "step": 3276500 }, { "epoch": 16.52, "learning_rate": 2.659285714285714e-05, "loss": 1.5257, "step": 3277000 }, { "epoch": 16.53, "learning_rate": 2.6589285714285715e-05, "loss": 1.5174, "step": 3277500 }, { "epoch": 16.53, "learning_rate": 2.6585714285714286e-05, "loss": 1.5369, "step": 3278000 }, { "epoch": 16.53, "learning_rate": 2.658214285714286e-05, "loss": 1.5362, "step": 3278500 }, { "epoch": 16.53, "learning_rate": 2.657857142857143e-05, "loss": 1.5089, "step": 3279000 }, { "epoch": 16.54, "learning_rate": 2.6575e-05, "loss": 1.5181, "step": 3279500 }, { "epoch": 16.54, "learning_rate": 2.6571428571428576e-05, "loss": 1.5193, "step": 3280000 }, { "epoch": 16.54, "learning_rate": 2.6567857142857144e-05, "loss": 1.5201, "step": 3280500 }, { "epoch": 16.55, "learning_rate": 2.6564285714285715e-05, "loss": 1.53, "step": 3281000 }, { "epoch": 16.55, "learning_rate": 2.656071428571429e-05, "loss": 1.5222, "step": 3281500 }, { "epoch": 16.55, "learning_rate": 2.655714285714286e-05, "loss": 1.5185, "step": 3282000 }, { "epoch": 16.55, "learning_rate": 2.6553571428571427e-05, "loss": 1.5271, "step": 3282500 }, { "epoch": 16.56, "learning_rate": 2.655e-05, "loss": 1.5249, "step": 3283000 }, { "epoch": 16.56, "learning_rate": 2.6546428571428572e-05, "loss": 1.5203, "step": 3283500 }, { "epoch": 16.56, "learning_rate": 2.654285714285714e-05, "loss": 1.5206, "step": 3284000 }, { "epoch": 16.56, "learning_rate": 2.6539285714285718e-05, "loss": 1.5174, "step": 3284500 }, { "epoch": 16.57, "learning_rate": 2.6535714285714285e-05, "loss": 1.518, "step": 3285000 }, { "epoch": 16.57, "learning_rate": 2.653214285714286e-05, "loss": 1.5243, "step": 3285500 }, { "epoch": 16.57, "learning_rate": 2.652857142857143e-05, "loss": 1.5272, "step": 3286000 }, { "epoch": 16.57, "learning_rate": 2.6525e-05, "loss": 1.5256, "step": 3286500 }, { "epoch": 16.58, "learning_rate": 2.6521428571428575e-05, "loss": 1.5249, "step": 3287000 }, { "epoch": 16.58, "learning_rate": 2.6517857142857143e-05, "loss": 1.5207, "step": 3287500 }, { "epoch": 16.58, "learning_rate": 2.6514285714285714e-05, "loss": 1.5416, "step": 3288000 }, { "epoch": 16.59, "learning_rate": 2.6510714285714288e-05, "loss": 1.525, "step": 3288500 }, { "epoch": 16.59, "learning_rate": 2.650714285714286e-05, "loss": 1.5122, "step": 3289000 }, { "epoch": 16.59, "learning_rate": 2.6503571428571426e-05, "loss": 1.5198, "step": 3289500 }, { "epoch": 16.59, "learning_rate": 2.6500000000000004e-05, "loss": 1.5104, "step": 3290000 }, { "epoch": 16.6, "learning_rate": 2.649642857142857e-05, "loss": 1.5223, "step": 3290500 }, { "epoch": 16.6, "learning_rate": 2.6492857142857142e-05, "loss": 1.5222, "step": 3291000 }, { "epoch": 16.6, "learning_rate": 2.6489285714285716e-05, "loss": 1.5242, "step": 3291500 }, { "epoch": 16.6, "learning_rate": 2.6485714285714287e-05, "loss": 1.5243, "step": 3292000 }, { "epoch": 16.61, "learning_rate": 2.648214285714286e-05, "loss": 1.5137, "step": 3292500 }, { "epoch": 16.61, "learning_rate": 2.647857142857143e-05, "loss": 1.5163, "step": 3293000 }, { "epoch": 16.61, "learning_rate": 2.6475e-05, "loss": 1.5173, "step": 3293500 }, { "epoch": 16.61, "learning_rate": 2.6471428571428574e-05, "loss": 1.5116, "step": 3294000 }, { "epoch": 16.62, "learning_rate": 2.6467857142857145e-05, "loss": 1.5218, "step": 3294500 }, { "epoch": 16.62, "learning_rate": 2.6464285714285712e-05, "loss": 1.5374, "step": 3295000 }, { "epoch": 16.62, "learning_rate": 2.646071428571429e-05, "loss": 1.526, "step": 3295500 }, { "epoch": 16.63, "learning_rate": 2.6457142857142857e-05, "loss": 1.5135, "step": 3296000 }, { "epoch": 16.63, "learning_rate": 2.645357142857143e-05, "loss": 1.5308, "step": 3296500 }, { "epoch": 16.63, "learning_rate": 2.6450000000000003e-05, "loss": 1.5168, "step": 3297000 }, { "epoch": 16.63, "learning_rate": 2.6446428571428573e-05, "loss": 1.5284, "step": 3297500 }, { "epoch": 16.64, "learning_rate": 2.644285714285714e-05, "loss": 1.5173, "step": 3298000 }, { "epoch": 16.64, "learning_rate": 2.6439285714285715e-05, "loss": 1.5215, "step": 3298500 }, { "epoch": 16.64, "learning_rate": 2.6435714285714286e-05, "loss": 1.5161, "step": 3299000 }, { "epoch": 16.64, "learning_rate": 2.643214285714286e-05, "loss": 1.5143, "step": 3299500 }, { "epoch": 16.65, "learning_rate": 2.642857142857143e-05, "loss": 1.5333, "step": 3300000 }, { "epoch": 16.65, "learning_rate": 2.6425e-05, "loss": 1.5279, "step": 3300500 }, { "epoch": 16.65, "learning_rate": 2.6421428571428576e-05, "loss": 1.5119, "step": 3301000 }, { "epoch": 16.65, "learning_rate": 2.6417857142857144e-05, "loss": 1.51, "step": 3301500 }, { "epoch": 16.66, "learning_rate": 2.6414285714285715e-05, "loss": 1.5165, "step": 3302000 }, { "epoch": 16.66, "learning_rate": 2.641071428571429e-05, "loss": 1.5169, "step": 3302500 }, { "epoch": 16.66, "learning_rate": 2.640714285714286e-05, "loss": 1.5164, "step": 3303000 }, { "epoch": 16.66, "learning_rate": 2.6403571428571427e-05, "loss": 1.5212, "step": 3303500 }, { "epoch": 16.67, "learning_rate": 2.64e-05, "loss": 1.5266, "step": 3304000 }, { "epoch": 16.67, "learning_rate": 2.6396428571428572e-05, "loss": 1.5256, "step": 3304500 }, { "epoch": 16.67, "learning_rate": 2.6392857142857143e-05, "loss": 1.5197, "step": 3305000 }, { "epoch": 16.68, "learning_rate": 2.6389285714285717e-05, "loss": 1.527, "step": 3305500 }, { "epoch": 16.68, "learning_rate": 2.6385714285714285e-05, "loss": 1.5352, "step": 3306000 }, { "epoch": 16.68, "learning_rate": 2.6382142857142862e-05, "loss": 1.5216, "step": 3306500 }, { "epoch": 16.68, "learning_rate": 2.637857142857143e-05, "loss": 1.518, "step": 3307000 }, { "epoch": 16.69, "learning_rate": 2.6375e-05, "loss": 1.5136, "step": 3307500 }, { "epoch": 16.69, "learning_rate": 2.6371428571428575e-05, "loss": 1.525, "step": 3308000 }, { "epoch": 16.69, "learning_rate": 2.6367857142857142e-05, "loss": 1.5263, "step": 3308500 }, { "epoch": 16.69, "learning_rate": 2.6364285714285713e-05, "loss": 1.5181, "step": 3309000 }, { "epoch": 16.7, "learning_rate": 2.6360714285714288e-05, "loss": 1.5251, "step": 3309500 }, { "epoch": 16.7, "learning_rate": 2.635714285714286e-05, "loss": 1.5058, "step": 3310000 }, { "epoch": 16.7, "learning_rate": 2.6353571428571426e-05, "loss": 1.5113, "step": 3310500 }, { "epoch": 16.7, "learning_rate": 2.6350000000000004e-05, "loss": 1.5258, "step": 3311000 }, { "epoch": 16.71, "learning_rate": 2.634642857142857e-05, "loss": 1.5166, "step": 3311500 }, { "epoch": 16.71, "learning_rate": 2.6342857142857142e-05, "loss": 1.529, "step": 3312000 }, { "epoch": 16.71, "learning_rate": 2.6339285714285716e-05, "loss": 1.5143, "step": 3312500 }, { "epoch": 16.72, "learning_rate": 2.6335714285714287e-05, "loss": 1.5198, "step": 3313000 }, { "epoch": 16.72, "learning_rate": 2.633214285714286e-05, "loss": 1.5196, "step": 3313500 }, { "epoch": 16.72, "learning_rate": 2.632857142857143e-05, "loss": 1.5164, "step": 3314000 }, { "epoch": 16.72, "learning_rate": 2.6325e-05, "loss": 1.5225, "step": 3314500 }, { "epoch": 16.73, "learning_rate": 2.6321428571428574e-05, "loss": 1.5216, "step": 3315000 }, { "epoch": 16.73, "learning_rate": 2.6317857142857145e-05, "loss": 1.5136, "step": 3315500 }, { "epoch": 16.73, "learning_rate": 2.6314285714285712e-05, "loss": 1.5123, "step": 3316000 }, { "epoch": 16.73, "learning_rate": 2.631071428571429e-05, "loss": 1.5232, "step": 3316500 }, { "epoch": 16.74, "learning_rate": 2.6307142857142857e-05, "loss": 1.5139, "step": 3317000 }, { "epoch": 16.74, "learning_rate": 2.6303571428571428e-05, "loss": 1.5281, "step": 3317500 }, { "epoch": 16.74, "learning_rate": 2.6300000000000002e-05, "loss": 1.5181, "step": 3318000 }, { "epoch": 16.74, "learning_rate": 2.6296428571428573e-05, "loss": 1.5162, "step": 3318500 }, { "epoch": 16.75, "learning_rate": 2.6292857142857147e-05, "loss": 1.5184, "step": 3319000 }, { "epoch": 16.75, "learning_rate": 2.6289285714285715e-05, "loss": 1.5244, "step": 3319500 }, { "epoch": 16.75, "learning_rate": 2.6285714285714286e-05, "loss": 1.5173, "step": 3320000 }, { "epoch": 16.76, "learning_rate": 2.628214285714286e-05, "loss": 1.53, "step": 3320500 }, { "epoch": 16.76, "learning_rate": 2.627857142857143e-05, "loss": 1.5166, "step": 3321000 }, { "epoch": 16.76, "learning_rate": 2.6275e-05, "loss": 1.5258, "step": 3321500 }, { "epoch": 16.76, "learning_rate": 2.6271428571428576e-05, "loss": 1.5136, "step": 3322000 }, { "epoch": 16.77, "learning_rate": 2.6267857142857143e-05, "loss": 1.5259, "step": 3322500 }, { "epoch": 16.77, "learning_rate": 2.6264285714285714e-05, "loss": 1.5232, "step": 3323000 }, { "epoch": 16.77, "learning_rate": 2.626071428571429e-05, "loss": 1.5103, "step": 3323500 }, { "epoch": 16.77, "learning_rate": 2.625714285714286e-05, "loss": 1.5227, "step": 3324000 }, { "epoch": 16.78, "learning_rate": 2.6253571428571427e-05, "loss": 1.5281, "step": 3324500 }, { "epoch": 16.78, "learning_rate": 2.625e-05, "loss": 1.5181, "step": 3325000 }, { "epoch": 16.78, "learning_rate": 2.6246428571428572e-05, "loss": 1.5196, "step": 3325500 }, { "epoch": 16.78, "learning_rate": 2.6242857142857146e-05, "loss": 1.5237, "step": 3326000 }, { "epoch": 16.79, "learning_rate": 2.6239285714285717e-05, "loss": 1.5348, "step": 3326500 }, { "epoch": 16.79, "learning_rate": 2.6235714285714285e-05, "loss": 1.5144, "step": 3327000 }, { "epoch": 16.79, "learning_rate": 2.6232142857142862e-05, "loss": 1.5257, "step": 3327500 }, { "epoch": 16.8, "learning_rate": 2.622857142857143e-05, "loss": 1.5137, "step": 3328000 }, { "epoch": 16.8, "learning_rate": 2.6225e-05, "loss": 1.5134, "step": 3328500 }, { "epoch": 16.8, "learning_rate": 2.6221428571428575e-05, "loss": 1.5182, "step": 3329000 }, { "epoch": 16.8, "learning_rate": 2.6217857142857146e-05, "loss": 1.5132, "step": 3329500 }, { "epoch": 16.81, "learning_rate": 2.6214285714285713e-05, "loss": 1.5091, "step": 3330000 }, { "epoch": 16.81, "learning_rate": 2.6210714285714287e-05, "loss": 1.5236, "step": 3330500 }, { "epoch": 16.81, "learning_rate": 2.6207142857142858e-05, "loss": 1.5198, "step": 3331000 }, { "epoch": 16.81, "learning_rate": 2.620357142857143e-05, "loss": 1.5222, "step": 3331500 }, { "epoch": 16.82, "learning_rate": 2.6200000000000003e-05, "loss": 1.5101, "step": 3332000 }, { "epoch": 16.82, "learning_rate": 2.619642857142857e-05, "loss": 1.5136, "step": 3332500 }, { "epoch": 16.82, "learning_rate": 2.619285714285715e-05, "loss": 1.5165, "step": 3333000 }, { "epoch": 16.82, "learning_rate": 2.6189285714285716e-05, "loss": 1.5162, "step": 3333500 }, { "epoch": 16.83, "learning_rate": 2.6185714285714287e-05, "loss": 1.5181, "step": 3334000 }, { "epoch": 16.83, "learning_rate": 2.618214285714286e-05, "loss": 1.5203, "step": 3334500 }, { "epoch": 16.83, "learning_rate": 2.617857142857143e-05, "loss": 1.5223, "step": 3335000 }, { "epoch": 16.84, "learning_rate": 2.6175e-05, "loss": 1.5185, "step": 3335500 }, { "epoch": 16.84, "learning_rate": 2.6171428571428574e-05, "loss": 1.5192, "step": 3336000 }, { "epoch": 16.84, "learning_rate": 2.6167857142857144e-05, "loss": 1.5028, "step": 3336500 }, { "epoch": 16.84, "learning_rate": 2.6164285714285712e-05, "loss": 1.5243, "step": 3337000 }, { "epoch": 16.85, "learning_rate": 2.616071428571429e-05, "loss": 1.5152, "step": 3337500 }, { "epoch": 16.85, "learning_rate": 2.6157142857142857e-05, "loss": 1.5137, "step": 3338000 }, { "epoch": 16.85, "learning_rate": 2.6153571428571428e-05, "loss": 1.5217, "step": 3338500 }, { "epoch": 16.85, "learning_rate": 2.6150000000000002e-05, "loss": 1.518, "step": 3339000 }, { "epoch": 16.86, "learning_rate": 2.6146428571428573e-05, "loss": 1.5203, "step": 3339500 }, { "epoch": 16.86, "learning_rate": 2.6142857142857147e-05, "loss": 1.5218, "step": 3340000 }, { "epoch": 16.86, "learning_rate": 2.6139285714285715e-05, "loss": 1.5191, "step": 3340500 }, { "epoch": 16.86, "learning_rate": 2.6135714285714285e-05, "loss": 1.5113, "step": 3341000 }, { "epoch": 16.87, "learning_rate": 2.613214285714286e-05, "loss": 1.5155, "step": 3341500 }, { "epoch": 16.87, "learning_rate": 2.612857142857143e-05, "loss": 1.5225, "step": 3342000 }, { "epoch": 16.87, "learning_rate": 2.6124999999999998e-05, "loss": 1.5111, "step": 3342500 }, { "epoch": 16.88, "learning_rate": 2.6121428571428576e-05, "loss": 1.5069, "step": 3343000 }, { "epoch": 16.88, "learning_rate": 2.6117857142857143e-05, "loss": 1.5147, "step": 3343500 }, { "epoch": 16.88, "learning_rate": 2.6114285714285714e-05, "loss": 1.5237, "step": 3344000 }, { "epoch": 16.88, "learning_rate": 2.6110714285714288e-05, "loss": 1.5181, "step": 3344500 }, { "epoch": 16.89, "learning_rate": 2.610714285714286e-05, "loss": 1.5129, "step": 3345000 }, { "epoch": 16.89, "learning_rate": 2.6103571428571427e-05, "loss": 1.5201, "step": 3345500 }, { "epoch": 16.89, "learning_rate": 2.61e-05, "loss": 1.5104, "step": 3346000 }, { "epoch": 16.89, "learning_rate": 2.609642857142857e-05, "loss": 1.5094, "step": 3346500 }, { "epoch": 16.9, "learning_rate": 2.6092857142857146e-05, "loss": 1.5221, "step": 3347000 }, { "epoch": 16.9, "learning_rate": 2.6089285714285717e-05, "loss": 1.5094, "step": 3347500 }, { "epoch": 16.9, "learning_rate": 2.6085714285714284e-05, "loss": 1.5094, "step": 3348000 }, { "epoch": 16.9, "learning_rate": 2.6082142857142862e-05, "loss": 1.5147, "step": 3348500 }, { "epoch": 16.91, "learning_rate": 2.607857142857143e-05, "loss": 1.5198, "step": 3349000 }, { "epoch": 16.91, "learning_rate": 2.6075e-05, "loss": 1.5159, "step": 3349500 }, { "epoch": 16.91, "learning_rate": 2.6071428571428574e-05, "loss": 1.5139, "step": 3350000 }, { "epoch": 16.92, "learning_rate": 2.6067857142857145e-05, "loss": 1.5131, "step": 3350500 }, { "epoch": 16.92, "learning_rate": 2.6064285714285713e-05, "loss": 1.5099, "step": 3351000 }, { "epoch": 16.92, "learning_rate": 2.6060714285714287e-05, "loss": 1.5168, "step": 3351500 }, { "epoch": 16.92, "learning_rate": 2.6057142857142858e-05, "loss": 1.5148, "step": 3352000 }, { "epoch": 16.93, "learning_rate": 2.605357142857143e-05, "loss": 1.5154, "step": 3352500 }, { "epoch": 16.93, "learning_rate": 2.6050000000000003e-05, "loss": 1.5084, "step": 3353000 }, { "epoch": 16.93, "learning_rate": 2.604642857142857e-05, "loss": 1.5128, "step": 3353500 }, { "epoch": 16.93, "learning_rate": 2.6042857142857148e-05, "loss": 1.5188, "step": 3354000 }, { "epoch": 16.94, "learning_rate": 2.6039285714285716e-05, "loss": 1.5246, "step": 3354500 }, { "epoch": 16.94, "learning_rate": 2.6035714285714286e-05, "loss": 1.5069, "step": 3355000 }, { "epoch": 16.94, "learning_rate": 2.603214285714286e-05, "loss": 1.519, "step": 3355500 }, { "epoch": 16.94, "learning_rate": 2.602857142857143e-05, "loss": 1.5188, "step": 3356000 }, { "epoch": 16.95, "learning_rate": 2.6025e-05, "loss": 1.5134, "step": 3356500 }, { "epoch": 16.95, "learning_rate": 2.6021428571428573e-05, "loss": 1.5143, "step": 3357000 }, { "epoch": 16.95, "learning_rate": 2.6017857142857144e-05, "loss": 1.5174, "step": 3357500 }, { "epoch": 16.96, "learning_rate": 2.601428571428571e-05, "loss": 1.5171, "step": 3358000 }, { "epoch": 16.96, "learning_rate": 2.601071428571429e-05, "loss": 1.5118, "step": 3358500 }, { "epoch": 16.96, "learning_rate": 2.6007142857142857e-05, "loss": 1.5113, "step": 3359000 }, { "epoch": 16.96, "learning_rate": 2.6003571428571428e-05, "loss": 1.5196, "step": 3359500 }, { "epoch": 16.97, "learning_rate": 2.6000000000000002e-05, "loss": 1.5162, "step": 3360000 }, { "epoch": 16.97, "learning_rate": 2.5996428571428573e-05, "loss": 1.5207, "step": 3360500 }, { "epoch": 16.97, "learning_rate": 2.5992857142857147e-05, "loss": 1.5042, "step": 3361000 }, { "epoch": 16.97, "learning_rate": 2.5989285714285714e-05, "loss": 1.5226, "step": 3361500 }, { "epoch": 16.98, "learning_rate": 2.5985714285714285e-05, "loss": 1.5161, "step": 3362000 }, { "epoch": 16.98, "learning_rate": 2.598214285714286e-05, "loss": 1.5119, "step": 3362500 }, { "epoch": 16.98, "learning_rate": 2.597857142857143e-05, "loss": 1.5213, "step": 3363000 }, { "epoch": 16.98, "learning_rate": 2.5974999999999998e-05, "loss": 1.5093, "step": 3363500 }, { "epoch": 16.99, "learning_rate": 2.5971428571428575e-05, "loss": 1.5195, "step": 3364000 }, { "epoch": 16.99, "learning_rate": 2.5967857142857143e-05, "loss": 1.5197, "step": 3364500 }, { "epoch": 16.99, "learning_rate": 2.5964285714285714e-05, "loss": 1.5152, "step": 3365000 }, { "epoch": 17.0, "learning_rate": 2.5960714285714288e-05, "loss": 1.5214, "step": 3365500 }, { "epoch": 17.0, "learning_rate": 2.595714285714286e-05, "loss": 1.5186, "step": 3366000 }, { "epoch": 17.0, "learning_rate": 2.5953571428571426e-05, "loss": 1.513, "step": 3366500 }, { "epoch": 17.0, "learning_rate": 2.595e-05, "loss": 1.5065, "step": 3367000 }, { "epoch": 17.01, "learning_rate": 2.594642857142857e-05, "loss": 1.5174, "step": 3367500 }, { "epoch": 17.01, "learning_rate": 2.5942857142857146e-05, "loss": 1.5164, "step": 3368000 }, { "epoch": 17.01, "learning_rate": 2.5939285714285717e-05, "loss": 1.5089, "step": 3368500 }, { "epoch": 17.01, "learning_rate": 2.5935714285714284e-05, "loss": 1.4996, "step": 3369000 }, { "epoch": 17.02, "learning_rate": 2.593214285714286e-05, "loss": 1.516, "step": 3369500 }, { "epoch": 17.02, "learning_rate": 2.592857142857143e-05, "loss": 1.5121, "step": 3370000 }, { "epoch": 17.02, "learning_rate": 2.5925e-05, "loss": 1.5157, "step": 3370500 }, { "epoch": 17.02, "learning_rate": 2.5921428571428574e-05, "loss": 1.5038, "step": 3371000 }, { "epoch": 17.03, "learning_rate": 2.5917857142857145e-05, "loss": 1.5022, "step": 3371500 }, { "epoch": 17.03, "learning_rate": 2.5914285714285713e-05, "loss": 1.5053, "step": 3372000 }, { "epoch": 17.03, "learning_rate": 2.5910714285714287e-05, "loss": 1.5074, "step": 3372500 }, { "epoch": 17.04, "learning_rate": 2.5907142857142858e-05, "loss": 1.511, "step": 3373000 }, { "epoch": 17.04, "learning_rate": 2.590357142857143e-05, "loss": 1.5123, "step": 3373500 }, { "epoch": 17.04, "learning_rate": 2.5900000000000003e-05, "loss": 1.5169, "step": 3374000 }, { "epoch": 17.04, "learning_rate": 2.589642857142857e-05, "loss": 1.5199, "step": 3374500 }, { "epoch": 17.05, "learning_rate": 2.5892857142857148e-05, "loss": 1.5111, "step": 3375000 }, { "epoch": 17.05, "learning_rate": 2.5889285714285715e-05, "loss": 1.4998, "step": 3375500 }, { "epoch": 17.05, "learning_rate": 2.5885714285714286e-05, "loss": 1.5132, "step": 3376000 }, { "epoch": 17.05, "learning_rate": 2.588214285714286e-05, "loss": 1.5097, "step": 3376500 }, { "epoch": 17.06, "learning_rate": 2.587857142857143e-05, "loss": 1.5154, "step": 3377000 }, { "epoch": 17.06, "learning_rate": 2.5875e-05, "loss": 1.509, "step": 3377500 }, { "epoch": 17.06, "learning_rate": 2.5871428571428573e-05, "loss": 1.5051, "step": 3378000 }, { "epoch": 17.06, "learning_rate": 2.5867857142857144e-05, "loss": 1.5193, "step": 3378500 }, { "epoch": 17.07, "learning_rate": 2.5864285714285715e-05, "loss": 1.5041, "step": 3379000 }, { "epoch": 17.07, "learning_rate": 2.586071428571429e-05, "loss": 1.4923, "step": 3379500 }, { "epoch": 17.07, "learning_rate": 2.5857142857142856e-05, "loss": 1.5151, "step": 3380000 }, { "epoch": 17.07, "learning_rate": 2.5853571428571434e-05, "loss": 1.5133, "step": 3380500 }, { "epoch": 17.08, "learning_rate": 2.585e-05, "loss": 1.5168, "step": 3381000 }, { "epoch": 17.08, "learning_rate": 2.5846428571428572e-05, "loss": 1.5052, "step": 3381500 }, { "epoch": 17.08, "learning_rate": 2.5842857142857147e-05, "loss": 1.5152, "step": 3382000 }, { "epoch": 17.09, "learning_rate": 2.5839285714285717e-05, "loss": 1.5121, "step": 3382500 }, { "epoch": 17.09, "learning_rate": 2.5835714285714285e-05, "loss": 1.5112, "step": 3383000 }, { "epoch": 17.09, "learning_rate": 2.583214285714286e-05, "loss": 1.5088, "step": 3383500 }, { "epoch": 17.09, "learning_rate": 2.582857142857143e-05, "loss": 1.5147, "step": 3384000 }, { "epoch": 17.1, "learning_rate": 2.5824999999999998e-05, "loss": 1.5183, "step": 3384500 }, { "epoch": 17.1, "learning_rate": 2.5821428571428575e-05, "loss": 1.5143, "step": 3385000 }, { "epoch": 17.1, "learning_rate": 2.5817857142857143e-05, "loss": 1.5071, "step": 3385500 }, { "epoch": 17.1, "learning_rate": 2.5814285714285713e-05, "loss": 1.5198, "step": 3386000 }, { "epoch": 17.11, "learning_rate": 2.5810714285714288e-05, "loss": 1.5179, "step": 3386500 }, { "epoch": 17.11, "learning_rate": 2.580714285714286e-05, "loss": 1.5084, "step": 3387000 }, { "epoch": 17.11, "learning_rate": 2.5803571428571433e-05, "loss": 1.5104, "step": 3387500 }, { "epoch": 17.11, "learning_rate": 2.58e-05, "loss": 1.512, "step": 3388000 }, { "epoch": 17.12, "learning_rate": 2.579642857142857e-05, "loss": 1.5127, "step": 3388500 }, { "epoch": 17.12, "learning_rate": 2.5792857142857145e-05, "loss": 1.5172, "step": 3389000 }, { "epoch": 17.12, "learning_rate": 2.5789285714285716e-05, "loss": 1.51, "step": 3389500 }, { "epoch": 17.13, "learning_rate": 2.5785714285714284e-05, "loss": 1.513, "step": 3390000 }, { "epoch": 17.13, "learning_rate": 2.578214285714286e-05, "loss": 1.5044, "step": 3390500 }, { "epoch": 17.13, "learning_rate": 2.577857142857143e-05, "loss": 1.5129, "step": 3391000 }, { "epoch": 17.13, "learning_rate": 2.5775e-05, "loss": 1.5171, "step": 3391500 }, { "epoch": 17.14, "learning_rate": 2.5771428571428574e-05, "loss": 1.5125, "step": 3392000 }, { "epoch": 17.14, "learning_rate": 2.5767857142857145e-05, "loss": 1.5088, "step": 3392500 }, { "epoch": 17.14, "learning_rate": 2.5764285714285712e-05, "loss": 1.5052, "step": 3393000 }, { "epoch": 17.14, "learning_rate": 2.5760714285714287e-05, "loss": 1.51, "step": 3393500 }, { "epoch": 17.15, "learning_rate": 2.5757142857142857e-05, "loss": 1.504, "step": 3394000 }, { "epoch": 17.15, "learning_rate": 2.575357142857143e-05, "loss": 1.5165, "step": 3394500 }, { "epoch": 17.15, "learning_rate": 2.5750000000000002e-05, "loss": 1.5035, "step": 3395000 }, { "epoch": 17.15, "learning_rate": 2.574642857142857e-05, "loss": 1.5133, "step": 3395500 }, { "epoch": 17.16, "learning_rate": 2.5742857142857148e-05, "loss": 1.504, "step": 3396000 }, { "epoch": 17.16, "learning_rate": 2.5739285714285715e-05, "loss": 1.5049, "step": 3396500 }, { "epoch": 17.16, "learning_rate": 2.5735714285714286e-05, "loss": 1.5071, "step": 3397000 }, { "epoch": 17.17, "learning_rate": 2.573214285714286e-05, "loss": 1.5158, "step": 3397500 }, { "epoch": 17.17, "learning_rate": 2.572857142857143e-05, "loss": 1.5045, "step": 3398000 }, { "epoch": 17.17, "learning_rate": 2.5725e-05, "loss": 1.5093, "step": 3398500 }, { "epoch": 17.17, "learning_rate": 2.5721428571428573e-05, "loss": 1.5122, "step": 3399000 }, { "epoch": 17.18, "learning_rate": 2.5717857142857144e-05, "loss": 1.5182, "step": 3399500 }, { "epoch": 17.18, "learning_rate": 2.5714285714285714e-05, "loss": 1.5098, "step": 3400000 }, { "epoch": 17.18, "learning_rate": 2.571071428571429e-05, "loss": 1.507, "step": 3400500 }, { "epoch": 17.18, "learning_rate": 2.5707142857142856e-05, "loss": 1.5039, "step": 3401000 }, { "epoch": 17.19, "learning_rate": 2.5703571428571434e-05, "loss": 1.5212, "step": 3401500 }, { "epoch": 17.19, "learning_rate": 2.57e-05, "loss": 1.5181, "step": 3402000 }, { "epoch": 17.19, "learning_rate": 2.5696428571428572e-05, "loss": 1.5151, "step": 3402500 }, { "epoch": 17.19, "learning_rate": 2.5692857142857146e-05, "loss": 1.5061, "step": 3403000 }, { "epoch": 17.2, "learning_rate": 2.5689285714285717e-05, "loss": 1.4972, "step": 3403500 }, { "epoch": 17.2, "learning_rate": 2.5685714285714285e-05, "loss": 1.5165, "step": 3404000 }, { "epoch": 17.2, "learning_rate": 2.568214285714286e-05, "loss": 1.5003, "step": 3404500 }, { "epoch": 17.21, "learning_rate": 2.567857142857143e-05, "loss": 1.5217, "step": 3405000 }, { "epoch": 17.21, "learning_rate": 2.5675e-05, "loss": 1.5092, "step": 3405500 }, { "epoch": 17.21, "learning_rate": 2.5671428571428575e-05, "loss": 1.5306, "step": 3406000 }, { "epoch": 17.21, "learning_rate": 2.5667857142857142e-05, "loss": 1.518, "step": 3406500 }, { "epoch": 17.22, "learning_rate": 2.5664285714285713e-05, "loss": 1.514, "step": 3407000 }, { "epoch": 17.22, "learning_rate": 2.5660714285714287e-05, "loss": 1.5114, "step": 3407500 }, { "epoch": 17.22, "learning_rate": 2.565714285714286e-05, "loss": 1.5217, "step": 3408000 }, { "epoch": 17.22, "learning_rate": 2.5653571428571433e-05, "loss": 1.5062, "step": 3408500 }, { "epoch": 17.23, "learning_rate": 2.5650000000000003e-05, "loss": 1.5084, "step": 3409000 }, { "epoch": 17.23, "learning_rate": 2.564642857142857e-05, "loss": 1.508, "step": 3409500 }, { "epoch": 17.23, "learning_rate": 2.5642857142857145e-05, "loss": 1.5129, "step": 3410000 }, { "epoch": 17.23, "learning_rate": 2.5639285714285716e-05, "loss": 1.5045, "step": 3410500 }, { "epoch": 17.24, "learning_rate": 2.5635714285714283e-05, "loss": 1.5116, "step": 3411000 }, { "epoch": 17.24, "learning_rate": 2.563214285714286e-05, "loss": 1.5139, "step": 3411500 }, { "epoch": 17.24, "learning_rate": 2.562857142857143e-05, "loss": 1.5097, "step": 3412000 }, { "epoch": 17.25, "learning_rate": 2.5625e-05, "loss": 1.5175, "step": 3412500 }, { "epoch": 17.25, "learning_rate": 2.5621428571428574e-05, "loss": 1.5111, "step": 3413000 }, { "epoch": 17.25, "learning_rate": 2.5617857142857145e-05, "loss": 1.5054, "step": 3413500 }, { "epoch": 17.25, "learning_rate": 2.5614285714285712e-05, "loss": 1.4995, "step": 3414000 }, { "epoch": 17.26, "learning_rate": 2.5610714285714286e-05, "loss": 1.5096, "step": 3414500 }, { "epoch": 17.26, "learning_rate": 2.5607142857142857e-05, "loss": 1.506, "step": 3415000 }, { "epoch": 17.26, "learning_rate": 2.560357142857143e-05, "loss": 1.5075, "step": 3415500 }, { "epoch": 17.26, "learning_rate": 2.5600000000000002e-05, "loss": 1.5185, "step": 3416000 }, { "epoch": 17.27, "learning_rate": 2.559642857142857e-05, "loss": 1.5144, "step": 3416500 }, { "epoch": 17.27, "learning_rate": 2.5592857142857147e-05, "loss": 1.5096, "step": 3417000 }, { "epoch": 17.27, "learning_rate": 2.5589285714285715e-05, "loss": 1.5098, "step": 3417500 }, { "epoch": 17.27, "learning_rate": 2.5585714285714286e-05, "loss": 1.5093, "step": 3418000 }, { "epoch": 17.28, "learning_rate": 2.558214285714286e-05, "loss": 1.514, "step": 3418500 }, { "epoch": 17.28, "learning_rate": 2.557857142857143e-05, "loss": 1.5138, "step": 3419000 }, { "epoch": 17.28, "learning_rate": 2.5574999999999998e-05, "loss": 1.5096, "step": 3419500 }, { "epoch": 17.29, "learning_rate": 2.5571428571428572e-05, "loss": 1.4998, "step": 3420000 }, { "epoch": 17.29, "learning_rate": 2.5567857142857143e-05, "loss": 1.496, "step": 3420500 }, { "epoch": 17.29, "learning_rate": 2.5564285714285714e-05, "loss": 1.4979, "step": 3421000 }, { "epoch": 17.29, "learning_rate": 2.556071428571429e-05, "loss": 1.5131, "step": 3421500 }, { "epoch": 17.3, "learning_rate": 2.5557142857142856e-05, "loss": 1.5039, "step": 3422000 }, { "epoch": 17.3, "learning_rate": 2.5553571428571434e-05, "loss": 1.5108, "step": 3422500 }, { "epoch": 17.3, "learning_rate": 2.555e-05, "loss": 1.5227, "step": 3423000 }, { "epoch": 17.3, "learning_rate": 2.5546428571428572e-05, "loss": 1.5094, "step": 3423500 }, { "epoch": 17.31, "learning_rate": 2.5542857142857146e-05, "loss": 1.5038, "step": 3424000 }, { "epoch": 17.31, "learning_rate": 2.5539285714285717e-05, "loss": 1.503, "step": 3424500 }, { "epoch": 17.31, "learning_rate": 2.5535714285714284e-05, "loss": 1.5071, "step": 3425000 }, { "epoch": 17.31, "learning_rate": 2.553214285714286e-05, "loss": 1.5032, "step": 3425500 }, { "epoch": 17.32, "learning_rate": 2.552857142857143e-05, "loss": 1.5171, "step": 3426000 }, { "epoch": 17.32, "learning_rate": 2.5525e-05, "loss": 1.5007, "step": 3426500 }, { "epoch": 17.32, "learning_rate": 2.5521428571428575e-05, "loss": 1.5068, "step": 3427000 }, { "epoch": 17.33, "learning_rate": 2.5517857142857142e-05, "loss": 1.5148, "step": 3427500 }, { "epoch": 17.33, "learning_rate": 2.5514285714285713e-05, "loss": 1.5083, "step": 3428000 }, { "epoch": 17.33, "learning_rate": 2.5510714285714287e-05, "loss": 1.4977, "step": 3428500 }, { "epoch": 17.33, "learning_rate": 2.5507142857142858e-05, "loss": 1.5061, "step": 3429000 }, { "epoch": 17.34, "learning_rate": 2.5503571428571432e-05, "loss": 1.503, "step": 3429500 }, { "epoch": 17.34, "learning_rate": 2.5500000000000003e-05, "loss": 1.5081, "step": 3430000 }, { "epoch": 17.34, "learning_rate": 2.549642857142857e-05, "loss": 1.5017, "step": 3430500 }, { "epoch": 17.34, "learning_rate": 2.5492857142857145e-05, "loss": 1.5149, "step": 3431000 }, { "epoch": 17.35, "learning_rate": 2.5489285714285716e-05, "loss": 1.5059, "step": 3431500 }, { "epoch": 17.35, "learning_rate": 2.5485714285714287e-05, "loss": 1.5124, "step": 3432000 }, { "epoch": 17.35, "learning_rate": 2.548214285714286e-05, "loss": 1.5083, "step": 3432500 }, { "epoch": 17.35, "learning_rate": 2.547857142857143e-05, "loss": 1.5151, "step": 3433000 }, { "epoch": 17.36, "learning_rate": 2.5475e-05, "loss": 1.5187, "step": 3433500 }, { "epoch": 17.36, "learning_rate": 2.5471428571428573e-05, "loss": 1.5179, "step": 3434000 }, { "epoch": 17.36, "learning_rate": 2.5467857142857144e-05, "loss": 1.5066, "step": 3434500 }, { "epoch": 17.37, "learning_rate": 2.5464285714285712e-05, "loss": 1.505, "step": 3435000 }, { "epoch": 17.37, "learning_rate": 2.546071428571429e-05, "loss": 1.5037, "step": 3435500 }, { "epoch": 17.37, "learning_rate": 2.5457142857142857e-05, "loss": 1.5084, "step": 3436000 }, { "epoch": 17.37, "learning_rate": 2.545357142857143e-05, "loss": 1.5067, "step": 3436500 }, { "epoch": 17.38, "learning_rate": 2.5450000000000002e-05, "loss": 1.5123, "step": 3437000 }, { "epoch": 17.38, "learning_rate": 2.544642857142857e-05, "loss": 1.5053, "step": 3437500 }, { "epoch": 17.38, "learning_rate": 2.5442857142857147e-05, "loss": 1.5118, "step": 3438000 }, { "epoch": 17.38, "learning_rate": 2.5439285714285715e-05, "loss": 1.5175, "step": 3438500 }, { "epoch": 17.39, "learning_rate": 2.5435714285714285e-05, "loss": 1.5138, "step": 3439000 }, { "epoch": 17.39, "learning_rate": 2.543214285714286e-05, "loss": 1.5106, "step": 3439500 }, { "epoch": 17.39, "learning_rate": 2.542857142857143e-05, "loss": 1.5097, "step": 3440000 }, { "epoch": 17.39, "learning_rate": 2.5424999999999998e-05, "loss": 1.506, "step": 3440500 }, { "epoch": 17.4, "learning_rate": 2.5421428571428572e-05, "loss": 1.5107, "step": 3441000 }, { "epoch": 17.4, "learning_rate": 2.5417857142857143e-05, "loss": 1.5145, "step": 3441500 }, { "epoch": 17.4, "learning_rate": 2.5414285714285717e-05, "loss": 1.5164, "step": 3442000 }, { "epoch": 17.41, "learning_rate": 2.5410714285714288e-05, "loss": 1.5036, "step": 3442500 }, { "epoch": 17.41, "learning_rate": 2.5407142857142856e-05, "loss": 1.5109, "step": 3443000 }, { "epoch": 17.41, "learning_rate": 2.5403571428571433e-05, "loss": 1.5025, "step": 3443500 }, { "epoch": 17.41, "learning_rate": 2.54e-05, "loss": 1.5153, "step": 3444000 }, { "epoch": 17.42, "learning_rate": 2.539642857142857e-05, "loss": 1.5055, "step": 3444500 }, { "epoch": 17.42, "learning_rate": 2.5392857142857146e-05, "loss": 1.5084, "step": 3445000 }, { "epoch": 17.42, "learning_rate": 2.5389285714285717e-05, "loss": 1.4977, "step": 3445500 }, { "epoch": 17.42, "learning_rate": 2.5385714285714284e-05, "loss": 1.5171, "step": 3446000 }, { "epoch": 17.43, "learning_rate": 2.538214285714286e-05, "loss": 1.5184, "step": 3446500 }, { "epoch": 17.43, "learning_rate": 2.537857142857143e-05, "loss": 1.5038, "step": 3447000 }, { "epoch": 17.43, "learning_rate": 2.5375e-05, "loss": 1.5098, "step": 3447500 }, { "epoch": 17.43, "learning_rate": 2.5371428571428574e-05, "loss": 1.5184, "step": 3448000 }, { "epoch": 17.44, "learning_rate": 2.5367857142857142e-05, "loss": 1.5145, "step": 3448500 }, { "epoch": 17.44, "learning_rate": 2.536428571428572e-05, "loss": 1.5066, "step": 3449000 }, { "epoch": 17.44, "learning_rate": 2.5360714285714287e-05, "loss": 1.5108, "step": 3449500 }, { "epoch": 17.45, "learning_rate": 2.5357142857142858e-05, "loss": 1.5215, "step": 3450000 }, { "epoch": 17.45, "learning_rate": 2.5353571428571432e-05, "loss": 1.5104, "step": 3450500 }, { "epoch": 17.45, "learning_rate": 2.5350000000000003e-05, "loss": 1.5093, "step": 3451000 }, { "epoch": 17.45, "learning_rate": 2.534642857142857e-05, "loss": 1.5104, "step": 3451500 }, { "epoch": 17.46, "learning_rate": 2.5342857142857145e-05, "loss": 1.5098, "step": 3452000 }, { "epoch": 17.46, "learning_rate": 2.5339285714285716e-05, "loss": 1.505, "step": 3452500 }, { "epoch": 17.46, "learning_rate": 2.5335714285714286e-05, "loss": 1.5215, "step": 3453000 }, { "epoch": 17.46, "learning_rate": 2.533214285714286e-05, "loss": 1.5169, "step": 3453500 }, { "epoch": 17.47, "learning_rate": 2.5328571428571428e-05, "loss": 1.5071, "step": 3454000 }, { "epoch": 17.47, "learning_rate": 2.5325e-05, "loss": 1.5087, "step": 3454500 }, { "epoch": 17.47, "learning_rate": 2.5321428571428573e-05, "loss": 1.5024, "step": 3455000 }, { "epoch": 17.47, "learning_rate": 2.5317857142857144e-05, "loss": 1.5137, "step": 3455500 }, { "epoch": 17.48, "learning_rate": 2.5314285714285718e-05, "loss": 1.5101, "step": 3456000 }, { "epoch": 17.48, "learning_rate": 2.531071428571429e-05, "loss": 1.5065, "step": 3456500 }, { "epoch": 17.48, "learning_rate": 2.5307142857142857e-05, "loss": 1.5127, "step": 3457000 }, { "epoch": 17.48, "learning_rate": 2.530357142857143e-05, "loss": 1.5172, "step": 3457500 }, { "epoch": 17.49, "learning_rate": 2.5300000000000002e-05, "loss": 1.5186, "step": 3458000 }, { "epoch": 17.49, "learning_rate": 2.5296428571428573e-05, "loss": 1.5128, "step": 3458500 }, { "epoch": 17.49, "learning_rate": 2.5292857142857147e-05, "loss": 1.5077, "step": 3459000 }, { "epoch": 17.5, "learning_rate": 2.5289285714285714e-05, "loss": 1.5027, "step": 3459500 }, { "epoch": 17.5, "learning_rate": 2.5285714285714285e-05, "loss": 1.5047, "step": 3460000 }, { "epoch": 17.5, "learning_rate": 2.528214285714286e-05, "loss": 1.5135, "step": 3460500 }, { "epoch": 17.5, "learning_rate": 2.527857142857143e-05, "loss": 1.5142, "step": 3461000 }, { "epoch": 17.51, "learning_rate": 2.5274999999999998e-05, "loss": 1.5065, "step": 3461500 }, { "epoch": 17.51, "learning_rate": 2.5271428571428575e-05, "loss": 1.5121, "step": 3462000 }, { "epoch": 17.51, "learning_rate": 2.5267857142857143e-05, "loss": 1.5077, "step": 3462500 }, { "epoch": 17.51, "learning_rate": 2.5264285714285717e-05, "loss": 1.5098, "step": 3463000 }, { "epoch": 17.52, "learning_rate": 2.5260714285714288e-05, "loss": 1.5121, "step": 3463500 }, { "epoch": 17.52, "learning_rate": 2.5257142857142855e-05, "loss": 1.5113, "step": 3464000 }, { "epoch": 17.52, "learning_rate": 2.5253571428571433e-05, "loss": 1.5041, "step": 3464500 }, { "epoch": 17.52, "learning_rate": 2.525e-05, "loss": 1.5155, "step": 3465000 }, { "epoch": 17.53, "learning_rate": 2.524642857142857e-05, "loss": 1.5014, "step": 3465500 }, { "epoch": 17.53, "learning_rate": 2.5242857142857146e-05, "loss": 1.519, "step": 3466000 }, { "epoch": 17.53, "learning_rate": 2.5239285714285716e-05, "loss": 1.502, "step": 3466500 }, { "epoch": 17.54, "learning_rate": 2.5235714285714284e-05, "loss": 1.5049, "step": 3467000 }, { "epoch": 17.54, "learning_rate": 2.5232142857142858e-05, "loss": 1.5017, "step": 3467500 }, { "epoch": 17.54, "learning_rate": 2.522857142857143e-05, "loss": 1.5071, "step": 3468000 }, { "epoch": 17.54, "learning_rate": 2.5225e-05, "loss": 1.5127, "step": 3468500 }, { "epoch": 17.55, "learning_rate": 2.5221428571428574e-05, "loss": 1.5012, "step": 3469000 }, { "epoch": 17.55, "learning_rate": 2.521785714285714e-05, "loss": 1.5172, "step": 3469500 }, { "epoch": 17.55, "learning_rate": 2.521428571428572e-05, "loss": 1.5045, "step": 3470000 }, { "epoch": 17.55, "learning_rate": 2.5210714285714287e-05, "loss": 1.5141, "step": 3470500 }, { "epoch": 17.56, "learning_rate": 2.5207142857142858e-05, "loss": 1.508, "step": 3471000 }, { "epoch": 17.56, "learning_rate": 2.5203571428571432e-05, "loss": 1.51, "step": 3471500 }, { "epoch": 17.56, "learning_rate": 2.5200000000000003e-05, "loss": 1.5094, "step": 3472000 }, { "epoch": 17.56, "learning_rate": 2.519642857142857e-05, "loss": 1.5056, "step": 3472500 }, { "epoch": 17.57, "learning_rate": 2.5192857142857144e-05, "loss": 1.5002, "step": 3473000 }, { "epoch": 17.57, "learning_rate": 2.5189285714285715e-05, "loss": 1.5015, "step": 3473500 }, { "epoch": 17.57, "learning_rate": 2.5185714285714286e-05, "loss": 1.513, "step": 3474000 }, { "epoch": 17.58, "learning_rate": 2.518214285714286e-05, "loss": 1.5095, "step": 3474500 }, { "epoch": 17.58, "learning_rate": 2.5178571428571428e-05, "loss": 1.5075, "step": 3475000 }, { "epoch": 17.58, "learning_rate": 2.5175e-05, "loss": 1.5005, "step": 3475500 }, { "epoch": 17.58, "learning_rate": 2.5171428571428573e-05, "loss": 1.5073, "step": 3476000 }, { "epoch": 17.59, "learning_rate": 2.5167857142857144e-05, "loss": 1.5003, "step": 3476500 }, { "epoch": 17.59, "learning_rate": 2.5164285714285718e-05, "loss": 1.5077, "step": 3477000 }, { "epoch": 17.59, "learning_rate": 2.516071428571429e-05, "loss": 1.5009, "step": 3477500 }, { "epoch": 17.59, "learning_rate": 2.5157142857142856e-05, "loss": 1.5208, "step": 3478000 }, { "epoch": 17.6, "learning_rate": 2.515357142857143e-05, "loss": 1.5051, "step": 3478500 }, { "epoch": 17.6, "learning_rate": 2.515e-05, "loss": 1.4975, "step": 3479000 }, { "epoch": 17.6, "learning_rate": 2.5146428571428572e-05, "loss": 1.5036, "step": 3479500 }, { "epoch": 17.6, "learning_rate": 2.5142857142857147e-05, "loss": 1.5113, "step": 3480000 }, { "epoch": 17.61, "learning_rate": 2.5139285714285714e-05, "loss": 1.5067, "step": 3480500 }, { "epoch": 17.61, "learning_rate": 2.5135714285714285e-05, "loss": 1.5145, "step": 3481000 }, { "epoch": 17.61, "learning_rate": 2.513214285714286e-05, "loss": 1.5069, "step": 3481500 }, { "epoch": 17.62, "learning_rate": 2.512857142857143e-05, "loss": 1.5074, "step": 3482000 }, { "epoch": 17.62, "learning_rate": 2.5124999999999997e-05, "loss": 1.5061, "step": 3482500 }, { "epoch": 17.62, "learning_rate": 2.5121428571428575e-05, "loss": 1.5036, "step": 3483000 }, { "epoch": 17.62, "learning_rate": 2.5117857142857143e-05, "loss": 1.5115, "step": 3483500 }, { "epoch": 17.63, "learning_rate": 2.5114285714285717e-05, "loss": 1.5109, "step": 3484000 }, { "epoch": 17.63, "learning_rate": 2.5110714285714288e-05, "loss": 1.5121, "step": 3484500 }, { "epoch": 17.63, "learning_rate": 2.510714285714286e-05, "loss": 1.5018, "step": 3485000 }, { "epoch": 17.63, "learning_rate": 2.5103571428571433e-05, "loss": 1.5124, "step": 3485500 }, { "epoch": 17.64, "learning_rate": 2.51e-05, "loss": 1.5103, "step": 3486000 }, { "epoch": 17.64, "learning_rate": 2.509642857142857e-05, "loss": 1.5091, "step": 3486500 }, { "epoch": 17.64, "learning_rate": 2.5092857142857145e-05, "loss": 1.5065, "step": 3487000 }, { "epoch": 17.64, "learning_rate": 2.5089285714285716e-05, "loss": 1.4953, "step": 3487500 }, { "epoch": 17.65, "learning_rate": 2.5085714285714284e-05, "loss": 1.5148, "step": 3488000 }, { "epoch": 17.65, "learning_rate": 2.5082142857142858e-05, "loss": 1.5162, "step": 3488500 }, { "epoch": 17.65, "learning_rate": 2.507857142857143e-05, "loss": 1.5075, "step": 3489000 }, { "epoch": 17.66, "learning_rate": 2.5075e-05, "loss": 1.5079, "step": 3489500 }, { "epoch": 17.66, "learning_rate": 2.5071428571428574e-05, "loss": 1.5065, "step": 3490000 }, { "epoch": 17.66, "learning_rate": 2.506785714285714e-05, "loss": 1.5024, "step": 3490500 }, { "epoch": 17.66, "learning_rate": 2.506428571428572e-05, "loss": 1.5147, "step": 3491000 }, { "epoch": 17.67, "learning_rate": 2.5060714285714286e-05, "loss": 1.5012, "step": 3491500 }, { "epoch": 17.67, "learning_rate": 2.5057142857142857e-05, "loss": 1.5014, "step": 3492000 }, { "epoch": 17.67, "learning_rate": 2.505357142857143e-05, "loss": 1.505, "step": 3492500 }, { "epoch": 17.67, "learning_rate": 2.5050000000000002e-05, "loss": 1.5102, "step": 3493000 }, { "epoch": 17.68, "learning_rate": 2.504642857142857e-05, "loss": 1.502, "step": 3493500 }, { "epoch": 17.68, "learning_rate": 2.5042857142857144e-05, "loss": 1.5071, "step": 3494000 }, { "epoch": 17.68, "learning_rate": 2.5039285714285715e-05, "loss": 1.5082, "step": 3494500 }, { "epoch": 17.68, "learning_rate": 2.5035714285714286e-05, "loss": 1.4978, "step": 3495000 }, { "epoch": 17.69, "learning_rate": 2.503214285714286e-05, "loss": 1.5078, "step": 3495500 }, { "epoch": 17.69, "learning_rate": 2.5028571428571428e-05, "loss": 1.4932, "step": 3496000 }, { "epoch": 17.69, "learning_rate": 2.5025e-05, "loss": 1.5059, "step": 3496500 }, { "epoch": 17.7, "learning_rate": 2.5021428571428573e-05, "loss": 1.5052, "step": 3497000 }, { "epoch": 17.7, "learning_rate": 2.5017857142857144e-05, "loss": 1.5184, "step": 3497500 }, { "epoch": 17.7, "learning_rate": 2.5014285714285718e-05, "loss": 1.4989, "step": 3498000 }, { "epoch": 17.7, "learning_rate": 2.501071428571429e-05, "loss": 1.5117, "step": 3498500 }, { "epoch": 17.71, "learning_rate": 2.5007142857142856e-05, "loss": 1.5048, "step": 3499000 }, { "epoch": 17.71, "learning_rate": 2.500357142857143e-05, "loss": 1.5043, "step": 3499500 }, { "epoch": 17.71, "learning_rate": 2.5e-05, "loss": 1.5039, "step": 3500000 }, { "epoch": 17.71, "learning_rate": 2.4996428571428572e-05, "loss": 1.4989, "step": 3500500 }, { "epoch": 17.72, "learning_rate": 2.4992857142857143e-05, "loss": 1.5039, "step": 3501000 }, { "epoch": 17.72, "learning_rate": 2.4989285714285714e-05, "loss": 1.4996, "step": 3501500 }, { "epoch": 17.72, "learning_rate": 2.4985714285714288e-05, "loss": 1.5127, "step": 3502000 }, { "epoch": 17.72, "learning_rate": 2.498214285714286e-05, "loss": 1.505, "step": 3502500 }, { "epoch": 17.73, "learning_rate": 2.497857142857143e-05, "loss": 1.4943, "step": 3503000 }, { "epoch": 17.73, "learning_rate": 2.4975e-05, "loss": 1.5025, "step": 3503500 }, { "epoch": 17.73, "learning_rate": 2.4971428571428575e-05, "loss": 1.5084, "step": 3504000 }, { "epoch": 17.74, "learning_rate": 2.4967857142857142e-05, "loss": 1.4958, "step": 3504500 }, { "epoch": 17.74, "learning_rate": 2.4964285714285717e-05, "loss": 1.5002, "step": 3505000 }, { "epoch": 17.74, "learning_rate": 2.4960714285714287e-05, "loss": 1.5001, "step": 3505500 }, { "epoch": 17.74, "learning_rate": 2.4957142857142858e-05, "loss": 1.5111, "step": 3506000 }, { "epoch": 17.75, "learning_rate": 2.495357142857143e-05, "loss": 1.506, "step": 3506500 }, { "epoch": 17.75, "learning_rate": 2.495e-05, "loss": 1.5009, "step": 3507000 }, { "epoch": 17.75, "learning_rate": 2.4946428571428574e-05, "loss": 1.5141, "step": 3507500 }, { "epoch": 17.75, "learning_rate": 2.4942857142857142e-05, "loss": 1.5054, "step": 3508000 }, { "epoch": 17.76, "learning_rate": 2.4939285714285716e-05, "loss": 1.5024, "step": 3508500 }, { "epoch": 17.76, "learning_rate": 2.4935714285714287e-05, "loss": 1.5058, "step": 3509000 }, { "epoch": 17.76, "learning_rate": 2.493214285714286e-05, "loss": 1.5029, "step": 3509500 }, { "epoch": 17.76, "learning_rate": 2.492857142857143e-05, "loss": 1.5035, "step": 3510000 }, { "epoch": 17.77, "learning_rate": 2.4925000000000003e-05, "loss": 1.5177, "step": 3510500 }, { "epoch": 17.77, "learning_rate": 2.4921428571428574e-05, "loss": 1.5021, "step": 3511000 }, { "epoch": 17.77, "learning_rate": 2.4917857142857144e-05, "loss": 1.5071, "step": 3511500 }, { "epoch": 17.78, "learning_rate": 2.4914285714285715e-05, "loss": 1.5053, "step": 3512000 }, { "epoch": 17.78, "learning_rate": 2.4910714285714286e-05, "loss": 1.5064, "step": 3512500 }, { "epoch": 17.78, "learning_rate": 2.490714285714286e-05, "loss": 1.5044, "step": 3513000 }, { "epoch": 17.78, "learning_rate": 2.4903571428571428e-05, "loss": 1.5043, "step": 3513500 }, { "epoch": 17.79, "learning_rate": 2.4900000000000002e-05, "loss": 1.5003, "step": 3514000 }, { "epoch": 17.79, "learning_rate": 2.4896428571428573e-05, "loss": 1.5068, "step": 3514500 }, { "epoch": 17.79, "learning_rate": 2.4892857142857144e-05, "loss": 1.5106, "step": 3515000 }, { "epoch": 17.79, "learning_rate": 2.4889285714285715e-05, "loss": 1.5041, "step": 3515500 }, { "epoch": 17.8, "learning_rate": 2.4885714285714286e-05, "loss": 1.511, "step": 3516000 }, { "epoch": 17.8, "learning_rate": 2.488214285714286e-05, "loss": 1.4999, "step": 3516500 }, { "epoch": 17.8, "learning_rate": 2.4878571428571427e-05, "loss": 1.5045, "step": 3517000 }, { "epoch": 17.8, "learning_rate": 2.4875e-05, "loss": 1.4997, "step": 3517500 }, { "epoch": 17.81, "learning_rate": 2.4871428571428572e-05, "loss": 1.5053, "step": 3518000 }, { "epoch": 17.81, "learning_rate": 2.4867857142857143e-05, "loss": 1.5094, "step": 3518500 }, { "epoch": 17.81, "learning_rate": 2.4864285714285714e-05, "loss": 1.5001, "step": 3519000 }, { "epoch": 17.82, "learning_rate": 2.486071428571429e-05, "loss": 1.4963, "step": 3519500 }, { "epoch": 17.82, "learning_rate": 2.485714285714286e-05, "loss": 1.5013, "step": 3520000 }, { "epoch": 17.82, "learning_rate": 2.485357142857143e-05, "loss": 1.5189, "step": 3520500 }, { "epoch": 17.82, "learning_rate": 2.485e-05, "loss": 1.5095, "step": 3521000 }, { "epoch": 17.83, "learning_rate": 2.4846428571428572e-05, "loss": 1.4987, "step": 3521500 }, { "epoch": 17.83, "learning_rate": 2.4842857142857143e-05, "loss": 1.5017, "step": 3522000 }, { "epoch": 17.83, "learning_rate": 2.4839285714285714e-05, "loss": 1.5072, "step": 3522500 }, { "epoch": 17.83, "learning_rate": 2.4835714285714288e-05, "loss": 1.4985, "step": 3523000 }, { "epoch": 17.84, "learning_rate": 2.483214285714286e-05, "loss": 1.5008, "step": 3523500 }, { "epoch": 17.84, "learning_rate": 2.482857142857143e-05, "loss": 1.5094, "step": 3524000 }, { "epoch": 17.84, "learning_rate": 2.4825e-05, "loss": 1.5143, "step": 3524500 }, { "epoch": 17.84, "learning_rate": 2.4821428571428575e-05, "loss": 1.503, "step": 3525000 }, { "epoch": 17.85, "learning_rate": 2.4817857142857142e-05, "loss": 1.5052, "step": 3525500 }, { "epoch": 17.85, "learning_rate": 2.4814285714285716e-05, "loss": 1.5026, "step": 3526000 }, { "epoch": 17.85, "learning_rate": 2.4810714285714287e-05, "loss": 1.5027, "step": 3526500 }, { "epoch": 17.86, "learning_rate": 2.4807142857142858e-05, "loss": 1.5077, "step": 3527000 }, { "epoch": 17.86, "learning_rate": 2.480357142857143e-05, "loss": 1.5045, "step": 3527500 }, { "epoch": 17.86, "learning_rate": 2.48e-05, "loss": 1.5081, "step": 3528000 }, { "epoch": 17.86, "learning_rate": 2.4796428571428574e-05, "loss": 1.5027, "step": 3528500 }, { "epoch": 17.87, "learning_rate": 2.479285714285714e-05, "loss": 1.5037, "step": 3529000 }, { "epoch": 17.87, "learning_rate": 2.4789285714285716e-05, "loss": 1.4943, "step": 3529500 }, { "epoch": 17.87, "learning_rate": 2.4785714285714287e-05, "loss": 1.5222, "step": 3530000 }, { "epoch": 17.87, "learning_rate": 2.478214285714286e-05, "loss": 1.5149, "step": 3530500 }, { "epoch": 17.88, "learning_rate": 2.4778571428571428e-05, "loss": 1.5007, "step": 3531000 }, { "epoch": 17.88, "learning_rate": 2.4775000000000003e-05, "loss": 1.5089, "step": 3531500 }, { "epoch": 17.88, "learning_rate": 2.4771428571428573e-05, "loss": 1.5057, "step": 3532000 }, { "epoch": 17.88, "learning_rate": 2.4767857142857144e-05, "loss": 1.5024, "step": 3532500 }, { "epoch": 17.89, "learning_rate": 2.4764285714285715e-05, "loss": 1.5037, "step": 3533000 }, { "epoch": 17.89, "learning_rate": 2.4760714285714286e-05, "loss": 1.508, "step": 3533500 }, { "epoch": 17.89, "learning_rate": 2.475714285714286e-05, "loss": 1.511, "step": 3534000 }, { "epoch": 17.89, "learning_rate": 2.4753571428571428e-05, "loss": 1.501, "step": 3534500 }, { "epoch": 17.9, "learning_rate": 2.4750000000000002e-05, "loss": 1.5072, "step": 3535000 }, { "epoch": 17.9, "learning_rate": 2.4746428571428573e-05, "loss": 1.4959, "step": 3535500 }, { "epoch": 17.9, "learning_rate": 2.4742857142857147e-05, "loss": 1.4977, "step": 3536000 }, { "epoch": 17.91, "learning_rate": 2.4739285714285714e-05, "loss": 1.5015, "step": 3536500 }, { "epoch": 17.91, "learning_rate": 2.473571428571429e-05, "loss": 1.5077, "step": 3537000 }, { "epoch": 17.91, "learning_rate": 2.473214285714286e-05, "loss": 1.4996, "step": 3537500 }, { "epoch": 17.91, "learning_rate": 2.4728571428571427e-05, "loss": 1.5043, "step": 3538000 }, { "epoch": 17.92, "learning_rate": 2.4725e-05, "loss": 1.5022, "step": 3538500 }, { "epoch": 17.92, "learning_rate": 2.4721428571428572e-05, "loss": 1.5026, "step": 3539000 }, { "epoch": 17.92, "learning_rate": 2.4717857142857146e-05, "loss": 1.497, "step": 3539500 }, { "epoch": 17.92, "learning_rate": 2.4714285714285714e-05, "loss": 1.4943, "step": 3540000 }, { "epoch": 17.93, "learning_rate": 2.4710714285714288e-05, "loss": 1.4981, "step": 3540500 }, { "epoch": 17.93, "learning_rate": 2.470714285714286e-05, "loss": 1.5148, "step": 3541000 }, { "epoch": 17.93, "learning_rate": 2.470357142857143e-05, "loss": 1.5047, "step": 3541500 }, { "epoch": 17.93, "learning_rate": 2.47e-05, "loss": 1.5024, "step": 3542000 }, { "epoch": 17.94, "learning_rate": 2.469642857142857e-05, "loss": 1.5149, "step": 3542500 }, { "epoch": 17.94, "learning_rate": 2.4692857142857146e-05, "loss": 1.5038, "step": 3543000 }, { "epoch": 17.94, "learning_rate": 2.4689285714285713e-05, "loss": 1.5131, "step": 3543500 }, { "epoch": 17.95, "learning_rate": 2.4685714285714288e-05, "loss": 1.5003, "step": 3544000 }, { "epoch": 17.95, "learning_rate": 2.468214285714286e-05, "loss": 1.4964, "step": 3544500 }, { "epoch": 17.95, "learning_rate": 2.467857142857143e-05, "loss": 1.51, "step": 3545000 }, { "epoch": 17.95, "learning_rate": 2.4675e-05, "loss": 1.5036, "step": 3545500 }, { "epoch": 17.96, "learning_rate": 2.4671428571428574e-05, "loss": 1.5081, "step": 3546000 }, { "epoch": 17.96, "learning_rate": 2.4667857142857145e-05, "loss": 1.5075, "step": 3546500 }, { "epoch": 17.96, "learning_rate": 2.4664285714285716e-05, "loss": 1.5078, "step": 3547000 }, { "epoch": 17.96, "learning_rate": 2.4660714285714287e-05, "loss": 1.5169, "step": 3547500 }, { "epoch": 17.97, "learning_rate": 2.4657142857142858e-05, "loss": 1.5052, "step": 3548000 }, { "epoch": 17.97, "learning_rate": 2.465357142857143e-05, "loss": 1.4958, "step": 3548500 }, { "epoch": 17.97, "learning_rate": 2.465e-05, "loss": 1.5022, "step": 3549000 }, { "epoch": 17.97, "learning_rate": 2.4646428571428574e-05, "loss": 1.5049, "step": 3549500 }, { "epoch": 17.98, "learning_rate": 2.4642857142857145e-05, "loss": 1.4931, "step": 3550000 }, { "epoch": 17.98, "learning_rate": 2.4639285714285715e-05, "loss": 1.5016, "step": 3550500 }, { "epoch": 17.98, "learning_rate": 2.4635714285714286e-05, "loss": 1.5008, "step": 3551000 }, { "epoch": 17.99, "learning_rate": 2.463214285714286e-05, "loss": 1.5031, "step": 3551500 }, { "epoch": 17.99, "learning_rate": 2.4628571428571428e-05, "loss": 1.4976, "step": 3552000 }, { "epoch": 17.99, "learning_rate": 2.4625000000000002e-05, "loss": 1.5114, "step": 3552500 }, { "epoch": 17.99, "learning_rate": 2.4621428571428573e-05, "loss": 1.4993, "step": 3553000 }, { "epoch": 18.0, "learning_rate": 2.4617857142857144e-05, "loss": 1.5017, "step": 3553500 }, { "epoch": 18.0, "learning_rate": 2.4614285714285715e-05, "loss": 1.5111, "step": 3554000 }, { "epoch": 18.0, "learning_rate": 2.4610714285714286e-05, "loss": 1.5074, "step": 3554500 }, { "epoch": 18.0, "learning_rate": 2.460714285714286e-05, "loss": 1.492, "step": 3555000 }, { "epoch": 18.01, "learning_rate": 2.4603571428571427e-05, "loss": 1.4979, "step": 3555500 }, { "epoch": 18.01, "learning_rate": 2.46e-05, "loss": 1.5007, "step": 3556000 }, { "epoch": 18.01, "learning_rate": 2.4596428571428572e-05, "loss": 1.5054, "step": 3556500 }, { "epoch": 18.01, "learning_rate": 2.4592857142857147e-05, "loss": 1.5018, "step": 3557000 }, { "epoch": 18.02, "learning_rate": 2.4589285714285714e-05, "loss": 1.4867, "step": 3557500 }, { "epoch": 18.02, "learning_rate": 2.458571428571429e-05, "loss": 1.5009, "step": 3558000 }, { "epoch": 18.02, "learning_rate": 2.458214285714286e-05, "loss": 1.4917, "step": 3558500 }, { "epoch": 18.03, "learning_rate": 2.457857142857143e-05, "loss": 1.5038, "step": 3559000 }, { "epoch": 18.03, "learning_rate": 2.4575e-05, "loss": 1.4945, "step": 3559500 }, { "epoch": 18.03, "learning_rate": 2.4571428571428572e-05, "loss": 1.4981, "step": 3560000 }, { "epoch": 18.03, "learning_rate": 2.4567857142857146e-05, "loss": 1.5047, "step": 3560500 }, { "epoch": 18.04, "learning_rate": 2.4564285714285714e-05, "loss": 1.4917, "step": 3561000 }, { "epoch": 18.04, "learning_rate": 2.4560714285714288e-05, "loss": 1.4964, "step": 3561500 }, { "epoch": 18.04, "learning_rate": 2.455714285714286e-05, "loss": 1.4955, "step": 3562000 }, { "epoch": 18.04, "learning_rate": 2.455357142857143e-05, "loss": 1.498, "step": 3562500 }, { "epoch": 18.05, "learning_rate": 2.455e-05, "loss": 1.4961, "step": 3563000 }, { "epoch": 18.05, "learning_rate": 2.4546428571428575e-05, "loss": 1.5028, "step": 3563500 }, { "epoch": 18.05, "learning_rate": 2.4542857142857146e-05, "loss": 1.5002, "step": 3564000 }, { "epoch": 18.05, "learning_rate": 2.4539285714285713e-05, "loss": 1.4891, "step": 3564500 }, { "epoch": 18.06, "learning_rate": 2.4535714285714287e-05, "loss": 1.4911, "step": 3565000 }, { "epoch": 18.06, "learning_rate": 2.4532142857142858e-05, "loss": 1.495, "step": 3565500 }, { "epoch": 18.06, "learning_rate": 2.452857142857143e-05, "loss": 1.4923, "step": 3566000 }, { "epoch": 18.07, "learning_rate": 2.4525e-05, "loss": 1.4963, "step": 3566500 }, { "epoch": 18.07, "learning_rate": 2.4521428571428574e-05, "loss": 1.4976, "step": 3567000 }, { "epoch": 18.07, "learning_rate": 2.4517857142857145e-05, "loss": 1.4924, "step": 3567500 }, { "epoch": 18.07, "learning_rate": 2.4514285714285716e-05, "loss": 1.505, "step": 3568000 }, { "epoch": 18.08, "learning_rate": 2.4510714285714287e-05, "loss": 1.4977, "step": 3568500 }, { "epoch": 18.08, "learning_rate": 2.4507142857142857e-05, "loss": 1.4958, "step": 3569000 }, { "epoch": 18.08, "learning_rate": 2.450357142857143e-05, "loss": 1.4988, "step": 3569500 }, { "epoch": 18.08, "learning_rate": 2.45e-05, "loss": 1.4953, "step": 3570000 }, { "epoch": 18.09, "learning_rate": 2.4496428571428573e-05, "loss": 1.5019, "step": 3570500 }, { "epoch": 18.09, "learning_rate": 2.4492857142857144e-05, "loss": 1.4973, "step": 3571000 }, { "epoch": 18.09, "learning_rate": 2.4489285714285715e-05, "loss": 1.499, "step": 3571500 }, { "epoch": 18.09, "learning_rate": 2.4485714285714286e-05, "loss": 1.5088, "step": 3572000 }, { "epoch": 18.1, "learning_rate": 2.448214285714286e-05, "loss": 1.4985, "step": 3572500 }, { "epoch": 18.1, "learning_rate": 2.4478571428571428e-05, "loss": 1.5043, "step": 3573000 }, { "epoch": 18.1, "learning_rate": 2.4475000000000002e-05, "loss": 1.4966, "step": 3573500 }, { "epoch": 18.11, "learning_rate": 2.4471428571428573e-05, "loss": 1.4965, "step": 3574000 }, { "epoch": 18.11, "learning_rate": 2.4467857142857144e-05, "loss": 1.5007, "step": 3574500 }, { "epoch": 18.11, "learning_rate": 2.4464285714285715e-05, "loss": 1.4933, "step": 3575000 }, { "epoch": 18.11, "learning_rate": 2.4460714285714285e-05, "loss": 1.5004, "step": 3575500 }, { "epoch": 18.12, "learning_rate": 2.445714285714286e-05, "loss": 1.4989, "step": 3576000 }, { "epoch": 18.12, "learning_rate": 2.4453571428571427e-05, "loss": 1.4966, "step": 3576500 }, { "epoch": 18.12, "learning_rate": 2.445e-05, "loss": 1.4961, "step": 3577000 }, { "epoch": 18.12, "learning_rate": 2.4446428571428572e-05, "loss": 1.5024, "step": 3577500 }, { "epoch": 18.13, "learning_rate": 2.4442857142857146e-05, "loss": 1.503, "step": 3578000 }, { "epoch": 18.13, "learning_rate": 2.4439285714285714e-05, "loss": 1.5, "step": 3578500 }, { "epoch": 18.13, "learning_rate": 2.4435714285714288e-05, "loss": 1.5028, "step": 3579000 }, { "epoch": 18.13, "learning_rate": 2.443214285714286e-05, "loss": 1.4994, "step": 3579500 }, { "epoch": 18.14, "learning_rate": 2.442857142857143e-05, "loss": 1.4889, "step": 3580000 }, { "epoch": 18.14, "learning_rate": 2.4425e-05, "loss": 1.5108, "step": 3580500 }, { "epoch": 18.14, "learning_rate": 2.442142857142857e-05, "loss": 1.4922, "step": 3581000 }, { "epoch": 18.15, "learning_rate": 2.4417857142857146e-05, "loss": 1.4959, "step": 3581500 }, { "epoch": 18.15, "learning_rate": 2.4414285714285713e-05, "loss": 1.4977, "step": 3582000 }, { "epoch": 18.15, "learning_rate": 2.4410714285714288e-05, "loss": 1.4908, "step": 3582500 }, { "epoch": 18.15, "learning_rate": 2.440714285714286e-05, "loss": 1.4918, "step": 3583000 }, { "epoch": 18.16, "learning_rate": 2.440357142857143e-05, "loss": 1.5061, "step": 3583500 }, { "epoch": 18.16, "learning_rate": 2.44e-05, "loss": 1.5049, "step": 3584000 }, { "epoch": 18.16, "learning_rate": 2.4396428571428574e-05, "loss": 1.5006, "step": 3584500 }, { "epoch": 18.16, "learning_rate": 2.4392857142857145e-05, "loss": 1.4988, "step": 3585000 }, { "epoch": 18.17, "learning_rate": 2.4389285714285716e-05, "loss": 1.4893, "step": 3585500 }, { "epoch": 18.17, "learning_rate": 2.4385714285714287e-05, "loss": 1.5002, "step": 3586000 }, { "epoch": 18.17, "learning_rate": 2.4382142857142858e-05, "loss": 1.4979, "step": 3586500 }, { "epoch": 18.17, "learning_rate": 2.437857142857143e-05, "loss": 1.4959, "step": 3587000 }, { "epoch": 18.18, "learning_rate": 2.4375e-05, "loss": 1.4973, "step": 3587500 }, { "epoch": 18.18, "learning_rate": 2.4371428571428574e-05, "loss": 1.4881, "step": 3588000 }, { "epoch": 18.18, "learning_rate": 2.4367857142857145e-05, "loss": 1.5019, "step": 3588500 }, { "epoch": 18.19, "learning_rate": 2.4364285714285716e-05, "loss": 1.4911, "step": 3589000 }, { "epoch": 18.19, "learning_rate": 2.4360714285714286e-05, "loss": 1.4914, "step": 3589500 }, { "epoch": 18.19, "learning_rate": 2.4357142857142857e-05, "loss": 1.4953, "step": 3590000 }, { "epoch": 18.19, "learning_rate": 2.4353571428571428e-05, "loss": 1.499, "step": 3590500 }, { "epoch": 18.2, "learning_rate": 2.435e-05, "loss": 1.4903, "step": 3591000 }, { "epoch": 18.2, "learning_rate": 2.4346428571428573e-05, "loss": 1.4965, "step": 3591500 }, { "epoch": 18.2, "learning_rate": 2.4342857142857144e-05, "loss": 1.4973, "step": 3592000 }, { "epoch": 18.2, "learning_rate": 2.4339285714285715e-05, "loss": 1.5052, "step": 3592500 }, { "epoch": 18.21, "learning_rate": 2.4335714285714286e-05, "loss": 1.4901, "step": 3593000 }, { "epoch": 18.21, "learning_rate": 2.433214285714286e-05, "loss": 1.4959, "step": 3593500 }, { "epoch": 18.21, "learning_rate": 2.432857142857143e-05, "loss": 1.4997, "step": 3594000 }, { "epoch": 18.21, "learning_rate": 2.4325000000000002e-05, "loss": 1.4965, "step": 3594500 }, { "epoch": 18.22, "learning_rate": 2.4321428571428573e-05, "loss": 1.4933, "step": 3595000 }, { "epoch": 18.22, "learning_rate": 2.4317857142857143e-05, "loss": 1.5083, "step": 3595500 }, { "epoch": 18.22, "learning_rate": 2.4314285714285714e-05, "loss": 1.504, "step": 3596000 }, { "epoch": 18.23, "learning_rate": 2.4310714285714285e-05, "loss": 1.4984, "step": 3596500 }, { "epoch": 18.23, "learning_rate": 2.430714285714286e-05, "loss": 1.4928, "step": 3597000 }, { "epoch": 18.23, "learning_rate": 2.430357142857143e-05, "loss": 1.4949, "step": 3597500 }, { "epoch": 18.23, "learning_rate": 2.43e-05, "loss": 1.4926, "step": 3598000 }, { "epoch": 18.24, "learning_rate": 2.4296428571428572e-05, "loss": 1.5012, "step": 3598500 }, { "epoch": 18.24, "learning_rate": 2.4292857142857146e-05, "loss": 1.4781, "step": 3599000 }, { "epoch": 18.24, "learning_rate": 2.4289285714285714e-05, "loss": 1.5018, "step": 3599500 }, { "epoch": 18.24, "learning_rate": 2.4285714285714288e-05, "loss": 1.5009, "step": 3600000 }, { "epoch": 18.25, "learning_rate": 2.428214285714286e-05, "loss": 1.5024, "step": 3600500 }, { "epoch": 18.25, "learning_rate": 2.427857142857143e-05, "loss": 1.4989, "step": 3601000 }, { "epoch": 18.25, "learning_rate": 2.4275e-05, "loss": 1.5001, "step": 3601500 }, { "epoch": 18.25, "learning_rate": 2.427142857142857e-05, "loss": 1.5016, "step": 3602000 }, { "epoch": 18.26, "learning_rate": 2.4267857142857146e-05, "loss": 1.4989, "step": 3602500 }, { "epoch": 18.26, "learning_rate": 2.4264285714285713e-05, "loss": 1.49, "step": 3603000 }, { "epoch": 18.26, "learning_rate": 2.4260714285714287e-05, "loss": 1.5009, "step": 3603500 }, { "epoch": 18.26, "learning_rate": 2.4257142857142858e-05, "loss": 1.4979, "step": 3604000 }, { "epoch": 18.27, "learning_rate": 2.4253571428571432e-05, "loss": 1.4981, "step": 3604500 }, { "epoch": 18.27, "learning_rate": 2.425e-05, "loss": 1.4976, "step": 3605000 }, { "epoch": 18.27, "learning_rate": 2.4246428571428574e-05, "loss": 1.5043, "step": 3605500 }, { "epoch": 18.28, "learning_rate": 2.4242857142857145e-05, "loss": 1.4962, "step": 3606000 }, { "epoch": 18.28, "learning_rate": 2.4239285714285716e-05, "loss": 1.5057, "step": 3606500 }, { "epoch": 18.28, "learning_rate": 2.4235714285714287e-05, "loss": 1.4966, "step": 3607000 }, { "epoch": 18.28, "learning_rate": 2.4232142857142858e-05, "loss": 1.4987, "step": 3607500 }, { "epoch": 18.29, "learning_rate": 2.4228571428571432e-05, "loss": 1.4973, "step": 3608000 }, { "epoch": 18.29, "learning_rate": 2.4225e-05, "loss": 1.4972, "step": 3608500 }, { "epoch": 18.29, "learning_rate": 2.4221428571428574e-05, "loss": 1.4908, "step": 3609000 }, { "epoch": 18.29, "learning_rate": 2.4217857142857144e-05, "loss": 1.4943, "step": 3609500 }, { "epoch": 18.3, "learning_rate": 2.4214285714285715e-05, "loss": 1.5006, "step": 3610000 }, { "epoch": 18.3, "learning_rate": 2.4210714285714286e-05, "loss": 1.5041, "step": 3610500 }, { "epoch": 18.3, "learning_rate": 2.420714285714286e-05, "loss": 1.498, "step": 3611000 }, { "epoch": 18.3, "learning_rate": 2.420357142857143e-05, "loss": 1.4888, "step": 3611500 }, { "epoch": 18.31, "learning_rate": 2.4200000000000002e-05, "loss": 1.497, "step": 3612000 }, { "epoch": 18.31, "learning_rate": 2.4196428571428573e-05, "loss": 1.5075, "step": 3612500 }, { "epoch": 18.31, "learning_rate": 2.4192857142857144e-05, "loss": 1.4966, "step": 3613000 }, { "epoch": 18.32, "learning_rate": 2.4189285714285715e-05, "loss": 1.498, "step": 3613500 }, { "epoch": 18.32, "learning_rate": 2.4185714285714286e-05, "loss": 1.4949, "step": 3614000 }, { "epoch": 18.32, "learning_rate": 2.418214285714286e-05, "loss": 1.4988, "step": 3614500 }, { "epoch": 18.32, "learning_rate": 2.417857142857143e-05, "loss": 1.4888, "step": 3615000 }, { "epoch": 18.33, "learning_rate": 2.4175e-05, "loss": 1.488, "step": 3615500 }, { "epoch": 18.33, "learning_rate": 2.4171428571428572e-05, "loss": 1.497, "step": 3616000 }, { "epoch": 18.33, "learning_rate": 2.4167857142857143e-05, "loss": 1.4942, "step": 3616500 }, { "epoch": 18.33, "learning_rate": 2.4164285714285714e-05, "loss": 1.4929, "step": 3617000 }, { "epoch": 18.34, "learning_rate": 2.4160714285714285e-05, "loss": 1.5039, "step": 3617500 }, { "epoch": 18.34, "learning_rate": 2.415714285714286e-05, "loss": 1.4941, "step": 3618000 }, { "epoch": 18.34, "learning_rate": 2.415357142857143e-05, "loss": 1.4964, "step": 3618500 }, { "epoch": 18.34, "learning_rate": 2.415e-05, "loss": 1.494, "step": 3619000 }, { "epoch": 18.35, "learning_rate": 2.4146428571428572e-05, "loss": 1.5045, "step": 3619500 }, { "epoch": 18.35, "learning_rate": 2.4142857142857146e-05, "loss": 1.4954, "step": 3620000 }, { "epoch": 18.35, "learning_rate": 2.4139285714285713e-05, "loss": 1.5067, "step": 3620500 }, { "epoch": 18.36, "learning_rate": 2.4135714285714288e-05, "loss": 1.4976, "step": 3621000 }, { "epoch": 18.36, "learning_rate": 2.413214285714286e-05, "loss": 1.5019, "step": 3621500 }, { "epoch": 18.36, "learning_rate": 2.412857142857143e-05, "loss": 1.5001, "step": 3622000 }, { "epoch": 18.36, "learning_rate": 2.4125e-05, "loss": 1.4967, "step": 3622500 }, { "epoch": 18.37, "learning_rate": 2.412142857142857e-05, "loss": 1.4889, "step": 3623000 }, { "epoch": 18.37, "learning_rate": 2.4117857142857145e-05, "loss": 1.4906, "step": 3623500 }, { "epoch": 18.37, "learning_rate": 2.4114285714285713e-05, "loss": 1.4945, "step": 3624000 }, { "epoch": 18.37, "learning_rate": 2.4110714285714287e-05, "loss": 1.4957, "step": 3624500 }, { "epoch": 18.38, "learning_rate": 2.4107142857142858e-05, "loss": 1.4898, "step": 3625000 }, { "epoch": 18.38, "learning_rate": 2.4103571428571432e-05, "loss": 1.4913, "step": 3625500 }, { "epoch": 18.38, "learning_rate": 2.41e-05, "loss": 1.4952, "step": 3626000 }, { "epoch": 18.38, "learning_rate": 2.4096428571428574e-05, "loss": 1.4986, "step": 3626500 }, { "epoch": 18.39, "learning_rate": 2.4092857142857145e-05, "loss": 1.4843, "step": 3627000 }, { "epoch": 18.39, "learning_rate": 2.4089285714285716e-05, "loss": 1.5046, "step": 3627500 }, { "epoch": 18.39, "learning_rate": 2.4085714285714286e-05, "loss": 1.5043, "step": 3628000 }, { "epoch": 18.4, "learning_rate": 2.4082142857142857e-05, "loss": 1.4907, "step": 3628500 }, { "epoch": 18.4, "learning_rate": 2.407857142857143e-05, "loss": 1.493, "step": 3629000 }, { "epoch": 18.4, "learning_rate": 2.4075e-05, "loss": 1.4983, "step": 3629500 }, { "epoch": 18.4, "learning_rate": 2.4071428571428573e-05, "loss": 1.4945, "step": 3630000 }, { "epoch": 18.41, "learning_rate": 2.4067857142857144e-05, "loss": 1.4948, "step": 3630500 }, { "epoch": 18.41, "learning_rate": 2.4064285714285715e-05, "loss": 1.4908, "step": 3631000 }, { "epoch": 18.41, "learning_rate": 2.4060714285714286e-05, "loss": 1.4949, "step": 3631500 }, { "epoch": 18.41, "learning_rate": 2.405714285714286e-05, "loss": 1.4818, "step": 3632000 }, { "epoch": 18.42, "learning_rate": 2.405357142857143e-05, "loss": 1.4962, "step": 3632500 }, { "epoch": 18.42, "learning_rate": 2.4050000000000002e-05, "loss": 1.4972, "step": 3633000 }, { "epoch": 18.42, "learning_rate": 2.4046428571428573e-05, "loss": 1.5016, "step": 3633500 }, { "epoch": 18.42, "learning_rate": 2.4042857142857144e-05, "loss": 1.4925, "step": 3634000 }, { "epoch": 18.43, "learning_rate": 2.4039285714285714e-05, "loss": 1.5001, "step": 3634500 }, { "epoch": 18.43, "learning_rate": 2.4035714285714285e-05, "loss": 1.4936, "step": 3635000 }, { "epoch": 18.43, "learning_rate": 2.403214285714286e-05, "loss": 1.4892, "step": 3635500 }, { "epoch": 18.44, "learning_rate": 2.402857142857143e-05, "loss": 1.497, "step": 3636000 }, { "epoch": 18.44, "learning_rate": 2.4025e-05, "loss": 1.491, "step": 3636500 }, { "epoch": 18.44, "learning_rate": 2.4021428571428572e-05, "loss": 1.4919, "step": 3637000 }, { "epoch": 18.44, "learning_rate": 2.4017857142857146e-05, "loss": 1.4928, "step": 3637500 }, { "epoch": 18.45, "learning_rate": 2.4014285714285714e-05, "loss": 1.4991, "step": 3638000 }, { "epoch": 18.45, "learning_rate": 2.4010714285714288e-05, "loss": 1.4966, "step": 3638500 }, { "epoch": 18.45, "learning_rate": 2.400714285714286e-05, "loss": 1.4935, "step": 3639000 }, { "epoch": 18.45, "learning_rate": 2.400357142857143e-05, "loss": 1.4942, "step": 3639500 }, { "epoch": 18.46, "learning_rate": 2.4e-05, "loss": 1.4942, "step": 3640000 }, { "epoch": 18.46, "learning_rate": 2.399642857142857e-05, "loss": 1.5016, "step": 3640500 }, { "epoch": 18.46, "learning_rate": 2.3992857142857146e-05, "loss": 1.502, "step": 3641000 }, { "epoch": 18.46, "learning_rate": 2.3989285714285713e-05, "loss": 1.4971, "step": 3641500 }, { "epoch": 18.47, "learning_rate": 2.3985714285714287e-05, "loss": 1.503, "step": 3642000 }, { "epoch": 18.47, "learning_rate": 2.3982142857142858e-05, "loss": 1.5051, "step": 3642500 }, { "epoch": 18.47, "learning_rate": 2.397857142857143e-05, "loss": 1.5004, "step": 3643000 }, { "epoch": 18.48, "learning_rate": 2.3975e-05, "loss": 1.4959, "step": 3643500 }, { "epoch": 18.48, "learning_rate": 2.397142857142857e-05, "loss": 1.4858, "step": 3644000 }, { "epoch": 18.48, "learning_rate": 2.3967857142857145e-05, "loss": 1.486, "step": 3644500 }, { "epoch": 18.48, "learning_rate": 2.3964285714285713e-05, "loss": 1.4942, "step": 3645000 }, { "epoch": 18.49, "learning_rate": 2.3960714285714287e-05, "loss": 1.4892, "step": 3645500 }, { "epoch": 18.49, "learning_rate": 2.3957142857142858e-05, "loss": 1.4912, "step": 3646000 }, { "epoch": 18.49, "learning_rate": 2.3953571428571432e-05, "loss": 1.4917, "step": 3646500 }, { "epoch": 18.49, "learning_rate": 2.395e-05, "loss": 1.4958, "step": 3647000 }, { "epoch": 18.5, "learning_rate": 2.3946428571428574e-05, "loss": 1.4909, "step": 3647500 }, { "epoch": 18.5, "learning_rate": 2.3942857142857144e-05, "loss": 1.5056, "step": 3648000 }, { "epoch": 18.5, "learning_rate": 2.3939285714285715e-05, "loss": 1.4963, "step": 3648500 }, { "epoch": 18.5, "learning_rate": 2.3935714285714286e-05, "loss": 1.5001, "step": 3649000 }, { "epoch": 18.51, "learning_rate": 2.3932142857142857e-05, "loss": 1.5032, "step": 3649500 }, { "epoch": 18.51, "learning_rate": 2.392857142857143e-05, "loss": 1.4896, "step": 3650000 }, { "epoch": 18.51, "learning_rate": 2.3925e-05, "loss": 1.4885, "step": 3650500 }, { "epoch": 18.52, "learning_rate": 2.3921428571428573e-05, "loss": 1.5039, "step": 3651000 }, { "epoch": 18.52, "learning_rate": 2.3917857142857144e-05, "loss": 1.4885, "step": 3651500 }, { "epoch": 18.52, "learning_rate": 2.3914285714285715e-05, "loss": 1.4955, "step": 3652000 }, { "epoch": 18.52, "learning_rate": 2.3910714285714286e-05, "loss": 1.4952, "step": 3652500 }, { "epoch": 18.53, "learning_rate": 2.390714285714286e-05, "loss": 1.5001, "step": 3653000 }, { "epoch": 18.53, "learning_rate": 2.390357142857143e-05, "loss": 1.4851, "step": 3653500 }, { "epoch": 18.53, "learning_rate": 2.39e-05, "loss": 1.4933, "step": 3654000 }, { "epoch": 18.53, "learning_rate": 2.3896428571428572e-05, "loss": 1.4877, "step": 3654500 }, { "epoch": 18.54, "learning_rate": 2.3892857142857143e-05, "loss": 1.486, "step": 3655000 }, { "epoch": 18.54, "learning_rate": 2.3889285714285714e-05, "loss": 1.4983, "step": 3655500 }, { "epoch": 18.54, "learning_rate": 2.3885714285714285e-05, "loss": 1.4896, "step": 3656000 }, { "epoch": 18.54, "learning_rate": 2.388214285714286e-05, "loss": 1.5, "step": 3656500 }, { "epoch": 18.55, "learning_rate": 2.387857142857143e-05, "loss": 1.4897, "step": 3657000 }, { "epoch": 18.55, "learning_rate": 2.3875e-05, "loss": 1.4872, "step": 3657500 }, { "epoch": 18.55, "learning_rate": 2.3871428571428572e-05, "loss": 1.4841, "step": 3658000 }, { "epoch": 18.56, "learning_rate": 2.3867857142857146e-05, "loss": 1.5012, "step": 3658500 }, { "epoch": 18.56, "learning_rate": 2.3864285714285717e-05, "loss": 1.4891, "step": 3659000 }, { "epoch": 18.56, "learning_rate": 2.3860714285714288e-05, "loss": 1.499, "step": 3659500 }, { "epoch": 18.56, "learning_rate": 2.385714285714286e-05, "loss": 1.4979, "step": 3660000 }, { "epoch": 18.57, "learning_rate": 2.385357142857143e-05, "loss": 1.4978, "step": 3660500 }, { "epoch": 18.57, "learning_rate": 2.385e-05, "loss": 1.4922, "step": 3661000 }, { "epoch": 18.57, "learning_rate": 2.384642857142857e-05, "loss": 1.4888, "step": 3661500 }, { "epoch": 18.57, "learning_rate": 2.3842857142857145e-05, "loss": 1.4891, "step": 3662000 }, { "epoch": 18.58, "learning_rate": 2.3839285714285716e-05, "loss": 1.4891, "step": 3662500 }, { "epoch": 18.58, "learning_rate": 2.3835714285714287e-05, "loss": 1.4926, "step": 3663000 }, { "epoch": 18.58, "learning_rate": 2.3832142857142858e-05, "loss": 1.4829, "step": 3663500 }, { "epoch": 18.58, "learning_rate": 2.3828571428571432e-05, "loss": 1.4895, "step": 3664000 }, { "epoch": 18.59, "learning_rate": 2.3825e-05, "loss": 1.5004, "step": 3664500 }, { "epoch": 18.59, "learning_rate": 2.3821428571428574e-05, "loss": 1.4955, "step": 3665000 }, { "epoch": 18.59, "learning_rate": 2.3817857142857145e-05, "loss": 1.4955, "step": 3665500 }, { "epoch": 18.6, "learning_rate": 2.3814285714285716e-05, "loss": 1.4941, "step": 3666000 }, { "epoch": 18.6, "learning_rate": 2.3810714285714287e-05, "loss": 1.4932, "step": 3666500 }, { "epoch": 18.6, "learning_rate": 2.3807142857142857e-05, "loss": 1.4944, "step": 3667000 }, { "epoch": 18.6, "learning_rate": 2.380357142857143e-05, "loss": 1.4925, "step": 3667500 }, { "epoch": 18.61, "learning_rate": 2.38e-05, "loss": 1.4932, "step": 3668000 }, { "epoch": 18.61, "learning_rate": 2.3796428571428573e-05, "loss": 1.4893, "step": 3668500 }, { "epoch": 18.61, "learning_rate": 2.3792857142857144e-05, "loss": 1.4857, "step": 3669000 }, { "epoch": 18.61, "learning_rate": 2.3789285714285715e-05, "loss": 1.5047, "step": 3669500 }, { "epoch": 18.62, "learning_rate": 2.3785714285714286e-05, "loss": 1.4888, "step": 3670000 }, { "epoch": 18.62, "learning_rate": 2.3782142857142857e-05, "loss": 1.4929, "step": 3670500 }, { "epoch": 18.62, "learning_rate": 2.377857142857143e-05, "loss": 1.4987, "step": 3671000 }, { "epoch": 18.62, "learning_rate": 2.3775e-05, "loss": 1.4917, "step": 3671500 }, { "epoch": 18.63, "learning_rate": 2.3771428571428573e-05, "loss": 1.5015, "step": 3672000 }, { "epoch": 18.63, "learning_rate": 2.3767857142857144e-05, "loss": 1.4863, "step": 3672500 }, { "epoch": 18.63, "learning_rate": 2.3764285714285718e-05, "loss": 1.4995, "step": 3673000 }, { "epoch": 18.64, "learning_rate": 2.3760714285714285e-05, "loss": 1.4875, "step": 3673500 }, { "epoch": 18.64, "learning_rate": 2.375714285714286e-05, "loss": 1.4987, "step": 3674000 }, { "epoch": 18.64, "learning_rate": 2.375357142857143e-05, "loss": 1.5037, "step": 3674500 }, { "epoch": 18.64, "learning_rate": 2.375e-05, "loss": 1.5021, "step": 3675000 }, { "epoch": 18.65, "learning_rate": 2.3746428571428572e-05, "loss": 1.4882, "step": 3675500 }, { "epoch": 18.65, "learning_rate": 2.3742857142857143e-05, "loss": 1.4909, "step": 3676000 }, { "epoch": 18.65, "learning_rate": 2.3739285714285717e-05, "loss": 1.4933, "step": 3676500 }, { "epoch": 18.65, "learning_rate": 2.3735714285714285e-05, "loss": 1.4896, "step": 3677000 }, { "epoch": 18.66, "learning_rate": 2.373214285714286e-05, "loss": 1.4991, "step": 3677500 }, { "epoch": 18.66, "learning_rate": 2.372857142857143e-05, "loss": 1.5059, "step": 3678000 }, { "epoch": 18.66, "learning_rate": 2.3725e-05, "loss": 1.5062, "step": 3678500 }, { "epoch": 18.66, "learning_rate": 2.372142857142857e-05, "loss": 1.4889, "step": 3679000 }, { "epoch": 18.67, "learning_rate": 2.3717857142857146e-05, "loss": 1.4934, "step": 3679500 }, { "epoch": 18.67, "learning_rate": 2.3714285714285717e-05, "loss": 1.4917, "step": 3680000 }, { "epoch": 18.67, "learning_rate": 2.3710714285714288e-05, "loss": 1.4991, "step": 3680500 }, { "epoch": 18.67, "learning_rate": 2.370714285714286e-05, "loss": 1.4946, "step": 3681000 }, { "epoch": 18.68, "learning_rate": 2.370357142857143e-05, "loss": 1.4899, "step": 3681500 }, { "epoch": 18.68, "learning_rate": 2.37e-05, "loss": 1.4924, "step": 3682000 }, { "epoch": 18.68, "learning_rate": 2.369642857142857e-05, "loss": 1.4869, "step": 3682500 }, { "epoch": 18.69, "learning_rate": 2.3692857142857145e-05, "loss": 1.4936, "step": 3683000 }, { "epoch": 18.69, "learning_rate": 2.3689285714285716e-05, "loss": 1.4862, "step": 3683500 }, { "epoch": 18.69, "learning_rate": 2.3685714285714287e-05, "loss": 1.4918, "step": 3684000 }, { "epoch": 18.69, "learning_rate": 2.3682142857142858e-05, "loss": 1.4921, "step": 3684500 }, { "epoch": 18.7, "learning_rate": 2.3678571428571432e-05, "loss": 1.4902, "step": 3685000 }, { "epoch": 18.7, "learning_rate": 2.3675e-05, "loss": 1.4972, "step": 3685500 }, { "epoch": 18.7, "learning_rate": 2.3671428571428574e-05, "loss": 1.497, "step": 3686000 }, { "epoch": 18.7, "learning_rate": 2.3667857142857145e-05, "loss": 1.4864, "step": 3686500 }, { "epoch": 18.71, "learning_rate": 2.3664285714285715e-05, "loss": 1.4928, "step": 3687000 }, { "epoch": 18.71, "learning_rate": 2.3660714285714286e-05, "loss": 1.4865, "step": 3687500 }, { "epoch": 18.71, "learning_rate": 2.3657142857142857e-05, "loss": 1.4958, "step": 3688000 }, { "epoch": 18.71, "learning_rate": 2.365357142857143e-05, "loss": 1.4879, "step": 3688500 }, { "epoch": 18.72, "learning_rate": 2.365e-05, "loss": 1.4967, "step": 3689000 }, { "epoch": 18.72, "learning_rate": 2.3646428571428573e-05, "loss": 1.4873, "step": 3689500 }, { "epoch": 18.72, "learning_rate": 2.3642857142857144e-05, "loss": 1.4878, "step": 3690000 }, { "epoch": 18.73, "learning_rate": 2.3639285714285718e-05, "loss": 1.4881, "step": 3690500 }, { "epoch": 18.73, "learning_rate": 2.3635714285714286e-05, "loss": 1.4859, "step": 3691000 }, { "epoch": 18.73, "learning_rate": 2.363214285714286e-05, "loss": 1.4874, "step": 3691500 }, { "epoch": 18.73, "learning_rate": 2.362857142857143e-05, "loss": 1.4932, "step": 3692000 }, { "epoch": 18.74, "learning_rate": 2.3624999999999998e-05, "loss": 1.4932, "step": 3692500 }, { "epoch": 18.74, "learning_rate": 2.3621428571428573e-05, "loss": 1.4947, "step": 3693000 }, { "epoch": 18.74, "learning_rate": 2.3617857142857143e-05, "loss": 1.5015, "step": 3693500 }, { "epoch": 18.74, "learning_rate": 2.3614285714285718e-05, "loss": 1.4817, "step": 3694000 }, { "epoch": 18.75, "learning_rate": 2.3610714285714285e-05, "loss": 1.4933, "step": 3694500 }, { "epoch": 18.75, "learning_rate": 2.360714285714286e-05, "loss": 1.4974, "step": 3695000 }, { "epoch": 18.75, "learning_rate": 2.360357142857143e-05, "loss": 1.4972, "step": 3695500 }, { "epoch": 18.75, "learning_rate": 2.36e-05, "loss": 1.4971, "step": 3696000 }, { "epoch": 18.76, "learning_rate": 2.3596428571428572e-05, "loss": 1.4934, "step": 3696500 }, { "epoch": 18.76, "learning_rate": 2.3592857142857143e-05, "loss": 1.4888, "step": 3697000 }, { "epoch": 18.76, "learning_rate": 2.3589285714285717e-05, "loss": 1.49, "step": 3697500 }, { "epoch": 18.77, "learning_rate": 2.3585714285714284e-05, "loss": 1.4873, "step": 3698000 }, { "epoch": 18.77, "learning_rate": 2.358214285714286e-05, "loss": 1.4906, "step": 3698500 }, { "epoch": 18.77, "learning_rate": 2.357857142857143e-05, "loss": 1.4862, "step": 3699000 }, { "epoch": 18.77, "learning_rate": 2.3575e-05, "loss": 1.5, "step": 3699500 }, { "epoch": 18.78, "learning_rate": 2.357142857142857e-05, "loss": 1.4908, "step": 3700000 }, { "epoch": 18.78, "learning_rate": 2.3567857142857146e-05, "loss": 1.491, "step": 3700500 }, { "epoch": 18.78, "learning_rate": 2.3564285714285716e-05, "loss": 1.4909, "step": 3701000 }, { "epoch": 18.78, "learning_rate": 2.3560714285714287e-05, "loss": 1.4945, "step": 3701500 }, { "epoch": 18.79, "learning_rate": 2.3557142857142858e-05, "loss": 1.4867, "step": 3702000 }, { "epoch": 18.79, "learning_rate": 2.355357142857143e-05, "loss": 1.4913, "step": 3702500 }, { "epoch": 18.79, "learning_rate": 2.355e-05, "loss": 1.4969, "step": 3703000 }, { "epoch": 18.79, "learning_rate": 2.354642857142857e-05, "loss": 1.494, "step": 3703500 }, { "epoch": 18.8, "learning_rate": 2.3542857142857145e-05, "loss": 1.4983, "step": 3704000 }, { "epoch": 18.8, "learning_rate": 2.3539285714285716e-05, "loss": 1.4943, "step": 3704500 }, { "epoch": 18.8, "learning_rate": 2.3535714285714287e-05, "loss": 1.4917, "step": 3705000 }, { "epoch": 18.81, "learning_rate": 2.3532142857142858e-05, "loss": 1.4913, "step": 3705500 }, { "epoch": 18.81, "learning_rate": 2.3528571428571432e-05, "loss": 1.4874, "step": 3706000 }, { "epoch": 18.81, "learning_rate": 2.3525e-05, "loss": 1.491, "step": 3706500 }, { "epoch": 18.81, "learning_rate": 2.3521428571428573e-05, "loss": 1.5073, "step": 3707000 }, { "epoch": 18.82, "learning_rate": 2.3517857142857144e-05, "loss": 1.4861, "step": 3707500 }, { "epoch": 18.82, "learning_rate": 2.3514285714285715e-05, "loss": 1.4882, "step": 3708000 }, { "epoch": 18.82, "learning_rate": 2.3510714285714286e-05, "loss": 1.4858, "step": 3708500 }, { "epoch": 18.82, "learning_rate": 2.3507142857142857e-05, "loss": 1.4879, "step": 3709000 }, { "epoch": 18.83, "learning_rate": 2.350357142857143e-05, "loss": 1.4919, "step": 3709500 }, { "epoch": 18.83, "learning_rate": 2.35e-05, "loss": 1.4922, "step": 3710000 }, { "epoch": 18.83, "learning_rate": 2.3496428571428573e-05, "loss": 1.4977, "step": 3710500 }, { "epoch": 18.83, "learning_rate": 2.3492857142857144e-05, "loss": 1.485, "step": 3711000 }, { "epoch": 18.84, "learning_rate": 2.3489285714285718e-05, "loss": 1.5033, "step": 3711500 }, { "epoch": 18.84, "learning_rate": 2.3485714285714285e-05, "loss": 1.5078, "step": 3712000 }, { "epoch": 18.84, "learning_rate": 2.348214285714286e-05, "loss": 1.5128, "step": 3712500 }, { "epoch": 18.85, "learning_rate": 2.347857142857143e-05, "loss": 1.5033, "step": 3713000 }, { "epoch": 18.85, "learning_rate": 2.3475e-05, "loss": 1.507, "step": 3713500 }, { "epoch": 18.85, "learning_rate": 2.3471428571428572e-05, "loss": 1.5001, "step": 3714000 }, { "epoch": 18.85, "learning_rate": 2.3467857142857143e-05, "loss": 1.4993, "step": 3714500 }, { "epoch": 18.86, "learning_rate": 2.3464285714285717e-05, "loss": 1.4922, "step": 3715000 }, { "epoch": 18.86, "learning_rate": 2.3460714285714285e-05, "loss": 1.4878, "step": 3715500 }, { "epoch": 18.86, "learning_rate": 2.345714285714286e-05, "loss": 1.4966, "step": 3716000 }, { "epoch": 18.86, "learning_rate": 2.345357142857143e-05, "loss": 1.4839, "step": 3716500 }, { "epoch": 18.87, "learning_rate": 2.345e-05, "loss": 1.4949, "step": 3717000 }, { "epoch": 18.87, "learning_rate": 2.344642857142857e-05, "loss": 1.4976, "step": 3717500 }, { "epoch": 18.87, "learning_rate": 2.3442857142857143e-05, "loss": 1.5072, "step": 3718000 }, { "epoch": 18.87, "learning_rate": 2.3439285714285717e-05, "loss": 1.5073, "step": 3718500 }, { "epoch": 18.88, "learning_rate": 2.3435714285714284e-05, "loss": 1.5059, "step": 3719000 }, { "epoch": 18.88, "learning_rate": 2.343214285714286e-05, "loss": 1.5063, "step": 3719500 }, { "epoch": 18.88, "learning_rate": 2.342857142857143e-05, "loss": 1.4796, "step": 3720000 }, { "epoch": 18.89, "learning_rate": 2.3425000000000004e-05, "loss": 1.5042, "step": 3720500 }, { "epoch": 18.89, "learning_rate": 2.342142857142857e-05, "loss": 1.4796, "step": 3721000 }, { "epoch": 18.89, "learning_rate": 2.3417857142857145e-05, "loss": 1.494, "step": 3721500 }, { "epoch": 18.89, "learning_rate": 2.3414285714285716e-05, "loss": 1.5024, "step": 3722000 }, { "epoch": 18.9, "learning_rate": 2.3410714285714287e-05, "loss": 1.4936, "step": 3722500 }, { "epoch": 18.9, "learning_rate": 2.3407142857142858e-05, "loss": 1.508, "step": 3723000 }, { "epoch": 18.9, "learning_rate": 2.340357142857143e-05, "loss": 1.4913, "step": 3723500 }, { "epoch": 18.9, "learning_rate": 2.3400000000000003e-05, "loss": 1.491, "step": 3724000 }, { "epoch": 18.91, "learning_rate": 2.339642857142857e-05, "loss": 1.4978, "step": 3724500 }, { "epoch": 18.91, "learning_rate": 2.3392857142857145e-05, "loss": 1.4807, "step": 3725000 }, { "epoch": 18.91, "learning_rate": 2.3389285714285716e-05, "loss": 1.4986, "step": 3725500 }, { "epoch": 18.91, "learning_rate": 2.3385714285714286e-05, "loss": 1.5033, "step": 3726000 }, { "epoch": 18.92, "learning_rate": 2.3382142857142857e-05, "loss": 1.503, "step": 3726500 }, { "epoch": 18.92, "learning_rate": 2.337857142857143e-05, "loss": 1.4985, "step": 3727000 }, { "epoch": 18.92, "learning_rate": 2.3375000000000002e-05, "loss": 1.4954, "step": 3727500 }, { "epoch": 18.93, "learning_rate": 2.3371428571428573e-05, "loss": 1.4935, "step": 3728000 }, { "epoch": 18.93, "learning_rate": 2.3367857142857144e-05, "loss": 1.4944, "step": 3728500 }, { "epoch": 18.93, "learning_rate": 2.3364285714285715e-05, "loss": 1.4948, "step": 3729000 }, { "epoch": 18.93, "learning_rate": 2.3360714285714286e-05, "loss": 1.4933, "step": 3729500 }, { "epoch": 18.94, "learning_rate": 2.3357142857142857e-05, "loss": 1.4908, "step": 3730000 }, { "epoch": 18.94, "learning_rate": 2.335357142857143e-05, "loss": 1.4954, "step": 3730500 }, { "epoch": 18.94, "learning_rate": 2.3350000000000002e-05, "loss": 1.4952, "step": 3731000 }, { "epoch": 18.94, "learning_rate": 2.3346428571428573e-05, "loss": 1.4897, "step": 3731500 }, { "epoch": 18.95, "learning_rate": 2.3342857142857143e-05, "loss": 1.4933, "step": 3732000 }, { "epoch": 18.95, "learning_rate": 2.3339285714285718e-05, "loss": 1.4987, "step": 3732500 }, { "epoch": 18.95, "learning_rate": 2.3335714285714285e-05, "loss": 1.4951, "step": 3733000 }, { "epoch": 18.95, "learning_rate": 2.333214285714286e-05, "loss": 1.491, "step": 3733500 }, { "epoch": 18.96, "learning_rate": 2.332857142857143e-05, "loss": 1.5088, "step": 3734000 }, { "epoch": 18.96, "learning_rate": 2.3325e-05, "loss": 1.4885, "step": 3734500 }, { "epoch": 18.96, "learning_rate": 2.3321428571428572e-05, "loss": 1.4943, "step": 3735000 }, { "epoch": 18.97, "learning_rate": 2.3317857142857143e-05, "loss": 1.5006, "step": 3735500 }, { "epoch": 18.97, "learning_rate": 2.3314285714285717e-05, "loss": 1.4762, "step": 3736000 }, { "epoch": 18.97, "learning_rate": 2.3310714285714285e-05, "loss": 1.4911, "step": 3736500 }, { "epoch": 18.97, "learning_rate": 2.330714285714286e-05, "loss": 1.4969, "step": 3737000 }, { "epoch": 18.98, "learning_rate": 2.330357142857143e-05, "loss": 1.4841, "step": 3737500 }, { "epoch": 18.98, "learning_rate": 2.3300000000000004e-05, "loss": 1.4875, "step": 3738000 }, { "epoch": 18.98, "learning_rate": 2.329642857142857e-05, "loss": 1.4948, "step": 3738500 }, { "epoch": 18.98, "learning_rate": 2.3292857142857146e-05, "loss": 1.4932, "step": 3739000 }, { "epoch": 18.99, "learning_rate": 2.3289285714285716e-05, "loss": 1.4836, "step": 3739500 }, { "epoch": 18.99, "learning_rate": 2.3285714285714287e-05, "loss": 1.4954, "step": 3740000 }, { "epoch": 18.99, "learning_rate": 2.3282142857142858e-05, "loss": 1.4823, "step": 3740500 }, { "epoch": 18.99, "learning_rate": 2.327857142857143e-05, "loss": 1.484, "step": 3741000 }, { "epoch": 19.0, "learning_rate": 2.3275000000000003e-05, "loss": 1.49, "step": 3741500 }, { "epoch": 19.0, "learning_rate": 2.327142857142857e-05, "loss": 1.4966, "step": 3742000 }, { "epoch": 19.0, "learning_rate": 2.3267857142857145e-05, "loss": 1.4871, "step": 3742500 }, { "epoch": 19.01, "learning_rate": 2.3264285714285716e-05, "loss": 1.4845, "step": 3743000 }, { "epoch": 19.01, "learning_rate": 2.3260714285714287e-05, "loss": 1.4887, "step": 3743500 }, { "epoch": 19.01, "learning_rate": 2.3257142857142858e-05, "loss": 1.4895, "step": 3744000 }, { "epoch": 19.01, "learning_rate": 2.325357142857143e-05, "loss": 1.482, "step": 3744500 }, { "epoch": 19.02, "learning_rate": 2.3250000000000003e-05, "loss": 1.4837, "step": 3745000 }, { "epoch": 19.02, "learning_rate": 2.324642857142857e-05, "loss": 1.4758, "step": 3745500 }, { "epoch": 19.02, "learning_rate": 2.3242857142857144e-05, "loss": 1.4904, "step": 3746000 }, { "epoch": 19.02, "learning_rate": 2.3239285714285715e-05, "loss": 1.4893, "step": 3746500 }, { "epoch": 19.03, "learning_rate": 2.3235714285714286e-05, "loss": 1.4866, "step": 3747000 }, { "epoch": 19.03, "learning_rate": 2.3232142857142857e-05, "loss": 1.4905, "step": 3747500 }, { "epoch": 19.03, "learning_rate": 2.322857142857143e-05, "loss": 1.4784, "step": 3748000 }, { "epoch": 19.03, "learning_rate": 2.3225000000000002e-05, "loss": 1.487, "step": 3748500 }, { "epoch": 19.04, "learning_rate": 2.3221428571428573e-05, "loss": 1.485, "step": 3749000 }, { "epoch": 19.04, "learning_rate": 2.3217857142857144e-05, "loss": 1.4799, "step": 3749500 }, { "epoch": 19.04, "learning_rate": 2.3214285714285715e-05, "loss": 1.497, "step": 3750000 }, { "epoch": 19.05, "learning_rate": 2.3210714285714286e-05, "loss": 1.4887, "step": 3750500 }, { "epoch": 19.05, "learning_rate": 2.3207142857142856e-05, "loss": 1.4917, "step": 3751000 }, { "epoch": 19.05, "learning_rate": 2.320357142857143e-05, "loss": 1.4849, "step": 3751500 }, { "epoch": 19.05, "learning_rate": 2.32e-05, "loss": 1.4896, "step": 3752000 }, { "epoch": 19.06, "learning_rate": 2.3196428571428572e-05, "loss": 1.4865, "step": 3752500 }, { "epoch": 19.06, "learning_rate": 2.3192857142857143e-05, "loss": 1.4816, "step": 3753000 }, { "epoch": 19.06, "learning_rate": 2.3189285714285717e-05, "loss": 1.484, "step": 3753500 }, { "epoch": 19.06, "learning_rate": 2.3185714285714285e-05, "loss": 1.4881, "step": 3754000 }, { "epoch": 19.07, "learning_rate": 2.318214285714286e-05, "loss": 1.4829, "step": 3754500 }, { "epoch": 19.07, "learning_rate": 2.317857142857143e-05, "loss": 1.4717, "step": 3755000 }, { "epoch": 19.07, "learning_rate": 2.3175e-05, "loss": 1.4906, "step": 3755500 }, { "epoch": 19.07, "learning_rate": 2.3171428571428572e-05, "loss": 1.4788, "step": 3756000 }, { "epoch": 19.08, "learning_rate": 2.3167857142857143e-05, "loss": 1.4811, "step": 3756500 }, { "epoch": 19.08, "learning_rate": 2.3164285714285717e-05, "loss": 1.4782, "step": 3757000 }, { "epoch": 19.08, "learning_rate": 2.3160714285714284e-05, "loss": 1.485, "step": 3757500 }, { "epoch": 19.08, "learning_rate": 2.315714285714286e-05, "loss": 1.4873, "step": 3758000 }, { "epoch": 19.09, "learning_rate": 2.315357142857143e-05, "loss": 1.4898, "step": 3758500 }, { "epoch": 19.09, "learning_rate": 2.3150000000000004e-05, "loss": 1.4849, "step": 3759000 }, { "epoch": 19.09, "learning_rate": 2.314642857142857e-05, "loss": 1.478, "step": 3759500 }, { "epoch": 19.1, "learning_rate": 2.3142857142857145e-05, "loss": 1.4866, "step": 3760000 }, { "epoch": 19.1, "learning_rate": 2.3139285714285716e-05, "loss": 1.4842, "step": 3760500 }, { "epoch": 19.1, "learning_rate": 2.3135714285714287e-05, "loss": 1.4823, "step": 3761000 }, { "epoch": 19.1, "learning_rate": 2.3132142857142858e-05, "loss": 1.4878, "step": 3761500 }, { "epoch": 19.11, "learning_rate": 2.312857142857143e-05, "loss": 1.4707, "step": 3762000 }, { "epoch": 19.11, "learning_rate": 2.3125000000000003e-05, "loss": 1.4896, "step": 3762500 }, { "epoch": 19.11, "learning_rate": 2.312142857142857e-05, "loss": 1.4856, "step": 3763000 }, { "epoch": 19.11, "learning_rate": 2.3117857142857145e-05, "loss": 1.486, "step": 3763500 }, { "epoch": 19.12, "learning_rate": 2.3114285714285716e-05, "loss": 1.4721, "step": 3764000 }, { "epoch": 19.12, "learning_rate": 2.3110714285714286e-05, "loss": 1.4741, "step": 3764500 }, { "epoch": 19.12, "learning_rate": 2.3107142857142857e-05, "loss": 1.4775, "step": 3765000 }, { "epoch": 19.12, "learning_rate": 2.310357142857143e-05, "loss": 1.483, "step": 3765500 }, { "epoch": 19.13, "learning_rate": 2.3100000000000002e-05, "loss": 1.4868, "step": 3766000 }, { "epoch": 19.13, "learning_rate": 2.3096428571428573e-05, "loss": 1.4848, "step": 3766500 }, { "epoch": 19.13, "learning_rate": 2.3092857142857144e-05, "loss": 1.4812, "step": 3767000 }, { "epoch": 19.14, "learning_rate": 2.3089285714285715e-05, "loss": 1.4915, "step": 3767500 }, { "epoch": 19.14, "learning_rate": 2.3085714285714286e-05, "loss": 1.4848, "step": 3768000 }, { "epoch": 19.14, "learning_rate": 2.3082142857142857e-05, "loss": 1.4865, "step": 3768500 }, { "epoch": 19.14, "learning_rate": 2.307857142857143e-05, "loss": 1.4875, "step": 3769000 }, { "epoch": 19.15, "learning_rate": 2.3075000000000002e-05, "loss": 1.4848, "step": 3769500 }, { "epoch": 19.15, "learning_rate": 2.3071428571428573e-05, "loss": 1.4905, "step": 3770000 }, { "epoch": 19.15, "learning_rate": 2.3067857142857144e-05, "loss": 1.4957, "step": 3770500 }, { "epoch": 19.15, "learning_rate": 2.3064285714285714e-05, "loss": 1.4861, "step": 3771000 }, { "epoch": 19.16, "learning_rate": 2.3060714285714285e-05, "loss": 1.4829, "step": 3771500 }, { "epoch": 19.16, "learning_rate": 2.3057142857142856e-05, "loss": 1.4809, "step": 3772000 }, { "epoch": 19.16, "learning_rate": 2.305357142857143e-05, "loss": 1.4828, "step": 3772500 }, { "epoch": 19.16, "learning_rate": 2.305e-05, "loss": 1.4795, "step": 3773000 }, { "epoch": 19.17, "learning_rate": 2.3046428571428572e-05, "loss": 1.4842, "step": 3773500 }, { "epoch": 19.17, "learning_rate": 2.3042857142857143e-05, "loss": 1.4872, "step": 3774000 }, { "epoch": 19.17, "learning_rate": 2.3039285714285717e-05, "loss": 1.4941, "step": 3774500 }, { "epoch": 19.18, "learning_rate": 2.3035714285714285e-05, "loss": 1.4904, "step": 3775000 }, { "epoch": 19.18, "learning_rate": 2.303214285714286e-05, "loss": 1.4811, "step": 3775500 }, { "epoch": 19.18, "learning_rate": 2.302857142857143e-05, "loss": 1.4724, "step": 3776000 }, { "epoch": 19.18, "learning_rate": 2.3025e-05, "loss": 1.4809, "step": 3776500 }, { "epoch": 19.19, "learning_rate": 2.302142857142857e-05, "loss": 1.4829, "step": 3777000 }, { "epoch": 19.19, "learning_rate": 2.3017857142857142e-05, "loss": 1.4734, "step": 3777500 }, { "epoch": 19.19, "learning_rate": 2.3014285714285717e-05, "loss": 1.4833, "step": 3778000 }, { "epoch": 19.19, "learning_rate": 2.3010714285714284e-05, "loss": 1.4817, "step": 3778500 }, { "epoch": 19.2, "learning_rate": 2.3007142857142858e-05, "loss": 1.4811, "step": 3779000 }, { "epoch": 19.2, "learning_rate": 2.300357142857143e-05, "loss": 1.4822, "step": 3779500 }, { "epoch": 19.2, "learning_rate": 2.3000000000000003e-05, "loss": 1.4842, "step": 3780000 }, { "epoch": 19.2, "learning_rate": 2.299642857142857e-05, "loss": 1.4771, "step": 3780500 }, { "epoch": 19.21, "learning_rate": 2.2992857142857145e-05, "loss": 1.4828, "step": 3781000 }, { "epoch": 19.21, "learning_rate": 2.2989285714285716e-05, "loss": 1.4838, "step": 3781500 }, { "epoch": 19.21, "learning_rate": 2.2985714285714287e-05, "loss": 1.485, "step": 3782000 }, { "epoch": 19.22, "learning_rate": 2.2982142857142858e-05, "loss": 1.4765, "step": 3782500 }, { "epoch": 19.22, "learning_rate": 2.297857142857143e-05, "loss": 1.4835, "step": 3783000 }, { "epoch": 19.22, "learning_rate": 2.2975000000000003e-05, "loss": 1.4926, "step": 3783500 }, { "epoch": 19.22, "learning_rate": 2.297142857142857e-05, "loss": 1.4731, "step": 3784000 }, { "epoch": 19.23, "learning_rate": 2.2967857142857145e-05, "loss": 1.4836, "step": 3784500 }, { "epoch": 19.23, "learning_rate": 2.2964285714285715e-05, "loss": 1.4735, "step": 3785000 }, { "epoch": 19.23, "learning_rate": 2.296071428571429e-05, "loss": 1.4939, "step": 3785500 }, { "epoch": 19.23, "learning_rate": 2.2957142857142857e-05, "loss": 1.5004, "step": 3786000 }, { "epoch": 19.24, "learning_rate": 2.295357142857143e-05, "loss": 1.4801, "step": 3786500 }, { "epoch": 19.24, "learning_rate": 2.2950000000000002e-05, "loss": 1.4823, "step": 3787000 }, { "epoch": 19.24, "learning_rate": 2.2946428571428573e-05, "loss": 1.472, "step": 3787500 }, { "epoch": 19.24, "learning_rate": 2.2942857142857144e-05, "loss": 1.487, "step": 3788000 }, { "epoch": 19.25, "learning_rate": 2.2939285714285715e-05, "loss": 1.4854, "step": 3788500 }, { "epoch": 19.25, "learning_rate": 2.293571428571429e-05, "loss": 1.4838, "step": 3789000 }, { "epoch": 19.25, "learning_rate": 2.2932142857142856e-05, "loss": 1.4825, "step": 3789500 }, { "epoch": 19.26, "learning_rate": 2.292857142857143e-05, "loss": 1.4775, "step": 3790000 }, { "epoch": 19.26, "learning_rate": 2.2925e-05, "loss": 1.4894, "step": 3790500 }, { "epoch": 19.26, "learning_rate": 2.2921428571428572e-05, "loss": 1.4722, "step": 3791000 }, { "epoch": 19.26, "learning_rate": 2.2917857142857143e-05, "loss": 1.4886, "step": 3791500 }, { "epoch": 19.27, "learning_rate": 2.2914285714285718e-05, "loss": 1.5032, "step": 3792000 }, { "epoch": 19.27, "learning_rate": 2.291071428571429e-05, "loss": 1.4881, "step": 3792500 }, { "epoch": 19.27, "learning_rate": 2.290714285714286e-05, "loss": 1.4867, "step": 3793000 }, { "epoch": 19.27, "learning_rate": 2.290357142857143e-05, "loss": 1.4922, "step": 3793500 }, { "epoch": 19.28, "learning_rate": 2.29e-05, "loss": 1.4793, "step": 3794000 }, { "epoch": 19.28, "learning_rate": 2.2896428571428572e-05, "loss": 1.4799, "step": 3794500 }, { "epoch": 19.28, "learning_rate": 2.2892857142857143e-05, "loss": 1.4907, "step": 3795000 }, { "epoch": 19.28, "learning_rate": 2.2889285714285717e-05, "loss": 1.488, "step": 3795500 }, { "epoch": 19.29, "learning_rate": 2.2885714285714288e-05, "loss": 1.4961, "step": 3796000 }, { "epoch": 19.29, "learning_rate": 2.288214285714286e-05, "loss": 1.4938, "step": 3796500 }, { "epoch": 19.29, "learning_rate": 2.287857142857143e-05, "loss": 1.4837, "step": 3797000 }, { "epoch": 19.3, "learning_rate": 2.2875e-05, "loss": 1.491, "step": 3797500 }, { "epoch": 19.3, "learning_rate": 2.287142857142857e-05, "loss": 1.489, "step": 3798000 }, { "epoch": 19.3, "learning_rate": 2.2867857142857142e-05, "loss": 1.4892, "step": 3798500 }, { "epoch": 19.3, "learning_rate": 2.2864285714285716e-05, "loss": 1.4855, "step": 3799000 }, { "epoch": 19.31, "learning_rate": 2.2860714285714287e-05, "loss": 1.4823, "step": 3799500 }, { "epoch": 19.31, "learning_rate": 2.2857142857142858e-05, "loss": 1.4844, "step": 3800000 }, { "epoch": 19.31, "learning_rate": 2.285357142857143e-05, "loss": 1.4848, "step": 3800500 }, { "epoch": 19.31, "learning_rate": 2.2850000000000003e-05, "loss": 1.4854, "step": 3801000 }, { "epoch": 19.32, "learning_rate": 2.284642857142857e-05, "loss": 1.4823, "step": 3801500 }, { "epoch": 19.32, "learning_rate": 2.2842857142857145e-05, "loss": 1.4835, "step": 3802000 }, { "epoch": 19.32, "learning_rate": 2.2839285714285716e-05, "loss": 1.4801, "step": 3802500 }, { "epoch": 19.32, "learning_rate": 2.2835714285714287e-05, "loss": 1.4874, "step": 3803000 }, { "epoch": 19.33, "learning_rate": 2.2832142857142857e-05, "loss": 1.4885, "step": 3803500 }, { "epoch": 19.33, "learning_rate": 2.2828571428571428e-05, "loss": 1.4892, "step": 3804000 }, { "epoch": 19.33, "learning_rate": 2.2825000000000003e-05, "loss": 1.4841, "step": 3804500 }, { "epoch": 19.34, "learning_rate": 2.282142857142857e-05, "loss": 1.4835, "step": 3805000 }, { "epoch": 19.34, "learning_rate": 2.2817857142857144e-05, "loss": 1.4893, "step": 3805500 }, { "epoch": 19.34, "learning_rate": 2.2814285714285715e-05, "loss": 1.4738, "step": 3806000 }, { "epoch": 19.34, "learning_rate": 2.281071428571429e-05, "loss": 1.4774, "step": 3806500 }, { "epoch": 19.35, "learning_rate": 2.2807142857142857e-05, "loss": 1.4868, "step": 3807000 }, { "epoch": 19.35, "learning_rate": 2.280357142857143e-05, "loss": 1.4855, "step": 3807500 }, { "epoch": 19.35, "learning_rate": 2.2800000000000002e-05, "loss": 1.4891, "step": 3808000 }, { "epoch": 19.35, "learning_rate": 2.2796428571428573e-05, "loss": 1.4868, "step": 3808500 }, { "epoch": 19.36, "learning_rate": 2.2792857142857144e-05, "loss": 1.4768, "step": 3809000 }, { "epoch": 19.36, "learning_rate": 2.2789285714285715e-05, "loss": 1.4753, "step": 3809500 }, { "epoch": 19.36, "learning_rate": 2.278571428571429e-05, "loss": 1.4684, "step": 3810000 }, { "epoch": 19.36, "learning_rate": 2.2782142857142856e-05, "loss": 1.4784, "step": 3810500 }, { "epoch": 19.37, "learning_rate": 2.277857142857143e-05, "loss": 1.4891, "step": 3811000 }, { "epoch": 19.37, "learning_rate": 2.2775e-05, "loss": 1.4846, "step": 3811500 }, { "epoch": 19.37, "learning_rate": 2.2771428571428572e-05, "loss": 1.472, "step": 3812000 }, { "epoch": 19.38, "learning_rate": 2.2767857142857143e-05, "loss": 1.4911, "step": 3812500 }, { "epoch": 19.38, "learning_rate": 2.2764285714285717e-05, "loss": 1.4858, "step": 3813000 }, { "epoch": 19.38, "learning_rate": 2.2760714285714288e-05, "loss": 1.4741, "step": 3813500 }, { "epoch": 19.38, "learning_rate": 2.275714285714286e-05, "loss": 1.4768, "step": 3814000 }, { "epoch": 19.39, "learning_rate": 2.275357142857143e-05, "loss": 1.4802, "step": 3814500 }, { "epoch": 19.39, "learning_rate": 2.275e-05, "loss": 1.4799, "step": 3815000 }, { "epoch": 19.39, "learning_rate": 2.274642857142857e-05, "loss": 1.4943, "step": 3815500 }, { "epoch": 19.39, "learning_rate": 2.2742857142857142e-05, "loss": 1.4962, "step": 3816000 }, { "epoch": 19.4, "learning_rate": 2.2739285714285717e-05, "loss": 1.4825, "step": 3816500 }, { "epoch": 19.4, "learning_rate": 2.2735714285714288e-05, "loss": 1.4857, "step": 3817000 }, { "epoch": 19.4, "learning_rate": 2.273214285714286e-05, "loss": 1.4811, "step": 3817500 }, { "epoch": 19.4, "learning_rate": 2.272857142857143e-05, "loss": 1.4805, "step": 3818000 }, { "epoch": 19.41, "learning_rate": 2.2725000000000003e-05, "loss": 1.4875, "step": 3818500 }, { "epoch": 19.41, "learning_rate": 2.272142857142857e-05, "loss": 1.4797, "step": 3819000 }, { "epoch": 19.41, "learning_rate": 2.2717857142857142e-05, "loss": 1.4905, "step": 3819500 }, { "epoch": 19.42, "learning_rate": 2.2714285714285716e-05, "loss": 1.4794, "step": 3820000 }, { "epoch": 19.42, "learning_rate": 2.2710714285714287e-05, "loss": 1.4685, "step": 3820500 }, { "epoch": 19.42, "learning_rate": 2.2707142857142858e-05, "loss": 1.4754, "step": 3821000 }, { "epoch": 19.42, "learning_rate": 2.270357142857143e-05, "loss": 1.4817, "step": 3821500 }, { "epoch": 19.43, "learning_rate": 2.2700000000000003e-05, "loss": 1.476, "step": 3822000 }, { "epoch": 19.43, "learning_rate": 2.269642857142857e-05, "loss": 1.481, "step": 3822500 }, { "epoch": 19.43, "learning_rate": 2.2692857142857145e-05, "loss": 1.4882, "step": 3823000 }, { "epoch": 19.43, "learning_rate": 2.2689285714285715e-05, "loss": 1.4738, "step": 3823500 }, { "epoch": 19.44, "learning_rate": 2.2685714285714286e-05, "loss": 1.4909, "step": 3824000 }, { "epoch": 19.44, "learning_rate": 2.2682142857142857e-05, "loss": 1.4859, "step": 3824500 }, { "epoch": 19.44, "learning_rate": 2.2678571428571428e-05, "loss": 1.473, "step": 3825000 }, { "epoch": 19.44, "learning_rate": 2.2675000000000002e-05, "loss": 1.4798, "step": 3825500 }, { "epoch": 19.45, "learning_rate": 2.267142857142857e-05, "loss": 1.4893, "step": 3826000 }, { "epoch": 19.45, "learning_rate": 2.2667857142857144e-05, "loss": 1.4773, "step": 3826500 }, { "epoch": 19.45, "learning_rate": 2.2664285714285715e-05, "loss": 1.4712, "step": 3827000 }, { "epoch": 19.46, "learning_rate": 2.266071428571429e-05, "loss": 1.4806, "step": 3827500 }, { "epoch": 19.46, "learning_rate": 2.2657142857142857e-05, "loss": 1.4868, "step": 3828000 }, { "epoch": 19.46, "learning_rate": 2.265357142857143e-05, "loss": 1.4773, "step": 3828500 }, { "epoch": 19.46, "learning_rate": 2.265e-05, "loss": 1.4821, "step": 3829000 }, { "epoch": 19.47, "learning_rate": 2.2646428571428573e-05, "loss": 1.4786, "step": 3829500 }, { "epoch": 19.47, "learning_rate": 2.2642857142857143e-05, "loss": 1.4842, "step": 3830000 }, { "epoch": 19.47, "learning_rate": 2.2639285714285714e-05, "loss": 1.4831, "step": 3830500 }, { "epoch": 19.47, "learning_rate": 2.263571428571429e-05, "loss": 1.4792, "step": 3831000 }, { "epoch": 19.48, "learning_rate": 2.2632142857142856e-05, "loss": 1.4768, "step": 3831500 }, { "epoch": 19.48, "learning_rate": 2.262857142857143e-05, "loss": 1.4782, "step": 3832000 }, { "epoch": 19.48, "learning_rate": 2.2625e-05, "loss": 1.4791, "step": 3832500 }, { "epoch": 19.48, "learning_rate": 2.2621428571428572e-05, "loss": 1.4861, "step": 3833000 }, { "epoch": 19.49, "learning_rate": 2.2617857142857143e-05, "loss": 1.4703, "step": 3833500 }, { "epoch": 19.49, "learning_rate": 2.2614285714285717e-05, "loss": 1.4748, "step": 3834000 }, { "epoch": 19.49, "learning_rate": 2.2610714285714288e-05, "loss": 1.4861, "step": 3834500 }, { "epoch": 19.49, "learning_rate": 2.260714285714286e-05, "loss": 1.4807, "step": 3835000 }, { "epoch": 19.5, "learning_rate": 2.260357142857143e-05, "loss": 1.4858, "step": 3835500 }, { "epoch": 19.5, "learning_rate": 2.26e-05, "loss": 1.4766, "step": 3836000 }, { "epoch": 19.5, "learning_rate": 2.259642857142857e-05, "loss": 1.4864, "step": 3836500 }, { "epoch": 19.51, "learning_rate": 2.2592857142857142e-05, "loss": 1.478, "step": 3837000 }, { "epoch": 19.51, "learning_rate": 2.2589285714285716e-05, "loss": 1.4851, "step": 3837500 }, { "epoch": 19.51, "learning_rate": 2.2585714285714287e-05, "loss": 1.4914, "step": 3838000 }, { "epoch": 19.51, "learning_rate": 2.2582142857142858e-05, "loss": 1.48, "step": 3838500 }, { "epoch": 19.52, "learning_rate": 2.257857142857143e-05, "loss": 1.4798, "step": 3839000 }, { "epoch": 19.52, "learning_rate": 2.2575000000000003e-05, "loss": 1.4828, "step": 3839500 }, { "epoch": 19.52, "learning_rate": 2.257142857142857e-05, "loss": 1.4874, "step": 3840000 }, { "epoch": 19.52, "learning_rate": 2.2567857142857145e-05, "loss": 1.4847, "step": 3840500 }, { "epoch": 19.53, "learning_rate": 2.2564285714285716e-05, "loss": 1.4808, "step": 3841000 }, { "epoch": 19.53, "learning_rate": 2.2560714285714287e-05, "loss": 1.4851, "step": 3841500 }, { "epoch": 19.53, "learning_rate": 2.2557142857142858e-05, "loss": 1.487, "step": 3842000 }, { "epoch": 19.53, "learning_rate": 2.255357142857143e-05, "loss": 1.4876, "step": 3842500 }, { "epoch": 19.54, "learning_rate": 2.2550000000000003e-05, "loss": 1.4734, "step": 3843000 }, { "epoch": 19.54, "learning_rate": 2.254642857142857e-05, "loss": 1.4698, "step": 3843500 }, { "epoch": 19.54, "learning_rate": 2.2542857142857144e-05, "loss": 1.4845, "step": 3844000 }, { "epoch": 19.55, "learning_rate": 2.2539285714285715e-05, "loss": 1.4847, "step": 3844500 }, { "epoch": 19.55, "learning_rate": 2.253571428571429e-05, "loss": 1.4836, "step": 3845000 }, { "epoch": 19.55, "learning_rate": 2.2532142857142857e-05, "loss": 1.483, "step": 3845500 }, { "epoch": 19.55, "learning_rate": 2.2528571428571428e-05, "loss": 1.4756, "step": 3846000 }, { "epoch": 19.56, "learning_rate": 2.2525000000000002e-05, "loss": 1.4856, "step": 3846500 }, { "epoch": 19.56, "learning_rate": 2.2521428571428573e-05, "loss": 1.478, "step": 3847000 }, { "epoch": 19.56, "learning_rate": 2.2517857142857144e-05, "loss": 1.4878, "step": 3847500 }, { "epoch": 19.56, "learning_rate": 2.2514285714285715e-05, "loss": 1.4848, "step": 3848000 }, { "epoch": 19.57, "learning_rate": 2.251071428571429e-05, "loss": 1.4733, "step": 3848500 }, { "epoch": 19.57, "learning_rate": 2.2507142857142856e-05, "loss": 1.4864, "step": 3849000 }, { "epoch": 19.57, "learning_rate": 2.250357142857143e-05, "loss": 1.4901, "step": 3849500 }, { "epoch": 19.57, "learning_rate": 2.25e-05, "loss": 1.476, "step": 3850000 }, { "epoch": 19.58, "learning_rate": 2.2496428571428572e-05, "loss": 1.4682, "step": 3850500 }, { "epoch": 19.58, "learning_rate": 2.2492857142857143e-05, "loss": 1.4926, "step": 3851000 }, { "epoch": 19.58, "learning_rate": 2.2489285714285714e-05, "loss": 1.4887, "step": 3851500 }, { "epoch": 19.59, "learning_rate": 2.2485714285714288e-05, "loss": 1.48, "step": 3852000 }, { "epoch": 19.59, "learning_rate": 2.2482142857142856e-05, "loss": 1.4833, "step": 3852500 }, { "epoch": 19.59, "learning_rate": 2.247857142857143e-05, "loss": 1.4771, "step": 3853000 }, { "epoch": 19.59, "learning_rate": 2.2475e-05, "loss": 1.4804, "step": 3853500 }, { "epoch": 19.6, "learning_rate": 2.2471428571428575e-05, "loss": 1.4833, "step": 3854000 }, { "epoch": 19.6, "learning_rate": 2.2467857142857143e-05, "loss": 1.4788, "step": 3854500 }, { "epoch": 19.6, "learning_rate": 2.2464285714285717e-05, "loss": 1.4815, "step": 3855000 }, { "epoch": 19.6, "learning_rate": 2.2460714285714288e-05, "loss": 1.482, "step": 3855500 }, { "epoch": 19.61, "learning_rate": 2.245714285714286e-05, "loss": 1.4842, "step": 3856000 }, { "epoch": 19.61, "learning_rate": 2.245357142857143e-05, "loss": 1.4865, "step": 3856500 }, { "epoch": 19.61, "learning_rate": 2.245e-05, "loss": 1.48, "step": 3857000 }, { "epoch": 19.61, "learning_rate": 2.2446428571428574e-05, "loss": 1.4815, "step": 3857500 }, { "epoch": 19.62, "learning_rate": 2.2442857142857142e-05, "loss": 1.4853, "step": 3858000 }, { "epoch": 19.62, "learning_rate": 2.2439285714285716e-05, "loss": 1.4845, "step": 3858500 }, { "epoch": 19.62, "learning_rate": 2.2435714285714287e-05, "loss": 1.4853, "step": 3859000 }, { "epoch": 19.63, "learning_rate": 2.2432142857142858e-05, "loss": 1.4784, "step": 3859500 }, { "epoch": 19.63, "learning_rate": 2.242857142857143e-05, "loss": 1.4783, "step": 3860000 }, { "epoch": 19.63, "learning_rate": 2.2425000000000003e-05, "loss": 1.4888, "step": 3860500 }, { "epoch": 19.63, "learning_rate": 2.2421428571428574e-05, "loss": 1.472, "step": 3861000 }, { "epoch": 19.64, "learning_rate": 2.2417857142857145e-05, "loss": 1.4773, "step": 3861500 }, { "epoch": 19.64, "learning_rate": 2.2414285714285716e-05, "loss": 1.4813, "step": 3862000 }, { "epoch": 19.64, "learning_rate": 2.2410714285714286e-05, "loss": 1.477, "step": 3862500 }, { "epoch": 19.64, "learning_rate": 2.2407142857142857e-05, "loss": 1.4797, "step": 3863000 }, { "epoch": 19.65, "learning_rate": 2.2403571428571428e-05, "loss": 1.4795, "step": 3863500 }, { "epoch": 19.65, "learning_rate": 2.2400000000000002e-05, "loss": 1.4873, "step": 3864000 }, { "epoch": 19.65, "learning_rate": 2.2396428571428573e-05, "loss": 1.4753, "step": 3864500 }, { "epoch": 19.65, "learning_rate": 2.2392857142857144e-05, "loss": 1.478, "step": 3865000 }, { "epoch": 19.66, "learning_rate": 2.2389285714285715e-05, "loss": 1.4858, "step": 3865500 }, { "epoch": 19.66, "learning_rate": 2.238571428571429e-05, "loss": 1.4835, "step": 3866000 }, { "epoch": 19.66, "learning_rate": 2.2382142857142857e-05, "loss": 1.4747, "step": 3866500 }, { "epoch": 19.67, "learning_rate": 2.237857142857143e-05, "loss": 1.4832, "step": 3867000 }, { "epoch": 19.67, "learning_rate": 2.2375000000000002e-05, "loss": 1.4692, "step": 3867500 }, { "epoch": 19.67, "learning_rate": 2.2371428571428573e-05, "loss": 1.4774, "step": 3868000 }, { "epoch": 19.67, "learning_rate": 2.2367857142857143e-05, "loss": 1.4823, "step": 3868500 }, { "epoch": 19.68, "learning_rate": 2.2364285714285714e-05, "loss": 1.4674, "step": 3869000 }, { "epoch": 19.68, "learning_rate": 2.236071428571429e-05, "loss": 1.4719, "step": 3869500 }, { "epoch": 19.68, "learning_rate": 2.2357142857142856e-05, "loss": 1.4834, "step": 3870000 }, { "epoch": 19.68, "learning_rate": 2.235357142857143e-05, "loss": 1.4787, "step": 3870500 }, { "epoch": 19.69, "learning_rate": 2.235e-05, "loss": 1.4773, "step": 3871000 }, { "epoch": 19.69, "learning_rate": 2.2346428571428575e-05, "loss": 1.483, "step": 3871500 }, { "epoch": 19.69, "learning_rate": 2.2342857142857143e-05, "loss": 1.4741, "step": 3872000 }, { "epoch": 19.69, "learning_rate": 2.2339285714285714e-05, "loss": 1.4715, "step": 3872500 }, { "epoch": 19.7, "learning_rate": 2.2335714285714288e-05, "loss": 1.4777, "step": 3873000 }, { "epoch": 19.7, "learning_rate": 2.2332142857142855e-05, "loss": 1.4665, "step": 3873500 }, { "epoch": 19.7, "learning_rate": 2.232857142857143e-05, "loss": 1.4892, "step": 3874000 }, { "epoch": 19.71, "learning_rate": 2.2325e-05, "loss": 1.4756, "step": 3874500 }, { "epoch": 19.71, "learning_rate": 2.2321428571428575e-05, "loss": 1.4887, "step": 3875000 }, { "epoch": 19.71, "learning_rate": 2.2317857142857142e-05, "loss": 1.4828, "step": 3875500 }, { "epoch": 19.71, "learning_rate": 2.2314285714285717e-05, "loss": 1.4885, "step": 3876000 }, { "epoch": 19.72, "learning_rate": 2.2310714285714287e-05, "loss": 1.4704, "step": 3876500 }, { "epoch": 19.72, "learning_rate": 2.2307142857142858e-05, "loss": 1.4769, "step": 3877000 }, { "epoch": 19.72, "learning_rate": 2.230357142857143e-05, "loss": 1.4793, "step": 3877500 }, { "epoch": 19.72, "learning_rate": 2.23e-05, "loss": 1.4792, "step": 3878000 }, { "epoch": 19.73, "learning_rate": 2.2296428571428574e-05, "loss": 1.4885, "step": 3878500 }, { "epoch": 19.73, "learning_rate": 2.229285714285714e-05, "loss": 1.4792, "step": 3879000 }, { "epoch": 19.73, "learning_rate": 2.2289285714285716e-05, "loss": 1.4832, "step": 3879500 }, { "epoch": 19.73, "learning_rate": 2.2285714285714287e-05, "loss": 1.4861, "step": 3880000 }, { "epoch": 19.74, "learning_rate": 2.2282142857142858e-05, "loss": 1.4784, "step": 3880500 }, { "epoch": 19.74, "learning_rate": 2.227857142857143e-05, "loss": 1.4834, "step": 3881000 }, { "epoch": 19.74, "learning_rate": 2.2275000000000003e-05, "loss": 1.4841, "step": 3881500 }, { "epoch": 19.75, "learning_rate": 2.2271428571428574e-05, "loss": 1.4822, "step": 3882000 }, { "epoch": 19.75, "learning_rate": 2.2267857142857144e-05, "loss": 1.4818, "step": 3882500 }, { "epoch": 19.75, "learning_rate": 2.2264285714285715e-05, "loss": 1.4815, "step": 3883000 }, { "epoch": 19.75, "learning_rate": 2.2260714285714286e-05, "loss": 1.4828, "step": 3883500 }, { "epoch": 19.76, "learning_rate": 2.2257142857142857e-05, "loss": 1.4963, "step": 3884000 }, { "epoch": 19.76, "learning_rate": 2.2253571428571428e-05, "loss": 1.4819, "step": 3884500 }, { "epoch": 19.76, "learning_rate": 2.2250000000000002e-05, "loss": 1.5271, "step": 3885000 }, { "epoch": 19.76, "learning_rate": 2.2246428571428573e-05, "loss": 1.4828, "step": 3885500 }, { "epoch": 19.77, "learning_rate": 2.2242857142857144e-05, "loss": 1.4847, "step": 3886000 }, { "epoch": 19.77, "learning_rate": 2.2239285714285715e-05, "loss": 1.4875, "step": 3886500 }, { "epoch": 19.77, "learning_rate": 2.223571428571429e-05, "loss": 1.4902, "step": 3887000 }, { "epoch": 19.77, "learning_rate": 2.2232142857142856e-05, "loss": 1.4801, "step": 3887500 }, { "epoch": 19.78, "learning_rate": 2.222857142857143e-05, "loss": 1.4773, "step": 3888000 }, { "epoch": 19.78, "learning_rate": 2.2225e-05, "loss": 1.4867, "step": 3888500 }, { "epoch": 19.78, "learning_rate": 2.2221428571428572e-05, "loss": 1.4842, "step": 3889000 }, { "epoch": 19.79, "learning_rate": 2.2217857142857143e-05, "loss": 1.4833, "step": 3889500 }, { "epoch": 19.79, "learning_rate": 2.2214285714285714e-05, "loss": 1.4882, "step": 3890000 }, { "epoch": 19.79, "learning_rate": 2.221071428571429e-05, "loss": 1.4854, "step": 3890500 }, { "epoch": 19.79, "learning_rate": 2.2207142857142856e-05, "loss": 1.4821, "step": 3891000 }, { "epoch": 19.8, "learning_rate": 2.220357142857143e-05, "loss": 1.4782, "step": 3891500 }, { "epoch": 19.8, "learning_rate": 2.22e-05, "loss": 1.4696, "step": 3892000 }, { "epoch": 19.8, "learning_rate": 2.2196428571428575e-05, "loss": 1.4671, "step": 3892500 }, { "epoch": 19.8, "learning_rate": 2.2192857142857143e-05, "loss": 1.4796, "step": 3893000 }, { "epoch": 19.81, "learning_rate": 2.2189285714285717e-05, "loss": 1.4834, "step": 3893500 }, { "epoch": 19.81, "learning_rate": 2.2185714285714288e-05, "loss": 1.4841, "step": 3894000 }, { "epoch": 19.81, "learning_rate": 2.218214285714286e-05, "loss": 1.4856, "step": 3894500 }, { "epoch": 19.81, "learning_rate": 2.217857142857143e-05, "loss": 1.4852, "step": 3895000 }, { "epoch": 19.82, "learning_rate": 2.2175e-05, "loss": 1.486, "step": 3895500 }, { "epoch": 19.82, "learning_rate": 2.2171428571428575e-05, "loss": 1.4777, "step": 3896000 }, { "epoch": 19.82, "learning_rate": 2.2167857142857142e-05, "loss": 1.4838, "step": 3896500 }, { "epoch": 19.83, "learning_rate": 2.2164285714285716e-05, "loss": 1.4869, "step": 3897000 }, { "epoch": 19.83, "learning_rate": 2.2160714285714287e-05, "loss": 1.4731, "step": 3897500 }, { "epoch": 19.83, "learning_rate": 2.2157142857142858e-05, "loss": 1.4808, "step": 3898000 }, { "epoch": 19.83, "learning_rate": 2.215357142857143e-05, "loss": 1.4744, "step": 3898500 }, { "epoch": 19.84, "learning_rate": 2.215e-05, "loss": 1.4709, "step": 3899000 }, { "epoch": 19.84, "learning_rate": 2.2146428571428574e-05, "loss": 1.4796, "step": 3899500 }, { "epoch": 19.84, "learning_rate": 2.214285714285714e-05, "loss": 1.4831, "step": 3900000 }, { "epoch": 19.84, "learning_rate": 2.2139285714285716e-05, "loss": 1.471, "step": 3900500 }, { "epoch": 19.85, "learning_rate": 2.2135714285714287e-05, "loss": 1.4914, "step": 3901000 }, { "epoch": 19.85, "learning_rate": 2.2132142857142857e-05, "loss": 1.4876, "step": 3901500 }, { "epoch": 19.85, "learning_rate": 2.2128571428571428e-05, "loss": 1.4769, "step": 3902000 }, { "epoch": 19.85, "learning_rate": 2.2125000000000002e-05, "loss": 1.4836, "step": 3902500 }, { "epoch": 19.86, "learning_rate": 2.2121428571428573e-05, "loss": 1.4774, "step": 3903000 }, { "epoch": 19.86, "learning_rate": 2.2117857142857144e-05, "loss": 1.4796, "step": 3903500 }, { "epoch": 19.86, "learning_rate": 2.2114285714285715e-05, "loss": 1.4873, "step": 3904000 }, { "epoch": 19.87, "learning_rate": 2.2110714285714286e-05, "loss": 1.4898, "step": 3904500 }, { "epoch": 19.87, "learning_rate": 2.2107142857142857e-05, "loss": 1.4853, "step": 3905000 }, { "epoch": 19.87, "learning_rate": 2.2103571428571428e-05, "loss": 1.4918, "step": 3905500 }, { "epoch": 19.87, "learning_rate": 2.2100000000000002e-05, "loss": 1.482, "step": 3906000 }, { "epoch": 19.88, "learning_rate": 2.2096428571428573e-05, "loss": 1.4947, "step": 3906500 }, { "epoch": 19.88, "learning_rate": 2.2092857142857144e-05, "loss": 1.4876, "step": 3907000 }, { "epoch": 19.88, "learning_rate": 2.2089285714285714e-05, "loss": 1.4843, "step": 3907500 }, { "epoch": 19.88, "learning_rate": 2.208571428571429e-05, "loss": 1.4811, "step": 3908000 }, { "epoch": 19.89, "learning_rate": 2.208214285714286e-05, "loss": 1.4816, "step": 3908500 }, { "epoch": 19.89, "learning_rate": 2.207857142857143e-05, "loss": 1.488, "step": 3909000 }, { "epoch": 19.89, "learning_rate": 2.2075e-05, "loss": 1.489, "step": 3909500 }, { "epoch": 19.89, "learning_rate": 2.2071428571428572e-05, "loss": 1.4737, "step": 3910000 }, { "epoch": 19.9, "learning_rate": 2.2067857142857143e-05, "loss": 1.4806, "step": 3910500 }, { "epoch": 19.9, "learning_rate": 2.2064285714285714e-05, "loss": 1.477, "step": 3911000 }, { "epoch": 19.9, "learning_rate": 2.2060714285714288e-05, "loss": 1.4772, "step": 3911500 }, { "epoch": 19.9, "learning_rate": 2.205714285714286e-05, "loss": 1.4901, "step": 3912000 }, { "epoch": 19.91, "learning_rate": 2.205357142857143e-05, "loss": 1.4929, "step": 3912500 }, { "epoch": 19.91, "learning_rate": 2.205e-05, "loss": 1.4723, "step": 3913000 }, { "epoch": 19.91, "learning_rate": 2.2046428571428575e-05, "loss": 1.4869, "step": 3913500 }, { "epoch": 19.92, "learning_rate": 2.2042857142857142e-05, "loss": 1.4781, "step": 3914000 }, { "epoch": 19.92, "learning_rate": 2.2039285714285717e-05, "loss": 1.4794, "step": 3914500 }, { "epoch": 19.92, "learning_rate": 2.2035714285714287e-05, "loss": 1.4772, "step": 3915000 }, { "epoch": 19.92, "learning_rate": 2.203214285714286e-05, "loss": 1.4757, "step": 3915500 }, { "epoch": 19.93, "learning_rate": 2.202857142857143e-05, "loss": 1.4776, "step": 3916000 }, { "epoch": 19.93, "learning_rate": 2.2025e-05, "loss": 1.483, "step": 3916500 }, { "epoch": 19.93, "learning_rate": 2.2021428571428574e-05, "loss": 1.4845, "step": 3917000 }, { "epoch": 19.93, "learning_rate": 2.2017857142857142e-05, "loss": 1.4796, "step": 3917500 }, { "epoch": 19.94, "learning_rate": 2.2014285714285716e-05, "loss": 1.4815, "step": 3918000 }, { "epoch": 19.94, "learning_rate": 2.2010714285714287e-05, "loss": 1.4851, "step": 3918500 }, { "epoch": 19.94, "learning_rate": 2.200714285714286e-05, "loss": 1.486, "step": 3919000 }, { "epoch": 19.94, "learning_rate": 2.200357142857143e-05, "loss": 1.4823, "step": 3919500 }, { "epoch": 19.95, "learning_rate": 2.2000000000000003e-05, "loss": 1.4712, "step": 3920000 }, { "epoch": 19.95, "learning_rate": 2.1996428571428574e-05, "loss": 1.4644, "step": 3920500 }, { "epoch": 19.95, "learning_rate": 2.199285714285714e-05, "loss": 1.4863, "step": 3921000 }, { "epoch": 19.96, "learning_rate": 2.1989285714285715e-05, "loss": 1.478, "step": 3921500 }, { "epoch": 19.96, "learning_rate": 2.1985714285714286e-05, "loss": 1.4828, "step": 3922000 }, { "epoch": 19.96, "learning_rate": 2.198214285714286e-05, "loss": 1.4807, "step": 3922500 }, { "epoch": 19.96, "learning_rate": 2.1978571428571428e-05, "loss": 1.4896, "step": 3923000 }, { "epoch": 19.97, "learning_rate": 2.1975000000000002e-05, "loss": 1.474, "step": 3923500 }, { "epoch": 19.97, "learning_rate": 2.1971428571428573e-05, "loss": 1.482, "step": 3924000 }, { "epoch": 19.97, "learning_rate": 2.1967857142857144e-05, "loss": 1.474, "step": 3924500 }, { "epoch": 19.97, "learning_rate": 2.1964285714285715e-05, "loss": 1.4868, "step": 3925000 }, { "epoch": 19.98, "learning_rate": 2.1960714285714286e-05, "loss": 1.4782, "step": 3925500 }, { "epoch": 19.98, "learning_rate": 2.195714285714286e-05, "loss": 1.4824, "step": 3926000 }, { "epoch": 19.98, "learning_rate": 2.1953571428571427e-05, "loss": 1.4821, "step": 3926500 }, { "epoch": 19.98, "learning_rate": 2.195e-05, "loss": 1.4741, "step": 3927000 }, { "epoch": 19.99, "learning_rate": 2.1946428571428572e-05, "loss": 1.4704, "step": 3927500 }, { "epoch": 19.99, "learning_rate": 2.1942857142857143e-05, "loss": 1.4828, "step": 3928000 }, { "epoch": 19.99, "learning_rate": 2.1939285714285714e-05, "loss": 1.4775, "step": 3928500 }, { "epoch": 20.0, "learning_rate": 2.193571428571429e-05, "loss": 1.4694, "step": 3929000 }, { "epoch": 20.0, "learning_rate": 2.193214285714286e-05, "loss": 1.4889, "step": 3929500 }, { "epoch": 20.0, "learning_rate": 2.192857142857143e-05, "loss": 1.4801, "step": 3930000 }, { "epoch": 20.0, "learning_rate": 2.1925e-05, "loss": 1.4721, "step": 3930500 }, { "epoch": 20.01, "learning_rate": 2.1921428571428572e-05, "loss": 1.4772, "step": 3931000 }, { "epoch": 20.01, "learning_rate": 2.1917857142857143e-05, "loss": 1.4803, "step": 3931500 }, { "epoch": 20.01, "learning_rate": 2.1914285714285714e-05, "loss": 1.4851, "step": 3932000 }, { "epoch": 20.01, "learning_rate": 2.1910714285714288e-05, "loss": 1.473, "step": 3932500 }, { "epoch": 20.02, "learning_rate": 2.190714285714286e-05, "loss": 1.4817, "step": 3933000 }, { "epoch": 20.02, "learning_rate": 2.190357142857143e-05, "loss": 1.4734, "step": 3933500 }, { "epoch": 20.02, "learning_rate": 2.19e-05, "loss": 1.4775, "step": 3934000 }, { "epoch": 20.02, "learning_rate": 2.1896428571428575e-05, "loss": 1.4707, "step": 3934500 }, { "epoch": 20.03, "learning_rate": 2.1892857142857142e-05, "loss": 1.4714, "step": 3935000 }, { "epoch": 20.03, "learning_rate": 2.1889285714285716e-05, "loss": 1.4773, "step": 3935500 }, { "epoch": 20.03, "learning_rate": 2.1885714285714287e-05, "loss": 1.4745, "step": 3936000 }, { "epoch": 20.04, "learning_rate": 2.1882142857142858e-05, "loss": 1.4726, "step": 3936500 }, { "epoch": 20.04, "learning_rate": 2.187857142857143e-05, "loss": 1.4876, "step": 3937000 }, { "epoch": 20.04, "learning_rate": 2.1875e-05, "loss": 1.4833, "step": 3937500 }, { "epoch": 20.04, "learning_rate": 2.1871428571428574e-05, "loss": 1.4791, "step": 3938000 }, { "epoch": 20.05, "learning_rate": 2.186785714285714e-05, "loss": 1.4782, "step": 3938500 }, { "epoch": 20.05, "learning_rate": 2.1864285714285716e-05, "loss": 1.471, "step": 3939000 }, { "epoch": 20.05, "learning_rate": 2.1860714285714287e-05, "loss": 1.4716, "step": 3939500 }, { "epoch": 20.05, "learning_rate": 2.185714285714286e-05, "loss": 1.4675, "step": 3940000 }, { "epoch": 20.06, "learning_rate": 2.185357142857143e-05, "loss": 1.4782, "step": 3940500 }, { "epoch": 20.06, "learning_rate": 2.1850000000000003e-05, "loss": 1.4744, "step": 3941000 }, { "epoch": 20.06, "learning_rate": 2.1846428571428573e-05, "loss": 1.4723, "step": 3941500 }, { "epoch": 20.06, "learning_rate": 2.1842857142857144e-05, "loss": 1.4714, "step": 3942000 }, { "epoch": 20.07, "learning_rate": 2.1839285714285715e-05, "loss": 1.4837, "step": 3942500 }, { "epoch": 20.07, "learning_rate": 2.1835714285714286e-05, "loss": 1.4781, "step": 3943000 }, { "epoch": 20.07, "learning_rate": 2.183214285714286e-05, "loss": 1.4757, "step": 3943500 }, { "epoch": 20.08, "learning_rate": 2.1828571428571428e-05, "loss": 1.4672, "step": 3944000 }, { "epoch": 20.08, "learning_rate": 2.1825000000000002e-05, "loss": 1.4654, "step": 3944500 }, { "epoch": 20.08, "learning_rate": 2.1821428571428573e-05, "loss": 1.4763, "step": 3945000 }, { "epoch": 20.08, "learning_rate": 2.1817857142857144e-05, "loss": 1.4745, "step": 3945500 }, { "epoch": 20.09, "learning_rate": 2.1814285714285715e-05, "loss": 1.472, "step": 3946000 }, { "epoch": 20.09, "learning_rate": 2.181071428571429e-05, "loss": 1.4644, "step": 3946500 }, { "epoch": 20.09, "learning_rate": 2.180714285714286e-05, "loss": 1.482, "step": 3947000 }, { "epoch": 20.09, "learning_rate": 2.1803571428571427e-05, "loss": 1.4792, "step": 3947500 }, { "epoch": 20.1, "learning_rate": 2.18e-05, "loss": 1.4761, "step": 3948000 }, { "epoch": 20.1, "learning_rate": 2.1796428571428572e-05, "loss": 1.4714, "step": 3948500 }, { "epoch": 20.1, "learning_rate": 2.1792857142857143e-05, "loss": 1.466, "step": 3949000 }, { "epoch": 20.1, "learning_rate": 2.1789285714285714e-05, "loss": 1.4677, "step": 3949500 }, { "epoch": 20.11, "learning_rate": 2.1785714285714288e-05, "loss": 1.4752, "step": 3950000 }, { "epoch": 20.11, "learning_rate": 2.178214285714286e-05, "loss": 1.472, "step": 3950500 }, { "epoch": 20.11, "learning_rate": 2.177857142857143e-05, "loss": 1.4743, "step": 3951000 }, { "epoch": 20.12, "learning_rate": 2.1775e-05, "loss": 1.4735, "step": 3951500 }, { "epoch": 20.12, "learning_rate": 2.177142857142857e-05, "loss": 1.4727, "step": 3952000 }, { "epoch": 20.12, "learning_rate": 2.1767857142857142e-05, "loss": 1.4778, "step": 3952500 }, { "epoch": 20.12, "learning_rate": 2.1764285714285713e-05, "loss": 1.4747, "step": 3953000 }, { "epoch": 20.13, "learning_rate": 2.1760714285714288e-05, "loss": 1.4743, "step": 3953500 }, { "epoch": 20.13, "learning_rate": 2.175714285714286e-05, "loss": 1.468, "step": 3954000 }, { "epoch": 20.13, "learning_rate": 2.175357142857143e-05, "loss": 1.4776, "step": 3954500 }, { "epoch": 20.13, "learning_rate": 2.175e-05, "loss": 1.4664, "step": 3955000 }, { "epoch": 20.14, "learning_rate": 2.1746428571428574e-05, "loss": 1.4855, "step": 3955500 }, { "epoch": 20.14, "learning_rate": 2.1742857142857142e-05, "loss": 1.4751, "step": 3956000 }, { "epoch": 20.14, "learning_rate": 2.1739285714285716e-05, "loss": 1.469, "step": 3956500 }, { "epoch": 20.14, "learning_rate": 2.1735714285714287e-05, "loss": 1.4662, "step": 3957000 }, { "epoch": 20.15, "learning_rate": 2.1732142857142858e-05, "loss": 1.4812, "step": 3957500 }, { "epoch": 20.15, "learning_rate": 2.172857142857143e-05, "loss": 1.4681, "step": 3958000 }, { "epoch": 20.15, "learning_rate": 2.1725e-05, "loss": 1.468, "step": 3958500 }, { "epoch": 20.16, "learning_rate": 2.1721428571428574e-05, "loss": 1.4673, "step": 3959000 }, { "epoch": 20.16, "learning_rate": 2.171785714285714e-05, "loss": 1.4694, "step": 3959500 }, { "epoch": 20.16, "learning_rate": 2.1714285714285715e-05, "loss": 1.4765, "step": 3960000 }, { "epoch": 20.16, "learning_rate": 2.1710714285714286e-05, "loss": 1.4762, "step": 3960500 }, { "epoch": 20.17, "learning_rate": 2.170714285714286e-05, "loss": 1.4665, "step": 3961000 }, { "epoch": 20.17, "learning_rate": 2.1703571428571428e-05, "loss": 1.473, "step": 3961500 }, { "epoch": 20.17, "learning_rate": 2.1700000000000002e-05, "loss": 1.4703, "step": 3962000 }, { "epoch": 20.17, "learning_rate": 2.1696428571428573e-05, "loss": 1.4817, "step": 3962500 }, { "epoch": 20.18, "learning_rate": 2.1692857142857144e-05, "loss": 1.472, "step": 3963000 }, { "epoch": 20.18, "learning_rate": 2.1689285714285715e-05, "loss": 1.4703, "step": 3963500 }, { "epoch": 20.18, "learning_rate": 2.1685714285714286e-05, "loss": 1.4729, "step": 3964000 }, { "epoch": 20.18, "learning_rate": 2.168214285714286e-05, "loss": 1.4698, "step": 3964500 }, { "epoch": 20.19, "learning_rate": 2.1678571428571427e-05, "loss": 1.4691, "step": 3965000 }, { "epoch": 20.19, "learning_rate": 2.1675e-05, "loss": 1.4687, "step": 3965500 }, { "epoch": 20.19, "learning_rate": 2.1671428571428573e-05, "loss": 1.4649, "step": 3966000 }, { "epoch": 20.2, "learning_rate": 2.1667857142857143e-05, "loss": 1.4687, "step": 3966500 }, { "epoch": 20.2, "learning_rate": 2.1664285714285714e-05, "loss": 1.4809, "step": 3967000 }, { "epoch": 20.2, "learning_rate": 2.166071428571429e-05, "loss": 1.4764, "step": 3967500 }, { "epoch": 20.2, "learning_rate": 2.165714285714286e-05, "loss": 1.4744, "step": 3968000 }, { "epoch": 20.21, "learning_rate": 2.165357142857143e-05, "loss": 1.4785, "step": 3968500 }, { "epoch": 20.21, "learning_rate": 2.165e-05, "loss": 1.4715, "step": 3969000 }, { "epoch": 20.21, "learning_rate": 2.1646428571428572e-05, "loss": 1.4725, "step": 3969500 }, { "epoch": 20.21, "learning_rate": 2.1642857142857146e-05, "loss": 1.4645, "step": 3970000 }, { "epoch": 20.22, "learning_rate": 2.1639285714285714e-05, "loss": 1.4748, "step": 3970500 }, { "epoch": 20.22, "learning_rate": 2.1635714285714288e-05, "loss": 1.4762, "step": 3971000 }, { "epoch": 20.22, "learning_rate": 2.163214285714286e-05, "loss": 1.4781, "step": 3971500 }, { "epoch": 20.22, "learning_rate": 2.162857142857143e-05, "loss": 1.4637, "step": 3972000 }, { "epoch": 20.23, "learning_rate": 2.1625e-05, "loss": 1.4619, "step": 3972500 }, { "epoch": 20.23, "learning_rate": 2.1621428571428575e-05, "loss": 1.4691, "step": 3973000 }, { "epoch": 20.23, "learning_rate": 2.1617857142857146e-05, "loss": 1.461, "step": 3973500 }, { "epoch": 20.24, "learning_rate": 2.1614285714285713e-05, "loss": 1.4699, "step": 3974000 }, { "epoch": 20.24, "learning_rate": 2.1610714285714287e-05, "loss": 1.4776, "step": 3974500 }, { "epoch": 20.24, "learning_rate": 2.1607142857142858e-05, "loss": 1.4795, "step": 3975000 }, { "epoch": 20.24, "learning_rate": 2.160357142857143e-05, "loss": 1.4713, "step": 3975500 }, { "epoch": 20.25, "learning_rate": 2.16e-05, "loss": 1.4724, "step": 3976000 }, { "epoch": 20.25, "learning_rate": 2.1596428571428574e-05, "loss": 1.483, "step": 3976500 }, { "epoch": 20.25, "learning_rate": 2.1592857142857145e-05, "loss": 1.4683, "step": 3977000 }, { "epoch": 20.25, "learning_rate": 2.1589285714285716e-05, "loss": 1.4738, "step": 3977500 }, { "epoch": 20.26, "learning_rate": 2.1585714285714287e-05, "loss": 1.4571, "step": 3978000 }, { "epoch": 20.26, "learning_rate": 2.1582142857142858e-05, "loss": 1.4774, "step": 3978500 }, { "epoch": 20.26, "learning_rate": 2.157857142857143e-05, "loss": 1.4719, "step": 3979000 }, { "epoch": 20.26, "learning_rate": 2.1575e-05, "loss": 1.4617, "step": 3979500 }, { "epoch": 20.27, "learning_rate": 2.1571428571428574e-05, "loss": 1.4655, "step": 3980000 }, { "epoch": 20.27, "learning_rate": 2.1567857142857144e-05, "loss": 1.4799, "step": 3980500 }, { "epoch": 20.27, "learning_rate": 2.1564285714285715e-05, "loss": 1.4748, "step": 3981000 }, { "epoch": 20.27, "learning_rate": 2.1560714285714286e-05, "loss": 1.4797, "step": 3981500 }, { "epoch": 20.28, "learning_rate": 2.155714285714286e-05, "loss": 1.4773, "step": 3982000 }, { "epoch": 20.28, "learning_rate": 2.1553571428571428e-05, "loss": 1.4663, "step": 3982500 }, { "epoch": 20.28, "learning_rate": 2.1550000000000002e-05, "loss": 1.479, "step": 3983000 }, { "epoch": 20.29, "learning_rate": 2.1546428571428573e-05, "loss": 1.4691, "step": 3983500 }, { "epoch": 20.29, "learning_rate": 2.1542857142857144e-05, "loss": 1.4708, "step": 3984000 }, { "epoch": 20.29, "learning_rate": 2.1539285714285715e-05, "loss": 1.4777, "step": 3984500 }, { "epoch": 20.29, "learning_rate": 2.1535714285714285e-05, "loss": 1.4779, "step": 3985000 }, { "epoch": 20.3, "learning_rate": 2.153214285714286e-05, "loss": 1.4784, "step": 3985500 }, { "epoch": 20.3, "learning_rate": 2.1528571428571427e-05, "loss": 1.4823, "step": 3986000 }, { "epoch": 20.3, "learning_rate": 2.1525e-05, "loss": 1.4819, "step": 3986500 }, { "epoch": 20.3, "learning_rate": 2.1521428571428572e-05, "loss": 1.4714, "step": 3987000 }, { "epoch": 20.31, "learning_rate": 2.1517857142857147e-05, "loss": 1.4684, "step": 3987500 }, { "epoch": 20.31, "learning_rate": 2.1514285714285714e-05, "loss": 1.4744, "step": 3988000 }, { "epoch": 20.31, "learning_rate": 2.1510714285714288e-05, "loss": 1.463, "step": 3988500 }, { "epoch": 20.31, "learning_rate": 2.150714285714286e-05, "loss": 1.4729, "step": 3989000 }, { "epoch": 20.32, "learning_rate": 2.150357142857143e-05, "loss": 1.4722, "step": 3989500 }, { "epoch": 20.32, "learning_rate": 2.15e-05, "loss": 1.4832, "step": 3990000 }, { "epoch": 20.32, "learning_rate": 2.149642857142857e-05, "loss": 1.4733, "step": 3990500 }, { "epoch": 20.33, "learning_rate": 2.1492857142857146e-05, "loss": 1.462, "step": 3991000 }, { "epoch": 20.33, "learning_rate": 2.1489285714285713e-05, "loss": 1.468, "step": 3991500 }, { "epoch": 20.33, "learning_rate": 2.1485714285714288e-05, "loss": 1.4741, "step": 3992000 }, { "epoch": 20.33, "learning_rate": 2.148214285714286e-05, "loss": 1.4705, "step": 3992500 }, { "epoch": 20.34, "learning_rate": 2.147857142857143e-05, "loss": 1.4692, "step": 3993000 }, { "epoch": 20.34, "learning_rate": 2.1475e-05, "loss": 1.4604, "step": 3993500 }, { "epoch": 20.34, "learning_rate": 2.1471428571428574e-05, "loss": 1.4725, "step": 3994000 }, { "epoch": 20.34, "learning_rate": 2.1467857142857145e-05, "loss": 1.4691, "step": 3994500 }, { "epoch": 20.35, "learning_rate": 2.1464285714285716e-05, "loss": 1.4719, "step": 3995000 }, { "epoch": 20.35, "learning_rate": 2.1460714285714287e-05, "loss": 1.4777, "step": 3995500 }, { "epoch": 20.35, "learning_rate": 2.1457142857142858e-05, "loss": 1.4694, "step": 3996000 }, { "epoch": 20.35, "learning_rate": 2.145357142857143e-05, "loss": 1.4796, "step": 3996500 }, { "epoch": 20.36, "learning_rate": 2.145e-05, "loss": 1.4614, "step": 3997000 }, { "epoch": 20.36, "learning_rate": 2.1446428571428574e-05, "loss": 1.4712, "step": 3997500 }, { "epoch": 20.36, "learning_rate": 2.1442857142857145e-05, "loss": 1.4746, "step": 3998000 }, { "epoch": 20.37, "learning_rate": 2.1439285714285716e-05, "loss": 1.4748, "step": 3998500 }, { "epoch": 20.37, "learning_rate": 2.1435714285714286e-05, "loss": 1.4649, "step": 3999000 }, { "epoch": 20.37, "learning_rate": 2.1432142857142857e-05, "loss": 1.4708, "step": 3999500 }, { "epoch": 20.37, "learning_rate": 2.1428571428571428e-05, "loss": 1.4705, "step": 4000000 }, { "epoch": 20.38, "learning_rate": 2.1425e-05, "loss": 1.4713, "step": 4000500 }, { "epoch": 20.38, "learning_rate": 2.1421428571428573e-05, "loss": 1.4698, "step": 4001000 }, { "epoch": 20.38, "learning_rate": 2.1417857142857144e-05, "loss": 1.4724, "step": 4001500 }, { "epoch": 20.38, "learning_rate": 2.1414285714285715e-05, "loss": 1.4741, "step": 4002000 }, { "epoch": 20.39, "learning_rate": 2.1410714285714286e-05, "loss": 1.4838, "step": 4002500 }, { "epoch": 20.39, "learning_rate": 2.140714285714286e-05, "loss": 1.4693, "step": 4003000 }, { "epoch": 20.39, "learning_rate": 2.1403571428571428e-05, "loss": 1.4716, "step": 4003500 }, { "epoch": 20.39, "learning_rate": 2.1400000000000002e-05, "loss": 1.4695, "step": 4004000 }, { "epoch": 20.4, "learning_rate": 2.1396428571428573e-05, "loss": 1.4831, "step": 4004500 }, { "epoch": 20.4, "learning_rate": 2.1392857142857143e-05, "loss": 1.4708, "step": 4005000 }, { "epoch": 20.4, "learning_rate": 2.1389285714285714e-05, "loss": 1.4687, "step": 4005500 }, { "epoch": 20.41, "learning_rate": 2.1385714285714285e-05, "loss": 1.4701, "step": 4006000 }, { "epoch": 20.41, "learning_rate": 2.138214285714286e-05, "loss": 1.4702, "step": 4006500 }, { "epoch": 20.41, "learning_rate": 2.1378571428571427e-05, "loss": 1.4747, "step": 4007000 }, { "epoch": 20.41, "learning_rate": 2.1375e-05, "loss": 1.4744, "step": 4007500 }, { "epoch": 20.42, "learning_rate": 2.1371428571428572e-05, "loss": 1.4707, "step": 4008000 }, { "epoch": 20.42, "learning_rate": 2.1367857142857146e-05, "loss": 1.485, "step": 4008500 }, { "epoch": 20.42, "learning_rate": 2.1364285714285714e-05, "loss": 1.464, "step": 4009000 }, { "epoch": 20.42, "learning_rate": 2.1360714285714288e-05, "loss": 1.4686, "step": 4009500 }, { "epoch": 20.43, "learning_rate": 2.135714285714286e-05, "loss": 1.4583, "step": 4010000 }, { "epoch": 20.43, "learning_rate": 2.135357142857143e-05, "loss": 1.4646, "step": 4010500 }, { "epoch": 20.43, "learning_rate": 2.135e-05, "loss": 1.4708, "step": 4011000 }, { "epoch": 20.43, "learning_rate": 2.134642857142857e-05, "loss": 1.4747, "step": 4011500 }, { "epoch": 20.44, "learning_rate": 2.1342857142857146e-05, "loss": 1.4586, "step": 4012000 }, { "epoch": 20.44, "learning_rate": 2.1339285714285713e-05, "loss": 1.4654, "step": 4012500 }, { "epoch": 20.44, "learning_rate": 2.1335714285714287e-05, "loss": 1.4768, "step": 4013000 }, { "epoch": 20.45, "learning_rate": 2.1332142857142858e-05, "loss": 1.4673, "step": 4013500 }, { "epoch": 20.45, "learning_rate": 2.132857142857143e-05, "loss": 1.4743, "step": 4014000 }, { "epoch": 20.45, "learning_rate": 2.1325e-05, "loss": 1.465, "step": 4014500 }, { "epoch": 20.45, "learning_rate": 2.1321428571428574e-05, "loss": 1.4759, "step": 4015000 }, { "epoch": 20.46, "learning_rate": 2.1317857142857145e-05, "loss": 1.4739, "step": 4015500 }, { "epoch": 20.46, "learning_rate": 2.1314285714285716e-05, "loss": 1.4696, "step": 4016000 }, { "epoch": 20.46, "learning_rate": 2.1310714285714287e-05, "loss": 1.4714, "step": 4016500 }, { "epoch": 20.46, "learning_rate": 2.1307142857142858e-05, "loss": 1.4662, "step": 4017000 }, { "epoch": 20.47, "learning_rate": 2.130357142857143e-05, "loss": 1.4679, "step": 4017500 }, { "epoch": 20.47, "learning_rate": 2.13e-05, "loss": 1.4686, "step": 4018000 }, { "epoch": 20.47, "learning_rate": 2.1296428571428574e-05, "loss": 1.4796, "step": 4018500 }, { "epoch": 20.47, "learning_rate": 2.1292857142857144e-05, "loss": 1.4797, "step": 4019000 }, { "epoch": 20.48, "learning_rate": 2.1289285714285715e-05, "loss": 1.4759, "step": 4019500 }, { "epoch": 20.48, "learning_rate": 2.1285714285714286e-05, "loss": 1.4695, "step": 4020000 }, { "epoch": 20.48, "learning_rate": 2.128214285714286e-05, "loss": 1.4731, "step": 4020500 }, { "epoch": 20.49, "learning_rate": 2.1278571428571428e-05, "loss": 1.4611, "step": 4021000 }, { "epoch": 20.49, "learning_rate": 2.1275000000000002e-05, "loss": 1.4662, "step": 4021500 }, { "epoch": 20.49, "learning_rate": 2.1271428571428573e-05, "loss": 1.4661, "step": 4022000 }, { "epoch": 20.49, "learning_rate": 2.1267857142857144e-05, "loss": 1.4685, "step": 4022500 }, { "epoch": 20.5, "learning_rate": 2.1264285714285715e-05, "loss": 1.4683, "step": 4023000 }, { "epoch": 20.5, "learning_rate": 2.1260714285714286e-05, "loss": 1.4719, "step": 4023500 }, { "epoch": 20.5, "learning_rate": 2.125714285714286e-05, "loss": 1.4745, "step": 4024000 }, { "epoch": 20.5, "learning_rate": 2.1253571428571427e-05, "loss": 1.4711, "step": 4024500 }, { "epoch": 20.51, "learning_rate": 2.125e-05, "loss": 1.4779, "step": 4025000 }, { "epoch": 20.51, "learning_rate": 2.1246428571428572e-05, "loss": 1.4618, "step": 4025500 }, { "epoch": 20.51, "learning_rate": 2.1242857142857143e-05, "loss": 1.4748, "step": 4026000 }, { "epoch": 20.51, "learning_rate": 2.1239285714285714e-05, "loss": 1.4741, "step": 4026500 }, { "epoch": 20.52, "learning_rate": 2.1235714285714285e-05, "loss": 1.461, "step": 4027000 }, { "epoch": 20.52, "learning_rate": 2.123214285714286e-05, "loss": 1.4702, "step": 4027500 }, { "epoch": 20.52, "learning_rate": 2.1228571428571427e-05, "loss": 1.4795, "step": 4028000 }, { "epoch": 20.53, "learning_rate": 2.1225e-05, "loss": 1.4757, "step": 4028500 }, { "epoch": 20.53, "learning_rate": 2.1221428571428572e-05, "loss": 1.4727, "step": 4029000 }, { "epoch": 20.53, "learning_rate": 2.1217857142857146e-05, "loss": 1.4755, "step": 4029500 }, { "epoch": 20.53, "learning_rate": 2.1214285714285713e-05, "loss": 1.4814, "step": 4030000 }, { "epoch": 20.54, "learning_rate": 2.1210714285714288e-05, "loss": 1.4724, "step": 4030500 }, { "epoch": 20.54, "learning_rate": 2.120714285714286e-05, "loss": 1.4755, "step": 4031000 }, { "epoch": 20.54, "learning_rate": 2.120357142857143e-05, "loss": 1.4681, "step": 4031500 }, { "epoch": 20.54, "learning_rate": 2.12e-05, "loss": 1.4654, "step": 4032000 }, { "epoch": 20.55, "learning_rate": 2.119642857142857e-05, "loss": 1.4623, "step": 4032500 }, { "epoch": 20.55, "learning_rate": 2.1192857142857145e-05, "loss": 1.468, "step": 4033000 }, { "epoch": 20.55, "learning_rate": 2.1189285714285713e-05, "loss": 1.4738, "step": 4033500 }, { "epoch": 20.55, "learning_rate": 2.1185714285714287e-05, "loss": 1.4698, "step": 4034000 }, { "epoch": 20.56, "learning_rate": 2.1182142857142858e-05, "loss": 1.4683, "step": 4034500 }, { "epoch": 20.56, "learning_rate": 2.1178571428571432e-05, "loss": 1.4716, "step": 4035000 }, { "epoch": 20.56, "learning_rate": 2.1175e-05, "loss": 1.4696, "step": 4035500 }, { "epoch": 20.57, "learning_rate": 2.1171428571428574e-05, "loss": 1.4704, "step": 4036000 }, { "epoch": 20.57, "learning_rate": 2.1167857142857145e-05, "loss": 1.4665, "step": 4036500 }, { "epoch": 20.57, "learning_rate": 2.1164285714285716e-05, "loss": 1.4664, "step": 4037000 }, { "epoch": 20.57, "learning_rate": 2.1160714285714287e-05, "loss": 1.4752, "step": 4037500 }, { "epoch": 20.58, "learning_rate": 2.1157142857142857e-05, "loss": 1.4698, "step": 4038000 }, { "epoch": 20.58, "learning_rate": 2.115357142857143e-05, "loss": 1.4726, "step": 4038500 }, { "epoch": 20.58, "learning_rate": 2.115e-05, "loss": 1.4676, "step": 4039000 }, { "epoch": 20.58, "learning_rate": 2.1146428571428573e-05, "loss": 1.4706, "step": 4039500 }, { "epoch": 20.59, "learning_rate": 2.1142857142857144e-05, "loss": 1.4704, "step": 4040000 }, { "epoch": 20.59, "learning_rate": 2.1139285714285715e-05, "loss": 1.4662, "step": 4040500 }, { "epoch": 20.59, "learning_rate": 2.1135714285714286e-05, "loss": 1.4825, "step": 4041000 }, { "epoch": 20.59, "learning_rate": 2.113214285714286e-05, "loss": 1.4707, "step": 4041500 }, { "epoch": 20.6, "learning_rate": 2.112857142857143e-05, "loss": 1.4653, "step": 4042000 }, { "epoch": 20.6, "learning_rate": 2.1125000000000002e-05, "loss": 1.4742, "step": 4042500 }, { "epoch": 20.6, "learning_rate": 2.1121428571428573e-05, "loss": 1.4638, "step": 4043000 }, { "epoch": 20.61, "learning_rate": 2.1117857142857144e-05, "loss": 1.4689, "step": 4043500 }, { "epoch": 20.61, "learning_rate": 2.1114285714285714e-05, "loss": 1.4707, "step": 4044000 }, { "epoch": 20.61, "learning_rate": 2.1110714285714285e-05, "loss": 1.4611, "step": 4044500 }, { "epoch": 20.61, "learning_rate": 2.110714285714286e-05, "loss": 1.4631, "step": 4045000 }, { "epoch": 20.62, "learning_rate": 2.110357142857143e-05, "loss": 1.4662, "step": 4045500 }, { "epoch": 20.62, "learning_rate": 2.11e-05, "loss": 1.4697, "step": 4046000 }, { "epoch": 20.62, "learning_rate": 2.1096428571428572e-05, "loss": 1.4721, "step": 4046500 }, { "epoch": 20.62, "learning_rate": 2.1092857142857146e-05, "loss": 1.4765, "step": 4047000 }, { "epoch": 20.63, "learning_rate": 2.1089285714285714e-05, "loss": 1.4725, "step": 4047500 }, { "epoch": 20.63, "learning_rate": 2.1085714285714288e-05, "loss": 1.4719, "step": 4048000 }, { "epoch": 20.63, "learning_rate": 2.108214285714286e-05, "loss": 1.4724, "step": 4048500 }, { "epoch": 20.63, "learning_rate": 2.107857142857143e-05, "loss": 1.4705, "step": 4049000 }, { "epoch": 20.64, "learning_rate": 2.1075e-05, "loss": 1.4698, "step": 4049500 }, { "epoch": 20.64, "learning_rate": 2.107142857142857e-05, "loss": 1.4666, "step": 4050000 }, { "epoch": 20.64, "learning_rate": 2.1067857142857146e-05, "loss": 1.4754, "step": 4050500 }, { "epoch": 20.65, "learning_rate": 2.1064285714285713e-05, "loss": 1.4662, "step": 4051000 }, { "epoch": 20.65, "learning_rate": 2.1060714285714287e-05, "loss": 1.4655, "step": 4051500 }, { "epoch": 20.65, "learning_rate": 2.105714285714286e-05, "loss": 1.4659, "step": 4052000 }, { "epoch": 20.65, "learning_rate": 2.105357142857143e-05, "loss": 1.4668, "step": 4052500 }, { "epoch": 20.66, "learning_rate": 2.105e-05, "loss": 1.4711, "step": 4053000 }, { "epoch": 20.66, "learning_rate": 2.104642857142857e-05, "loss": 1.4745, "step": 4053500 }, { "epoch": 20.66, "learning_rate": 2.1042857142857145e-05, "loss": 1.4652, "step": 4054000 }, { "epoch": 20.66, "learning_rate": 2.1039285714285713e-05, "loss": 1.4572, "step": 4054500 }, { "epoch": 20.67, "learning_rate": 2.1035714285714287e-05, "loss": 1.4704, "step": 4055000 }, { "epoch": 20.67, "learning_rate": 2.1032142857142858e-05, "loss": 1.4608, "step": 4055500 }, { "epoch": 20.67, "learning_rate": 2.1028571428571432e-05, "loss": 1.4603, "step": 4056000 }, { "epoch": 20.67, "learning_rate": 2.1025e-05, "loss": 1.4717, "step": 4056500 }, { "epoch": 20.68, "learning_rate": 2.1021428571428574e-05, "loss": 1.4636, "step": 4057000 }, { "epoch": 20.68, "learning_rate": 2.1017857142857145e-05, "loss": 1.473, "step": 4057500 }, { "epoch": 20.68, "learning_rate": 2.1014285714285715e-05, "loss": 1.4697, "step": 4058000 }, { "epoch": 20.68, "learning_rate": 2.1010714285714286e-05, "loss": 1.4756, "step": 4058500 }, { "epoch": 20.69, "learning_rate": 2.1007142857142857e-05, "loss": 1.4739, "step": 4059000 }, { "epoch": 20.69, "learning_rate": 2.100357142857143e-05, "loss": 1.4736, "step": 4059500 }, { "epoch": 20.69, "learning_rate": 2.1e-05, "loss": 1.4701, "step": 4060000 }, { "epoch": 20.7, "learning_rate": 2.0996428571428573e-05, "loss": 1.4671, "step": 4060500 }, { "epoch": 20.7, "learning_rate": 2.0992857142857144e-05, "loss": 1.479, "step": 4061000 }, { "epoch": 20.7, "learning_rate": 2.0989285714285715e-05, "loss": 1.4671, "step": 4061500 }, { "epoch": 20.7, "learning_rate": 2.0985714285714286e-05, "loss": 1.4726, "step": 4062000 }, { "epoch": 20.71, "learning_rate": 2.098214285714286e-05, "loss": 1.478, "step": 4062500 }, { "epoch": 20.71, "learning_rate": 2.097857142857143e-05, "loss": 1.4652, "step": 4063000 }, { "epoch": 20.71, "learning_rate": 2.0975e-05, "loss": 1.4719, "step": 4063500 }, { "epoch": 20.71, "learning_rate": 2.0971428571428572e-05, "loss": 1.4685, "step": 4064000 }, { "epoch": 20.72, "learning_rate": 2.0967857142857143e-05, "loss": 1.4747, "step": 4064500 }, { "epoch": 20.72, "learning_rate": 2.0964285714285714e-05, "loss": 1.4702, "step": 4065000 }, { "epoch": 20.72, "learning_rate": 2.0960714285714285e-05, "loss": 1.4873, "step": 4065500 }, { "epoch": 20.72, "learning_rate": 2.095714285714286e-05, "loss": 1.4836, "step": 4066000 }, { "epoch": 20.73, "learning_rate": 2.095357142857143e-05, "loss": 1.4745, "step": 4066500 }, { "epoch": 20.73, "learning_rate": 2.095e-05, "loss": 1.4719, "step": 4067000 }, { "epoch": 20.73, "learning_rate": 2.0946428571428572e-05, "loss": 1.4733, "step": 4067500 }, { "epoch": 20.74, "learning_rate": 2.0942857142857146e-05, "loss": 1.4702, "step": 4068000 }, { "epoch": 20.74, "learning_rate": 2.0939285714285714e-05, "loss": 1.4635, "step": 4068500 }, { "epoch": 20.74, "learning_rate": 2.0935714285714288e-05, "loss": 1.4647, "step": 4069000 }, { "epoch": 20.74, "learning_rate": 2.093214285714286e-05, "loss": 1.4639, "step": 4069500 }, { "epoch": 20.75, "learning_rate": 2.092857142857143e-05, "loss": 1.4578, "step": 4070000 }, { "epoch": 20.75, "learning_rate": 2.0925e-05, "loss": 1.471, "step": 4070500 }, { "epoch": 20.75, "learning_rate": 2.092142857142857e-05, "loss": 1.471, "step": 4071000 }, { "epoch": 20.75, "learning_rate": 2.0917857142857146e-05, "loss": 1.4683, "step": 4071500 }, { "epoch": 20.76, "learning_rate": 2.0914285714285713e-05, "loss": 1.4643, "step": 4072000 }, { "epoch": 20.76, "learning_rate": 2.0910714285714287e-05, "loss": 1.4656, "step": 4072500 }, { "epoch": 20.76, "learning_rate": 2.0907142857142858e-05, "loss": 1.4677, "step": 4073000 }, { "epoch": 20.76, "learning_rate": 2.0903571428571432e-05, "loss": 1.4631, "step": 4073500 }, { "epoch": 20.77, "learning_rate": 2.09e-05, "loss": 1.4705, "step": 4074000 }, { "epoch": 20.77, "learning_rate": 2.0896428571428574e-05, "loss": 1.4705, "step": 4074500 }, { "epoch": 20.77, "learning_rate": 2.0892857142857145e-05, "loss": 1.468, "step": 4075000 }, { "epoch": 20.78, "learning_rate": 2.0889285714285712e-05, "loss": 1.4624, "step": 4075500 }, { "epoch": 20.78, "learning_rate": 2.0885714285714287e-05, "loss": 1.461, "step": 4076000 }, { "epoch": 20.78, "learning_rate": 2.0882142857142857e-05, "loss": 1.4669, "step": 4076500 }, { "epoch": 20.78, "learning_rate": 2.0878571428571432e-05, "loss": 1.4728, "step": 4077000 }, { "epoch": 20.79, "learning_rate": 2.0875e-05, "loss": 1.4695, "step": 4077500 }, { "epoch": 20.79, "learning_rate": 2.0871428571428573e-05, "loss": 1.4649, "step": 4078000 }, { "epoch": 20.79, "learning_rate": 2.0867857142857144e-05, "loss": 1.4759, "step": 4078500 }, { "epoch": 20.79, "learning_rate": 2.0864285714285715e-05, "loss": 1.4642, "step": 4079000 }, { "epoch": 20.8, "learning_rate": 2.0860714285714286e-05, "loss": 1.4777, "step": 4079500 }, { "epoch": 20.8, "learning_rate": 2.0857142857142857e-05, "loss": 1.4787, "step": 4080000 }, { "epoch": 20.8, "learning_rate": 2.085357142857143e-05, "loss": 1.4699, "step": 4080500 }, { "epoch": 20.8, "learning_rate": 2.085e-05, "loss": 1.4645, "step": 4081000 }, { "epoch": 20.81, "learning_rate": 2.0846428571428573e-05, "loss": 1.4731, "step": 4081500 }, { "epoch": 20.81, "learning_rate": 2.0842857142857144e-05, "loss": 1.4717, "step": 4082000 }, { "epoch": 20.81, "learning_rate": 2.0839285714285715e-05, "loss": 1.4693, "step": 4082500 }, { "epoch": 20.82, "learning_rate": 2.0835714285714285e-05, "loss": 1.4774, "step": 4083000 }, { "epoch": 20.82, "learning_rate": 2.083214285714286e-05, "loss": 1.4726, "step": 4083500 }, { "epoch": 20.82, "learning_rate": 2.082857142857143e-05, "loss": 1.4705, "step": 4084000 }, { "epoch": 20.82, "learning_rate": 2.0825e-05, "loss": 1.4743, "step": 4084500 }, { "epoch": 20.83, "learning_rate": 2.0821428571428572e-05, "loss": 1.4734, "step": 4085000 }, { "epoch": 20.83, "learning_rate": 2.0817857142857143e-05, "loss": 1.4778, "step": 4085500 }, { "epoch": 20.83, "learning_rate": 2.0814285714285714e-05, "loss": 1.4576, "step": 4086000 }, { "epoch": 20.83, "learning_rate": 2.0810714285714285e-05, "loss": 1.4635, "step": 4086500 }, { "epoch": 20.84, "learning_rate": 2.080714285714286e-05, "loss": 1.4643, "step": 4087000 }, { "epoch": 20.84, "learning_rate": 2.080357142857143e-05, "loss": 1.4761, "step": 4087500 }, { "epoch": 20.84, "learning_rate": 2.08e-05, "loss": 1.4615, "step": 4088000 }, { "epoch": 20.84, "learning_rate": 2.079642857142857e-05, "loss": 1.47, "step": 4088500 }, { "epoch": 20.85, "learning_rate": 2.0792857142857146e-05, "loss": 1.4694, "step": 4089000 }, { "epoch": 20.85, "learning_rate": 2.0789285714285713e-05, "loss": 1.4639, "step": 4089500 }, { "epoch": 20.85, "learning_rate": 2.0785714285714288e-05, "loss": 1.4746, "step": 4090000 }, { "epoch": 20.86, "learning_rate": 2.078214285714286e-05, "loss": 1.4732, "step": 4090500 }, { "epoch": 20.86, "learning_rate": 2.077857142857143e-05, "loss": 1.4626, "step": 4091000 }, { "epoch": 20.86, "learning_rate": 2.0775e-05, "loss": 1.4615, "step": 4091500 }, { "epoch": 20.86, "learning_rate": 2.077142857142857e-05, "loss": 1.4641, "step": 4092000 }, { "epoch": 20.87, "learning_rate": 2.0767857142857145e-05, "loss": 1.4716, "step": 4092500 }, { "epoch": 20.87, "learning_rate": 2.0764285714285713e-05, "loss": 1.4695, "step": 4093000 }, { "epoch": 20.87, "learning_rate": 2.0760714285714287e-05, "loss": 1.4651, "step": 4093500 }, { "epoch": 20.87, "learning_rate": 2.0757142857142858e-05, "loss": 1.4666, "step": 4094000 }, { "epoch": 20.88, "learning_rate": 2.0753571428571432e-05, "loss": 1.4796, "step": 4094500 }, { "epoch": 20.88, "learning_rate": 2.075e-05, "loss": 1.4575, "step": 4095000 }, { "epoch": 20.88, "learning_rate": 2.0746428571428574e-05, "loss": 1.4654, "step": 4095500 }, { "epoch": 20.88, "learning_rate": 2.0742857142857145e-05, "loss": 1.4684, "step": 4096000 }, { "epoch": 20.89, "learning_rate": 2.0739285714285715e-05, "loss": 1.4731, "step": 4096500 }, { "epoch": 20.89, "learning_rate": 2.0735714285714286e-05, "loss": 1.4705, "step": 4097000 }, { "epoch": 20.89, "learning_rate": 2.0732142857142857e-05, "loss": 1.4729, "step": 4097500 }, { "epoch": 20.9, "learning_rate": 2.072857142857143e-05, "loss": 1.4705, "step": 4098000 }, { "epoch": 20.9, "learning_rate": 2.0725e-05, "loss": 1.4679, "step": 4098500 }, { "epoch": 20.9, "learning_rate": 2.0721428571428573e-05, "loss": 1.4706, "step": 4099000 }, { "epoch": 20.9, "learning_rate": 2.0717857142857144e-05, "loss": 1.4695, "step": 4099500 }, { "epoch": 20.91, "learning_rate": 2.0714285714285718e-05, "loss": 1.4668, "step": 4100000 }, { "epoch": 20.91, "learning_rate": 2.0710714285714286e-05, "loss": 1.455, "step": 4100500 }, { "epoch": 20.91, "learning_rate": 2.0707142857142857e-05, "loss": 1.4695, "step": 4101000 }, { "epoch": 20.91, "learning_rate": 2.070357142857143e-05, "loss": 1.4664, "step": 4101500 }, { "epoch": 20.92, "learning_rate": 2.07e-05, "loss": 1.4732, "step": 4102000 }, { "epoch": 20.92, "learning_rate": 2.0696428571428573e-05, "loss": 1.4651, "step": 4102500 }, { "epoch": 20.92, "learning_rate": 2.0692857142857143e-05, "loss": 1.4549, "step": 4103000 }, { "epoch": 20.92, "learning_rate": 2.0689285714285718e-05, "loss": 1.4689, "step": 4103500 }, { "epoch": 20.93, "learning_rate": 2.0685714285714285e-05, "loss": 1.4595, "step": 4104000 }, { "epoch": 20.93, "learning_rate": 2.068214285714286e-05, "loss": 1.4693, "step": 4104500 }, { "epoch": 20.93, "learning_rate": 2.067857142857143e-05, "loss": 1.4655, "step": 4105000 }, { "epoch": 20.94, "learning_rate": 2.0675e-05, "loss": 1.4689, "step": 4105500 }, { "epoch": 20.94, "learning_rate": 2.0671428571428572e-05, "loss": 1.4568, "step": 4106000 }, { "epoch": 20.94, "learning_rate": 2.0667857142857143e-05, "loss": 1.4671, "step": 4106500 }, { "epoch": 20.94, "learning_rate": 2.0664285714285717e-05, "loss": 1.4685, "step": 4107000 }, { "epoch": 20.95, "learning_rate": 2.0660714285714285e-05, "loss": 1.4625, "step": 4107500 }, { "epoch": 20.95, "learning_rate": 2.065714285714286e-05, "loss": 1.4696, "step": 4108000 }, { "epoch": 20.95, "learning_rate": 2.065357142857143e-05, "loss": 1.457, "step": 4108500 }, { "epoch": 20.95, "learning_rate": 2.065e-05, "loss": 1.4752, "step": 4109000 }, { "epoch": 20.96, "learning_rate": 2.064642857142857e-05, "loss": 1.4638, "step": 4109500 }, { "epoch": 20.96, "learning_rate": 2.0642857142857146e-05, "loss": 1.4604, "step": 4110000 }, { "epoch": 20.96, "learning_rate": 2.0639285714285716e-05, "loss": 1.4649, "step": 4110500 }, { "epoch": 20.96, "learning_rate": 2.0635714285714287e-05, "loss": 1.4769, "step": 4111000 }, { "epoch": 20.97, "learning_rate": 2.0632142857142858e-05, "loss": 1.4655, "step": 4111500 }, { "epoch": 20.97, "learning_rate": 2.062857142857143e-05, "loss": 1.4649, "step": 4112000 }, { "epoch": 20.97, "learning_rate": 2.0625e-05, "loss": 1.4733, "step": 4112500 }, { "epoch": 20.98, "learning_rate": 2.062142857142857e-05, "loss": 1.4653, "step": 4113000 }, { "epoch": 20.98, "learning_rate": 2.0617857142857145e-05, "loss": 1.4601, "step": 4113500 }, { "epoch": 20.98, "learning_rate": 2.0614285714285716e-05, "loss": 1.474, "step": 4114000 }, { "epoch": 20.98, "learning_rate": 2.0610714285714287e-05, "loss": 1.4646, "step": 4114500 }, { "epoch": 20.99, "learning_rate": 2.0607142857142858e-05, "loss": 1.458, "step": 4115000 }, { "epoch": 20.99, "learning_rate": 2.0603571428571432e-05, "loss": 1.465, "step": 4115500 }, { "epoch": 20.99, "learning_rate": 2.06e-05, "loss": 1.4577, "step": 4116000 }, { "epoch": 20.99, "learning_rate": 2.0596428571428574e-05, "loss": 1.4665, "step": 4116500 }, { "epoch": 21.0, "learning_rate": 2.0592857142857144e-05, "loss": 1.4696, "step": 4117000 }, { "epoch": 21.0, "learning_rate": 2.0589285714285715e-05, "loss": 1.4618, "step": 4117500 }, { "epoch": 21.0, "learning_rate": 2.0585714285714286e-05, "loss": 1.4572, "step": 4118000 }, { "epoch": 21.0, "learning_rate": 2.0582142857142857e-05, "loss": 1.4706, "step": 4118500 }, { "epoch": 21.01, "learning_rate": 2.057857142857143e-05, "loss": 1.4579, "step": 4119000 }, { "epoch": 21.01, "learning_rate": 2.0575e-05, "loss": 1.4574, "step": 4119500 }, { "epoch": 21.01, "learning_rate": 2.0571428571428573e-05, "loss": 1.4646, "step": 4120000 }, { "epoch": 21.02, "learning_rate": 2.0567857142857144e-05, "loss": 1.4565, "step": 4120500 }, { "epoch": 21.02, "learning_rate": 2.0564285714285718e-05, "loss": 1.4581, "step": 4121000 }, { "epoch": 21.02, "learning_rate": 2.0560714285714285e-05, "loss": 1.4645, "step": 4121500 }, { "epoch": 21.02, "learning_rate": 2.055714285714286e-05, "loss": 1.4583, "step": 4122000 }, { "epoch": 21.03, "learning_rate": 2.055357142857143e-05, "loss": 1.4621, "step": 4122500 }, { "epoch": 21.03, "learning_rate": 2.055e-05, "loss": 1.4611, "step": 4123000 }, { "epoch": 21.03, "learning_rate": 2.0546428571428572e-05, "loss": 1.4528, "step": 4123500 }, { "epoch": 21.03, "learning_rate": 2.0542857142857143e-05, "loss": 1.4632, "step": 4124000 }, { "epoch": 21.04, "learning_rate": 2.0539285714285717e-05, "loss": 1.4688, "step": 4124500 }, { "epoch": 21.04, "learning_rate": 2.0535714285714285e-05, "loss": 1.4566, "step": 4125000 }, { "epoch": 21.04, "learning_rate": 2.053214285714286e-05, "loss": 1.4568, "step": 4125500 }, { "epoch": 21.04, "learning_rate": 2.052857142857143e-05, "loss": 1.4602, "step": 4126000 }, { "epoch": 21.05, "learning_rate": 2.0525e-05, "loss": 1.4649, "step": 4126500 }, { "epoch": 21.05, "learning_rate": 2.0521428571428572e-05, "loss": 1.4562, "step": 4127000 }, { "epoch": 21.05, "learning_rate": 2.0517857142857143e-05, "loss": 1.4652, "step": 4127500 }, { "epoch": 21.06, "learning_rate": 2.0514285714285717e-05, "loss": 1.4577, "step": 4128000 }, { "epoch": 21.06, "learning_rate": 2.0510714285714284e-05, "loss": 1.4534, "step": 4128500 }, { "epoch": 21.06, "learning_rate": 2.050714285714286e-05, "loss": 1.4571, "step": 4129000 }, { "epoch": 21.06, "learning_rate": 2.050357142857143e-05, "loss": 1.4603, "step": 4129500 }, { "epoch": 21.07, "learning_rate": 2.05e-05, "loss": 1.4698, "step": 4130000 }, { "epoch": 21.07, "learning_rate": 2.049642857142857e-05, "loss": 1.4718, "step": 4130500 }, { "epoch": 21.07, "learning_rate": 2.0492857142857145e-05, "loss": 1.465, "step": 4131000 }, { "epoch": 21.07, "learning_rate": 2.0489285714285716e-05, "loss": 1.4871, "step": 4131500 }, { "epoch": 21.08, "learning_rate": 2.0485714285714287e-05, "loss": 1.4703, "step": 4132000 }, { "epoch": 21.08, "learning_rate": 2.0482142857142858e-05, "loss": 1.4698, "step": 4132500 }, { "epoch": 21.08, "learning_rate": 2.047857142857143e-05, "loss": 1.4697, "step": 4133000 }, { "epoch": 21.08, "learning_rate": 2.0475e-05, "loss": 1.4595, "step": 4133500 }, { "epoch": 21.09, "learning_rate": 2.047142857142857e-05, "loss": 1.4634, "step": 4134000 }, { "epoch": 21.09, "learning_rate": 2.0467857142857145e-05, "loss": 1.4623, "step": 4134500 }, { "epoch": 21.09, "learning_rate": 2.0464285714285716e-05, "loss": 1.4615, "step": 4135000 }, { "epoch": 21.09, "learning_rate": 2.0460714285714286e-05, "loss": 1.4549, "step": 4135500 }, { "epoch": 21.1, "learning_rate": 2.0457142857142857e-05, "loss": 1.4651, "step": 4136000 }, { "epoch": 21.1, "learning_rate": 2.045357142857143e-05, "loss": 1.4638, "step": 4136500 }, { "epoch": 21.1, "learning_rate": 2.045e-05, "loss": 1.4616, "step": 4137000 }, { "epoch": 21.11, "learning_rate": 2.0446428571428573e-05, "loss": 1.4608, "step": 4137500 }, { "epoch": 21.11, "learning_rate": 2.0442857142857144e-05, "loss": 1.4599, "step": 4138000 }, { "epoch": 21.11, "learning_rate": 2.0439285714285715e-05, "loss": 1.4589, "step": 4138500 }, { "epoch": 21.11, "learning_rate": 2.0435714285714286e-05, "loss": 1.4494, "step": 4139000 }, { "epoch": 21.12, "learning_rate": 2.0432142857142857e-05, "loss": 1.4639, "step": 4139500 }, { "epoch": 21.12, "learning_rate": 2.042857142857143e-05, "loss": 1.4524, "step": 4140000 }, { "epoch": 21.12, "learning_rate": 2.0425e-05, "loss": 1.4623, "step": 4140500 }, { "epoch": 21.12, "learning_rate": 2.0421428571428573e-05, "loss": 1.463, "step": 4141000 }, { "epoch": 21.13, "learning_rate": 2.0417857142857144e-05, "loss": 1.4574, "step": 4141500 }, { "epoch": 21.13, "learning_rate": 2.0414285714285718e-05, "loss": 1.4654, "step": 4142000 }, { "epoch": 21.13, "learning_rate": 2.0410714285714285e-05, "loss": 1.4583, "step": 4142500 }, { "epoch": 21.13, "learning_rate": 2.040714285714286e-05, "loss": 1.4634, "step": 4143000 }, { "epoch": 21.14, "learning_rate": 2.040357142857143e-05, "loss": 1.4586, "step": 4143500 }, { "epoch": 21.14, "learning_rate": 2.04e-05, "loss": 1.4608, "step": 4144000 }, { "epoch": 21.14, "learning_rate": 2.0396428571428572e-05, "loss": 1.4661, "step": 4144500 }, { "epoch": 21.15, "learning_rate": 2.0392857142857143e-05, "loss": 1.4689, "step": 4145000 }, { "epoch": 21.15, "learning_rate": 2.0389285714285717e-05, "loss": 1.47, "step": 4145500 }, { "epoch": 21.15, "learning_rate": 2.0385714285714285e-05, "loss": 1.4548, "step": 4146000 }, { "epoch": 21.15, "learning_rate": 2.038214285714286e-05, "loss": 1.4633, "step": 4146500 }, { "epoch": 21.16, "learning_rate": 2.037857142857143e-05, "loss": 1.4575, "step": 4147000 }, { "epoch": 21.16, "learning_rate": 2.0375e-05, "loss": 1.4589, "step": 4147500 }, { "epoch": 21.16, "learning_rate": 2.037142857142857e-05, "loss": 1.4587, "step": 4148000 }, { "epoch": 21.16, "learning_rate": 2.0367857142857146e-05, "loss": 1.4589, "step": 4148500 }, { "epoch": 21.17, "learning_rate": 2.0364285714285717e-05, "loss": 1.4617, "step": 4149000 }, { "epoch": 21.17, "learning_rate": 2.0360714285714287e-05, "loss": 1.4553, "step": 4149500 }, { "epoch": 21.17, "learning_rate": 2.0357142857142858e-05, "loss": 1.4691, "step": 4150000 }, { "epoch": 21.17, "learning_rate": 2.035357142857143e-05, "loss": 1.4539, "step": 4150500 }, { "epoch": 21.18, "learning_rate": 2.035e-05, "loss": 1.4614, "step": 4151000 }, { "epoch": 21.18, "learning_rate": 2.034642857142857e-05, "loss": 1.459, "step": 4151500 }, { "epoch": 21.18, "learning_rate": 2.0342857142857145e-05, "loss": 1.4599, "step": 4152000 }, { "epoch": 21.19, "learning_rate": 2.0339285714285716e-05, "loss": 1.4646, "step": 4152500 }, { "epoch": 21.19, "learning_rate": 2.0335714285714287e-05, "loss": 1.4664, "step": 4153000 }, { "epoch": 21.19, "learning_rate": 2.0332142857142858e-05, "loss": 1.4624, "step": 4153500 }, { "epoch": 21.19, "learning_rate": 2.032857142857143e-05, "loss": 1.447, "step": 4154000 }, { "epoch": 21.2, "learning_rate": 2.0325e-05, "loss": 1.4683, "step": 4154500 }, { "epoch": 21.2, "learning_rate": 2.032142857142857e-05, "loss": 1.462, "step": 4155000 }, { "epoch": 21.2, "learning_rate": 2.0317857142857144e-05, "loss": 1.4642, "step": 4155500 }, { "epoch": 21.2, "learning_rate": 2.0314285714285715e-05, "loss": 1.4527, "step": 4156000 }, { "epoch": 21.21, "learning_rate": 2.0310714285714286e-05, "loss": 1.468, "step": 4156500 }, { "epoch": 21.21, "learning_rate": 2.0307142857142857e-05, "loss": 1.456, "step": 4157000 }, { "epoch": 21.21, "learning_rate": 2.030357142857143e-05, "loss": 1.4558, "step": 4157500 }, { "epoch": 21.21, "learning_rate": 2.0300000000000002e-05, "loss": 1.4679, "step": 4158000 }, { "epoch": 21.22, "learning_rate": 2.0296428571428573e-05, "loss": 1.4581, "step": 4158500 }, { "epoch": 21.22, "learning_rate": 2.0292857142857144e-05, "loss": 1.4629, "step": 4159000 }, { "epoch": 21.22, "learning_rate": 2.0289285714285715e-05, "loss": 1.4588, "step": 4159500 }, { "epoch": 21.23, "learning_rate": 2.0285714285714286e-05, "loss": 1.458, "step": 4160000 }, { "epoch": 21.23, "learning_rate": 2.0282142857142856e-05, "loss": 1.4611, "step": 4160500 }, { "epoch": 21.23, "learning_rate": 2.027857142857143e-05, "loss": 1.4685, "step": 4161000 }, { "epoch": 21.23, "learning_rate": 2.0275e-05, "loss": 1.4643, "step": 4161500 }, { "epoch": 21.24, "learning_rate": 2.0271428571428572e-05, "loss": 1.4574, "step": 4162000 }, { "epoch": 21.24, "learning_rate": 2.0267857142857143e-05, "loss": 1.4597, "step": 4162500 }, { "epoch": 21.24, "learning_rate": 2.0264285714285718e-05, "loss": 1.4654, "step": 4163000 }, { "epoch": 21.24, "learning_rate": 2.0260714285714285e-05, "loss": 1.4621, "step": 4163500 }, { "epoch": 21.25, "learning_rate": 2.025714285714286e-05, "loss": 1.466, "step": 4164000 }, { "epoch": 21.25, "learning_rate": 2.025357142857143e-05, "loss": 1.4468, "step": 4164500 }, { "epoch": 21.25, "learning_rate": 2.025e-05, "loss": 1.4637, "step": 4165000 }, { "epoch": 21.25, "learning_rate": 2.0246428571428572e-05, "loss": 1.4528, "step": 4165500 }, { "epoch": 21.26, "learning_rate": 2.0242857142857143e-05, "loss": 1.4689, "step": 4166000 }, { "epoch": 21.26, "learning_rate": 2.0239285714285717e-05, "loss": 1.4515, "step": 4166500 }, { "epoch": 21.26, "learning_rate": 2.0235714285714284e-05, "loss": 1.4612, "step": 4167000 }, { "epoch": 21.27, "learning_rate": 2.023214285714286e-05, "loss": 1.4586, "step": 4167500 }, { "epoch": 21.27, "learning_rate": 2.022857142857143e-05, "loss": 1.4817, "step": 4168000 }, { "epoch": 21.27, "learning_rate": 2.0225000000000004e-05, "loss": 1.4669, "step": 4168500 }, { "epoch": 21.27, "learning_rate": 2.022142857142857e-05, "loss": 1.4646, "step": 4169000 }, { "epoch": 21.28, "learning_rate": 2.0217857142857145e-05, "loss": 1.4491, "step": 4169500 }, { "epoch": 21.28, "learning_rate": 2.0214285714285716e-05, "loss": 1.4514, "step": 4170000 }, { "epoch": 21.28, "learning_rate": 2.0210714285714287e-05, "loss": 1.4558, "step": 4170500 }, { "epoch": 21.28, "learning_rate": 2.0207142857142858e-05, "loss": 1.4618, "step": 4171000 }, { "epoch": 21.29, "learning_rate": 2.020357142857143e-05, "loss": 1.4747, "step": 4171500 }, { "epoch": 21.29, "learning_rate": 2.0200000000000003e-05, "loss": 1.4735, "step": 4172000 }, { "epoch": 21.29, "learning_rate": 2.019642857142857e-05, "loss": 1.4629, "step": 4172500 }, { "epoch": 21.29, "learning_rate": 2.0192857142857145e-05, "loss": 1.4597, "step": 4173000 }, { "epoch": 21.3, "learning_rate": 2.0189285714285716e-05, "loss": 1.4426, "step": 4173500 }, { "epoch": 21.3, "learning_rate": 2.0185714285714287e-05, "loss": 1.4612, "step": 4174000 }, { "epoch": 21.3, "learning_rate": 2.0182142857142857e-05, "loss": 1.4758, "step": 4174500 }, { "epoch": 21.31, "learning_rate": 2.017857142857143e-05, "loss": 1.4499, "step": 4175000 }, { "epoch": 21.31, "learning_rate": 2.0175000000000003e-05, "loss": 1.4602, "step": 4175500 }, { "epoch": 21.31, "learning_rate": 2.0171428571428573e-05, "loss": 1.4674, "step": 4176000 }, { "epoch": 21.31, "learning_rate": 2.0167857142857144e-05, "loss": 1.4688, "step": 4176500 }, { "epoch": 21.32, "learning_rate": 2.0164285714285715e-05, "loss": 1.4708, "step": 4177000 }, { "epoch": 21.32, "learning_rate": 2.0160714285714286e-05, "loss": 1.4663, "step": 4177500 }, { "epoch": 21.32, "learning_rate": 2.0157142857142857e-05, "loss": 1.4562, "step": 4178000 }, { "epoch": 21.32, "learning_rate": 2.015357142857143e-05, "loss": 1.4619, "step": 4178500 }, { "epoch": 21.33, "learning_rate": 2.0150000000000002e-05, "loss": 1.4567, "step": 4179000 }, { "epoch": 21.33, "learning_rate": 2.0146428571428573e-05, "loss": 1.4621, "step": 4179500 }, { "epoch": 21.33, "learning_rate": 2.0142857142857144e-05, "loss": 1.4533, "step": 4180000 }, { "epoch": 21.33, "learning_rate": 2.0139285714285714e-05, "loss": 1.4506, "step": 4180500 }, { "epoch": 21.34, "learning_rate": 2.0135714285714285e-05, "loss": 1.4675, "step": 4181000 }, { "epoch": 21.34, "learning_rate": 2.0132142857142856e-05, "loss": 1.4503, "step": 4181500 }, { "epoch": 21.34, "learning_rate": 2.012857142857143e-05, "loss": 1.4571, "step": 4182000 }, { "epoch": 21.35, "learning_rate": 2.0125e-05, "loss": 1.4627, "step": 4182500 }, { "epoch": 21.35, "learning_rate": 2.0121428571428572e-05, "loss": 1.4602, "step": 4183000 }, { "epoch": 21.35, "learning_rate": 2.0117857142857143e-05, "loss": 1.4597, "step": 4183500 }, { "epoch": 21.35, "learning_rate": 2.0114285714285717e-05, "loss": 1.4624, "step": 4184000 }, { "epoch": 21.36, "learning_rate": 2.0110714285714285e-05, "loss": 1.4689, "step": 4184500 }, { "epoch": 21.36, "learning_rate": 2.010714285714286e-05, "loss": 1.4601, "step": 4185000 }, { "epoch": 21.36, "learning_rate": 2.010357142857143e-05, "loss": 1.4663, "step": 4185500 }, { "epoch": 21.36, "learning_rate": 2.01e-05, "loss": 1.4526, "step": 4186000 }, { "epoch": 21.37, "learning_rate": 2.009642857142857e-05, "loss": 1.4587, "step": 4186500 }, { "epoch": 21.37, "learning_rate": 2.0092857142857142e-05, "loss": 1.4527, "step": 4187000 }, { "epoch": 21.37, "learning_rate": 2.0089285714285717e-05, "loss": 1.4593, "step": 4187500 }, { "epoch": 21.37, "learning_rate": 2.0085714285714284e-05, "loss": 1.4637, "step": 4188000 }, { "epoch": 21.38, "learning_rate": 2.008214285714286e-05, "loss": 1.4457, "step": 4188500 }, { "epoch": 21.38, "learning_rate": 2.007857142857143e-05, "loss": 1.4526, "step": 4189000 }, { "epoch": 21.38, "learning_rate": 2.0075000000000003e-05, "loss": 1.4461, "step": 4189500 }, { "epoch": 21.39, "learning_rate": 2.007142857142857e-05, "loss": 1.4654, "step": 4190000 }, { "epoch": 21.39, "learning_rate": 2.0067857142857145e-05, "loss": 1.4661, "step": 4190500 }, { "epoch": 21.39, "learning_rate": 2.0064285714285716e-05, "loss": 1.4606, "step": 4191000 }, { "epoch": 21.39, "learning_rate": 2.0060714285714287e-05, "loss": 1.4582, "step": 4191500 }, { "epoch": 21.4, "learning_rate": 2.0057142857142858e-05, "loss": 1.4591, "step": 4192000 }, { "epoch": 21.4, "learning_rate": 2.005357142857143e-05, "loss": 1.4563, "step": 4192500 }, { "epoch": 21.4, "learning_rate": 2.0050000000000003e-05, "loss": 1.4489, "step": 4193000 }, { "epoch": 21.4, "learning_rate": 2.004642857142857e-05, "loss": 1.4567, "step": 4193500 }, { "epoch": 21.41, "learning_rate": 2.0042857142857145e-05, "loss": 1.4584, "step": 4194000 }, { "epoch": 21.41, "learning_rate": 2.0039285714285715e-05, "loss": 1.462, "step": 4194500 }, { "epoch": 21.41, "learning_rate": 2.0035714285714286e-05, "loss": 1.4579, "step": 4195000 }, { "epoch": 21.41, "learning_rate": 2.0032142857142857e-05, "loss": 1.4663, "step": 4195500 }, { "epoch": 21.42, "learning_rate": 2.002857142857143e-05, "loss": 1.459, "step": 4196000 }, { "epoch": 21.42, "learning_rate": 2.0025000000000002e-05, "loss": 1.4633, "step": 4196500 }, { "epoch": 21.42, "learning_rate": 2.0021428571428573e-05, "loss": 1.468, "step": 4197000 }, { "epoch": 21.43, "learning_rate": 2.0017857142857144e-05, "loss": 1.4528, "step": 4197500 }, { "epoch": 21.43, "learning_rate": 2.0014285714285715e-05, "loss": 1.4674, "step": 4198000 }, { "epoch": 21.43, "learning_rate": 2.0010714285714286e-05, "loss": 1.4623, "step": 4198500 }, { "epoch": 21.43, "learning_rate": 2.0007142857142857e-05, "loss": 1.4636, "step": 4199000 }, { "epoch": 21.44, "learning_rate": 2.000357142857143e-05, "loss": 1.4581, "step": 4199500 }, { "epoch": 21.44, "learning_rate": 2e-05, "loss": 1.4569, "step": 4200000 }, { "epoch": 21.44, "learning_rate": 1.9996428571428572e-05, "loss": 1.4509, "step": 4200500 }, { "epoch": 21.44, "learning_rate": 1.9992857142857143e-05, "loss": 1.4568, "step": 4201000 }, { "epoch": 21.45, "learning_rate": 1.9989285714285718e-05, "loss": 1.454, "step": 4201500 }, { "epoch": 21.45, "learning_rate": 1.9985714285714285e-05, "loss": 1.4579, "step": 4202000 }, { "epoch": 21.45, "learning_rate": 1.9982142857142856e-05, "loss": 1.4629, "step": 4202500 }, { "epoch": 21.45, "learning_rate": 1.997857142857143e-05, "loss": 1.4547, "step": 4203000 }, { "epoch": 21.46, "learning_rate": 1.9975e-05, "loss": 1.4492, "step": 4203500 }, { "epoch": 21.46, "learning_rate": 1.9971428571428572e-05, "loss": 1.4586, "step": 4204000 }, { "epoch": 21.46, "learning_rate": 1.9967857142857143e-05, "loss": 1.4527, "step": 4204500 }, { "epoch": 21.47, "learning_rate": 1.9964285714285717e-05, "loss": 1.4571, "step": 4205000 }, { "epoch": 21.47, "learning_rate": 1.9960714285714284e-05, "loss": 1.453, "step": 4205500 }, { "epoch": 21.47, "learning_rate": 1.995714285714286e-05, "loss": 1.4648, "step": 4206000 }, { "epoch": 21.47, "learning_rate": 1.995357142857143e-05, "loss": 1.4729, "step": 4206500 }, { "epoch": 21.48, "learning_rate": 1.995e-05, "loss": 1.456, "step": 4207000 }, { "epoch": 21.48, "learning_rate": 1.994642857142857e-05, "loss": 1.4547, "step": 4207500 }, { "epoch": 21.48, "learning_rate": 1.9942857142857142e-05, "loss": 1.458, "step": 4208000 }, { "epoch": 21.48, "learning_rate": 1.9939285714285716e-05, "loss": 1.4688, "step": 4208500 }, { "epoch": 21.49, "learning_rate": 1.9935714285714284e-05, "loss": 1.4629, "step": 4209000 }, { "epoch": 21.49, "learning_rate": 1.9932142857142858e-05, "loss": 1.4579, "step": 4209500 }, { "epoch": 21.49, "learning_rate": 1.992857142857143e-05, "loss": 1.4522, "step": 4210000 }, { "epoch": 21.49, "learning_rate": 1.9925000000000003e-05, "loss": 1.4627, "step": 4210500 }, { "epoch": 21.5, "learning_rate": 1.992142857142857e-05, "loss": 1.458, "step": 4211000 }, { "epoch": 21.5, "learning_rate": 1.9917857142857145e-05, "loss": 1.4654, "step": 4211500 }, { "epoch": 21.5, "learning_rate": 1.9914285714285716e-05, "loss": 1.46, "step": 4212000 }, { "epoch": 21.5, "learning_rate": 1.9910714285714287e-05, "loss": 1.4667, "step": 4212500 }, { "epoch": 21.51, "learning_rate": 1.9907142857142857e-05, "loss": 1.4519, "step": 4213000 }, { "epoch": 21.51, "learning_rate": 1.990357142857143e-05, "loss": 1.4583, "step": 4213500 }, { "epoch": 21.51, "learning_rate": 1.9900000000000003e-05, "loss": 1.462, "step": 4214000 }, { "epoch": 21.52, "learning_rate": 1.989642857142857e-05, "loss": 1.4611, "step": 4214500 }, { "epoch": 21.52, "learning_rate": 1.9892857142857144e-05, "loss": 1.4559, "step": 4215000 }, { "epoch": 21.52, "learning_rate": 1.9889285714285715e-05, "loss": 1.4582, "step": 4215500 }, { "epoch": 21.52, "learning_rate": 1.9885714285714286e-05, "loss": 1.4629, "step": 4216000 }, { "epoch": 21.53, "learning_rate": 1.9882142857142857e-05, "loss": 1.4521, "step": 4216500 }, { "epoch": 21.53, "learning_rate": 1.987857142857143e-05, "loss": 1.4668, "step": 4217000 }, { "epoch": 21.53, "learning_rate": 1.9875000000000002e-05, "loss": 1.4701, "step": 4217500 }, { "epoch": 21.53, "learning_rate": 1.9871428571428573e-05, "loss": 1.4495, "step": 4218000 }, { "epoch": 21.54, "learning_rate": 1.9867857142857144e-05, "loss": 1.4576, "step": 4218500 }, { "epoch": 21.54, "learning_rate": 1.9864285714285715e-05, "loss": 1.453, "step": 4219000 }, { "epoch": 21.54, "learning_rate": 1.986071428571429e-05, "loss": 1.4503, "step": 4219500 }, { "epoch": 21.54, "learning_rate": 1.9857142857142856e-05, "loss": 1.453, "step": 4220000 }, { "epoch": 21.55, "learning_rate": 1.985357142857143e-05, "loss": 1.4543, "step": 4220500 }, { "epoch": 21.55, "learning_rate": 1.985e-05, "loss": 1.4516, "step": 4221000 }, { "epoch": 21.55, "learning_rate": 1.9846428571428572e-05, "loss": 1.458, "step": 4221500 }, { "epoch": 21.56, "learning_rate": 1.9842857142857143e-05, "loss": 1.4564, "step": 4222000 }, { "epoch": 21.56, "learning_rate": 1.9839285714285717e-05, "loss": 1.4641, "step": 4222500 }, { "epoch": 21.56, "learning_rate": 1.9835714285714288e-05, "loss": 1.4629, "step": 4223000 }, { "epoch": 21.56, "learning_rate": 1.983214285714286e-05, "loss": 1.4595, "step": 4223500 }, { "epoch": 21.57, "learning_rate": 1.982857142857143e-05, "loss": 1.4558, "step": 4224000 }, { "epoch": 21.57, "learning_rate": 1.9825e-05, "loss": 1.4583, "step": 4224500 }, { "epoch": 21.57, "learning_rate": 1.982142857142857e-05, "loss": 1.46, "step": 4225000 }, { "epoch": 21.57, "learning_rate": 1.9817857142857142e-05, "loss": 1.4486, "step": 4225500 }, { "epoch": 21.58, "learning_rate": 1.9814285714285717e-05, "loss": 1.4495, "step": 4226000 }, { "epoch": 21.58, "learning_rate": 1.9810714285714288e-05, "loss": 1.4601, "step": 4226500 }, { "epoch": 21.58, "learning_rate": 1.980714285714286e-05, "loss": 1.4651, "step": 4227000 }, { "epoch": 21.58, "learning_rate": 1.980357142857143e-05, "loss": 1.4555, "step": 4227500 }, { "epoch": 21.59, "learning_rate": 1.9800000000000004e-05, "loss": 1.4513, "step": 4228000 }, { "epoch": 21.59, "learning_rate": 1.979642857142857e-05, "loss": 1.4633, "step": 4228500 }, { "epoch": 21.59, "learning_rate": 1.9792857142857142e-05, "loss": 1.4556, "step": 4229000 }, { "epoch": 21.6, "learning_rate": 1.9789285714285716e-05, "loss": 1.4613, "step": 4229500 }, { "epoch": 21.6, "learning_rate": 1.9785714285714287e-05, "loss": 1.4621, "step": 4230000 }, { "epoch": 21.6, "learning_rate": 1.9782142857142858e-05, "loss": 1.4594, "step": 4230500 }, { "epoch": 21.6, "learning_rate": 1.977857142857143e-05, "loss": 1.4545, "step": 4231000 }, { "epoch": 21.61, "learning_rate": 1.9775000000000003e-05, "loss": 1.4627, "step": 4231500 }, { "epoch": 21.61, "learning_rate": 1.977142857142857e-05, "loss": 1.459, "step": 4232000 }, { "epoch": 21.61, "learning_rate": 1.9767857142857145e-05, "loss": 1.4592, "step": 4232500 }, { "epoch": 21.61, "learning_rate": 1.9764285714285716e-05, "loss": 1.4653, "step": 4233000 }, { "epoch": 21.62, "learning_rate": 1.9760714285714286e-05, "loss": 1.4669, "step": 4233500 }, { "epoch": 21.62, "learning_rate": 1.9757142857142857e-05, "loss": 1.4592, "step": 4234000 }, { "epoch": 21.62, "learning_rate": 1.9753571428571428e-05, "loss": 1.4546, "step": 4234500 }, { "epoch": 21.62, "learning_rate": 1.9750000000000002e-05, "loss": 1.4575, "step": 4235000 }, { "epoch": 21.63, "learning_rate": 1.974642857142857e-05, "loss": 1.4566, "step": 4235500 }, { "epoch": 21.63, "learning_rate": 1.9742857142857144e-05, "loss": 1.4602, "step": 4236000 }, { "epoch": 21.63, "learning_rate": 1.9739285714285715e-05, "loss": 1.4581, "step": 4236500 }, { "epoch": 21.64, "learning_rate": 1.973571428571429e-05, "loss": 1.4497, "step": 4237000 }, { "epoch": 21.64, "learning_rate": 1.9732142857142857e-05, "loss": 1.4642, "step": 4237500 }, { "epoch": 21.64, "learning_rate": 1.972857142857143e-05, "loss": 1.4608, "step": 4238000 }, { "epoch": 21.64, "learning_rate": 1.9725000000000002e-05, "loss": 1.4574, "step": 4238500 }, { "epoch": 21.65, "learning_rate": 1.9721428571428573e-05, "loss": 1.4547, "step": 4239000 }, { "epoch": 21.65, "learning_rate": 1.9717857142857143e-05, "loss": 1.4726, "step": 4239500 }, { "epoch": 21.65, "learning_rate": 1.9714285714285714e-05, "loss": 1.4576, "step": 4240000 }, { "epoch": 21.65, "learning_rate": 1.971071428571429e-05, "loss": 1.4618, "step": 4240500 }, { "epoch": 21.66, "learning_rate": 1.9707142857142856e-05, "loss": 1.4616, "step": 4241000 }, { "epoch": 21.66, "learning_rate": 1.970357142857143e-05, "loss": 1.4619, "step": 4241500 }, { "epoch": 21.66, "learning_rate": 1.97e-05, "loss": 1.4565, "step": 4242000 }, { "epoch": 21.66, "learning_rate": 1.9696428571428572e-05, "loss": 1.4536, "step": 4242500 }, { "epoch": 21.67, "learning_rate": 1.9692857142857143e-05, "loss": 1.4574, "step": 4243000 }, { "epoch": 21.67, "learning_rate": 1.9689285714285717e-05, "loss": 1.4601, "step": 4243500 }, { "epoch": 21.67, "learning_rate": 1.9685714285714288e-05, "loss": 1.4513, "step": 4244000 }, { "epoch": 21.68, "learning_rate": 1.968214285714286e-05, "loss": 1.4625, "step": 4244500 }, { "epoch": 21.68, "learning_rate": 1.967857142857143e-05, "loss": 1.4669, "step": 4245000 }, { "epoch": 21.68, "learning_rate": 1.9675e-05, "loss": 1.4578, "step": 4245500 }, { "epoch": 21.68, "learning_rate": 1.967142857142857e-05, "loss": 1.458, "step": 4246000 }, { "epoch": 21.69, "learning_rate": 1.9667857142857142e-05, "loss": 1.4562, "step": 4246500 }, { "epoch": 21.69, "learning_rate": 1.9664285714285716e-05, "loss": 1.4531, "step": 4247000 }, { "epoch": 21.69, "learning_rate": 1.9660714285714287e-05, "loss": 1.4605, "step": 4247500 }, { "epoch": 21.69, "learning_rate": 1.9657142857142858e-05, "loss": 1.4597, "step": 4248000 }, { "epoch": 21.7, "learning_rate": 1.965357142857143e-05, "loss": 1.4551, "step": 4248500 }, { "epoch": 21.7, "learning_rate": 1.9650000000000003e-05, "loss": 1.4619, "step": 4249000 }, { "epoch": 21.7, "learning_rate": 1.964642857142857e-05, "loss": 1.4577, "step": 4249500 }, { "epoch": 21.7, "learning_rate": 1.9642857142857145e-05, "loss": 1.4577, "step": 4250000 }, { "epoch": 21.71, "learning_rate": 1.9639285714285716e-05, "loss": 1.4657, "step": 4250500 }, { "epoch": 21.71, "learning_rate": 1.9635714285714287e-05, "loss": 1.4621, "step": 4251000 }, { "epoch": 21.71, "learning_rate": 1.9632142857142858e-05, "loss": 1.4525, "step": 4251500 }, { "epoch": 21.72, "learning_rate": 1.962857142857143e-05, "loss": 1.4537, "step": 4252000 }, { "epoch": 21.72, "learning_rate": 1.9625000000000003e-05, "loss": 1.4668, "step": 4252500 }, { "epoch": 21.72, "learning_rate": 1.962142857142857e-05, "loss": 1.4615, "step": 4253000 }, { "epoch": 21.72, "learning_rate": 1.9617857142857144e-05, "loss": 1.4505, "step": 4253500 }, { "epoch": 21.73, "learning_rate": 1.9614285714285715e-05, "loss": 1.4594, "step": 4254000 }, { "epoch": 21.73, "learning_rate": 1.961071428571429e-05, "loss": 1.4558, "step": 4254500 }, { "epoch": 21.73, "learning_rate": 1.9607142857142857e-05, "loss": 1.4522, "step": 4255000 }, { "epoch": 21.73, "learning_rate": 1.9603571428571428e-05, "loss": 1.4572, "step": 4255500 }, { "epoch": 21.74, "learning_rate": 1.9600000000000002e-05, "loss": 1.4501, "step": 4256000 }, { "epoch": 21.74, "learning_rate": 1.959642857142857e-05, "loss": 1.4647, "step": 4256500 }, { "epoch": 21.74, "learning_rate": 1.9592857142857144e-05, "loss": 1.458, "step": 4257000 }, { "epoch": 21.74, "learning_rate": 1.9589285714285715e-05, "loss": 1.4612, "step": 4257500 }, { "epoch": 21.75, "learning_rate": 1.958571428571429e-05, "loss": 1.4503, "step": 4258000 }, { "epoch": 21.75, "learning_rate": 1.9582142857142856e-05, "loss": 1.4491, "step": 4258500 }, { "epoch": 21.75, "learning_rate": 1.957857142857143e-05, "loss": 1.4485, "step": 4259000 }, { "epoch": 21.76, "learning_rate": 1.9575e-05, "loss": 1.459, "step": 4259500 }, { "epoch": 21.76, "learning_rate": 1.9571428571428572e-05, "loss": 1.4566, "step": 4260000 }, { "epoch": 21.76, "learning_rate": 1.9567857142857143e-05, "loss": 1.4599, "step": 4260500 }, { "epoch": 21.76, "learning_rate": 1.9564285714285714e-05, "loss": 1.4502, "step": 4261000 }, { "epoch": 21.77, "learning_rate": 1.9560714285714288e-05, "loss": 1.4516, "step": 4261500 }, { "epoch": 21.77, "learning_rate": 1.9557142857142856e-05, "loss": 1.4604, "step": 4262000 }, { "epoch": 21.77, "learning_rate": 1.955357142857143e-05, "loss": 1.4602, "step": 4262500 }, { "epoch": 21.77, "learning_rate": 1.955e-05, "loss": 1.4591, "step": 4263000 }, { "epoch": 21.78, "learning_rate": 1.9546428571428572e-05, "loss": 1.4626, "step": 4263500 }, { "epoch": 21.78, "learning_rate": 1.9542857142857143e-05, "loss": 1.452, "step": 4264000 }, { "epoch": 21.78, "learning_rate": 1.9539285714285717e-05, "loss": 1.4543, "step": 4264500 }, { "epoch": 21.78, "learning_rate": 1.9535714285714288e-05, "loss": 1.4549, "step": 4265000 }, { "epoch": 21.79, "learning_rate": 1.953214285714286e-05, "loss": 1.4524, "step": 4265500 }, { "epoch": 21.79, "learning_rate": 1.952857142857143e-05, "loss": 1.4546, "step": 4266000 }, { "epoch": 21.79, "learning_rate": 1.9525e-05, "loss": 1.4519, "step": 4266500 }, { "epoch": 21.8, "learning_rate": 1.952142857142857e-05, "loss": 1.4522, "step": 4267000 }, { "epoch": 21.8, "learning_rate": 1.9517857142857142e-05, "loss": 1.4586, "step": 4267500 }, { "epoch": 21.8, "learning_rate": 1.9514285714285716e-05, "loss": 1.459, "step": 4268000 }, { "epoch": 21.8, "learning_rate": 1.9510714285714287e-05, "loss": 1.4637, "step": 4268500 }, { "epoch": 21.81, "learning_rate": 1.9507142857142858e-05, "loss": 1.4557, "step": 4269000 }, { "epoch": 21.81, "learning_rate": 1.950357142857143e-05, "loss": 1.4631, "step": 4269500 }, { "epoch": 21.81, "learning_rate": 1.9500000000000003e-05, "loss": 1.4734, "step": 4270000 }, { "epoch": 21.81, "learning_rate": 1.949642857142857e-05, "loss": 1.4611, "step": 4270500 }, { "epoch": 21.82, "learning_rate": 1.9492857142857145e-05, "loss": 1.4575, "step": 4271000 }, { "epoch": 21.82, "learning_rate": 1.9489285714285716e-05, "loss": 1.4616, "step": 4271500 }, { "epoch": 21.82, "learning_rate": 1.9485714285714286e-05, "loss": 1.4612, "step": 4272000 }, { "epoch": 21.82, "learning_rate": 1.9482142857142857e-05, "loss": 1.4602, "step": 4272500 }, { "epoch": 21.83, "learning_rate": 1.9478571428571428e-05, "loss": 1.4612, "step": 4273000 }, { "epoch": 21.83, "learning_rate": 1.9475000000000002e-05, "loss": 1.4619, "step": 4273500 }, { "epoch": 21.83, "learning_rate": 1.947142857142857e-05, "loss": 1.4558, "step": 4274000 }, { "epoch": 21.84, "learning_rate": 1.9467857142857144e-05, "loss": 1.4596, "step": 4274500 }, { "epoch": 21.84, "learning_rate": 1.9464285714285715e-05, "loss": 1.4607, "step": 4275000 }, { "epoch": 21.84, "learning_rate": 1.946071428571429e-05, "loss": 1.4578, "step": 4275500 }, { "epoch": 21.84, "learning_rate": 1.9457142857142857e-05, "loss": 1.4507, "step": 4276000 }, { "epoch": 21.85, "learning_rate": 1.945357142857143e-05, "loss": 1.4588, "step": 4276500 }, { "epoch": 21.85, "learning_rate": 1.9450000000000002e-05, "loss": 1.4595, "step": 4277000 }, { "epoch": 21.85, "learning_rate": 1.9446428571428573e-05, "loss": 1.4598, "step": 4277500 }, { "epoch": 21.85, "learning_rate": 1.9442857142857144e-05, "loss": 1.4599, "step": 4278000 }, { "epoch": 21.86, "learning_rate": 1.9439285714285714e-05, "loss": 1.4601, "step": 4278500 }, { "epoch": 21.86, "learning_rate": 1.943571428571429e-05, "loss": 1.4518, "step": 4279000 }, { "epoch": 21.86, "learning_rate": 1.9432142857142856e-05, "loss": 1.4587, "step": 4279500 }, { "epoch": 21.86, "learning_rate": 1.942857142857143e-05, "loss": 1.4578, "step": 4280000 }, { "epoch": 21.87, "learning_rate": 1.9425e-05, "loss": 1.4645, "step": 4280500 }, { "epoch": 21.87, "learning_rate": 1.9421428571428572e-05, "loss": 1.4535, "step": 4281000 }, { "epoch": 21.87, "learning_rate": 1.9417857142857143e-05, "loss": 1.4576, "step": 4281500 }, { "epoch": 21.88, "learning_rate": 1.9414285714285714e-05, "loss": 1.4559, "step": 4282000 }, { "epoch": 21.88, "learning_rate": 1.9410714285714288e-05, "loss": 1.4578, "step": 4282500 }, { "epoch": 21.88, "learning_rate": 1.9407142857142856e-05, "loss": 1.4545, "step": 4283000 }, { "epoch": 21.88, "learning_rate": 1.940357142857143e-05, "loss": 1.4558, "step": 4283500 }, { "epoch": 21.89, "learning_rate": 1.94e-05, "loss": 1.4625, "step": 4284000 }, { "epoch": 21.89, "learning_rate": 1.9396428571428575e-05, "loss": 1.4544, "step": 4284500 }, { "epoch": 21.89, "learning_rate": 1.9392857142857142e-05, "loss": 1.4569, "step": 4285000 }, { "epoch": 21.89, "learning_rate": 1.9389285714285717e-05, "loss": 1.4567, "step": 4285500 }, { "epoch": 21.9, "learning_rate": 1.9385714285714287e-05, "loss": 1.4562, "step": 4286000 }, { "epoch": 21.9, "learning_rate": 1.9382142857142858e-05, "loss": 1.4442, "step": 4286500 }, { "epoch": 21.9, "learning_rate": 1.937857142857143e-05, "loss": 1.4493, "step": 4287000 }, { "epoch": 21.9, "learning_rate": 1.9375e-05, "loss": 1.453, "step": 4287500 }, { "epoch": 21.91, "learning_rate": 1.9371428571428574e-05, "loss": 1.451, "step": 4288000 }, { "epoch": 21.91, "learning_rate": 1.9367857142857142e-05, "loss": 1.462, "step": 4288500 }, { "epoch": 21.91, "learning_rate": 1.9364285714285716e-05, "loss": 1.4627, "step": 4289000 }, { "epoch": 21.91, "learning_rate": 1.9360714285714287e-05, "loss": 1.4557, "step": 4289500 }, { "epoch": 21.92, "learning_rate": 1.9357142857142858e-05, "loss": 1.4584, "step": 4290000 }, { "epoch": 21.92, "learning_rate": 1.935357142857143e-05, "loss": 1.4594, "step": 4290500 }, { "epoch": 21.92, "learning_rate": 1.9350000000000003e-05, "loss": 1.4603, "step": 4291000 }, { "epoch": 21.93, "learning_rate": 1.9346428571428574e-05, "loss": 1.4598, "step": 4291500 }, { "epoch": 21.93, "learning_rate": 1.9342857142857144e-05, "loss": 1.4567, "step": 4292000 }, { "epoch": 21.93, "learning_rate": 1.9339285714285715e-05, "loss": 1.4583, "step": 4292500 }, { "epoch": 21.93, "learning_rate": 1.9335714285714286e-05, "loss": 1.4644, "step": 4293000 }, { "epoch": 21.94, "learning_rate": 1.9332142857142857e-05, "loss": 1.4504, "step": 4293500 }, { "epoch": 21.94, "learning_rate": 1.9328571428571428e-05, "loss": 1.4552, "step": 4294000 }, { "epoch": 21.94, "learning_rate": 1.9325000000000002e-05, "loss": 1.4535, "step": 4294500 }, { "epoch": 21.94, "learning_rate": 1.9321428571428573e-05, "loss": 1.4664, "step": 4295000 }, { "epoch": 21.95, "learning_rate": 1.9317857142857144e-05, "loss": 1.4619, "step": 4295500 }, { "epoch": 21.95, "learning_rate": 1.9314285714285715e-05, "loss": 1.4643, "step": 4296000 }, { "epoch": 21.95, "learning_rate": 1.931071428571429e-05, "loss": 1.4618, "step": 4296500 }, { "epoch": 21.95, "learning_rate": 1.9307142857142856e-05, "loss": 1.4722, "step": 4297000 }, { "epoch": 21.96, "learning_rate": 1.930357142857143e-05, "loss": 1.4623, "step": 4297500 }, { "epoch": 21.96, "learning_rate": 1.93e-05, "loss": 1.456, "step": 4298000 }, { "epoch": 21.96, "learning_rate": 1.9296428571428572e-05, "loss": 1.4532, "step": 4298500 }, { "epoch": 21.97, "learning_rate": 1.9292857142857143e-05, "loss": 1.4607, "step": 4299000 }, { "epoch": 21.97, "learning_rate": 1.9289285714285714e-05, "loss": 1.4393, "step": 4299500 }, { "epoch": 21.97, "learning_rate": 1.928571428571429e-05, "loss": 1.4567, "step": 4300000 }, { "epoch": 21.97, "learning_rate": 1.9282142857142856e-05, "loss": 1.4411, "step": 4300500 }, { "epoch": 21.98, "learning_rate": 1.927857142857143e-05, "loss": 1.4512, "step": 4301000 }, { "epoch": 21.98, "learning_rate": 1.9275e-05, "loss": 1.455, "step": 4301500 }, { "epoch": 21.98, "learning_rate": 1.9271428571428575e-05, "loss": 1.4492, "step": 4302000 }, { "epoch": 21.98, "learning_rate": 1.9267857142857143e-05, "loss": 1.4449, "step": 4302500 }, { "epoch": 21.99, "learning_rate": 1.9264285714285717e-05, "loss": 1.4588, "step": 4303000 }, { "epoch": 21.99, "learning_rate": 1.9260714285714288e-05, "loss": 1.4582, "step": 4303500 }, { "epoch": 21.99, "learning_rate": 1.9257142857142855e-05, "loss": 1.4452, "step": 4304000 }, { "epoch": 21.99, "learning_rate": 1.925357142857143e-05, "loss": 1.4609, "step": 4304500 }, { "epoch": 22.0, "learning_rate": 1.925e-05, "loss": 1.4497, "step": 4305000 }, { "epoch": 22.0, "learning_rate": 1.9246428571428575e-05, "loss": 1.461, "step": 4305500 }, { "epoch": 22.0, "learning_rate": 1.9242857142857142e-05, "loss": 1.443, "step": 4306000 }, { "epoch": 22.01, "learning_rate": 1.9239285714285716e-05, "loss": 1.4526, "step": 4306500 }, { "epoch": 22.01, "learning_rate": 1.9235714285714287e-05, "loss": 1.4484, "step": 4307000 }, { "epoch": 22.01, "learning_rate": 1.9232142857142858e-05, "loss": 1.4527, "step": 4307500 }, { "epoch": 22.01, "learning_rate": 1.922857142857143e-05, "loss": 1.4584, "step": 4308000 }, { "epoch": 22.02, "learning_rate": 1.9225e-05, "loss": 1.4624, "step": 4308500 }, { "epoch": 22.02, "learning_rate": 1.9221428571428574e-05, "loss": 1.4618, "step": 4309000 }, { "epoch": 22.02, "learning_rate": 1.921785714285714e-05, "loss": 1.4514, "step": 4309500 }, { "epoch": 22.02, "learning_rate": 1.9214285714285716e-05, "loss": 1.4554, "step": 4310000 }, { "epoch": 22.03, "learning_rate": 1.9210714285714287e-05, "loss": 1.4492, "step": 4310500 }, { "epoch": 22.03, "learning_rate": 1.9207142857142857e-05, "loss": 1.4469, "step": 4311000 }, { "epoch": 22.03, "learning_rate": 1.9203571428571428e-05, "loss": 1.4522, "step": 4311500 }, { "epoch": 22.03, "learning_rate": 1.9200000000000003e-05, "loss": 1.4422, "step": 4312000 }, { "epoch": 22.04, "learning_rate": 1.9196428571428573e-05, "loss": 1.4517, "step": 4312500 }, { "epoch": 22.04, "learning_rate": 1.9192857142857144e-05, "loss": 1.4432, "step": 4313000 }, { "epoch": 22.04, "learning_rate": 1.9189285714285715e-05, "loss": 1.466, "step": 4313500 }, { "epoch": 22.05, "learning_rate": 1.9185714285714286e-05, "loss": 1.4615, "step": 4314000 }, { "epoch": 22.05, "learning_rate": 1.9182142857142857e-05, "loss": 1.4478, "step": 4314500 }, { "epoch": 22.05, "learning_rate": 1.9178571428571428e-05, "loss": 1.4487, "step": 4315000 }, { "epoch": 22.05, "learning_rate": 1.9175000000000002e-05, "loss": 1.4567, "step": 4315500 }, { "epoch": 22.06, "learning_rate": 1.9171428571428573e-05, "loss": 1.4558, "step": 4316000 }, { "epoch": 22.06, "learning_rate": 1.9167857142857144e-05, "loss": 1.446, "step": 4316500 }, { "epoch": 22.06, "learning_rate": 1.9164285714285714e-05, "loss": 1.4448, "step": 4317000 }, { "epoch": 22.06, "learning_rate": 1.916071428571429e-05, "loss": 1.4486, "step": 4317500 }, { "epoch": 22.07, "learning_rate": 1.9157142857142856e-05, "loss": 1.4555, "step": 4318000 }, { "epoch": 22.07, "learning_rate": 1.915357142857143e-05, "loss": 1.45, "step": 4318500 }, { "epoch": 22.07, "learning_rate": 1.915e-05, "loss": 1.4552, "step": 4319000 }, { "epoch": 22.07, "learning_rate": 1.9146428571428572e-05, "loss": 1.4565, "step": 4319500 }, { "epoch": 22.08, "learning_rate": 1.9142857142857143e-05, "loss": 1.4544, "step": 4320000 }, { "epoch": 22.08, "learning_rate": 1.9139285714285714e-05, "loss": 1.4503, "step": 4320500 }, { "epoch": 22.08, "learning_rate": 1.9135714285714288e-05, "loss": 1.4525, "step": 4321000 }, { "epoch": 22.09, "learning_rate": 1.9132142857142856e-05, "loss": 1.4508, "step": 4321500 }, { "epoch": 22.09, "learning_rate": 1.912857142857143e-05, "loss": 1.4556, "step": 4322000 }, { "epoch": 22.09, "learning_rate": 1.9125e-05, "loss": 1.4493, "step": 4322500 }, { "epoch": 22.09, "learning_rate": 1.9121428571428575e-05, "loss": 1.4488, "step": 4323000 }, { "epoch": 22.1, "learning_rate": 1.9117857142857142e-05, "loss": 1.4485, "step": 4323500 }, { "epoch": 22.1, "learning_rate": 1.9114285714285717e-05, "loss": 1.4539, "step": 4324000 }, { "epoch": 22.1, "learning_rate": 1.9110714285714288e-05, "loss": 1.4491, "step": 4324500 }, { "epoch": 22.1, "learning_rate": 1.910714285714286e-05, "loss": 1.4519, "step": 4325000 }, { "epoch": 22.11, "learning_rate": 1.910357142857143e-05, "loss": 1.4471, "step": 4325500 }, { "epoch": 22.11, "learning_rate": 1.91e-05, "loss": 1.4511, "step": 4326000 }, { "epoch": 22.11, "learning_rate": 1.9096428571428574e-05, "loss": 1.447, "step": 4326500 }, { "epoch": 22.11, "learning_rate": 1.9092857142857142e-05, "loss": 1.4531, "step": 4327000 }, { "epoch": 22.12, "learning_rate": 1.9089285714285716e-05, "loss": 1.4492, "step": 4327500 }, { "epoch": 22.12, "learning_rate": 1.9085714285714287e-05, "loss": 1.4506, "step": 4328000 }, { "epoch": 22.12, "learning_rate": 1.9082142857142858e-05, "loss": 1.4401, "step": 4328500 }, { "epoch": 22.13, "learning_rate": 1.907857142857143e-05, "loss": 1.4478, "step": 4329000 }, { "epoch": 22.13, "learning_rate": 1.9075000000000003e-05, "loss": 1.456, "step": 4329500 }, { "epoch": 22.13, "learning_rate": 1.9071428571428574e-05, "loss": 1.4511, "step": 4330000 }, { "epoch": 22.13, "learning_rate": 1.906785714285714e-05, "loss": 1.4482, "step": 4330500 }, { "epoch": 22.14, "learning_rate": 1.9064285714285715e-05, "loss": 1.4489, "step": 4331000 }, { "epoch": 22.14, "learning_rate": 1.9060714285714286e-05, "loss": 1.4539, "step": 4331500 }, { "epoch": 22.14, "learning_rate": 1.9057142857142857e-05, "loss": 1.4496, "step": 4332000 }, { "epoch": 22.14, "learning_rate": 1.9053571428571428e-05, "loss": 1.4552, "step": 4332500 }, { "epoch": 22.15, "learning_rate": 1.9050000000000002e-05, "loss": 1.4474, "step": 4333000 }, { "epoch": 22.15, "learning_rate": 1.9046428571428573e-05, "loss": 1.4494, "step": 4333500 }, { "epoch": 22.15, "learning_rate": 1.9042857142857144e-05, "loss": 1.4558, "step": 4334000 }, { "epoch": 22.15, "learning_rate": 1.9039285714285715e-05, "loss": 1.4577, "step": 4334500 }, { "epoch": 22.16, "learning_rate": 1.9035714285714286e-05, "loss": 1.4477, "step": 4335000 }, { "epoch": 22.16, "learning_rate": 1.9032142857142857e-05, "loss": 1.4468, "step": 4335500 }, { "epoch": 22.16, "learning_rate": 1.9028571428571427e-05, "loss": 1.4558, "step": 4336000 }, { "epoch": 22.17, "learning_rate": 1.9025e-05, "loss": 1.4511, "step": 4336500 }, { "epoch": 22.17, "learning_rate": 1.9021428571428573e-05, "loss": 1.4462, "step": 4337000 }, { "epoch": 22.17, "learning_rate": 1.9017857142857143e-05, "loss": 1.4526, "step": 4337500 }, { "epoch": 22.17, "learning_rate": 1.9014285714285714e-05, "loss": 1.4503, "step": 4338000 }, { "epoch": 22.18, "learning_rate": 1.901071428571429e-05, "loss": 1.4495, "step": 4338500 }, { "epoch": 22.18, "learning_rate": 1.9007142857142856e-05, "loss": 1.4476, "step": 4339000 }, { "epoch": 22.18, "learning_rate": 1.900357142857143e-05, "loss": 1.4542, "step": 4339500 }, { "epoch": 22.18, "learning_rate": 1.9e-05, "loss": 1.4483, "step": 4340000 }, { "epoch": 22.19, "learning_rate": 1.8996428571428572e-05, "loss": 1.4434, "step": 4340500 }, { "epoch": 22.19, "learning_rate": 1.8992857142857143e-05, "loss": 1.466, "step": 4341000 }, { "epoch": 22.19, "learning_rate": 1.8989285714285714e-05, "loss": 1.447, "step": 4341500 }, { "epoch": 22.19, "learning_rate": 1.8985714285714288e-05, "loss": 1.4536, "step": 4342000 }, { "epoch": 22.2, "learning_rate": 1.8982142857142855e-05, "loss": 1.4393, "step": 4342500 }, { "epoch": 22.2, "learning_rate": 1.897857142857143e-05, "loss": 1.4516, "step": 4343000 }, { "epoch": 22.2, "learning_rate": 1.8975e-05, "loss": 1.4553, "step": 4343500 }, { "epoch": 22.21, "learning_rate": 1.8971428571428575e-05, "loss": 1.4489, "step": 4344000 }, { "epoch": 22.21, "learning_rate": 1.8967857142857142e-05, "loss": 1.4605, "step": 4344500 }, { "epoch": 22.21, "learning_rate": 1.8964285714285716e-05, "loss": 1.4415, "step": 4345000 }, { "epoch": 22.21, "learning_rate": 1.8960714285714287e-05, "loss": 1.4526, "step": 4345500 }, { "epoch": 22.22, "learning_rate": 1.8957142857142858e-05, "loss": 1.4587, "step": 4346000 }, { "epoch": 22.22, "learning_rate": 1.895357142857143e-05, "loss": 1.4443, "step": 4346500 }, { "epoch": 22.22, "learning_rate": 1.895e-05, "loss": 1.4398, "step": 4347000 }, { "epoch": 22.22, "learning_rate": 1.8946428571428574e-05, "loss": 1.449, "step": 4347500 }, { "epoch": 22.23, "learning_rate": 1.894285714285714e-05, "loss": 1.4638, "step": 4348000 }, { "epoch": 22.23, "learning_rate": 1.8939285714285716e-05, "loss": 1.4452, "step": 4348500 }, { "epoch": 22.23, "learning_rate": 1.8935714285714287e-05, "loss": 1.4417, "step": 4349000 }, { "epoch": 22.23, "learning_rate": 1.893214285714286e-05, "loss": 1.4511, "step": 4349500 }, { "epoch": 22.24, "learning_rate": 1.892857142857143e-05, "loss": 1.4473, "step": 4350000 }, { "epoch": 22.24, "learning_rate": 1.8925000000000003e-05, "loss": 1.4472, "step": 4350500 }, { "epoch": 22.24, "learning_rate": 1.8921428571428573e-05, "loss": 1.4496, "step": 4351000 }, { "epoch": 22.25, "learning_rate": 1.8917857142857144e-05, "loss": 1.451, "step": 4351500 }, { "epoch": 22.25, "learning_rate": 1.8914285714285715e-05, "loss": 1.4447, "step": 4352000 }, { "epoch": 22.25, "learning_rate": 1.8910714285714286e-05, "loss": 1.4464, "step": 4352500 }, { "epoch": 22.25, "learning_rate": 1.890714285714286e-05, "loss": 1.4495, "step": 4353000 }, { "epoch": 22.26, "learning_rate": 1.8903571428571428e-05, "loss": 1.4572, "step": 4353500 }, { "epoch": 22.26, "learning_rate": 1.8900000000000002e-05, "loss": 1.4507, "step": 4354000 }, { "epoch": 22.26, "learning_rate": 1.8896428571428573e-05, "loss": 1.4536, "step": 4354500 }, { "epoch": 22.26, "learning_rate": 1.8892857142857144e-05, "loss": 1.4563, "step": 4355000 }, { "epoch": 22.27, "learning_rate": 1.8889285714285715e-05, "loss": 1.4495, "step": 4355500 }, { "epoch": 22.27, "learning_rate": 1.888571428571429e-05, "loss": 1.4405, "step": 4356000 }, { "epoch": 22.27, "learning_rate": 1.888214285714286e-05, "loss": 1.4551, "step": 4356500 }, { "epoch": 22.27, "learning_rate": 1.8878571428571427e-05, "loss": 1.4554, "step": 4357000 }, { "epoch": 22.28, "learning_rate": 1.8875e-05, "loss": 1.4458, "step": 4357500 }, { "epoch": 22.28, "learning_rate": 1.8871428571428572e-05, "loss": 1.4616, "step": 4358000 }, { "epoch": 22.28, "learning_rate": 1.8867857142857143e-05, "loss": 1.4453, "step": 4358500 }, { "epoch": 22.28, "learning_rate": 1.8864285714285714e-05, "loss": 1.4524, "step": 4359000 }, { "epoch": 22.29, "learning_rate": 1.8860714285714288e-05, "loss": 1.4463, "step": 4359500 }, { "epoch": 22.29, "learning_rate": 1.885714285714286e-05, "loss": 1.4405, "step": 4360000 }, { "epoch": 22.29, "learning_rate": 1.885357142857143e-05, "loss": 1.4522, "step": 4360500 }, { "epoch": 22.3, "learning_rate": 1.885e-05, "loss": 1.4531, "step": 4361000 }, { "epoch": 22.3, "learning_rate": 1.884642857142857e-05, "loss": 1.4503, "step": 4361500 }, { "epoch": 22.3, "learning_rate": 1.8842857142857143e-05, "loss": 1.4463, "step": 4362000 }, { "epoch": 22.3, "learning_rate": 1.8839285714285713e-05, "loss": 1.4495, "step": 4362500 }, { "epoch": 22.31, "learning_rate": 1.8835714285714288e-05, "loss": 1.4529, "step": 4363000 }, { "epoch": 22.31, "learning_rate": 1.883214285714286e-05, "loss": 1.4481, "step": 4363500 }, { "epoch": 22.31, "learning_rate": 1.882857142857143e-05, "loss": 1.4482, "step": 4364000 }, { "epoch": 22.31, "learning_rate": 1.8825e-05, "loss": 1.447, "step": 4364500 }, { "epoch": 22.32, "learning_rate": 1.8821428571428574e-05, "loss": 1.4491, "step": 4365000 }, { "epoch": 22.32, "learning_rate": 1.8817857142857142e-05, "loss": 1.4351, "step": 4365500 }, { "epoch": 22.32, "learning_rate": 1.8814285714285716e-05, "loss": 1.4493, "step": 4366000 }, { "epoch": 22.32, "learning_rate": 1.8810714285714287e-05, "loss": 1.4477, "step": 4366500 }, { "epoch": 22.33, "learning_rate": 1.8807142857142858e-05, "loss": 1.4453, "step": 4367000 }, { "epoch": 22.33, "learning_rate": 1.880357142857143e-05, "loss": 1.4429, "step": 4367500 }, { "epoch": 22.33, "learning_rate": 1.88e-05, "loss": 1.4461, "step": 4368000 }, { "epoch": 22.34, "learning_rate": 1.8796428571428574e-05, "loss": 1.4513, "step": 4368500 }, { "epoch": 22.34, "learning_rate": 1.879285714285714e-05, "loss": 1.4528, "step": 4369000 }, { "epoch": 22.34, "learning_rate": 1.8789285714285716e-05, "loss": 1.4354, "step": 4369500 }, { "epoch": 22.34, "learning_rate": 1.8785714285714286e-05, "loss": 1.4426, "step": 4370000 }, { "epoch": 22.35, "learning_rate": 1.878214285714286e-05, "loss": 1.4516, "step": 4370500 }, { "epoch": 22.35, "learning_rate": 1.8778571428571428e-05, "loss": 1.4472, "step": 4371000 }, { "epoch": 22.35, "learning_rate": 1.8775000000000002e-05, "loss": 1.4523, "step": 4371500 }, { "epoch": 22.35, "learning_rate": 1.8771428571428573e-05, "loss": 1.449, "step": 4372000 }, { "epoch": 22.36, "learning_rate": 1.8767857142857144e-05, "loss": 1.4455, "step": 4372500 }, { "epoch": 22.36, "learning_rate": 1.8764285714285715e-05, "loss": 1.4383, "step": 4373000 }, { "epoch": 22.36, "learning_rate": 1.8760714285714286e-05, "loss": 1.4513, "step": 4373500 }, { "epoch": 22.36, "learning_rate": 1.875714285714286e-05, "loss": 1.4512, "step": 4374000 }, { "epoch": 22.37, "learning_rate": 1.8753571428571427e-05, "loss": 1.4495, "step": 4374500 }, { "epoch": 22.37, "learning_rate": 1.8750000000000002e-05, "loss": 1.4443, "step": 4375000 }, { "epoch": 22.37, "learning_rate": 1.8746428571428573e-05, "loss": 1.4516, "step": 4375500 }, { "epoch": 22.38, "learning_rate": 1.8742857142857143e-05, "loss": 1.4494, "step": 4376000 }, { "epoch": 22.38, "learning_rate": 1.8739285714285714e-05, "loss": 1.4486, "step": 4376500 }, { "epoch": 22.38, "learning_rate": 1.873571428571429e-05, "loss": 1.453, "step": 4377000 }, { "epoch": 22.38, "learning_rate": 1.873214285714286e-05, "loss": 1.4467, "step": 4377500 }, { "epoch": 22.39, "learning_rate": 1.872857142857143e-05, "loss": 1.4472, "step": 4378000 }, { "epoch": 22.39, "learning_rate": 1.8725e-05, "loss": 1.4485, "step": 4378500 }, { "epoch": 22.39, "learning_rate": 1.8721428571428572e-05, "loss": 1.4374, "step": 4379000 }, { "epoch": 22.39, "learning_rate": 1.8717857142857143e-05, "loss": 1.4458, "step": 4379500 }, { "epoch": 22.4, "learning_rate": 1.8714285714285714e-05, "loss": 1.4415, "step": 4380000 }, { "epoch": 22.4, "learning_rate": 1.8710714285714288e-05, "loss": 1.4464, "step": 4380500 }, { "epoch": 22.4, "learning_rate": 1.870714285714286e-05, "loss": 1.4537, "step": 4381000 }, { "epoch": 22.4, "learning_rate": 1.870357142857143e-05, "loss": 1.4461, "step": 4381500 }, { "epoch": 22.41, "learning_rate": 1.87e-05, "loss": 1.4464, "step": 4382000 }, { "epoch": 22.41, "learning_rate": 1.869642857142857e-05, "loss": 1.447, "step": 4382500 }, { "epoch": 22.41, "learning_rate": 1.8692857142857142e-05, "loss": 1.4545, "step": 4383000 }, { "epoch": 22.42, "learning_rate": 1.8689285714285713e-05, "loss": 1.4435, "step": 4383500 }, { "epoch": 22.42, "learning_rate": 1.8685714285714287e-05, "loss": 1.4493, "step": 4384000 }, { "epoch": 22.42, "learning_rate": 1.8682142857142858e-05, "loss": 1.4545, "step": 4384500 }, { "epoch": 22.42, "learning_rate": 1.867857142857143e-05, "loss": 1.4524, "step": 4385000 }, { "epoch": 22.43, "learning_rate": 1.8675e-05, "loss": 1.4478, "step": 4385500 }, { "epoch": 22.43, "learning_rate": 1.8671428571428574e-05, "loss": 1.4552, "step": 4386000 }, { "epoch": 22.43, "learning_rate": 1.866785714285714e-05, "loss": 1.4427, "step": 4386500 }, { "epoch": 22.43, "learning_rate": 1.8664285714285716e-05, "loss": 1.4348, "step": 4387000 }, { "epoch": 22.44, "learning_rate": 1.8660714285714287e-05, "loss": 1.4527, "step": 4387500 }, { "epoch": 22.44, "learning_rate": 1.8657142857142858e-05, "loss": 1.4498, "step": 4388000 }, { "epoch": 22.44, "learning_rate": 1.865357142857143e-05, "loss": 1.4477, "step": 4388500 }, { "epoch": 22.44, "learning_rate": 1.865e-05, "loss": 1.4454, "step": 4389000 }, { "epoch": 22.45, "learning_rate": 1.8646428571428574e-05, "loss": 1.4659, "step": 4389500 }, { "epoch": 22.45, "learning_rate": 1.864285714285714e-05, "loss": 1.4471, "step": 4390000 }, { "epoch": 22.45, "learning_rate": 1.8639285714285715e-05, "loss": 1.4576, "step": 4390500 }, { "epoch": 22.46, "learning_rate": 1.8635714285714286e-05, "loss": 1.445, "step": 4391000 }, { "epoch": 22.46, "learning_rate": 1.863214285714286e-05, "loss": 1.4458, "step": 4391500 }, { "epoch": 22.46, "learning_rate": 1.8628571428571428e-05, "loss": 1.4424, "step": 4392000 }, { "epoch": 22.46, "learning_rate": 1.8625000000000002e-05, "loss": 1.441, "step": 4392500 }, { "epoch": 22.47, "learning_rate": 1.8621428571428573e-05, "loss": 1.4513, "step": 4393000 }, { "epoch": 22.47, "learning_rate": 1.8617857142857144e-05, "loss": 1.4578, "step": 4393500 }, { "epoch": 22.47, "learning_rate": 1.8614285714285715e-05, "loss": 1.4475, "step": 4394000 }, { "epoch": 22.47, "learning_rate": 1.8610714285714286e-05, "loss": 1.4527, "step": 4394500 }, { "epoch": 22.48, "learning_rate": 1.860714285714286e-05, "loss": 1.4432, "step": 4395000 }, { "epoch": 22.48, "learning_rate": 1.8603571428571427e-05, "loss": 1.4469, "step": 4395500 }, { "epoch": 22.48, "learning_rate": 1.86e-05, "loss": 1.4502, "step": 4396000 }, { "epoch": 22.48, "learning_rate": 1.8596428571428572e-05, "loss": 1.4433, "step": 4396500 }, { "epoch": 22.49, "learning_rate": 1.8592857142857143e-05, "loss": 1.4407, "step": 4397000 }, { "epoch": 22.49, "learning_rate": 1.8589285714285714e-05, "loss": 1.442, "step": 4397500 }, { "epoch": 22.49, "learning_rate": 1.858571428571429e-05, "loss": 1.4339, "step": 4398000 }, { "epoch": 22.5, "learning_rate": 1.858214285714286e-05, "loss": 1.4461, "step": 4398500 }, { "epoch": 22.5, "learning_rate": 1.857857142857143e-05, "loss": 1.4459, "step": 4399000 }, { "epoch": 22.5, "learning_rate": 1.8575e-05, "loss": 1.4496, "step": 4399500 }, { "epoch": 22.5, "learning_rate": 1.8571428571428572e-05, "loss": 1.4473, "step": 4400000 }, { "epoch": 22.51, "learning_rate": 1.8567857142857143e-05, "loss": 1.4417, "step": 4400500 }, { "epoch": 22.51, "learning_rate": 1.8564285714285713e-05, "loss": 1.4507, "step": 4401000 }, { "epoch": 22.51, "learning_rate": 1.8560714285714288e-05, "loss": 1.4379, "step": 4401500 }, { "epoch": 22.51, "learning_rate": 1.855714285714286e-05, "loss": 1.4425, "step": 4402000 }, { "epoch": 22.52, "learning_rate": 1.855357142857143e-05, "loss": 1.446, "step": 4402500 }, { "epoch": 22.52, "learning_rate": 1.855e-05, "loss": 1.4437, "step": 4403000 }, { "epoch": 22.52, "learning_rate": 1.8546428571428575e-05, "loss": 1.4456, "step": 4403500 }, { "epoch": 22.52, "learning_rate": 1.8542857142857142e-05, "loss": 1.451, "step": 4404000 }, { "epoch": 22.53, "learning_rate": 1.8539285714285716e-05, "loss": 1.4511, "step": 4404500 }, { "epoch": 22.53, "learning_rate": 1.8535714285714287e-05, "loss": 1.4534, "step": 4405000 }, { "epoch": 22.53, "learning_rate": 1.8532142857142858e-05, "loss": 1.4425, "step": 4405500 }, { "epoch": 22.54, "learning_rate": 1.852857142857143e-05, "loss": 1.4473, "step": 4406000 }, { "epoch": 22.54, "learning_rate": 1.8525e-05, "loss": 1.4467, "step": 4406500 }, { "epoch": 22.54, "learning_rate": 1.8521428571428574e-05, "loss": 1.4421, "step": 4407000 }, { "epoch": 22.54, "learning_rate": 1.8517857142857145e-05, "loss": 1.4383, "step": 4407500 }, { "epoch": 22.55, "learning_rate": 1.8514285714285716e-05, "loss": 1.4421, "step": 4408000 }, { "epoch": 22.55, "learning_rate": 1.8510714285714286e-05, "loss": 1.4446, "step": 4408500 }, { "epoch": 22.55, "learning_rate": 1.8507142857142857e-05, "loss": 1.4458, "step": 4409000 }, { "epoch": 22.55, "learning_rate": 1.8503571428571428e-05, "loss": 1.4526, "step": 4409500 }, { "epoch": 22.56, "learning_rate": 1.85e-05, "loss": 1.4379, "step": 4410000 }, { "epoch": 22.56, "learning_rate": 1.8496428571428573e-05, "loss": 1.4513, "step": 4410500 }, { "epoch": 22.56, "learning_rate": 1.8492857142857144e-05, "loss": 1.4384, "step": 4411000 }, { "epoch": 22.56, "learning_rate": 1.8489285714285715e-05, "loss": 1.453, "step": 4411500 }, { "epoch": 22.57, "learning_rate": 1.8485714285714286e-05, "loss": 1.4416, "step": 4412000 }, { "epoch": 22.57, "learning_rate": 1.848214285714286e-05, "loss": 1.4481, "step": 4412500 }, { "epoch": 22.57, "learning_rate": 1.8478571428571428e-05, "loss": 1.4526, "step": 4413000 }, { "epoch": 22.58, "learning_rate": 1.8475000000000002e-05, "loss": 1.4532, "step": 4413500 }, { "epoch": 22.58, "learning_rate": 1.8471428571428573e-05, "loss": 1.4479, "step": 4414000 }, { "epoch": 22.58, "learning_rate": 1.8467857142857144e-05, "loss": 1.4595, "step": 4414500 }, { "epoch": 22.58, "learning_rate": 1.8464285714285714e-05, "loss": 1.4528, "step": 4415000 }, { "epoch": 22.59, "learning_rate": 1.8460714285714285e-05, "loss": 1.4515, "step": 4415500 }, { "epoch": 22.59, "learning_rate": 1.845714285714286e-05, "loss": 1.4445, "step": 4416000 }, { "epoch": 22.59, "learning_rate": 1.8453571428571427e-05, "loss": 1.4525, "step": 4416500 }, { "epoch": 22.59, "learning_rate": 1.845e-05, "loss": 1.4428, "step": 4417000 }, { "epoch": 22.6, "learning_rate": 1.8446428571428572e-05, "loss": 1.4522, "step": 4417500 }, { "epoch": 22.6, "learning_rate": 1.8442857142857146e-05, "loss": 1.4429, "step": 4418000 }, { "epoch": 22.6, "learning_rate": 1.8439285714285714e-05, "loss": 1.4586, "step": 4418500 }, { "epoch": 22.6, "learning_rate": 1.8435714285714288e-05, "loss": 1.4379, "step": 4419000 }, { "epoch": 22.61, "learning_rate": 1.843214285714286e-05, "loss": 1.4418, "step": 4419500 }, { "epoch": 22.61, "learning_rate": 1.842857142857143e-05, "loss": 1.4415, "step": 4420000 }, { "epoch": 22.61, "learning_rate": 1.8425e-05, "loss": 1.4371, "step": 4420500 }, { "epoch": 22.62, "learning_rate": 1.842142857142857e-05, "loss": 1.4489, "step": 4421000 }, { "epoch": 22.62, "learning_rate": 1.8417857142857146e-05, "loss": 1.4384, "step": 4421500 }, { "epoch": 22.62, "learning_rate": 1.8414285714285713e-05, "loss": 1.4565, "step": 4422000 }, { "epoch": 22.62, "learning_rate": 1.8410714285714287e-05, "loss": 1.4512, "step": 4422500 }, { "epoch": 22.63, "learning_rate": 1.8407142857142858e-05, "loss": 1.4411, "step": 4423000 }, { "epoch": 22.63, "learning_rate": 1.840357142857143e-05, "loss": 1.4442, "step": 4423500 }, { "epoch": 22.63, "learning_rate": 1.84e-05, "loss": 1.4477, "step": 4424000 }, { "epoch": 22.63, "learning_rate": 1.8396428571428574e-05, "loss": 1.4443, "step": 4424500 }, { "epoch": 22.64, "learning_rate": 1.8392857142857145e-05, "loss": 1.4496, "step": 4425000 }, { "epoch": 22.64, "learning_rate": 1.8389285714285716e-05, "loss": 1.4509, "step": 4425500 }, { "epoch": 22.64, "learning_rate": 1.8385714285714287e-05, "loss": 1.4462, "step": 4426000 }, { "epoch": 22.64, "learning_rate": 1.8382142857142858e-05, "loss": 1.4524, "step": 4426500 }, { "epoch": 22.65, "learning_rate": 1.837857142857143e-05, "loss": 1.455, "step": 4427000 }, { "epoch": 22.65, "learning_rate": 1.8375e-05, "loss": 1.4412, "step": 4427500 }, { "epoch": 22.65, "learning_rate": 1.8371428571428574e-05, "loss": 1.4517, "step": 4428000 }, { "epoch": 22.66, "learning_rate": 1.8367857142857145e-05, "loss": 1.4418, "step": 4428500 }, { "epoch": 22.66, "learning_rate": 1.8364285714285715e-05, "loss": 1.4419, "step": 4429000 }, { "epoch": 22.66, "learning_rate": 1.8360714285714286e-05, "loss": 1.4508, "step": 4429500 }, { "epoch": 22.66, "learning_rate": 1.835714285714286e-05, "loss": 1.436, "step": 4430000 }, { "epoch": 22.67, "learning_rate": 1.8353571428571428e-05, "loss": 1.4456, "step": 4430500 }, { "epoch": 22.67, "learning_rate": 1.8350000000000002e-05, "loss": 1.4489, "step": 4431000 }, { "epoch": 22.67, "learning_rate": 1.8346428571428573e-05, "loss": 1.445, "step": 4431500 }, { "epoch": 22.67, "learning_rate": 1.8342857142857144e-05, "loss": 1.441, "step": 4432000 }, { "epoch": 22.68, "learning_rate": 1.8339285714285715e-05, "loss": 1.4536, "step": 4432500 }, { "epoch": 22.68, "learning_rate": 1.8335714285714286e-05, "loss": 1.4448, "step": 4433000 }, { "epoch": 22.68, "learning_rate": 1.833214285714286e-05, "loss": 1.4374, "step": 4433500 }, { "epoch": 22.68, "learning_rate": 1.8328571428571427e-05, "loss": 1.4452, "step": 4434000 }, { "epoch": 22.69, "learning_rate": 1.8325e-05, "loss": 1.4387, "step": 4434500 }, { "epoch": 22.69, "learning_rate": 1.8321428571428572e-05, "loss": 1.4429, "step": 4435000 }, { "epoch": 22.69, "learning_rate": 1.8317857142857143e-05, "loss": 1.4471, "step": 4435500 }, { "epoch": 22.69, "learning_rate": 1.8314285714285714e-05, "loss": 1.4439, "step": 4436000 }, { "epoch": 22.7, "learning_rate": 1.8310714285714285e-05, "loss": 1.4375, "step": 4436500 }, { "epoch": 22.7, "learning_rate": 1.830714285714286e-05, "loss": 1.4471, "step": 4437000 }, { "epoch": 22.7, "learning_rate": 1.8303571428571427e-05, "loss": 1.4459, "step": 4437500 }, { "epoch": 22.71, "learning_rate": 1.83e-05, "loss": 1.4419, "step": 4438000 }, { "epoch": 22.71, "learning_rate": 1.8296428571428572e-05, "loss": 1.4371, "step": 4438500 }, { "epoch": 22.71, "learning_rate": 1.8292857142857146e-05, "loss": 1.4496, "step": 4439000 }, { "epoch": 22.71, "learning_rate": 1.8289285714285714e-05, "loss": 1.4487, "step": 4439500 }, { "epoch": 22.72, "learning_rate": 1.8285714285714288e-05, "loss": 1.4452, "step": 4440000 }, { "epoch": 22.72, "learning_rate": 1.828214285714286e-05, "loss": 1.4436, "step": 4440500 }, { "epoch": 22.72, "learning_rate": 1.827857142857143e-05, "loss": 1.4543, "step": 4441000 }, { "epoch": 22.72, "learning_rate": 1.8275e-05, "loss": 1.4442, "step": 4441500 }, { "epoch": 22.73, "learning_rate": 1.827142857142857e-05, "loss": 1.4504, "step": 4442000 }, { "epoch": 22.73, "learning_rate": 1.8267857142857145e-05, "loss": 1.4474, "step": 4442500 }, { "epoch": 22.73, "learning_rate": 1.8264285714285713e-05, "loss": 1.4479, "step": 4443000 }, { "epoch": 22.73, "learning_rate": 1.8260714285714287e-05, "loss": 1.4434, "step": 4443500 }, { "epoch": 22.74, "learning_rate": 1.8257142857142858e-05, "loss": 1.4447, "step": 4444000 }, { "epoch": 22.74, "learning_rate": 1.825357142857143e-05, "loss": 1.4446, "step": 4444500 }, { "epoch": 22.74, "learning_rate": 1.825e-05, "loss": 1.4338, "step": 4445000 }, { "epoch": 22.75, "learning_rate": 1.8246428571428574e-05, "loss": 1.4405, "step": 4445500 }, { "epoch": 22.75, "learning_rate": 1.8242857142857145e-05, "loss": 1.4424, "step": 4446000 }, { "epoch": 22.75, "learning_rate": 1.8239285714285716e-05, "loss": 1.4532, "step": 4446500 }, { "epoch": 22.75, "learning_rate": 1.8235714285714287e-05, "loss": 1.4401, "step": 4447000 }, { "epoch": 22.76, "learning_rate": 1.8232142857142857e-05, "loss": 1.4488, "step": 4447500 }, { "epoch": 22.76, "learning_rate": 1.8228571428571428e-05, "loss": 1.4407, "step": 4448000 }, { "epoch": 22.76, "learning_rate": 1.8225e-05, "loss": 1.4441, "step": 4448500 }, { "epoch": 22.76, "learning_rate": 1.8221428571428573e-05, "loss": 1.4525, "step": 4449000 }, { "epoch": 22.77, "learning_rate": 1.8217857142857144e-05, "loss": 1.4419, "step": 4449500 }, { "epoch": 22.77, "learning_rate": 1.8214285714285715e-05, "loss": 1.4619, "step": 4450000 }, { "epoch": 22.77, "learning_rate": 1.8210714285714286e-05, "loss": 1.4568, "step": 4450500 }, { "epoch": 22.77, "learning_rate": 1.820714285714286e-05, "loss": 1.4446, "step": 4451000 }, { "epoch": 22.78, "learning_rate": 1.8203571428571428e-05, "loss": 1.4416, "step": 4451500 }, { "epoch": 22.78, "learning_rate": 1.8200000000000002e-05, "loss": 1.4424, "step": 4452000 }, { "epoch": 22.78, "learning_rate": 1.8196428571428573e-05, "loss": 1.4414, "step": 4452500 }, { "epoch": 22.79, "learning_rate": 1.8192857142857144e-05, "loss": 1.4525, "step": 4453000 }, { "epoch": 22.79, "learning_rate": 1.8189285714285715e-05, "loss": 1.4409, "step": 4453500 }, { "epoch": 22.79, "learning_rate": 1.8185714285714285e-05, "loss": 1.4418, "step": 4454000 }, { "epoch": 22.79, "learning_rate": 1.818214285714286e-05, "loss": 1.4531, "step": 4454500 }, { "epoch": 22.8, "learning_rate": 1.8178571428571427e-05, "loss": 1.4482, "step": 4455000 }, { "epoch": 22.8, "learning_rate": 1.8175e-05, "loss": 1.4468, "step": 4455500 }, { "epoch": 22.8, "learning_rate": 1.8171428571428572e-05, "loss": 1.4481, "step": 4456000 }, { "epoch": 22.8, "learning_rate": 1.8167857142857146e-05, "loss": 1.4451, "step": 4456500 }, { "epoch": 22.81, "learning_rate": 1.8164285714285714e-05, "loss": 1.4469, "step": 4457000 }, { "epoch": 22.81, "learning_rate": 1.8160714285714288e-05, "loss": 1.4484, "step": 4457500 }, { "epoch": 22.81, "learning_rate": 1.815714285714286e-05, "loss": 1.4724, "step": 4458000 }, { "epoch": 22.81, "learning_rate": 1.8153571428571426e-05, "loss": 1.4508, "step": 4458500 }, { "epoch": 22.82, "learning_rate": 1.815e-05, "loss": 1.4583, "step": 4459000 }, { "epoch": 22.82, "learning_rate": 1.814642857142857e-05, "loss": 1.4467, "step": 4459500 }, { "epoch": 22.82, "learning_rate": 1.8142857142857146e-05, "loss": 1.4469, "step": 4460000 }, { "epoch": 22.83, "learning_rate": 1.8139285714285713e-05, "loss": 1.4441, "step": 4460500 }, { "epoch": 22.83, "learning_rate": 1.8135714285714288e-05, "loss": 1.4508, "step": 4461000 }, { "epoch": 22.83, "learning_rate": 1.813214285714286e-05, "loss": 1.4507, "step": 4461500 }, { "epoch": 22.83, "learning_rate": 1.812857142857143e-05, "loss": 1.4481, "step": 4462000 }, { "epoch": 22.84, "learning_rate": 1.8125e-05, "loss": 1.4362, "step": 4462500 }, { "epoch": 22.84, "learning_rate": 1.812142857142857e-05, "loss": 1.4498, "step": 4463000 }, { "epoch": 22.84, "learning_rate": 1.8117857142857145e-05, "loss": 1.4404, "step": 4463500 }, { "epoch": 22.84, "learning_rate": 1.8114285714285713e-05, "loss": 1.4453, "step": 4464000 }, { "epoch": 22.85, "learning_rate": 1.8110714285714287e-05, "loss": 1.4484, "step": 4464500 }, { "epoch": 22.85, "learning_rate": 1.8107142857142858e-05, "loss": 1.4513, "step": 4465000 }, { "epoch": 22.85, "learning_rate": 1.810357142857143e-05, "loss": 1.4413, "step": 4465500 }, { "epoch": 22.85, "learning_rate": 1.81e-05, "loss": 1.4445, "step": 4466000 }, { "epoch": 22.86, "learning_rate": 1.8096428571428574e-05, "loss": 1.451, "step": 4466500 }, { "epoch": 22.86, "learning_rate": 1.8092857142857145e-05, "loss": 1.4474, "step": 4467000 }, { "epoch": 22.86, "learning_rate": 1.8089285714285715e-05, "loss": 1.4458, "step": 4467500 }, { "epoch": 22.87, "learning_rate": 1.8085714285714286e-05, "loss": 1.4448, "step": 4468000 }, { "epoch": 22.87, "learning_rate": 1.8082142857142857e-05, "loss": 1.4491, "step": 4468500 }, { "epoch": 22.87, "learning_rate": 1.807857142857143e-05, "loss": 1.4421, "step": 4469000 }, { "epoch": 22.87, "learning_rate": 1.8075e-05, "loss": 1.4449, "step": 4469500 }, { "epoch": 22.88, "learning_rate": 1.8071428571428573e-05, "loss": 1.4523, "step": 4470000 }, { "epoch": 22.88, "learning_rate": 1.8067857142857144e-05, "loss": 1.4477, "step": 4470500 }, { "epoch": 22.88, "learning_rate": 1.8064285714285715e-05, "loss": 1.4412, "step": 4471000 }, { "epoch": 22.88, "learning_rate": 1.8060714285714286e-05, "loss": 1.4436, "step": 4471500 }, { "epoch": 22.89, "learning_rate": 1.805714285714286e-05, "loss": 1.4474, "step": 4472000 }, { "epoch": 22.89, "learning_rate": 1.805357142857143e-05, "loss": 1.4425, "step": 4472500 }, { "epoch": 22.89, "learning_rate": 1.805e-05, "loss": 1.4372, "step": 4473000 }, { "epoch": 22.89, "learning_rate": 1.8046428571428573e-05, "loss": 1.4467, "step": 4473500 }, { "epoch": 22.9, "learning_rate": 1.8042857142857143e-05, "loss": 1.4457, "step": 4474000 }, { "epoch": 22.9, "learning_rate": 1.8039285714285714e-05, "loss": 1.4504, "step": 4474500 }, { "epoch": 22.9, "learning_rate": 1.8035714285714285e-05, "loss": 1.4406, "step": 4475000 }, { "epoch": 22.91, "learning_rate": 1.803214285714286e-05, "loss": 1.4523, "step": 4475500 }, { "epoch": 22.91, "learning_rate": 1.802857142857143e-05, "loss": 1.4375, "step": 4476000 }, { "epoch": 22.91, "learning_rate": 1.8025e-05, "loss": 1.4447, "step": 4476500 }, { "epoch": 22.91, "learning_rate": 1.8021428571428572e-05, "loss": 1.4515, "step": 4477000 }, { "epoch": 22.92, "learning_rate": 1.8017857142857146e-05, "loss": 1.441, "step": 4477500 }, { "epoch": 22.92, "learning_rate": 1.8014285714285714e-05, "loss": 1.456, "step": 4478000 }, { "epoch": 22.92, "learning_rate": 1.8010714285714288e-05, "loss": 1.4468, "step": 4478500 }, { "epoch": 22.92, "learning_rate": 1.800714285714286e-05, "loss": 1.4423, "step": 4479000 }, { "epoch": 22.93, "learning_rate": 1.800357142857143e-05, "loss": 1.4418, "step": 4479500 }, { "epoch": 22.93, "learning_rate": 1.8e-05, "loss": 1.4545, "step": 4480000 }, { "epoch": 22.93, "learning_rate": 1.799642857142857e-05, "loss": 1.4433, "step": 4480500 }, { "epoch": 22.93, "learning_rate": 1.7992857142857146e-05, "loss": 1.4464, "step": 4481000 }, { "epoch": 22.94, "learning_rate": 1.7989285714285713e-05, "loss": 1.4402, "step": 4481500 }, { "epoch": 22.94, "learning_rate": 1.7985714285714287e-05, "loss": 1.4554, "step": 4482000 }, { "epoch": 22.94, "learning_rate": 1.7982142857142858e-05, "loss": 1.4651, "step": 4482500 }, { "epoch": 22.95, "learning_rate": 1.7978571428571432e-05, "loss": 1.4576, "step": 4483000 }, { "epoch": 22.95, "learning_rate": 1.7975e-05, "loss": 1.4558, "step": 4483500 }, { "epoch": 22.95, "learning_rate": 1.797142857142857e-05, "loss": 1.4398, "step": 4484000 }, { "epoch": 22.95, "learning_rate": 1.7967857142857145e-05, "loss": 1.452, "step": 4484500 }, { "epoch": 22.96, "learning_rate": 1.7964285714285712e-05, "loss": 1.4462, "step": 4485000 }, { "epoch": 22.96, "learning_rate": 1.7960714285714287e-05, "loss": 1.4507, "step": 4485500 }, { "epoch": 22.96, "learning_rate": 1.7957142857142858e-05, "loss": 1.4437, "step": 4486000 }, { "epoch": 22.96, "learning_rate": 1.7953571428571432e-05, "loss": 1.4473, "step": 4486500 }, { "epoch": 22.97, "learning_rate": 1.795e-05, "loss": 1.4534, "step": 4487000 }, { "epoch": 22.97, "learning_rate": 1.7946428571428573e-05, "loss": 1.4451, "step": 4487500 }, { "epoch": 22.97, "learning_rate": 1.7942857142857144e-05, "loss": 1.4366, "step": 4488000 }, { "epoch": 22.97, "learning_rate": 1.7939285714285715e-05, "loss": 1.4453, "step": 4488500 }, { "epoch": 22.98, "learning_rate": 1.7935714285714286e-05, "loss": 1.4372, "step": 4489000 }, { "epoch": 22.98, "learning_rate": 1.7932142857142857e-05, "loss": 1.4451, "step": 4489500 }, { "epoch": 22.98, "learning_rate": 1.792857142857143e-05, "loss": 1.4479, "step": 4490000 }, { "epoch": 22.99, "learning_rate": 1.7925e-05, "loss": 1.4406, "step": 4490500 }, { "epoch": 22.99, "learning_rate": 1.7921428571428573e-05, "loss": 1.445, "step": 4491000 }, { "epoch": 22.99, "learning_rate": 1.7917857142857144e-05, "loss": 1.4508, "step": 4491500 }, { "epoch": 22.99, "learning_rate": 1.7914285714285715e-05, "loss": 1.4549, "step": 4492000 }, { "epoch": 23.0, "learning_rate": 1.7910714285714285e-05, "loss": 1.446, "step": 4492500 }, { "epoch": 23.0, "learning_rate": 1.790714285714286e-05, "loss": 1.4509, "step": 4493000 }, { "epoch": 23.0, "learning_rate": 1.790357142857143e-05, "loss": 1.4451, "step": 4493500 }, { "epoch": 23.0, "learning_rate": 1.79e-05, "loss": 1.4411, "step": 4494000 }, { "epoch": 23.01, "learning_rate": 1.7896428571428572e-05, "loss": 1.4354, "step": 4494500 }, { "epoch": 23.01, "learning_rate": 1.7892857142857143e-05, "loss": 1.4499, "step": 4495000 }, { "epoch": 23.01, "learning_rate": 1.7889285714285714e-05, "loss": 1.4395, "step": 4495500 }, { "epoch": 23.01, "learning_rate": 1.7885714285714285e-05, "loss": 1.4468, "step": 4496000 }, { "epoch": 23.02, "learning_rate": 1.788214285714286e-05, "loss": 1.4435, "step": 4496500 }, { "epoch": 23.02, "learning_rate": 1.787857142857143e-05, "loss": 1.4399, "step": 4497000 }, { "epoch": 23.02, "learning_rate": 1.7875e-05, "loss": 1.4499, "step": 4497500 }, { "epoch": 23.03, "learning_rate": 1.787142857142857e-05, "loss": 1.4418, "step": 4498000 }, { "epoch": 23.03, "learning_rate": 1.7867857142857146e-05, "loss": 1.4306, "step": 4498500 }, { "epoch": 23.03, "learning_rate": 1.7864285714285713e-05, "loss": 1.432, "step": 4499000 }, { "epoch": 23.03, "learning_rate": 1.7860714285714288e-05, "loss": 1.4323, "step": 4499500 }, { "epoch": 23.04, "learning_rate": 1.785714285714286e-05, "loss": 1.4415, "step": 4500000 }, { "epoch": 23.04, "learning_rate": 1.785357142857143e-05, "loss": 1.4382, "step": 4500500 }, { "epoch": 23.04, "learning_rate": 1.785e-05, "loss": 1.4421, "step": 4501000 }, { "epoch": 23.04, "learning_rate": 1.784642857142857e-05, "loss": 1.4368, "step": 4501500 }, { "epoch": 23.05, "learning_rate": 1.7842857142857145e-05, "loss": 1.4629, "step": 4502000 }, { "epoch": 23.05, "learning_rate": 1.7839285714285713e-05, "loss": 1.4429, "step": 4502500 }, { "epoch": 23.05, "learning_rate": 1.7835714285714287e-05, "loss": 1.4485, "step": 4503000 }, { "epoch": 23.05, "learning_rate": 1.7832142857142858e-05, "loss": 1.4394, "step": 4503500 }, { "epoch": 23.06, "learning_rate": 1.7828571428571432e-05, "loss": 1.4552, "step": 4504000 }, { "epoch": 23.06, "learning_rate": 1.7825e-05, "loss": 1.4595, "step": 4504500 }, { "epoch": 23.06, "learning_rate": 1.7821428571428574e-05, "loss": 1.4458, "step": 4505000 }, { "epoch": 23.07, "learning_rate": 1.7817857142857145e-05, "loss": 1.4425, "step": 4505500 }, { "epoch": 23.07, "learning_rate": 1.7814285714285716e-05, "loss": 1.4454, "step": 4506000 }, { "epoch": 23.07, "learning_rate": 1.7810714285714286e-05, "loss": 1.4482, "step": 4506500 }, { "epoch": 23.07, "learning_rate": 1.7807142857142857e-05, "loss": 1.4381, "step": 4507000 }, { "epoch": 23.08, "learning_rate": 1.780357142857143e-05, "loss": 1.4452, "step": 4507500 }, { "epoch": 23.08, "learning_rate": 1.78e-05, "loss": 1.4585, "step": 4508000 }, { "epoch": 23.08, "learning_rate": 1.7796428571428573e-05, "loss": 1.4438, "step": 4508500 }, { "epoch": 23.08, "learning_rate": 1.7792857142857144e-05, "loss": 1.4429, "step": 4509000 }, { "epoch": 23.09, "learning_rate": 1.7789285714285715e-05, "loss": 1.4382, "step": 4509500 }, { "epoch": 23.09, "learning_rate": 1.7785714285714286e-05, "loss": 1.4481, "step": 4510000 }, { "epoch": 23.09, "learning_rate": 1.7782142857142857e-05, "loss": 1.435, "step": 4510500 }, { "epoch": 23.09, "learning_rate": 1.777857142857143e-05, "loss": 1.4582, "step": 4511000 }, { "epoch": 23.1, "learning_rate": 1.7775e-05, "loss": 1.4448, "step": 4511500 }, { "epoch": 23.1, "learning_rate": 1.7771428571428573e-05, "loss": 1.4444, "step": 4512000 }, { "epoch": 23.1, "learning_rate": 1.7767857142857143e-05, "loss": 1.449, "step": 4512500 }, { "epoch": 23.1, "learning_rate": 1.7764285714285714e-05, "loss": 1.437, "step": 4513000 }, { "epoch": 23.11, "learning_rate": 1.7760714285714285e-05, "loss": 1.4391, "step": 4513500 }, { "epoch": 23.11, "learning_rate": 1.775714285714286e-05, "loss": 1.4514, "step": 4514000 }, { "epoch": 23.11, "learning_rate": 1.775357142857143e-05, "loss": 1.4467, "step": 4514500 }, { "epoch": 23.12, "learning_rate": 1.775e-05, "loss": 1.4424, "step": 4515000 }, { "epoch": 23.12, "learning_rate": 1.7746428571428572e-05, "loss": 1.4289, "step": 4515500 }, { "epoch": 23.12, "learning_rate": 1.7742857142857143e-05, "loss": 1.4404, "step": 4516000 }, { "epoch": 23.12, "learning_rate": 1.7739285714285714e-05, "loss": 1.454, "step": 4516500 }, { "epoch": 23.13, "learning_rate": 1.7735714285714285e-05, "loss": 1.431, "step": 4517000 }, { "epoch": 23.13, "learning_rate": 1.773214285714286e-05, "loss": 1.4388, "step": 4517500 }, { "epoch": 23.13, "learning_rate": 1.772857142857143e-05, "loss": 1.4484, "step": 4518000 }, { "epoch": 23.13, "learning_rate": 1.7725e-05, "loss": 1.446, "step": 4518500 }, { "epoch": 23.14, "learning_rate": 1.772142857142857e-05, "loss": 1.4425, "step": 4519000 }, { "epoch": 23.14, "learning_rate": 1.7717857142857146e-05, "loss": 1.4383, "step": 4519500 }, { "epoch": 23.14, "learning_rate": 1.7714285714285713e-05, "loss": 1.4411, "step": 4520000 }, { "epoch": 23.14, "learning_rate": 1.7710714285714287e-05, "loss": 1.4458, "step": 4520500 }, { "epoch": 23.15, "learning_rate": 1.7707142857142858e-05, "loss": 1.4445, "step": 4521000 }, { "epoch": 23.15, "learning_rate": 1.770357142857143e-05, "loss": 1.4515, "step": 4521500 }, { "epoch": 23.15, "learning_rate": 1.77e-05, "loss": 1.451, "step": 4522000 }, { "epoch": 23.16, "learning_rate": 1.769642857142857e-05, "loss": 1.4527, "step": 4522500 }, { "epoch": 23.16, "learning_rate": 1.7692857142857145e-05, "loss": 1.4416, "step": 4523000 }, { "epoch": 23.16, "learning_rate": 1.7689285714285713e-05, "loss": 1.4452, "step": 4523500 }, { "epoch": 23.16, "learning_rate": 1.7685714285714287e-05, "loss": 1.4378, "step": 4524000 }, { "epoch": 23.17, "learning_rate": 1.7682142857142858e-05, "loss": 1.4415, "step": 4524500 }, { "epoch": 23.17, "learning_rate": 1.7678571428571432e-05, "loss": 1.4409, "step": 4525000 }, { "epoch": 23.17, "learning_rate": 1.7675e-05, "loss": 1.4542, "step": 4525500 }, { "epoch": 23.17, "learning_rate": 1.7671428571428574e-05, "loss": 1.4446, "step": 4526000 }, { "epoch": 23.18, "learning_rate": 1.7667857142857144e-05, "loss": 1.4599, "step": 4526500 }, { "epoch": 23.18, "learning_rate": 1.7664285714285715e-05, "loss": 1.4384, "step": 4527000 }, { "epoch": 23.18, "learning_rate": 1.7660714285714286e-05, "loss": 1.4432, "step": 4527500 }, { "epoch": 23.18, "learning_rate": 1.7657142857142857e-05, "loss": 1.4384, "step": 4528000 }, { "epoch": 23.19, "learning_rate": 1.765357142857143e-05, "loss": 1.4401, "step": 4528500 }, { "epoch": 23.19, "learning_rate": 1.765e-05, "loss": 1.4533, "step": 4529000 }, { "epoch": 23.19, "learning_rate": 1.7646428571428573e-05, "loss": 1.4475, "step": 4529500 }, { "epoch": 23.2, "learning_rate": 1.7642857142857144e-05, "loss": 1.4486, "step": 4530000 }, { "epoch": 23.2, "learning_rate": 1.7639285714285715e-05, "loss": 1.441, "step": 4530500 }, { "epoch": 23.2, "learning_rate": 1.7635714285714286e-05, "loss": 1.4489, "step": 4531000 }, { "epoch": 23.2, "learning_rate": 1.763214285714286e-05, "loss": 1.4462, "step": 4531500 }, { "epoch": 23.21, "learning_rate": 1.762857142857143e-05, "loss": 1.4406, "step": 4532000 }, { "epoch": 23.21, "learning_rate": 1.7625e-05, "loss": 1.4421, "step": 4532500 }, { "epoch": 23.21, "learning_rate": 1.7621428571428572e-05, "loss": 1.4403, "step": 4533000 }, { "epoch": 23.21, "learning_rate": 1.7617857142857143e-05, "loss": 1.4394, "step": 4533500 }, { "epoch": 23.22, "learning_rate": 1.7614285714285717e-05, "loss": 1.4406, "step": 4534000 }, { "epoch": 23.22, "learning_rate": 1.7610714285714285e-05, "loss": 1.4456, "step": 4534500 }, { "epoch": 23.22, "learning_rate": 1.760714285714286e-05, "loss": 1.439, "step": 4535000 }, { "epoch": 23.22, "learning_rate": 1.760357142857143e-05, "loss": 1.4501, "step": 4535500 }, { "epoch": 23.23, "learning_rate": 1.76e-05, "loss": 1.4355, "step": 4536000 }, { "epoch": 23.23, "learning_rate": 1.7596428571428572e-05, "loss": 1.4532, "step": 4536500 }, { "epoch": 23.23, "learning_rate": 1.7592857142857143e-05, "loss": 1.4547, "step": 4537000 }, { "epoch": 23.24, "learning_rate": 1.7589285714285717e-05, "loss": 1.454, "step": 4537500 }, { "epoch": 23.24, "learning_rate": 1.7585714285714284e-05, "loss": 1.4402, "step": 4538000 }, { "epoch": 23.24, "learning_rate": 1.758214285714286e-05, "loss": 1.4411, "step": 4538500 }, { "epoch": 23.24, "learning_rate": 1.757857142857143e-05, "loss": 1.4344, "step": 4539000 }, { "epoch": 23.25, "learning_rate": 1.7575e-05, "loss": 1.4489, "step": 4539500 }, { "epoch": 23.25, "learning_rate": 1.757142857142857e-05, "loss": 1.4416, "step": 4540000 }, { "epoch": 23.25, "learning_rate": 1.7567857142857145e-05, "loss": 1.4479, "step": 4540500 }, { "epoch": 23.25, "learning_rate": 1.7564285714285716e-05, "loss": 1.4525, "step": 4541000 }, { "epoch": 23.26, "learning_rate": 1.7560714285714287e-05, "loss": 1.4576, "step": 4541500 }, { "epoch": 23.26, "learning_rate": 1.7557142857142858e-05, "loss": 1.501, "step": 4542000 }, { "epoch": 23.26, "learning_rate": 1.755357142857143e-05, "loss": 1.4464, "step": 4542500 }, { "epoch": 23.26, "learning_rate": 1.755e-05, "loss": 1.4499, "step": 4543000 }, { "epoch": 23.27, "learning_rate": 1.754642857142857e-05, "loss": 1.4376, "step": 4543500 }, { "epoch": 23.27, "learning_rate": 1.7542857142857145e-05, "loss": 1.4366, "step": 4544000 }, { "epoch": 23.27, "learning_rate": 1.7539285714285716e-05, "loss": 1.4494, "step": 4544500 }, { "epoch": 23.28, "learning_rate": 1.7535714285714287e-05, "loss": 1.4461, "step": 4545000 }, { "epoch": 23.28, "learning_rate": 1.7532142857142857e-05, "loss": 1.4507, "step": 4545500 }, { "epoch": 23.28, "learning_rate": 1.752857142857143e-05, "loss": 1.442, "step": 4546000 }, { "epoch": 23.28, "learning_rate": 1.7525e-05, "loss": 1.4504, "step": 4546500 }, { "epoch": 23.29, "learning_rate": 1.7521428571428573e-05, "loss": 1.4362, "step": 4547000 }, { "epoch": 23.29, "learning_rate": 1.7517857142857144e-05, "loss": 1.4316, "step": 4547500 }, { "epoch": 23.29, "learning_rate": 1.7514285714285715e-05, "loss": 1.4419, "step": 4548000 }, { "epoch": 23.29, "learning_rate": 1.7510714285714286e-05, "loss": 1.4362, "step": 4548500 }, { "epoch": 23.3, "learning_rate": 1.7507142857142857e-05, "loss": 1.4345, "step": 4549000 }, { "epoch": 23.3, "learning_rate": 1.750357142857143e-05, "loss": 1.4427, "step": 4549500 }, { "epoch": 23.3, "learning_rate": 1.75e-05, "loss": 1.4429, "step": 4550000 }, { "epoch": 23.3, "learning_rate": 1.7496428571428573e-05, "loss": 1.4493, "step": 4550500 }, { "epoch": 23.31, "learning_rate": 1.7492857142857144e-05, "loss": 1.4376, "step": 4551000 }, { "epoch": 23.31, "learning_rate": 1.7489285714285718e-05, "loss": 1.4425, "step": 4551500 }, { "epoch": 23.31, "learning_rate": 1.7485714285714285e-05, "loss": 1.4363, "step": 4552000 }, { "epoch": 23.32, "learning_rate": 1.748214285714286e-05, "loss": 1.4338, "step": 4552500 }, { "epoch": 23.32, "learning_rate": 1.747857142857143e-05, "loss": 1.4388, "step": 4553000 }, { "epoch": 23.32, "learning_rate": 1.7475e-05, "loss": 1.4393, "step": 4553500 }, { "epoch": 23.32, "learning_rate": 1.7471428571428572e-05, "loss": 1.4389, "step": 4554000 }, { "epoch": 23.33, "learning_rate": 1.7467857142857143e-05, "loss": 1.4357, "step": 4554500 }, { "epoch": 23.33, "learning_rate": 1.7464285714285717e-05, "loss": 1.4368, "step": 4555000 }, { "epoch": 23.33, "learning_rate": 1.7460714285714285e-05, "loss": 1.439, "step": 4555500 }, { "epoch": 23.33, "learning_rate": 1.745714285714286e-05, "loss": 1.4522, "step": 4556000 }, { "epoch": 23.34, "learning_rate": 1.745357142857143e-05, "loss": 1.4425, "step": 4556500 }, { "epoch": 23.34, "learning_rate": 1.745e-05, "loss": 1.4469, "step": 4557000 }, { "epoch": 23.34, "learning_rate": 1.744642857142857e-05, "loss": 1.4464, "step": 4557500 }, { "epoch": 23.34, "learning_rate": 1.7442857142857146e-05, "loss": 1.4426, "step": 4558000 }, { "epoch": 23.35, "learning_rate": 1.7439285714285717e-05, "loss": 1.4408, "step": 4558500 }, { "epoch": 23.35, "learning_rate": 1.7435714285714287e-05, "loss": 1.4444, "step": 4559000 }, { "epoch": 23.35, "learning_rate": 1.743214285714286e-05, "loss": 1.4428, "step": 4559500 }, { "epoch": 23.36, "learning_rate": 1.742857142857143e-05, "loss": 1.4329, "step": 4560000 }, { "epoch": 23.36, "learning_rate": 1.7425e-05, "loss": 1.4385, "step": 4560500 }, { "epoch": 23.36, "learning_rate": 1.742142857142857e-05, "loss": 1.4383, "step": 4561000 }, { "epoch": 23.36, "learning_rate": 1.7417857142857145e-05, "loss": 1.4286, "step": 4561500 }, { "epoch": 23.37, "learning_rate": 1.7414285714285716e-05, "loss": 1.4395, "step": 4562000 }, { "epoch": 23.37, "learning_rate": 1.7410714285714287e-05, "loss": 1.4402, "step": 4562500 }, { "epoch": 23.37, "learning_rate": 1.7407142857142858e-05, "loss": 1.4307, "step": 4563000 }, { "epoch": 23.37, "learning_rate": 1.740357142857143e-05, "loss": 1.4474, "step": 4563500 }, { "epoch": 23.38, "learning_rate": 1.74e-05, "loss": 1.4321, "step": 4564000 }, { "epoch": 23.38, "learning_rate": 1.739642857142857e-05, "loss": 1.4363, "step": 4564500 }, { "epoch": 23.38, "learning_rate": 1.7392857142857145e-05, "loss": 1.4334, "step": 4565000 }, { "epoch": 23.38, "learning_rate": 1.7389285714285715e-05, "loss": 1.4509, "step": 4565500 }, { "epoch": 23.39, "learning_rate": 1.7385714285714286e-05, "loss": 1.4464, "step": 4566000 }, { "epoch": 23.39, "learning_rate": 1.7382142857142857e-05, "loss": 1.4339, "step": 4566500 }, { "epoch": 23.39, "learning_rate": 1.737857142857143e-05, "loss": 1.4487, "step": 4567000 }, { "epoch": 23.4, "learning_rate": 1.7375e-05, "loss": 1.441, "step": 4567500 }, { "epoch": 23.4, "learning_rate": 1.7371428571428573e-05, "loss": 1.4485, "step": 4568000 }, { "epoch": 23.4, "learning_rate": 1.7367857142857144e-05, "loss": 1.4327, "step": 4568500 }, { "epoch": 23.4, "learning_rate": 1.7364285714285715e-05, "loss": 1.4356, "step": 4569000 }, { "epoch": 23.41, "learning_rate": 1.7360714285714286e-05, "loss": 1.4246, "step": 4569500 }, { "epoch": 23.41, "learning_rate": 1.7357142857142856e-05, "loss": 1.4468, "step": 4570000 }, { "epoch": 23.41, "learning_rate": 1.735357142857143e-05, "loss": 1.4325, "step": 4570500 }, { "epoch": 23.41, "learning_rate": 1.7349999999999998e-05, "loss": 1.4391, "step": 4571000 }, { "epoch": 23.42, "learning_rate": 1.7346428571428572e-05, "loss": 1.4419, "step": 4571500 }, { "epoch": 23.42, "learning_rate": 1.7342857142857143e-05, "loss": 1.4333, "step": 4572000 }, { "epoch": 23.42, "learning_rate": 1.7339285714285718e-05, "loss": 1.4367, "step": 4572500 }, { "epoch": 23.42, "learning_rate": 1.7335714285714285e-05, "loss": 1.4465, "step": 4573000 }, { "epoch": 23.43, "learning_rate": 1.733214285714286e-05, "loss": 1.4359, "step": 4573500 }, { "epoch": 23.43, "learning_rate": 1.732857142857143e-05, "loss": 1.4317, "step": 4574000 }, { "epoch": 23.43, "learning_rate": 1.7325e-05, "loss": 1.4294, "step": 4574500 }, { "epoch": 23.44, "learning_rate": 1.7321428571428572e-05, "loss": 1.4413, "step": 4575000 }, { "epoch": 23.44, "learning_rate": 1.7317857142857143e-05, "loss": 1.4295, "step": 4575500 }, { "epoch": 23.44, "learning_rate": 1.7314285714285717e-05, "loss": 1.4451, "step": 4576000 }, { "epoch": 23.44, "learning_rate": 1.7310714285714284e-05, "loss": 1.4299, "step": 4576500 }, { "epoch": 23.45, "learning_rate": 1.730714285714286e-05, "loss": 1.4417, "step": 4577000 }, { "epoch": 23.45, "learning_rate": 1.730357142857143e-05, "loss": 1.4326, "step": 4577500 }, { "epoch": 23.45, "learning_rate": 1.73e-05, "loss": 1.4352, "step": 4578000 }, { "epoch": 23.45, "learning_rate": 1.729642857142857e-05, "loss": 1.4385, "step": 4578500 }, { "epoch": 23.46, "learning_rate": 1.7292857142857145e-05, "loss": 1.4395, "step": 4579000 }, { "epoch": 23.46, "learning_rate": 1.7289285714285716e-05, "loss": 1.4405, "step": 4579500 }, { "epoch": 23.46, "learning_rate": 1.7285714285714287e-05, "loss": 1.4382, "step": 4580000 }, { "epoch": 23.46, "learning_rate": 1.7282142857142858e-05, "loss": 1.4443, "step": 4580500 }, { "epoch": 23.47, "learning_rate": 1.727857142857143e-05, "loss": 1.4409, "step": 4581000 }, { "epoch": 23.47, "learning_rate": 1.7275e-05, "loss": 1.4433, "step": 4581500 }, { "epoch": 23.47, "learning_rate": 1.727142857142857e-05, "loss": 1.4394, "step": 4582000 }, { "epoch": 23.48, "learning_rate": 1.7267857142857145e-05, "loss": 1.4398, "step": 4582500 }, { "epoch": 23.48, "learning_rate": 1.7264285714285716e-05, "loss": 1.4355, "step": 4583000 }, { "epoch": 23.48, "learning_rate": 1.7260714285714287e-05, "loss": 1.4424, "step": 4583500 }, { "epoch": 23.48, "learning_rate": 1.7257142857142857e-05, "loss": 1.4383, "step": 4584000 }, { "epoch": 23.49, "learning_rate": 1.7253571428571432e-05, "loss": 1.4339, "step": 4584500 }, { "epoch": 23.49, "learning_rate": 1.725e-05, "loss": 1.4408, "step": 4585000 }, { "epoch": 23.49, "learning_rate": 1.724642857142857e-05, "loss": 1.4377, "step": 4585500 }, { "epoch": 23.49, "learning_rate": 1.7242857142857144e-05, "loss": 1.4287, "step": 4586000 }, { "epoch": 23.5, "learning_rate": 1.7239285714285715e-05, "loss": 1.4389, "step": 4586500 }, { "epoch": 23.5, "learning_rate": 1.7235714285714286e-05, "loss": 1.44, "step": 4587000 }, { "epoch": 23.5, "learning_rate": 1.7232142857142857e-05, "loss": 1.442, "step": 4587500 }, { "epoch": 23.5, "learning_rate": 1.722857142857143e-05, "loss": 1.4304, "step": 4588000 }, { "epoch": 23.51, "learning_rate": 1.7225e-05, "loss": 1.4433, "step": 4588500 }, { "epoch": 23.51, "learning_rate": 1.7221428571428573e-05, "loss": 1.4454, "step": 4589000 }, { "epoch": 23.51, "learning_rate": 1.7217857142857144e-05, "loss": 1.4367, "step": 4589500 }, { "epoch": 23.51, "learning_rate": 1.7214285714285715e-05, "loss": 1.4427, "step": 4590000 }, { "epoch": 23.52, "learning_rate": 1.7210714285714285e-05, "loss": 1.4394, "step": 4590500 }, { "epoch": 23.52, "learning_rate": 1.7207142857142856e-05, "loss": 1.4349, "step": 4591000 }, { "epoch": 23.52, "learning_rate": 1.720357142857143e-05, "loss": 1.4393, "step": 4591500 }, { "epoch": 23.53, "learning_rate": 1.7199999999999998e-05, "loss": 1.4394, "step": 4592000 }, { "epoch": 23.53, "learning_rate": 1.7196428571428572e-05, "loss": 1.431, "step": 4592500 }, { "epoch": 23.53, "learning_rate": 1.7192857142857143e-05, "loss": 1.4416, "step": 4593000 }, { "epoch": 23.53, "learning_rate": 1.7189285714285717e-05, "loss": 1.4384, "step": 4593500 }, { "epoch": 23.54, "learning_rate": 1.7185714285714285e-05, "loss": 1.4345, "step": 4594000 }, { "epoch": 23.54, "learning_rate": 1.718214285714286e-05, "loss": 1.4351, "step": 4594500 }, { "epoch": 23.54, "learning_rate": 1.717857142857143e-05, "loss": 1.4471, "step": 4595000 }, { "epoch": 23.54, "learning_rate": 1.7175e-05, "loss": 1.4368, "step": 4595500 }, { "epoch": 23.55, "learning_rate": 1.717142857142857e-05, "loss": 1.4317, "step": 4596000 }, { "epoch": 23.55, "learning_rate": 1.7167857142857142e-05, "loss": 1.4346, "step": 4596500 }, { "epoch": 23.55, "learning_rate": 1.7164285714285717e-05, "loss": 1.4345, "step": 4597000 }, { "epoch": 23.55, "learning_rate": 1.7160714285714284e-05, "loss": 1.4448, "step": 4597500 }, { "epoch": 23.56, "learning_rate": 1.715714285714286e-05, "loss": 1.4432, "step": 4598000 }, { "epoch": 23.56, "learning_rate": 1.715357142857143e-05, "loss": 1.4421, "step": 4598500 }, { "epoch": 23.56, "learning_rate": 1.7150000000000004e-05, "loss": 1.4418, "step": 4599000 }, { "epoch": 23.57, "learning_rate": 1.714642857142857e-05, "loss": 1.4477, "step": 4599500 }, { "epoch": 23.57, "learning_rate": 1.7142857142857145e-05, "loss": 1.4485, "step": 4600000 }, { "epoch": 23.57, "learning_rate": 1.7139285714285716e-05, "loss": 1.4533, "step": 4600500 }, { "epoch": 23.57, "learning_rate": 1.7135714285714287e-05, "loss": 1.4447, "step": 4601000 }, { "epoch": 23.58, "learning_rate": 1.7132142857142858e-05, "loss": 1.4427, "step": 4601500 }, { "epoch": 23.58, "learning_rate": 1.712857142857143e-05, "loss": 1.4231, "step": 4602000 }, { "epoch": 23.58, "learning_rate": 1.7125000000000003e-05, "loss": 1.4348, "step": 4602500 }, { "epoch": 23.58, "learning_rate": 1.712142857142857e-05, "loss": 1.4372, "step": 4603000 }, { "epoch": 23.59, "learning_rate": 1.7117857142857145e-05, "loss": 1.4439, "step": 4603500 }, { "epoch": 23.59, "learning_rate": 1.7114285714285715e-05, "loss": 1.436, "step": 4604000 }, { "epoch": 23.59, "learning_rate": 1.7110714285714286e-05, "loss": 1.4425, "step": 4604500 }, { "epoch": 23.59, "learning_rate": 1.7107142857142857e-05, "loss": 1.4408, "step": 4605000 }, { "epoch": 23.6, "learning_rate": 1.710357142857143e-05, "loss": 1.4375, "step": 4605500 }, { "epoch": 23.6, "learning_rate": 1.7100000000000002e-05, "loss": 1.4356, "step": 4606000 }, { "epoch": 23.6, "learning_rate": 1.7096428571428573e-05, "loss": 1.4332, "step": 4606500 }, { "epoch": 23.61, "learning_rate": 1.7092857142857144e-05, "loss": 1.4399, "step": 4607000 }, { "epoch": 23.61, "learning_rate": 1.7089285714285715e-05, "loss": 1.4373, "step": 4607500 }, { "epoch": 23.61, "learning_rate": 1.7085714285714286e-05, "loss": 1.4372, "step": 4608000 }, { "epoch": 23.61, "learning_rate": 1.7082142857142857e-05, "loss": 1.4479, "step": 4608500 }, { "epoch": 23.62, "learning_rate": 1.707857142857143e-05, "loss": 1.4224, "step": 4609000 }, { "epoch": 23.62, "learning_rate": 1.7075e-05, "loss": 1.433, "step": 4609500 }, { "epoch": 23.62, "learning_rate": 1.7071428571428573e-05, "loss": 1.4419, "step": 4610000 }, { "epoch": 23.62, "learning_rate": 1.7067857142857143e-05, "loss": 1.4346, "step": 4610500 }, { "epoch": 23.63, "learning_rate": 1.7064285714285718e-05, "loss": 1.4419, "step": 4611000 }, { "epoch": 23.63, "learning_rate": 1.7060714285714285e-05, "loss": 1.4443, "step": 4611500 }, { "epoch": 23.63, "learning_rate": 1.7057142857142856e-05, "loss": 1.4409, "step": 4612000 }, { "epoch": 23.63, "learning_rate": 1.705357142857143e-05, "loss": 1.4382, "step": 4612500 }, { "epoch": 23.64, "learning_rate": 1.705e-05, "loss": 1.4296, "step": 4613000 }, { "epoch": 23.64, "learning_rate": 1.7046428571428572e-05, "loss": 1.4399, "step": 4613500 }, { "epoch": 23.64, "learning_rate": 1.7042857142857143e-05, "loss": 1.4456, "step": 4614000 }, { "epoch": 23.65, "learning_rate": 1.7039285714285717e-05, "loss": 1.442, "step": 4614500 }, { "epoch": 23.65, "learning_rate": 1.7035714285714285e-05, "loss": 1.4414, "step": 4615000 }, { "epoch": 23.65, "learning_rate": 1.703214285714286e-05, "loss": 1.4436, "step": 4615500 }, { "epoch": 23.65, "learning_rate": 1.702857142857143e-05, "loss": 1.4442, "step": 4616000 }, { "epoch": 23.66, "learning_rate": 1.7025e-05, "loss": 1.4367, "step": 4616500 }, { "epoch": 23.66, "learning_rate": 1.702142857142857e-05, "loss": 1.4403, "step": 4617000 }, { "epoch": 23.66, "learning_rate": 1.7017857142857142e-05, "loss": 1.4293, "step": 4617500 }, { "epoch": 23.66, "learning_rate": 1.7014285714285716e-05, "loss": 1.4396, "step": 4618000 }, { "epoch": 23.67, "learning_rate": 1.7010714285714284e-05, "loss": 1.4298, "step": 4618500 }, { "epoch": 23.67, "learning_rate": 1.7007142857142858e-05, "loss": 1.4407, "step": 4619000 }, { "epoch": 23.67, "learning_rate": 1.700357142857143e-05, "loss": 1.4294, "step": 4619500 }, { "epoch": 23.67, "learning_rate": 1.7000000000000003e-05, "loss": 1.4373, "step": 4620000 }, { "epoch": 23.68, "learning_rate": 1.699642857142857e-05, "loss": 1.4357, "step": 4620500 }, { "epoch": 23.68, "learning_rate": 1.6992857142857145e-05, "loss": 1.4369, "step": 4621000 }, { "epoch": 23.68, "learning_rate": 1.6989285714285716e-05, "loss": 1.4337, "step": 4621500 }, { "epoch": 23.69, "learning_rate": 1.6985714285714287e-05, "loss": 1.4294, "step": 4622000 }, { "epoch": 23.69, "learning_rate": 1.6982142857142858e-05, "loss": 1.4344, "step": 4622500 }, { "epoch": 23.69, "learning_rate": 1.697857142857143e-05, "loss": 1.44, "step": 4623000 }, { "epoch": 23.69, "learning_rate": 1.6975000000000003e-05, "loss": 1.4399, "step": 4623500 }, { "epoch": 23.7, "learning_rate": 1.697142857142857e-05, "loss": 1.4391, "step": 4624000 }, { "epoch": 23.7, "learning_rate": 1.6967857142857144e-05, "loss": 1.4402, "step": 4624500 }, { "epoch": 23.7, "learning_rate": 1.6964285714285715e-05, "loss": 1.4325, "step": 4625000 }, { "epoch": 23.7, "learning_rate": 1.6960714285714286e-05, "loss": 1.4337, "step": 4625500 }, { "epoch": 23.71, "learning_rate": 1.6957142857142857e-05, "loss": 1.4316, "step": 4626000 }, { "epoch": 23.71, "learning_rate": 1.695357142857143e-05, "loss": 1.4468, "step": 4626500 }, { "epoch": 23.71, "learning_rate": 1.6950000000000002e-05, "loss": 1.4371, "step": 4627000 }, { "epoch": 23.71, "learning_rate": 1.6946428571428573e-05, "loss": 1.4301, "step": 4627500 }, { "epoch": 23.72, "learning_rate": 1.6942857142857144e-05, "loss": 1.4468, "step": 4628000 }, { "epoch": 23.72, "learning_rate": 1.6939285714285715e-05, "loss": 1.4352, "step": 4628500 }, { "epoch": 23.72, "learning_rate": 1.6935714285714285e-05, "loss": 1.433, "step": 4629000 }, { "epoch": 23.73, "learning_rate": 1.6932142857142856e-05, "loss": 1.4305, "step": 4629500 }, { "epoch": 23.73, "learning_rate": 1.692857142857143e-05, "loss": 1.4396, "step": 4630000 }, { "epoch": 23.73, "learning_rate": 1.6925e-05, "loss": 1.4422, "step": 4630500 }, { "epoch": 23.73, "learning_rate": 1.6921428571428572e-05, "loss": 1.446, "step": 4631000 }, { "epoch": 23.74, "learning_rate": 1.6917857142857143e-05, "loss": 1.4407, "step": 4631500 }, { "epoch": 23.74, "learning_rate": 1.6914285714285717e-05, "loss": 1.4482, "step": 4632000 }, { "epoch": 23.74, "learning_rate": 1.6910714285714285e-05, "loss": 1.4376, "step": 4632500 }, { "epoch": 23.74, "learning_rate": 1.690714285714286e-05, "loss": 1.4382, "step": 4633000 }, { "epoch": 23.75, "learning_rate": 1.690357142857143e-05, "loss": 1.4331, "step": 4633500 }, { "epoch": 23.75, "learning_rate": 1.69e-05, "loss": 1.4383, "step": 4634000 }, { "epoch": 23.75, "learning_rate": 1.689642857142857e-05, "loss": 1.4397, "step": 4634500 }, { "epoch": 23.75, "learning_rate": 1.6892857142857143e-05, "loss": 1.441, "step": 4635000 }, { "epoch": 23.76, "learning_rate": 1.6889285714285717e-05, "loss": 1.4395, "step": 4635500 }, { "epoch": 23.76, "learning_rate": 1.6885714285714284e-05, "loss": 1.4343, "step": 4636000 }, { "epoch": 23.76, "learning_rate": 1.688214285714286e-05, "loss": 1.4289, "step": 4636500 }, { "epoch": 23.77, "learning_rate": 1.687857142857143e-05, "loss": 1.4336, "step": 4637000 }, { "epoch": 23.77, "learning_rate": 1.6875000000000004e-05, "loss": 1.4349, "step": 4637500 }, { "epoch": 23.77, "learning_rate": 1.687142857142857e-05, "loss": 1.4434, "step": 4638000 }, { "epoch": 23.77, "learning_rate": 1.6867857142857142e-05, "loss": 1.4548, "step": 4638500 }, { "epoch": 23.78, "learning_rate": 1.6864285714285716e-05, "loss": 1.4409, "step": 4639000 }, { "epoch": 23.78, "learning_rate": 1.6860714285714284e-05, "loss": 1.4344, "step": 4639500 }, { "epoch": 23.78, "learning_rate": 1.6857142857142858e-05, "loss": 1.4448, "step": 4640000 }, { "epoch": 23.78, "learning_rate": 1.685357142857143e-05, "loss": 1.4383, "step": 4640500 }, { "epoch": 23.79, "learning_rate": 1.6850000000000003e-05, "loss": 1.4431, "step": 4641000 }, { "epoch": 23.79, "learning_rate": 1.684642857142857e-05, "loss": 1.4412, "step": 4641500 }, { "epoch": 23.79, "learning_rate": 1.6842857142857145e-05, "loss": 1.4323, "step": 4642000 }, { "epoch": 23.79, "learning_rate": 1.6839285714285716e-05, "loss": 1.4449, "step": 4642500 }, { "epoch": 23.8, "learning_rate": 1.6835714285714286e-05, "loss": 1.4345, "step": 4643000 }, { "epoch": 23.8, "learning_rate": 1.6832142857142857e-05, "loss": 1.4416, "step": 4643500 }, { "epoch": 23.8, "learning_rate": 1.6828571428571428e-05, "loss": 1.435, "step": 4644000 }, { "epoch": 23.81, "learning_rate": 1.6825000000000002e-05, "loss": 1.4334, "step": 4644500 }, { "epoch": 23.81, "learning_rate": 1.682142857142857e-05, "loss": 1.4396, "step": 4645000 }, { "epoch": 23.81, "learning_rate": 1.6817857142857144e-05, "loss": 1.4329, "step": 4645500 }, { "epoch": 23.81, "learning_rate": 1.6814285714285715e-05, "loss": 1.4403, "step": 4646000 }, { "epoch": 23.82, "learning_rate": 1.6810714285714286e-05, "loss": 1.4376, "step": 4646500 }, { "epoch": 23.82, "learning_rate": 1.6807142857142857e-05, "loss": 1.427, "step": 4647000 }, { "epoch": 23.82, "learning_rate": 1.680357142857143e-05, "loss": 1.4271, "step": 4647500 }, { "epoch": 23.82, "learning_rate": 1.6800000000000002e-05, "loss": 1.4427, "step": 4648000 }, { "epoch": 23.83, "learning_rate": 1.6796428571428573e-05, "loss": 1.4425, "step": 4648500 }, { "epoch": 23.83, "learning_rate": 1.6792857142857143e-05, "loss": 1.4336, "step": 4649000 }, { "epoch": 23.83, "learning_rate": 1.6789285714285714e-05, "loss": 1.4349, "step": 4649500 }, { "epoch": 23.83, "learning_rate": 1.6785714285714285e-05, "loss": 1.442, "step": 4650000 }, { "epoch": 23.84, "learning_rate": 1.6782142857142856e-05, "loss": 1.4364, "step": 4650500 }, { "epoch": 23.84, "learning_rate": 1.677857142857143e-05, "loss": 1.4323, "step": 4651000 }, { "epoch": 23.84, "learning_rate": 1.6775e-05, "loss": 1.436, "step": 4651500 }, { "epoch": 23.85, "learning_rate": 1.6771428571428572e-05, "loss": 1.4318, "step": 4652000 }, { "epoch": 23.85, "learning_rate": 1.6767857142857143e-05, "loss": 1.4326, "step": 4652500 }, { "epoch": 23.85, "learning_rate": 1.6764285714285717e-05, "loss": 1.4324, "step": 4653000 }, { "epoch": 23.85, "learning_rate": 1.6760714285714285e-05, "loss": 1.4386, "step": 4653500 }, { "epoch": 23.86, "learning_rate": 1.675714285714286e-05, "loss": 1.4329, "step": 4654000 }, { "epoch": 23.86, "learning_rate": 1.675357142857143e-05, "loss": 1.4352, "step": 4654500 }, { "epoch": 23.86, "learning_rate": 1.675e-05, "loss": 1.4296, "step": 4655000 }, { "epoch": 23.86, "learning_rate": 1.674642857142857e-05, "loss": 1.4404, "step": 4655500 }, { "epoch": 23.87, "learning_rate": 1.6742857142857142e-05, "loss": 1.4367, "step": 4656000 }, { "epoch": 23.87, "learning_rate": 1.6739285714285717e-05, "loss": 1.4445, "step": 4656500 }, { "epoch": 23.87, "learning_rate": 1.6735714285714287e-05, "loss": 1.4262, "step": 4657000 }, { "epoch": 23.87, "learning_rate": 1.6732142857142858e-05, "loss": 1.4325, "step": 4657500 }, { "epoch": 23.88, "learning_rate": 1.672857142857143e-05, "loss": 1.4342, "step": 4658000 }, { "epoch": 23.88, "learning_rate": 1.6725000000000003e-05, "loss": 1.434, "step": 4658500 }, { "epoch": 23.88, "learning_rate": 1.672142857142857e-05, "loss": 1.4356, "step": 4659000 }, { "epoch": 23.89, "learning_rate": 1.6717857142857145e-05, "loss": 1.4397, "step": 4659500 }, { "epoch": 23.89, "learning_rate": 1.6714285714285716e-05, "loss": 1.4309, "step": 4660000 }, { "epoch": 23.89, "learning_rate": 1.6710714285714287e-05, "loss": 1.4379, "step": 4660500 }, { "epoch": 23.89, "learning_rate": 1.6707142857142858e-05, "loss": 1.421, "step": 4661000 }, { "epoch": 23.9, "learning_rate": 1.670357142857143e-05, "loss": 1.438, "step": 4661500 }, { "epoch": 23.9, "learning_rate": 1.6700000000000003e-05, "loss": 1.4374, "step": 4662000 }, { "epoch": 23.9, "learning_rate": 1.669642857142857e-05, "loss": 1.4463, "step": 4662500 }, { "epoch": 23.9, "learning_rate": 1.6692857142857144e-05, "loss": 1.4334, "step": 4663000 }, { "epoch": 23.91, "learning_rate": 1.6689285714285715e-05, "loss": 1.426, "step": 4663500 }, { "epoch": 23.91, "learning_rate": 1.6685714285714286e-05, "loss": 1.4268, "step": 4664000 }, { "epoch": 23.91, "learning_rate": 1.6682142857142857e-05, "loss": 1.4259, "step": 4664500 }, { "epoch": 23.91, "learning_rate": 1.6678571428571428e-05, "loss": 1.4401, "step": 4665000 }, { "epoch": 23.92, "learning_rate": 1.6675000000000002e-05, "loss": 1.4371, "step": 4665500 }, { "epoch": 23.92, "learning_rate": 1.667142857142857e-05, "loss": 1.4387, "step": 4666000 }, { "epoch": 23.92, "learning_rate": 1.6667857142857144e-05, "loss": 1.4371, "step": 4666500 }, { "epoch": 23.92, "learning_rate": 1.6664285714285715e-05, "loss": 1.4304, "step": 4667000 }, { "epoch": 23.93, "learning_rate": 1.666071428571429e-05, "loss": 1.4231, "step": 4667500 }, { "epoch": 23.93, "learning_rate": 1.6657142857142856e-05, "loss": 1.4364, "step": 4668000 }, { "epoch": 23.93, "learning_rate": 1.665357142857143e-05, "loss": 1.4358, "step": 4668500 }, { "epoch": 23.94, "learning_rate": 1.665e-05, "loss": 1.4367, "step": 4669000 }, { "epoch": 23.94, "learning_rate": 1.6646428571428572e-05, "loss": 1.4328, "step": 4669500 }, { "epoch": 23.94, "learning_rate": 1.6642857142857143e-05, "loss": 1.4319, "step": 4670000 }, { "epoch": 23.94, "learning_rate": 1.6639285714285714e-05, "loss": 1.4301, "step": 4670500 }, { "epoch": 23.95, "learning_rate": 1.663571428571429e-05, "loss": 1.436, "step": 4671000 }, { "epoch": 23.95, "learning_rate": 1.6632142857142856e-05, "loss": 1.4246, "step": 4671500 }, { "epoch": 23.95, "learning_rate": 1.662857142857143e-05, "loss": 1.4416, "step": 4672000 }, { "epoch": 23.95, "learning_rate": 1.6625e-05, "loss": 1.4374, "step": 4672500 }, { "epoch": 23.96, "learning_rate": 1.6621428571428572e-05, "loss": 1.4397, "step": 4673000 }, { "epoch": 23.96, "learning_rate": 1.6617857142857143e-05, "loss": 1.4288, "step": 4673500 }, { "epoch": 23.96, "learning_rate": 1.6614285714285717e-05, "loss": 1.4399, "step": 4674000 }, { "epoch": 23.96, "learning_rate": 1.6610714285714288e-05, "loss": 1.442, "step": 4674500 }, { "epoch": 23.97, "learning_rate": 1.660714285714286e-05, "loss": 1.4354, "step": 4675000 }, { "epoch": 23.97, "learning_rate": 1.660357142857143e-05, "loss": 1.427, "step": 4675500 }, { "epoch": 23.97, "learning_rate": 1.66e-05, "loss": 1.4285, "step": 4676000 }, { "epoch": 23.98, "learning_rate": 1.659642857142857e-05, "loss": 1.4297, "step": 4676500 }, { "epoch": 23.98, "learning_rate": 1.6592857142857142e-05, "loss": 1.4365, "step": 4677000 }, { "epoch": 23.98, "learning_rate": 1.6589285714285716e-05, "loss": 1.4248, "step": 4677500 }, { "epoch": 23.98, "learning_rate": 1.6585714285714287e-05, "loss": 1.4415, "step": 4678000 }, { "epoch": 23.99, "learning_rate": 1.6582142857142858e-05, "loss": 1.4299, "step": 4678500 }, { "epoch": 23.99, "learning_rate": 1.657857142857143e-05, "loss": 1.4349, "step": 4679000 }, { "epoch": 23.99, "learning_rate": 1.6575000000000003e-05, "loss": 1.4374, "step": 4679500 }, { "epoch": 23.99, "learning_rate": 1.657142857142857e-05, "loss": 1.4291, "step": 4680000 }, { "epoch": 24.0, "learning_rate": 1.6567857142857145e-05, "loss": 1.4344, "step": 4680500 }, { "epoch": 24.0, "learning_rate": 1.6564285714285716e-05, "loss": 1.4389, "step": 4681000 }, { "epoch": 24.0, "learning_rate": 1.6560714285714287e-05, "loss": 1.4221, "step": 4681500 }, { "epoch": 24.0, "learning_rate": 1.6557142857142857e-05, "loss": 1.4235, "step": 4682000 }, { "epoch": 24.01, "learning_rate": 1.6553571428571428e-05, "loss": 1.4333, "step": 4682500 }, { "epoch": 24.01, "learning_rate": 1.6550000000000002e-05, "loss": 1.4409, "step": 4683000 }, { "epoch": 24.01, "learning_rate": 1.654642857142857e-05, "loss": 1.4306, "step": 4683500 }, { "epoch": 24.02, "learning_rate": 1.6542857142857144e-05, "loss": 1.4227, "step": 4684000 }, { "epoch": 24.02, "learning_rate": 1.6539285714285715e-05, "loss": 1.4267, "step": 4684500 }, { "epoch": 24.02, "learning_rate": 1.653571428571429e-05, "loss": 1.4225, "step": 4685000 }, { "epoch": 24.02, "learning_rate": 1.6532142857142857e-05, "loss": 1.4346, "step": 4685500 }, { "epoch": 24.03, "learning_rate": 1.652857142857143e-05, "loss": 1.426, "step": 4686000 }, { "epoch": 24.03, "learning_rate": 1.6525000000000002e-05, "loss": 1.4207, "step": 4686500 }, { "epoch": 24.03, "learning_rate": 1.652142857142857e-05, "loss": 1.4356, "step": 4687000 }, { "epoch": 24.03, "learning_rate": 1.6517857142857144e-05, "loss": 1.4269, "step": 4687500 }, { "epoch": 24.04, "learning_rate": 1.6514285714285714e-05, "loss": 1.4274, "step": 4688000 }, { "epoch": 24.04, "learning_rate": 1.651071428571429e-05, "loss": 1.4247, "step": 4688500 }, { "epoch": 24.04, "learning_rate": 1.6507142857142856e-05, "loss": 1.4325, "step": 4689000 }, { "epoch": 24.04, "learning_rate": 1.650357142857143e-05, "loss": 1.4146, "step": 4689500 }, { "epoch": 24.05, "learning_rate": 1.65e-05, "loss": 1.4331, "step": 4690000 }, { "epoch": 24.05, "learning_rate": 1.6496428571428572e-05, "loss": 1.4206, "step": 4690500 }, { "epoch": 24.05, "learning_rate": 1.6492857142857143e-05, "loss": 1.4315, "step": 4691000 }, { "epoch": 24.06, "learning_rate": 1.6489285714285714e-05, "loss": 1.4303, "step": 4691500 }, { "epoch": 24.06, "learning_rate": 1.6485714285714288e-05, "loss": 1.4209, "step": 4692000 }, { "epoch": 24.06, "learning_rate": 1.6482142857142856e-05, "loss": 1.4384, "step": 4692500 }, { "epoch": 24.06, "learning_rate": 1.647857142857143e-05, "loss": 1.4249, "step": 4693000 }, { "epoch": 24.07, "learning_rate": 1.6475e-05, "loss": 1.4252, "step": 4693500 }, { "epoch": 24.07, "learning_rate": 1.647142857142857e-05, "loss": 1.4266, "step": 4694000 }, { "epoch": 24.07, "learning_rate": 1.6467857142857142e-05, "loss": 1.4189, "step": 4694500 }, { "epoch": 24.07, "learning_rate": 1.6464285714285717e-05, "loss": 1.4271, "step": 4695000 }, { "epoch": 24.08, "learning_rate": 1.6460714285714287e-05, "loss": 1.4266, "step": 4695500 }, { "epoch": 24.08, "learning_rate": 1.645714285714286e-05, "loss": 1.427, "step": 4696000 }, { "epoch": 24.08, "learning_rate": 1.645357142857143e-05, "loss": 1.4328, "step": 4696500 }, { "epoch": 24.08, "learning_rate": 1.645e-05, "loss": 1.4252, "step": 4697000 }, { "epoch": 24.09, "learning_rate": 1.644642857142857e-05, "loss": 1.4317, "step": 4697500 }, { "epoch": 24.09, "learning_rate": 1.6442857142857142e-05, "loss": 1.4163, "step": 4698000 }, { "epoch": 24.09, "learning_rate": 1.6439285714285716e-05, "loss": 1.4422, "step": 4698500 }, { "epoch": 24.1, "learning_rate": 1.6435714285714287e-05, "loss": 1.4342, "step": 4699000 }, { "epoch": 24.1, "learning_rate": 1.6432142857142858e-05, "loss": 1.4236, "step": 4699500 }, { "epoch": 24.1, "learning_rate": 1.642857142857143e-05, "loss": 1.4244, "step": 4700000 }, { "epoch": 24.1, "learning_rate": 1.6425000000000003e-05, "loss": 1.4348, "step": 4700500 }, { "epoch": 24.11, "learning_rate": 1.642142857142857e-05, "loss": 1.4244, "step": 4701000 }, { "epoch": 24.11, "learning_rate": 1.6417857142857145e-05, "loss": 1.4265, "step": 4701500 }, { "epoch": 24.11, "learning_rate": 1.6414285714285715e-05, "loss": 1.429, "step": 4702000 }, { "epoch": 24.11, "learning_rate": 1.6410714285714286e-05, "loss": 1.4293, "step": 4702500 }, { "epoch": 24.12, "learning_rate": 1.6407142857142857e-05, "loss": 1.4311, "step": 4703000 }, { "epoch": 24.12, "learning_rate": 1.6403571428571428e-05, "loss": 1.4205, "step": 4703500 }, { "epoch": 24.12, "learning_rate": 1.6400000000000002e-05, "loss": 1.4268, "step": 4704000 }, { "epoch": 24.12, "learning_rate": 1.639642857142857e-05, "loss": 1.4256, "step": 4704500 }, { "epoch": 24.13, "learning_rate": 1.6392857142857144e-05, "loss": 1.435, "step": 4705000 }, { "epoch": 24.13, "learning_rate": 1.6389285714285715e-05, "loss": 1.4306, "step": 4705500 }, { "epoch": 24.13, "learning_rate": 1.638571428571429e-05, "loss": 1.423, "step": 4706000 }, { "epoch": 24.14, "learning_rate": 1.6382142857142857e-05, "loss": 1.4297, "step": 4706500 }, { "epoch": 24.14, "learning_rate": 1.637857142857143e-05, "loss": 1.4295, "step": 4707000 }, { "epoch": 24.14, "learning_rate": 1.6375e-05, "loss": 1.4298, "step": 4707500 }, { "epoch": 24.14, "learning_rate": 1.6371428571428572e-05, "loss": 1.4367, "step": 4708000 }, { "epoch": 24.15, "learning_rate": 1.6367857142857143e-05, "loss": 1.4345, "step": 4708500 }, { "epoch": 24.15, "learning_rate": 1.6364285714285714e-05, "loss": 1.4327, "step": 4709000 }, { "epoch": 24.15, "learning_rate": 1.636071428571429e-05, "loss": 1.4369, "step": 4709500 }, { "epoch": 24.15, "learning_rate": 1.6357142857142856e-05, "loss": 1.4334, "step": 4710000 }, { "epoch": 24.16, "learning_rate": 1.635357142857143e-05, "loss": 1.4252, "step": 4710500 }, { "epoch": 24.16, "learning_rate": 1.635e-05, "loss": 1.4378, "step": 4711000 }, { "epoch": 24.16, "learning_rate": 1.6346428571428572e-05, "loss": 1.4308, "step": 4711500 }, { "epoch": 24.16, "learning_rate": 1.6342857142857143e-05, "loss": 1.4357, "step": 4712000 }, { "epoch": 24.17, "learning_rate": 1.6339285714285717e-05, "loss": 1.4245, "step": 4712500 }, { "epoch": 24.17, "learning_rate": 1.6335714285714288e-05, "loss": 1.4291, "step": 4713000 }, { "epoch": 24.17, "learning_rate": 1.6332142857142855e-05, "loss": 1.4319, "step": 4713500 }, { "epoch": 24.18, "learning_rate": 1.632857142857143e-05, "loss": 1.4315, "step": 4714000 }, { "epoch": 24.18, "learning_rate": 1.6325e-05, "loss": 1.4408, "step": 4714500 }, { "epoch": 24.18, "learning_rate": 1.632142857142857e-05, "loss": 1.4294, "step": 4715000 }, { "epoch": 24.18, "learning_rate": 1.6317857142857142e-05, "loss": 1.4319, "step": 4715500 }, { "epoch": 24.19, "learning_rate": 1.6314285714285716e-05, "loss": 1.4317, "step": 4716000 }, { "epoch": 24.19, "learning_rate": 1.6310714285714287e-05, "loss": 1.4275, "step": 4716500 }, { "epoch": 24.19, "learning_rate": 1.6307142857142858e-05, "loss": 1.4221, "step": 4717000 }, { "epoch": 24.19, "learning_rate": 1.630357142857143e-05, "loss": 1.4255, "step": 4717500 }, { "epoch": 24.2, "learning_rate": 1.63e-05, "loss": 1.4355, "step": 4718000 }, { "epoch": 24.2, "learning_rate": 1.629642857142857e-05, "loss": 1.438, "step": 4718500 }, { "epoch": 24.2, "learning_rate": 1.629285714285714e-05, "loss": 1.4299, "step": 4719000 }, { "epoch": 24.2, "learning_rate": 1.6289285714285716e-05, "loss": 1.4306, "step": 4719500 }, { "epoch": 24.21, "learning_rate": 1.6285714285714287e-05, "loss": 1.433, "step": 4720000 }, { "epoch": 24.21, "learning_rate": 1.6282142857142857e-05, "loss": 1.4368, "step": 4720500 }, { "epoch": 24.21, "learning_rate": 1.627857142857143e-05, "loss": 1.4331, "step": 4721000 }, { "epoch": 24.22, "learning_rate": 1.6275000000000003e-05, "loss": 1.431, "step": 4721500 }, { "epoch": 24.22, "learning_rate": 1.6271428571428573e-05, "loss": 1.4397, "step": 4722000 }, { "epoch": 24.22, "learning_rate": 1.6267857142857144e-05, "loss": 1.4381, "step": 4722500 }, { "epoch": 24.22, "learning_rate": 1.6264285714285715e-05, "loss": 1.4384, "step": 4723000 }, { "epoch": 24.23, "learning_rate": 1.6260714285714286e-05, "loss": 1.4269, "step": 4723500 }, { "epoch": 24.23, "learning_rate": 1.6257142857142857e-05, "loss": 1.4217, "step": 4724000 }, { "epoch": 24.23, "learning_rate": 1.6253571428571428e-05, "loss": 1.4358, "step": 4724500 }, { "epoch": 24.23, "learning_rate": 1.6250000000000002e-05, "loss": 1.4377, "step": 4725000 }, { "epoch": 24.24, "learning_rate": 1.6246428571428573e-05, "loss": 1.431, "step": 4725500 }, { "epoch": 24.24, "learning_rate": 1.6242857142857144e-05, "loss": 1.4299, "step": 4726000 }, { "epoch": 24.24, "learning_rate": 1.6239285714285715e-05, "loss": 1.4368, "step": 4726500 }, { "epoch": 24.24, "learning_rate": 1.623571428571429e-05, "loss": 1.4327, "step": 4727000 }, { "epoch": 24.25, "learning_rate": 1.6232142857142856e-05, "loss": 1.4341, "step": 4727500 }, { "epoch": 24.25, "learning_rate": 1.622857142857143e-05, "loss": 1.427, "step": 4728000 }, { "epoch": 24.25, "learning_rate": 1.6225e-05, "loss": 1.4313, "step": 4728500 }, { "epoch": 24.26, "learning_rate": 1.6221428571428572e-05, "loss": 1.4345, "step": 4729000 }, { "epoch": 24.26, "learning_rate": 1.6217857142857143e-05, "loss": 1.4366, "step": 4729500 }, { "epoch": 24.26, "learning_rate": 1.6214285714285714e-05, "loss": 1.4323, "step": 4730000 }, { "epoch": 24.26, "learning_rate": 1.6210714285714288e-05, "loss": 1.4345, "step": 4730500 }, { "epoch": 24.27, "learning_rate": 1.6207142857142856e-05, "loss": 1.4321, "step": 4731000 }, { "epoch": 24.27, "learning_rate": 1.620357142857143e-05, "loss": 1.4349, "step": 4731500 }, { "epoch": 24.27, "learning_rate": 1.62e-05, "loss": 1.4272, "step": 4732000 }, { "epoch": 24.27, "learning_rate": 1.6196428571428575e-05, "loss": 1.4294, "step": 4732500 }, { "epoch": 24.28, "learning_rate": 1.6192857142857142e-05, "loss": 1.4341, "step": 4733000 }, { "epoch": 24.28, "learning_rate": 1.6189285714285717e-05, "loss": 1.4473, "step": 4733500 }, { "epoch": 24.28, "learning_rate": 1.6185714285714288e-05, "loss": 1.4374, "step": 4734000 }, { "epoch": 24.28, "learning_rate": 1.618214285714286e-05, "loss": 1.4331, "step": 4734500 }, { "epoch": 24.29, "learning_rate": 1.617857142857143e-05, "loss": 1.4259, "step": 4735000 }, { "epoch": 24.29, "learning_rate": 1.6175e-05, "loss": 1.438, "step": 4735500 }, { "epoch": 24.29, "learning_rate": 1.6171428571428574e-05, "loss": 1.4306, "step": 4736000 }, { "epoch": 24.29, "learning_rate": 1.6167857142857142e-05, "loss": 1.4346, "step": 4736500 }, { "epoch": 24.3, "learning_rate": 1.6164285714285716e-05, "loss": 1.4273, "step": 4737000 }, { "epoch": 24.3, "learning_rate": 1.6160714285714287e-05, "loss": 1.4434, "step": 4737500 }, { "epoch": 24.3, "learning_rate": 1.6157142857142858e-05, "loss": 1.4388, "step": 4738000 }, { "epoch": 24.31, "learning_rate": 1.615357142857143e-05, "loss": 1.4255, "step": 4738500 }, { "epoch": 24.31, "learning_rate": 1.6150000000000003e-05, "loss": 1.4241, "step": 4739000 }, { "epoch": 24.31, "learning_rate": 1.6146428571428574e-05, "loss": 1.4285, "step": 4739500 }, { "epoch": 24.31, "learning_rate": 1.614285714285714e-05, "loss": 1.4321, "step": 4740000 }, { "epoch": 24.32, "learning_rate": 1.6139285714285715e-05, "loss": 1.4326, "step": 4740500 }, { "epoch": 24.32, "learning_rate": 1.6135714285714286e-05, "loss": 1.4367, "step": 4741000 }, { "epoch": 24.32, "learning_rate": 1.6132142857142857e-05, "loss": 1.4351, "step": 4741500 }, { "epoch": 24.32, "learning_rate": 1.6128571428571428e-05, "loss": 1.4357, "step": 4742000 }, { "epoch": 24.33, "learning_rate": 1.6125000000000002e-05, "loss": 1.4268, "step": 4742500 }, { "epoch": 24.33, "learning_rate": 1.6121428571428573e-05, "loss": 1.4316, "step": 4743000 }, { "epoch": 24.33, "learning_rate": 1.6117857142857144e-05, "loss": 1.4226, "step": 4743500 }, { "epoch": 24.33, "learning_rate": 1.6114285714285715e-05, "loss": 1.4266, "step": 4744000 }, { "epoch": 24.34, "learning_rate": 1.6110714285714286e-05, "loss": 1.433, "step": 4744500 }, { "epoch": 24.34, "learning_rate": 1.6107142857142857e-05, "loss": 1.4251, "step": 4745000 }, { "epoch": 24.34, "learning_rate": 1.6103571428571427e-05, "loss": 1.4273, "step": 4745500 }, { "epoch": 24.35, "learning_rate": 1.6100000000000002e-05, "loss": 1.4343, "step": 4746000 }, { "epoch": 24.35, "learning_rate": 1.6096428571428573e-05, "loss": 1.4315, "step": 4746500 }, { "epoch": 24.35, "learning_rate": 1.6092857142857143e-05, "loss": 1.445, "step": 4747000 }, { "epoch": 24.35, "learning_rate": 1.6089285714285714e-05, "loss": 1.4307, "step": 4747500 }, { "epoch": 24.36, "learning_rate": 1.608571428571429e-05, "loss": 1.4374, "step": 4748000 }, { "epoch": 24.36, "learning_rate": 1.6082142857142856e-05, "loss": 1.434, "step": 4748500 }, { "epoch": 24.36, "learning_rate": 1.607857142857143e-05, "loss": 1.4296, "step": 4749000 }, { "epoch": 24.36, "learning_rate": 1.6075e-05, "loss": 1.4447, "step": 4749500 }, { "epoch": 24.37, "learning_rate": 1.6071428571428572e-05, "loss": 1.4303, "step": 4750000 }, { "epoch": 24.37, "learning_rate": 1.6067857142857143e-05, "loss": 1.4303, "step": 4750500 }, { "epoch": 24.37, "learning_rate": 1.6064285714285714e-05, "loss": 1.4197, "step": 4751000 }, { "epoch": 24.37, "learning_rate": 1.6060714285714288e-05, "loss": 1.4339, "step": 4751500 }, { "epoch": 24.38, "learning_rate": 1.6057142857142855e-05, "loss": 1.4393, "step": 4752000 }, { "epoch": 24.38, "learning_rate": 1.605357142857143e-05, "loss": 1.425, "step": 4752500 }, { "epoch": 24.38, "learning_rate": 1.605e-05, "loss": 1.4294, "step": 4753000 }, { "epoch": 24.39, "learning_rate": 1.6046428571428575e-05, "loss": 1.4253, "step": 4753500 }, { "epoch": 24.39, "learning_rate": 1.6042857142857142e-05, "loss": 1.4327, "step": 4754000 }, { "epoch": 24.39, "learning_rate": 1.6039285714285716e-05, "loss": 1.4322, "step": 4754500 }, { "epoch": 24.39, "learning_rate": 1.6035714285714287e-05, "loss": 1.4354, "step": 4755000 }, { "epoch": 24.4, "learning_rate": 1.6032142857142858e-05, "loss": 1.4249, "step": 4755500 }, { "epoch": 24.4, "learning_rate": 1.602857142857143e-05, "loss": 1.4378, "step": 4756000 }, { "epoch": 24.4, "learning_rate": 1.6025e-05, "loss": 1.4348, "step": 4756500 }, { "epoch": 24.4, "learning_rate": 1.6021428571428574e-05, "loss": 1.4292, "step": 4757000 }, { "epoch": 24.41, "learning_rate": 1.601785714285714e-05, "loss": 1.4406, "step": 4757500 }, { "epoch": 24.41, "learning_rate": 1.6014285714285716e-05, "loss": 1.4331, "step": 4758000 }, { "epoch": 24.41, "learning_rate": 1.6010714285714287e-05, "loss": 1.4277, "step": 4758500 }, { "epoch": 24.41, "learning_rate": 1.6007142857142858e-05, "loss": 1.4308, "step": 4759000 }, { "epoch": 24.42, "learning_rate": 1.600357142857143e-05, "loss": 1.4371, "step": 4759500 }, { "epoch": 24.42, "learning_rate": 1.6000000000000003e-05, "loss": 1.4457, "step": 4760000 }, { "epoch": 24.42, "learning_rate": 1.5996428571428574e-05, "loss": 1.4392, "step": 4760500 }, { "epoch": 24.43, "learning_rate": 1.5992857142857144e-05, "loss": 1.4284, "step": 4761000 }, { "epoch": 24.43, "learning_rate": 1.5989285714285715e-05, "loss": 1.4326, "step": 4761500 }, { "epoch": 24.43, "learning_rate": 1.5985714285714286e-05, "loss": 1.4238, "step": 4762000 }, { "epoch": 24.43, "learning_rate": 1.5982142857142857e-05, "loss": 1.4352, "step": 4762500 }, { "epoch": 24.44, "learning_rate": 1.5978571428571428e-05, "loss": 1.4345, "step": 4763000 }, { "epoch": 24.44, "learning_rate": 1.5975000000000002e-05, "loss": 1.4326, "step": 4763500 }, { "epoch": 24.44, "learning_rate": 1.5971428571428573e-05, "loss": 1.4152, "step": 4764000 }, { "epoch": 24.44, "learning_rate": 1.5967857142857144e-05, "loss": 1.4224, "step": 4764500 }, { "epoch": 24.45, "learning_rate": 1.5964285714285715e-05, "loss": 1.427, "step": 4765000 }, { "epoch": 24.45, "learning_rate": 1.5960714285714285e-05, "loss": 1.4227, "step": 4765500 }, { "epoch": 24.45, "learning_rate": 1.5957142857142856e-05, "loss": 1.4369, "step": 4766000 }, { "epoch": 24.45, "learning_rate": 1.5953571428571427e-05, "loss": 1.4369, "step": 4766500 }, { "epoch": 24.46, "learning_rate": 1.595e-05, "loss": 1.4257, "step": 4767000 }, { "epoch": 24.46, "learning_rate": 1.5946428571428572e-05, "loss": 1.4283, "step": 4767500 }, { "epoch": 24.46, "learning_rate": 1.5942857142857143e-05, "loss": 1.4197, "step": 4768000 }, { "epoch": 24.47, "learning_rate": 1.5939285714285714e-05, "loss": 1.4292, "step": 4768500 }, { "epoch": 24.47, "learning_rate": 1.5935714285714288e-05, "loss": 1.426, "step": 4769000 }, { "epoch": 24.47, "learning_rate": 1.5932142857142856e-05, "loss": 1.429, "step": 4769500 }, { "epoch": 24.47, "learning_rate": 1.592857142857143e-05, "loss": 1.434, "step": 4770000 }, { "epoch": 24.48, "learning_rate": 1.5925e-05, "loss": 1.4373, "step": 4770500 }, { "epoch": 24.48, "learning_rate": 1.5921428571428572e-05, "loss": 1.4279, "step": 4771000 }, { "epoch": 24.48, "learning_rate": 1.5917857142857143e-05, "loss": 1.4291, "step": 4771500 }, { "epoch": 24.48, "learning_rate": 1.5914285714285713e-05, "loss": 1.4291, "step": 4772000 }, { "epoch": 24.49, "learning_rate": 1.5910714285714288e-05, "loss": 1.4296, "step": 4772500 }, { "epoch": 24.49, "learning_rate": 1.5907142857142855e-05, "loss": 1.4267, "step": 4773000 }, { "epoch": 24.49, "learning_rate": 1.590357142857143e-05, "loss": 1.4389, "step": 4773500 }, { "epoch": 24.49, "learning_rate": 1.59e-05, "loss": 1.4321, "step": 4774000 }, { "epoch": 24.5, "learning_rate": 1.5896428571428574e-05, "loss": 1.4321, "step": 4774500 }, { "epoch": 24.5, "learning_rate": 1.5892857142857142e-05, "loss": 1.4335, "step": 4775000 }, { "epoch": 24.5, "learning_rate": 1.5889285714285716e-05, "loss": 1.4185, "step": 4775500 }, { "epoch": 24.51, "learning_rate": 1.5885714285714287e-05, "loss": 1.4295, "step": 4776000 }, { "epoch": 24.51, "learning_rate": 1.5882142857142858e-05, "loss": 1.4323, "step": 4776500 }, { "epoch": 24.51, "learning_rate": 1.587857142857143e-05, "loss": 1.428, "step": 4777000 }, { "epoch": 24.51, "learning_rate": 1.5875e-05, "loss": 1.4317, "step": 4777500 }, { "epoch": 24.52, "learning_rate": 1.5871428571428574e-05, "loss": 1.4245, "step": 4778000 }, { "epoch": 24.52, "learning_rate": 1.586785714285714e-05, "loss": 1.4332, "step": 4778500 }, { "epoch": 24.52, "learning_rate": 1.5864285714285716e-05, "loss": 1.4256, "step": 4779000 }, { "epoch": 24.52, "learning_rate": 1.5860714285714286e-05, "loss": 1.4258, "step": 4779500 }, { "epoch": 24.53, "learning_rate": 1.5857142857142857e-05, "loss": 1.4347, "step": 4780000 }, { "epoch": 24.53, "learning_rate": 1.5853571428571428e-05, "loss": 1.4139, "step": 4780500 }, { "epoch": 24.53, "learning_rate": 1.5850000000000002e-05, "loss": 1.4277, "step": 4781000 }, { "epoch": 24.53, "learning_rate": 1.5846428571428573e-05, "loss": 1.4355, "step": 4781500 }, { "epoch": 24.54, "learning_rate": 1.5842857142857144e-05, "loss": 1.4284, "step": 4782000 }, { "epoch": 24.54, "learning_rate": 1.5839285714285715e-05, "loss": 1.4261, "step": 4782500 }, { "epoch": 24.54, "learning_rate": 1.5835714285714286e-05, "loss": 1.4247, "step": 4783000 }, { "epoch": 24.55, "learning_rate": 1.583214285714286e-05, "loss": 1.4302, "step": 4783500 }, { "epoch": 24.55, "learning_rate": 1.5828571428571428e-05, "loss": 1.4341, "step": 4784000 }, { "epoch": 24.55, "learning_rate": 1.5825000000000002e-05, "loss": 1.4291, "step": 4784500 }, { "epoch": 24.55, "learning_rate": 1.5821428571428573e-05, "loss": 1.4287, "step": 4785000 }, { "epoch": 24.56, "learning_rate": 1.5817857142857144e-05, "loss": 1.4363, "step": 4785500 }, { "epoch": 24.56, "learning_rate": 1.5814285714285714e-05, "loss": 1.4233, "step": 4786000 }, { "epoch": 24.56, "learning_rate": 1.581071428571429e-05, "loss": 1.4319, "step": 4786500 }, { "epoch": 24.56, "learning_rate": 1.580714285714286e-05, "loss": 1.431, "step": 4787000 }, { "epoch": 24.57, "learning_rate": 1.580357142857143e-05, "loss": 1.4266, "step": 4787500 }, { "epoch": 24.57, "learning_rate": 1.58e-05, "loss": 1.4277, "step": 4788000 }, { "epoch": 24.57, "learning_rate": 1.5796428571428572e-05, "loss": 1.4365, "step": 4788500 }, { "epoch": 24.57, "learning_rate": 1.5792857142857143e-05, "loss": 1.4327, "step": 4789000 }, { "epoch": 24.58, "learning_rate": 1.5789285714285714e-05, "loss": 1.4304, "step": 4789500 }, { "epoch": 24.58, "learning_rate": 1.5785714285714288e-05, "loss": 1.4282, "step": 4790000 }, { "epoch": 24.58, "learning_rate": 1.578214285714286e-05, "loss": 1.4203, "step": 4790500 }, { "epoch": 24.59, "learning_rate": 1.577857142857143e-05, "loss": 1.4288, "step": 4791000 }, { "epoch": 24.59, "learning_rate": 1.5775e-05, "loss": 1.431, "step": 4791500 }, { "epoch": 24.59, "learning_rate": 1.577142857142857e-05, "loss": 1.4372, "step": 4792000 }, { "epoch": 24.59, "learning_rate": 1.5767857142857142e-05, "loss": 1.4275, "step": 4792500 }, { "epoch": 24.6, "learning_rate": 1.5764285714285713e-05, "loss": 1.4225, "step": 4793000 }, { "epoch": 24.6, "learning_rate": 1.5760714285714287e-05, "loss": 1.4213, "step": 4793500 }, { "epoch": 24.6, "learning_rate": 1.5757142857142858e-05, "loss": 1.4366, "step": 4794000 }, { "epoch": 24.6, "learning_rate": 1.575357142857143e-05, "loss": 1.4167, "step": 4794500 }, { "epoch": 24.61, "learning_rate": 1.575e-05, "loss": 1.4311, "step": 4795000 }, { "epoch": 24.61, "learning_rate": 1.5746428571428574e-05, "loss": 1.4277, "step": 4795500 }, { "epoch": 24.61, "learning_rate": 1.574285714285714e-05, "loss": 1.4261, "step": 4796000 }, { "epoch": 24.61, "learning_rate": 1.5739285714285716e-05, "loss": 1.4209, "step": 4796500 }, { "epoch": 24.62, "learning_rate": 1.5735714285714287e-05, "loss": 1.4161, "step": 4797000 }, { "epoch": 24.62, "learning_rate": 1.5732142857142858e-05, "loss": 1.4277, "step": 4797500 }, { "epoch": 24.62, "learning_rate": 1.572857142857143e-05, "loss": 1.4307, "step": 4798000 }, { "epoch": 24.63, "learning_rate": 1.5725e-05, "loss": 1.4404, "step": 4798500 }, { "epoch": 24.63, "learning_rate": 1.5721428571428574e-05, "loss": 1.4291, "step": 4799000 }, { "epoch": 24.63, "learning_rate": 1.571785714285714e-05, "loss": 1.4304, "step": 4799500 }, { "epoch": 24.63, "learning_rate": 1.5714285714285715e-05, "loss": 1.427, "step": 4800000 }, { "epoch": 24.64, "learning_rate": 1.5710714285714286e-05, "loss": 1.4272, "step": 4800500 }, { "epoch": 24.64, "learning_rate": 1.570714285714286e-05, "loss": 1.4388, "step": 4801000 }, { "epoch": 24.64, "learning_rate": 1.5703571428571428e-05, "loss": 1.4289, "step": 4801500 }, { "epoch": 24.64, "learning_rate": 1.5700000000000002e-05, "loss": 1.4182, "step": 4802000 }, { "epoch": 24.65, "learning_rate": 1.5696428571428573e-05, "loss": 1.432, "step": 4802500 }, { "epoch": 24.65, "learning_rate": 1.5692857142857144e-05, "loss": 1.4176, "step": 4803000 }, { "epoch": 24.65, "learning_rate": 1.5689285714285715e-05, "loss": 1.4237, "step": 4803500 }, { "epoch": 24.65, "learning_rate": 1.5685714285714286e-05, "loss": 1.4237, "step": 4804000 }, { "epoch": 24.66, "learning_rate": 1.568214285714286e-05, "loss": 1.4198, "step": 4804500 }, { "epoch": 24.66, "learning_rate": 1.5678571428571427e-05, "loss": 1.4288, "step": 4805000 }, { "epoch": 24.66, "learning_rate": 1.5675e-05, "loss": 1.4323, "step": 4805500 }, { "epoch": 24.67, "learning_rate": 1.5671428571428572e-05, "loss": 1.424, "step": 4806000 }, { "epoch": 24.67, "learning_rate": 1.5667857142857143e-05, "loss": 1.4204, "step": 4806500 }, { "epoch": 24.67, "learning_rate": 1.5664285714285714e-05, "loss": 1.423, "step": 4807000 }, { "epoch": 24.67, "learning_rate": 1.566071428571429e-05, "loss": 1.4251, "step": 4807500 }, { "epoch": 24.68, "learning_rate": 1.565714285714286e-05, "loss": 1.4293, "step": 4808000 }, { "epoch": 24.68, "learning_rate": 1.565357142857143e-05, "loss": 1.4233, "step": 4808500 }, { "epoch": 24.68, "learning_rate": 1.565e-05, "loss": 1.4315, "step": 4809000 }, { "epoch": 24.68, "learning_rate": 1.5646428571428572e-05, "loss": 1.4318, "step": 4809500 }, { "epoch": 24.69, "learning_rate": 1.5642857142857143e-05, "loss": 1.424, "step": 4810000 }, { "epoch": 24.69, "learning_rate": 1.5639285714285714e-05, "loss": 1.4279, "step": 4810500 }, { "epoch": 24.69, "learning_rate": 1.5635714285714288e-05, "loss": 1.4293, "step": 4811000 }, { "epoch": 24.69, "learning_rate": 1.563214285714286e-05, "loss": 1.4315, "step": 4811500 }, { "epoch": 24.7, "learning_rate": 1.562857142857143e-05, "loss": 1.4279, "step": 4812000 }, { "epoch": 24.7, "learning_rate": 1.5625e-05, "loss": 1.4293, "step": 4812500 }, { "epoch": 24.7, "learning_rate": 1.5621428571428575e-05, "loss": 1.4427, "step": 4813000 }, { "epoch": 24.7, "learning_rate": 1.5617857142857142e-05, "loss": 1.4391, "step": 4813500 }, { "epoch": 24.71, "learning_rate": 1.5614285714285716e-05, "loss": 1.4315, "step": 4814000 }, { "epoch": 24.71, "learning_rate": 1.5610714285714287e-05, "loss": 1.4479, "step": 4814500 }, { "epoch": 24.71, "learning_rate": 1.5607142857142858e-05, "loss": 1.4569, "step": 4815000 }, { "epoch": 24.72, "learning_rate": 1.560357142857143e-05, "loss": 1.4352, "step": 4815500 }, { "epoch": 24.72, "learning_rate": 1.56e-05, "loss": 1.4212, "step": 4816000 }, { "epoch": 24.72, "learning_rate": 1.5596428571428574e-05, "loss": 1.4281, "step": 4816500 }, { "epoch": 24.72, "learning_rate": 1.559285714285714e-05, "loss": 1.4335, "step": 4817000 }, { "epoch": 24.73, "learning_rate": 1.5589285714285716e-05, "loss": 1.4293, "step": 4817500 }, { "epoch": 24.73, "learning_rate": 1.5585714285714287e-05, "loss": 1.4324, "step": 4818000 }, { "epoch": 24.73, "learning_rate": 1.5582142857142857e-05, "loss": 1.4381, "step": 4818500 }, { "epoch": 24.73, "learning_rate": 1.5578571428571428e-05, "loss": 1.4368, "step": 4819000 }, { "epoch": 24.74, "learning_rate": 1.5575e-05, "loss": 1.4346, "step": 4819500 }, { "epoch": 24.74, "learning_rate": 1.5571428571428573e-05, "loss": 1.4366, "step": 4820000 }, { "epoch": 24.74, "learning_rate": 1.556785714285714e-05, "loss": 1.4267, "step": 4820500 }, { "epoch": 24.74, "learning_rate": 1.5564285714285715e-05, "loss": 1.4151, "step": 4821000 }, { "epoch": 24.75, "learning_rate": 1.5560714285714286e-05, "loss": 1.4359, "step": 4821500 }, { "epoch": 24.75, "learning_rate": 1.555714285714286e-05, "loss": 1.4176, "step": 4822000 }, { "epoch": 24.75, "learning_rate": 1.5553571428571428e-05, "loss": 1.4359, "step": 4822500 }, { "epoch": 24.76, "learning_rate": 1.5550000000000002e-05, "loss": 1.4144, "step": 4823000 }, { "epoch": 24.76, "learning_rate": 1.5546428571428573e-05, "loss": 1.4264, "step": 4823500 }, { "epoch": 24.76, "learning_rate": 1.5542857142857144e-05, "loss": 1.4407, "step": 4824000 }, { "epoch": 24.76, "learning_rate": 1.5539285714285714e-05, "loss": 1.4318, "step": 4824500 }, { "epoch": 24.77, "learning_rate": 1.5535714285714285e-05, "loss": 1.4315, "step": 4825000 }, { "epoch": 24.77, "learning_rate": 1.553214285714286e-05, "loss": 1.4365, "step": 4825500 }, { "epoch": 24.77, "learning_rate": 1.5528571428571427e-05, "loss": 1.4202, "step": 4826000 }, { "epoch": 24.77, "learning_rate": 1.5525e-05, "loss": 1.417, "step": 4826500 }, { "epoch": 24.78, "learning_rate": 1.5521428571428572e-05, "loss": 1.4281, "step": 4827000 }, { "epoch": 24.78, "learning_rate": 1.5517857142857143e-05, "loss": 1.4363, "step": 4827500 }, { "epoch": 24.78, "learning_rate": 1.5514285714285714e-05, "loss": 1.4297, "step": 4828000 }, { "epoch": 24.78, "learning_rate": 1.5510714285714288e-05, "loss": 1.43, "step": 4828500 }, { "epoch": 24.79, "learning_rate": 1.550714285714286e-05, "loss": 1.4282, "step": 4829000 }, { "epoch": 24.79, "learning_rate": 1.550357142857143e-05, "loss": 1.4232, "step": 4829500 }, { "epoch": 24.79, "learning_rate": 1.55e-05, "loss": 1.4289, "step": 4830000 }, { "epoch": 24.8, "learning_rate": 1.549642857142857e-05, "loss": 1.4302, "step": 4830500 }, { "epoch": 24.8, "learning_rate": 1.5492857142857142e-05, "loss": 1.4264, "step": 4831000 }, { "epoch": 24.8, "learning_rate": 1.5489285714285713e-05, "loss": 1.4177, "step": 4831500 }, { "epoch": 24.8, "learning_rate": 1.5485714285714287e-05, "loss": 1.4328, "step": 4832000 }, { "epoch": 24.81, "learning_rate": 1.548214285714286e-05, "loss": 1.422, "step": 4832500 }, { "epoch": 24.81, "learning_rate": 1.547857142857143e-05, "loss": 1.4186, "step": 4833000 }, { "epoch": 24.81, "learning_rate": 1.5475e-05, "loss": 1.4306, "step": 4833500 }, { "epoch": 24.81, "learning_rate": 1.5471428571428574e-05, "loss": 1.4328, "step": 4834000 }, { "epoch": 24.82, "learning_rate": 1.5467857142857142e-05, "loss": 1.4329, "step": 4834500 }, { "epoch": 24.82, "learning_rate": 1.5464285714285716e-05, "loss": 1.4304, "step": 4835000 }, { "epoch": 24.82, "learning_rate": 1.5460714285714287e-05, "loss": 1.4288, "step": 4835500 }, { "epoch": 24.82, "learning_rate": 1.5457142857142858e-05, "loss": 1.4244, "step": 4836000 }, { "epoch": 24.83, "learning_rate": 1.545357142857143e-05, "loss": 1.4202, "step": 4836500 }, { "epoch": 24.83, "learning_rate": 1.545e-05, "loss": 1.4247, "step": 4837000 }, { "epoch": 24.83, "learning_rate": 1.5446428571428574e-05, "loss": 1.4268, "step": 4837500 }, { "epoch": 24.84, "learning_rate": 1.544285714285714e-05, "loss": 1.4294, "step": 4838000 }, { "epoch": 24.84, "learning_rate": 1.5439285714285715e-05, "loss": 1.421, "step": 4838500 }, { "epoch": 24.84, "learning_rate": 1.5435714285714286e-05, "loss": 1.4183, "step": 4839000 }, { "epoch": 24.84, "learning_rate": 1.543214285714286e-05, "loss": 1.4266, "step": 4839500 }, { "epoch": 24.85, "learning_rate": 1.5428571428571428e-05, "loss": 1.4333, "step": 4840000 }, { "epoch": 24.85, "learning_rate": 1.5425000000000002e-05, "loss": 1.4314, "step": 4840500 }, { "epoch": 24.85, "learning_rate": 1.5421428571428573e-05, "loss": 1.4285, "step": 4841000 }, { "epoch": 24.85, "learning_rate": 1.541785714285714e-05, "loss": 1.4272, "step": 4841500 }, { "epoch": 24.86, "learning_rate": 1.5414285714285715e-05, "loss": 1.4375, "step": 4842000 }, { "epoch": 24.86, "learning_rate": 1.5410714285714286e-05, "loss": 1.4284, "step": 4842500 }, { "epoch": 24.86, "learning_rate": 1.540714285714286e-05, "loss": 1.4226, "step": 4843000 }, { "epoch": 24.86, "learning_rate": 1.5403571428571427e-05, "loss": 1.4285, "step": 4843500 }, { "epoch": 24.87, "learning_rate": 1.54e-05, "loss": 1.4332, "step": 4844000 }, { "epoch": 24.87, "learning_rate": 1.5396428571428572e-05, "loss": 1.4326, "step": 4844500 }, { "epoch": 24.87, "learning_rate": 1.5392857142857143e-05, "loss": 1.4316, "step": 4845000 }, { "epoch": 24.88, "learning_rate": 1.5389285714285714e-05, "loss": 1.4368, "step": 4845500 }, { "epoch": 24.88, "learning_rate": 1.5385714285714285e-05, "loss": 1.4341, "step": 4846000 }, { "epoch": 24.88, "learning_rate": 1.538214285714286e-05, "loss": 1.4362, "step": 4846500 }, { "epoch": 24.88, "learning_rate": 1.5378571428571427e-05, "loss": 1.4195, "step": 4847000 }, { "epoch": 24.89, "learning_rate": 1.5375e-05, "loss": 1.4256, "step": 4847500 }, { "epoch": 24.89, "learning_rate": 1.5371428571428572e-05, "loss": 1.4192, "step": 4848000 }, { "epoch": 24.89, "learning_rate": 1.5367857142857146e-05, "loss": 1.4298, "step": 4848500 }, { "epoch": 24.89, "learning_rate": 1.5364285714285714e-05, "loss": 1.4199, "step": 4849000 }, { "epoch": 24.9, "learning_rate": 1.5360714285714288e-05, "loss": 1.4297, "step": 4849500 }, { "epoch": 24.9, "learning_rate": 1.535714285714286e-05, "loss": 1.4239, "step": 4850000 }, { "epoch": 24.9, "learning_rate": 1.535357142857143e-05, "loss": 1.4279, "step": 4850500 }, { "epoch": 24.9, "learning_rate": 1.535e-05, "loss": 1.4151, "step": 4851000 }, { "epoch": 24.91, "learning_rate": 1.534642857142857e-05, "loss": 1.4196, "step": 4851500 }, { "epoch": 24.91, "learning_rate": 1.5342857142857146e-05, "loss": 1.4151, "step": 4852000 }, { "epoch": 24.91, "learning_rate": 1.5339285714285713e-05, "loss": 1.4351, "step": 4852500 }, { "epoch": 24.92, "learning_rate": 1.5335714285714287e-05, "loss": 1.4215, "step": 4853000 }, { "epoch": 24.92, "learning_rate": 1.5332142857142858e-05, "loss": 1.4275, "step": 4853500 }, { "epoch": 24.92, "learning_rate": 1.532857142857143e-05, "loss": 1.4237, "step": 4854000 }, { "epoch": 24.92, "learning_rate": 1.5325e-05, "loss": 1.4227, "step": 4854500 }, { "epoch": 24.93, "learning_rate": 1.5321428571428574e-05, "loss": 1.4342, "step": 4855000 }, { "epoch": 24.93, "learning_rate": 1.5317857142857145e-05, "loss": 1.42, "step": 4855500 }, { "epoch": 24.93, "learning_rate": 1.5314285714285716e-05, "loss": 1.4207, "step": 4856000 }, { "epoch": 24.93, "learning_rate": 1.5310714285714287e-05, "loss": 1.4177, "step": 4856500 }, { "epoch": 24.94, "learning_rate": 1.5307142857142857e-05, "loss": 1.4318, "step": 4857000 }, { "epoch": 24.94, "learning_rate": 1.530357142857143e-05, "loss": 1.4381, "step": 4857500 }, { "epoch": 24.94, "learning_rate": 1.53e-05, "loss": 1.4312, "step": 4858000 }, { "epoch": 24.94, "learning_rate": 1.5296428571428573e-05, "loss": 1.4355, "step": 4858500 }, { "epoch": 24.95, "learning_rate": 1.5292857142857144e-05, "loss": 1.4267, "step": 4859000 }, { "epoch": 24.95, "learning_rate": 1.5289285714285715e-05, "loss": 1.4193, "step": 4859500 }, { "epoch": 24.95, "learning_rate": 1.5285714285714286e-05, "loss": 1.4339, "step": 4860000 }, { "epoch": 24.96, "learning_rate": 1.528214285714286e-05, "loss": 1.4364, "step": 4860500 }, { "epoch": 24.96, "learning_rate": 1.5278571428571428e-05, "loss": 1.433, "step": 4861000 }, { "epoch": 24.96, "learning_rate": 1.5275000000000002e-05, "loss": 1.4283, "step": 4861500 }, { "epoch": 24.96, "learning_rate": 1.5271428571428573e-05, "loss": 1.4342, "step": 4862000 }, { "epoch": 24.97, "learning_rate": 1.5267857142857144e-05, "loss": 1.4264, "step": 4862500 }, { "epoch": 24.97, "learning_rate": 1.5264285714285715e-05, "loss": 1.4113, "step": 4863000 }, { "epoch": 24.97, "learning_rate": 1.5260714285714285e-05, "loss": 1.4341, "step": 4863500 }, { "epoch": 24.97, "learning_rate": 1.5257142857142858e-05, "loss": 1.4269, "step": 4864000 }, { "epoch": 24.98, "learning_rate": 1.5253571428571429e-05, "loss": 1.4326, "step": 4864500 }, { "epoch": 24.98, "learning_rate": 1.525e-05, "loss": 1.4254, "step": 4865000 }, { "epoch": 24.98, "learning_rate": 1.5246428571428572e-05, "loss": 1.4226, "step": 4865500 }, { "epoch": 24.98, "learning_rate": 1.5242857142857145e-05, "loss": 1.4386, "step": 4866000 }, { "epoch": 24.99, "learning_rate": 1.5239285714285714e-05, "loss": 1.4305, "step": 4866500 }, { "epoch": 24.99, "learning_rate": 1.5235714285714286e-05, "loss": 1.4255, "step": 4867000 }, { "epoch": 24.99, "learning_rate": 1.5232142857142859e-05, "loss": 1.4258, "step": 4867500 }, { "epoch": 25.0, "learning_rate": 1.5228571428571428e-05, "loss": 1.419, "step": 4868000 }, { "epoch": 25.0, "learning_rate": 1.5225e-05, "loss": 1.4297, "step": 4868500 }, { "epoch": 25.0, "learning_rate": 1.5221428571428573e-05, "loss": 1.4298, "step": 4869000 }, { "epoch": 25.0, "learning_rate": 1.5217857142857144e-05, "loss": 1.4129, "step": 4869500 }, { "epoch": 25.01, "learning_rate": 1.5214285714285715e-05, "loss": 1.427, "step": 4870000 }, { "epoch": 25.01, "learning_rate": 1.5210714285714286e-05, "loss": 1.4235, "step": 4870500 }, { "epoch": 25.01, "learning_rate": 1.5207142857142858e-05, "loss": 1.4166, "step": 4871000 }, { "epoch": 25.01, "learning_rate": 1.5203571428571428e-05, "loss": 1.425, "step": 4871500 }, { "epoch": 25.02, "learning_rate": 1.52e-05, "loss": 1.4199, "step": 4872000 }, { "epoch": 25.02, "learning_rate": 1.5196428571428573e-05, "loss": 1.4169, "step": 4872500 }, { "epoch": 25.02, "learning_rate": 1.5192857142857145e-05, "loss": 1.4267, "step": 4873000 }, { "epoch": 25.02, "learning_rate": 1.5189285714285714e-05, "loss": 1.4151, "step": 4873500 }, { "epoch": 25.03, "learning_rate": 1.5185714285714287e-05, "loss": 1.4279, "step": 4874000 }, { "epoch": 25.03, "learning_rate": 1.5182142857142858e-05, "loss": 1.4266, "step": 4874500 }, { "epoch": 25.03, "learning_rate": 1.5178571428571429e-05, "loss": 1.4208, "step": 4875000 }, { "epoch": 25.04, "learning_rate": 1.5175e-05, "loss": 1.4261, "step": 4875500 }, { "epoch": 25.04, "learning_rate": 1.5171428571428572e-05, "loss": 1.4143, "step": 4876000 }, { "epoch": 25.04, "learning_rate": 1.5167857142857145e-05, "loss": 1.4251, "step": 4876500 }, { "epoch": 25.04, "learning_rate": 1.5164285714285714e-05, "loss": 1.4177, "step": 4877000 }, { "epoch": 25.05, "learning_rate": 1.5160714285714286e-05, "loss": 1.4187, "step": 4877500 }, { "epoch": 25.05, "learning_rate": 1.5157142857142859e-05, "loss": 1.4203, "step": 4878000 }, { "epoch": 25.05, "learning_rate": 1.5153571428571428e-05, "loss": 1.4202, "step": 4878500 }, { "epoch": 25.05, "learning_rate": 1.515e-05, "loss": 1.4317, "step": 4879000 }, { "epoch": 25.06, "learning_rate": 1.5146428571428573e-05, "loss": 1.4203, "step": 4879500 }, { "epoch": 25.06, "learning_rate": 1.5142857142857144e-05, "loss": 1.4168, "step": 4880000 }, { "epoch": 25.06, "learning_rate": 1.5139285714285715e-05, "loss": 1.4281, "step": 4880500 }, { "epoch": 25.06, "learning_rate": 1.5135714285714286e-05, "loss": 1.4067, "step": 4881000 }, { "epoch": 25.07, "learning_rate": 1.5132142857142858e-05, "loss": 1.4244, "step": 4881500 }, { "epoch": 25.07, "learning_rate": 1.5128571428571427e-05, "loss": 1.4157, "step": 4882000 }, { "epoch": 25.07, "learning_rate": 1.5125e-05, "loss": 1.4116, "step": 4882500 }, { "epoch": 25.08, "learning_rate": 1.5121428571428573e-05, "loss": 1.4152, "step": 4883000 }, { "epoch": 25.08, "learning_rate": 1.5117857142857145e-05, "loss": 1.4208, "step": 4883500 }, { "epoch": 25.08, "learning_rate": 1.5114285714285714e-05, "loss": 1.4183, "step": 4884000 }, { "epoch": 25.08, "learning_rate": 1.5110714285714287e-05, "loss": 1.4027, "step": 4884500 }, { "epoch": 25.09, "learning_rate": 1.510714285714286e-05, "loss": 1.4171, "step": 4885000 }, { "epoch": 25.09, "learning_rate": 1.5103571428571429e-05, "loss": 1.4168, "step": 4885500 }, { "epoch": 25.09, "learning_rate": 1.51e-05, "loss": 1.408, "step": 4886000 }, { "epoch": 25.09, "learning_rate": 1.5096428571428572e-05, "loss": 1.4168, "step": 4886500 }, { "epoch": 25.1, "learning_rate": 1.5092857142857145e-05, "loss": 1.4172, "step": 4887000 }, { "epoch": 25.1, "learning_rate": 1.5089285714285714e-05, "loss": 1.4274, "step": 4887500 }, { "epoch": 25.1, "learning_rate": 1.5085714285714286e-05, "loss": 1.4108, "step": 4888000 }, { "epoch": 25.1, "learning_rate": 1.5082142857142859e-05, "loss": 1.4164, "step": 4888500 }, { "epoch": 25.11, "learning_rate": 1.5078571428571428e-05, "loss": 1.4265, "step": 4889000 }, { "epoch": 25.11, "learning_rate": 1.5075e-05, "loss": 1.4178, "step": 4889500 }, { "epoch": 25.11, "learning_rate": 1.5071428571428573e-05, "loss": 1.4096, "step": 4890000 }, { "epoch": 25.11, "learning_rate": 1.5067857142857144e-05, "loss": 1.4128, "step": 4890500 }, { "epoch": 25.12, "learning_rate": 1.5064285714285715e-05, "loss": 1.4094, "step": 4891000 }, { "epoch": 25.12, "learning_rate": 1.5060714285714286e-05, "loss": 1.4211, "step": 4891500 }, { "epoch": 25.12, "learning_rate": 1.5057142857142858e-05, "loss": 1.4197, "step": 4892000 }, { "epoch": 25.13, "learning_rate": 1.5053571428571427e-05, "loss": 1.4122, "step": 4892500 }, { "epoch": 25.13, "learning_rate": 1.505e-05, "loss": 1.4298, "step": 4893000 }, { "epoch": 25.13, "learning_rate": 1.5046428571428572e-05, "loss": 1.4281, "step": 4893500 }, { "epoch": 25.13, "learning_rate": 1.5042857142857145e-05, "loss": 1.4168, "step": 4894000 }, { "epoch": 25.14, "learning_rate": 1.5039285714285714e-05, "loss": 1.4146, "step": 4894500 }, { "epoch": 25.14, "learning_rate": 1.5035714285714287e-05, "loss": 1.422, "step": 4895000 }, { "epoch": 25.14, "learning_rate": 1.503214285714286e-05, "loss": 1.4191, "step": 4895500 }, { "epoch": 25.14, "learning_rate": 1.5028571428571428e-05, "loss": 1.4175, "step": 4896000 }, { "epoch": 25.15, "learning_rate": 1.5025000000000001e-05, "loss": 1.4108, "step": 4896500 }, { "epoch": 25.15, "learning_rate": 1.5021428571428572e-05, "loss": 1.4225, "step": 4897000 }, { "epoch": 25.15, "learning_rate": 1.5017857142857144e-05, "loss": 1.4256, "step": 4897500 }, { "epoch": 25.15, "learning_rate": 1.5014285714285714e-05, "loss": 1.4046, "step": 4898000 }, { "epoch": 25.16, "learning_rate": 1.5010714285714286e-05, "loss": 1.4202, "step": 4898500 }, { "epoch": 25.16, "learning_rate": 1.5007142857142859e-05, "loss": 1.4103, "step": 4899000 }, { "epoch": 25.16, "learning_rate": 1.5003571428571428e-05, "loss": 1.4195, "step": 4899500 }, { "epoch": 25.17, "learning_rate": 1.5e-05, "loss": 1.4201, "step": 4900000 }, { "epoch": 25.17, "learning_rate": 1.4996428571428573e-05, "loss": 1.4206, "step": 4900500 }, { "epoch": 25.17, "learning_rate": 1.4992857142857144e-05, "loss": 1.4184, "step": 4901000 }, { "epoch": 25.17, "learning_rate": 1.4989285714285715e-05, "loss": 1.4136, "step": 4901500 }, { "epoch": 25.18, "learning_rate": 1.4985714285714286e-05, "loss": 1.4132, "step": 4902000 }, { "epoch": 25.18, "learning_rate": 1.4982142857142858e-05, "loss": 1.4131, "step": 4902500 }, { "epoch": 25.18, "learning_rate": 1.4978571428571427e-05, "loss": 1.4217, "step": 4903000 }, { "epoch": 25.18, "learning_rate": 1.4975e-05, "loss": 1.4151, "step": 4903500 }, { "epoch": 25.19, "learning_rate": 1.4971428571428572e-05, "loss": 1.4152, "step": 4904000 }, { "epoch": 25.19, "learning_rate": 1.4967857142857145e-05, "loss": 1.4159, "step": 4904500 }, { "epoch": 25.19, "learning_rate": 1.4964285714285714e-05, "loss": 1.4158, "step": 4905000 }, { "epoch": 25.19, "learning_rate": 1.4960714285714287e-05, "loss": 1.4224, "step": 4905500 }, { "epoch": 25.2, "learning_rate": 1.4957142857142859e-05, "loss": 1.4182, "step": 4906000 }, { "epoch": 25.2, "learning_rate": 1.495357142857143e-05, "loss": 1.4271, "step": 4906500 }, { "epoch": 25.2, "learning_rate": 1.4950000000000001e-05, "loss": 1.4094, "step": 4907000 }, { "epoch": 25.21, "learning_rate": 1.4946428571428572e-05, "loss": 1.4233, "step": 4907500 }, { "epoch": 25.21, "learning_rate": 1.4942857142857144e-05, "loss": 1.4103, "step": 4908000 }, { "epoch": 25.21, "learning_rate": 1.4939285714285713e-05, "loss": 1.4155, "step": 4908500 }, { "epoch": 25.21, "learning_rate": 1.4935714285714286e-05, "loss": 1.4117, "step": 4909000 }, { "epoch": 25.22, "learning_rate": 1.4932142857142859e-05, "loss": 1.4277, "step": 4909500 }, { "epoch": 25.22, "learning_rate": 1.4928571428571431e-05, "loss": 1.4234, "step": 4910000 }, { "epoch": 25.22, "learning_rate": 1.4925e-05, "loss": 1.4193, "step": 4910500 }, { "epoch": 25.22, "learning_rate": 1.4921428571428573e-05, "loss": 1.4218, "step": 4911000 }, { "epoch": 25.23, "learning_rate": 1.4917857142857145e-05, "loss": 1.4292, "step": 4911500 }, { "epoch": 25.23, "learning_rate": 1.4914285714285715e-05, "loss": 1.416, "step": 4912000 }, { "epoch": 25.23, "learning_rate": 1.4910714285714285e-05, "loss": 1.422, "step": 4912500 }, { "epoch": 25.23, "learning_rate": 1.4907142857142858e-05, "loss": 1.4175, "step": 4913000 }, { "epoch": 25.24, "learning_rate": 1.490357142857143e-05, "loss": 1.425, "step": 4913500 }, { "epoch": 25.24, "learning_rate": 1.49e-05, "loss": 1.4208, "step": 4914000 }, { "epoch": 25.24, "learning_rate": 1.4896428571428572e-05, "loss": 1.4077, "step": 4914500 }, { "epoch": 25.25, "learning_rate": 1.4892857142857145e-05, "loss": 1.4196, "step": 4915000 }, { "epoch": 25.25, "learning_rate": 1.4889285714285714e-05, "loss": 1.4138, "step": 4915500 }, { "epoch": 25.25, "learning_rate": 1.4885714285714286e-05, "loss": 1.4198, "step": 4916000 }, { "epoch": 25.25, "learning_rate": 1.4882142857142859e-05, "loss": 1.4213, "step": 4916500 }, { "epoch": 25.26, "learning_rate": 1.487857142857143e-05, "loss": 1.4278, "step": 4917000 }, { "epoch": 25.26, "learning_rate": 1.4875e-05, "loss": 1.4138, "step": 4917500 }, { "epoch": 25.26, "learning_rate": 1.4871428571428572e-05, "loss": 1.4216, "step": 4918000 }, { "epoch": 25.26, "learning_rate": 1.4867857142857144e-05, "loss": 1.4192, "step": 4918500 }, { "epoch": 25.27, "learning_rate": 1.4864285714285713e-05, "loss": 1.4259, "step": 4919000 }, { "epoch": 25.27, "learning_rate": 1.4860714285714286e-05, "loss": 1.4259, "step": 4919500 }, { "epoch": 25.27, "learning_rate": 1.4857142857142858e-05, "loss": 1.4169, "step": 4920000 }, { "epoch": 25.27, "learning_rate": 1.4853571428571431e-05, "loss": 1.4324, "step": 4920500 }, { "epoch": 25.28, "learning_rate": 1.485e-05, "loss": 1.4174, "step": 4921000 }, { "epoch": 25.28, "learning_rate": 1.4846428571428573e-05, "loss": 1.4139, "step": 4921500 }, { "epoch": 25.28, "learning_rate": 1.4842857142857145e-05, "loss": 1.4243, "step": 4922000 }, { "epoch": 25.29, "learning_rate": 1.4839285714285714e-05, "loss": 1.4163, "step": 4922500 }, { "epoch": 25.29, "learning_rate": 1.4835714285714287e-05, "loss": 1.4239, "step": 4923000 }, { "epoch": 25.29, "learning_rate": 1.4832142857142858e-05, "loss": 1.4236, "step": 4923500 }, { "epoch": 25.29, "learning_rate": 1.482857142857143e-05, "loss": 1.4302, "step": 4924000 }, { "epoch": 25.3, "learning_rate": 1.4825e-05, "loss": 1.4147, "step": 4924500 }, { "epoch": 25.3, "learning_rate": 1.4821428571428572e-05, "loss": 1.433, "step": 4925000 }, { "epoch": 25.3, "learning_rate": 1.4817857142857145e-05, "loss": 1.4173, "step": 4925500 }, { "epoch": 25.3, "learning_rate": 1.4814285714285714e-05, "loss": 1.4287, "step": 4926000 }, { "epoch": 25.31, "learning_rate": 1.4810714285714286e-05, "loss": 1.4265, "step": 4926500 }, { "epoch": 25.31, "learning_rate": 1.4807142857142859e-05, "loss": 1.4184, "step": 4927000 }, { "epoch": 25.31, "learning_rate": 1.480357142857143e-05, "loss": 1.4199, "step": 4927500 }, { "epoch": 25.31, "learning_rate": 1.48e-05, "loss": 1.426, "step": 4928000 }, { "epoch": 25.32, "learning_rate": 1.4796428571428571e-05, "loss": 1.4159, "step": 4928500 }, { "epoch": 25.32, "learning_rate": 1.4792857142857144e-05, "loss": 1.4164, "step": 4929000 }, { "epoch": 25.32, "learning_rate": 1.4789285714285713e-05, "loss": 1.4305, "step": 4929500 }, { "epoch": 25.33, "learning_rate": 1.4785714285714286e-05, "loss": 1.4169, "step": 4930000 }, { "epoch": 25.33, "learning_rate": 1.4782142857142858e-05, "loss": 1.4171, "step": 4930500 }, { "epoch": 25.33, "learning_rate": 1.477857142857143e-05, "loss": 1.4129, "step": 4931000 }, { "epoch": 25.33, "learning_rate": 1.4775e-05, "loss": 1.4241, "step": 4931500 }, { "epoch": 25.34, "learning_rate": 1.4771428571428573e-05, "loss": 1.4142, "step": 4932000 }, { "epoch": 25.34, "learning_rate": 1.4767857142857145e-05, "loss": 1.4159, "step": 4932500 }, { "epoch": 25.34, "learning_rate": 1.4764285714285714e-05, "loss": 1.4213, "step": 4933000 }, { "epoch": 25.34, "learning_rate": 1.4760714285714287e-05, "loss": 1.4096, "step": 4933500 }, { "epoch": 25.35, "learning_rate": 1.4757142857142858e-05, "loss": 1.4257, "step": 4934000 }, { "epoch": 25.35, "learning_rate": 1.475357142857143e-05, "loss": 1.434, "step": 4934500 }, { "epoch": 25.35, "learning_rate": 1.475e-05, "loss": 1.4172, "step": 4935000 }, { "epoch": 25.35, "learning_rate": 1.4746428571428572e-05, "loss": 1.4196, "step": 4935500 }, { "epoch": 25.36, "learning_rate": 1.4742857142857144e-05, "loss": 1.4126, "step": 4936000 }, { "epoch": 25.36, "learning_rate": 1.4739285714285714e-05, "loss": 1.4154, "step": 4936500 }, { "epoch": 25.36, "learning_rate": 1.4735714285714286e-05, "loss": 1.4051, "step": 4937000 }, { "epoch": 25.37, "learning_rate": 1.4732142857142859e-05, "loss": 1.4205, "step": 4937500 }, { "epoch": 25.37, "learning_rate": 1.4728571428571431e-05, "loss": 1.4258, "step": 4938000 }, { "epoch": 25.37, "learning_rate": 1.4725e-05, "loss": 1.4218, "step": 4938500 }, { "epoch": 25.37, "learning_rate": 1.4721428571428571e-05, "loss": 1.4174, "step": 4939000 }, { "epoch": 25.38, "learning_rate": 1.4717857142857144e-05, "loss": 1.4176, "step": 4939500 }, { "epoch": 25.38, "learning_rate": 1.4714285714285713e-05, "loss": 1.4197, "step": 4940000 }, { "epoch": 25.38, "learning_rate": 1.4710714285714286e-05, "loss": 1.4173, "step": 4940500 }, { "epoch": 25.38, "learning_rate": 1.4707142857142858e-05, "loss": 1.42, "step": 4941000 }, { "epoch": 25.39, "learning_rate": 1.470357142857143e-05, "loss": 1.4219, "step": 4941500 }, { "epoch": 25.39, "learning_rate": 1.47e-05, "loss": 1.4191, "step": 4942000 }, { "epoch": 25.39, "learning_rate": 1.4696428571428572e-05, "loss": 1.4176, "step": 4942500 }, { "epoch": 25.39, "learning_rate": 1.4692857142857145e-05, "loss": 1.4221, "step": 4943000 }, { "epoch": 25.4, "learning_rate": 1.4689285714285714e-05, "loss": 1.4268, "step": 4943500 }, { "epoch": 25.4, "learning_rate": 1.4685714285714287e-05, "loss": 1.4123, "step": 4944000 }, { "epoch": 25.4, "learning_rate": 1.4682142857142858e-05, "loss": 1.4208, "step": 4944500 }, { "epoch": 25.41, "learning_rate": 1.467857142857143e-05, "loss": 1.4153, "step": 4945000 }, { "epoch": 25.41, "learning_rate": 1.4675e-05, "loss": 1.4179, "step": 4945500 }, { "epoch": 25.41, "learning_rate": 1.4671428571428572e-05, "loss": 1.4171, "step": 4946000 }, { "epoch": 25.41, "learning_rate": 1.4667857142857144e-05, "loss": 1.4154, "step": 4946500 }, { "epoch": 25.42, "learning_rate": 1.4664285714285714e-05, "loss": 1.4163, "step": 4947000 }, { "epoch": 25.42, "learning_rate": 1.4660714285714286e-05, "loss": 1.4166, "step": 4947500 }, { "epoch": 25.42, "learning_rate": 1.4657142857142859e-05, "loss": 1.426, "step": 4948000 }, { "epoch": 25.42, "learning_rate": 1.4653571428571431e-05, "loss": 1.4118, "step": 4948500 }, { "epoch": 25.43, "learning_rate": 1.465e-05, "loss": 1.4161, "step": 4949000 }, { "epoch": 25.43, "learning_rate": 1.4646428571428573e-05, "loss": 1.4179, "step": 4949500 }, { "epoch": 25.43, "learning_rate": 1.4642857142857144e-05, "loss": 1.418, "step": 4950000 }, { "epoch": 25.43, "learning_rate": 1.4639285714285713e-05, "loss": 1.4139, "step": 4950500 }, { "epoch": 25.44, "learning_rate": 1.4635714285714285e-05, "loss": 1.4264, "step": 4951000 }, { "epoch": 25.44, "learning_rate": 1.4632142857142858e-05, "loss": 1.4193, "step": 4951500 }, { "epoch": 25.44, "learning_rate": 1.462857142857143e-05, "loss": 1.416, "step": 4952000 }, { "epoch": 25.45, "learning_rate": 1.4625e-05, "loss": 1.4174, "step": 4952500 }, { "epoch": 25.45, "learning_rate": 1.4621428571428572e-05, "loss": 1.415, "step": 4953000 }, { "epoch": 25.45, "learning_rate": 1.4617857142857145e-05, "loss": 1.4111, "step": 4953500 }, { "epoch": 25.45, "learning_rate": 1.4614285714285714e-05, "loss": 1.4175, "step": 4954000 }, { "epoch": 25.46, "learning_rate": 1.4610714285714287e-05, "loss": 1.4171, "step": 4954500 }, { "epoch": 25.46, "learning_rate": 1.4607142857142857e-05, "loss": 1.4152, "step": 4955000 }, { "epoch": 25.46, "learning_rate": 1.460357142857143e-05, "loss": 1.4154, "step": 4955500 }, { "epoch": 25.46, "learning_rate": 1.4599999999999999e-05, "loss": 1.4101, "step": 4956000 }, { "epoch": 25.47, "learning_rate": 1.4596428571428572e-05, "loss": 1.4118, "step": 4956500 }, { "epoch": 25.47, "learning_rate": 1.4592857142857144e-05, "loss": 1.4171, "step": 4957000 }, { "epoch": 25.47, "learning_rate": 1.4589285714285713e-05, "loss": 1.4176, "step": 4957500 }, { "epoch": 25.47, "learning_rate": 1.4585714285714286e-05, "loss": 1.4153, "step": 4958000 }, { "epoch": 25.48, "learning_rate": 1.4582142857142858e-05, "loss": 1.4274, "step": 4958500 }, { "epoch": 25.48, "learning_rate": 1.4578571428571431e-05, "loss": 1.4262, "step": 4959000 }, { "epoch": 25.48, "learning_rate": 1.4575e-05, "loss": 1.4179, "step": 4959500 }, { "epoch": 25.49, "learning_rate": 1.4571428571428573e-05, "loss": 1.4199, "step": 4960000 }, { "epoch": 25.49, "learning_rate": 1.4567857142857144e-05, "loss": 1.4248, "step": 4960500 }, { "epoch": 25.49, "learning_rate": 1.4564285714285714e-05, "loss": 1.4162, "step": 4961000 }, { "epoch": 25.49, "learning_rate": 1.4560714285714285e-05, "loss": 1.4189, "step": 4961500 }, { "epoch": 25.5, "learning_rate": 1.4557142857142858e-05, "loss": 1.4169, "step": 4962000 }, { "epoch": 25.5, "learning_rate": 1.455357142857143e-05, "loss": 1.4182, "step": 4962500 }, { "epoch": 25.5, "learning_rate": 1.455e-05, "loss": 1.4077, "step": 4963000 }, { "epoch": 25.5, "learning_rate": 1.4546428571428572e-05, "loss": 1.4115, "step": 4963500 }, { "epoch": 25.51, "learning_rate": 1.4542857142857145e-05, "loss": 1.4151, "step": 4964000 }, { "epoch": 25.51, "learning_rate": 1.4539285714285714e-05, "loss": 1.426, "step": 4964500 }, { "epoch": 25.51, "learning_rate": 1.4535714285714286e-05, "loss": 1.4225, "step": 4965000 }, { "epoch": 25.51, "learning_rate": 1.4532142857142857e-05, "loss": 1.4271, "step": 4965500 }, { "epoch": 25.52, "learning_rate": 1.452857142857143e-05, "loss": 1.4131, "step": 4966000 }, { "epoch": 25.52, "learning_rate": 1.4524999999999999e-05, "loss": 1.4096, "step": 4966500 }, { "epoch": 25.52, "learning_rate": 1.4521428571428572e-05, "loss": 1.4117, "step": 4967000 }, { "epoch": 25.52, "learning_rate": 1.4517857142857144e-05, "loss": 1.4189, "step": 4967500 }, { "epoch": 25.53, "learning_rate": 1.4514285714285713e-05, "loss": 1.4162, "step": 4968000 }, { "epoch": 25.53, "learning_rate": 1.4510714285714286e-05, "loss": 1.4146, "step": 4968500 }, { "epoch": 25.53, "learning_rate": 1.4507142857142858e-05, "loss": 1.4261, "step": 4969000 }, { "epoch": 25.54, "learning_rate": 1.4503571428571431e-05, "loss": 1.4127, "step": 4969500 }, { "epoch": 25.54, "learning_rate": 1.45e-05, "loss": 1.4279, "step": 4970000 }, { "epoch": 25.54, "learning_rate": 1.4496428571428573e-05, "loss": 1.4194, "step": 4970500 }, { "epoch": 25.54, "learning_rate": 1.4492857142857143e-05, "loss": 1.4146, "step": 4971000 }, { "epoch": 25.55, "learning_rate": 1.4489285714285716e-05, "loss": 1.4211, "step": 4971500 }, { "epoch": 25.55, "learning_rate": 1.4485714285714285e-05, "loss": 1.4142, "step": 4972000 }, { "epoch": 25.55, "learning_rate": 1.4482142857142858e-05, "loss": 1.4167, "step": 4972500 }, { "epoch": 25.55, "learning_rate": 1.447857142857143e-05, "loss": 1.4198, "step": 4973000 }, { "epoch": 25.56, "learning_rate": 1.4475e-05, "loss": 1.4203, "step": 4973500 }, { "epoch": 25.56, "learning_rate": 1.4471428571428572e-05, "loss": 1.4192, "step": 4974000 }, { "epoch": 25.56, "learning_rate": 1.4467857142857145e-05, "loss": 1.4141, "step": 4974500 }, { "epoch": 25.56, "learning_rate": 1.4464285714285717e-05, "loss": 1.4221, "step": 4975000 }, { "epoch": 25.57, "learning_rate": 1.4460714285714286e-05, "loss": 1.429, "step": 4975500 }, { "epoch": 25.57, "learning_rate": 1.4457142857142857e-05, "loss": 1.4148, "step": 4976000 }, { "epoch": 25.57, "learning_rate": 1.445357142857143e-05, "loss": 1.4217, "step": 4976500 }, { "epoch": 25.58, "learning_rate": 1.4449999999999999e-05, "loss": 1.4218, "step": 4977000 }, { "epoch": 25.58, "learning_rate": 1.4446428571428571e-05, "loss": 1.4179, "step": 4977500 }, { "epoch": 25.58, "learning_rate": 1.4442857142857144e-05, "loss": 1.4262, "step": 4978000 }, { "epoch": 25.58, "learning_rate": 1.4439285714285717e-05, "loss": 1.4209, "step": 4978500 }, { "epoch": 25.59, "learning_rate": 1.4435714285714286e-05, "loss": 1.4187, "step": 4979000 }, { "epoch": 25.59, "learning_rate": 1.4432142857142858e-05, "loss": 1.4179, "step": 4979500 }, { "epoch": 25.59, "learning_rate": 1.442857142857143e-05, "loss": 1.4188, "step": 4980000 }, { "epoch": 25.59, "learning_rate": 1.4425e-05, "loss": 1.4184, "step": 4980500 }, { "epoch": 25.6, "learning_rate": 1.4421428571428573e-05, "loss": 1.4145, "step": 4981000 }, { "epoch": 25.6, "learning_rate": 1.4417857142857143e-05, "loss": 1.4159, "step": 4981500 }, { "epoch": 25.6, "learning_rate": 1.4414285714285716e-05, "loss": 1.4182, "step": 4982000 }, { "epoch": 25.6, "learning_rate": 1.4410714285714285e-05, "loss": 1.4094, "step": 4982500 }, { "epoch": 25.61, "learning_rate": 1.4407142857142858e-05, "loss": 1.4319, "step": 4983000 }, { "epoch": 25.61, "learning_rate": 1.440357142857143e-05, "loss": 1.4261, "step": 4983500 }, { "epoch": 25.61, "learning_rate": 1.44e-05, "loss": 1.4246, "step": 4984000 }, { "epoch": 25.62, "learning_rate": 1.4396428571428572e-05, "loss": 1.4379, "step": 4984500 }, { "epoch": 25.62, "learning_rate": 1.4392857142857144e-05, "loss": 1.4168, "step": 4985000 }, { "epoch": 25.62, "learning_rate": 1.4389285714285717e-05, "loss": 1.4156, "step": 4985500 }, { "epoch": 25.62, "learning_rate": 1.4385714285714286e-05, "loss": 1.4175, "step": 4986000 }, { "epoch": 25.63, "learning_rate": 1.4382142857142859e-05, "loss": 1.4168, "step": 4986500 }, { "epoch": 25.63, "learning_rate": 1.437857142857143e-05, "loss": 1.4166, "step": 4987000 }, { "epoch": 25.63, "learning_rate": 1.4374999999999999e-05, "loss": 1.4209, "step": 4987500 }, { "epoch": 25.63, "learning_rate": 1.4371428571428571e-05, "loss": 1.4273, "step": 4988000 }, { "epoch": 25.64, "learning_rate": 1.4367857142857144e-05, "loss": 1.418, "step": 4988500 }, { "epoch": 25.64, "learning_rate": 1.4364285714285716e-05, "loss": 1.42, "step": 4989000 }, { "epoch": 25.64, "learning_rate": 1.4360714285714286e-05, "loss": 1.42, "step": 4989500 }, { "epoch": 25.64, "learning_rate": 1.4357142857142858e-05, "loss": 1.4143, "step": 4990000 }, { "epoch": 25.65, "learning_rate": 1.435357142857143e-05, "loss": 1.4169, "step": 4990500 }, { "epoch": 25.65, "learning_rate": 1.435e-05, "loss": 1.4134, "step": 4991000 }, { "epoch": 25.65, "learning_rate": 1.4346428571428572e-05, "loss": 1.4193, "step": 4991500 }, { "epoch": 25.66, "learning_rate": 1.4342857142857143e-05, "loss": 1.417, "step": 4992000 }, { "epoch": 25.66, "learning_rate": 1.4339285714285716e-05, "loss": 1.4212, "step": 4992500 }, { "epoch": 25.66, "learning_rate": 1.4335714285714285e-05, "loss": 1.4099, "step": 4993000 }, { "epoch": 25.66, "learning_rate": 1.4332142857142858e-05, "loss": 1.4167, "step": 4993500 }, { "epoch": 25.67, "learning_rate": 1.432857142857143e-05, "loss": 1.4131, "step": 4994000 }, { "epoch": 25.67, "learning_rate": 1.4325e-05, "loss": 1.4096, "step": 4994500 }, { "epoch": 25.67, "learning_rate": 1.4321428571428572e-05, "loss": 1.4183, "step": 4995000 }, { "epoch": 25.67, "learning_rate": 1.4317857142857144e-05, "loss": 1.4105, "step": 4995500 }, { "epoch": 25.68, "learning_rate": 1.4314285714285717e-05, "loss": 1.4048, "step": 4996000 }, { "epoch": 25.68, "learning_rate": 1.4310714285714286e-05, "loss": 1.4104, "step": 4996500 }, { "epoch": 25.68, "learning_rate": 1.4307142857142859e-05, "loss": 1.4222, "step": 4997000 }, { "epoch": 25.68, "learning_rate": 1.430357142857143e-05, "loss": 1.4058, "step": 4997500 }, { "epoch": 25.69, "learning_rate": 1.43e-05, "loss": 1.4233, "step": 4998000 }, { "epoch": 25.69, "learning_rate": 1.4296428571428571e-05, "loss": 1.4125, "step": 4998500 }, { "epoch": 25.69, "learning_rate": 1.4292857142857144e-05, "loss": 1.4133, "step": 4999000 }, { "epoch": 25.7, "learning_rate": 1.4289285714285716e-05, "loss": 1.4175, "step": 4999500 }, { "epoch": 25.7, "learning_rate": 1.4285714285714285e-05, "loss": 1.4165, "step": 5000000 }, { "epoch": 25.7, "learning_rate": 1.4282142857142858e-05, "loss": 1.4166, "step": 5000500 }, { "epoch": 25.7, "learning_rate": 1.427857142857143e-05, "loss": 1.4188, "step": 5001000 }, { "epoch": 25.71, "learning_rate": 1.4275e-05, "loss": 1.4196, "step": 5001500 }, { "epoch": 25.71, "learning_rate": 1.4271428571428572e-05, "loss": 1.4198, "step": 5002000 }, { "epoch": 25.71, "learning_rate": 1.4267857142857143e-05, "loss": 1.4187, "step": 5002500 }, { "epoch": 25.71, "learning_rate": 1.4264285714285716e-05, "loss": 1.4159, "step": 5003000 }, { "epoch": 25.72, "learning_rate": 1.4260714285714285e-05, "loss": 1.4172, "step": 5003500 }, { "epoch": 25.72, "learning_rate": 1.4257142857142857e-05, "loss": 1.4217, "step": 5004000 }, { "epoch": 25.72, "learning_rate": 1.425357142857143e-05, "loss": 1.4102, "step": 5004500 }, { "epoch": 25.72, "learning_rate": 1.4249999999999999e-05, "loss": 1.4095, "step": 5005000 }, { "epoch": 25.73, "learning_rate": 1.4246428571428572e-05, "loss": 1.4043, "step": 5005500 }, { "epoch": 25.73, "learning_rate": 1.4242857142857144e-05, "loss": 1.4145, "step": 5006000 }, { "epoch": 25.73, "learning_rate": 1.4239285714285717e-05, "loss": 1.4119, "step": 5006500 }, { "epoch": 25.74, "learning_rate": 1.4235714285714286e-05, "loss": 1.4065, "step": 5007000 }, { "epoch": 25.74, "learning_rate": 1.4232142857142858e-05, "loss": 1.4049, "step": 5007500 }, { "epoch": 25.74, "learning_rate": 1.422857142857143e-05, "loss": 1.4112, "step": 5008000 }, { "epoch": 25.74, "learning_rate": 1.4225e-05, "loss": 1.4264, "step": 5008500 }, { "epoch": 25.75, "learning_rate": 1.4221428571428571e-05, "loss": 1.4215, "step": 5009000 }, { "epoch": 25.75, "learning_rate": 1.4217857142857144e-05, "loss": 1.4144, "step": 5009500 }, { "epoch": 25.75, "learning_rate": 1.4214285714285716e-05, "loss": 1.4124, "step": 5010000 }, { "epoch": 25.75, "learning_rate": 1.4210714285714285e-05, "loss": 1.4204, "step": 5010500 }, { "epoch": 25.76, "learning_rate": 1.4207142857142858e-05, "loss": 1.4036, "step": 5011000 }, { "epoch": 25.76, "learning_rate": 1.420357142857143e-05, "loss": 1.4189, "step": 5011500 }, { "epoch": 25.76, "learning_rate": 1.42e-05, "loss": 1.4226, "step": 5012000 }, { "epoch": 25.76, "learning_rate": 1.4196428571428572e-05, "loss": 1.407, "step": 5012500 }, { "epoch": 25.77, "learning_rate": 1.4192857142857145e-05, "loss": 1.4168, "step": 5013000 }, { "epoch": 25.77, "learning_rate": 1.4189285714285716e-05, "loss": 1.4146, "step": 5013500 }, { "epoch": 25.77, "learning_rate": 1.4185714285714285e-05, "loss": 1.4136, "step": 5014000 }, { "epoch": 25.78, "learning_rate": 1.4182142857142857e-05, "loss": 1.4092, "step": 5014500 }, { "epoch": 25.78, "learning_rate": 1.417857142857143e-05, "loss": 1.4104, "step": 5015000 }, { "epoch": 25.78, "learning_rate": 1.4174999999999999e-05, "loss": 1.4106, "step": 5015500 }, { "epoch": 25.78, "learning_rate": 1.4171428571428572e-05, "loss": 1.4189, "step": 5016000 }, { "epoch": 25.79, "learning_rate": 1.4167857142857144e-05, "loss": 1.4181, "step": 5016500 }, { "epoch": 25.79, "learning_rate": 1.4164285714285717e-05, "loss": 1.4129, "step": 5017000 }, { "epoch": 25.79, "learning_rate": 1.4160714285714286e-05, "loss": 1.4227, "step": 5017500 }, { "epoch": 25.79, "learning_rate": 1.4157142857142858e-05, "loss": 1.4076, "step": 5018000 }, { "epoch": 25.8, "learning_rate": 1.415357142857143e-05, "loss": 1.4219, "step": 5018500 }, { "epoch": 25.8, "learning_rate": 1.415e-05, "loss": 1.4165, "step": 5019000 }, { "epoch": 25.8, "learning_rate": 1.4146428571428571e-05, "loss": 1.4087, "step": 5019500 }, { "epoch": 25.8, "learning_rate": 1.4142857142857143e-05, "loss": 1.4161, "step": 5020000 }, { "epoch": 25.81, "learning_rate": 1.4139285714285716e-05, "loss": 1.4217, "step": 5020500 }, { "epoch": 25.81, "learning_rate": 1.4135714285714285e-05, "loss": 1.4182, "step": 5021000 }, { "epoch": 25.81, "learning_rate": 1.4132142857142858e-05, "loss": 1.419, "step": 5021500 }, { "epoch": 25.82, "learning_rate": 1.412857142857143e-05, "loss": 1.4117, "step": 5022000 }, { "epoch": 25.82, "learning_rate": 1.4125e-05, "loss": 1.4092, "step": 5022500 }, { "epoch": 25.82, "learning_rate": 1.4121428571428572e-05, "loss": 1.4213, "step": 5023000 }, { "epoch": 25.82, "learning_rate": 1.4117857142857145e-05, "loss": 1.4165, "step": 5023500 }, { "epoch": 25.83, "learning_rate": 1.4114285714285715e-05, "loss": 1.4175, "step": 5024000 }, { "epoch": 25.83, "learning_rate": 1.4110714285714286e-05, "loss": 1.417, "step": 5024500 }, { "epoch": 25.83, "learning_rate": 1.4107142857142857e-05, "loss": 1.4171, "step": 5025000 }, { "epoch": 25.83, "learning_rate": 1.410357142857143e-05, "loss": 1.4132, "step": 5025500 }, { "epoch": 25.84, "learning_rate": 1.4099999999999999e-05, "loss": 1.4107, "step": 5026000 }, { "epoch": 25.84, "learning_rate": 1.4096428571428571e-05, "loss": 1.4124, "step": 5026500 }, { "epoch": 25.84, "learning_rate": 1.4092857142857144e-05, "loss": 1.4202, "step": 5027000 }, { "epoch": 25.84, "learning_rate": 1.4089285714285716e-05, "loss": 1.4111, "step": 5027500 }, { "epoch": 25.85, "learning_rate": 1.4085714285714286e-05, "loss": 1.4212, "step": 5028000 }, { "epoch": 25.85, "learning_rate": 1.4082142857142858e-05, "loss": 1.4163, "step": 5028500 }, { "epoch": 25.85, "learning_rate": 1.4078571428571429e-05, "loss": 1.4101, "step": 5029000 }, { "epoch": 25.86, "learning_rate": 1.4075e-05, "loss": 1.418, "step": 5029500 }, { "epoch": 25.86, "learning_rate": 1.407142857142857e-05, "loss": 1.4038, "step": 5030000 }, { "epoch": 25.86, "learning_rate": 1.4067857142857143e-05, "loss": 1.421, "step": 5030500 }, { "epoch": 25.86, "learning_rate": 1.4064285714285716e-05, "loss": 1.416, "step": 5031000 }, { "epoch": 25.87, "learning_rate": 1.4060714285714285e-05, "loss": 1.4219, "step": 5031500 }, { "epoch": 25.87, "learning_rate": 1.4057142857142858e-05, "loss": 1.4149, "step": 5032000 }, { "epoch": 25.87, "learning_rate": 1.405357142857143e-05, "loss": 1.4283, "step": 5032500 }, { "epoch": 25.87, "learning_rate": 1.4050000000000003e-05, "loss": 1.4256, "step": 5033000 }, { "epoch": 25.88, "learning_rate": 1.4046428571428572e-05, "loss": 1.4168, "step": 5033500 }, { "epoch": 25.88, "learning_rate": 1.4042857142857144e-05, "loss": 1.4146, "step": 5034000 }, { "epoch": 25.88, "learning_rate": 1.4039285714285715e-05, "loss": 1.4199, "step": 5034500 }, { "epoch": 25.88, "learning_rate": 1.4035714285714286e-05, "loss": 1.4186, "step": 5035000 }, { "epoch": 25.89, "learning_rate": 1.4032142857142857e-05, "loss": 1.4155, "step": 5035500 }, { "epoch": 25.89, "learning_rate": 1.402857142857143e-05, "loss": 1.4128, "step": 5036000 }, { "epoch": 25.89, "learning_rate": 1.4025000000000002e-05, "loss": 1.4184, "step": 5036500 }, { "epoch": 25.9, "learning_rate": 1.4021428571428571e-05, "loss": 1.4388, "step": 5037000 }, { "epoch": 25.9, "learning_rate": 1.4017857142857144e-05, "loss": 1.4301, "step": 5037500 }, { "epoch": 25.9, "learning_rate": 1.4014285714285716e-05, "loss": 1.4256, "step": 5038000 }, { "epoch": 25.9, "learning_rate": 1.4010714285714286e-05, "loss": 1.432, "step": 5038500 }, { "epoch": 25.91, "learning_rate": 1.4007142857142858e-05, "loss": 1.4114, "step": 5039000 }, { "epoch": 25.91, "learning_rate": 1.400357142857143e-05, "loss": 1.4099, "step": 5039500 }, { "epoch": 25.91, "learning_rate": 1.4000000000000001e-05, "loss": 1.4142, "step": 5040000 }, { "epoch": 25.91, "learning_rate": 1.399642857142857e-05, "loss": 1.411, "step": 5040500 }, { "epoch": 25.92, "learning_rate": 1.3992857142857143e-05, "loss": 1.4213, "step": 5041000 }, { "epoch": 25.92, "learning_rate": 1.3989285714285716e-05, "loss": 1.4226, "step": 5041500 }, { "epoch": 25.92, "learning_rate": 1.3985714285714285e-05, "loss": 1.4186, "step": 5042000 }, { "epoch": 25.92, "learning_rate": 1.3982142857142857e-05, "loss": 1.4177, "step": 5042500 }, { "epoch": 25.93, "learning_rate": 1.397857142857143e-05, "loss": 1.4137, "step": 5043000 }, { "epoch": 25.93, "learning_rate": 1.3975000000000003e-05, "loss": 1.4197, "step": 5043500 }, { "epoch": 25.93, "learning_rate": 1.3971428571428572e-05, "loss": 1.4245, "step": 5044000 }, { "epoch": 25.93, "learning_rate": 1.3967857142857144e-05, "loss": 1.4061, "step": 5044500 }, { "epoch": 25.94, "learning_rate": 1.3964285714285715e-05, "loss": 1.4179, "step": 5045000 }, { "epoch": 25.94, "learning_rate": 1.3960714285714286e-05, "loss": 1.4236, "step": 5045500 }, { "epoch": 25.94, "learning_rate": 1.3957142857142857e-05, "loss": 1.4127, "step": 5046000 }, { "epoch": 25.95, "learning_rate": 1.395357142857143e-05, "loss": 1.4071, "step": 5046500 }, { "epoch": 25.95, "learning_rate": 1.3950000000000002e-05, "loss": 1.4208, "step": 5047000 }, { "epoch": 25.95, "learning_rate": 1.3946428571428571e-05, "loss": 1.4187, "step": 5047500 }, { "epoch": 25.95, "learning_rate": 1.3942857142857144e-05, "loss": 1.4101, "step": 5048000 }, { "epoch": 25.96, "learning_rate": 1.3939285714285716e-05, "loss": 1.4223, "step": 5048500 }, { "epoch": 25.96, "learning_rate": 1.3935714285714285e-05, "loss": 1.4124, "step": 5049000 }, { "epoch": 25.96, "learning_rate": 1.3932142857142858e-05, "loss": 1.4021, "step": 5049500 }, { "epoch": 25.96, "learning_rate": 1.392857142857143e-05, "loss": 1.4289, "step": 5050000 }, { "epoch": 25.97, "learning_rate": 1.3925000000000001e-05, "loss": 1.4142, "step": 5050500 }, { "epoch": 25.97, "learning_rate": 1.3921428571428572e-05, "loss": 1.4233, "step": 5051000 }, { "epoch": 25.97, "learning_rate": 1.3917857142857143e-05, "loss": 1.4234, "step": 5051500 }, { "epoch": 25.97, "learning_rate": 1.3914285714285716e-05, "loss": 1.4157, "step": 5052000 }, { "epoch": 25.98, "learning_rate": 1.3910714285714285e-05, "loss": 1.4165, "step": 5052500 }, { "epoch": 25.98, "learning_rate": 1.3907142857142857e-05, "loss": 1.4139, "step": 5053000 }, { "epoch": 25.98, "learning_rate": 1.390357142857143e-05, "loss": 1.4238, "step": 5053500 }, { "epoch": 25.99, "learning_rate": 1.3900000000000002e-05, "loss": 1.4256, "step": 5054000 }, { "epoch": 25.99, "learning_rate": 1.3896428571428572e-05, "loss": 1.4281, "step": 5054500 }, { "epoch": 25.99, "learning_rate": 1.3892857142857144e-05, "loss": 1.4353, "step": 5055000 }, { "epoch": 25.99, "learning_rate": 1.3889285714285715e-05, "loss": 1.419, "step": 5055500 }, { "epoch": 26.0, "learning_rate": 1.3885714285714286e-05, "loss": 1.428, "step": 5056000 }, { "epoch": 26.0, "learning_rate": 1.3882142857142857e-05, "loss": 1.4241, "step": 5056500 }, { "epoch": 26.0, "learning_rate": 1.387857142857143e-05, "loss": 1.414, "step": 5057000 }, { "epoch": 26.0, "learning_rate": 1.3875000000000002e-05, "loss": 1.407, "step": 5057500 }, { "epoch": 26.01, "learning_rate": 1.3871428571428571e-05, "loss": 1.4179, "step": 5058000 }, { "epoch": 26.01, "learning_rate": 1.3867857142857144e-05, "loss": 1.4047, "step": 5058500 }, { "epoch": 26.01, "learning_rate": 1.3864285714285716e-05, "loss": 1.4044, "step": 5059000 }, { "epoch": 26.01, "learning_rate": 1.3860714285714285e-05, "loss": 1.409, "step": 5059500 }, { "epoch": 26.02, "learning_rate": 1.3857142857142858e-05, "loss": 1.4134, "step": 5060000 }, { "epoch": 26.02, "learning_rate": 1.385357142857143e-05, "loss": 1.4214, "step": 5060500 }, { "epoch": 26.02, "learning_rate": 1.3850000000000001e-05, "loss": 1.4123, "step": 5061000 }, { "epoch": 26.03, "learning_rate": 1.3846428571428572e-05, "loss": 1.4177, "step": 5061500 }, { "epoch": 26.03, "learning_rate": 1.3842857142857143e-05, "loss": 1.4064, "step": 5062000 }, { "epoch": 26.03, "learning_rate": 1.3839285714285715e-05, "loss": 1.4033, "step": 5062500 }, { "epoch": 26.03, "learning_rate": 1.3835714285714285e-05, "loss": 1.4111, "step": 5063000 }, { "epoch": 26.04, "learning_rate": 1.3832142857142857e-05, "loss": 1.3995, "step": 5063500 }, { "epoch": 26.04, "learning_rate": 1.382857142857143e-05, "loss": 1.4118, "step": 5064000 }, { "epoch": 26.04, "learning_rate": 1.3825000000000002e-05, "loss": 1.4021, "step": 5064500 }, { "epoch": 26.04, "learning_rate": 1.3821428571428571e-05, "loss": 1.4176, "step": 5065000 }, { "epoch": 26.05, "learning_rate": 1.3817857142857144e-05, "loss": 1.414, "step": 5065500 }, { "epoch": 26.05, "learning_rate": 1.3814285714285715e-05, "loss": 1.4109, "step": 5066000 }, { "epoch": 26.05, "learning_rate": 1.3810714285714286e-05, "loss": 1.4045, "step": 5066500 }, { "epoch": 26.05, "learning_rate": 1.3807142857142857e-05, "loss": 1.4106, "step": 5067000 }, { "epoch": 26.06, "learning_rate": 1.380357142857143e-05, "loss": 1.4215, "step": 5067500 }, { "epoch": 26.06, "learning_rate": 1.3800000000000002e-05, "loss": 1.4142, "step": 5068000 }, { "epoch": 26.06, "learning_rate": 1.3796428571428571e-05, "loss": 1.4163, "step": 5068500 }, { "epoch": 26.07, "learning_rate": 1.3792857142857143e-05, "loss": 1.4193, "step": 5069000 }, { "epoch": 26.07, "learning_rate": 1.3789285714285716e-05, "loss": 1.4154, "step": 5069500 }, { "epoch": 26.07, "learning_rate": 1.3785714285714285e-05, "loss": 1.4107, "step": 5070000 }, { "epoch": 26.07, "learning_rate": 1.3782142857142858e-05, "loss": 1.4139, "step": 5070500 }, { "epoch": 26.08, "learning_rate": 1.377857142857143e-05, "loss": 1.4101, "step": 5071000 }, { "epoch": 26.08, "learning_rate": 1.3775000000000001e-05, "loss": 1.4184, "step": 5071500 }, { "epoch": 26.08, "learning_rate": 1.3771428571428572e-05, "loss": 1.4199, "step": 5072000 }, { "epoch": 26.08, "learning_rate": 1.3767857142857143e-05, "loss": 1.416, "step": 5072500 }, { "epoch": 26.09, "learning_rate": 1.3764285714285715e-05, "loss": 1.4042, "step": 5073000 }, { "epoch": 26.09, "learning_rate": 1.3760714285714285e-05, "loss": 1.4164, "step": 5073500 }, { "epoch": 26.09, "learning_rate": 1.3757142857142857e-05, "loss": 1.4177, "step": 5074000 }, { "epoch": 26.09, "learning_rate": 1.375357142857143e-05, "loss": 1.4525, "step": 5074500 }, { "epoch": 26.1, "learning_rate": 1.3750000000000002e-05, "loss": 1.426, "step": 5075000 }, { "epoch": 26.1, "learning_rate": 1.3746428571428571e-05, "loss": 1.4244, "step": 5075500 }, { "epoch": 26.1, "learning_rate": 1.3742857142857144e-05, "loss": 1.4564, "step": 5076000 }, { "epoch": 26.11, "learning_rate": 1.3739285714285716e-05, "loss": 1.4612, "step": 5076500 }, { "epoch": 26.11, "learning_rate": 1.3735714285714286e-05, "loss": 1.4766, "step": 5077000 }, { "epoch": 26.11, "learning_rate": 1.3732142857142856e-05, "loss": 1.4313, "step": 5077500 }, { "epoch": 26.11, "learning_rate": 1.3728571428571429e-05, "loss": 1.4157, "step": 5078000 }, { "epoch": 26.12, "learning_rate": 1.3725000000000002e-05, "loss": 1.4283, "step": 5078500 }, { "epoch": 26.12, "learning_rate": 1.372142857142857e-05, "loss": 1.4187, "step": 5079000 }, { "epoch": 26.12, "learning_rate": 1.3717857142857143e-05, "loss": 1.4172, "step": 5079500 }, { "epoch": 26.12, "learning_rate": 1.3714285714285716e-05, "loss": 1.4147, "step": 5080000 }, { "epoch": 26.13, "learning_rate": 1.3710714285714285e-05, "loss": 1.4116, "step": 5080500 }, { "epoch": 26.13, "learning_rate": 1.3707142857142858e-05, "loss": 1.4087, "step": 5081000 }, { "epoch": 26.13, "learning_rate": 1.370357142857143e-05, "loss": 1.4077, "step": 5081500 }, { "epoch": 26.13, "learning_rate": 1.3700000000000001e-05, "loss": 1.4171, "step": 5082000 }, { "epoch": 26.14, "learning_rate": 1.3696428571428572e-05, "loss": 1.4094, "step": 5082500 }, { "epoch": 26.14, "learning_rate": 1.3692857142857143e-05, "loss": 1.4183, "step": 5083000 }, { "epoch": 26.14, "learning_rate": 1.3689285714285715e-05, "loss": 1.4131, "step": 5083500 }, { "epoch": 26.15, "learning_rate": 1.3685714285714284e-05, "loss": 1.4074, "step": 5084000 }, { "epoch": 26.15, "learning_rate": 1.3682142857142857e-05, "loss": 1.4122, "step": 5084500 }, { "epoch": 26.15, "learning_rate": 1.367857142857143e-05, "loss": 1.4094, "step": 5085000 }, { "epoch": 26.15, "learning_rate": 1.3675000000000002e-05, "loss": 1.4117, "step": 5085500 }, { "epoch": 26.16, "learning_rate": 1.3671428571428571e-05, "loss": 1.4025, "step": 5086000 }, { "epoch": 26.16, "learning_rate": 1.3667857142857144e-05, "loss": 1.4091, "step": 5086500 }, { "epoch": 26.16, "learning_rate": 1.3664285714285716e-05, "loss": 1.4128, "step": 5087000 }, { "epoch": 26.16, "learning_rate": 1.3660714285714285e-05, "loss": 1.4101, "step": 5087500 }, { "epoch": 26.17, "learning_rate": 1.3657142857142858e-05, "loss": 1.4152, "step": 5088000 }, { "epoch": 26.17, "learning_rate": 1.3653571428571429e-05, "loss": 1.4065, "step": 5088500 }, { "epoch": 26.17, "learning_rate": 1.3650000000000001e-05, "loss": 1.4041, "step": 5089000 }, { "epoch": 26.17, "learning_rate": 1.364642857142857e-05, "loss": 1.4159, "step": 5089500 }, { "epoch": 26.18, "learning_rate": 1.3642857142857143e-05, "loss": 1.4145, "step": 5090000 }, { "epoch": 26.18, "learning_rate": 1.3639285714285716e-05, "loss": 1.4086, "step": 5090500 }, { "epoch": 26.18, "learning_rate": 1.3635714285714285e-05, "loss": 1.4181, "step": 5091000 }, { "epoch": 26.19, "learning_rate": 1.3632142857142857e-05, "loss": 1.4112, "step": 5091500 }, { "epoch": 26.19, "learning_rate": 1.362857142857143e-05, "loss": 1.4058, "step": 5092000 }, { "epoch": 26.19, "learning_rate": 1.3625e-05, "loss": 1.4032, "step": 5092500 }, { "epoch": 26.19, "learning_rate": 1.3621428571428572e-05, "loss": 1.4075, "step": 5093000 }, { "epoch": 26.2, "learning_rate": 1.3617857142857143e-05, "loss": 1.4087, "step": 5093500 }, { "epoch": 26.2, "learning_rate": 1.3614285714285715e-05, "loss": 1.409, "step": 5094000 }, { "epoch": 26.2, "learning_rate": 1.3610714285714288e-05, "loss": 1.3996, "step": 5094500 }, { "epoch": 26.2, "learning_rate": 1.3607142857142857e-05, "loss": 1.4063, "step": 5095000 }, { "epoch": 26.21, "learning_rate": 1.360357142857143e-05, "loss": 1.4008, "step": 5095500 }, { "epoch": 26.21, "learning_rate": 1.3600000000000002e-05, "loss": 1.4098, "step": 5096000 }, { "epoch": 26.21, "learning_rate": 1.3596428571428571e-05, "loss": 1.4118, "step": 5096500 }, { "epoch": 26.21, "learning_rate": 1.3592857142857144e-05, "loss": 1.4195, "step": 5097000 }, { "epoch": 26.22, "learning_rate": 1.3589285714285716e-05, "loss": 1.4049, "step": 5097500 }, { "epoch": 26.22, "learning_rate": 1.3585714285714287e-05, "loss": 1.4136, "step": 5098000 }, { "epoch": 26.22, "learning_rate": 1.3582142857142858e-05, "loss": 1.4173, "step": 5098500 }, { "epoch": 26.23, "learning_rate": 1.3578571428571429e-05, "loss": 1.4254, "step": 5099000 }, { "epoch": 26.23, "learning_rate": 1.3575000000000001e-05, "loss": 1.4047, "step": 5099500 }, { "epoch": 26.23, "learning_rate": 1.357142857142857e-05, "loss": 1.403, "step": 5100000 }, { "epoch": 26.23, "learning_rate": 1.3567857142857143e-05, "loss": 1.4008, "step": 5100500 }, { "epoch": 26.24, "learning_rate": 1.3564285714285716e-05, "loss": 1.415, "step": 5101000 }, { "epoch": 26.24, "learning_rate": 1.3560714285714288e-05, "loss": 1.4038, "step": 5101500 }, { "epoch": 26.24, "learning_rate": 1.3557142857142857e-05, "loss": 1.4168, "step": 5102000 }, { "epoch": 26.24, "learning_rate": 1.355357142857143e-05, "loss": 1.4175, "step": 5102500 }, { "epoch": 26.25, "learning_rate": 1.3550000000000002e-05, "loss": 1.4078, "step": 5103000 }, { "epoch": 26.25, "learning_rate": 1.3546428571428572e-05, "loss": 1.4073, "step": 5103500 }, { "epoch": 26.25, "learning_rate": 1.3542857142857142e-05, "loss": 1.398, "step": 5104000 }, { "epoch": 26.25, "learning_rate": 1.3539285714285715e-05, "loss": 1.4037, "step": 5104500 }, { "epoch": 26.26, "learning_rate": 1.3535714285714288e-05, "loss": 1.4018, "step": 5105000 }, { "epoch": 26.26, "learning_rate": 1.3532142857142857e-05, "loss": 1.4026, "step": 5105500 }, { "epoch": 26.26, "learning_rate": 1.352857142857143e-05, "loss": 1.3987, "step": 5106000 }, { "epoch": 26.27, "learning_rate": 1.3525000000000002e-05, "loss": 1.4133, "step": 5106500 }, { "epoch": 26.27, "learning_rate": 1.3521428571428571e-05, "loss": 1.4128, "step": 5107000 }, { "epoch": 26.27, "learning_rate": 1.3517857142857144e-05, "loss": 1.4186, "step": 5107500 }, { "epoch": 26.27, "learning_rate": 1.3514285714285716e-05, "loss": 1.406, "step": 5108000 }, { "epoch": 26.28, "learning_rate": 1.3510714285714287e-05, "loss": 1.403, "step": 5108500 }, { "epoch": 26.28, "learning_rate": 1.3507142857142858e-05, "loss": 1.4125, "step": 5109000 }, { "epoch": 26.28, "learning_rate": 1.3503571428571429e-05, "loss": 1.4153, "step": 5109500 }, { "epoch": 26.28, "learning_rate": 1.3500000000000001e-05, "loss": 1.4067, "step": 5110000 }, { "epoch": 26.29, "learning_rate": 1.349642857142857e-05, "loss": 1.3995, "step": 5110500 }, { "epoch": 26.29, "learning_rate": 1.3492857142857143e-05, "loss": 1.4081, "step": 5111000 }, { "epoch": 26.29, "learning_rate": 1.3489285714285715e-05, "loss": 1.4035, "step": 5111500 }, { "epoch": 26.29, "learning_rate": 1.3485714285714288e-05, "loss": 1.4079, "step": 5112000 }, { "epoch": 26.3, "learning_rate": 1.3482142857142857e-05, "loss": 1.4089, "step": 5112500 }, { "epoch": 26.3, "learning_rate": 1.347857142857143e-05, "loss": 1.4098, "step": 5113000 }, { "epoch": 26.3, "learning_rate": 1.3475000000000002e-05, "loss": 1.4076, "step": 5113500 }, { "epoch": 26.3, "learning_rate": 1.3471428571428571e-05, "loss": 1.4005, "step": 5114000 }, { "epoch": 26.31, "learning_rate": 1.3467857142857144e-05, "loss": 1.4108, "step": 5114500 }, { "epoch": 26.31, "learning_rate": 1.3464285714285715e-05, "loss": 1.4101, "step": 5115000 }, { "epoch": 26.31, "learning_rate": 1.3460714285714287e-05, "loss": 1.3977, "step": 5115500 }, { "epoch": 26.32, "learning_rate": 1.3457142857142857e-05, "loss": 1.4108, "step": 5116000 }, { "epoch": 26.32, "learning_rate": 1.3453571428571429e-05, "loss": 1.4073, "step": 5116500 }, { "epoch": 26.32, "learning_rate": 1.3450000000000002e-05, "loss": 1.4032, "step": 5117000 }, { "epoch": 26.32, "learning_rate": 1.344642857142857e-05, "loss": 1.4002, "step": 5117500 }, { "epoch": 26.33, "learning_rate": 1.3442857142857143e-05, "loss": 1.4042, "step": 5118000 }, { "epoch": 26.33, "learning_rate": 1.3439285714285716e-05, "loss": 1.4096, "step": 5118500 }, { "epoch": 26.33, "learning_rate": 1.3435714285714287e-05, "loss": 1.399, "step": 5119000 }, { "epoch": 26.33, "learning_rate": 1.3432142857142858e-05, "loss": 1.3982, "step": 5119500 }, { "epoch": 26.34, "learning_rate": 1.3428571428571429e-05, "loss": 1.4163, "step": 5120000 }, { "epoch": 26.34, "learning_rate": 1.3425000000000001e-05, "loss": 1.4032, "step": 5120500 }, { "epoch": 26.34, "learning_rate": 1.342142857142857e-05, "loss": 1.3959, "step": 5121000 }, { "epoch": 26.34, "learning_rate": 1.3417857142857143e-05, "loss": 1.4063, "step": 5121500 }, { "epoch": 26.35, "learning_rate": 1.3414285714285715e-05, "loss": 1.3978, "step": 5122000 }, { "epoch": 26.35, "learning_rate": 1.3410714285714288e-05, "loss": 1.4078, "step": 5122500 }, { "epoch": 26.35, "learning_rate": 1.3407142857142857e-05, "loss": 1.4077, "step": 5123000 }, { "epoch": 26.36, "learning_rate": 1.340357142857143e-05, "loss": 1.4123, "step": 5123500 }, { "epoch": 26.36, "learning_rate": 1.3400000000000002e-05, "loss": 1.4164, "step": 5124000 }, { "epoch": 26.36, "learning_rate": 1.3396428571428571e-05, "loss": 1.4086, "step": 5124500 }, { "epoch": 26.36, "learning_rate": 1.3392857142857144e-05, "loss": 1.4062, "step": 5125000 }, { "epoch": 26.37, "learning_rate": 1.3389285714285715e-05, "loss": 1.4073, "step": 5125500 }, { "epoch": 26.37, "learning_rate": 1.3385714285714287e-05, "loss": 1.4062, "step": 5126000 }, { "epoch": 26.37, "learning_rate": 1.3382142857142856e-05, "loss": 1.4105, "step": 5126500 }, { "epoch": 26.37, "learning_rate": 1.3378571428571429e-05, "loss": 1.4131, "step": 5127000 }, { "epoch": 26.38, "learning_rate": 1.3375000000000002e-05, "loss": 1.4, "step": 5127500 }, { "epoch": 26.38, "learning_rate": 1.337142857142857e-05, "loss": 1.4191, "step": 5128000 }, { "epoch": 26.38, "learning_rate": 1.3367857142857143e-05, "loss": 1.4073, "step": 5128500 }, { "epoch": 26.38, "learning_rate": 1.3364285714285716e-05, "loss": 1.4113, "step": 5129000 }, { "epoch": 26.39, "learning_rate": 1.3360714285714288e-05, "loss": 1.3937, "step": 5129500 }, { "epoch": 26.39, "learning_rate": 1.3357142857142858e-05, "loss": 1.4146, "step": 5130000 }, { "epoch": 26.39, "learning_rate": 1.3353571428571428e-05, "loss": 1.4126, "step": 5130500 }, { "epoch": 26.4, "learning_rate": 1.3350000000000001e-05, "loss": 1.4015, "step": 5131000 }, { "epoch": 26.4, "learning_rate": 1.334642857142857e-05, "loss": 1.4067, "step": 5131500 }, { "epoch": 26.4, "learning_rate": 1.3342857142857143e-05, "loss": 1.415, "step": 5132000 }, { "epoch": 26.4, "learning_rate": 1.3339285714285715e-05, "loss": 1.4119, "step": 5132500 }, { "epoch": 26.41, "learning_rate": 1.3335714285714288e-05, "loss": 1.4103, "step": 5133000 }, { "epoch": 26.41, "learning_rate": 1.3332142857142857e-05, "loss": 1.4169, "step": 5133500 }, { "epoch": 26.41, "learning_rate": 1.332857142857143e-05, "loss": 1.4166, "step": 5134000 }, { "epoch": 26.41, "learning_rate": 1.3325000000000002e-05, "loss": 1.4051, "step": 5134500 }, { "epoch": 26.42, "learning_rate": 1.3321428571428571e-05, "loss": 1.4067, "step": 5135000 }, { "epoch": 26.42, "learning_rate": 1.3317857142857144e-05, "loss": 1.4078, "step": 5135500 }, { "epoch": 26.42, "learning_rate": 1.3314285714285715e-05, "loss": 1.4019, "step": 5136000 }, { "epoch": 26.42, "learning_rate": 1.3310714285714287e-05, "loss": 1.4101, "step": 5136500 }, { "epoch": 26.43, "learning_rate": 1.3307142857142856e-05, "loss": 1.4144, "step": 5137000 }, { "epoch": 26.43, "learning_rate": 1.3303571428571429e-05, "loss": 1.4066, "step": 5137500 }, { "epoch": 26.43, "learning_rate": 1.3300000000000001e-05, "loss": 1.4193, "step": 5138000 }, { "epoch": 26.44, "learning_rate": 1.329642857142857e-05, "loss": 1.4023, "step": 5138500 }, { "epoch": 26.44, "learning_rate": 1.3292857142857143e-05, "loss": 1.4026, "step": 5139000 }, { "epoch": 26.44, "learning_rate": 1.3289285714285716e-05, "loss": 1.3978, "step": 5139500 }, { "epoch": 26.44, "learning_rate": 1.3285714285714288e-05, "loss": 1.4099, "step": 5140000 }, { "epoch": 26.45, "learning_rate": 1.3282142857142857e-05, "loss": 1.4202, "step": 5140500 }, { "epoch": 26.45, "learning_rate": 1.327857142857143e-05, "loss": 1.4107, "step": 5141000 }, { "epoch": 26.45, "learning_rate": 1.3275e-05, "loss": 1.4097, "step": 5141500 }, { "epoch": 26.45, "learning_rate": 1.327142857142857e-05, "loss": 1.4107, "step": 5142000 }, { "epoch": 26.46, "learning_rate": 1.3267857142857143e-05, "loss": 1.4072, "step": 5142500 }, { "epoch": 26.46, "learning_rate": 1.3264285714285715e-05, "loss": 1.4156, "step": 5143000 }, { "epoch": 26.46, "learning_rate": 1.3260714285714288e-05, "loss": 1.4068, "step": 5143500 }, { "epoch": 26.46, "learning_rate": 1.3257142857142857e-05, "loss": 1.4046, "step": 5144000 }, { "epoch": 26.47, "learning_rate": 1.325357142857143e-05, "loss": 1.3997, "step": 5144500 }, { "epoch": 26.47, "learning_rate": 1.3250000000000002e-05, "loss": 1.4085, "step": 5145000 }, { "epoch": 26.47, "learning_rate": 1.3246428571428571e-05, "loss": 1.3997, "step": 5145500 }, { "epoch": 26.48, "learning_rate": 1.3242857142857144e-05, "loss": 1.4027, "step": 5146000 }, { "epoch": 26.48, "learning_rate": 1.3239285714285714e-05, "loss": 1.4114, "step": 5146500 }, { "epoch": 26.48, "learning_rate": 1.3235714285714287e-05, "loss": 1.4115, "step": 5147000 }, { "epoch": 26.48, "learning_rate": 1.3232142857142856e-05, "loss": 1.4138, "step": 5147500 }, { "epoch": 26.49, "learning_rate": 1.3228571428571429e-05, "loss": 1.3987, "step": 5148000 }, { "epoch": 26.49, "learning_rate": 1.3225000000000001e-05, "loss": 1.4095, "step": 5148500 }, { "epoch": 26.49, "learning_rate": 1.322142857142857e-05, "loss": 1.4088, "step": 5149000 }, { "epoch": 26.49, "learning_rate": 1.3217857142857143e-05, "loss": 1.4107, "step": 5149500 }, { "epoch": 26.5, "learning_rate": 1.3214285714285716e-05, "loss": 1.4098, "step": 5150000 }, { "epoch": 26.5, "learning_rate": 1.3210714285714288e-05, "loss": 1.4051, "step": 5150500 }, { "epoch": 26.5, "learning_rate": 1.3207142857142857e-05, "loss": 1.415, "step": 5151000 }, { "epoch": 26.5, "learning_rate": 1.320357142857143e-05, "loss": 1.4088, "step": 5151500 }, { "epoch": 26.51, "learning_rate": 1.32e-05, "loss": 1.416, "step": 5152000 }, { "epoch": 26.51, "learning_rate": 1.3196428571428572e-05, "loss": 1.4131, "step": 5152500 }, { "epoch": 26.51, "learning_rate": 1.3192857142857142e-05, "loss": 1.4048, "step": 5153000 }, { "epoch": 26.52, "learning_rate": 1.3189285714285715e-05, "loss": 1.3999, "step": 5153500 }, { "epoch": 26.52, "learning_rate": 1.3185714285714287e-05, "loss": 1.4124, "step": 5154000 }, { "epoch": 26.52, "learning_rate": 1.3182142857142857e-05, "loss": 1.4114, "step": 5154500 }, { "epoch": 26.52, "learning_rate": 1.317857142857143e-05, "loss": 1.418, "step": 5155000 }, { "epoch": 26.53, "learning_rate": 1.3175000000000002e-05, "loss": 1.4114, "step": 5155500 }, { "epoch": 26.53, "learning_rate": 1.3171428571428571e-05, "loss": 1.4137, "step": 5156000 }, { "epoch": 26.53, "learning_rate": 1.3167857142857143e-05, "loss": 1.4104, "step": 5156500 }, { "epoch": 26.53, "learning_rate": 1.3164285714285714e-05, "loss": 1.4154, "step": 5157000 }, { "epoch": 26.54, "learning_rate": 1.3160714285714287e-05, "loss": 1.406, "step": 5157500 }, { "epoch": 26.54, "learning_rate": 1.3157142857142856e-05, "loss": 1.402, "step": 5158000 }, { "epoch": 26.54, "learning_rate": 1.3153571428571429e-05, "loss": 1.4186, "step": 5158500 }, { "epoch": 26.54, "learning_rate": 1.3150000000000001e-05, "loss": 1.4118, "step": 5159000 }, { "epoch": 26.55, "learning_rate": 1.3146428571428574e-05, "loss": 1.411, "step": 5159500 }, { "epoch": 26.55, "learning_rate": 1.3142857142857143e-05, "loss": 1.3996, "step": 5160000 }, { "epoch": 26.55, "learning_rate": 1.3139285714285715e-05, "loss": 1.4198, "step": 5160500 }, { "epoch": 26.56, "learning_rate": 1.3135714285714288e-05, "loss": 1.4175, "step": 5161000 }, { "epoch": 26.56, "learning_rate": 1.3132142857142857e-05, "loss": 1.4031, "step": 5161500 }, { "epoch": 26.56, "learning_rate": 1.312857142857143e-05, "loss": 1.4057, "step": 5162000 }, { "epoch": 26.56, "learning_rate": 1.3125e-05, "loss": 1.4185, "step": 5162500 }, { "epoch": 26.57, "learning_rate": 1.3121428571428573e-05, "loss": 1.4203, "step": 5163000 }, { "epoch": 26.57, "learning_rate": 1.3117857142857142e-05, "loss": 1.4097, "step": 5163500 }, { "epoch": 26.57, "learning_rate": 1.3114285714285715e-05, "loss": 1.4139, "step": 5164000 }, { "epoch": 26.57, "learning_rate": 1.3110714285714287e-05, "loss": 1.401, "step": 5164500 }, { "epoch": 26.58, "learning_rate": 1.3107142857142857e-05, "loss": 1.409, "step": 5165000 }, { "epoch": 26.58, "learning_rate": 1.3103571428571429e-05, "loss": 1.4118, "step": 5165500 }, { "epoch": 26.58, "learning_rate": 1.3100000000000002e-05, "loss": 1.4065, "step": 5166000 }, { "epoch": 26.58, "learning_rate": 1.3096428571428574e-05, "loss": 1.4113, "step": 5166500 }, { "epoch": 26.59, "learning_rate": 1.3092857142857143e-05, "loss": 1.4023, "step": 5167000 }, { "epoch": 26.59, "learning_rate": 1.3089285714285714e-05, "loss": 1.4059, "step": 5167500 }, { "epoch": 26.59, "learning_rate": 1.3085714285714287e-05, "loss": 1.4077, "step": 5168000 }, { "epoch": 26.6, "learning_rate": 1.3082142857142856e-05, "loss": 1.4101, "step": 5168500 }, { "epoch": 26.6, "learning_rate": 1.3078571428571428e-05, "loss": 1.4016, "step": 5169000 }, { "epoch": 26.6, "learning_rate": 1.3075000000000001e-05, "loss": 1.407, "step": 5169500 }, { "epoch": 26.6, "learning_rate": 1.3071428571428574e-05, "loss": 1.4027, "step": 5170000 }, { "epoch": 26.61, "learning_rate": 1.3067857142857143e-05, "loss": 1.4028, "step": 5170500 }, { "epoch": 26.61, "learning_rate": 1.3064285714285715e-05, "loss": 1.4017, "step": 5171000 }, { "epoch": 26.61, "learning_rate": 1.3060714285714288e-05, "loss": 1.3977, "step": 5171500 }, { "epoch": 26.61, "learning_rate": 1.3057142857142857e-05, "loss": 1.3998, "step": 5172000 }, { "epoch": 26.62, "learning_rate": 1.305357142857143e-05, "loss": 1.4064, "step": 5172500 }, { "epoch": 26.62, "learning_rate": 1.305e-05, "loss": 1.4085, "step": 5173000 }, { "epoch": 26.62, "learning_rate": 1.3046428571428573e-05, "loss": 1.4051, "step": 5173500 }, { "epoch": 26.62, "learning_rate": 1.3042857142857142e-05, "loss": 1.4129, "step": 5174000 }, { "epoch": 26.63, "learning_rate": 1.3039285714285715e-05, "loss": 1.4018, "step": 5174500 }, { "epoch": 26.63, "learning_rate": 1.3035714285714287e-05, "loss": 1.4033, "step": 5175000 }, { "epoch": 26.63, "learning_rate": 1.3032142857142856e-05, "loss": 1.4116, "step": 5175500 }, { "epoch": 26.64, "learning_rate": 1.3028571428571429e-05, "loss": 1.415, "step": 5176000 }, { "epoch": 26.64, "learning_rate": 1.3025000000000002e-05, "loss": 1.4043, "step": 5176500 }, { "epoch": 26.64, "learning_rate": 1.3021428571428574e-05, "loss": 1.4134, "step": 5177000 }, { "epoch": 26.64, "learning_rate": 1.3017857142857143e-05, "loss": 1.4048, "step": 5177500 }, { "epoch": 26.65, "learning_rate": 1.3014285714285716e-05, "loss": 1.411, "step": 5178000 }, { "epoch": 26.65, "learning_rate": 1.3010714285714287e-05, "loss": 1.4142, "step": 5178500 }, { "epoch": 26.65, "learning_rate": 1.3007142857142856e-05, "loss": 1.4066, "step": 5179000 }, { "epoch": 26.65, "learning_rate": 1.3003571428571428e-05, "loss": 1.4061, "step": 5179500 }, { "epoch": 26.66, "learning_rate": 1.3000000000000001e-05, "loss": 1.4087, "step": 5180000 }, { "epoch": 26.66, "learning_rate": 1.2996428571428573e-05, "loss": 1.4016, "step": 5180500 }, { "epoch": 26.66, "learning_rate": 1.2992857142857143e-05, "loss": 1.4078, "step": 5181000 }, { "epoch": 26.66, "learning_rate": 1.2989285714285715e-05, "loss": 1.4162, "step": 5181500 }, { "epoch": 26.67, "learning_rate": 1.2985714285714288e-05, "loss": 1.4092, "step": 5182000 }, { "epoch": 26.67, "learning_rate": 1.2982142857142857e-05, "loss": 1.4073, "step": 5182500 }, { "epoch": 26.67, "learning_rate": 1.297857142857143e-05, "loss": 1.4224, "step": 5183000 }, { "epoch": 26.68, "learning_rate": 1.2975e-05, "loss": 1.414, "step": 5183500 }, { "epoch": 26.68, "learning_rate": 1.2971428571428573e-05, "loss": 1.4046, "step": 5184000 }, { "epoch": 26.68, "learning_rate": 1.2967857142857142e-05, "loss": 1.4093, "step": 5184500 }, { "epoch": 26.68, "learning_rate": 1.2964285714285715e-05, "loss": 1.416, "step": 5185000 }, { "epoch": 26.69, "learning_rate": 1.2960714285714287e-05, "loss": 1.4066, "step": 5185500 }, { "epoch": 26.69, "learning_rate": 1.2957142857142856e-05, "loss": 1.4078, "step": 5186000 }, { "epoch": 26.69, "learning_rate": 1.2953571428571429e-05, "loss": 1.4026, "step": 5186500 }, { "epoch": 26.69, "learning_rate": 1.2950000000000001e-05, "loss": 1.4041, "step": 5187000 }, { "epoch": 26.7, "learning_rate": 1.2946428571428574e-05, "loss": 1.4125, "step": 5187500 }, { "epoch": 26.7, "learning_rate": 1.2942857142857143e-05, "loss": 1.4186, "step": 5188000 }, { "epoch": 26.7, "learning_rate": 1.2939285714285716e-05, "loss": 1.4232, "step": 5188500 }, { "epoch": 26.7, "learning_rate": 1.2935714285714286e-05, "loss": 1.4173, "step": 5189000 }, { "epoch": 26.71, "learning_rate": 1.2932142857142857e-05, "loss": 1.4067, "step": 5189500 }, { "epoch": 26.71, "learning_rate": 1.2928571428571428e-05, "loss": 1.4069, "step": 5190000 }, { "epoch": 26.71, "learning_rate": 1.2925e-05, "loss": 1.4082, "step": 5190500 }, { "epoch": 26.71, "learning_rate": 1.2921428571428573e-05, "loss": 1.4031, "step": 5191000 }, { "epoch": 26.72, "learning_rate": 1.2917857142857142e-05, "loss": 1.4034, "step": 5191500 }, { "epoch": 26.72, "learning_rate": 1.2914285714285715e-05, "loss": 1.4047, "step": 5192000 }, { "epoch": 26.72, "learning_rate": 1.2910714285714288e-05, "loss": 1.4143, "step": 5192500 }, { "epoch": 26.73, "learning_rate": 1.2907142857142857e-05, "loss": 1.4026, "step": 5193000 }, { "epoch": 26.73, "learning_rate": 1.290357142857143e-05, "loss": 1.4153, "step": 5193500 }, { "epoch": 26.73, "learning_rate": 1.29e-05, "loss": 1.4038, "step": 5194000 }, { "epoch": 26.73, "learning_rate": 1.2896428571428573e-05, "loss": 1.402, "step": 5194500 }, { "epoch": 26.74, "learning_rate": 1.2892857142857142e-05, "loss": 1.4094, "step": 5195000 }, { "epoch": 26.74, "learning_rate": 1.2889285714285714e-05, "loss": 1.4065, "step": 5195500 }, { "epoch": 26.74, "learning_rate": 1.2885714285714287e-05, "loss": 1.4036, "step": 5196000 }, { "epoch": 26.74, "learning_rate": 1.2882142857142856e-05, "loss": 1.4068, "step": 5196500 }, { "epoch": 26.75, "learning_rate": 1.2878571428571429e-05, "loss": 1.4069, "step": 5197000 }, { "epoch": 26.75, "learning_rate": 1.2875000000000001e-05, "loss": 1.4115, "step": 5197500 }, { "epoch": 26.75, "learning_rate": 1.2871428571428574e-05, "loss": 1.4122, "step": 5198000 }, { "epoch": 26.75, "learning_rate": 1.2867857142857143e-05, "loss": 1.4172, "step": 5198500 }, { "epoch": 26.76, "learning_rate": 1.2864285714285716e-05, "loss": 1.4073, "step": 5199000 }, { "epoch": 26.76, "learning_rate": 1.2860714285714286e-05, "loss": 1.4091, "step": 5199500 }, { "epoch": 26.76, "learning_rate": 1.2857142857142857e-05, "loss": 1.4093, "step": 5200000 }, { "epoch": 26.77, "learning_rate": 1.2853571428571428e-05, "loss": 1.4227, "step": 5200500 }, { "epoch": 26.77, "learning_rate": 1.285e-05, "loss": 1.4271, "step": 5201000 }, { "epoch": 26.77, "learning_rate": 1.2846428571428573e-05, "loss": 1.4368, "step": 5201500 }, { "epoch": 26.77, "learning_rate": 1.2842857142857142e-05, "loss": 1.4271, "step": 5202000 }, { "epoch": 26.78, "learning_rate": 1.2839285714285715e-05, "loss": 1.432, "step": 5202500 }, { "epoch": 26.78, "learning_rate": 1.2835714285714287e-05, "loss": 1.4267, "step": 5203000 }, { "epoch": 26.78, "learning_rate": 1.2832142857142857e-05, "loss": 1.4189, "step": 5203500 }, { "epoch": 26.78, "learning_rate": 1.282857142857143e-05, "loss": 1.4713, "step": 5204000 }, { "epoch": 26.79, "learning_rate": 1.2825000000000002e-05, "loss": 1.445, "step": 5204500 }, { "epoch": 26.79, "learning_rate": 1.2821428571428573e-05, "loss": 1.4891, "step": 5205000 }, { "epoch": 26.79, "learning_rate": 1.2817857142857142e-05, "loss": 1.4571, "step": 5205500 }, { "epoch": 26.79, "learning_rate": 1.2814285714285714e-05, "loss": 1.5627, "step": 5206000 }, { "epoch": 26.8, "learning_rate": 1.2810714285714287e-05, "loss": 1.517, "step": 5206500 }, { "epoch": 26.8, "learning_rate": 1.2807142857142856e-05, "loss": 1.5535, "step": 5207000 }, { "epoch": 26.8, "learning_rate": 1.2803571428571429e-05, "loss": 1.4708, "step": 5207500 }, { "epoch": 26.81, "learning_rate": 1.2800000000000001e-05, "loss": 1.4641, "step": 5208000 }, { "epoch": 26.81, "learning_rate": 1.2796428571428574e-05, "loss": 1.4501, "step": 5208500 }, { "epoch": 26.81, "learning_rate": 1.2792857142857143e-05, "loss": 1.4349, "step": 5209000 }, { "epoch": 26.81, "learning_rate": 1.2789285714285715e-05, "loss": 1.4219, "step": 5209500 }, { "epoch": 26.82, "learning_rate": 1.2785714285714286e-05, "loss": 1.4238, "step": 5210000 }, { "epoch": 26.82, "learning_rate": 1.2782142857142857e-05, "loss": 1.4782, "step": 5210500 }, { "epoch": 26.82, "learning_rate": 1.2778571428571428e-05, "loss": 1.5255, "step": 5211000 }, { "epoch": 26.82, "learning_rate": 1.2775e-05, "loss": 1.5452, "step": 5211500 }, { "epoch": 26.83, "learning_rate": 1.2771428571428573e-05, "loss": 1.611, "step": 5212000 }, { "epoch": 26.83, "learning_rate": 1.2767857142857142e-05, "loss": 1.541, "step": 5212500 }, { "epoch": 26.83, "learning_rate": 1.2764285714285715e-05, "loss": 1.6697, "step": 5213000 }, { "epoch": 26.83, "learning_rate": 1.2760714285714287e-05, "loss": 1.593, "step": 5213500 }, { "epoch": 26.84, "learning_rate": 1.2757142857142856e-05, "loss": 1.7346, "step": 5214000 }, { "epoch": 26.84, "learning_rate": 1.2753571428571429e-05, "loss": 1.737, "step": 5214500 }, { "epoch": 26.84, "learning_rate": 1.2750000000000002e-05, "loss": 1.6946, "step": 5215000 }, { "epoch": 26.85, "learning_rate": 1.2746428571428572e-05, "loss": 1.5413, "step": 5215500 }, { "epoch": 26.85, "learning_rate": 1.2742857142857143e-05, "loss": 1.7376, "step": 5216000 }, { "epoch": 26.85, "learning_rate": 1.2739285714285714e-05, "loss": 1.8201, "step": 5216500 }, { "epoch": 26.85, "learning_rate": 1.2735714285714287e-05, "loss": 1.6965, "step": 5217000 }, { "epoch": 26.86, "learning_rate": 1.2732142857142856e-05, "loss": 1.8302, "step": 5217500 }, { "epoch": 26.86, "learning_rate": 1.2728571428571428e-05, "loss": 1.7103, "step": 5218000 }, { "epoch": 26.86, "learning_rate": 1.2725000000000001e-05, "loss": 1.7166, "step": 5218500 }, { "epoch": 26.86, "learning_rate": 1.2721428571428574e-05, "loss": 1.5891, "step": 5219000 }, { "epoch": 26.87, "learning_rate": 1.2717857142857143e-05, "loss": 1.6588, "step": 5219500 }, { "epoch": 26.87, "learning_rate": 1.2714285714285715e-05, "loss": 1.5229, "step": 5220000 }, { "epoch": 26.87, "learning_rate": 1.2710714285714286e-05, "loss": 1.5661, "step": 5220500 }, { "epoch": 26.87, "learning_rate": 1.2707142857142859e-05, "loss": 1.6274, "step": 5221000 }, { "epoch": 26.88, "learning_rate": 1.2703571428571428e-05, "loss": 1.7377, "step": 5221500 }, { "epoch": 26.88, "learning_rate": 1.27e-05, "loss": 1.8826, "step": 5222000 }, { "epoch": 26.88, "learning_rate": 1.2696428571428573e-05, "loss": 1.7349, "step": 5222500 }, { "epoch": 26.89, "learning_rate": 1.2692857142857142e-05, "loss": 1.8844, "step": 5223000 }, { "epoch": 26.89, "learning_rate": 1.2689285714285715e-05, "loss": 1.738, "step": 5223500 }, { "epoch": 26.89, "learning_rate": 1.2685714285714287e-05, "loss": 1.7501, "step": 5224000 }, { "epoch": 26.89, "learning_rate": 1.268214285714286e-05, "loss": 1.7547, "step": 5224500 }, { "epoch": 26.9, "learning_rate": 1.2678571428571429e-05, "loss": 1.7572, "step": 5225000 }, { "epoch": 26.9, "learning_rate": 1.2675000000000001e-05, "loss": 2.0594, "step": 5225500 }, { "epoch": 26.9, "learning_rate": 1.2671428571428572e-05, "loss": 1.9909, "step": 5226000 }, { "epoch": 26.9, "learning_rate": 1.2667857142857143e-05, "loss": 2.0232, "step": 5226500 }, { "epoch": 26.91, "learning_rate": 1.2664285714285714e-05, "loss": 2.0814, "step": 5227000 }, { "epoch": 26.91, "learning_rate": 1.2660714285714287e-05, "loss": 2.2741, "step": 5227500 }, { "epoch": 26.91, "learning_rate": 1.2657142857142859e-05, "loss": 2.1519, "step": 5228000 }, { "epoch": 26.91, "learning_rate": 1.2653571428571428e-05, "loss": 2.4739, "step": 5228500 }, { "epoch": 26.92, "learning_rate": 1.2650000000000001e-05, "loss": 2.3656, "step": 5229000 }, { "epoch": 26.92, "learning_rate": 1.2646428571428573e-05, "loss": 2.2544, "step": 5229500 }, { "epoch": 26.92, "learning_rate": 1.2642857142857143e-05, "loss": 2.2587, "step": 5230000 }, { "epoch": 26.93, "learning_rate": 1.2639285714285715e-05, "loss": 2.2744, "step": 5230500 }, { "epoch": 26.93, "learning_rate": 1.2635714285714288e-05, "loss": 2.3338, "step": 5231000 }, { "epoch": 26.93, "learning_rate": 1.2632142857142859e-05, "loss": 2.3652, "step": 5231500 }, { "epoch": 26.93, "learning_rate": 1.2628571428571428e-05, "loss": 2.594, "step": 5232000 }, { "epoch": 26.94, "learning_rate": 1.2625e-05, "loss": 2.5775, "step": 5232500 }, { "epoch": 26.94, "learning_rate": 1.2621428571428573e-05, "loss": 2.562, "step": 5233000 }, { "epoch": 26.94, "learning_rate": 1.2617857142857142e-05, "loss": 2.5782, "step": 5233500 }, { "epoch": 26.94, "learning_rate": 1.2614285714285715e-05, "loss": 2.9932, "step": 5234000 }, { "epoch": 26.95, "learning_rate": 1.2610714285714287e-05, "loss": 2.6483, "step": 5234500 }, { "epoch": 26.95, "learning_rate": 1.260714285714286e-05, "loss": 2.6895, "step": 5235000 }, { "epoch": 26.95, "learning_rate": 1.2603571428571429e-05, "loss": 2.6471, "step": 5235500 }, { "epoch": 26.95, "learning_rate": 1.2600000000000001e-05, "loss": 2.8266, "step": 5236000 }, { "epoch": 26.96, "learning_rate": 1.2596428571428572e-05, "loss": 2.9139, "step": 5236500 }, { "epoch": 26.96, "learning_rate": 1.2592857142857143e-05, "loss": 2.8292, "step": 5237000 }, { "epoch": 26.96, "learning_rate": 1.2589285714285714e-05, "loss": 3.3785, "step": 5237500 }, { "epoch": 26.97, "learning_rate": 1.2585714285714286e-05, "loss": 3.2004, "step": 5238000 }, { "epoch": 26.97, "learning_rate": 1.2582142857142859e-05, "loss": 3.2038, "step": 5238500 }, { "epoch": 26.97, "learning_rate": 1.2578571428571428e-05, "loss": 3.2764, "step": 5239000 }, { "epoch": 26.97, "learning_rate": 1.2575e-05, "loss": 3.2322, "step": 5239500 }, { "epoch": 26.98, "learning_rate": 1.2571428571428573e-05, "loss": 2.9601, "step": 5240000 }, { "epoch": 26.98, "learning_rate": 1.2567857142857142e-05, "loss": 2.9856, "step": 5240500 }, { "epoch": 26.98, "learning_rate": 1.2564285714285715e-05, "loss": 2.5532, "step": 5241000 }, { "epoch": 26.98, "learning_rate": 1.2560714285714288e-05, "loss": 3.1641, "step": 5241500 }, { "epoch": 26.99, "learning_rate": 1.2557142857142858e-05, "loss": 3.4347, "step": 5242000 }, { "epoch": 26.99, "learning_rate": 1.255357142857143e-05, "loss": 3.2068, "step": 5242500 }, { "epoch": 26.99, "learning_rate": 1.255e-05, "loss": 3.2654, "step": 5243000 }, { "epoch": 26.99, "learning_rate": 1.2546428571428573e-05, "loss": 3.1156, "step": 5243500 }, { "epoch": 27.0, "learning_rate": 1.2542857142857142e-05, "loss": 3.1957, "step": 5244000 }, { "epoch": 27.0, "learning_rate": 1.2539285714285714e-05, "loss": 3.3082, "step": 5244500 }, { "epoch": 27.0, "learning_rate": 1.2535714285714287e-05, "loss": 2.863, "step": 5245000 }, { "epoch": 27.01, "learning_rate": 1.253214285714286e-05, "loss": 3.0659, "step": 5245500 }, { "epoch": 27.01, "learning_rate": 1.2528571428571429e-05, "loss": 2.6355, "step": 5246000 }, { "epoch": 27.01, "learning_rate": 1.2525000000000001e-05, "loss": 3.0151, "step": 5246500 }, { "epoch": 27.01, "learning_rate": 1.2521428571428572e-05, "loss": 3.2886, "step": 5247000 }, { "epoch": 27.02, "learning_rate": 1.2517857142857143e-05, "loss": 3.1842, "step": 5247500 }, { "epoch": 27.02, "learning_rate": 1.2514285714285714e-05, "loss": 2.8998, "step": 5248000 }, { "epoch": 27.02, "learning_rate": 1.2510714285714286e-05, "loss": 3.136, "step": 5248500 }, { "epoch": 27.02, "learning_rate": 1.2507142857142859e-05, "loss": 3.253, "step": 5249000 }, { "epoch": 27.03, "learning_rate": 1.2503571428571428e-05, "loss": 3.1368, "step": 5249500 }, { "epoch": 27.03, "learning_rate": 1.25e-05, "loss": 2.8234, "step": 5250000 }, { "epoch": 27.03, "learning_rate": 1.2496428571428571e-05, "loss": 2.8897, "step": 5250500 }, { "epoch": 27.03, "learning_rate": 1.2492857142857144e-05, "loss": 3.0668, "step": 5251000 }, { "epoch": 27.04, "learning_rate": 1.2489285714285715e-05, "loss": 3.0286, "step": 5251500 }, { "epoch": 27.04, "learning_rate": 1.2485714285714287e-05, "loss": 2.733, "step": 5252000 }, { "epoch": 27.04, "learning_rate": 1.2482142857142858e-05, "loss": 2.7963, "step": 5252500 }, { "epoch": 27.05, "learning_rate": 1.2478571428571429e-05, "loss": 2.5626, "step": 5253000 }, { "epoch": 27.05, "learning_rate": 1.2475e-05, "loss": 2.9253, "step": 5253500 }, { "epoch": 27.05, "learning_rate": 1.2471428571428571e-05, "loss": 2.8412, "step": 5254000 }, { "epoch": 27.05, "learning_rate": 1.2467857142857143e-05, "loss": 2.7317, "step": 5254500 }, { "epoch": 27.06, "learning_rate": 1.2464285714285714e-05, "loss": 3.3238, "step": 5255000 }, { "epoch": 27.06, "learning_rate": 1.2460714285714287e-05, "loss": 3.0241, "step": 5255500 }, { "epoch": 27.06, "learning_rate": 1.2457142857142858e-05, "loss": 3.0503, "step": 5256000 }, { "epoch": 27.06, "learning_rate": 1.245357142857143e-05, "loss": 2.6212, "step": 5256500 }, { "epoch": 27.07, "learning_rate": 1.2450000000000001e-05, "loss": 2.8229, "step": 5257000 }, { "epoch": 27.07, "learning_rate": 1.2446428571428572e-05, "loss": 2.6336, "step": 5257500 }, { "epoch": 27.07, "learning_rate": 1.2442857142857143e-05, "loss": 2.4667, "step": 5258000 }, { "epoch": 27.07, "learning_rate": 1.2439285714285714e-05, "loss": 2.6735, "step": 5258500 }, { "epoch": 27.08, "learning_rate": 1.2435714285714286e-05, "loss": 2.2935, "step": 5259000 }, { "epoch": 27.08, "learning_rate": 1.2432142857142857e-05, "loss": 2.4337, "step": 5259500 }, { "epoch": 27.08, "learning_rate": 1.242857142857143e-05, "loss": 2.5662, "step": 5260000 }, { "epoch": 27.09, "learning_rate": 1.2425e-05, "loss": 2.3622, "step": 5260500 }, { "epoch": 27.09, "learning_rate": 1.2421428571428571e-05, "loss": 2.5113, "step": 5261000 }, { "epoch": 27.09, "learning_rate": 1.2417857142857144e-05, "loss": 2.5498, "step": 5261500 }, { "epoch": 27.09, "learning_rate": 1.2414285714285715e-05, "loss": 2.3514, "step": 5262000 }, { "epoch": 27.1, "learning_rate": 1.2410714285714287e-05, "loss": 2.0587, "step": 5262500 }, { "epoch": 27.1, "learning_rate": 1.2407142857142858e-05, "loss": 2.2986, "step": 5263000 }, { "epoch": 27.1, "learning_rate": 1.2403571428571429e-05, "loss": 2.2076, "step": 5263500 }, { "epoch": 27.1, "learning_rate": 1.24e-05, "loss": 2.5729, "step": 5264000 }, { "epoch": 27.11, "learning_rate": 1.239642857142857e-05, "loss": 2.2578, "step": 5264500 }, { "epoch": 27.11, "learning_rate": 1.2392857142857143e-05, "loss": 2.3416, "step": 5265000 }, { "epoch": 27.11, "learning_rate": 1.2389285714285714e-05, "loss": 2.1061, "step": 5265500 }, { "epoch": 27.11, "learning_rate": 1.2385714285714287e-05, "loss": 2.0692, "step": 5266000 }, { "epoch": 27.12, "learning_rate": 1.2382142857142858e-05, "loss": 2.1835, "step": 5266500 }, { "epoch": 27.12, "learning_rate": 1.237857142857143e-05, "loss": 1.9899, "step": 5267000 }, { "epoch": 27.12, "learning_rate": 1.2375000000000001e-05, "loss": 1.8072, "step": 5267500 }, { "epoch": 27.12, "learning_rate": 1.2371428571428574e-05, "loss": 1.9581, "step": 5268000 }, { "epoch": 27.13, "learning_rate": 1.2367857142857144e-05, "loss": 1.9659, "step": 5268500 }, { "epoch": 27.13, "learning_rate": 1.2364285714285714e-05, "loss": 2.0243, "step": 5269000 }, { "epoch": 27.13, "learning_rate": 1.2360714285714286e-05, "loss": 2.1591, "step": 5269500 }, { "epoch": 27.14, "learning_rate": 1.2357142857142857e-05, "loss": 2.2621, "step": 5270000 }, { "epoch": 27.14, "learning_rate": 1.235357142857143e-05, "loss": 1.9527, "step": 5270500 }, { "epoch": 27.14, "learning_rate": 1.235e-05, "loss": 1.6543, "step": 5271000 }, { "epoch": 27.14, "learning_rate": 1.2346428571428573e-05, "loss": 1.7997, "step": 5271500 }, { "epoch": 27.15, "learning_rate": 1.2342857142857144e-05, "loss": 1.6395, "step": 5272000 }, { "epoch": 27.15, "learning_rate": 1.2339285714285715e-05, "loss": 1.7162, "step": 5272500 }, { "epoch": 27.15, "learning_rate": 1.2335714285714287e-05, "loss": 1.7944, "step": 5273000 }, { "epoch": 27.15, "learning_rate": 1.2332142857142858e-05, "loss": 1.9113, "step": 5273500 }, { "epoch": 27.16, "learning_rate": 1.2328571428571429e-05, "loss": 1.7232, "step": 5274000 }, { "epoch": 27.16, "learning_rate": 1.2325e-05, "loss": 1.8883, "step": 5274500 }, { "epoch": 27.16, "learning_rate": 1.2321428571428572e-05, "loss": 1.8692, "step": 5275000 }, { "epoch": 27.16, "learning_rate": 1.2317857142857143e-05, "loss": 1.9153, "step": 5275500 }, { "epoch": 27.17, "learning_rate": 1.2314285714285714e-05, "loss": 1.8697, "step": 5276000 }, { "epoch": 27.17, "learning_rate": 1.2310714285714287e-05, "loss": 1.7861, "step": 5276500 }, { "epoch": 27.17, "learning_rate": 1.2307142857142857e-05, "loss": 1.7873, "step": 5277000 }, { "epoch": 27.18, "learning_rate": 1.230357142857143e-05, "loss": 1.8072, "step": 5277500 }, { "epoch": 27.18, "learning_rate": 1.23e-05, "loss": 1.8646, "step": 5278000 }, { "epoch": 27.18, "learning_rate": 1.2296428571428573e-05, "loss": 2.2012, "step": 5278500 }, { "epoch": 27.18, "learning_rate": 1.2292857142857144e-05, "loss": 2.0648, "step": 5279000 }, { "epoch": 27.19, "learning_rate": 1.2289285714285715e-05, "loss": 2.1155, "step": 5279500 }, { "epoch": 27.19, "learning_rate": 1.2285714285714286e-05, "loss": 2.3295, "step": 5280000 }, { "epoch": 27.19, "learning_rate": 1.2282142857142857e-05, "loss": 1.9241, "step": 5280500 }, { "epoch": 27.19, "learning_rate": 1.227857142857143e-05, "loss": 2.0666, "step": 5281000 }, { "epoch": 27.2, "learning_rate": 1.2275e-05, "loss": 2.0609, "step": 5281500 }, { "epoch": 27.2, "learning_rate": 1.2271428571428573e-05, "loss": 1.8983, "step": 5282000 }, { "epoch": 27.2, "learning_rate": 1.2267857142857144e-05, "loss": 1.8302, "step": 5282500 }, { "epoch": 27.2, "learning_rate": 1.2264285714285714e-05, "loss": 1.7532, "step": 5283000 }, { "epoch": 27.21, "learning_rate": 1.2260714285714287e-05, "loss": 1.7547, "step": 5283500 }, { "epoch": 27.21, "learning_rate": 1.2257142857142858e-05, "loss": 1.91, "step": 5284000 }, { "epoch": 27.21, "learning_rate": 1.2253571428571429e-05, "loss": 1.657, "step": 5284500 }, { "epoch": 27.22, "learning_rate": 1.225e-05, "loss": 1.6826, "step": 5285000 }, { "epoch": 27.22, "learning_rate": 1.2246428571428572e-05, "loss": 1.6283, "step": 5285500 }, { "epoch": 27.22, "learning_rate": 1.2242857142857143e-05, "loss": 1.5168, "step": 5286000 }, { "epoch": 27.22, "learning_rate": 1.2239285714285714e-05, "loss": 1.5972, "step": 5286500 }, { "epoch": 27.23, "learning_rate": 1.2235714285714286e-05, "loss": 1.5527, "step": 5287000 }, { "epoch": 27.23, "learning_rate": 1.2232142857142857e-05, "loss": 1.5102, "step": 5287500 }, { "epoch": 27.23, "learning_rate": 1.222857142857143e-05, "loss": 1.4826, "step": 5288000 }, { "epoch": 27.23, "learning_rate": 1.2225e-05, "loss": 1.6998, "step": 5288500 }, { "epoch": 27.24, "learning_rate": 1.2221428571428573e-05, "loss": 1.5122, "step": 5289000 }, { "epoch": 27.24, "learning_rate": 1.2217857142857144e-05, "loss": 1.7057, "step": 5289500 }, { "epoch": 27.24, "learning_rate": 1.2214285714285715e-05, "loss": 1.7632, "step": 5290000 }, { "epoch": 27.24, "learning_rate": 1.2210714285714286e-05, "loss": 1.5941, "step": 5290500 }, { "epoch": 27.25, "learning_rate": 1.2207142857142857e-05, "loss": 1.4951, "step": 5291000 }, { "epoch": 27.25, "learning_rate": 1.220357142857143e-05, "loss": 1.6285, "step": 5291500 }, { "epoch": 27.25, "learning_rate": 1.22e-05, "loss": 1.5072, "step": 5292000 }, { "epoch": 27.26, "learning_rate": 1.2196428571428573e-05, "loss": 1.5478, "step": 5292500 }, { "epoch": 27.26, "learning_rate": 1.2192857142857143e-05, "loss": 1.5471, "step": 5293000 }, { "epoch": 27.26, "learning_rate": 1.2189285714285714e-05, "loss": 1.5952, "step": 5293500 }, { "epoch": 27.26, "learning_rate": 1.2185714285714287e-05, "loss": 1.5077, "step": 5294000 }, { "epoch": 27.27, "learning_rate": 1.2182142857142858e-05, "loss": 1.5058, "step": 5294500 }, { "epoch": 27.27, "learning_rate": 1.2178571428571429e-05, "loss": 1.5016, "step": 5295000 }, { "epoch": 27.27, "learning_rate": 1.2175e-05, "loss": 1.5869, "step": 5295500 }, { "epoch": 27.27, "learning_rate": 1.2171428571428572e-05, "loss": 1.487, "step": 5296000 }, { "epoch": 27.28, "learning_rate": 1.2167857142857143e-05, "loss": 1.4458, "step": 5296500 }, { "epoch": 27.28, "learning_rate": 1.2164285714285715e-05, "loss": 1.496, "step": 5297000 }, { "epoch": 27.28, "learning_rate": 1.2160714285714286e-05, "loss": 1.5077, "step": 5297500 }, { "epoch": 27.28, "learning_rate": 1.2157142857142857e-05, "loss": 1.4927, "step": 5298000 }, { "epoch": 27.29, "learning_rate": 1.215357142857143e-05, "loss": 1.5802, "step": 5298500 }, { "epoch": 27.29, "learning_rate": 1.215e-05, "loss": 1.6134, "step": 5299000 }, { "epoch": 27.29, "learning_rate": 1.2146428571428573e-05, "loss": 1.4647, "step": 5299500 }, { "epoch": 27.3, "learning_rate": 1.2142857142857144e-05, "loss": 1.4451, "step": 5300000 }, { "epoch": 27.3, "learning_rate": 1.2139285714285715e-05, "loss": 1.5108, "step": 5300500 }, { "epoch": 27.3, "learning_rate": 1.2135714285714286e-05, "loss": 1.6473, "step": 5301000 }, { "epoch": 27.3, "learning_rate": 1.2132142857142857e-05, "loss": 1.6026, "step": 5301500 }, { "epoch": 27.31, "learning_rate": 1.2128571428571429e-05, "loss": 1.6063, "step": 5302000 }, { "epoch": 27.31, "learning_rate": 1.2125e-05, "loss": 1.5707, "step": 5302500 }, { "epoch": 27.31, "learning_rate": 1.2121428571428573e-05, "loss": 1.4879, "step": 5303000 }, { "epoch": 27.31, "learning_rate": 1.2117857142857143e-05, "loss": 1.5653, "step": 5303500 }, { "epoch": 27.32, "learning_rate": 1.2114285714285716e-05, "loss": 1.4813, "step": 5304000 }, { "epoch": 27.32, "learning_rate": 1.2110714285714287e-05, "loss": 1.4352, "step": 5304500 }, { "epoch": 27.32, "learning_rate": 1.2107142857142858e-05, "loss": 1.4313, "step": 5305000 }, { "epoch": 27.32, "learning_rate": 1.210357142857143e-05, "loss": 1.4957, "step": 5305500 }, { "epoch": 27.33, "learning_rate": 1.2100000000000001e-05, "loss": 1.4931, "step": 5306000 }, { "epoch": 27.33, "learning_rate": 1.2096428571428572e-05, "loss": 1.4472, "step": 5306500 }, { "epoch": 27.33, "learning_rate": 1.2092857142857143e-05, "loss": 1.5049, "step": 5307000 }, { "epoch": 27.34, "learning_rate": 1.2089285714285715e-05, "loss": 1.464, "step": 5307500 }, { "epoch": 27.34, "learning_rate": 1.2085714285714286e-05, "loss": 1.4712, "step": 5308000 }, { "epoch": 27.34, "learning_rate": 1.2082142857142857e-05, "loss": 1.5169, "step": 5308500 }, { "epoch": 27.34, "learning_rate": 1.207857142857143e-05, "loss": 1.4544, "step": 5309000 }, { "epoch": 27.35, "learning_rate": 1.2075e-05, "loss": 1.4531, "step": 5309500 }, { "epoch": 27.35, "learning_rate": 1.2071428571428573e-05, "loss": 1.4993, "step": 5310000 }, { "epoch": 27.35, "learning_rate": 1.2067857142857144e-05, "loss": 1.5327, "step": 5310500 }, { "epoch": 27.35, "learning_rate": 1.2064285714285715e-05, "loss": 1.4744, "step": 5311000 }, { "epoch": 27.36, "learning_rate": 1.2060714285714286e-05, "loss": 1.4422, "step": 5311500 }, { "epoch": 27.36, "learning_rate": 1.2057142857142856e-05, "loss": 1.4416, "step": 5312000 }, { "epoch": 27.36, "learning_rate": 1.2053571428571429e-05, "loss": 1.4341, "step": 5312500 }, { "epoch": 27.36, "learning_rate": 1.205e-05, "loss": 1.4379, "step": 5313000 }, { "epoch": 27.37, "learning_rate": 1.2046428571428572e-05, "loss": 1.4942, "step": 5313500 }, { "epoch": 27.37, "learning_rate": 1.2042857142857143e-05, "loss": 1.4944, "step": 5314000 }, { "epoch": 27.37, "learning_rate": 1.2039285714285716e-05, "loss": 1.469, "step": 5314500 }, { "epoch": 27.38, "learning_rate": 1.2035714285714287e-05, "loss": 1.4604, "step": 5315000 }, { "epoch": 27.38, "learning_rate": 1.2032142857142858e-05, "loss": 1.4752, "step": 5315500 }, { "epoch": 27.38, "learning_rate": 1.202857142857143e-05, "loss": 1.4997, "step": 5316000 }, { "epoch": 27.38, "learning_rate": 1.2025000000000001e-05, "loss": 1.5432, "step": 5316500 }, { "epoch": 27.39, "learning_rate": 1.2021428571428572e-05, "loss": 1.4786, "step": 5317000 }, { "epoch": 27.39, "learning_rate": 1.2017857142857143e-05, "loss": 1.4655, "step": 5317500 }, { "epoch": 27.39, "learning_rate": 1.2014285714285715e-05, "loss": 1.4279, "step": 5318000 }, { "epoch": 27.39, "learning_rate": 1.2010714285714286e-05, "loss": 1.4378, "step": 5318500 }, { "epoch": 27.4, "learning_rate": 1.2007142857142857e-05, "loss": 1.4403, "step": 5319000 }, { "epoch": 27.4, "learning_rate": 1.200357142857143e-05, "loss": 1.4333, "step": 5319500 }, { "epoch": 27.4, "learning_rate": 1.2e-05, "loss": 1.4188, "step": 5320000 }, { "epoch": 27.4, "learning_rate": 1.1996428571428573e-05, "loss": 1.42, "step": 5320500 }, { "epoch": 27.41, "learning_rate": 1.1992857142857144e-05, "loss": 1.4413, "step": 5321000 }, { "epoch": 27.41, "learning_rate": 1.1989285714285715e-05, "loss": 1.4532, "step": 5321500 }, { "epoch": 27.41, "learning_rate": 1.1985714285714285e-05, "loss": 1.4398, "step": 5322000 }, { "epoch": 27.42, "learning_rate": 1.1982142857142856e-05, "loss": 1.4607, "step": 5322500 }, { "epoch": 27.42, "learning_rate": 1.1978571428571429e-05, "loss": 1.5219, "step": 5323000 }, { "epoch": 27.42, "learning_rate": 1.1975e-05, "loss": 1.5241, "step": 5323500 }, { "epoch": 27.42, "learning_rate": 1.1971428571428572e-05, "loss": 1.4404, "step": 5324000 }, { "epoch": 27.43, "learning_rate": 1.1967857142857143e-05, "loss": 1.4379, "step": 5324500 }, { "epoch": 27.43, "learning_rate": 1.1964285714285716e-05, "loss": 1.4299, "step": 5325000 }, { "epoch": 27.43, "learning_rate": 1.1960714285714287e-05, "loss": 1.4279, "step": 5325500 }, { "epoch": 27.43, "learning_rate": 1.1957142857142857e-05, "loss": 1.4288, "step": 5326000 }, { "epoch": 27.44, "learning_rate": 1.195357142857143e-05, "loss": 1.4797, "step": 5326500 }, { "epoch": 27.44, "learning_rate": 1.195e-05, "loss": 1.5376, "step": 5327000 }, { "epoch": 27.44, "learning_rate": 1.1946428571428572e-05, "loss": 1.5531, "step": 5327500 }, { "epoch": 27.44, "learning_rate": 1.1942857142857142e-05, "loss": 1.4647, "step": 5328000 }, { "epoch": 27.45, "learning_rate": 1.1939285714285715e-05, "loss": 1.5101, "step": 5328500 }, { "epoch": 27.45, "learning_rate": 1.1935714285714286e-05, "loss": 1.5824, "step": 5329000 }, { "epoch": 27.45, "learning_rate": 1.1932142857142858e-05, "loss": 1.4407, "step": 5329500 }, { "epoch": 27.46, "learning_rate": 1.192857142857143e-05, "loss": 1.4221, "step": 5330000 }, { "epoch": 27.46, "learning_rate": 1.1925e-05, "loss": 1.4245, "step": 5330500 }, { "epoch": 27.46, "learning_rate": 1.1921428571428573e-05, "loss": 1.454, "step": 5331000 }, { "epoch": 27.46, "learning_rate": 1.1917857142857144e-05, "loss": 1.4584, "step": 5331500 }, { "epoch": 27.47, "learning_rate": 1.1914285714285716e-05, "loss": 1.4603, "step": 5332000 }, { "epoch": 27.47, "learning_rate": 1.1910714285714287e-05, "loss": 1.4759, "step": 5332500 }, { "epoch": 27.47, "learning_rate": 1.1907142857142858e-05, "loss": 1.4361, "step": 5333000 }, { "epoch": 27.47, "learning_rate": 1.1903571428571429e-05, "loss": 1.4446, "step": 5333500 }, { "epoch": 27.48, "learning_rate": 1.19e-05, "loss": 1.4585, "step": 5334000 }, { "epoch": 27.48, "learning_rate": 1.1896428571428572e-05, "loss": 1.4765, "step": 5334500 }, { "epoch": 27.48, "learning_rate": 1.1892857142857143e-05, "loss": 1.485, "step": 5335000 }, { "epoch": 27.48, "learning_rate": 1.1889285714285716e-05, "loss": 1.4955, "step": 5335500 }, { "epoch": 27.49, "learning_rate": 1.1885714285714286e-05, "loss": 1.4642, "step": 5336000 }, { "epoch": 27.49, "learning_rate": 1.1882142857142859e-05, "loss": 1.5632, "step": 5336500 }, { "epoch": 27.49, "learning_rate": 1.187857142857143e-05, "loss": 1.5593, "step": 5337000 }, { "epoch": 27.5, "learning_rate": 1.1875e-05, "loss": 1.4447, "step": 5337500 }, { "epoch": 27.5, "learning_rate": 1.1871428571428572e-05, "loss": 1.4185, "step": 5338000 }, { "epoch": 27.5, "learning_rate": 1.1867857142857142e-05, "loss": 1.4233, "step": 5338500 }, { "epoch": 27.5, "learning_rate": 1.1864285714285715e-05, "loss": 1.4389, "step": 5339000 }, { "epoch": 27.51, "learning_rate": 1.1860714285714286e-05, "loss": 1.4287, "step": 5339500 }, { "epoch": 27.51, "learning_rate": 1.1857142857142858e-05, "loss": 1.4281, "step": 5340000 }, { "epoch": 27.51, "learning_rate": 1.185357142857143e-05, "loss": 1.4625, "step": 5340500 }, { "epoch": 27.51, "learning_rate": 1.185e-05, "loss": 1.4526, "step": 5341000 }, { "epoch": 27.52, "learning_rate": 1.1846428571428573e-05, "loss": 1.4161, "step": 5341500 }, { "epoch": 27.52, "learning_rate": 1.1842857142857143e-05, "loss": 1.4285, "step": 5342000 }, { "epoch": 27.52, "learning_rate": 1.1839285714285716e-05, "loss": 1.4399, "step": 5342500 }, { "epoch": 27.52, "learning_rate": 1.1835714285714287e-05, "loss": 1.4763, "step": 5343000 }, { "epoch": 27.53, "learning_rate": 1.1832142857142858e-05, "loss": 1.527, "step": 5343500 }, { "epoch": 27.53, "learning_rate": 1.1828571428571429e-05, "loss": 1.4662, "step": 5344000 }, { "epoch": 27.53, "learning_rate": 1.1825e-05, "loss": 1.4411, "step": 5344500 }, { "epoch": 27.53, "learning_rate": 1.1821428571428572e-05, "loss": 1.433, "step": 5345000 }, { "epoch": 27.54, "learning_rate": 1.1817857142857143e-05, "loss": 1.4346, "step": 5345500 }, { "epoch": 27.54, "learning_rate": 1.1814285714285715e-05, "loss": 1.4437, "step": 5346000 }, { "epoch": 27.54, "learning_rate": 1.1810714285714286e-05, "loss": 1.4199, "step": 5346500 }, { "epoch": 27.55, "learning_rate": 1.1807142857142859e-05, "loss": 1.4233, "step": 5347000 }, { "epoch": 27.55, "learning_rate": 1.180357142857143e-05, "loss": 1.4152, "step": 5347500 }, { "epoch": 27.55, "learning_rate": 1.18e-05, "loss": 1.4242, "step": 5348000 }, { "epoch": 27.55, "learning_rate": 1.1796428571428571e-05, "loss": 1.4529, "step": 5348500 }, { "epoch": 27.56, "learning_rate": 1.1792857142857142e-05, "loss": 1.4578, "step": 5349000 }, { "epoch": 27.56, "learning_rate": 1.1789285714285715e-05, "loss": 1.4198, "step": 5349500 }, { "epoch": 27.56, "learning_rate": 1.1785714285714286e-05, "loss": 1.419, "step": 5350000 }, { "epoch": 27.56, "learning_rate": 1.1782142857142858e-05, "loss": 1.4285, "step": 5350500 }, { "epoch": 27.57, "learning_rate": 1.1778571428571429e-05, "loss": 1.463, "step": 5351000 }, { "epoch": 27.57, "learning_rate": 1.1775e-05, "loss": 1.5006, "step": 5351500 }, { "epoch": 27.57, "learning_rate": 1.1771428571428572e-05, "loss": 1.4423, "step": 5352000 }, { "epoch": 27.57, "learning_rate": 1.1767857142857143e-05, "loss": 1.427, "step": 5352500 }, { "epoch": 27.58, "learning_rate": 1.1764285714285716e-05, "loss": 1.4363, "step": 5353000 }, { "epoch": 27.58, "learning_rate": 1.1760714285714287e-05, "loss": 1.4661, "step": 5353500 }, { "epoch": 27.58, "learning_rate": 1.1757142857142858e-05, "loss": 1.4804, "step": 5354000 }, { "epoch": 27.59, "learning_rate": 1.1753571428571428e-05, "loss": 1.4503, "step": 5354500 }, { "epoch": 27.59, "learning_rate": 1.175e-05, "loss": 1.4623, "step": 5355000 }, { "epoch": 27.59, "learning_rate": 1.1746428571428572e-05, "loss": 1.4713, "step": 5355500 }, { "epoch": 27.59, "learning_rate": 1.1742857142857143e-05, "loss": 1.5123, "step": 5356000 }, { "epoch": 27.6, "learning_rate": 1.1739285714285715e-05, "loss": 1.5877, "step": 5356500 }, { "epoch": 27.6, "learning_rate": 1.1735714285714286e-05, "loss": 1.4933, "step": 5357000 }, { "epoch": 27.6, "learning_rate": 1.1732142857142859e-05, "loss": 1.4573, "step": 5357500 }, { "epoch": 27.6, "learning_rate": 1.172857142857143e-05, "loss": 1.4953, "step": 5358000 }, { "epoch": 27.61, "learning_rate": 1.1725e-05, "loss": 1.5096, "step": 5358500 }, { "epoch": 27.61, "learning_rate": 1.1721428571428571e-05, "loss": 1.5574, "step": 5359000 }, { "epoch": 27.61, "learning_rate": 1.1717857142857142e-05, "loss": 1.5655, "step": 5359500 }, { "epoch": 27.61, "learning_rate": 1.1714285714285715e-05, "loss": 1.499, "step": 5360000 }, { "epoch": 27.62, "learning_rate": 1.1710714285714286e-05, "loss": 1.4531, "step": 5360500 }, { "epoch": 27.62, "learning_rate": 1.1707142857142858e-05, "loss": 1.5097, "step": 5361000 }, { "epoch": 27.62, "learning_rate": 1.1703571428571429e-05, "loss": 1.5263, "step": 5361500 }, { "epoch": 27.63, "learning_rate": 1.1700000000000001e-05, "loss": 1.4479, "step": 5362000 }, { "epoch": 27.63, "learning_rate": 1.1696428571428572e-05, "loss": 1.4249, "step": 5362500 }, { "epoch": 27.63, "learning_rate": 1.1692857142857143e-05, "loss": 1.4198, "step": 5363000 }, { "epoch": 27.63, "learning_rate": 1.1689285714285716e-05, "loss": 1.4308, "step": 5363500 }, { "epoch": 27.64, "learning_rate": 1.1685714285714287e-05, "loss": 1.4272, "step": 5364000 }, { "epoch": 27.64, "learning_rate": 1.1682142857142857e-05, "loss": 1.5071, "step": 5364500 }, { "epoch": 27.64, "learning_rate": 1.1678571428571428e-05, "loss": 1.5767, "step": 5365000 }, { "epoch": 27.64, "learning_rate": 1.1675000000000001e-05, "loss": 1.5057, "step": 5365500 }, { "epoch": 27.65, "learning_rate": 1.1671428571428572e-05, "loss": 1.4376, "step": 5366000 }, { "epoch": 27.65, "learning_rate": 1.1667857142857143e-05, "loss": 1.479, "step": 5366500 }, { "epoch": 27.65, "learning_rate": 1.1664285714285715e-05, "loss": 1.4653, "step": 5367000 }, { "epoch": 27.65, "learning_rate": 1.1660714285714286e-05, "loss": 1.429, "step": 5367500 }, { "epoch": 27.66, "learning_rate": 1.1657142857142859e-05, "loss": 1.441, "step": 5368000 }, { "epoch": 27.66, "learning_rate": 1.165357142857143e-05, "loss": 1.4252, "step": 5368500 }, { "epoch": 27.66, "learning_rate": 1.1650000000000002e-05, "loss": 1.465, "step": 5369000 }, { "epoch": 27.67, "learning_rate": 1.1646428571428573e-05, "loss": 1.4419, "step": 5369500 }, { "epoch": 27.67, "learning_rate": 1.1642857142857144e-05, "loss": 1.4642, "step": 5370000 }, { "epoch": 27.67, "learning_rate": 1.1639285714285715e-05, "loss": 1.4397, "step": 5370500 }, { "epoch": 27.67, "learning_rate": 1.1635714285714285e-05, "loss": 1.4264, "step": 5371000 }, { "epoch": 27.68, "learning_rate": 1.1632142857142858e-05, "loss": 1.4151, "step": 5371500 }, { "epoch": 27.68, "learning_rate": 1.1628571428571429e-05, "loss": 1.4324, "step": 5372000 }, { "epoch": 27.68, "learning_rate": 1.1625000000000001e-05, "loss": 1.4182, "step": 5372500 }, { "epoch": 27.68, "learning_rate": 1.1621428571428572e-05, "loss": 1.4186, "step": 5373000 }, { "epoch": 27.69, "learning_rate": 1.1617857142857143e-05, "loss": 1.4339, "step": 5373500 }, { "epoch": 27.69, "learning_rate": 1.1614285714285716e-05, "loss": 1.4083, "step": 5374000 }, { "epoch": 27.69, "learning_rate": 1.1610714285714286e-05, "loss": 1.4201, "step": 5374500 }, { "epoch": 27.69, "learning_rate": 1.1607142857142857e-05, "loss": 1.4678, "step": 5375000 }, { "epoch": 27.7, "learning_rate": 1.1603571428571428e-05, "loss": 1.4201, "step": 5375500 }, { "epoch": 27.7, "learning_rate": 1.16e-05, "loss": 1.4368, "step": 5376000 }, { "epoch": 27.7, "learning_rate": 1.1596428571428572e-05, "loss": 1.4458, "step": 5376500 }, { "epoch": 27.71, "learning_rate": 1.1592857142857142e-05, "loss": 1.4668, "step": 5377000 }, { "epoch": 27.71, "learning_rate": 1.1589285714285715e-05, "loss": 1.4419, "step": 5377500 }, { "epoch": 27.71, "learning_rate": 1.1585714285714286e-05, "loss": 1.4365, "step": 5378000 }, { "epoch": 27.71, "learning_rate": 1.1582142857142858e-05, "loss": 1.4163, "step": 5378500 }, { "epoch": 27.72, "learning_rate": 1.157857142857143e-05, "loss": 1.4315, "step": 5379000 }, { "epoch": 27.72, "learning_rate": 1.1575000000000002e-05, "loss": 1.4107, "step": 5379500 }, { "epoch": 27.72, "learning_rate": 1.1571428571428573e-05, "loss": 1.416, "step": 5380000 }, { "epoch": 27.72, "learning_rate": 1.1567857142857144e-05, "loss": 1.4154, "step": 5380500 }, { "epoch": 27.73, "learning_rate": 1.1564285714285714e-05, "loss": 1.415, "step": 5381000 }, { "epoch": 27.73, "learning_rate": 1.1560714285714285e-05, "loss": 1.4119, "step": 5381500 }, { "epoch": 27.73, "learning_rate": 1.1557142857142858e-05, "loss": 1.4137, "step": 5382000 }, { "epoch": 27.73, "learning_rate": 1.1553571428571429e-05, "loss": 1.4409, "step": 5382500 }, { "epoch": 27.74, "learning_rate": 1.1550000000000001e-05, "loss": 1.4331, "step": 5383000 }, { "epoch": 27.74, "learning_rate": 1.1546428571428572e-05, "loss": 1.4299, "step": 5383500 }, { "epoch": 27.74, "learning_rate": 1.1542857142857143e-05, "loss": 1.536, "step": 5384000 }, { "epoch": 27.75, "learning_rate": 1.1539285714285715e-05, "loss": 1.447, "step": 5384500 }, { "epoch": 27.75, "learning_rate": 1.1535714285714286e-05, "loss": 1.4104, "step": 5385000 }, { "epoch": 27.75, "learning_rate": 1.1532142857142857e-05, "loss": 1.4173, "step": 5385500 }, { "epoch": 27.75, "learning_rate": 1.1528571428571428e-05, "loss": 1.4126, "step": 5386000 }, { "epoch": 27.76, "learning_rate": 1.1525e-05, "loss": 1.4448, "step": 5386500 }, { "epoch": 27.76, "learning_rate": 1.1521428571428571e-05, "loss": 1.4568, "step": 5387000 }, { "epoch": 27.76, "learning_rate": 1.1517857142857142e-05, "loss": 1.4165, "step": 5387500 }, { "epoch": 27.76, "learning_rate": 1.1514285714285715e-05, "loss": 1.4184, "step": 5388000 }, { "epoch": 27.77, "learning_rate": 1.1510714285714286e-05, "loss": 1.4107, "step": 5388500 }, { "epoch": 27.77, "learning_rate": 1.1507142857142858e-05, "loss": 1.4503, "step": 5389000 }, { "epoch": 27.77, "learning_rate": 1.1503571428571429e-05, "loss": 1.4341, "step": 5389500 }, { "epoch": 27.77, "learning_rate": 1.1500000000000002e-05, "loss": 1.4417, "step": 5390000 }, { "epoch": 27.78, "learning_rate": 1.1496428571428573e-05, "loss": 1.4285, "step": 5390500 }, { "epoch": 27.78, "learning_rate": 1.1492857142857143e-05, "loss": 1.4137, "step": 5391000 }, { "epoch": 27.78, "learning_rate": 1.1489285714285714e-05, "loss": 1.4108, "step": 5391500 }, { "epoch": 27.79, "learning_rate": 1.1485714285714285e-05, "loss": 1.4073, "step": 5392000 }, { "epoch": 27.79, "learning_rate": 1.1482142857142858e-05, "loss": 1.4347, "step": 5392500 }, { "epoch": 27.79, "learning_rate": 1.1478571428571429e-05, "loss": 1.4387, "step": 5393000 }, { "epoch": 27.79, "learning_rate": 1.1475000000000001e-05, "loss": 1.4642, "step": 5393500 }, { "epoch": 27.8, "learning_rate": 1.1471428571428572e-05, "loss": 1.4909, "step": 5394000 }, { "epoch": 27.8, "learning_rate": 1.1467857142857145e-05, "loss": 1.5547, "step": 5394500 }, { "epoch": 27.8, "learning_rate": 1.1464285714285715e-05, "loss": 1.5624, "step": 5395000 }, { "epoch": 27.8, "learning_rate": 1.1460714285714286e-05, "loss": 1.4508, "step": 5395500 }, { "epoch": 27.81, "learning_rate": 1.1457142857142859e-05, "loss": 1.4314, "step": 5396000 }, { "epoch": 27.81, "learning_rate": 1.145357142857143e-05, "loss": 1.4758, "step": 5396500 }, { "epoch": 27.81, "learning_rate": 1.145e-05, "loss": 1.4425, "step": 5397000 }, { "epoch": 27.81, "learning_rate": 1.1446428571428571e-05, "loss": 1.4151, "step": 5397500 }, { "epoch": 27.82, "learning_rate": 1.1442857142857144e-05, "loss": 1.4108, "step": 5398000 }, { "epoch": 27.82, "learning_rate": 1.1439285714285715e-05, "loss": 1.4163, "step": 5398500 }, { "epoch": 27.82, "learning_rate": 1.1435714285714286e-05, "loss": 1.4083, "step": 5399000 }, { "epoch": 27.83, "learning_rate": 1.1432142857142858e-05, "loss": 1.4285, "step": 5399500 }, { "epoch": 27.83, "learning_rate": 1.1428571428571429e-05, "loss": 1.4158, "step": 5400000 }, { "epoch": 27.83, "learning_rate": 1.1425000000000002e-05, "loss": 1.4098, "step": 5400500 }, { "epoch": 27.83, "learning_rate": 1.1421428571428572e-05, "loss": 1.4239, "step": 5401000 }, { "epoch": 27.84, "learning_rate": 1.1417857142857143e-05, "loss": 1.4434, "step": 5401500 }, { "epoch": 27.84, "learning_rate": 1.1414285714285714e-05, "loss": 1.4328, "step": 5402000 }, { "epoch": 27.84, "learning_rate": 1.1410714285714285e-05, "loss": 1.4744, "step": 5402500 }, { "epoch": 27.84, "learning_rate": 1.1407142857142858e-05, "loss": 1.4335, "step": 5403000 }, { "epoch": 27.85, "learning_rate": 1.1403571428571428e-05, "loss": 1.4155, "step": 5403500 }, { "epoch": 27.85, "learning_rate": 1.1400000000000001e-05, "loss": 1.4235, "step": 5404000 }, { "epoch": 27.85, "learning_rate": 1.1396428571428572e-05, "loss": 1.4134, "step": 5404500 }, { "epoch": 27.85, "learning_rate": 1.1392857142857144e-05, "loss": 1.4021, "step": 5405000 }, { "epoch": 27.86, "learning_rate": 1.1389285714285715e-05, "loss": 1.414, "step": 5405500 }, { "epoch": 27.86, "learning_rate": 1.1385714285714286e-05, "loss": 1.4055, "step": 5406000 }, { "epoch": 27.86, "learning_rate": 1.1382142857142859e-05, "loss": 1.4494, "step": 5406500 }, { "epoch": 27.87, "learning_rate": 1.137857142857143e-05, "loss": 1.4482, "step": 5407000 }, { "epoch": 27.87, "learning_rate": 1.1375e-05, "loss": 1.4156, "step": 5407500 }, { "epoch": 27.87, "learning_rate": 1.1371428571428571e-05, "loss": 1.4135, "step": 5408000 }, { "epoch": 27.87, "learning_rate": 1.1367857142857144e-05, "loss": 1.4046, "step": 5408500 }, { "epoch": 27.88, "learning_rate": 1.1364285714285715e-05, "loss": 1.4065, "step": 5409000 }, { "epoch": 27.88, "learning_rate": 1.1360714285714285e-05, "loss": 1.4077, "step": 5409500 }, { "epoch": 27.88, "learning_rate": 1.1357142857142858e-05, "loss": 1.4042, "step": 5410000 }, { "epoch": 27.88, "learning_rate": 1.1353571428571429e-05, "loss": 1.4194, "step": 5410500 }, { "epoch": 27.89, "learning_rate": 1.1350000000000001e-05, "loss": 1.4107, "step": 5411000 }, { "epoch": 27.89, "learning_rate": 1.1346428571428572e-05, "loss": 1.403, "step": 5411500 }, { "epoch": 27.89, "learning_rate": 1.1342857142857143e-05, "loss": 1.4174, "step": 5412000 }, { "epoch": 27.89, "learning_rate": 1.1339285714285714e-05, "loss": 1.4236, "step": 5412500 }, { "epoch": 27.9, "learning_rate": 1.1335714285714285e-05, "loss": 1.4232, "step": 5413000 }, { "epoch": 27.9, "learning_rate": 1.1332142857142857e-05, "loss": 1.4332, "step": 5413500 }, { "epoch": 27.9, "learning_rate": 1.1328571428571428e-05, "loss": 1.4109, "step": 5414000 }, { "epoch": 27.91, "learning_rate": 1.1325e-05, "loss": 1.4149, "step": 5414500 }, { "epoch": 27.91, "learning_rate": 1.1321428571428572e-05, "loss": 1.4084, "step": 5415000 }, { "epoch": 27.91, "learning_rate": 1.1317857142857144e-05, "loss": 1.4154, "step": 5415500 }, { "epoch": 27.91, "learning_rate": 1.1314285714285715e-05, "loss": 1.4257, "step": 5416000 }, { "epoch": 27.92, "learning_rate": 1.1310714285714286e-05, "loss": 1.4611, "step": 5416500 }, { "epoch": 27.92, "learning_rate": 1.1307142857142859e-05, "loss": 1.4336, "step": 5417000 }, { "epoch": 27.92, "learning_rate": 1.130357142857143e-05, "loss": 1.4196, "step": 5417500 }, { "epoch": 27.92, "learning_rate": 1.13e-05, "loss": 1.4081, "step": 5418000 }, { "epoch": 27.93, "learning_rate": 1.1296428571428571e-05, "loss": 1.4052, "step": 5418500 }, { "epoch": 27.93, "learning_rate": 1.1292857142857144e-05, "loss": 1.4152, "step": 5419000 }, { "epoch": 27.93, "learning_rate": 1.1289285714285714e-05, "loss": 1.4095, "step": 5419500 }, { "epoch": 27.93, "learning_rate": 1.1285714285714285e-05, "loss": 1.4219, "step": 5420000 }, { "epoch": 27.94, "learning_rate": 1.1282142857142858e-05, "loss": 1.4108, "step": 5420500 }, { "epoch": 27.94, "learning_rate": 1.1278571428571429e-05, "loss": 1.4128, "step": 5421000 }, { "epoch": 27.94, "learning_rate": 1.1275000000000001e-05, "loss": 1.4162, "step": 5421500 }, { "epoch": 27.94, "learning_rate": 1.1271428571428572e-05, "loss": 1.4085, "step": 5422000 }, { "epoch": 27.95, "learning_rate": 1.1267857142857145e-05, "loss": 1.4124, "step": 5422500 }, { "epoch": 27.95, "learning_rate": 1.1264285714285714e-05, "loss": 1.4155, "step": 5423000 }, { "epoch": 27.95, "learning_rate": 1.1260714285714286e-05, "loss": 1.4285, "step": 5423500 }, { "epoch": 27.96, "learning_rate": 1.1257142857142857e-05, "loss": 1.419, "step": 5424000 }, { "epoch": 27.96, "learning_rate": 1.1253571428571428e-05, "loss": 1.417, "step": 5424500 }, { "epoch": 27.96, "learning_rate": 1.125e-05, "loss": 1.4503, "step": 5425000 }, { "epoch": 27.96, "learning_rate": 1.1246428571428572e-05, "loss": 1.4241, "step": 5425500 }, { "epoch": 27.97, "learning_rate": 1.1242857142857144e-05, "loss": 1.4028, "step": 5426000 }, { "epoch": 27.97, "learning_rate": 1.1239285714285715e-05, "loss": 1.403, "step": 5426500 }, { "epoch": 27.97, "learning_rate": 1.1235714285714288e-05, "loss": 1.4169, "step": 5427000 }, { "epoch": 27.97, "learning_rate": 1.1232142857142858e-05, "loss": 1.41, "step": 5427500 }, { "epoch": 27.98, "learning_rate": 1.122857142857143e-05, "loss": 1.4006, "step": 5428000 }, { "epoch": 27.98, "learning_rate": 1.1225e-05, "loss": 1.4075, "step": 5428500 }, { "epoch": 27.98, "learning_rate": 1.1221428571428571e-05, "loss": 1.4007, "step": 5429000 }, { "epoch": 27.98, "learning_rate": 1.1217857142857144e-05, "loss": 1.4023, "step": 5429500 }, { "epoch": 27.99, "learning_rate": 1.1214285714285714e-05, "loss": 1.3993, "step": 5430000 }, { "epoch": 27.99, "learning_rate": 1.1210714285714287e-05, "loss": 1.4072, "step": 5430500 }, { "epoch": 27.99, "learning_rate": 1.1207142857142858e-05, "loss": 1.408, "step": 5431000 }, { "epoch": 28.0, "learning_rate": 1.1203571428571429e-05, "loss": 1.4099, "step": 5431500 }, { "epoch": 28.0, "learning_rate": 1.1200000000000001e-05, "loss": 1.4019, "step": 5432000 }, { "epoch": 28.0, "learning_rate": 1.1196428571428572e-05, "loss": 1.4077, "step": 5432500 }, { "epoch": 28.0, "learning_rate": 1.1192857142857145e-05, "loss": 1.4012, "step": 5433000 }, { "epoch": 28.01, "learning_rate": 1.1189285714285715e-05, "loss": 1.4045, "step": 5433500 }, { "epoch": 28.01, "learning_rate": 1.1185714285714286e-05, "loss": 1.4319, "step": 5434000 }, { "epoch": 28.01, "learning_rate": 1.1182142857142857e-05, "loss": 1.4229, "step": 5434500 }, { "epoch": 28.01, "learning_rate": 1.1178571428571428e-05, "loss": 1.443, "step": 5435000 }, { "epoch": 28.02, "learning_rate": 1.1175e-05, "loss": 1.4369, "step": 5435500 }, { "epoch": 28.02, "learning_rate": 1.1171428571428571e-05, "loss": 1.4143, "step": 5436000 }, { "epoch": 28.02, "learning_rate": 1.1167857142857144e-05, "loss": 1.4067, "step": 5436500 }, { "epoch": 28.02, "learning_rate": 1.1164285714285715e-05, "loss": 1.4142, "step": 5437000 }, { "epoch": 28.03, "learning_rate": 1.1160714285714287e-05, "loss": 1.4497, "step": 5437500 }, { "epoch": 28.03, "learning_rate": 1.1157142857142858e-05, "loss": 1.415, "step": 5438000 }, { "epoch": 28.03, "learning_rate": 1.1153571428571429e-05, "loss": 1.4034, "step": 5438500 }, { "epoch": 28.04, "learning_rate": 1.115e-05, "loss": 1.4019, "step": 5439000 }, { "epoch": 28.04, "learning_rate": 1.114642857142857e-05, "loss": 1.4052, "step": 5439500 }, { "epoch": 28.04, "learning_rate": 1.1142857142857143e-05, "loss": 1.4036, "step": 5440000 }, { "epoch": 28.04, "learning_rate": 1.1139285714285714e-05, "loss": 1.4017, "step": 5440500 }, { "epoch": 28.05, "learning_rate": 1.1135714285714287e-05, "loss": 1.4393, "step": 5441000 }, { "epoch": 28.05, "learning_rate": 1.1132142857142858e-05, "loss": 1.4242, "step": 5441500 }, { "epoch": 28.05, "learning_rate": 1.1128571428571429e-05, "loss": 1.4041, "step": 5442000 }, { "epoch": 28.05, "learning_rate": 1.1125000000000001e-05, "loss": 1.3982, "step": 5442500 }, { "epoch": 28.06, "learning_rate": 1.1121428571428572e-05, "loss": 1.3978, "step": 5443000 }, { "epoch": 28.06, "learning_rate": 1.1117857142857144e-05, "loss": 1.4146, "step": 5443500 }, { "epoch": 28.06, "learning_rate": 1.1114285714285715e-05, "loss": 1.3994, "step": 5444000 }, { "epoch": 28.06, "learning_rate": 1.1110714285714286e-05, "loss": 1.4002, "step": 5444500 }, { "epoch": 28.07, "learning_rate": 1.1107142857142857e-05, "loss": 1.3921, "step": 5445000 }, { "epoch": 28.07, "learning_rate": 1.1103571428571428e-05, "loss": 1.407, "step": 5445500 }, { "epoch": 28.07, "learning_rate": 1.11e-05, "loss": 1.4116, "step": 5446000 }, { "epoch": 28.08, "learning_rate": 1.1096428571428571e-05, "loss": 1.4201, "step": 5446500 }, { "epoch": 28.08, "learning_rate": 1.1092857142857144e-05, "loss": 1.4284, "step": 5447000 }, { "epoch": 28.08, "learning_rate": 1.1089285714285715e-05, "loss": 1.4104, "step": 5447500 }, { "epoch": 28.08, "learning_rate": 1.1085714285714287e-05, "loss": 1.4625, "step": 5448000 }, { "epoch": 28.09, "learning_rate": 1.1082142857142858e-05, "loss": 1.4417, "step": 5448500 }, { "epoch": 28.09, "learning_rate": 1.1078571428571429e-05, "loss": 1.4274, "step": 5449000 }, { "epoch": 28.09, "learning_rate": 1.1075e-05, "loss": 1.4263, "step": 5449500 }, { "epoch": 28.09, "learning_rate": 1.107142857142857e-05, "loss": 1.4294, "step": 5450000 }, { "epoch": 28.1, "learning_rate": 1.1067857142857143e-05, "loss": 1.4137, "step": 5450500 }, { "epoch": 28.1, "learning_rate": 1.1064285714285714e-05, "loss": 1.4028, "step": 5451000 }, { "epoch": 28.1, "learning_rate": 1.1060714285714287e-05, "loss": 1.454, "step": 5451500 }, { "epoch": 28.1, "learning_rate": 1.1057142857142858e-05, "loss": 1.4448, "step": 5452000 }, { "epoch": 28.11, "learning_rate": 1.1053571428571428e-05, "loss": 1.4217, "step": 5452500 }, { "epoch": 28.11, "learning_rate": 1.1050000000000001e-05, "loss": 1.4147, "step": 5453000 }, { "epoch": 28.11, "learning_rate": 1.1046428571428572e-05, "loss": 1.4117, "step": 5453500 }, { "epoch": 28.12, "learning_rate": 1.1042857142857144e-05, "loss": 1.4459, "step": 5454000 }, { "epoch": 28.12, "learning_rate": 1.1039285714285715e-05, "loss": 1.4169, "step": 5454500 }, { "epoch": 28.12, "learning_rate": 1.1035714285714286e-05, "loss": 1.4056, "step": 5455000 }, { "epoch": 28.12, "learning_rate": 1.1032142857142857e-05, "loss": 1.4083, "step": 5455500 }, { "epoch": 28.13, "learning_rate": 1.102857142857143e-05, "loss": 1.4189, "step": 5456000 }, { "epoch": 28.13, "learning_rate": 1.1025e-05, "loss": 1.4422, "step": 5456500 }, { "epoch": 28.13, "learning_rate": 1.1021428571428571e-05, "loss": 1.4046, "step": 5457000 }, { "epoch": 28.13, "learning_rate": 1.1017857142857144e-05, "loss": 1.4531, "step": 5457500 }, { "epoch": 28.14, "learning_rate": 1.1014285714285715e-05, "loss": 1.4278, "step": 5458000 }, { "epoch": 28.14, "learning_rate": 1.1010714285714287e-05, "loss": 1.4214, "step": 5458500 }, { "epoch": 28.14, "learning_rate": 1.1007142857142858e-05, "loss": 1.4109, "step": 5459000 }, { "epoch": 28.14, "learning_rate": 1.100357142857143e-05, "loss": 1.4365, "step": 5459500 }, { "epoch": 28.15, "learning_rate": 1.1000000000000001e-05, "loss": 1.4184, "step": 5460000 }, { "epoch": 28.15, "learning_rate": 1.099642857142857e-05, "loss": 1.4208, "step": 5460500 }, { "epoch": 28.15, "learning_rate": 1.0992857142857143e-05, "loss": 1.4166, "step": 5461000 }, { "epoch": 28.16, "learning_rate": 1.0989285714285714e-05, "loss": 1.4092, "step": 5461500 }, { "epoch": 28.16, "learning_rate": 1.0985714285714287e-05, "loss": 1.3993, "step": 5462000 }, { "epoch": 28.16, "learning_rate": 1.0982142857142857e-05, "loss": 1.4025, "step": 5462500 }, { "epoch": 28.16, "learning_rate": 1.097857142857143e-05, "loss": 1.4172, "step": 5463000 }, { "epoch": 28.17, "learning_rate": 1.0975e-05, "loss": 1.413, "step": 5463500 }, { "epoch": 28.17, "learning_rate": 1.0971428571428572e-05, "loss": 1.4141, "step": 5464000 }, { "epoch": 28.17, "learning_rate": 1.0967857142857144e-05, "loss": 1.3977, "step": 5464500 }, { "epoch": 28.17, "learning_rate": 1.0964285714285715e-05, "loss": 1.4173, "step": 5465000 }, { "epoch": 28.18, "learning_rate": 1.0960714285714286e-05, "loss": 1.4173, "step": 5465500 }, { "epoch": 28.18, "learning_rate": 1.0957142857142857e-05, "loss": 1.4088, "step": 5466000 }, { "epoch": 28.18, "learning_rate": 1.095357142857143e-05, "loss": 1.4076, "step": 5466500 }, { "epoch": 28.18, "learning_rate": 1.095e-05, "loss": 1.4121, "step": 5467000 }, { "epoch": 28.19, "learning_rate": 1.0946428571428571e-05, "loss": 1.4084, "step": 5467500 }, { "epoch": 28.19, "learning_rate": 1.0942857142857144e-05, "loss": 1.3999, "step": 5468000 }, { "epoch": 28.19, "learning_rate": 1.0939285714285714e-05, "loss": 1.4009, "step": 5468500 }, { "epoch": 28.2, "learning_rate": 1.0935714285714287e-05, "loss": 1.4136, "step": 5469000 }, { "epoch": 28.2, "learning_rate": 1.0932142857142858e-05, "loss": 1.4302, "step": 5469500 }, { "epoch": 28.2, "learning_rate": 1.092857142857143e-05, "loss": 1.4308, "step": 5470000 }, { "epoch": 28.2, "learning_rate": 1.0925000000000001e-05, "loss": 1.4395, "step": 5470500 }, { "epoch": 28.21, "learning_rate": 1.0921428571428572e-05, "loss": 1.4218, "step": 5471000 }, { "epoch": 28.21, "learning_rate": 1.0917857142857143e-05, "loss": 1.4219, "step": 5471500 }, { "epoch": 28.21, "learning_rate": 1.0914285714285714e-05, "loss": 1.4042, "step": 5472000 }, { "epoch": 28.21, "learning_rate": 1.0910714285714286e-05, "loss": 1.4062, "step": 5472500 }, { "epoch": 28.22, "learning_rate": 1.0907142857142857e-05, "loss": 1.4117, "step": 5473000 }, { "epoch": 28.22, "learning_rate": 1.090357142857143e-05, "loss": 1.4388, "step": 5473500 }, { "epoch": 28.22, "learning_rate": 1.09e-05, "loss": 1.4305, "step": 5474000 }, { "epoch": 28.22, "learning_rate": 1.0896428571428572e-05, "loss": 1.45, "step": 5474500 }, { "epoch": 28.23, "learning_rate": 1.0892857142857144e-05, "loss": 1.4716, "step": 5475000 }, { "epoch": 28.23, "learning_rate": 1.0889285714285715e-05, "loss": 1.4242, "step": 5475500 }, { "epoch": 28.23, "learning_rate": 1.0885714285714286e-05, "loss": 1.4144, "step": 5476000 }, { "epoch": 28.24, "learning_rate": 1.0882142857142857e-05, "loss": 1.4225, "step": 5476500 }, { "epoch": 28.24, "learning_rate": 1.087857142857143e-05, "loss": 1.4278, "step": 5477000 }, { "epoch": 28.24, "learning_rate": 1.0875e-05, "loss": 1.4425, "step": 5477500 }, { "epoch": 28.24, "learning_rate": 1.0871428571428571e-05, "loss": 1.4107, "step": 5478000 }, { "epoch": 28.25, "learning_rate": 1.0867857142857143e-05, "loss": 1.4269, "step": 5478500 }, { "epoch": 28.25, "learning_rate": 1.0864285714285714e-05, "loss": 1.4414, "step": 5479000 }, { "epoch": 28.25, "learning_rate": 1.0860714285714287e-05, "loss": 1.4479, "step": 5479500 }, { "epoch": 28.25, "learning_rate": 1.0857142857142858e-05, "loss": 1.4724, "step": 5480000 }, { "epoch": 28.26, "learning_rate": 1.085357142857143e-05, "loss": 1.4537, "step": 5480500 }, { "epoch": 28.26, "learning_rate": 1.0850000000000001e-05, "loss": 1.4446, "step": 5481000 }, { "epoch": 28.26, "learning_rate": 1.0846428571428572e-05, "loss": 1.4335, "step": 5481500 }, { "epoch": 28.26, "learning_rate": 1.0842857142857143e-05, "loss": 1.4294, "step": 5482000 }, { "epoch": 28.27, "learning_rate": 1.0839285714285714e-05, "loss": 1.4106, "step": 5482500 }, { "epoch": 28.27, "learning_rate": 1.0835714285714286e-05, "loss": 1.4211, "step": 5483000 }, { "epoch": 28.27, "learning_rate": 1.0832142857142857e-05, "loss": 1.4171, "step": 5483500 }, { "epoch": 28.28, "learning_rate": 1.082857142857143e-05, "loss": 1.4327, "step": 5484000 }, { "epoch": 28.28, "learning_rate": 1.0825e-05, "loss": 1.5694, "step": 5484500 }, { "epoch": 28.28, "learning_rate": 1.0821428571428573e-05, "loss": 1.4949, "step": 5485000 }, { "epoch": 28.28, "learning_rate": 1.0817857142857144e-05, "loss": 1.4245, "step": 5485500 }, { "epoch": 28.29, "learning_rate": 1.0814285714285715e-05, "loss": 1.4192, "step": 5486000 }, { "epoch": 28.29, "learning_rate": 1.0810714285714287e-05, "loss": 1.4568, "step": 5486500 }, { "epoch": 28.29, "learning_rate": 1.0807142857142857e-05, "loss": 1.4776, "step": 5487000 }, { "epoch": 28.29, "learning_rate": 1.0803571428571429e-05, "loss": 1.4702, "step": 5487500 }, { "epoch": 28.3, "learning_rate": 1.08e-05, "loss": 1.4661, "step": 5488000 }, { "epoch": 28.3, "learning_rate": 1.0796428571428572e-05, "loss": 1.4502, "step": 5488500 }, { "epoch": 28.3, "learning_rate": 1.0792857142857143e-05, "loss": 1.448, "step": 5489000 }, { "epoch": 28.3, "learning_rate": 1.0789285714285714e-05, "loss": 1.4293, "step": 5489500 }, { "epoch": 28.31, "learning_rate": 1.0785714285714287e-05, "loss": 1.4215, "step": 5490000 }, { "epoch": 28.31, "learning_rate": 1.0782142857142858e-05, "loss": 1.4396, "step": 5490500 }, { "epoch": 28.31, "learning_rate": 1.077857142857143e-05, "loss": 1.4229, "step": 5491000 }, { "epoch": 28.31, "learning_rate": 1.0775000000000001e-05, "loss": 1.4779, "step": 5491500 }, { "epoch": 28.32, "learning_rate": 1.0771428571428572e-05, "loss": 1.4285, "step": 5492000 }, { "epoch": 28.32, "learning_rate": 1.0767857142857143e-05, "loss": 1.4265, "step": 5492500 }, { "epoch": 28.32, "learning_rate": 1.0764285714285714e-05, "loss": 1.3971, "step": 5493000 }, { "epoch": 28.33, "learning_rate": 1.0760714285714286e-05, "loss": 1.4006, "step": 5493500 }, { "epoch": 28.33, "learning_rate": 1.0757142857142857e-05, "loss": 1.4002, "step": 5494000 }, { "epoch": 28.33, "learning_rate": 1.075357142857143e-05, "loss": 1.4068, "step": 5494500 }, { "epoch": 28.33, "learning_rate": 1.075e-05, "loss": 1.4005, "step": 5495000 }, { "epoch": 28.34, "learning_rate": 1.0746428571428573e-05, "loss": 1.3903, "step": 5495500 }, { "epoch": 28.34, "learning_rate": 1.0742857142857144e-05, "loss": 1.3974, "step": 5496000 }, { "epoch": 28.34, "learning_rate": 1.0739285714285715e-05, "loss": 1.4107, "step": 5496500 }, { "epoch": 28.34, "learning_rate": 1.0735714285714287e-05, "loss": 1.4148, "step": 5497000 }, { "epoch": 28.35, "learning_rate": 1.0732142857142858e-05, "loss": 1.4148, "step": 5497500 }, { "epoch": 28.35, "learning_rate": 1.0728571428571429e-05, "loss": 1.4228, "step": 5498000 }, { "epoch": 28.35, "learning_rate": 1.0725e-05, "loss": 1.4095, "step": 5498500 }, { "epoch": 28.35, "learning_rate": 1.0721428571428572e-05, "loss": 1.4002, "step": 5499000 }, { "epoch": 28.36, "learning_rate": 1.0717857142857143e-05, "loss": 1.4127, "step": 5499500 }, { "epoch": 28.36, "learning_rate": 1.0714285714285714e-05, "loss": 1.4068, "step": 5500000 }, { "epoch": 28.36, "learning_rate": 1.0710714285714287e-05, "loss": 1.3954, "step": 5500500 }, { "epoch": 28.37, "learning_rate": 1.0707142857142857e-05, "loss": 1.3995, "step": 5501000 }, { "epoch": 28.37, "learning_rate": 1.070357142857143e-05, "loss": 1.4124, "step": 5501500 }, { "epoch": 28.37, "learning_rate": 1.0700000000000001e-05, "loss": 1.4055, "step": 5502000 }, { "epoch": 28.37, "learning_rate": 1.0696428571428572e-05, "loss": 1.4011, "step": 5502500 }, { "epoch": 28.38, "learning_rate": 1.0692857142857143e-05, "loss": 1.4035, "step": 5503000 }, { "epoch": 28.38, "learning_rate": 1.0689285714285713e-05, "loss": 1.4034, "step": 5503500 }, { "epoch": 28.38, "learning_rate": 1.0685714285714286e-05, "loss": 1.3989, "step": 5504000 }, { "epoch": 28.38, "learning_rate": 1.0682142857142857e-05, "loss": 1.4103, "step": 5504500 }, { "epoch": 28.39, "learning_rate": 1.067857142857143e-05, "loss": 1.407, "step": 5505000 }, { "epoch": 28.39, "learning_rate": 1.0675e-05, "loss": 1.4031, "step": 5505500 }, { "epoch": 28.39, "learning_rate": 1.0671428571428573e-05, "loss": 1.4054, "step": 5506000 }, { "epoch": 28.39, "learning_rate": 1.0667857142857144e-05, "loss": 1.425, "step": 5506500 }, { "epoch": 28.4, "learning_rate": 1.0664285714285715e-05, "loss": 1.4343, "step": 5507000 }, { "epoch": 28.4, "learning_rate": 1.0660714285714287e-05, "loss": 1.4179, "step": 5507500 }, { "epoch": 28.4, "learning_rate": 1.0657142857142858e-05, "loss": 1.4041, "step": 5508000 }, { "epoch": 28.41, "learning_rate": 1.0653571428571429e-05, "loss": 1.4105, "step": 5508500 }, { "epoch": 28.41, "learning_rate": 1.065e-05, "loss": 1.4204, "step": 5509000 }, { "epoch": 28.41, "learning_rate": 1.0646428571428572e-05, "loss": 1.4086, "step": 5509500 }, { "epoch": 28.41, "learning_rate": 1.0642857142857143e-05, "loss": 1.4241, "step": 5510000 }, { "epoch": 28.42, "learning_rate": 1.0639285714285714e-05, "loss": 1.423, "step": 5510500 }, { "epoch": 28.42, "learning_rate": 1.0635714285714286e-05, "loss": 1.4201, "step": 5511000 }, { "epoch": 28.42, "learning_rate": 1.0632142857142857e-05, "loss": 1.4147, "step": 5511500 }, { "epoch": 28.42, "learning_rate": 1.062857142857143e-05, "loss": 1.4174, "step": 5512000 }, { "epoch": 28.43, "learning_rate": 1.0625e-05, "loss": 1.4203, "step": 5512500 }, { "epoch": 28.43, "learning_rate": 1.0621428571428572e-05, "loss": 1.4045, "step": 5513000 }, { "epoch": 28.43, "learning_rate": 1.0617857142857142e-05, "loss": 1.4142, "step": 5513500 }, { "epoch": 28.43, "learning_rate": 1.0614285714285713e-05, "loss": 1.4126, "step": 5514000 }, { "epoch": 28.44, "learning_rate": 1.0610714285714286e-05, "loss": 1.4076, "step": 5514500 }, { "epoch": 28.44, "learning_rate": 1.0607142857142857e-05, "loss": 1.4263, "step": 5515000 }, { "epoch": 28.44, "learning_rate": 1.060357142857143e-05, "loss": 1.4002, "step": 5515500 }, { "epoch": 28.45, "learning_rate": 1.06e-05, "loss": 1.4057, "step": 5516000 }, { "epoch": 28.45, "learning_rate": 1.0596428571428573e-05, "loss": 1.4322, "step": 5516500 }, { "epoch": 28.45, "learning_rate": 1.0592857142857144e-05, "loss": 1.4103, "step": 5517000 }, { "epoch": 28.45, "learning_rate": 1.0589285714285716e-05, "loss": 1.434, "step": 5517500 }, { "epoch": 28.46, "learning_rate": 1.0585714285714287e-05, "loss": 1.4324, "step": 5518000 }, { "epoch": 28.46, "learning_rate": 1.0582142857142858e-05, "loss": 1.4317, "step": 5518500 }, { "epoch": 28.46, "learning_rate": 1.0578571428571429e-05, "loss": 1.4063, "step": 5519000 }, { "epoch": 28.46, "learning_rate": 1.0575e-05, "loss": 1.4163, "step": 5519500 }, { "epoch": 28.47, "learning_rate": 1.0571428571428572e-05, "loss": 1.4052, "step": 5520000 }, { "epoch": 28.47, "learning_rate": 1.0567857142857143e-05, "loss": 1.39, "step": 5520500 }, { "epoch": 28.47, "learning_rate": 1.0564285714285716e-05, "loss": 1.4036, "step": 5521000 }, { "epoch": 28.47, "learning_rate": 1.0560714285714286e-05, "loss": 1.403, "step": 5521500 }, { "epoch": 28.48, "learning_rate": 1.0557142857142857e-05, "loss": 1.3959, "step": 5522000 }, { "epoch": 28.48, "learning_rate": 1.055357142857143e-05, "loss": 1.4167, "step": 5522500 }, { "epoch": 28.48, "learning_rate": 1.055e-05, "loss": 1.4002, "step": 5523000 }, { "epoch": 28.49, "learning_rate": 1.0546428571428573e-05, "loss": 1.4051, "step": 5523500 }, { "epoch": 28.49, "learning_rate": 1.0542857142857144e-05, "loss": 1.4128, "step": 5524000 }, { "epoch": 28.49, "learning_rate": 1.0539285714285715e-05, "loss": 1.4179, "step": 5524500 }, { "epoch": 28.49, "learning_rate": 1.0535714285714286e-05, "loss": 1.4392, "step": 5525000 }, { "epoch": 28.5, "learning_rate": 1.0532142857142857e-05, "loss": 1.4385, "step": 5525500 }, { "epoch": 28.5, "learning_rate": 1.052857142857143e-05, "loss": 1.4249, "step": 5526000 }, { "epoch": 28.5, "learning_rate": 1.0525e-05, "loss": 1.4064, "step": 5526500 }, { "epoch": 28.5, "learning_rate": 1.0521428571428573e-05, "loss": 1.4088, "step": 5527000 }, { "epoch": 28.51, "learning_rate": 1.0517857142857143e-05, "loss": 1.4045, "step": 5527500 }, { "epoch": 28.51, "learning_rate": 1.0514285714285716e-05, "loss": 1.422, "step": 5528000 }, { "epoch": 28.51, "learning_rate": 1.0510714285714287e-05, "loss": 1.4128, "step": 5528500 }, { "epoch": 28.51, "learning_rate": 1.0507142857142858e-05, "loss": 1.4058, "step": 5529000 }, { "epoch": 28.52, "learning_rate": 1.0503571428571429e-05, "loss": 1.4268, "step": 5529500 }, { "epoch": 28.52, "learning_rate": 1.05e-05, "loss": 1.4469, "step": 5530000 }, { "epoch": 28.52, "learning_rate": 1.0496428571428572e-05, "loss": 1.4425, "step": 5530500 }, { "epoch": 28.53, "learning_rate": 1.0492857142857143e-05, "loss": 1.4133, "step": 5531000 }, { "epoch": 28.53, "learning_rate": 1.0489285714285715e-05, "loss": 1.4318, "step": 5531500 }, { "epoch": 28.53, "learning_rate": 1.0485714285714286e-05, "loss": 1.4256, "step": 5532000 }, { "epoch": 28.53, "learning_rate": 1.0482142857142857e-05, "loss": 1.4317, "step": 5532500 }, { "epoch": 28.54, "learning_rate": 1.047857142857143e-05, "loss": 1.4467, "step": 5533000 }, { "epoch": 28.54, "learning_rate": 1.0475e-05, "loss": 1.429, "step": 5533500 }, { "epoch": 28.54, "learning_rate": 1.0471428571428573e-05, "loss": 1.4654, "step": 5534000 }, { "epoch": 28.54, "learning_rate": 1.0467857142857144e-05, "loss": 1.4465, "step": 5534500 }, { "epoch": 28.55, "learning_rate": 1.0464285714285715e-05, "loss": 1.4577, "step": 5535000 }, { "epoch": 28.55, "learning_rate": 1.0460714285714286e-05, "loss": 1.4195, "step": 5535500 }, { "epoch": 28.55, "learning_rate": 1.0457142857142856e-05, "loss": 1.4284, "step": 5536000 }, { "epoch": 28.55, "learning_rate": 1.0453571428571429e-05, "loss": 1.4074, "step": 5536500 }, { "epoch": 28.56, "learning_rate": 1.045e-05, "loss": 1.4132, "step": 5537000 }, { "epoch": 28.56, "learning_rate": 1.0446428571428572e-05, "loss": 1.4052, "step": 5537500 }, { "epoch": 28.56, "learning_rate": 1.0442857142857143e-05, "loss": 1.4142, "step": 5538000 }, { "epoch": 28.57, "learning_rate": 1.0439285714285716e-05, "loss": 1.42, "step": 5538500 }, { "epoch": 28.57, "learning_rate": 1.0435714285714287e-05, "loss": 1.4056, "step": 5539000 }, { "epoch": 28.57, "learning_rate": 1.0432142857142858e-05, "loss": 1.4194, "step": 5539500 }, { "epoch": 28.57, "learning_rate": 1.0428571428571428e-05, "loss": 1.4025, "step": 5540000 }, { "epoch": 28.58, "learning_rate": 1.0425e-05, "loss": 1.3913, "step": 5540500 }, { "epoch": 28.58, "learning_rate": 1.0421428571428572e-05, "loss": 1.405, "step": 5541000 }, { "epoch": 28.58, "learning_rate": 1.0417857142857143e-05, "loss": 1.4064, "step": 5541500 }, { "epoch": 28.58, "learning_rate": 1.0414285714285715e-05, "loss": 1.406, "step": 5542000 }, { "epoch": 28.59, "learning_rate": 1.0410714285714286e-05, "loss": 1.4167, "step": 5542500 }, { "epoch": 28.59, "learning_rate": 1.0407142857142857e-05, "loss": 1.4095, "step": 5543000 }, { "epoch": 28.59, "learning_rate": 1.040357142857143e-05, "loss": 1.4123, "step": 5543500 }, { "epoch": 28.59, "learning_rate": 1.04e-05, "loss": 1.4078, "step": 5544000 }, { "epoch": 28.6, "learning_rate": 1.0396428571428573e-05, "loss": 1.4043, "step": 5544500 }, { "epoch": 28.6, "learning_rate": 1.0392857142857144e-05, "loss": 1.4041, "step": 5545000 }, { "epoch": 28.6, "learning_rate": 1.0389285714285715e-05, "loss": 1.4053, "step": 5545500 }, { "epoch": 28.61, "learning_rate": 1.0385714285714286e-05, "loss": 1.417, "step": 5546000 }, { "epoch": 28.61, "learning_rate": 1.0382142857142856e-05, "loss": 1.3993, "step": 5546500 }, { "epoch": 28.61, "learning_rate": 1.0378571428571429e-05, "loss": 1.4028, "step": 5547000 }, { "epoch": 28.61, "learning_rate": 1.0375e-05, "loss": 1.3944, "step": 5547500 }, { "epoch": 28.62, "learning_rate": 1.0371428571428572e-05, "loss": 1.4052, "step": 5548000 }, { "epoch": 28.62, "learning_rate": 1.0367857142857143e-05, "loss": 1.393, "step": 5548500 }, { "epoch": 28.62, "learning_rate": 1.0364285714285716e-05, "loss": 1.4027, "step": 5549000 }, { "epoch": 28.62, "learning_rate": 1.0360714285714287e-05, "loss": 1.3942, "step": 5549500 }, { "epoch": 28.63, "learning_rate": 1.0357142857142859e-05, "loss": 1.4021, "step": 5550000 }, { "epoch": 28.63, "learning_rate": 1.0353571428571428e-05, "loss": 1.3917, "step": 5550500 }, { "epoch": 28.63, "learning_rate": 1.035e-05, "loss": 1.4011, "step": 5551000 }, { "epoch": 28.63, "learning_rate": 1.0346428571428572e-05, "loss": 1.3973, "step": 5551500 }, { "epoch": 28.64, "learning_rate": 1.0342857142857143e-05, "loss": 1.4028, "step": 5552000 }, { "epoch": 28.64, "learning_rate": 1.0339285714285715e-05, "loss": 1.4086, "step": 5552500 }, { "epoch": 28.64, "learning_rate": 1.0335714285714286e-05, "loss": 1.4129, "step": 5553000 }, { "epoch": 28.65, "learning_rate": 1.0332142857142859e-05, "loss": 1.4058, "step": 5553500 }, { "epoch": 28.65, "learning_rate": 1.032857142857143e-05, "loss": 1.3953, "step": 5554000 }, { "epoch": 28.65, "learning_rate": 1.0325e-05, "loss": 1.4028, "step": 5554500 }, { "epoch": 28.65, "learning_rate": 1.0321428571428573e-05, "loss": 1.4074, "step": 5555000 }, { "epoch": 28.66, "learning_rate": 1.0317857142857144e-05, "loss": 1.4033, "step": 5555500 }, { "epoch": 28.66, "learning_rate": 1.0314285714285715e-05, "loss": 1.4035, "step": 5556000 }, { "epoch": 28.66, "learning_rate": 1.0310714285714285e-05, "loss": 1.4012, "step": 5556500 }, { "epoch": 28.66, "learning_rate": 1.0307142857142858e-05, "loss": 1.4008, "step": 5557000 }, { "epoch": 28.67, "learning_rate": 1.0303571428571429e-05, "loss": 1.406, "step": 5557500 }, { "epoch": 28.67, "learning_rate": 1.03e-05, "loss": 1.4035, "step": 5558000 }, { "epoch": 28.67, "learning_rate": 1.0296428571428572e-05, "loss": 1.414, "step": 5558500 }, { "epoch": 28.67, "learning_rate": 1.0292857142857143e-05, "loss": 1.4187, "step": 5559000 }, { "epoch": 28.68, "learning_rate": 1.0289285714285716e-05, "loss": 1.4258, "step": 5559500 }, { "epoch": 28.68, "learning_rate": 1.0285714285714286e-05, "loss": 1.428, "step": 5560000 }, { "epoch": 28.68, "learning_rate": 1.0282142857142859e-05, "loss": 1.4333, "step": 5560500 }, { "epoch": 28.69, "learning_rate": 1.027857142857143e-05, "loss": 1.4416, "step": 5561000 }, { "epoch": 28.69, "learning_rate": 1.0275e-05, "loss": 1.4469, "step": 5561500 }, { "epoch": 28.69, "learning_rate": 1.0271428571428572e-05, "loss": 1.4215, "step": 5562000 }, { "epoch": 28.69, "learning_rate": 1.0267857142857142e-05, "loss": 1.4367, "step": 5562500 }, { "epoch": 28.7, "learning_rate": 1.0264285714285715e-05, "loss": 1.4179, "step": 5563000 }, { "epoch": 28.7, "learning_rate": 1.0260714285714286e-05, "loss": 1.4101, "step": 5563500 }, { "epoch": 28.7, "learning_rate": 1.0257142857142858e-05, "loss": 1.4161, "step": 5564000 }, { "epoch": 28.7, "learning_rate": 1.025357142857143e-05, "loss": 1.4017, "step": 5564500 }, { "epoch": 28.71, "learning_rate": 1.025e-05, "loss": 1.4089, "step": 5565000 }, { "epoch": 28.71, "learning_rate": 1.0246428571428573e-05, "loss": 1.4019, "step": 5565500 }, { "epoch": 28.71, "learning_rate": 1.0242857142857144e-05, "loss": 1.4043, "step": 5566000 }, { "epoch": 28.71, "learning_rate": 1.0239285714285714e-05, "loss": 1.4073, "step": 5566500 }, { "epoch": 28.72, "learning_rate": 1.0235714285714285e-05, "loss": 1.4043, "step": 5567000 }, { "epoch": 28.72, "learning_rate": 1.0232142857142858e-05, "loss": 1.3995, "step": 5567500 }, { "epoch": 28.72, "learning_rate": 1.0228571428571429e-05, "loss": 1.3985, "step": 5568000 }, { "epoch": 28.72, "learning_rate": 1.0225e-05, "loss": 1.4032, "step": 5568500 }, { "epoch": 28.73, "learning_rate": 1.0221428571428572e-05, "loss": 1.4146, "step": 5569000 }, { "epoch": 28.73, "learning_rate": 1.0217857142857143e-05, "loss": 1.4052, "step": 5569500 }, { "epoch": 28.73, "learning_rate": 1.0214285714285715e-05, "loss": 1.4066, "step": 5570000 }, { "epoch": 28.74, "learning_rate": 1.0210714285714286e-05, "loss": 1.4097, "step": 5570500 }, { "epoch": 28.74, "learning_rate": 1.0207142857142859e-05, "loss": 1.3994, "step": 5571000 }, { "epoch": 28.74, "learning_rate": 1.020357142857143e-05, "loss": 1.3979, "step": 5571500 }, { "epoch": 28.74, "learning_rate": 1.02e-05, "loss": 1.4003, "step": 5572000 }, { "epoch": 28.75, "learning_rate": 1.0196428571428571e-05, "loss": 1.4206, "step": 5572500 }, { "epoch": 28.75, "learning_rate": 1.0192857142857142e-05, "loss": 1.4518, "step": 5573000 }, { "epoch": 28.75, "learning_rate": 1.0189285714285715e-05, "loss": 1.4263, "step": 5573500 }, { "epoch": 28.75, "learning_rate": 1.0185714285714286e-05, "loss": 1.4321, "step": 5574000 }, { "epoch": 28.76, "learning_rate": 1.0182142857142858e-05, "loss": 1.408, "step": 5574500 }, { "epoch": 28.76, "learning_rate": 1.0178571428571429e-05, "loss": 1.4211, "step": 5575000 }, { "epoch": 28.76, "learning_rate": 1.0175e-05, "loss": 1.4204, "step": 5575500 }, { "epoch": 28.76, "learning_rate": 1.0171428571428573e-05, "loss": 1.4201, "step": 5576000 }, { "epoch": 28.77, "learning_rate": 1.0167857142857143e-05, "loss": 1.4157, "step": 5576500 }, { "epoch": 28.77, "learning_rate": 1.0164285714285714e-05, "loss": 1.4195, "step": 5577000 }, { "epoch": 28.77, "learning_rate": 1.0160714285714285e-05, "loss": 1.4239, "step": 5577500 }, { "epoch": 28.78, "learning_rate": 1.0157142857142858e-05, "loss": 1.4209, "step": 5578000 }, { "epoch": 28.78, "learning_rate": 1.0153571428571429e-05, "loss": 1.4303, "step": 5578500 }, { "epoch": 28.78, "learning_rate": 1.0150000000000001e-05, "loss": 1.4091, "step": 5579000 }, { "epoch": 28.78, "learning_rate": 1.0146428571428572e-05, "loss": 1.4142, "step": 5579500 }, { "epoch": 28.79, "learning_rate": 1.0142857142857143e-05, "loss": 1.4204, "step": 5580000 }, { "epoch": 28.79, "learning_rate": 1.0139285714285715e-05, "loss": 1.4334, "step": 5580500 }, { "epoch": 28.79, "learning_rate": 1.0135714285714286e-05, "loss": 1.4878, "step": 5581000 }, { "epoch": 28.79, "learning_rate": 1.0132142857142859e-05, "loss": 1.4388, "step": 5581500 }, { "epoch": 28.8, "learning_rate": 1.012857142857143e-05, "loss": 1.4192, "step": 5582000 }, { "epoch": 28.8, "learning_rate": 1.0125e-05, "loss": 1.4229, "step": 5582500 }, { "epoch": 28.8, "learning_rate": 1.0121428571428571e-05, "loss": 1.4271, "step": 5583000 }, { "epoch": 28.8, "learning_rate": 1.0117857142857142e-05, "loss": 1.4042, "step": 5583500 }, { "epoch": 28.81, "learning_rate": 1.0114285714285715e-05, "loss": 1.4119, "step": 5584000 }, { "epoch": 28.81, "learning_rate": 1.0110714285714286e-05, "loss": 1.3901, "step": 5584500 }, { "epoch": 28.81, "learning_rate": 1.0107142857142858e-05, "loss": 1.3908, "step": 5585000 }, { "epoch": 28.82, "learning_rate": 1.0103571428571429e-05, "loss": 1.3914, "step": 5585500 }, { "epoch": 28.82, "learning_rate": 1.0100000000000002e-05, "loss": 1.3985, "step": 5586000 }, { "epoch": 28.82, "learning_rate": 1.0096428571428572e-05, "loss": 1.4021, "step": 5586500 }, { "epoch": 28.82, "learning_rate": 1.0092857142857143e-05, "loss": 1.4092, "step": 5587000 }, { "epoch": 28.83, "learning_rate": 1.0089285714285716e-05, "loss": 1.4139, "step": 5587500 }, { "epoch": 28.83, "learning_rate": 1.0085714285714287e-05, "loss": 1.3871, "step": 5588000 }, { "epoch": 28.83, "learning_rate": 1.0082142857142858e-05, "loss": 1.402, "step": 5588500 }, { "epoch": 28.83, "learning_rate": 1.0078571428571428e-05, "loss": 1.3876, "step": 5589000 }, { "epoch": 28.84, "learning_rate": 1.0075000000000001e-05, "loss": 1.3998, "step": 5589500 }, { "epoch": 28.84, "learning_rate": 1.0071428571428572e-05, "loss": 1.3975, "step": 5590000 }, { "epoch": 28.84, "learning_rate": 1.0067857142857143e-05, "loss": 1.3874, "step": 5590500 }, { "epoch": 28.84, "learning_rate": 1.0064285714285715e-05, "loss": 1.3933, "step": 5591000 }, { "epoch": 28.85, "learning_rate": 1.0060714285714286e-05, "loss": 1.392, "step": 5591500 }, { "epoch": 28.85, "learning_rate": 1.0057142857142859e-05, "loss": 1.3983, "step": 5592000 }, { "epoch": 28.85, "learning_rate": 1.005357142857143e-05, "loss": 1.3942, "step": 5592500 }, { "epoch": 28.86, "learning_rate": 1.005e-05, "loss": 1.3985, "step": 5593000 }, { "epoch": 28.86, "learning_rate": 1.0046428571428571e-05, "loss": 1.399, "step": 5593500 }, { "epoch": 28.86, "learning_rate": 1.0042857142857142e-05, "loss": 1.3935, "step": 5594000 }, { "epoch": 28.86, "learning_rate": 1.0039285714285715e-05, "loss": 1.4037, "step": 5594500 }, { "epoch": 28.87, "learning_rate": 1.0035714285714285e-05, "loss": 1.4266, "step": 5595000 }, { "epoch": 28.87, "learning_rate": 1.0032142857142858e-05, "loss": 1.4256, "step": 5595500 }, { "epoch": 28.87, "learning_rate": 1.0028571428571429e-05, "loss": 1.4236, "step": 5596000 }, { "epoch": 28.87, "learning_rate": 1.0025000000000001e-05, "loss": 1.4199, "step": 5596500 }, { "epoch": 28.88, "learning_rate": 1.0021428571428572e-05, "loss": 1.42, "step": 5597000 }, { "epoch": 28.88, "learning_rate": 1.0017857142857143e-05, "loss": 1.4099, "step": 5597500 }, { "epoch": 28.88, "learning_rate": 1.0014285714285716e-05, "loss": 1.4067, "step": 5598000 }, { "epoch": 28.88, "learning_rate": 1.0010714285714287e-05, "loss": 1.4421, "step": 5598500 }, { "epoch": 28.89, "learning_rate": 1.0007142857142857e-05, "loss": 1.4124, "step": 5599000 }, { "epoch": 28.89, "learning_rate": 1.0003571428571428e-05, "loss": 1.4748, "step": 5599500 }, { "epoch": 28.89, "learning_rate": 1e-05, "loss": 1.4156, "step": 5600000 }, { "epoch": 28.9, "learning_rate": 9.996428571428572e-06, "loss": 1.4458, "step": 5600500 }, { "epoch": 28.9, "learning_rate": 9.992857142857143e-06, "loss": 1.4575, "step": 5601000 }, { "epoch": 28.9, "learning_rate": 9.989285714285715e-06, "loss": 1.486, "step": 5601500 }, { "epoch": 28.9, "learning_rate": 9.985714285714286e-06, "loss": 1.5065, "step": 5602000 }, { "epoch": 28.91, "learning_rate": 9.982142857142858e-06, "loss": 1.4256, "step": 5602500 }, { "epoch": 28.91, "learning_rate": 9.97857142857143e-06, "loss": 1.4202, "step": 5603000 }, { "epoch": 28.91, "learning_rate": 9.975e-06, "loss": 1.4106, "step": 5603500 }, { "epoch": 28.91, "learning_rate": 9.971428571428571e-06, "loss": 1.4082, "step": 5604000 }, { "epoch": 28.92, "learning_rate": 9.967857142857142e-06, "loss": 1.3983, "step": 5604500 }, { "epoch": 28.92, "learning_rate": 9.964285714285714e-06, "loss": 1.4074, "step": 5605000 }, { "epoch": 28.92, "learning_rate": 9.960714285714285e-06, "loss": 1.4213, "step": 5605500 }, { "epoch": 28.92, "learning_rate": 9.957142857142858e-06, "loss": 1.4191, "step": 5606000 }, { "epoch": 28.93, "learning_rate": 9.953571428571429e-06, "loss": 1.4058, "step": 5606500 }, { "epoch": 28.93, "learning_rate": 9.950000000000001e-06, "loss": 1.3953, "step": 5607000 }, { "epoch": 28.93, "learning_rate": 9.946428571428572e-06, "loss": 1.393, "step": 5607500 }, { "epoch": 28.94, "learning_rate": 9.942857142857143e-06, "loss": 1.4088, "step": 5608000 }, { "epoch": 28.94, "learning_rate": 9.939285714285716e-06, "loss": 1.4114, "step": 5608500 }, { "epoch": 28.94, "learning_rate": 9.935714285714286e-06, "loss": 1.4144, "step": 5609000 }, { "epoch": 28.94, "learning_rate": 9.932142857142857e-06, "loss": 1.4053, "step": 5609500 }, { "epoch": 28.95, "learning_rate": 9.928571428571428e-06, "loss": 1.4036, "step": 5610000 }, { "epoch": 28.95, "learning_rate": 9.925e-06, "loss": 1.3994, "step": 5610500 }, { "epoch": 28.95, "learning_rate": 9.921428571428572e-06, "loss": 1.4085, "step": 5611000 }, { "epoch": 28.95, "learning_rate": 9.917857142857144e-06, "loss": 1.3979, "step": 5611500 }, { "epoch": 28.96, "learning_rate": 9.914285714285715e-06, "loss": 1.3911, "step": 5612000 }, { "epoch": 28.96, "learning_rate": 9.910714285714286e-06, "loss": 1.4003, "step": 5612500 }, { "epoch": 28.96, "learning_rate": 9.907142857142858e-06, "loss": 1.4007, "step": 5613000 }, { "epoch": 28.96, "learning_rate": 9.90357142857143e-06, "loss": 1.3937, "step": 5613500 }, { "epoch": 28.97, "learning_rate": 9.900000000000002e-06, "loss": 1.3995, "step": 5614000 }, { "epoch": 28.97, "learning_rate": 9.896428571428571e-06, "loss": 1.3959, "step": 5614500 }, { "epoch": 28.97, "learning_rate": 9.892857142857143e-06, "loss": 1.3951, "step": 5615000 }, { "epoch": 28.98, "learning_rate": 9.889285714285714e-06, "loss": 1.405, "step": 5615500 }, { "epoch": 28.98, "learning_rate": 9.885714285714285e-06, "loss": 1.3996, "step": 5616000 }, { "epoch": 28.98, "learning_rate": 9.882142857142858e-06, "loss": 1.4036, "step": 5616500 }, { "epoch": 28.98, "learning_rate": 9.878571428571429e-06, "loss": 1.3963, "step": 5617000 }, { "epoch": 28.99, "learning_rate": 9.875000000000001e-06, "loss": 1.3866, "step": 5617500 }, { "epoch": 28.99, "learning_rate": 9.871428571428572e-06, "loss": 1.3968, "step": 5618000 }, { "epoch": 28.99, "learning_rate": 9.867857142857145e-06, "loss": 1.3939, "step": 5618500 }, { "epoch": 28.99, "learning_rate": 9.864285714285715e-06, "loss": 1.4065, "step": 5619000 }, { "epoch": 29.0, "learning_rate": 9.860714285714286e-06, "loss": 1.4242, "step": 5619500 }, { "epoch": 29.0, "learning_rate": 9.857142857142857e-06, "loss": 1.4235, "step": 5620000 }, { "epoch": 29.0, "learning_rate": 9.853571428571428e-06, "loss": 1.417, "step": 5620500 }, { "epoch": 29.0, "learning_rate": 9.85e-06, "loss": 1.4074, "step": 5621000 }, { "epoch": 29.01, "learning_rate": 9.846428571428571e-06, "loss": 1.3951, "step": 5621500 }, { "epoch": 29.01, "learning_rate": 9.842857142857144e-06, "loss": 1.3875, "step": 5622000 }, { "epoch": 29.01, "learning_rate": 9.839285714285715e-06, "loss": 1.3956, "step": 5622500 }, { "epoch": 29.02, "learning_rate": 9.835714285714286e-06, "loss": 1.4023, "step": 5623000 }, { "epoch": 29.02, "learning_rate": 9.832142857142858e-06, "loss": 1.4137, "step": 5623500 }, { "epoch": 29.02, "learning_rate": 9.828571428571429e-06, "loss": 1.4096, "step": 5624000 }, { "epoch": 29.02, "learning_rate": 9.825000000000002e-06, "loss": 1.4008, "step": 5624500 }, { "epoch": 29.03, "learning_rate": 9.821428571428573e-06, "loss": 1.4043, "step": 5625000 }, { "epoch": 29.03, "learning_rate": 9.817857142857143e-06, "loss": 1.3885, "step": 5625500 }, { "epoch": 29.03, "learning_rate": 9.814285714285714e-06, "loss": 1.3923, "step": 5626000 }, { "epoch": 29.03, "learning_rate": 9.810714285714285e-06, "loss": 1.4034, "step": 5626500 }, { "epoch": 29.04, "learning_rate": 9.807142857142858e-06, "loss": 1.4037, "step": 5627000 }, { "epoch": 29.04, "learning_rate": 9.803571428571428e-06, "loss": 1.3981, "step": 5627500 }, { "epoch": 29.04, "learning_rate": 9.800000000000001e-06, "loss": 1.4169, "step": 5628000 }, { "epoch": 29.04, "learning_rate": 9.796428571428572e-06, "loss": 1.403, "step": 5628500 }, { "epoch": 29.05, "learning_rate": 9.792857142857144e-06, "loss": 1.395, "step": 5629000 }, { "epoch": 29.05, "learning_rate": 9.789285714285715e-06, "loss": 1.3904, "step": 5629500 }, { "epoch": 29.05, "learning_rate": 9.785714285714286e-06, "loss": 1.3984, "step": 5630000 }, { "epoch": 29.06, "learning_rate": 9.782142857142857e-06, "loss": 1.4124, "step": 5630500 }, { "epoch": 29.06, "learning_rate": 9.778571428571428e-06, "loss": 1.4245, "step": 5631000 }, { "epoch": 29.06, "learning_rate": 9.775e-06, "loss": 1.4057, "step": 5631500 }, { "epoch": 29.06, "learning_rate": 9.771428571428571e-06, "loss": 1.4308, "step": 5632000 }, { "epoch": 29.07, "learning_rate": 9.767857142857144e-06, "loss": 1.4668, "step": 5632500 }, { "epoch": 29.07, "learning_rate": 9.764285714285715e-06, "loss": 1.4172, "step": 5633000 }, { "epoch": 29.07, "learning_rate": 9.760714285714286e-06, "loss": 1.4027, "step": 5633500 }, { "epoch": 29.07, "learning_rate": 9.757142857142858e-06, "loss": 1.4011, "step": 5634000 }, { "epoch": 29.08, "learning_rate": 9.753571428571429e-06, "loss": 1.402, "step": 5634500 }, { "epoch": 29.08, "learning_rate": 9.750000000000002e-06, "loss": 1.4105, "step": 5635000 }, { "epoch": 29.08, "learning_rate": 9.746428571428572e-06, "loss": 1.4299, "step": 5635500 }, { "epoch": 29.08, "learning_rate": 9.742857142857143e-06, "loss": 1.438, "step": 5636000 }, { "epoch": 29.09, "learning_rate": 9.739285714285714e-06, "loss": 1.4075, "step": 5636500 }, { "epoch": 29.09, "learning_rate": 9.735714285714285e-06, "loss": 1.4198, "step": 5637000 }, { "epoch": 29.09, "learning_rate": 9.732142857142858e-06, "loss": 1.4135, "step": 5637500 }, { "epoch": 29.1, "learning_rate": 9.728571428571428e-06, "loss": 1.4422, "step": 5638000 }, { "epoch": 29.1, "learning_rate": 9.725000000000001e-06, "loss": 1.4196, "step": 5638500 }, { "epoch": 29.1, "learning_rate": 9.721428571428572e-06, "loss": 1.4276, "step": 5639000 }, { "epoch": 29.1, "learning_rate": 9.717857142857144e-06, "loss": 1.4306, "step": 5639500 }, { "epoch": 29.11, "learning_rate": 9.714285714285715e-06, "loss": 1.4229, "step": 5640000 }, { "epoch": 29.11, "learning_rate": 9.710714285714286e-06, "loss": 1.4254, "step": 5640500 }, { "epoch": 29.11, "learning_rate": 9.707142857142857e-06, "loss": 1.4028, "step": 5641000 }, { "epoch": 29.11, "learning_rate": 9.703571428571428e-06, "loss": 1.4066, "step": 5641500 }, { "epoch": 29.12, "learning_rate": 9.7e-06, "loss": 1.4026, "step": 5642000 }, { "epoch": 29.12, "learning_rate": 9.696428571428571e-06, "loss": 1.4302, "step": 5642500 }, { "epoch": 29.12, "learning_rate": 9.692857142857144e-06, "loss": 1.4218, "step": 5643000 }, { "epoch": 29.12, "learning_rate": 9.689285714285715e-06, "loss": 1.4019, "step": 5643500 }, { "epoch": 29.13, "learning_rate": 9.685714285714287e-06, "loss": 1.4011, "step": 5644000 }, { "epoch": 29.13, "learning_rate": 9.682142857142858e-06, "loss": 1.4372, "step": 5644500 }, { "epoch": 29.13, "learning_rate": 9.678571428571429e-06, "loss": 1.3944, "step": 5645000 }, { "epoch": 29.13, "learning_rate": 9.675000000000001e-06, "loss": 1.3956, "step": 5645500 }, { "epoch": 29.14, "learning_rate": 9.671428571428572e-06, "loss": 1.3959, "step": 5646000 }, { "epoch": 29.14, "learning_rate": 9.667857142857143e-06, "loss": 1.3972, "step": 5646500 }, { "epoch": 29.14, "learning_rate": 9.664285714285714e-06, "loss": 1.3937, "step": 5647000 }, { "epoch": 29.15, "learning_rate": 9.660714285714287e-06, "loss": 1.3874, "step": 5647500 }, { "epoch": 29.15, "learning_rate": 9.657142857142857e-06, "loss": 1.3949, "step": 5648000 }, { "epoch": 29.15, "learning_rate": 9.653571428571428e-06, "loss": 1.408, "step": 5648500 }, { "epoch": 29.15, "learning_rate": 9.65e-06, "loss": 1.3977, "step": 5649000 }, { "epoch": 29.16, "learning_rate": 9.646428571428572e-06, "loss": 1.4018, "step": 5649500 }, { "epoch": 29.16, "learning_rate": 9.642857142857144e-06, "loss": 1.3972, "step": 5650000 }, { "epoch": 29.16, "learning_rate": 9.639285714285715e-06, "loss": 1.3924, "step": 5650500 }, { "epoch": 29.16, "learning_rate": 9.635714285714288e-06, "loss": 1.395, "step": 5651000 }, { "epoch": 29.17, "learning_rate": 9.632142857142858e-06, "loss": 1.3953, "step": 5651500 }, { "epoch": 29.17, "learning_rate": 9.628571428571428e-06, "loss": 1.3964, "step": 5652000 }, { "epoch": 29.17, "learning_rate": 9.625e-06, "loss": 1.3928, "step": 5652500 }, { "epoch": 29.17, "learning_rate": 9.621428571428571e-06, "loss": 1.3889, "step": 5653000 }, { "epoch": 29.18, "learning_rate": 9.617857142857144e-06, "loss": 1.3938, "step": 5653500 }, { "epoch": 29.18, "learning_rate": 9.614285714285714e-06, "loss": 1.3959, "step": 5654000 }, { "epoch": 29.18, "learning_rate": 9.610714285714287e-06, "loss": 1.3927, "step": 5654500 }, { "epoch": 29.19, "learning_rate": 9.607142857142858e-06, "loss": 1.3949, "step": 5655000 }, { "epoch": 29.19, "learning_rate": 9.603571428571429e-06, "loss": 1.4043, "step": 5655500 }, { "epoch": 29.19, "learning_rate": 9.600000000000001e-06, "loss": 1.3974, "step": 5656000 }, { "epoch": 29.19, "learning_rate": 9.596428571428572e-06, "loss": 1.4039, "step": 5656500 }, { "epoch": 29.2, "learning_rate": 9.592857142857143e-06, "loss": 1.4051, "step": 5657000 }, { "epoch": 29.2, "learning_rate": 9.589285714285714e-06, "loss": 1.3917, "step": 5657500 }, { "epoch": 29.2, "learning_rate": 9.585714285714286e-06, "loss": 1.4101, "step": 5658000 }, { "epoch": 29.2, "learning_rate": 9.582142857142857e-06, "loss": 1.3888, "step": 5658500 }, { "epoch": 29.21, "learning_rate": 9.578571428571428e-06, "loss": 1.3957, "step": 5659000 }, { "epoch": 29.21, "learning_rate": 9.575e-06, "loss": 1.3935, "step": 5659500 }, { "epoch": 29.21, "learning_rate": 9.571428571428572e-06, "loss": 1.4103, "step": 5660000 }, { "epoch": 29.21, "learning_rate": 9.567857142857144e-06, "loss": 1.3838, "step": 5660500 }, { "epoch": 29.22, "learning_rate": 9.564285714285715e-06, "loss": 1.4004, "step": 5661000 }, { "epoch": 29.22, "learning_rate": 9.560714285714287e-06, "loss": 1.4039, "step": 5661500 }, { "epoch": 29.22, "learning_rate": 9.557142857142858e-06, "loss": 1.4001, "step": 5662000 }, { "epoch": 29.23, "learning_rate": 9.55357142857143e-06, "loss": 1.3882, "step": 5662500 }, { "epoch": 29.23, "learning_rate": 9.55e-06, "loss": 1.3945, "step": 5663000 }, { "epoch": 29.23, "learning_rate": 9.546428571428571e-06, "loss": 1.3997, "step": 5663500 }, { "epoch": 29.23, "learning_rate": 9.542857142857143e-06, "loss": 1.3816, "step": 5664000 }, { "epoch": 29.24, "learning_rate": 9.539285714285714e-06, "loss": 1.3854, "step": 5664500 }, { "epoch": 29.24, "learning_rate": 9.535714285714287e-06, "loss": 1.3937, "step": 5665000 }, { "epoch": 29.24, "learning_rate": 9.532142857142858e-06, "loss": 1.3907, "step": 5665500 }, { "epoch": 29.24, "learning_rate": 9.528571428571429e-06, "loss": 1.3939, "step": 5666000 }, { "epoch": 29.25, "learning_rate": 9.525000000000001e-06, "loss": 1.3917, "step": 5666500 }, { "epoch": 29.25, "learning_rate": 9.521428571428572e-06, "loss": 1.3961, "step": 5667000 }, { "epoch": 29.25, "learning_rate": 9.517857142857143e-06, "loss": 1.3897, "step": 5667500 }, { "epoch": 29.25, "learning_rate": 9.514285714285714e-06, "loss": 1.3951, "step": 5668000 }, { "epoch": 29.26, "learning_rate": 9.510714285714286e-06, "loss": 1.4012, "step": 5668500 }, { "epoch": 29.26, "learning_rate": 9.507142857142857e-06, "loss": 1.396, "step": 5669000 }, { "epoch": 29.26, "learning_rate": 9.503571428571428e-06, "loss": 1.3921, "step": 5669500 }, { "epoch": 29.27, "learning_rate": 9.5e-06, "loss": 1.3889, "step": 5670000 }, { "epoch": 29.27, "learning_rate": 9.496428571428571e-06, "loss": 1.4004, "step": 5670500 }, { "epoch": 29.27, "learning_rate": 9.492857142857144e-06, "loss": 1.4051, "step": 5671000 }, { "epoch": 29.27, "learning_rate": 9.489285714285715e-06, "loss": 1.4003, "step": 5671500 }, { "epoch": 29.28, "learning_rate": 9.485714285714287e-06, "loss": 1.3863, "step": 5672000 }, { "epoch": 29.28, "learning_rate": 9.482142857142858e-06, "loss": 1.4033, "step": 5672500 }, { "epoch": 29.28, "learning_rate": 9.478571428571429e-06, "loss": 1.4014, "step": 5673000 }, { "epoch": 29.28, "learning_rate": 9.475e-06, "loss": 1.388, "step": 5673500 }, { "epoch": 29.29, "learning_rate": 9.47142857142857e-06, "loss": 1.3862, "step": 5674000 }, { "epoch": 29.29, "learning_rate": 9.467857142857143e-06, "loss": 1.3947, "step": 5674500 }, { "epoch": 29.29, "learning_rate": 9.464285714285714e-06, "loss": 1.3968, "step": 5675000 }, { "epoch": 29.29, "learning_rate": 9.460714285714287e-06, "loss": 1.3838, "step": 5675500 }, { "epoch": 29.3, "learning_rate": 9.457142857142858e-06, "loss": 1.3883, "step": 5676000 }, { "epoch": 29.3, "learning_rate": 9.45357142857143e-06, "loss": 1.3962, "step": 5676500 }, { "epoch": 29.3, "learning_rate": 9.450000000000001e-06, "loss": 1.399, "step": 5677000 }, { "epoch": 29.31, "learning_rate": 9.446428571428572e-06, "loss": 1.393, "step": 5677500 }, { "epoch": 29.31, "learning_rate": 9.442857142857144e-06, "loss": 1.3993, "step": 5678000 }, { "epoch": 29.31, "learning_rate": 9.439285714285714e-06, "loss": 1.3854, "step": 5678500 }, { "epoch": 29.31, "learning_rate": 9.435714285714286e-06, "loss": 1.3982, "step": 5679000 }, { "epoch": 29.32, "learning_rate": 9.432142857142857e-06, "loss": 1.4103, "step": 5679500 }, { "epoch": 29.32, "learning_rate": 9.42857142857143e-06, "loss": 1.3966, "step": 5680000 }, { "epoch": 29.32, "learning_rate": 9.425e-06, "loss": 1.4055, "step": 5680500 }, { "epoch": 29.32, "learning_rate": 9.421428571428571e-06, "loss": 1.398, "step": 5681000 }, { "epoch": 29.33, "learning_rate": 9.417857142857144e-06, "loss": 1.3987, "step": 5681500 }, { "epoch": 29.33, "learning_rate": 9.414285714285715e-06, "loss": 1.3949, "step": 5682000 }, { "epoch": 29.33, "learning_rate": 9.410714285714287e-06, "loss": 1.3911, "step": 5682500 }, { "epoch": 29.33, "learning_rate": 9.407142857142858e-06, "loss": 1.3875, "step": 5683000 }, { "epoch": 29.34, "learning_rate": 9.403571428571429e-06, "loss": 1.4019, "step": 5683500 }, { "epoch": 29.34, "learning_rate": 9.4e-06, "loss": 1.3973, "step": 5684000 }, { "epoch": 29.34, "learning_rate": 9.39642857142857e-06, "loss": 1.3954, "step": 5684500 }, { "epoch": 29.35, "learning_rate": 9.392857142857143e-06, "loss": 1.3969, "step": 5685000 }, { "epoch": 29.35, "learning_rate": 9.389285714285714e-06, "loss": 1.3855, "step": 5685500 }, { "epoch": 29.35, "learning_rate": 9.385714285714287e-06, "loss": 1.3911, "step": 5686000 }, { "epoch": 29.35, "learning_rate": 9.382142857142857e-06, "loss": 1.3911, "step": 5686500 }, { "epoch": 29.36, "learning_rate": 9.37857142857143e-06, "loss": 1.3914, "step": 5687000 }, { "epoch": 29.36, "learning_rate": 9.375000000000001e-06, "loss": 1.3923, "step": 5687500 }, { "epoch": 29.36, "learning_rate": 9.371428571428572e-06, "loss": 1.4024, "step": 5688000 }, { "epoch": 29.36, "learning_rate": 9.367857142857144e-06, "loss": 1.3895, "step": 5688500 }, { "epoch": 29.37, "learning_rate": 9.364285714285715e-06, "loss": 1.3878, "step": 5689000 }, { "epoch": 29.37, "learning_rate": 9.360714285714286e-06, "loss": 1.3879, "step": 5689500 }, { "epoch": 29.37, "learning_rate": 9.357142857142857e-06, "loss": 1.3946, "step": 5690000 }, { "epoch": 29.37, "learning_rate": 9.35357142857143e-06, "loss": 1.4085, "step": 5690500 }, { "epoch": 29.38, "learning_rate": 9.35e-06, "loss": 1.3934, "step": 5691000 }, { "epoch": 29.38, "learning_rate": 9.346428571428571e-06, "loss": 1.3981, "step": 5691500 }, { "epoch": 29.38, "learning_rate": 9.342857142857144e-06, "loss": 1.3851, "step": 5692000 }, { "epoch": 29.39, "learning_rate": 9.339285714285715e-06, "loss": 1.3984, "step": 5692500 }, { "epoch": 29.39, "learning_rate": 9.335714285714287e-06, "loss": 1.4033, "step": 5693000 }, { "epoch": 29.39, "learning_rate": 9.332142857142858e-06, "loss": 1.3917, "step": 5693500 }, { "epoch": 29.39, "learning_rate": 9.328571428571429e-06, "loss": 1.3808, "step": 5694000 }, { "epoch": 29.4, "learning_rate": 9.325e-06, "loss": 1.3933, "step": 5694500 }, { "epoch": 29.4, "learning_rate": 9.32142857142857e-06, "loss": 1.4062, "step": 5695000 }, { "epoch": 29.4, "learning_rate": 9.317857142857143e-06, "loss": 1.4407, "step": 5695500 }, { "epoch": 29.4, "learning_rate": 9.314285714285714e-06, "loss": 1.4149, "step": 5696000 }, { "epoch": 29.41, "learning_rate": 9.310714285714286e-06, "loss": 1.4788, "step": 5696500 }, { "epoch": 29.41, "learning_rate": 9.307142857142857e-06, "loss": 1.4771, "step": 5697000 }, { "epoch": 29.41, "learning_rate": 9.30357142857143e-06, "loss": 1.4262, "step": 5697500 }, { "epoch": 29.41, "learning_rate": 9.3e-06, "loss": 1.4316, "step": 5698000 }, { "epoch": 29.42, "learning_rate": 9.296428571428572e-06, "loss": 1.4078, "step": 5698500 }, { "epoch": 29.42, "learning_rate": 9.292857142857144e-06, "loss": 1.4031, "step": 5699000 }, { "epoch": 29.42, "learning_rate": 9.289285714285715e-06, "loss": 1.3972, "step": 5699500 }, { "epoch": 29.43, "learning_rate": 9.285714285714286e-06, "loss": 1.3937, "step": 5700000 }, { "epoch": 29.43, "learning_rate": 9.282142857142857e-06, "loss": 1.3949, "step": 5700500 }, { "epoch": 29.43, "learning_rate": 9.27857142857143e-06, "loss": 1.396, "step": 5701000 }, { "epoch": 29.43, "learning_rate": 9.275e-06, "loss": 1.4146, "step": 5701500 }, { "epoch": 29.44, "learning_rate": 9.271428571428571e-06, "loss": 1.4171, "step": 5702000 }, { "epoch": 29.44, "learning_rate": 9.267857142857144e-06, "loss": 1.3997, "step": 5702500 }, { "epoch": 29.44, "learning_rate": 9.264285714285714e-06, "loss": 1.4114, "step": 5703000 }, { "epoch": 29.44, "learning_rate": 9.260714285714287e-06, "loss": 1.4216, "step": 5703500 }, { "epoch": 29.45, "learning_rate": 9.257142857142858e-06, "loss": 1.4565, "step": 5704000 }, { "epoch": 29.45, "learning_rate": 9.253571428571429e-06, "loss": 1.4184, "step": 5704500 }, { "epoch": 29.45, "learning_rate": 9.25e-06, "loss": 1.4099, "step": 5705000 }, { "epoch": 29.45, "learning_rate": 9.246428571428572e-06, "loss": 1.3971, "step": 5705500 }, { "epoch": 29.46, "learning_rate": 9.242857142857143e-06, "loss": 1.4022, "step": 5706000 }, { "epoch": 29.46, "learning_rate": 9.239285714285714e-06, "loss": 1.4352, "step": 5706500 }, { "epoch": 29.46, "learning_rate": 9.235714285714286e-06, "loss": 1.4044, "step": 5707000 }, { "epoch": 29.47, "learning_rate": 9.232142857142857e-06, "loss": 1.3928, "step": 5707500 }, { "epoch": 29.47, "learning_rate": 9.22857142857143e-06, "loss": 1.3938, "step": 5708000 }, { "epoch": 29.47, "learning_rate": 9.225e-06, "loss": 1.3898, "step": 5708500 }, { "epoch": 29.47, "learning_rate": 9.221428571428573e-06, "loss": 1.3909, "step": 5709000 }, { "epoch": 29.48, "learning_rate": 9.217857142857144e-06, "loss": 1.4018, "step": 5709500 }, { "epoch": 29.48, "learning_rate": 9.214285714285715e-06, "loss": 1.4091, "step": 5710000 }, { "epoch": 29.48, "learning_rate": 9.210714285714286e-06, "loss": 1.4, "step": 5710500 }, { "epoch": 29.48, "learning_rate": 9.207142857142857e-06, "loss": 1.3979, "step": 5711000 }, { "epoch": 29.49, "learning_rate": 9.203571428571429e-06, "loss": 1.421, "step": 5711500 }, { "epoch": 29.49, "learning_rate": 9.2e-06, "loss": 1.3993, "step": 5712000 }, { "epoch": 29.49, "learning_rate": 9.196428571428573e-06, "loss": 1.4039, "step": 5712500 }, { "epoch": 29.49, "learning_rate": 9.192857142857143e-06, "loss": 1.404, "step": 5713000 }, { "epoch": 29.5, "learning_rate": 9.189285714285714e-06, "loss": 1.4061, "step": 5713500 }, { "epoch": 29.5, "learning_rate": 9.185714285714287e-06, "loss": 1.4168, "step": 5714000 }, { "epoch": 29.5, "learning_rate": 9.182142857142858e-06, "loss": 1.4192, "step": 5714500 }, { "epoch": 29.51, "learning_rate": 9.17857142857143e-06, "loss": 1.4283, "step": 5715000 }, { "epoch": 29.51, "learning_rate": 9.175000000000001e-06, "loss": 1.3942, "step": 5715500 }, { "epoch": 29.51, "learning_rate": 9.171428571428572e-06, "loss": 1.3979, "step": 5716000 }, { "epoch": 29.51, "learning_rate": 9.167857142857143e-06, "loss": 1.4116, "step": 5716500 }, { "epoch": 29.52, "learning_rate": 9.164285714285714e-06, "loss": 1.3941, "step": 5717000 }, { "epoch": 29.52, "learning_rate": 9.160714285714286e-06, "loss": 1.4016, "step": 5717500 }, { "epoch": 29.52, "learning_rate": 9.157142857142857e-06, "loss": 1.4083, "step": 5718000 }, { "epoch": 29.52, "learning_rate": 9.15357142857143e-06, "loss": 1.4094, "step": 5718500 }, { "epoch": 29.53, "learning_rate": 9.15e-06, "loss": 1.4004, "step": 5719000 }, { "epoch": 29.53, "learning_rate": 9.146428571428573e-06, "loss": 1.4122, "step": 5719500 }, { "epoch": 29.53, "learning_rate": 9.142857142857144e-06, "loss": 1.3964, "step": 5720000 }, { "epoch": 29.53, "learning_rate": 9.139285714285715e-06, "loss": 1.4044, "step": 5720500 }, { "epoch": 29.54, "learning_rate": 9.135714285714286e-06, "loss": 1.4232, "step": 5721000 }, { "epoch": 29.54, "learning_rate": 9.132142857142856e-06, "loss": 1.4024, "step": 5721500 }, { "epoch": 29.54, "learning_rate": 9.128571428571429e-06, "loss": 1.4089, "step": 5722000 }, { "epoch": 29.54, "learning_rate": 9.125e-06, "loss": 1.3876, "step": 5722500 }, { "epoch": 29.55, "learning_rate": 9.121428571428572e-06, "loss": 1.4184, "step": 5723000 }, { "epoch": 29.55, "learning_rate": 9.117857142857143e-06, "loss": 1.443, "step": 5723500 }, { "epoch": 29.55, "learning_rate": 9.114285714285714e-06, "loss": 1.4206, "step": 5724000 }, { "epoch": 29.56, "learning_rate": 9.110714285714287e-06, "loss": 1.4173, "step": 5724500 }, { "epoch": 29.56, "learning_rate": 9.107142857142858e-06, "loss": 1.4117, "step": 5725000 }, { "epoch": 29.56, "learning_rate": 9.10357142857143e-06, "loss": 1.3987, "step": 5725500 }, { "epoch": 29.56, "learning_rate": 9.100000000000001e-06, "loss": 1.3866, "step": 5726000 }, { "epoch": 29.57, "learning_rate": 9.096428571428572e-06, "loss": 1.4003, "step": 5726500 }, { "epoch": 29.57, "learning_rate": 9.092857142857143e-06, "loss": 1.3985, "step": 5727000 }, { "epoch": 29.57, "learning_rate": 9.089285714285714e-06, "loss": 1.3919, "step": 5727500 }, { "epoch": 29.57, "learning_rate": 9.085714285714286e-06, "loss": 1.4042, "step": 5728000 }, { "epoch": 29.58, "learning_rate": 9.082142857142857e-06, "loss": 1.3977, "step": 5728500 }, { "epoch": 29.58, "learning_rate": 9.07857142857143e-06, "loss": 1.4008, "step": 5729000 }, { "epoch": 29.58, "learning_rate": 9.075e-06, "loss": 1.3939, "step": 5729500 }, { "epoch": 29.58, "learning_rate": 9.071428571428573e-06, "loss": 1.431, "step": 5730000 }, { "epoch": 29.59, "learning_rate": 9.067857142857144e-06, "loss": 1.5196, "step": 5730500 }, { "epoch": 29.59, "learning_rate": 9.064285714285715e-06, "loss": 1.4317, "step": 5731000 }, { "epoch": 29.59, "learning_rate": 9.060714285714285e-06, "loss": 1.3938, "step": 5731500 }, { "epoch": 29.6, "learning_rate": 9.057142857142856e-06, "loss": 1.3984, "step": 5732000 }, { "epoch": 29.6, "learning_rate": 9.053571428571429e-06, "loss": 1.3949, "step": 5732500 }, { "epoch": 29.6, "learning_rate": 9.05e-06, "loss": 1.407, "step": 5733000 }, { "epoch": 29.6, "learning_rate": 9.046428571428572e-06, "loss": 1.4071, "step": 5733500 }, { "epoch": 29.61, "learning_rate": 9.042857142857143e-06, "loss": 1.424, "step": 5734000 }, { "epoch": 29.61, "learning_rate": 9.039285714285716e-06, "loss": 1.4023, "step": 5734500 }, { "epoch": 29.61, "learning_rate": 9.035714285714287e-06, "loss": 1.3929, "step": 5735000 }, { "epoch": 29.61, "learning_rate": 9.032142857142857e-06, "loss": 1.3856, "step": 5735500 }, { "epoch": 29.62, "learning_rate": 9.02857142857143e-06, "loss": 1.3907, "step": 5736000 }, { "epoch": 29.62, "learning_rate": 9.025e-06, "loss": 1.4107, "step": 5736500 }, { "epoch": 29.62, "learning_rate": 9.021428571428572e-06, "loss": 1.4116, "step": 5737000 }, { "epoch": 29.62, "learning_rate": 9.017857142857143e-06, "loss": 1.3928, "step": 5737500 }, { "epoch": 29.63, "learning_rate": 9.014285714285715e-06, "loss": 1.3993, "step": 5738000 }, { "epoch": 29.63, "learning_rate": 9.010714285714286e-06, "loss": 1.3947, "step": 5738500 }, { "epoch": 29.63, "learning_rate": 9.007142857142857e-06, "loss": 1.4115, "step": 5739000 }, { "epoch": 29.64, "learning_rate": 9.00357142857143e-06, "loss": 1.4181, "step": 5739500 }, { "epoch": 29.64, "learning_rate": 9e-06, "loss": 1.4142, "step": 5740000 }, { "epoch": 29.64, "learning_rate": 8.996428571428573e-06, "loss": 1.3961, "step": 5740500 }, { "epoch": 29.64, "learning_rate": 8.992857142857144e-06, "loss": 1.4082, "step": 5741000 }, { "epoch": 29.65, "learning_rate": 8.989285714285716e-06, "loss": 1.4047, "step": 5741500 }, { "epoch": 29.65, "learning_rate": 8.985714285714285e-06, "loss": 1.4532, "step": 5742000 }, { "epoch": 29.65, "learning_rate": 8.982142857142856e-06, "loss": 1.4235, "step": 5742500 }, { "epoch": 29.65, "learning_rate": 8.978571428571429e-06, "loss": 1.4146, "step": 5743000 }, { "epoch": 29.66, "learning_rate": 8.975e-06, "loss": 1.4062, "step": 5743500 }, { "epoch": 29.66, "learning_rate": 8.971428571428572e-06, "loss": 1.4042, "step": 5744000 }, { "epoch": 29.66, "learning_rate": 8.967857142857143e-06, "loss": 1.4032, "step": 5744500 }, { "epoch": 29.66, "learning_rate": 8.964285714285716e-06, "loss": 1.39, "step": 5745000 }, { "epoch": 29.67, "learning_rate": 8.960714285714286e-06, "loss": 1.4678, "step": 5745500 }, { "epoch": 29.67, "learning_rate": 8.957142857142857e-06, "loss": 1.4223, "step": 5746000 }, { "epoch": 29.67, "learning_rate": 8.95357142857143e-06, "loss": 1.4414, "step": 5746500 }, { "epoch": 29.68, "learning_rate": 8.95e-06, "loss": 1.4019, "step": 5747000 }, { "epoch": 29.68, "learning_rate": 8.946428571428572e-06, "loss": 1.3866, "step": 5747500 }, { "epoch": 29.68, "learning_rate": 8.942857142857142e-06, "loss": 1.4028, "step": 5748000 }, { "epoch": 29.68, "learning_rate": 8.939285714285715e-06, "loss": 1.3981, "step": 5748500 }, { "epoch": 29.69, "learning_rate": 8.935714285714286e-06, "loss": 1.39, "step": 5749000 }, { "epoch": 29.69, "learning_rate": 8.932142857142857e-06, "loss": 1.3989, "step": 5749500 }, { "epoch": 29.69, "learning_rate": 8.92857142857143e-06, "loss": 1.4215, "step": 5750000 }, { "epoch": 29.69, "learning_rate": 8.925e-06, "loss": 1.4125, "step": 5750500 }, { "epoch": 29.7, "learning_rate": 8.921428571428573e-06, "loss": 1.4057, "step": 5751000 }, { "epoch": 29.7, "learning_rate": 8.917857142857144e-06, "loss": 1.4029, "step": 5751500 }, { "epoch": 29.7, "learning_rate": 8.914285714285716e-06, "loss": 1.4003, "step": 5752000 }, { "epoch": 29.7, "learning_rate": 8.910714285714287e-06, "loss": 1.4028, "step": 5752500 }, { "epoch": 29.71, "learning_rate": 8.907142857142858e-06, "loss": 1.4003, "step": 5753000 }, { "epoch": 29.71, "learning_rate": 8.903571428571429e-06, "loss": 1.4038, "step": 5753500 }, { "epoch": 29.71, "learning_rate": 8.9e-06, "loss": 1.3888, "step": 5754000 }, { "epoch": 29.72, "learning_rate": 8.896428571428572e-06, "loss": 1.4224, "step": 5754500 }, { "epoch": 29.72, "learning_rate": 8.892857142857143e-06, "loss": 1.4323, "step": 5755000 }, { "epoch": 29.72, "learning_rate": 8.889285714285715e-06, "loss": 1.4332, "step": 5755500 }, { "epoch": 29.72, "learning_rate": 8.885714285714286e-06, "loss": 1.4026, "step": 5756000 }, { "epoch": 29.73, "learning_rate": 8.882142857142857e-06, "loss": 1.4053, "step": 5756500 }, { "epoch": 29.73, "learning_rate": 8.87857142857143e-06, "loss": 1.4126, "step": 5757000 }, { "epoch": 29.73, "learning_rate": 8.875e-06, "loss": 1.3948, "step": 5757500 }, { "epoch": 29.73, "learning_rate": 8.871428571428571e-06, "loss": 1.3991, "step": 5758000 }, { "epoch": 29.74, "learning_rate": 8.867857142857142e-06, "loss": 1.4388, "step": 5758500 }, { "epoch": 29.74, "learning_rate": 8.864285714285715e-06, "loss": 1.4585, "step": 5759000 }, { "epoch": 29.74, "learning_rate": 8.860714285714286e-06, "loss": 1.4709, "step": 5759500 }, { "epoch": 29.74, "learning_rate": 8.857142857142857e-06, "loss": 1.4442, "step": 5760000 }, { "epoch": 29.75, "learning_rate": 8.853571428571429e-06, "loss": 1.4097, "step": 5760500 }, { "epoch": 29.75, "learning_rate": 8.85e-06, "loss": 1.3922, "step": 5761000 }, { "epoch": 29.75, "learning_rate": 8.846428571428573e-06, "loss": 1.3933, "step": 5761500 }, { "epoch": 29.76, "learning_rate": 8.842857142857143e-06, "loss": 1.3951, "step": 5762000 }, { "epoch": 29.76, "learning_rate": 8.839285714285716e-06, "loss": 1.4039, "step": 5762500 }, { "epoch": 29.76, "learning_rate": 8.835714285714287e-06, "loss": 1.4017, "step": 5763000 }, { "epoch": 29.76, "learning_rate": 8.832142857142858e-06, "loss": 1.3926, "step": 5763500 }, { "epoch": 29.77, "learning_rate": 8.828571428571429e-06, "loss": 1.3952, "step": 5764000 }, { "epoch": 29.77, "learning_rate": 8.825e-06, "loss": 1.3898, "step": 5764500 }, { "epoch": 29.77, "learning_rate": 8.821428571428572e-06, "loss": 1.4244, "step": 5765000 }, { "epoch": 29.77, "learning_rate": 8.817857142857143e-06, "loss": 1.4148, "step": 5765500 }, { "epoch": 29.78, "learning_rate": 8.814285714285715e-06, "loss": 1.4142, "step": 5766000 }, { "epoch": 29.78, "learning_rate": 8.810714285714286e-06, "loss": 1.3877, "step": 5766500 }, { "epoch": 29.78, "learning_rate": 8.807142857142859e-06, "loss": 1.3919, "step": 5767000 }, { "epoch": 29.78, "learning_rate": 8.80357142857143e-06, "loss": 1.39, "step": 5767500 }, { "epoch": 29.79, "learning_rate": 8.8e-06, "loss": 1.3907, "step": 5768000 }, { "epoch": 29.79, "learning_rate": 8.796428571428571e-06, "loss": 1.3964, "step": 5768500 }, { "epoch": 29.79, "learning_rate": 8.792857142857142e-06, "loss": 1.397, "step": 5769000 }, { "epoch": 29.8, "learning_rate": 8.789285714285715e-06, "loss": 1.3933, "step": 5769500 }, { "epoch": 29.8, "learning_rate": 8.785714285714286e-06, "loss": 1.3828, "step": 5770000 }, { "epoch": 29.8, "learning_rate": 8.782142857142858e-06, "loss": 1.3901, "step": 5770500 }, { "epoch": 29.8, "learning_rate": 8.778571428571429e-06, "loss": 1.3828, "step": 5771000 }, { "epoch": 29.81, "learning_rate": 8.775e-06, "loss": 1.3904, "step": 5771500 }, { "epoch": 29.81, "learning_rate": 8.771428571428572e-06, "loss": 1.3932, "step": 5772000 }, { "epoch": 29.81, "learning_rate": 8.767857142857143e-06, "loss": 1.3836, "step": 5772500 }, { "epoch": 29.81, "learning_rate": 8.764285714285716e-06, "loss": 1.3863, "step": 5773000 }, { "epoch": 29.82, "learning_rate": 8.760714285714287e-06, "loss": 1.3846, "step": 5773500 }, { "epoch": 29.82, "learning_rate": 8.757142857142858e-06, "loss": 1.3847, "step": 5774000 }, { "epoch": 29.82, "learning_rate": 8.753571428571428e-06, "loss": 1.3847, "step": 5774500 }, { "epoch": 29.82, "learning_rate": 8.75e-06, "loss": 1.3905, "step": 5775000 }, { "epoch": 29.83, "learning_rate": 8.746428571428572e-06, "loss": 1.3802, "step": 5775500 }, { "epoch": 29.83, "learning_rate": 8.742857142857143e-06, "loss": 1.3852, "step": 5776000 }, { "epoch": 29.83, "learning_rate": 8.739285714285715e-06, "loss": 1.3972, "step": 5776500 }, { "epoch": 29.84, "learning_rate": 8.735714285714286e-06, "loss": 1.3892, "step": 5777000 }, { "epoch": 29.84, "learning_rate": 8.732142857142859e-06, "loss": 1.3743, "step": 5777500 }, { "epoch": 29.84, "learning_rate": 8.72857142857143e-06, "loss": 1.3911, "step": 5778000 }, { "epoch": 29.84, "learning_rate": 8.725e-06, "loss": 1.3916, "step": 5778500 }, { "epoch": 29.85, "learning_rate": 8.721428571428573e-06, "loss": 1.3774, "step": 5779000 }, { "epoch": 29.85, "learning_rate": 8.717857142857144e-06, "loss": 1.3851, "step": 5779500 }, { "epoch": 29.85, "learning_rate": 8.714285714285715e-06, "loss": 1.3912, "step": 5780000 }, { "epoch": 29.85, "learning_rate": 8.710714285714285e-06, "loss": 1.3863, "step": 5780500 }, { "epoch": 29.86, "learning_rate": 8.707142857142858e-06, "loss": 1.3811, "step": 5781000 }, { "epoch": 29.86, "learning_rate": 8.703571428571429e-06, "loss": 1.3792, "step": 5781500 }, { "epoch": 29.86, "learning_rate": 8.7e-06, "loss": 1.3773, "step": 5782000 }, { "epoch": 29.86, "learning_rate": 8.696428571428572e-06, "loss": 1.3817, "step": 5782500 }, { "epoch": 29.87, "learning_rate": 8.692857142857143e-06, "loss": 1.3786, "step": 5783000 }, { "epoch": 29.87, "learning_rate": 8.689285714285716e-06, "loss": 1.3828, "step": 5783500 }, { "epoch": 29.87, "learning_rate": 8.685714285714287e-06, "loss": 1.3914, "step": 5784000 }, { "epoch": 29.88, "learning_rate": 8.682142857142857e-06, "loss": 1.3821, "step": 5784500 }, { "epoch": 29.88, "learning_rate": 8.678571428571428e-06, "loss": 1.3877, "step": 5785000 }, { "epoch": 29.88, "learning_rate": 8.674999999999999e-06, "loss": 1.3956, "step": 5785500 }, { "epoch": 29.88, "learning_rate": 8.671428571428572e-06, "loss": 1.41, "step": 5786000 }, { "epoch": 29.89, "learning_rate": 8.667857142857143e-06, "loss": 1.4243, "step": 5786500 }, { "epoch": 29.89, "learning_rate": 8.664285714285715e-06, "loss": 1.4155, "step": 5787000 }, { "epoch": 29.89, "learning_rate": 8.660714285714286e-06, "loss": 1.3979, "step": 5787500 }, { "epoch": 29.89, "learning_rate": 8.657142857142858e-06, "loss": 1.389, "step": 5788000 }, { "epoch": 29.9, "learning_rate": 8.65357142857143e-06, "loss": 1.388, "step": 5788500 }, { "epoch": 29.9, "learning_rate": 8.65e-06, "loss": 1.3852, "step": 5789000 }, { "epoch": 29.9, "learning_rate": 8.646428571428573e-06, "loss": 1.3877, "step": 5789500 }, { "epoch": 29.9, "learning_rate": 8.642857142857144e-06, "loss": 1.3965, "step": 5790000 }, { "epoch": 29.91, "learning_rate": 8.639285714285714e-06, "loss": 1.3901, "step": 5790500 }, { "epoch": 29.91, "learning_rate": 8.635714285714285e-06, "loss": 1.3908, "step": 5791000 }, { "epoch": 29.91, "learning_rate": 8.632142857142858e-06, "loss": 1.3867, "step": 5791500 }, { "epoch": 29.91, "learning_rate": 8.628571428571429e-06, "loss": 1.3907, "step": 5792000 }, { "epoch": 29.92, "learning_rate": 8.625e-06, "loss": 1.3871, "step": 5792500 }, { "epoch": 29.92, "learning_rate": 8.621428571428572e-06, "loss": 1.3849, "step": 5793000 }, { "epoch": 29.92, "learning_rate": 8.617857142857143e-06, "loss": 1.3904, "step": 5793500 }, { "epoch": 29.93, "learning_rate": 8.614285714285716e-06, "loss": 1.391, "step": 5794000 }, { "epoch": 29.93, "learning_rate": 8.610714285714286e-06, "loss": 1.3884, "step": 5794500 }, { "epoch": 29.93, "learning_rate": 8.607142857142857e-06, "loss": 1.3891, "step": 5795000 }, { "epoch": 29.93, "learning_rate": 8.603571428571428e-06, "loss": 1.3907, "step": 5795500 }, { "epoch": 29.94, "learning_rate": 8.599999999999999e-06, "loss": 1.3718, "step": 5796000 }, { "epoch": 29.94, "learning_rate": 8.596428571428572e-06, "loss": 1.3882, "step": 5796500 }, { "epoch": 29.94, "learning_rate": 8.592857142857142e-06, "loss": 1.3963, "step": 5797000 }, { "epoch": 29.94, "learning_rate": 8.589285714285715e-06, "loss": 1.3881, "step": 5797500 }, { "epoch": 29.95, "learning_rate": 8.585714285714286e-06, "loss": 1.3841, "step": 5798000 }, { "epoch": 29.95, "learning_rate": 8.582142857142858e-06, "loss": 1.393, "step": 5798500 }, { "epoch": 29.95, "learning_rate": 8.57857142857143e-06, "loss": 1.3778, "step": 5799000 }, { "epoch": 29.95, "learning_rate": 8.575000000000002e-06, "loss": 1.3973, "step": 5799500 }, { "epoch": 29.96, "learning_rate": 8.571428571428573e-06, "loss": 1.3938, "step": 5800000 }, { "epoch": 29.96, "learning_rate": 8.567857142857143e-06, "loss": 1.3778, "step": 5800500 }, { "epoch": 29.96, "learning_rate": 8.564285714285714e-06, "loss": 1.3839, "step": 5801000 }, { "epoch": 29.97, "learning_rate": 8.560714285714285e-06, "loss": 1.3874, "step": 5801500 }, { "epoch": 29.97, "learning_rate": 8.557142857142858e-06, "loss": 1.3977, "step": 5802000 }, { "epoch": 29.97, "learning_rate": 8.553571428571429e-06, "loss": 1.3806, "step": 5802500 }, { "epoch": 29.97, "learning_rate": 8.550000000000001e-06, "loss": 1.4065, "step": 5803000 }, { "epoch": 29.98, "learning_rate": 8.546428571428572e-06, "loss": 1.3994, "step": 5803500 }, { "epoch": 29.98, "learning_rate": 8.542857142857143e-06, "loss": 1.395, "step": 5804000 }, { "epoch": 29.98, "learning_rate": 8.539285714285715e-06, "loss": 1.396, "step": 5804500 }, { "epoch": 29.98, "learning_rate": 8.535714285714286e-06, "loss": 1.408, "step": 5805000 }, { "epoch": 29.99, "learning_rate": 8.532142857142859e-06, "loss": 1.3995, "step": 5805500 }, { "epoch": 29.99, "learning_rate": 8.528571428571428e-06, "loss": 1.394, "step": 5806000 }, { "epoch": 29.99, "learning_rate": 8.525e-06, "loss": 1.4022, "step": 5806500 }, { "epoch": 29.99, "learning_rate": 8.521428571428571e-06, "loss": 1.4143, "step": 5807000 }, { "epoch": 30.0, "learning_rate": 8.517857142857142e-06, "loss": 1.4033, "step": 5807500 }, { "epoch": 30.0, "learning_rate": 8.514285714285715e-06, "loss": 1.4031, "step": 5808000 }, { "epoch": 30.0, "learning_rate": 8.510714285714286e-06, "loss": 1.4327, "step": 5808500 }, { "epoch": 30.01, "learning_rate": 8.507142857142858e-06, "loss": 1.4323, "step": 5809000 }, { "epoch": 30.01, "learning_rate": 8.503571428571429e-06, "loss": 1.4444, "step": 5809500 }, { "epoch": 30.01, "learning_rate": 8.500000000000002e-06, "loss": 1.4188, "step": 5810000 }, { "epoch": 30.01, "learning_rate": 8.496428571428572e-06, "loss": 1.4451, "step": 5810500 }, { "epoch": 30.02, "learning_rate": 8.492857142857143e-06, "loss": 1.4303, "step": 5811000 }, { "epoch": 30.02, "learning_rate": 8.489285714285714e-06, "loss": 1.4496, "step": 5811500 }, { "epoch": 30.02, "learning_rate": 8.485714285714285e-06, "loss": 1.4298, "step": 5812000 }, { "epoch": 30.02, "learning_rate": 8.482142857142858e-06, "loss": 1.3959, "step": 5812500 }, { "epoch": 30.03, "learning_rate": 8.478571428571428e-06, "loss": 1.4097, "step": 5813000 }, { "epoch": 30.03, "learning_rate": 8.475000000000001e-06, "loss": 1.4234, "step": 5813500 }, { "epoch": 30.03, "learning_rate": 8.471428571428572e-06, "loss": 1.4077, "step": 5814000 }, { "epoch": 30.03, "learning_rate": 8.467857142857143e-06, "loss": 1.4095, "step": 5814500 }, { "epoch": 30.04, "learning_rate": 8.464285714285715e-06, "loss": 1.442, "step": 5815000 }, { "epoch": 30.04, "learning_rate": 8.460714285714286e-06, "loss": 1.4639, "step": 5815500 }, { "epoch": 30.04, "learning_rate": 8.457142857142859e-06, "loss": 1.4457, "step": 5816000 }, { "epoch": 30.05, "learning_rate": 8.45357142857143e-06, "loss": 1.4074, "step": 5816500 }, { "epoch": 30.05, "learning_rate": 8.45e-06, "loss": 1.3896, "step": 5817000 }, { "epoch": 30.05, "learning_rate": 8.446428571428571e-06, "loss": 1.4156, "step": 5817500 }, { "epoch": 30.05, "learning_rate": 8.442857142857142e-06, "loss": 1.4108, "step": 5818000 }, { "epoch": 30.06, "learning_rate": 8.439285714285715e-06, "loss": 1.3926, "step": 5818500 }, { "epoch": 30.06, "learning_rate": 8.435714285714286e-06, "loss": 1.4089, "step": 5819000 }, { "epoch": 30.06, "learning_rate": 8.432142857142858e-06, "loss": 1.395, "step": 5819500 }, { "epoch": 30.06, "learning_rate": 8.428571428571429e-06, "loss": 1.3816, "step": 5820000 }, { "epoch": 30.07, "learning_rate": 8.425000000000001e-06, "loss": 1.3869, "step": 5820500 }, { "epoch": 30.07, "learning_rate": 8.421428571428572e-06, "loss": 1.3922, "step": 5821000 }, { "epoch": 30.07, "learning_rate": 8.417857142857143e-06, "loss": 1.4242, "step": 5821500 }, { "epoch": 30.07, "learning_rate": 8.414285714285714e-06, "loss": 1.4173, "step": 5822000 }, { "epoch": 30.08, "learning_rate": 8.410714285714285e-06, "loss": 1.3803, "step": 5822500 }, { "epoch": 30.08, "learning_rate": 8.407142857142857e-06, "loss": 1.3888, "step": 5823000 }, { "epoch": 30.08, "learning_rate": 8.403571428571428e-06, "loss": 1.385, "step": 5823500 }, { "epoch": 30.09, "learning_rate": 8.400000000000001e-06, "loss": 1.3813, "step": 5824000 }, { "epoch": 30.09, "learning_rate": 8.396428571428572e-06, "loss": 1.3993, "step": 5824500 }, { "epoch": 30.09, "learning_rate": 8.392857142857143e-06, "loss": 1.3794, "step": 5825000 }, { "epoch": 30.09, "learning_rate": 8.389285714285715e-06, "loss": 1.3781, "step": 5825500 }, { "epoch": 30.1, "learning_rate": 8.385714285714286e-06, "loss": 1.3842, "step": 5826000 }, { "epoch": 30.1, "learning_rate": 8.382142857142859e-06, "loss": 1.3828, "step": 5826500 }, { "epoch": 30.1, "learning_rate": 8.37857142857143e-06, "loss": 1.384, "step": 5827000 }, { "epoch": 30.1, "learning_rate": 8.375e-06, "loss": 1.3787, "step": 5827500 }, { "epoch": 30.11, "learning_rate": 8.371428571428571e-06, "loss": 1.3868, "step": 5828000 }, { "epoch": 30.11, "learning_rate": 8.367857142857144e-06, "loss": 1.3802, "step": 5828500 }, { "epoch": 30.11, "learning_rate": 8.364285714285715e-06, "loss": 1.38, "step": 5829000 }, { "epoch": 30.11, "learning_rate": 8.360714285714285e-06, "loss": 1.3897, "step": 5829500 }, { "epoch": 30.12, "learning_rate": 8.357142857142858e-06, "loss": 1.3864, "step": 5830000 }, { "epoch": 30.12, "learning_rate": 8.353571428571429e-06, "loss": 1.409, "step": 5830500 }, { "epoch": 30.12, "learning_rate": 8.350000000000001e-06, "loss": 1.4041, "step": 5831000 }, { "epoch": 30.13, "learning_rate": 8.346428571428572e-06, "loss": 1.3995, "step": 5831500 }, { "epoch": 30.13, "learning_rate": 8.342857142857143e-06, "loss": 1.4138, "step": 5832000 }, { "epoch": 30.13, "learning_rate": 8.339285714285714e-06, "loss": 1.3921, "step": 5832500 }, { "epoch": 30.13, "learning_rate": 8.335714285714285e-06, "loss": 1.4068, "step": 5833000 }, { "epoch": 30.14, "learning_rate": 8.332142857142857e-06, "loss": 1.3922, "step": 5833500 }, { "epoch": 30.14, "learning_rate": 8.328571428571428e-06, "loss": 1.3821, "step": 5834000 }, { "epoch": 30.14, "learning_rate": 8.325e-06, "loss": 1.3894, "step": 5834500 }, { "epoch": 30.14, "learning_rate": 8.321428571428572e-06, "loss": 1.385, "step": 5835000 }, { "epoch": 30.15, "learning_rate": 8.317857142857144e-06, "loss": 1.3864, "step": 5835500 }, { "epoch": 30.15, "learning_rate": 8.314285714285715e-06, "loss": 1.3789, "step": 5836000 }, { "epoch": 30.15, "learning_rate": 8.310714285714286e-06, "loss": 1.3845, "step": 5836500 }, { "epoch": 30.15, "learning_rate": 8.307142857142858e-06, "loss": 1.3884, "step": 5837000 }, { "epoch": 30.16, "learning_rate": 8.30357142857143e-06, "loss": 1.3801, "step": 5837500 }, { "epoch": 30.16, "learning_rate": 8.3e-06, "loss": 1.3804, "step": 5838000 }, { "epoch": 30.16, "learning_rate": 8.296428571428571e-06, "loss": 1.3941, "step": 5838500 }, { "epoch": 30.17, "learning_rate": 8.292857142857144e-06, "loss": 1.3854, "step": 5839000 }, { "epoch": 30.17, "learning_rate": 8.289285714285714e-06, "loss": 1.3717, "step": 5839500 }, { "epoch": 30.17, "learning_rate": 8.285714285714285e-06, "loss": 1.3805, "step": 5840000 }, { "epoch": 30.17, "learning_rate": 8.282142857142858e-06, "loss": 1.3754, "step": 5840500 }, { "epoch": 30.18, "learning_rate": 8.278571428571429e-06, "loss": 1.3861, "step": 5841000 }, { "epoch": 30.18, "learning_rate": 8.275000000000001e-06, "loss": 1.3856, "step": 5841500 }, { "epoch": 30.18, "learning_rate": 8.271428571428572e-06, "loss": 1.3948, "step": 5842000 }, { "epoch": 30.18, "learning_rate": 8.267857142857145e-06, "loss": 1.386, "step": 5842500 }, { "epoch": 30.19, "learning_rate": 8.264285714285716e-06, "loss": 1.3846, "step": 5843000 }, { "epoch": 30.19, "learning_rate": 8.260714285714285e-06, "loss": 1.3815, "step": 5843500 }, { "epoch": 30.19, "learning_rate": 8.257142857142857e-06, "loss": 1.3842, "step": 5844000 }, { "epoch": 30.19, "learning_rate": 8.253571428571428e-06, "loss": 1.3737, "step": 5844500 }, { "epoch": 30.2, "learning_rate": 8.25e-06, "loss": 1.3872, "step": 5845000 }, { "epoch": 30.2, "learning_rate": 8.246428571428571e-06, "loss": 1.3758, "step": 5845500 }, { "epoch": 30.2, "learning_rate": 8.242857142857144e-06, "loss": 1.3887, "step": 5846000 }, { "epoch": 30.21, "learning_rate": 8.239285714285715e-06, "loss": 1.4, "step": 5846500 }, { "epoch": 30.21, "learning_rate": 8.235714285714286e-06, "loss": 1.3865, "step": 5847000 }, { "epoch": 30.21, "learning_rate": 8.232142857142858e-06, "loss": 1.3942, "step": 5847500 }, { "epoch": 30.21, "learning_rate": 8.22857142857143e-06, "loss": 1.3897, "step": 5848000 }, { "epoch": 30.22, "learning_rate": 8.225e-06, "loss": 1.3896, "step": 5848500 }, { "epoch": 30.22, "learning_rate": 8.221428571428571e-06, "loss": 1.3744, "step": 5849000 }, { "epoch": 30.22, "learning_rate": 8.217857142857143e-06, "loss": 1.3855, "step": 5849500 }, { "epoch": 30.22, "learning_rate": 8.214285714285714e-06, "loss": 1.381, "step": 5850000 }, { "epoch": 30.23, "learning_rate": 8.210714285714285e-06, "loss": 1.3981, "step": 5850500 }, { "epoch": 30.23, "learning_rate": 8.207142857142858e-06, "loss": 1.3987, "step": 5851000 }, { "epoch": 30.23, "learning_rate": 8.203571428571429e-06, "loss": 1.385, "step": 5851500 }, { "epoch": 30.23, "learning_rate": 8.200000000000001e-06, "loss": 1.3989, "step": 5852000 }, { "epoch": 30.24, "learning_rate": 8.196428571428572e-06, "loss": 1.3833, "step": 5852500 }, { "epoch": 30.24, "learning_rate": 8.192857142857145e-06, "loss": 1.3793, "step": 5853000 }, { "epoch": 30.24, "learning_rate": 8.189285714285715e-06, "loss": 1.386, "step": 5853500 }, { "epoch": 30.25, "learning_rate": 8.185714285714286e-06, "loss": 1.3712, "step": 5854000 }, { "epoch": 30.25, "learning_rate": 8.182142857142857e-06, "loss": 1.3871, "step": 5854500 }, { "epoch": 30.25, "learning_rate": 8.178571428571428e-06, "loss": 1.3772, "step": 5855000 }, { "epoch": 30.25, "learning_rate": 8.175e-06, "loss": 1.3779, "step": 5855500 }, { "epoch": 30.26, "learning_rate": 8.171428571428571e-06, "loss": 1.3886, "step": 5856000 }, { "epoch": 30.26, "learning_rate": 8.167857142857144e-06, "loss": 1.4023, "step": 5856500 }, { "epoch": 30.26, "learning_rate": 8.164285714285715e-06, "loss": 1.4115, "step": 5857000 }, { "epoch": 30.26, "learning_rate": 8.160714285714286e-06, "loss": 1.3848, "step": 5857500 }, { "epoch": 30.27, "learning_rate": 8.157142857142858e-06, "loss": 1.3817, "step": 5858000 }, { "epoch": 30.27, "learning_rate": 8.153571428571429e-06, "loss": 1.3825, "step": 5858500 }, { "epoch": 30.27, "learning_rate": 8.15e-06, "loss": 1.3716, "step": 5859000 }, { "epoch": 30.27, "learning_rate": 8.14642857142857e-06, "loss": 1.3894, "step": 5859500 }, { "epoch": 30.28, "learning_rate": 8.142857142857143e-06, "loss": 1.38, "step": 5860000 }, { "epoch": 30.28, "learning_rate": 8.139285714285714e-06, "loss": 1.3787, "step": 5860500 }, { "epoch": 30.28, "learning_rate": 8.135714285714287e-06, "loss": 1.3839, "step": 5861000 }, { "epoch": 30.29, "learning_rate": 8.132142857142858e-06, "loss": 1.3755, "step": 5861500 }, { "epoch": 30.29, "learning_rate": 8.128571428571428e-06, "loss": 1.3748, "step": 5862000 }, { "epoch": 30.29, "learning_rate": 8.125000000000001e-06, "loss": 1.3737, "step": 5862500 }, { "epoch": 30.29, "learning_rate": 8.121428571428572e-06, "loss": 1.3857, "step": 5863000 }, { "epoch": 30.3, "learning_rate": 8.117857142857144e-06, "loss": 1.3721, "step": 5863500 }, { "epoch": 30.3, "learning_rate": 8.114285714285715e-06, "loss": 1.3788, "step": 5864000 }, { "epoch": 30.3, "learning_rate": 8.110714285714286e-06, "loss": 1.3823, "step": 5864500 }, { "epoch": 30.3, "learning_rate": 8.107142857142857e-06, "loss": 1.388, "step": 5865000 }, { "epoch": 30.31, "learning_rate": 8.103571428571428e-06, "loss": 1.3809, "step": 5865500 }, { "epoch": 30.31, "learning_rate": 8.1e-06, "loss": 1.3809, "step": 5866000 }, { "epoch": 30.31, "learning_rate": 8.096428571428571e-06, "loss": 1.3964, "step": 5866500 }, { "epoch": 30.31, "learning_rate": 8.092857142857144e-06, "loss": 1.37, "step": 5867000 }, { "epoch": 30.32, "learning_rate": 8.089285714285715e-06, "loss": 1.3875, "step": 5867500 }, { "epoch": 30.32, "learning_rate": 8.085714285714287e-06, "loss": 1.3805, "step": 5868000 }, { "epoch": 30.32, "learning_rate": 8.082142857142858e-06, "loss": 1.3943, "step": 5868500 }, { "epoch": 30.32, "learning_rate": 8.078571428571429e-06, "loss": 1.3887, "step": 5869000 }, { "epoch": 30.33, "learning_rate": 8.075000000000001e-06, "loss": 1.4031, "step": 5869500 }, { "epoch": 30.33, "learning_rate": 8.07142857142857e-06, "loss": 1.3873, "step": 5870000 }, { "epoch": 30.33, "learning_rate": 8.067857142857143e-06, "loss": 1.3892, "step": 5870500 }, { "epoch": 30.34, "learning_rate": 8.064285714285714e-06, "loss": 1.3815, "step": 5871000 }, { "epoch": 30.34, "learning_rate": 8.060714285714287e-06, "loss": 1.3881, "step": 5871500 }, { "epoch": 30.34, "learning_rate": 8.057142857142857e-06, "loss": 1.3991, "step": 5872000 }, { "epoch": 30.34, "learning_rate": 8.053571428571428e-06, "loss": 1.3907, "step": 5872500 }, { "epoch": 30.35, "learning_rate": 8.050000000000001e-06, "loss": 1.3827, "step": 5873000 }, { "epoch": 30.35, "learning_rate": 8.046428571428572e-06, "loss": 1.3889, "step": 5873500 }, { "epoch": 30.35, "learning_rate": 8.042857142857144e-06, "loss": 1.3906, "step": 5874000 }, { "epoch": 30.35, "learning_rate": 8.039285714285715e-06, "loss": 1.3778, "step": 5874500 }, { "epoch": 30.36, "learning_rate": 8.035714285714286e-06, "loss": 1.3834, "step": 5875000 }, { "epoch": 30.36, "learning_rate": 8.032142857142857e-06, "loss": 1.3755, "step": 5875500 }, { "epoch": 30.36, "learning_rate": 8.028571428571428e-06, "loss": 1.3667, "step": 5876000 }, { "epoch": 30.36, "learning_rate": 8.025e-06, "loss": 1.3765, "step": 5876500 }, { "epoch": 30.37, "learning_rate": 8.021428571428571e-06, "loss": 1.3821, "step": 5877000 }, { "epoch": 30.37, "learning_rate": 8.017857142857144e-06, "loss": 1.3855, "step": 5877500 }, { "epoch": 30.37, "learning_rate": 8.014285714285715e-06, "loss": 1.3736, "step": 5878000 }, { "epoch": 30.38, "learning_rate": 8.010714285714287e-06, "loss": 1.3716, "step": 5878500 }, { "epoch": 30.38, "learning_rate": 8.007142857142858e-06, "loss": 1.3779, "step": 5879000 }, { "epoch": 30.38, "learning_rate": 8.003571428571429e-06, "loss": 1.3787, "step": 5879500 }, { "epoch": 30.38, "learning_rate": 8.000000000000001e-06, "loss": 1.3734, "step": 5880000 }, { "epoch": 30.39, "learning_rate": 7.996428571428572e-06, "loss": 1.3783, "step": 5880500 }, { "epoch": 30.39, "learning_rate": 7.992857142857143e-06, "loss": 1.3789, "step": 5881000 }, { "epoch": 30.39, "learning_rate": 7.989285714285714e-06, "loss": 1.3782, "step": 5881500 }, { "epoch": 30.39, "learning_rate": 7.985714285714286e-06, "loss": 1.3627, "step": 5882000 }, { "epoch": 30.4, "learning_rate": 7.982142857142857e-06, "loss": 1.3719, "step": 5882500 }, { "epoch": 30.4, "learning_rate": 7.978571428571428e-06, "loss": 1.37, "step": 5883000 }, { "epoch": 30.4, "learning_rate": 7.975e-06, "loss": 1.3735, "step": 5883500 }, { "epoch": 30.4, "learning_rate": 7.971428571428572e-06, "loss": 1.3829, "step": 5884000 }, { "epoch": 30.41, "learning_rate": 7.967857142857144e-06, "loss": 1.3719, "step": 5884500 }, { "epoch": 30.41, "learning_rate": 7.964285714285715e-06, "loss": 1.3731, "step": 5885000 }, { "epoch": 30.41, "learning_rate": 7.960714285714286e-06, "loss": 1.3701, "step": 5885500 }, { "epoch": 30.42, "learning_rate": 7.957142857142857e-06, "loss": 1.3773, "step": 5886000 }, { "epoch": 30.42, "learning_rate": 7.953571428571428e-06, "loss": 1.3819, "step": 5886500 }, { "epoch": 30.42, "learning_rate": 7.95e-06, "loss": 1.3743, "step": 5887000 }, { "epoch": 30.42, "learning_rate": 7.946428571428571e-06, "loss": 1.3954, "step": 5887500 }, { "epoch": 30.43, "learning_rate": 7.942857142857144e-06, "loss": 1.3895, "step": 5888000 }, { "epoch": 30.43, "learning_rate": 7.939285714285714e-06, "loss": 1.3771, "step": 5888500 }, { "epoch": 30.43, "learning_rate": 7.935714285714287e-06, "loss": 1.3877, "step": 5889000 }, { "epoch": 30.43, "learning_rate": 7.932142857142858e-06, "loss": 1.393, "step": 5889500 }, { "epoch": 30.44, "learning_rate": 7.928571428571429e-06, "loss": 1.3974, "step": 5890000 }, { "epoch": 30.44, "learning_rate": 7.925000000000001e-06, "loss": 1.3887, "step": 5890500 }, { "epoch": 30.44, "learning_rate": 7.921428571428572e-06, "loss": 1.3847, "step": 5891000 }, { "epoch": 30.44, "learning_rate": 7.917857142857143e-06, "loss": 1.3854, "step": 5891500 }, { "epoch": 30.45, "learning_rate": 7.914285714285714e-06, "loss": 1.3823, "step": 5892000 }, { "epoch": 30.45, "learning_rate": 7.910714285714286e-06, "loss": 1.3786, "step": 5892500 }, { "epoch": 30.45, "learning_rate": 7.907142857142857e-06, "loss": 1.3887, "step": 5893000 }, { "epoch": 30.46, "learning_rate": 7.90357142857143e-06, "loss": 1.39, "step": 5893500 }, { "epoch": 30.46, "learning_rate": 7.9e-06, "loss": 1.3822, "step": 5894000 }, { "epoch": 30.46, "learning_rate": 7.896428571428571e-06, "loss": 1.3887, "step": 5894500 }, { "epoch": 30.46, "learning_rate": 7.892857142857144e-06, "loss": 1.3674, "step": 5895000 }, { "epoch": 30.47, "learning_rate": 7.889285714285715e-06, "loss": 1.3803, "step": 5895500 }, { "epoch": 30.47, "learning_rate": 7.885714285714286e-06, "loss": 1.3888, "step": 5896000 }, { "epoch": 30.47, "learning_rate": 7.882142857142857e-06, "loss": 1.3882, "step": 5896500 }, { "epoch": 30.47, "learning_rate": 7.878571428571429e-06, "loss": 1.3719, "step": 5897000 }, { "epoch": 30.48, "learning_rate": 7.875e-06, "loss": 1.3778, "step": 5897500 }, { "epoch": 30.48, "learning_rate": 7.87142857142857e-06, "loss": 1.377, "step": 5898000 }, { "epoch": 30.48, "learning_rate": 7.867857142857143e-06, "loss": 1.3993, "step": 5898500 }, { "epoch": 30.48, "learning_rate": 7.864285714285714e-06, "loss": 1.3727, "step": 5899000 }, { "epoch": 30.49, "learning_rate": 7.860714285714287e-06, "loss": 1.3702, "step": 5899500 }, { "epoch": 30.49, "learning_rate": 7.857142857142858e-06, "loss": 1.3789, "step": 5900000 }, { "epoch": 30.49, "learning_rate": 7.85357142857143e-06, "loss": 1.3844, "step": 5900500 }, { "epoch": 30.5, "learning_rate": 7.850000000000001e-06, "loss": 1.3771, "step": 5901000 }, { "epoch": 30.5, "learning_rate": 7.846428571428572e-06, "loss": 1.3795, "step": 5901500 }, { "epoch": 30.5, "learning_rate": 7.842857142857143e-06, "loss": 1.3826, "step": 5902000 }, { "epoch": 30.5, "learning_rate": 7.839285714285714e-06, "loss": 1.3802, "step": 5902500 }, { "epoch": 30.51, "learning_rate": 7.835714285714286e-06, "loss": 1.3759, "step": 5903000 }, { "epoch": 30.51, "learning_rate": 7.832142857142857e-06, "loss": 1.3679, "step": 5903500 }, { "epoch": 30.51, "learning_rate": 7.82857142857143e-06, "loss": 1.3789, "step": 5904000 }, { "epoch": 30.51, "learning_rate": 7.825e-06, "loss": 1.383, "step": 5904500 }, { "epoch": 30.52, "learning_rate": 7.821428571428571e-06, "loss": 1.3848, "step": 5905000 }, { "epoch": 30.52, "learning_rate": 7.817857142857144e-06, "loss": 1.3831, "step": 5905500 }, { "epoch": 30.52, "learning_rate": 7.814285714285715e-06, "loss": 1.3796, "step": 5906000 }, { "epoch": 30.52, "learning_rate": 7.810714285714287e-06, "loss": 1.3857, "step": 5906500 }, { "epoch": 30.53, "learning_rate": 7.807142857142858e-06, "loss": 1.3736, "step": 5907000 }, { "epoch": 30.53, "learning_rate": 7.803571428571429e-06, "loss": 1.3878, "step": 5907500 }, { "epoch": 30.53, "learning_rate": 7.8e-06, "loss": 1.4024, "step": 5908000 }, { "epoch": 30.54, "learning_rate": 7.79642857142857e-06, "loss": 1.3991, "step": 5908500 }, { "epoch": 30.54, "learning_rate": 7.792857142857143e-06, "loss": 1.4045, "step": 5909000 }, { "epoch": 30.54, "learning_rate": 7.789285714285714e-06, "loss": 1.3934, "step": 5909500 }, { "epoch": 30.54, "learning_rate": 7.785714285714287e-06, "loss": 1.3889, "step": 5910000 }, { "epoch": 30.55, "learning_rate": 7.782142857142858e-06, "loss": 1.3758, "step": 5910500 }, { "epoch": 30.55, "learning_rate": 7.77857142857143e-06, "loss": 1.4001, "step": 5911000 }, { "epoch": 30.55, "learning_rate": 7.775000000000001e-06, "loss": 1.4028, "step": 5911500 }, { "epoch": 30.55, "learning_rate": 7.771428571428572e-06, "loss": 1.4078, "step": 5912000 }, { "epoch": 30.56, "learning_rate": 7.767857142857143e-06, "loss": 1.3947, "step": 5912500 }, { "epoch": 30.56, "learning_rate": 7.764285714285714e-06, "loss": 1.3898, "step": 5913000 }, { "epoch": 30.56, "learning_rate": 7.760714285714286e-06, "loss": 1.4078, "step": 5913500 }, { "epoch": 30.56, "learning_rate": 7.757142857142857e-06, "loss": 1.4289, "step": 5914000 }, { "epoch": 30.57, "learning_rate": 7.75357142857143e-06, "loss": 1.3909, "step": 5914500 }, { "epoch": 30.57, "learning_rate": 7.75e-06, "loss": 1.3917, "step": 5915000 }, { "epoch": 30.57, "learning_rate": 7.746428571428571e-06, "loss": 1.3842, "step": 5915500 }, { "epoch": 30.58, "learning_rate": 7.742857142857144e-06, "loss": 1.4032, "step": 5916000 }, { "epoch": 30.58, "learning_rate": 7.739285714285715e-06, "loss": 1.3848, "step": 5916500 }, { "epoch": 30.58, "learning_rate": 7.735714285714287e-06, "loss": 1.3863, "step": 5917000 }, { "epoch": 30.58, "learning_rate": 7.732142857142858e-06, "loss": 1.393, "step": 5917500 }, { "epoch": 30.59, "learning_rate": 7.728571428571429e-06, "loss": 1.3896, "step": 5918000 }, { "epoch": 30.59, "learning_rate": 7.725e-06, "loss": 1.3997, "step": 5918500 }, { "epoch": 30.59, "learning_rate": 7.72142857142857e-06, "loss": 1.3838, "step": 5919000 }, { "epoch": 30.59, "learning_rate": 7.717857142857143e-06, "loss": 1.3837, "step": 5919500 }, { "epoch": 30.6, "learning_rate": 7.714285714285714e-06, "loss": 1.388, "step": 5920000 }, { "epoch": 30.6, "learning_rate": 7.710714285714287e-06, "loss": 1.3876, "step": 5920500 }, { "epoch": 30.6, "learning_rate": 7.707142857142857e-06, "loss": 1.3883, "step": 5921000 }, { "epoch": 30.6, "learning_rate": 7.70357142857143e-06, "loss": 1.4037, "step": 5921500 }, { "epoch": 30.61, "learning_rate": 7.7e-06, "loss": 1.3804, "step": 5922000 }, { "epoch": 30.61, "learning_rate": 7.696428571428572e-06, "loss": 1.377, "step": 5922500 }, { "epoch": 30.61, "learning_rate": 7.692857142857143e-06, "loss": 1.3788, "step": 5923000 }, { "epoch": 30.62, "learning_rate": 7.689285714285713e-06, "loss": 1.3871, "step": 5923500 }, { "epoch": 30.62, "learning_rate": 7.685714285714286e-06, "loss": 1.3834, "step": 5924000 }, { "epoch": 30.62, "learning_rate": 7.682142857142857e-06, "loss": 1.383, "step": 5924500 }, { "epoch": 30.62, "learning_rate": 7.67857142857143e-06, "loss": 1.3918, "step": 5925000 }, { "epoch": 30.63, "learning_rate": 7.675e-06, "loss": 1.418, "step": 5925500 }, { "epoch": 30.63, "learning_rate": 7.671428571428573e-06, "loss": 1.4031, "step": 5926000 }, { "epoch": 30.63, "learning_rate": 7.667857142857144e-06, "loss": 1.3946, "step": 5926500 }, { "epoch": 30.63, "learning_rate": 7.664285714285714e-06, "loss": 1.3888, "step": 5927000 }, { "epoch": 30.64, "learning_rate": 7.660714285714287e-06, "loss": 1.3769, "step": 5927500 }, { "epoch": 30.64, "learning_rate": 7.657142857142858e-06, "loss": 1.3897, "step": 5928000 }, { "epoch": 30.64, "learning_rate": 7.653571428571429e-06, "loss": 1.3775, "step": 5928500 }, { "epoch": 30.64, "learning_rate": 7.65e-06, "loss": 1.3792, "step": 5929000 }, { "epoch": 30.65, "learning_rate": 7.646428571428572e-06, "loss": 1.3859, "step": 5929500 }, { "epoch": 30.65, "learning_rate": 7.642857142857143e-06, "loss": 1.3753, "step": 5930000 }, { "epoch": 30.65, "learning_rate": 7.639285714285714e-06, "loss": 1.3745, "step": 5930500 }, { "epoch": 30.66, "learning_rate": 7.635714285714286e-06, "loss": 1.3689, "step": 5931000 }, { "epoch": 30.66, "learning_rate": 7.632142857142857e-06, "loss": 1.3727, "step": 5931500 }, { "epoch": 30.66, "learning_rate": 7.628571428571429e-06, "loss": 1.3765, "step": 5932000 }, { "epoch": 30.66, "learning_rate": 7.625e-06, "loss": 1.366, "step": 5932500 }, { "epoch": 30.67, "learning_rate": 7.621428571428572e-06, "loss": 1.3755, "step": 5933000 }, { "epoch": 30.67, "learning_rate": 7.617857142857143e-06, "loss": 1.3828, "step": 5933500 }, { "epoch": 30.67, "learning_rate": 7.614285714285714e-06, "loss": 1.372, "step": 5934000 }, { "epoch": 30.67, "learning_rate": 7.610714285714287e-06, "loss": 1.3837, "step": 5934500 }, { "epoch": 30.68, "learning_rate": 7.6071428571428575e-06, "loss": 1.3623, "step": 5935000 }, { "epoch": 30.68, "learning_rate": 7.603571428571429e-06, "loss": 1.3669, "step": 5935500 }, { "epoch": 30.68, "learning_rate": 7.6e-06, "loss": 1.3761, "step": 5936000 }, { "epoch": 30.68, "learning_rate": 7.596428571428573e-06, "loss": 1.3748, "step": 5936500 }, { "epoch": 30.69, "learning_rate": 7.5928571428571435e-06, "loss": 1.3614, "step": 5937000 }, { "epoch": 30.69, "learning_rate": 7.589285714285714e-06, "loss": 1.374, "step": 5937500 }, { "epoch": 30.69, "learning_rate": 7.585714285714286e-06, "loss": 1.3737, "step": 5938000 }, { "epoch": 30.7, "learning_rate": 7.582142857142857e-06, "loss": 1.3863, "step": 5938500 }, { "epoch": 30.7, "learning_rate": 7.5785714285714295e-06, "loss": 1.3803, "step": 5939000 }, { "epoch": 30.7, "learning_rate": 7.575e-06, "loss": 1.3831, "step": 5939500 }, { "epoch": 30.7, "learning_rate": 7.571428571428572e-06, "loss": 1.3865, "step": 5940000 }, { "epoch": 30.71, "learning_rate": 7.567857142857143e-06, "loss": 1.3883, "step": 5940500 }, { "epoch": 30.71, "learning_rate": 7.564285714285714e-06, "loss": 1.3758, "step": 5941000 }, { "epoch": 30.71, "learning_rate": 7.560714285714286e-06, "loss": 1.3795, "step": 5941500 }, { "epoch": 30.71, "learning_rate": 7.557142857142857e-06, "loss": 1.374, "step": 5942000 }, { "epoch": 30.72, "learning_rate": 7.55357142857143e-06, "loss": 1.3753, "step": 5942500 }, { "epoch": 30.72, "learning_rate": 7.55e-06, "loss": 1.3764, "step": 5943000 }, { "epoch": 30.72, "learning_rate": 7.546428571428572e-06, "loss": 1.3778, "step": 5943500 }, { "epoch": 30.72, "learning_rate": 7.542857142857143e-06, "loss": 1.3709, "step": 5944000 }, { "epoch": 30.73, "learning_rate": 7.539285714285714e-06, "loss": 1.3687, "step": 5944500 }, { "epoch": 30.73, "learning_rate": 7.5357142857142865e-06, "loss": 1.374, "step": 5945000 }, { "epoch": 30.73, "learning_rate": 7.532142857142857e-06, "loss": 1.3747, "step": 5945500 }, { "epoch": 30.73, "learning_rate": 7.528571428571429e-06, "loss": 1.3861, "step": 5946000 }, { "epoch": 30.74, "learning_rate": 7.525e-06, "loss": 1.3929, "step": 5946500 }, { "epoch": 30.74, "learning_rate": 7.5214285714285725e-06, "loss": 1.3962, "step": 5947000 }, { "epoch": 30.74, "learning_rate": 7.517857142857143e-06, "loss": 1.3942, "step": 5947500 }, { "epoch": 30.75, "learning_rate": 7.514285714285714e-06, "loss": 1.3845, "step": 5948000 }, { "epoch": 30.75, "learning_rate": 7.510714285714286e-06, "loss": 1.3827, "step": 5948500 }, { "epoch": 30.75, "learning_rate": 7.507142857142857e-06, "loss": 1.3731, "step": 5949000 }, { "epoch": 30.75, "learning_rate": 7.503571428571429e-06, "loss": 1.3802, "step": 5949500 }, { "epoch": 30.76, "learning_rate": 7.5e-06, "loss": 1.3692, "step": 5950000 }, { "epoch": 30.76, "learning_rate": 7.496428571428572e-06, "loss": 1.3785, "step": 5950500 }, { "epoch": 30.76, "learning_rate": 7.492857142857143e-06, "loss": 1.3785, "step": 5951000 }, { "epoch": 30.76, "learning_rate": 7.489285714285714e-06, "loss": 1.3827, "step": 5951500 }, { "epoch": 30.77, "learning_rate": 7.485714285714286e-06, "loss": 1.38, "step": 5952000 }, { "epoch": 30.77, "learning_rate": 7.482142857142857e-06, "loss": 1.3727, "step": 5952500 }, { "epoch": 30.77, "learning_rate": 7.4785714285714296e-06, "loss": 1.3765, "step": 5953000 }, { "epoch": 30.77, "learning_rate": 7.4750000000000004e-06, "loss": 1.37, "step": 5953500 }, { "epoch": 30.78, "learning_rate": 7.471428571428572e-06, "loss": 1.3717, "step": 5954000 }, { "epoch": 30.78, "learning_rate": 7.467857142857143e-06, "loss": 1.3745, "step": 5954500 }, { "epoch": 30.78, "learning_rate": 7.4642857142857155e-06, "loss": 1.3791, "step": 5955000 }, { "epoch": 30.79, "learning_rate": 7.460714285714286e-06, "loss": 1.3857, "step": 5955500 }, { "epoch": 30.79, "learning_rate": 7.457142857142857e-06, "loss": 1.3911, "step": 5956000 }, { "epoch": 30.79, "learning_rate": 7.453571428571429e-06, "loss": 1.4057, "step": 5956500 }, { "epoch": 30.79, "learning_rate": 7.45e-06, "loss": 1.4283, "step": 5957000 }, { "epoch": 30.8, "learning_rate": 7.446428571428572e-06, "loss": 1.4271, "step": 5957500 }, { "epoch": 30.8, "learning_rate": 7.442857142857143e-06, "loss": 1.4062, "step": 5958000 }, { "epoch": 30.8, "learning_rate": 7.439285714285715e-06, "loss": 1.4096, "step": 5958500 }, { "epoch": 30.8, "learning_rate": 7.435714285714286e-06, "loss": 1.4042, "step": 5959000 }, { "epoch": 30.81, "learning_rate": 7.432142857142857e-06, "loss": 1.3753, "step": 5959500 }, { "epoch": 30.81, "learning_rate": 7.428571428571429e-06, "loss": 1.3677, "step": 5960000 }, { "epoch": 30.81, "learning_rate": 7.425e-06, "loss": 1.3746, "step": 5960500 }, { "epoch": 30.81, "learning_rate": 7.421428571428573e-06, "loss": 1.3751, "step": 5961000 }, { "epoch": 30.82, "learning_rate": 7.4178571428571435e-06, "loss": 1.3848, "step": 5961500 }, { "epoch": 30.82, "learning_rate": 7.414285714285715e-06, "loss": 1.3883, "step": 5962000 }, { "epoch": 30.82, "learning_rate": 7.410714285714286e-06, "loss": 1.3935, "step": 5962500 }, { "epoch": 30.83, "learning_rate": 7.407142857142857e-06, "loss": 1.3772, "step": 5963000 }, { "epoch": 30.83, "learning_rate": 7.4035714285714294e-06, "loss": 1.3898, "step": 5963500 }, { "epoch": 30.83, "learning_rate": 7.4e-06, "loss": 1.3808, "step": 5964000 }, { "epoch": 30.83, "learning_rate": 7.396428571428572e-06, "loss": 1.384, "step": 5964500 }, { "epoch": 30.84, "learning_rate": 7.392857142857143e-06, "loss": 1.3919, "step": 5965000 }, { "epoch": 30.84, "learning_rate": 7.389285714285715e-06, "loss": 1.3913, "step": 5965500 }, { "epoch": 30.84, "learning_rate": 7.385714285714286e-06, "loss": 1.3833, "step": 5966000 }, { "epoch": 30.84, "learning_rate": 7.382142857142857e-06, "loss": 1.3821, "step": 5966500 }, { "epoch": 30.85, "learning_rate": 7.378571428571429e-06, "loss": 1.3809, "step": 5967000 }, { "epoch": 30.85, "learning_rate": 7.375e-06, "loss": 1.3885, "step": 5967500 }, { "epoch": 30.85, "learning_rate": 7.371428571428572e-06, "loss": 1.3854, "step": 5968000 }, { "epoch": 30.85, "learning_rate": 7.367857142857143e-06, "loss": 1.3758, "step": 5968500 }, { "epoch": 30.86, "learning_rate": 7.364285714285716e-06, "loss": 1.388, "step": 5969000 }, { "epoch": 30.86, "learning_rate": 7.360714285714286e-06, "loss": 1.3872, "step": 5969500 }, { "epoch": 30.86, "learning_rate": 7.3571428571428565e-06, "loss": 1.3892, "step": 5970000 }, { "epoch": 30.87, "learning_rate": 7.353571428571429e-06, "loss": 1.3796, "step": 5970500 }, { "epoch": 30.87, "learning_rate": 7.35e-06, "loss": 1.3813, "step": 5971000 }, { "epoch": 30.87, "learning_rate": 7.3464285714285725e-06, "loss": 1.3718, "step": 5971500 }, { "epoch": 30.87, "learning_rate": 7.342857142857143e-06, "loss": 1.3747, "step": 5972000 }, { "epoch": 30.88, "learning_rate": 7.339285714285715e-06, "loss": 1.3727, "step": 5972500 }, { "epoch": 30.88, "learning_rate": 7.335714285714286e-06, "loss": 1.3745, "step": 5973000 }, { "epoch": 30.88, "learning_rate": 7.332142857142857e-06, "loss": 1.3927, "step": 5973500 }, { "epoch": 30.88, "learning_rate": 7.328571428571429e-06, "loss": 1.3917, "step": 5974000 }, { "epoch": 30.89, "learning_rate": 7.325e-06, "loss": 1.3851, "step": 5974500 }, { "epoch": 30.89, "learning_rate": 7.321428571428572e-06, "loss": 1.3819, "step": 5975000 }, { "epoch": 30.89, "learning_rate": 7.317857142857143e-06, "loss": 1.3757, "step": 5975500 }, { "epoch": 30.89, "learning_rate": 7.314285714285715e-06, "loss": 1.3896, "step": 5976000 }, { "epoch": 30.9, "learning_rate": 7.310714285714286e-06, "loss": 1.3746, "step": 5976500 }, { "epoch": 30.9, "learning_rate": 7.307142857142857e-06, "loss": 1.3746, "step": 5977000 }, { "epoch": 30.9, "learning_rate": 7.303571428571429e-06, "loss": 1.3726, "step": 5977500 }, { "epoch": 30.91, "learning_rate": 7.2999999999999996e-06, "loss": 1.3653, "step": 5978000 }, { "epoch": 30.91, "learning_rate": 7.296428571428572e-06, "loss": 1.3747, "step": 5978500 }, { "epoch": 30.91, "learning_rate": 7.292857142857143e-06, "loss": 1.3745, "step": 5979000 }, { "epoch": 30.91, "learning_rate": 7.2892857142857155e-06, "loss": 1.3725, "step": 5979500 }, { "epoch": 30.92, "learning_rate": 7.285714285714286e-06, "loss": 1.3763, "step": 5980000 }, { "epoch": 30.92, "learning_rate": 7.282142857142857e-06, "loss": 1.3871, "step": 5980500 }, { "epoch": 30.92, "learning_rate": 7.278571428571429e-06, "loss": 1.3784, "step": 5981000 }, { "epoch": 30.92, "learning_rate": 7.275e-06, "loss": 1.3709, "step": 5981500 }, { "epoch": 30.93, "learning_rate": 7.271428571428572e-06, "loss": 1.3811, "step": 5982000 }, { "epoch": 30.93, "learning_rate": 7.267857142857143e-06, "loss": 1.3847, "step": 5982500 }, { "epoch": 30.93, "learning_rate": 7.264285714285715e-06, "loss": 1.3789, "step": 5983000 }, { "epoch": 30.93, "learning_rate": 7.260714285714286e-06, "loss": 1.3848, "step": 5983500 }, { "epoch": 30.94, "learning_rate": 7.257142857142857e-06, "loss": 1.3948, "step": 5984000 }, { "epoch": 30.94, "learning_rate": 7.253571428571429e-06, "loss": 1.3677, "step": 5984500 }, { "epoch": 30.94, "learning_rate": 7.25e-06, "loss": 1.364, "step": 5985000 }, { "epoch": 30.95, "learning_rate": 7.246428571428572e-06, "loss": 1.3804, "step": 5985500 }, { "epoch": 30.95, "learning_rate": 7.242857142857143e-06, "loss": 1.3812, "step": 5986000 }, { "epoch": 30.95, "learning_rate": 7.239285714285715e-06, "loss": 1.3695, "step": 5986500 }, { "epoch": 30.95, "learning_rate": 7.235714285714286e-06, "loss": 1.3806, "step": 5987000 }, { "epoch": 30.96, "learning_rate": 7.2321428571428586e-06, "loss": 1.378, "step": 5987500 }, { "epoch": 30.96, "learning_rate": 7.228571428571429e-06, "loss": 1.3655, "step": 5988000 }, { "epoch": 30.96, "learning_rate": 7.2249999999999994e-06, "loss": 1.3727, "step": 5988500 }, { "epoch": 30.96, "learning_rate": 7.221428571428572e-06, "loss": 1.3705, "step": 5989000 }, { "epoch": 30.97, "learning_rate": 7.217857142857143e-06, "loss": 1.3669, "step": 5989500 }, { "epoch": 30.97, "learning_rate": 7.214285714285715e-06, "loss": 1.3773, "step": 5990000 }, { "epoch": 30.97, "learning_rate": 7.210714285714286e-06, "loss": 1.3758, "step": 5990500 }, { "epoch": 30.97, "learning_rate": 7.207142857142858e-06, "loss": 1.3854, "step": 5991000 }, { "epoch": 30.98, "learning_rate": 7.203571428571429e-06, "loss": 1.3702, "step": 5991500 }, { "epoch": 30.98, "learning_rate": 7.2e-06, "loss": 1.3667, "step": 5992000 }, { "epoch": 30.98, "learning_rate": 7.196428571428572e-06, "loss": 1.3731, "step": 5992500 }, { "epoch": 30.99, "learning_rate": 7.192857142857143e-06, "loss": 1.3765, "step": 5993000 }, { "epoch": 30.99, "learning_rate": 7.189285714285715e-06, "loss": 1.3794, "step": 5993500 }, { "epoch": 30.99, "learning_rate": 7.185714285714286e-06, "loss": 1.3665, "step": 5994000 }, { "epoch": 30.99, "learning_rate": 7.182142857142858e-06, "loss": 1.3788, "step": 5994500 }, { "epoch": 31.0, "learning_rate": 7.178571428571429e-06, "loss": 1.3755, "step": 5995000 }, { "epoch": 31.0, "learning_rate": 7.175e-06, "loss": 1.3813, "step": 5995500 }, { "epoch": 31.0, "learning_rate": 7.171428571428572e-06, "loss": 1.3676, "step": 5996000 }, { "epoch": 31.0, "learning_rate": 7.1678571428571425e-06, "loss": 1.3648, "step": 5996500 }, { "epoch": 31.01, "learning_rate": 7.164285714285715e-06, "loss": 1.3703, "step": 5997000 }, { "epoch": 31.01, "learning_rate": 7.160714285714286e-06, "loss": 1.3715, "step": 5997500 }, { "epoch": 31.01, "learning_rate": 7.1571428571428584e-06, "loss": 1.3678, "step": 5998000 }, { "epoch": 31.01, "learning_rate": 7.153571428571429e-06, "loss": 1.381, "step": 5998500 }, { "epoch": 31.02, "learning_rate": 7.15e-06, "loss": 1.368, "step": 5999000 }, { "epoch": 31.02, "learning_rate": 7.146428571428572e-06, "loss": 1.3639, "step": 5999500 }, { "epoch": 31.02, "learning_rate": 7.142857142857143e-06, "loss": 1.3701, "step": 6000000 }, { "epoch": 31.03, "learning_rate": 7.139285714285715e-06, "loss": 1.3676, "step": 6000500 }, { "epoch": 31.03, "learning_rate": 7.135714285714286e-06, "loss": 1.3703, "step": 6001000 }, { "epoch": 31.03, "learning_rate": 7.132142857142858e-06, "loss": 1.3706, "step": 6001500 }, { "epoch": 31.03, "learning_rate": 7.128571428571429e-06, "loss": 1.3798, "step": 6002000 }, { "epoch": 31.04, "learning_rate": 7.1249999999999995e-06, "loss": 1.3793, "step": 6002500 }, { "epoch": 31.04, "learning_rate": 7.121428571428572e-06, "loss": 1.3783, "step": 6003000 }, { "epoch": 31.04, "learning_rate": 7.117857142857143e-06, "loss": 1.3693, "step": 6003500 }, { "epoch": 31.04, "learning_rate": 7.114285714285715e-06, "loss": 1.3607, "step": 6004000 }, { "epoch": 31.05, "learning_rate": 7.1107142857142855e-06, "loss": 1.3782, "step": 6004500 }, { "epoch": 31.05, "learning_rate": 7.107142857142858e-06, "loss": 1.3741, "step": 6005000 }, { "epoch": 31.05, "learning_rate": 7.103571428571429e-06, "loss": 1.3629, "step": 6005500 }, { "epoch": 31.05, "learning_rate": 7.1e-06, "loss": 1.3712, "step": 6006000 }, { "epoch": 31.06, "learning_rate": 7.096428571428572e-06, "loss": 1.3714, "step": 6006500 }, { "epoch": 31.06, "learning_rate": 7.092857142857142e-06, "loss": 1.3705, "step": 6007000 }, { "epoch": 31.06, "learning_rate": 7.089285714285715e-06, "loss": 1.3819, "step": 6007500 }, { "epoch": 31.07, "learning_rate": 7.085714285714286e-06, "loss": 1.3919, "step": 6008000 }, { "epoch": 31.07, "learning_rate": 7.082142857142858e-06, "loss": 1.404, "step": 6008500 }, { "epoch": 31.07, "learning_rate": 7.078571428571429e-06, "loss": 1.4016, "step": 6009000 }, { "epoch": 31.07, "learning_rate": 7.075e-06, "loss": 1.4082, "step": 6009500 }, { "epoch": 31.08, "learning_rate": 7.071428571428572e-06, "loss": 1.3978, "step": 6010000 }, { "epoch": 31.08, "learning_rate": 7.067857142857143e-06, "loss": 1.4051, "step": 6010500 }, { "epoch": 31.08, "learning_rate": 7.064285714285715e-06, "loss": 1.4336, "step": 6011000 }, { "epoch": 31.08, "learning_rate": 7.060714285714286e-06, "loss": 1.393, "step": 6011500 }, { "epoch": 31.09, "learning_rate": 7.057142857142858e-06, "loss": 1.382, "step": 6012000 }, { "epoch": 31.09, "learning_rate": 7.0535714285714286e-06, "loss": 1.379, "step": 6012500 }, { "epoch": 31.09, "learning_rate": 7.049999999999999e-06, "loss": 1.3783, "step": 6013000 }, { "epoch": 31.09, "learning_rate": 7.046428571428572e-06, "loss": 1.3852, "step": 6013500 }, { "epoch": 31.1, "learning_rate": 7.042857142857143e-06, "loss": 1.3955, "step": 6014000 }, { "epoch": 31.1, "learning_rate": 7.0392857142857145e-06, "loss": 1.3753, "step": 6014500 }, { "epoch": 31.1, "learning_rate": 7.035714285714285e-06, "loss": 1.3802, "step": 6015000 }, { "epoch": 31.11, "learning_rate": 7.032142857142858e-06, "loss": 1.3815, "step": 6015500 }, { "epoch": 31.11, "learning_rate": 7.028571428571429e-06, "loss": 1.3748, "step": 6016000 }, { "epoch": 31.11, "learning_rate": 7.025000000000001e-06, "loss": 1.3772, "step": 6016500 }, { "epoch": 31.11, "learning_rate": 7.021428571428572e-06, "loss": 1.378, "step": 6017000 }, { "epoch": 31.12, "learning_rate": 7.017857142857143e-06, "loss": 1.378, "step": 6017500 }, { "epoch": 31.12, "learning_rate": 7.014285714285715e-06, "loss": 1.379, "step": 6018000 }, { "epoch": 31.12, "learning_rate": 7.010714285714286e-06, "loss": 1.3691, "step": 6018500 }, { "epoch": 31.12, "learning_rate": 7.007142857142858e-06, "loss": 1.3628, "step": 6019000 }, { "epoch": 31.13, "learning_rate": 7.003571428571429e-06, "loss": 1.3719, "step": 6019500 }, { "epoch": 31.13, "learning_rate": 7.000000000000001e-06, "loss": 1.3717, "step": 6020000 }, { "epoch": 31.13, "learning_rate": 6.996428571428572e-06, "loss": 1.3851, "step": 6020500 }, { "epoch": 31.13, "learning_rate": 6.9928571428571425e-06, "loss": 1.3926, "step": 6021000 }, { "epoch": 31.14, "learning_rate": 6.989285714285715e-06, "loss": 1.3842, "step": 6021500 }, { "epoch": 31.14, "learning_rate": 6.985714285714286e-06, "loss": 1.376, "step": 6022000 }, { "epoch": 31.14, "learning_rate": 6.9821428571428576e-06, "loss": 1.3772, "step": 6022500 }, { "epoch": 31.14, "learning_rate": 6.9785714285714284e-06, "loss": 1.3812, "step": 6023000 }, { "epoch": 31.15, "learning_rate": 6.975000000000001e-06, "loss": 1.3967, "step": 6023500 }, { "epoch": 31.15, "learning_rate": 6.971428571428572e-06, "loss": 1.382, "step": 6024000 }, { "epoch": 31.15, "learning_rate": 6.967857142857143e-06, "loss": 1.3778, "step": 6024500 }, { "epoch": 31.16, "learning_rate": 6.964285714285715e-06, "loss": 1.3805, "step": 6025000 }, { "epoch": 31.16, "learning_rate": 6.960714285714286e-06, "loss": 1.3743, "step": 6025500 }, { "epoch": 31.16, "learning_rate": 6.957142857142858e-06, "loss": 1.3678, "step": 6026000 }, { "epoch": 31.16, "learning_rate": 6.953571428571429e-06, "loss": 1.3718, "step": 6026500 }, { "epoch": 31.17, "learning_rate": 6.950000000000001e-06, "loss": 1.3679, "step": 6027000 }, { "epoch": 31.17, "learning_rate": 6.946428571428572e-06, "loss": 1.3644, "step": 6027500 }, { "epoch": 31.17, "learning_rate": 6.942857142857143e-06, "loss": 1.3768, "step": 6028000 }, { "epoch": 31.17, "learning_rate": 6.939285714285715e-06, "loss": 1.3806, "step": 6028500 }, { "epoch": 31.18, "learning_rate": 6.9357142857142855e-06, "loss": 1.3664, "step": 6029000 }, { "epoch": 31.18, "learning_rate": 6.932142857142858e-06, "loss": 1.3808, "step": 6029500 }, { "epoch": 31.18, "learning_rate": 6.928571428571429e-06, "loss": 1.3809, "step": 6030000 }, { "epoch": 31.18, "learning_rate": 6.925000000000001e-06, "loss": 1.3639, "step": 6030500 }, { "epoch": 31.19, "learning_rate": 6.9214285714285715e-06, "loss": 1.3666, "step": 6031000 }, { "epoch": 31.19, "learning_rate": 6.917857142857142e-06, "loss": 1.3673, "step": 6031500 }, { "epoch": 31.19, "learning_rate": 6.914285714285715e-06, "loss": 1.3726, "step": 6032000 }, { "epoch": 31.2, "learning_rate": 6.910714285714286e-06, "loss": 1.3681, "step": 6032500 }, { "epoch": 31.2, "learning_rate": 6.9071428571428574e-06, "loss": 1.3672, "step": 6033000 }, { "epoch": 31.2, "learning_rate": 6.903571428571428e-06, "loss": 1.3699, "step": 6033500 }, { "epoch": 31.2, "learning_rate": 6.900000000000001e-06, "loss": 1.3776, "step": 6034000 }, { "epoch": 31.21, "learning_rate": 6.896428571428572e-06, "loss": 1.3677, "step": 6034500 }, { "epoch": 31.21, "learning_rate": 6.8928571428571426e-06, "loss": 1.3657, "step": 6035000 }, { "epoch": 31.21, "learning_rate": 6.889285714285715e-06, "loss": 1.3854, "step": 6035500 }, { "epoch": 31.21, "learning_rate": 6.885714285714286e-06, "loss": 1.3909, "step": 6036000 }, { "epoch": 31.22, "learning_rate": 6.882142857142858e-06, "loss": 1.3777, "step": 6036500 }, { "epoch": 31.22, "learning_rate": 6.8785714285714285e-06, "loss": 1.3695, "step": 6037000 }, { "epoch": 31.22, "learning_rate": 6.875000000000001e-06, "loss": 1.3772, "step": 6037500 }, { "epoch": 31.22, "learning_rate": 6.871428571428572e-06, "loss": 1.3905, "step": 6038000 }, { "epoch": 31.23, "learning_rate": 6.867857142857143e-06, "loss": 1.3707, "step": 6038500 }, { "epoch": 31.23, "learning_rate": 6.8642857142857145e-06, "loss": 1.3689, "step": 6039000 }, { "epoch": 31.23, "learning_rate": 6.860714285714285e-06, "loss": 1.3755, "step": 6039500 }, { "epoch": 31.24, "learning_rate": 6.857142857142858e-06, "loss": 1.36, "step": 6040000 }, { "epoch": 31.24, "learning_rate": 6.853571428571429e-06, "loss": 1.3571, "step": 6040500 }, { "epoch": 31.24, "learning_rate": 6.8500000000000005e-06, "loss": 1.3635, "step": 6041000 }, { "epoch": 31.24, "learning_rate": 6.846428571428571e-06, "loss": 1.3842, "step": 6041500 }, { "epoch": 31.25, "learning_rate": 6.842857142857142e-06, "loss": 1.3839, "step": 6042000 }, { "epoch": 31.25, "learning_rate": 6.839285714285715e-06, "loss": 1.3911, "step": 6042500 }, { "epoch": 31.25, "learning_rate": 6.835714285714286e-06, "loss": 1.3958, "step": 6043000 }, { "epoch": 31.25, "learning_rate": 6.832142857142858e-06, "loss": 1.379, "step": 6043500 }, { "epoch": 31.26, "learning_rate": 6.828571428571429e-06, "loss": 1.3685, "step": 6044000 }, { "epoch": 31.26, "learning_rate": 6.825000000000001e-06, "loss": 1.3687, "step": 6044500 }, { "epoch": 31.26, "learning_rate": 6.821428571428572e-06, "loss": 1.3732, "step": 6045000 }, { "epoch": 31.26, "learning_rate": 6.8178571428571424e-06, "loss": 1.3673, "step": 6045500 }, { "epoch": 31.27, "learning_rate": 6.814285714285715e-06, "loss": 1.3682, "step": 6046000 }, { "epoch": 31.27, "learning_rate": 6.810714285714286e-06, "loss": 1.3652, "step": 6046500 }, { "epoch": 31.27, "learning_rate": 6.8071428571428576e-06, "loss": 1.3754, "step": 6047000 }, { "epoch": 31.28, "learning_rate": 6.803571428571428e-06, "loss": 1.3753, "step": 6047500 }, { "epoch": 31.28, "learning_rate": 6.800000000000001e-06, "loss": 1.3877, "step": 6048000 }, { "epoch": 31.28, "learning_rate": 6.796428571428572e-06, "loss": 1.3772, "step": 6048500 }, { "epoch": 31.28, "learning_rate": 6.7928571428571435e-06, "loss": 1.3726, "step": 6049000 }, { "epoch": 31.29, "learning_rate": 6.789285714285714e-06, "loss": 1.366, "step": 6049500 }, { "epoch": 31.29, "learning_rate": 6.785714285714285e-06, "loss": 1.3536, "step": 6050000 }, { "epoch": 31.29, "learning_rate": 6.782142857142858e-06, "loss": 1.3758, "step": 6050500 }, { "epoch": 31.29, "learning_rate": 6.778571428571429e-06, "loss": 1.3688, "step": 6051000 }, { "epoch": 31.3, "learning_rate": 6.775000000000001e-06, "loss": 1.3731, "step": 6051500 }, { "epoch": 31.3, "learning_rate": 6.771428571428571e-06, "loss": 1.3655, "step": 6052000 }, { "epoch": 31.3, "learning_rate": 6.767857142857144e-06, "loss": 1.363, "step": 6052500 }, { "epoch": 31.3, "learning_rate": 6.764285714285715e-06, "loss": 1.3643, "step": 6053000 }, { "epoch": 31.31, "learning_rate": 6.7607142857142855e-06, "loss": 1.3597, "step": 6053500 }, { "epoch": 31.31, "learning_rate": 6.757142857142858e-06, "loss": 1.3687, "step": 6054000 }, { "epoch": 31.31, "learning_rate": 6.753571428571429e-06, "loss": 1.3755, "step": 6054500 }, { "epoch": 31.32, "learning_rate": 6.750000000000001e-06, "loss": 1.3752, "step": 6055000 }, { "epoch": 31.32, "learning_rate": 6.7464285714285715e-06, "loss": 1.3661, "step": 6055500 }, { "epoch": 31.32, "learning_rate": 6.742857142857144e-06, "loss": 1.3639, "step": 6056000 }, { "epoch": 31.32, "learning_rate": 6.739285714285715e-06, "loss": 1.37, "step": 6056500 }, { "epoch": 31.33, "learning_rate": 6.735714285714286e-06, "loss": 1.3718, "step": 6057000 }, { "epoch": 31.33, "learning_rate": 6.7321428571428574e-06, "loss": 1.3662, "step": 6057500 }, { "epoch": 31.33, "learning_rate": 6.728571428571428e-06, "loss": 1.3654, "step": 6058000 }, { "epoch": 31.33, "learning_rate": 6.725000000000001e-06, "loss": 1.3645, "step": 6058500 }, { "epoch": 31.34, "learning_rate": 6.721428571428572e-06, "loss": 1.3671, "step": 6059000 }, { "epoch": 31.34, "learning_rate": 6.717857142857143e-06, "loss": 1.3656, "step": 6059500 }, { "epoch": 31.34, "learning_rate": 6.714285714285714e-06, "loss": 1.3698, "step": 6060000 }, { "epoch": 31.34, "learning_rate": 6.710714285714285e-06, "loss": 1.3683, "step": 6060500 }, { "epoch": 31.35, "learning_rate": 6.707142857142858e-06, "loss": 1.3725, "step": 6061000 }, { "epoch": 31.35, "learning_rate": 6.7035714285714285e-06, "loss": 1.3658, "step": 6061500 }, { "epoch": 31.35, "learning_rate": 6.700000000000001e-06, "loss": 1.3711, "step": 6062000 }, { "epoch": 31.36, "learning_rate": 6.696428571428572e-06, "loss": 1.3609, "step": 6062500 }, { "epoch": 31.36, "learning_rate": 6.692857142857144e-06, "loss": 1.3653, "step": 6063000 }, { "epoch": 31.36, "learning_rate": 6.6892857142857145e-06, "loss": 1.3629, "step": 6063500 }, { "epoch": 31.36, "learning_rate": 6.685714285714285e-06, "loss": 1.3653, "step": 6064000 }, { "epoch": 31.37, "learning_rate": 6.682142857142858e-06, "loss": 1.3635, "step": 6064500 }, { "epoch": 31.37, "learning_rate": 6.678571428571429e-06, "loss": 1.3721, "step": 6065000 }, { "epoch": 31.37, "learning_rate": 6.6750000000000005e-06, "loss": 1.3768, "step": 6065500 }, { "epoch": 31.37, "learning_rate": 6.671428571428571e-06, "loss": 1.3652, "step": 6066000 }, { "epoch": 31.38, "learning_rate": 6.667857142857144e-06, "loss": 1.3609, "step": 6066500 }, { "epoch": 31.38, "learning_rate": 6.664285714285715e-06, "loss": 1.3646, "step": 6067000 }, { "epoch": 31.38, "learning_rate": 6.660714285714286e-06, "loss": 1.3818, "step": 6067500 }, { "epoch": 31.38, "learning_rate": 6.657142857142857e-06, "loss": 1.3701, "step": 6068000 }, { "epoch": 31.39, "learning_rate": 6.653571428571428e-06, "loss": 1.3748, "step": 6068500 }, { "epoch": 31.39, "learning_rate": 6.650000000000001e-06, "loss": 1.3745, "step": 6069000 }, { "epoch": 31.39, "learning_rate": 6.6464285714285716e-06, "loss": 1.3674, "step": 6069500 }, { "epoch": 31.4, "learning_rate": 6.642857142857144e-06, "loss": 1.3658, "step": 6070000 }, { "epoch": 31.4, "learning_rate": 6.639285714285715e-06, "loss": 1.3655, "step": 6070500 }, { "epoch": 31.4, "learning_rate": 6.635714285714285e-06, "loss": 1.3719, "step": 6071000 }, { "epoch": 31.4, "learning_rate": 6.6321428571428575e-06, "loss": 1.3624, "step": 6071500 }, { "epoch": 31.41, "learning_rate": 6.628571428571428e-06, "loss": 1.3636, "step": 6072000 }, { "epoch": 31.41, "learning_rate": 6.625000000000001e-06, "loss": 1.3624, "step": 6072500 }, { "epoch": 31.41, "learning_rate": 6.621428571428572e-06, "loss": 1.3685, "step": 6073000 }, { "epoch": 31.41, "learning_rate": 6.6178571428571435e-06, "loss": 1.3629, "step": 6073500 }, { "epoch": 31.42, "learning_rate": 6.614285714285714e-06, "loss": 1.3716, "step": 6074000 }, { "epoch": 31.42, "learning_rate": 6.610714285714285e-06, "loss": 1.3554, "step": 6074500 }, { "epoch": 31.42, "learning_rate": 6.607142857142858e-06, "loss": 1.3679, "step": 6075000 }, { "epoch": 31.42, "learning_rate": 6.603571428571429e-06, "loss": 1.3702, "step": 6075500 }, { "epoch": 31.43, "learning_rate": 6.6e-06, "loss": 1.3677, "step": 6076000 }, { "epoch": 31.43, "learning_rate": 6.596428571428571e-06, "loss": 1.3682, "step": 6076500 }, { "epoch": 31.43, "learning_rate": 6.592857142857144e-06, "loss": 1.3678, "step": 6077000 }, { "epoch": 31.44, "learning_rate": 6.589285714285715e-06, "loss": 1.3704, "step": 6077500 }, { "epoch": 31.44, "learning_rate": 6.5857142857142855e-06, "loss": 1.3692, "step": 6078000 }, { "epoch": 31.44, "learning_rate": 6.582142857142857e-06, "loss": 1.3763, "step": 6078500 }, { "epoch": 31.44, "learning_rate": 6.578571428571428e-06, "loss": 1.3665, "step": 6079000 }, { "epoch": 31.45, "learning_rate": 6.5750000000000006e-06, "loss": 1.3833, "step": 6079500 }, { "epoch": 31.45, "learning_rate": 6.5714285714285714e-06, "loss": 1.371, "step": 6080000 }, { "epoch": 31.45, "learning_rate": 6.567857142857144e-06, "loss": 1.3671, "step": 6080500 }, { "epoch": 31.45, "learning_rate": 6.564285714285715e-06, "loss": 1.3761, "step": 6081000 }, { "epoch": 31.46, "learning_rate": 6.5607142857142865e-06, "loss": 1.3721, "step": 6081500 }, { "epoch": 31.46, "learning_rate": 6.557142857142857e-06, "loss": 1.3662, "step": 6082000 }, { "epoch": 31.46, "learning_rate": 6.553571428571428e-06, "loss": 1.3648, "step": 6082500 }, { "epoch": 31.46, "learning_rate": 6.550000000000001e-06, "loss": 1.3745, "step": 6083000 }, { "epoch": 31.47, "learning_rate": 6.546428571428572e-06, "loss": 1.3727, "step": 6083500 }, { "epoch": 31.47, "learning_rate": 6.542857142857143e-06, "loss": 1.3641, "step": 6084000 }, { "epoch": 31.47, "learning_rate": 6.539285714285714e-06, "loss": 1.3749, "step": 6084500 }, { "epoch": 31.48, "learning_rate": 6.535714285714287e-06, "loss": 1.3914, "step": 6085000 }, { "epoch": 31.48, "learning_rate": 6.532142857142858e-06, "loss": 1.3994, "step": 6085500 }, { "epoch": 31.48, "learning_rate": 6.5285714285714285e-06, "loss": 1.3837, "step": 6086000 }, { "epoch": 31.48, "learning_rate": 6.525e-06, "loss": 1.3803, "step": 6086500 }, { "epoch": 31.49, "learning_rate": 6.521428571428571e-06, "loss": 1.3724, "step": 6087000 }, { "epoch": 31.49, "learning_rate": 6.517857142857144e-06, "loss": 1.3889, "step": 6087500 }, { "epoch": 31.49, "learning_rate": 6.5142857142857145e-06, "loss": 1.368, "step": 6088000 }, { "epoch": 31.49, "learning_rate": 6.510714285714287e-06, "loss": 1.3672, "step": 6088500 }, { "epoch": 31.5, "learning_rate": 6.507142857142858e-06, "loss": 1.3674, "step": 6089000 }, { "epoch": 31.5, "learning_rate": 6.503571428571428e-06, "loss": 1.3862, "step": 6089500 }, { "epoch": 31.5, "learning_rate": 6.5000000000000004e-06, "loss": 1.3639, "step": 6090000 }, { "epoch": 31.5, "learning_rate": 6.496428571428571e-06, "loss": 1.3655, "step": 6090500 }, { "epoch": 31.51, "learning_rate": 6.492857142857144e-06, "loss": 1.3714, "step": 6091000 }, { "epoch": 31.51, "learning_rate": 6.489285714285715e-06, "loss": 1.3631, "step": 6091500 }, { "epoch": 31.51, "learning_rate": 6.485714285714286e-06, "loss": 1.3669, "step": 6092000 }, { "epoch": 31.52, "learning_rate": 6.482142857142857e-06, "loss": 1.3672, "step": 6092500 }, { "epoch": 31.52, "learning_rate": 6.478571428571428e-06, "loss": 1.3773, "step": 6093000 }, { "epoch": 31.52, "learning_rate": 6.475000000000001e-06, "loss": 1.3671, "step": 6093500 }, { "epoch": 31.52, "learning_rate": 6.4714285714285715e-06, "loss": 1.3691, "step": 6094000 }, { "epoch": 31.53, "learning_rate": 6.467857142857143e-06, "loss": 1.3723, "step": 6094500 }, { "epoch": 31.53, "learning_rate": 6.464285714285714e-06, "loss": 1.376, "step": 6095000 }, { "epoch": 31.53, "learning_rate": 6.460714285714287e-06, "loss": 1.3674, "step": 6095500 }, { "epoch": 31.53, "learning_rate": 6.4571428571428575e-06, "loss": 1.364, "step": 6096000 }, { "epoch": 31.54, "learning_rate": 6.453571428571428e-06, "loss": 1.3806, "step": 6096500 }, { "epoch": 31.54, "learning_rate": 6.45e-06, "loss": 1.364, "step": 6097000 }, { "epoch": 31.54, "learning_rate": 6.446428571428571e-06, "loss": 1.3652, "step": 6097500 }, { "epoch": 31.54, "learning_rate": 6.4428571428571435e-06, "loss": 1.3753, "step": 6098000 }, { "epoch": 31.55, "learning_rate": 6.439285714285714e-06, "loss": 1.3714, "step": 6098500 }, { "epoch": 31.55, "learning_rate": 6.435714285714287e-06, "loss": 1.3635, "step": 6099000 }, { "epoch": 31.55, "learning_rate": 6.432142857142858e-06, "loss": 1.3634, "step": 6099500 }, { "epoch": 31.55, "learning_rate": 6.428571428571429e-06, "loss": 1.3607, "step": 6100000 }, { "epoch": 31.56, "learning_rate": 6.425e-06, "loss": 1.3627, "step": 6100500 }, { "epoch": 31.56, "learning_rate": 6.421428571428571e-06, "loss": 1.3624, "step": 6101000 }, { "epoch": 31.56, "learning_rate": 6.417857142857144e-06, "loss": 1.3665, "step": 6101500 }, { "epoch": 31.57, "learning_rate": 6.414285714285715e-06, "loss": 1.3659, "step": 6102000 }, { "epoch": 31.57, "learning_rate": 6.410714285714286e-06, "loss": 1.3648, "step": 6102500 }, { "epoch": 31.57, "learning_rate": 6.407142857142857e-06, "loss": 1.3684, "step": 6103000 }, { "epoch": 31.57, "learning_rate": 6.403571428571428e-06, "loss": 1.3653, "step": 6103500 }, { "epoch": 31.58, "learning_rate": 6.4000000000000006e-06, "loss": 1.3543, "step": 6104000 }, { "epoch": 31.58, "learning_rate": 6.396428571428571e-06, "loss": 1.3732, "step": 6104500 }, { "epoch": 31.58, "learning_rate": 6.392857142857143e-06, "loss": 1.3655, "step": 6105000 }, { "epoch": 31.58, "learning_rate": 6.389285714285714e-06, "loss": 1.3676, "step": 6105500 }, { "epoch": 31.59, "learning_rate": 6.3857142857142865e-06, "loss": 1.3697, "step": 6106000 }, { "epoch": 31.59, "learning_rate": 6.382142857142857e-06, "loss": 1.3575, "step": 6106500 }, { "epoch": 31.59, "learning_rate": 6.378571428571428e-06, "loss": 1.3716, "step": 6107000 }, { "epoch": 31.59, "learning_rate": 6.375000000000001e-06, "loss": 1.3633, "step": 6107500 }, { "epoch": 31.6, "learning_rate": 6.371428571428572e-06, "loss": 1.3617, "step": 6108000 }, { "epoch": 31.6, "learning_rate": 6.367857142857143e-06, "loss": 1.3725, "step": 6108500 }, { "epoch": 31.6, "learning_rate": 6.364285714285714e-06, "loss": 1.3588, "step": 6109000 }, { "epoch": 31.61, "learning_rate": 6.360714285714287e-06, "loss": 1.3654, "step": 6109500 }, { "epoch": 31.61, "learning_rate": 6.357142857142858e-06, "loss": 1.3722, "step": 6110000 }, { "epoch": 31.61, "learning_rate": 6.353571428571429e-06, "loss": 1.369, "step": 6110500 }, { "epoch": 31.61, "learning_rate": 6.35e-06, "loss": 1.3659, "step": 6111000 }, { "epoch": 31.62, "learning_rate": 6.346428571428571e-06, "loss": 1.3663, "step": 6111500 }, { "epoch": 31.62, "learning_rate": 6.342857142857144e-06, "loss": 1.3629, "step": 6112000 }, { "epoch": 31.62, "learning_rate": 6.3392857142857145e-06, "loss": 1.3638, "step": 6112500 }, { "epoch": 31.62, "learning_rate": 6.335714285714286e-06, "loss": 1.3627, "step": 6113000 }, { "epoch": 31.63, "learning_rate": 6.332142857142857e-06, "loss": 1.3658, "step": 6113500 }, { "epoch": 31.63, "learning_rate": 6.3285714285714296e-06, "loss": 1.3745, "step": 6114000 }, { "epoch": 31.63, "learning_rate": 6.3250000000000004e-06, "loss": 1.3747, "step": 6114500 }, { "epoch": 31.63, "learning_rate": 6.321428571428571e-06, "loss": 1.3636, "step": 6115000 }, { "epoch": 31.64, "learning_rate": 6.317857142857144e-06, "loss": 1.3678, "step": 6115500 }, { "epoch": 31.64, "learning_rate": 6.314285714285714e-06, "loss": 1.3552, "step": 6116000 }, { "epoch": 31.64, "learning_rate": 6.310714285714286e-06, "loss": 1.3669, "step": 6116500 }, { "epoch": 31.65, "learning_rate": 6.307142857142857e-06, "loss": 1.3566, "step": 6117000 }, { "epoch": 31.65, "learning_rate": 6.30357142857143e-06, "loss": 1.3607, "step": 6117500 }, { "epoch": 31.65, "learning_rate": 6.300000000000001e-06, "loss": 1.3697, "step": 6118000 }, { "epoch": 31.65, "learning_rate": 6.2964285714285715e-06, "loss": 1.3641, "step": 6118500 }, { "epoch": 31.66, "learning_rate": 6.292857142857143e-06, "loss": 1.3648, "step": 6119000 }, { "epoch": 31.66, "learning_rate": 6.289285714285714e-06, "loss": 1.3653, "step": 6119500 }, { "epoch": 31.66, "learning_rate": 6.285714285714287e-06, "loss": 1.3732, "step": 6120000 }, { "epoch": 31.66, "learning_rate": 6.2821428571428575e-06, "loss": 1.3685, "step": 6120500 }, { "epoch": 31.67, "learning_rate": 6.278571428571429e-06, "loss": 1.3635, "step": 6121000 }, { "epoch": 31.67, "learning_rate": 6.275e-06, "loss": 1.3647, "step": 6121500 }, { "epoch": 31.67, "learning_rate": 6.271428571428571e-06, "loss": 1.3628, "step": 6122000 }, { "epoch": 31.67, "learning_rate": 6.2678571428571435e-06, "loss": 1.3546, "step": 6122500 }, { "epoch": 31.68, "learning_rate": 6.264285714285714e-06, "loss": 1.3643, "step": 6123000 }, { "epoch": 31.68, "learning_rate": 6.260714285714286e-06, "loss": 1.3673, "step": 6123500 }, { "epoch": 31.68, "learning_rate": 6.257142857142857e-06, "loss": 1.3675, "step": 6124000 }, { "epoch": 31.69, "learning_rate": 6.2535714285714294e-06, "loss": 1.3604, "step": 6124500 }, { "epoch": 31.69, "learning_rate": 6.25e-06, "loss": 1.3642, "step": 6125000 }, { "epoch": 31.69, "learning_rate": 6.246428571428572e-06, "loss": 1.3643, "step": 6125500 }, { "epoch": 31.69, "learning_rate": 6.242857142857144e-06, "loss": 1.3632, "step": 6126000 }, { "epoch": 31.7, "learning_rate": 6.2392857142857146e-06, "loss": 1.3693, "step": 6126500 }, { "epoch": 31.7, "learning_rate": 6.2357142857142854e-06, "loss": 1.375, "step": 6127000 }, { "epoch": 31.7, "learning_rate": 6.232142857142857e-06, "loss": 1.3621, "step": 6127500 }, { "epoch": 31.7, "learning_rate": 6.228571428571429e-06, "loss": 1.3705, "step": 6128000 }, { "epoch": 31.71, "learning_rate": 6.2250000000000005e-06, "loss": 1.3669, "step": 6128500 }, { "epoch": 31.71, "learning_rate": 6.221428571428571e-06, "loss": 1.3596, "step": 6129000 }, { "epoch": 31.71, "learning_rate": 6.217857142857143e-06, "loss": 1.3615, "step": 6129500 }, { "epoch": 31.71, "learning_rate": 6.214285714285715e-06, "loss": 1.3609, "step": 6130000 }, { "epoch": 31.72, "learning_rate": 6.210714285714286e-06, "loss": 1.3586, "step": 6130500 }, { "epoch": 31.72, "learning_rate": 6.207142857142857e-06, "loss": 1.3652, "step": 6131000 }, { "epoch": 31.72, "learning_rate": 6.203571428571429e-06, "loss": 1.3604, "step": 6131500 }, { "epoch": 31.73, "learning_rate": 6.2e-06, "loss": 1.3674, "step": 6132000 }, { "epoch": 31.73, "learning_rate": 6.196428571428572e-06, "loss": 1.3666, "step": 6132500 }, { "epoch": 31.73, "learning_rate": 6.192857142857143e-06, "loss": 1.3677, "step": 6133000 }, { "epoch": 31.73, "learning_rate": 6.189285714285715e-06, "loss": 1.3719, "step": 6133500 }, { "epoch": 31.74, "learning_rate": 6.185714285714287e-06, "loss": 1.3697, "step": 6134000 }, { "epoch": 31.74, "learning_rate": 6.182142857142857e-06, "loss": 1.3748, "step": 6134500 }, { "epoch": 31.74, "learning_rate": 6.1785714285714285e-06, "loss": 1.3649, "step": 6135000 }, { "epoch": 31.74, "learning_rate": 6.175e-06, "loss": 1.3798, "step": 6135500 }, { "epoch": 31.75, "learning_rate": 6.171428571428572e-06, "loss": 1.377, "step": 6136000 }, { "epoch": 31.75, "learning_rate": 6.167857142857144e-06, "loss": 1.3767, "step": 6136500 }, { "epoch": 31.75, "learning_rate": 6.1642857142857144e-06, "loss": 1.37, "step": 6137000 }, { "epoch": 31.75, "learning_rate": 6.160714285714286e-06, "loss": 1.3638, "step": 6137500 }, { "epoch": 31.76, "learning_rate": 6.157142857142857e-06, "loss": 1.3522, "step": 6138000 }, { "epoch": 31.76, "learning_rate": 6.153571428571429e-06, "loss": 1.3571, "step": 6138500 }, { "epoch": 31.76, "learning_rate": 6.15e-06, "loss": 1.3626, "step": 6139000 }, { "epoch": 31.77, "learning_rate": 6.146428571428572e-06, "loss": 1.3628, "step": 6139500 }, { "epoch": 31.77, "learning_rate": 6.142857142857143e-06, "loss": 1.3598, "step": 6140000 }, { "epoch": 31.77, "learning_rate": 6.139285714285715e-06, "loss": 1.373, "step": 6140500 }, { "epoch": 31.77, "learning_rate": 6.135714285714286e-06, "loss": 1.3699, "step": 6141000 }, { "epoch": 31.78, "learning_rate": 6.132142857142857e-06, "loss": 1.3666, "step": 6141500 }, { "epoch": 31.78, "learning_rate": 6.128571428571429e-06, "loss": 1.3626, "step": 6142000 }, { "epoch": 31.78, "learning_rate": 6.125e-06, "loss": 1.3633, "step": 6142500 }, { "epoch": 31.78, "learning_rate": 6.1214285714285715e-06, "loss": 1.3676, "step": 6143000 }, { "epoch": 31.79, "learning_rate": 6.117857142857143e-06, "loss": 1.3771, "step": 6143500 }, { "epoch": 31.79, "learning_rate": 6.114285714285715e-06, "loss": 1.3604, "step": 6144000 }, { "epoch": 31.79, "learning_rate": 6.110714285714287e-06, "loss": 1.3708, "step": 6144500 }, { "epoch": 31.79, "learning_rate": 6.1071428571428575e-06, "loss": 1.3712, "step": 6145000 }, { "epoch": 31.8, "learning_rate": 6.103571428571428e-06, "loss": 1.365, "step": 6145500 }, { "epoch": 31.8, "learning_rate": 6.1e-06, "loss": 1.3638, "step": 6146000 }, { "epoch": 31.8, "learning_rate": 6.096428571428572e-06, "loss": 1.3582, "step": 6146500 }, { "epoch": 31.81, "learning_rate": 6.0928571428571435e-06, "loss": 1.368, "step": 6147000 }, { "epoch": 31.81, "learning_rate": 6.089285714285714e-06, "loss": 1.3651, "step": 6147500 }, { "epoch": 31.81, "learning_rate": 6.085714285714286e-06, "loss": 1.3632, "step": 6148000 }, { "epoch": 31.81, "learning_rate": 6.082142857142858e-06, "loss": 1.3648, "step": 6148500 }, { "epoch": 31.82, "learning_rate": 6.0785714285714286e-06, "loss": 1.3614, "step": 6149000 }, { "epoch": 31.82, "learning_rate": 6.075e-06, "loss": 1.3641, "step": 6149500 }, { "epoch": 31.82, "learning_rate": 6.071428571428572e-06, "loss": 1.3566, "step": 6150000 }, { "epoch": 31.82, "learning_rate": 6.067857142857143e-06, "loss": 1.3601, "step": 6150500 }, { "epoch": 31.83, "learning_rate": 6.0642857142857145e-06, "loss": 1.3634, "step": 6151000 }, { "epoch": 31.83, "learning_rate": 6.060714285714286e-06, "loss": 1.3675, "step": 6151500 }, { "epoch": 31.83, "learning_rate": 6.057142857142858e-06, "loss": 1.3728, "step": 6152000 }, { "epoch": 31.83, "learning_rate": 6.053571428571429e-06, "loss": 1.3629, "step": 6152500 }, { "epoch": 31.84, "learning_rate": 6.0500000000000005e-06, "loss": 1.3695, "step": 6153000 }, { "epoch": 31.84, "learning_rate": 6.046428571428571e-06, "loss": 1.3621, "step": 6153500 }, { "epoch": 31.84, "learning_rate": 6.042857142857143e-06, "loss": 1.3618, "step": 6154000 }, { "epoch": 31.85, "learning_rate": 6.039285714285715e-06, "loss": 1.3634, "step": 6154500 }, { "epoch": 31.85, "learning_rate": 6.0357142857142865e-06, "loss": 1.36, "step": 6155000 }, { "epoch": 31.85, "learning_rate": 6.032142857142857e-06, "loss": 1.3639, "step": 6155500 }, { "epoch": 31.85, "learning_rate": 6.028571428571428e-06, "loss": 1.3684, "step": 6156000 }, { "epoch": 31.86, "learning_rate": 6.025e-06, "loss": 1.3591, "step": 6156500 }, { "epoch": 31.86, "learning_rate": 6.021428571428572e-06, "loss": 1.3685, "step": 6157000 }, { "epoch": 31.86, "learning_rate": 6.017857142857143e-06, "loss": 1.3633, "step": 6157500 }, { "epoch": 31.86, "learning_rate": 6.014285714285715e-06, "loss": 1.3637, "step": 6158000 }, { "epoch": 31.87, "learning_rate": 6.010714285714286e-06, "loss": 1.3673, "step": 6158500 }, { "epoch": 31.87, "learning_rate": 6.007142857142858e-06, "loss": 1.3782, "step": 6159000 }, { "epoch": 31.87, "learning_rate": 6.0035714285714284e-06, "loss": 1.3629, "step": 6159500 }, { "epoch": 31.87, "learning_rate": 6e-06, "loss": 1.3637, "step": 6160000 }, { "epoch": 31.88, "learning_rate": 5.996428571428572e-06, "loss": 1.3645, "step": 6160500 }, { "epoch": 31.88, "learning_rate": 5.992857142857143e-06, "loss": 1.364, "step": 6161000 }, { "epoch": 31.88, "learning_rate": 5.989285714285714e-06, "loss": 1.356, "step": 6161500 }, { "epoch": 31.89, "learning_rate": 5.985714285714286e-06, "loss": 1.3459, "step": 6162000 }, { "epoch": 31.89, "learning_rate": 5.982142857142858e-06, "loss": 1.3678, "step": 6162500 }, { "epoch": 31.89, "learning_rate": 5.978571428571429e-06, "loss": 1.3761, "step": 6163000 }, { "epoch": 31.89, "learning_rate": 5.975e-06, "loss": 1.3531, "step": 6163500 }, { "epoch": 31.9, "learning_rate": 5.971428571428571e-06, "loss": 1.3658, "step": 6164000 }, { "epoch": 31.9, "learning_rate": 5.967857142857143e-06, "loss": 1.368, "step": 6164500 }, { "epoch": 31.9, "learning_rate": 5.964285714285715e-06, "loss": 1.3661, "step": 6165000 }, { "epoch": 31.9, "learning_rate": 5.960714285714286e-06, "loss": 1.3672, "step": 6165500 }, { "epoch": 31.91, "learning_rate": 5.957142857142858e-06, "loss": 1.3622, "step": 6166000 }, { "epoch": 31.91, "learning_rate": 5.953571428571429e-06, "loss": 1.3578, "step": 6166500 }, { "epoch": 31.91, "learning_rate": 5.95e-06, "loss": 1.3666, "step": 6167000 }, { "epoch": 31.91, "learning_rate": 5.9464285714285715e-06, "loss": 1.3642, "step": 6167500 }, { "epoch": 31.92, "learning_rate": 5.942857142857143e-06, "loss": 1.3584, "step": 6168000 }, { "epoch": 31.92, "learning_rate": 5.939285714285715e-06, "loss": 1.3587, "step": 6168500 }, { "epoch": 31.92, "learning_rate": 5.935714285714286e-06, "loss": 1.3598, "step": 6169000 }, { "epoch": 31.92, "learning_rate": 5.9321428571428575e-06, "loss": 1.362, "step": 6169500 }, { "epoch": 31.93, "learning_rate": 5.928571428571429e-06, "loss": 1.3802, "step": 6170000 }, { "epoch": 31.93, "learning_rate": 5.925e-06, "loss": 1.3688, "step": 6170500 }, { "epoch": 31.93, "learning_rate": 5.921428571428572e-06, "loss": 1.3666, "step": 6171000 }, { "epoch": 31.94, "learning_rate": 5.9178571428571434e-06, "loss": 1.3718, "step": 6171500 }, { "epoch": 31.94, "learning_rate": 5.914285714285714e-06, "loss": 1.3669, "step": 6172000 }, { "epoch": 31.94, "learning_rate": 5.910714285714286e-06, "loss": 1.3657, "step": 6172500 }, { "epoch": 31.94, "learning_rate": 5.907142857142858e-06, "loss": 1.3676, "step": 6173000 }, { "epoch": 31.95, "learning_rate": 5.903571428571429e-06, "loss": 1.3631, "step": 6173500 }, { "epoch": 31.95, "learning_rate": 5.9e-06, "loss": 1.36, "step": 6174000 }, { "epoch": 31.95, "learning_rate": 5.896428571428571e-06, "loss": 1.3684, "step": 6174500 }, { "epoch": 31.95, "learning_rate": 5.892857142857143e-06, "loss": 1.354, "step": 6175000 }, { "epoch": 31.96, "learning_rate": 5.8892857142857145e-06, "loss": 1.3576, "step": 6175500 }, { "epoch": 31.96, "learning_rate": 5.885714285714286e-06, "loss": 1.38, "step": 6176000 }, { "epoch": 31.96, "learning_rate": 5.882142857142858e-06, "loss": 1.3657, "step": 6176500 }, { "epoch": 31.96, "learning_rate": 5.878571428571429e-06, "loss": 1.3725, "step": 6177000 }, { "epoch": 31.97, "learning_rate": 5.875e-06, "loss": 1.3589, "step": 6177500 }, { "epoch": 31.97, "learning_rate": 5.871428571428571e-06, "loss": 1.3606, "step": 6178000 }, { "epoch": 31.97, "learning_rate": 5.867857142857143e-06, "loss": 1.3563, "step": 6178500 }, { "epoch": 31.98, "learning_rate": 5.864285714285715e-06, "loss": 1.365, "step": 6179000 }, { "epoch": 31.98, "learning_rate": 5.860714285714286e-06, "loss": 1.3674, "step": 6179500 }, { "epoch": 31.98, "learning_rate": 5.857142857142857e-06, "loss": 1.3692, "step": 6180000 }, { "epoch": 31.98, "learning_rate": 5.853571428571429e-06, "loss": 1.3532, "step": 6180500 }, { "epoch": 31.99, "learning_rate": 5.850000000000001e-06, "loss": 1.3635, "step": 6181000 }, { "epoch": 31.99, "learning_rate": 5.846428571428572e-06, "loss": 1.3627, "step": 6181500 }, { "epoch": 31.99, "learning_rate": 5.842857142857143e-06, "loss": 1.3725, "step": 6182000 }, { "epoch": 31.99, "learning_rate": 5.839285714285714e-06, "loss": 1.3668, "step": 6182500 }, { "epoch": 32.0, "learning_rate": 5.835714285714286e-06, "loss": 1.377, "step": 6183000 }, { "epoch": 32.0, "learning_rate": 5.8321428571428576e-06, "loss": 1.367, "step": 6183500 }, { "epoch": 32.0, "learning_rate": 5.828571428571429e-06, "loss": 1.3563, "step": 6184000 }, { "epoch": 32.0, "learning_rate": 5.825000000000001e-06, "loss": 1.3758, "step": 6184500 }, { "epoch": 32.01, "learning_rate": 5.821428571428572e-06, "loss": 1.3652, "step": 6185000 }, { "epoch": 32.01, "learning_rate": 5.817857142857143e-06, "loss": 1.3643, "step": 6185500 }, { "epoch": 32.01, "learning_rate": 5.814285714285714e-06, "loss": 1.3582, "step": 6186000 }, { "epoch": 32.02, "learning_rate": 5.810714285714286e-06, "loss": 1.3534, "step": 6186500 }, { "epoch": 32.02, "learning_rate": 5.807142857142858e-06, "loss": 1.367, "step": 6187000 }, { "epoch": 32.02, "learning_rate": 5.803571428571429e-06, "loss": 1.3655, "step": 6187500 }, { "epoch": 32.02, "learning_rate": 5.8e-06, "loss": 1.364, "step": 6188000 }, { "epoch": 32.03, "learning_rate": 5.796428571428571e-06, "loss": 1.3622, "step": 6188500 }, { "epoch": 32.03, "learning_rate": 5.792857142857143e-06, "loss": 1.3627, "step": 6189000 }, { "epoch": 32.03, "learning_rate": 5.789285714285715e-06, "loss": 1.3654, "step": 6189500 }, { "epoch": 32.03, "learning_rate": 5.785714285714286e-06, "loss": 1.3566, "step": 6190000 }, { "epoch": 32.04, "learning_rate": 5.782142857142857e-06, "loss": 1.3556, "step": 6190500 }, { "epoch": 32.04, "learning_rate": 5.778571428571429e-06, "loss": 1.3605, "step": 6191000 }, { "epoch": 32.04, "learning_rate": 5.775000000000001e-06, "loss": 1.3624, "step": 6191500 }, { "epoch": 32.04, "learning_rate": 5.7714285714285715e-06, "loss": 1.3639, "step": 6192000 }, { "epoch": 32.05, "learning_rate": 5.767857142857143e-06, "loss": 1.3561, "step": 6192500 }, { "epoch": 32.05, "learning_rate": 5.764285714285714e-06, "loss": 1.3675, "step": 6193000 }, { "epoch": 32.05, "learning_rate": 5.760714285714286e-06, "loss": 1.3681, "step": 6193500 }, { "epoch": 32.06, "learning_rate": 5.7571428571428574e-06, "loss": 1.3634, "step": 6194000 }, { "epoch": 32.06, "learning_rate": 5.753571428571429e-06, "loss": 1.3625, "step": 6194500 }, { "epoch": 32.06, "learning_rate": 5.750000000000001e-06, "loss": 1.3696, "step": 6195000 }, { "epoch": 32.06, "learning_rate": 5.746428571428572e-06, "loss": 1.3622, "step": 6195500 }, { "epoch": 32.07, "learning_rate": 5.7428571428571426e-06, "loss": 1.3619, "step": 6196000 }, { "epoch": 32.07, "learning_rate": 5.739285714285714e-06, "loss": 1.3673, "step": 6196500 }, { "epoch": 32.07, "learning_rate": 5.735714285714286e-06, "loss": 1.3646, "step": 6197000 }, { "epoch": 32.07, "learning_rate": 5.732142857142858e-06, "loss": 1.3592, "step": 6197500 }, { "epoch": 32.08, "learning_rate": 5.728571428571429e-06, "loss": 1.3614, "step": 6198000 }, { "epoch": 32.08, "learning_rate": 5.725e-06, "loss": 1.3597, "step": 6198500 }, { "epoch": 32.08, "learning_rate": 5.721428571428572e-06, "loss": 1.3582, "step": 6199000 }, { "epoch": 32.08, "learning_rate": 5.717857142857143e-06, "loss": 1.3631, "step": 6199500 }, { "epoch": 32.09, "learning_rate": 5.7142857142857145e-06, "loss": 1.3728, "step": 6200000 }, { "epoch": 32.09, "learning_rate": 5.710714285714286e-06, "loss": 1.362, "step": 6200500 }, { "epoch": 32.09, "learning_rate": 5.707142857142857e-06, "loss": 1.3623, "step": 6201000 }, { "epoch": 32.1, "learning_rate": 5.703571428571429e-06, "loss": 1.359, "step": 6201500 }, { "epoch": 32.1, "learning_rate": 5.7000000000000005e-06, "loss": 1.3609, "step": 6202000 }, { "epoch": 32.1, "learning_rate": 5.696428571428572e-06, "loss": 1.3572, "step": 6202500 }, { "epoch": 32.1, "learning_rate": 5.692857142857143e-06, "loss": 1.3545, "step": 6203000 }, { "epoch": 32.11, "learning_rate": 5.689285714285715e-06, "loss": 1.3605, "step": 6203500 }, { "epoch": 32.11, "learning_rate": 5.685714285714286e-06, "loss": 1.3502, "step": 6204000 }, { "epoch": 32.11, "learning_rate": 5.682142857142857e-06, "loss": 1.3633, "step": 6204500 }, { "epoch": 32.11, "learning_rate": 5.678571428571429e-06, "loss": 1.3555, "step": 6205000 }, { "epoch": 32.12, "learning_rate": 5.675000000000001e-06, "loss": 1.3663, "step": 6205500 }, { "epoch": 32.12, "learning_rate": 5.671428571428572e-06, "loss": 1.3618, "step": 6206000 }, { "epoch": 32.12, "learning_rate": 5.6678571428571424e-06, "loss": 1.3514, "step": 6206500 }, { "epoch": 32.12, "learning_rate": 5.664285714285714e-06, "loss": 1.3512, "step": 6207000 }, { "epoch": 32.13, "learning_rate": 5.660714285714286e-06, "loss": 1.3621, "step": 6207500 }, { "epoch": 32.13, "learning_rate": 5.6571428571428576e-06, "loss": 1.356, "step": 6208000 }, { "epoch": 32.13, "learning_rate": 5.653571428571429e-06, "loss": 1.3477, "step": 6208500 }, { "epoch": 32.14, "learning_rate": 5.65e-06, "loss": 1.3532, "step": 6209000 }, { "epoch": 32.14, "learning_rate": 5.646428571428572e-06, "loss": 1.3606, "step": 6209500 }, { "epoch": 32.14, "learning_rate": 5.642857142857143e-06, "loss": 1.3603, "step": 6210000 }, { "epoch": 32.14, "learning_rate": 5.639285714285714e-06, "loss": 1.3625, "step": 6210500 }, { "epoch": 32.15, "learning_rate": 5.635714285714286e-06, "loss": 1.3575, "step": 6211000 }, { "epoch": 32.15, "learning_rate": 5.632142857142857e-06, "loss": 1.3641, "step": 6211500 }, { "epoch": 32.15, "learning_rate": 5.628571428571429e-06, "loss": 1.3623, "step": 6212000 }, { "epoch": 32.15, "learning_rate": 5.625e-06, "loss": 1.3549, "step": 6212500 }, { "epoch": 32.16, "learning_rate": 5.621428571428572e-06, "loss": 1.3568, "step": 6213000 }, { "epoch": 32.16, "learning_rate": 5.617857142857144e-06, "loss": 1.3638, "step": 6213500 }, { "epoch": 32.16, "learning_rate": 5.614285714285715e-06, "loss": 1.3585, "step": 6214000 }, { "epoch": 32.16, "learning_rate": 5.6107142857142855e-06, "loss": 1.3624, "step": 6214500 }, { "epoch": 32.17, "learning_rate": 5.607142857142857e-06, "loss": 1.3605, "step": 6215000 }, { "epoch": 32.17, "learning_rate": 5.603571428571429e-06, "loss": 1.3612, "step": 6215500 }, { "epoch": 32.17, "learning_rate": 5.600000000000001e-06, "loss": 1.3692, "step": 6216000 }, { "epoch": 32.18, "learning_rate": 5.596428571428572e-06, "loss": 1.364, "step": 6216500 }, { "epoch": 32.18, "learning_rate": 5.592857142857143e-06, "loss": 1.3619, "step": 6217000 }, { "epoch": 32.18, "learning_rate": 5.589285714285714e-06, "loss": 1.3605, "step": 6217500 }, { "epoch": 32.18, "learning_rate": 5.585714285714286e-06, "loss": 1.3624, "step": 6218000 }, { "epoch": 32.19, "learning_rate": 5.582142857142857e-06, "loss": 1.3708, "step": 6218500 }, { "epoch": 32.19, "learning_rate": 5.578571428571429e-06, "loss": 1.3559, "step": 6219000 }, { "epoch": 32.19, "learning_rate": 5.575e-06, "loss": 1.3594, "step": 6219500 }, { "epoch": 32.19, "learning_rate": 5.571428571428572e-06, "loss": 1.3563, "step": 6220000 }, { "epoch": 32.2, "learning_rate": 5.567857142857143e-06, "loss": 1.3626, "step": 6220500 }, { "epoch": 32.2, "learning_rate": 5.564285714285714e-06, "loss": 1.3663, "step": 6221000 }, { "epoch": 32.2, "learning_rate": 5.560714285714286e-06, "loss": 1.3594, "step": 6221500 }, { "epoch": 32.2, "learning_rate": 5.557142857142858e-06, "loss": 1.3627, "step": 6222000 }, { "epoch": 32.21, "learning_rate": 5.5535714285714285e-06, "loss": 1.3582, "step": 6222500 }, { "epoch": 32.21, "learning_rate": 5.55e-06, "loss": 1.3654, "step": 6223000 }, { "epoch": 32.21, "learning_rate": 5.546428571428572e-06, "loss": 1.3675, "step": 6223500 }, { "epoch": 32.22, "learning_rate": 5.542857142857144e-06, "loss": 1.3619, "step": 6224000 }, { "epoch": 32.22, "learning_rate": 5.5392857142857145e-06, "loss": 1.3629, "step": 6224500 }, { "epoch": 32.22, "learning_rate": 5.535714285714285e-06, "loss": 1.3594, "step": 6225000 }, { "epoch": 32.22, "learning_rate": 5.532142857142857e-06, "loss": 1.3539, "step": 6225500 }, { "epoch": 32.23, "learning_rate": 5.528571428571429e-06, "loss": 1.3627, "step": 6226000 }, { "epoch": 32.23, "learning_rate": 5.5250000000000005e-06, "loss": 1.3619, "step": 6226500 }, { "epoch": 32.23, "learning_rate": 5.521428571428572e-06, "loss": 1.3552, "step": 6227000 }, { "epoch": 32.23, "learning_rate": 5.517857142857143e-06, "loss": 1.3648, "step": 6227500 }, { "epoch": 32.24, "learning_rate": 5.514285714285715e-06, "loss": 1.3729, "step": 6228000 }, { "epoch": 32.24, "learning_rate": 5.510714285714286e-06, "loss": 1.3676, "step": 6228500 }, { "epoch": 32.24, "learning_rate": 5.507142857142857e-06, "loss": 1.3548, "step": 6229000 }, { "epoch": 32.24, "learning_rate": 5.503571428571429e-06, "loss": 1.3661, "step": 6229500 }, { "epoch": 32.25, "learning_rate": 5.500000000000001e-06, "loss": 1.354, "step": 6230000 }, { "epoch": 32.25, "learning_rate": 5.4964285714285716e-06, "loss": 1.3588, "step": 6230500 }, { "epoch": 32.25, "learning_rate": 5.492857142857143e-06, "loss": 1.3689, "step": 6231000 }, { "epoch": 32.26, "learning_rate": 5.489285714285715e-06, "loss": 1.3536, "step": 6231500 }, { "epoch": 32.26, "learning_rate": 5.485714285714286e-06, "loss": 1.3612, "step": 6232000 }, { "epoch": 32.26, "learning_rate": 5.4821428571428575e-06, "loss": 1.3612, "step": 6232500 }, { "epoch": 32.26, "learning_rate": 5.478571428571428e-06, "loss": 1.3684, "step": 6233000 }, { "epoch": 32.27, "learning_rate": 5.475e-06, "loss": 1.3639, "step": 6233500 }, { "epoch": 32.27, "learning_rate": 5.471428571428572e-06, "loss": 1.358, "step": 6234000 }, { "epoch": 32.27, "learning_rate": 5.4678571428571435e-06, "loss": 1.3622, "step": 6234500 }, { "epoch": 32.27, "learning_rate": 5.464285714285715e-06, "loss": 1.3567, "step": 6235000 }, { "epoch": 32.28, "learning_rate": 5.460714285714286e-06, "loss": 1.36, "step": 6235500 }, { "epoch": 32.28, "learning_rate": 5.457142857142857e-06, "loss": 1.3615, "step": 6236000 }, { "epoch": 32.28, "learning_rate": 5.453571428571429e-06, "loss": 1.3735, "step": 6236500 }, { "epoch": 32.28, "learning_rate": 5.45e-06, "loss": 1.3687, "step": 6237000 }, { "epoch": 32.29, "learning_rate": 5.446428571428572e-06, "loss": 1.3672, "step": 6237500 }, { "epoch": 32.29, "learning_rate": 5.442857142857143e-06, "loss": 1.3561, "step": 6238000 }, { "epoch": 32.29, "learning_rate": 5.439285714285715e-06, "loss": 1.3604, "step": 6238500 }, { "epoch": 32.3, "learning_rate": 5.4357142857142855e-06, "loss": 1.3687, "step": 6239000 }, { "epoch": 32.3, "learning_rate": 5.432142857142857e-06, "loss": 1.3593, "step": 6239500 }, { "epoch": 32.3, "learning_rate": 5.428571428571429e-06, "loss": 1.3629, "step": 6240000 }, { "epoch": 32.3, "learning_rate": 5.4250000000000006e-06, "loss": 1.3629, "step": 6240500 }, { "epoch": 32.31, "learning_rate": 5.4214285714285714e-06, "loss": 1.3656, "step": 6241000 }, { "epoch": 32.31, "learning_rate": 5.417857142857143e-06, "loss": 1.3608, "step": 6241500 }, { "epoch": 32.31, "learning_rate": 5.414285714285715e-06, "loss": 1.355, "step": 6242000 }, { "epoch": 32.31, "learning_rate": 5.4107142857142865e-06, "loss": 1.3641, "step": 6242500 }, { "epoch": 32.32, "learning_rate": 5.407142857142857e-06, "loss": 1.3573, "step": 6243000 }, { "epoch": 32.32, "learning_rate": 5.403571428571428e-06, "loss": 1.3595, "step": 6243500 }, { "epoch": 32.32, "learning_rate": 5.4e-06, "loss": 1.3605, "step": 6244000 }, { "epoch": 32.32, "learning_rate": 5.396428571428572e-06, "loss": 1.359, "step": 6244500 }, { "epoch": 32.33, "learning_rate": 5.392857142857143e-06, "loss": 1.3689, "step": 6245000 }, { "epoch": 32.33, "learning_rate": 5.389285714285715e-06, "loss": 1.3655, "step": 6245500 }, { "epoch": 32.33, "learning_rate": 5.385714285714286e-06, "loss": 1.3591, "step": 6246000 }, { "epoch": 32.33, "learning_rate": 5.382142857142857e-06, "loss": 1.3567, "step": 6246500 }, { "epoch": 32.34, "learning_rate": 5.3785714285714285e-06, "loss": 1.3627, "step": 6247000 }, { "epoch": 32.34, "learning_rate": 5.375e-06, "loss": 1.3531, "step": 6247500 }, { "epoch": 32.34, "learning_rate": 5.371428571428572e-06, "loss": 1.3552, "step": 6248000 }, { "epoch": 32.35, "learning_rate": 5.367857142857144e-06, "loss": 1.3568, "step": 6248500 }, { "epoch": 32.35, "learning_rate": 5.3642857142857145e-06, "loss": 1.356, "step": 6249000 }, { "epoch": 32.35, "learning_rate": 5.360714285714286e-06, "loss": 1.3575, "step": 6249500 }, { "epoch": 32.35, "learning_rate": 5.357142857142857e-06, "loss": 1.3637, "step": 6250000 }, { "epoch": 32.36, "learning_rate": 5.353571428571429e-06, "loss": 1.3534, "step": 6250500 }, { "epoch": 32.36, "learning_rate": 5.3500000000000004e-06, "loss": 1.3551, "step": 6251000 }, { "epoch": 32.36, "learning_rate": 5.346428571428571e-06, "loss": 1.3606, "step": 6251500 }, { "epoch": 32.36, "learning_rate": 5.342857142857143e-06, "loss": 1.3646, "step": 6252000 }, { "epoch": 32.37, "learning_rate": 5.339285714285715e-06, "loss": 1.3636, "step": 6252500 }, { "epoch": 32.37, "learning_rate": 5.335714285714286e-06, "loss": 1.3615, "step": 6253000 }, { "epoch": 32.37, "learning_rate": 5.332142857142857e-06, "loss": 1.3553, "step": 6253500 }, { "epoch": 32.37, "learning_rate": 5.328571428571429e-06, "loss": 1.3591, "step": 6254000 }, { "epoch": 32.38, "learning_rate": 5.325e-06, "loss": 1.3615, "step": 6254500 }, { "epoch": 32.38, "learning_rate": 5.3214285714285715e-06, "loss": 1.3714, "step": 6255000 }, { "epoch": 32.38, "learning_rate": 5.317857142857143e-06, "loss": 1.3591, "step": 6255500 }, { "epoch": 32.39, "learning_rate": 5.314285714285715e-06, "loss": 1.3631, "step": 6256000 }, { "epoch": 32.39, "learning_rate": 5.310714285714286e-06, "loss": 1.3592, "step": 6256500 }, { "epoch": 32.39, "learning_rate": 5.307142857142857e-06, "loss": 1.3496, "step": 6257000 }, { "epoch": 32.39, "learning_rate": 5.303571428571428e-06, "loss": 1.3501, "step": 6257500 }, { "epoch": 32.4, "learning_rate": 5.3e-06, "loss": 1.354, "step": 6258000 }, { "epoch": 32.4, "learning_rate": 5.296428571428572e-06, "loss": 1.3593, "step": 6258500 }, { "epoch": 32.4, "learning_rate": 5.2928571428571435e-06, "loss": 1.3453, "step": 6259000 }, { "epoch": 32.4, "learning_rate": 5.289285714285714e-06, "loss": 1.3622, "step": 6259500 }, { "epoch": 32.41, "learning_rate": 5.285714285714286e-06, "loss": 1.3579, "step": 6260000 }, { "epoch": 32.41, "learning_rate": 5.282142857142858e-06, "loss": 1.3569, "step": 6260500 }, { "epoch": 32.41, "learning_rate": 5.278571428571429e-06, "loss": 1.3561, "step": 6261000 }, { "epoch": 32.41, "learning_rate": 5.275e-06, "loss": 1.3507, "step": 6261500 }, { "epoch": 32.42, "learning_rate": 5.271428571428572e-06, "loss": 1.3478, "step": 6262000 }, { "epoch": 32.42, "learning_rate": 5.267857142857143e-06, "loss": 1.3543, "step": 6262500 }, { "epoch": 32.42, "learning_rate": 5.264285714285715e-06, "loss": 1.359, "step": 6263000 }, { "epoch": 32.43, "learning_rate": 5.260714285714286e-06, "loss": 1.3538, "step": 6263500 }, { "epoch": 32.43, "learning_rate": 5.257142857142858e-06, "loss": 1.3525, "step": 6264000 }, { "epoch": 32.43, "learning_rate": 5.253571428571429e-06, "loss": 1.3586, "step": 6264500 }, { "epoch": 32.43, "learning_rate": 5.25e-06, "loss": 1.3601, "step": 6265000 }, { "epoch": 32.44, "learning_rate": 5.246428571428571e-06, "loss": 1.3599, "step": 6265500 }, { "epoch": 32.44, "learning_rate": 5.242857142857143e-06, "loss": 1.3587, "step": 6266000 }, { "epoch": 32.44, "learning_rate": 5.239285714285715e-06, "loss": 1.3571, "step": 6266500 }, { "epoch": 32.44, "learning_rate": 5.2357142857142865e-06, "loss": 1.3542, "step": 6267000 }, { "epoch": 32.45, "learning_rate": 5.232142857142857e-06, "loss": 1.3569, "step": 6267500 }, { "epoch": 32.45, "learning_rate": 5.228571428571428e-06, "loss": 1.3601, "step": 6268000 }, { "epoch": 32.45, "learning_rate": 5.225e-06, "loss": 1.3562, "step": 6268500 }, { "epoch": 32.45, "learning_rate": 5.221428571428572e-06, "loss": 1.3561, "step": 6269000 }, { "epoch": 32.46, "learning_rate": 5.217857142857143e-06, "loss": 1.3613, "step": 6269500 }, { "epoch": 32.46, "learning_rate": 5.214285714285714e-06, "loss": 1.3563, "step": 6270000 }, { "epoch": 32.46, "learning_rate": 5.210714285714286e-06, "loss": 1.368, "step": 6270500 }, { "epoch": 32.47, "learning_rate": 5.207142857142858e-06, "loss": 1.3543, "step": 6271000 }, { "epoch": 32.47, "learning_rate": 5.2035714285714285e-06, "loss": 1.3595, "step": 6271500 }, { "epoch": 32.47, "learning_rate": 5.2e-06, "loss": 1.3559, "step": 6272000 }, { "epoch": 32.47, "learning_rate": 5.196428571428572e-06, "loss": 1.3597, "step": 6272500 }, { "epoch": 32.48, "learning_rate": 5.192857142857143e-06, "loss": 1.3584, "step": 6273000 }, { "epoch": 32.48, "learning_rate": 5.1892857142857145e-06, "loss": 1.3508, "step": 6273500 }, { "epoch": 32.48, "learning_rate": 5.185714285714286e-06, "loss": 1.362, "step": 6274000 }, { "epoch": 32.48, "learning_rate": 5.182142857142858e-06, "loss": 1.3523, "step": 6274500 }, { "epoch": 32.49, "learning_rate": 5.1785714285714296e-06, "loss": 1.3575, "step": 6275000 }, { "epoch": 32.49, "learning_rate": 5.175e-06, "loss": 1.362, "step": 6275500 }, { "epoch": 32.49, "learning_rate": 5.171428571428571e-06, "loss": 1.3541, "step": 6276000 }, { "epoch": 32.49, "learning_rate": 5.167857142857143e-06, "loss": 1.3615, "step": 6276500 }, { "epoch": 32.5, "learning_rate": 5.164285714285715e-06, "loss": 1.3671, "step": 6277000 }, { "epoch": 32.5, "learning_rate": 5.160714285714286e-06, "loss": 1.3581, "step": 6277500 }, { "epoch": 32.5, "learning_rate": 5.157142857142857e-06, "loss": 1.3661, "step": 6278000 }, { "epoch": 32.51, "learning_rate": 5.153571428571429e-06, "loss": 1.3481, "step": 6278500 }, { "epoch": 32.51, "learning_rate": 5.15e-06, "loss": 1.3632, "step": 6279000 }, { "epoch": 32.51, "learning_rate": 5.1464285714285715e-06, "loss": 1.3591, "step": 6279500 }, { "epoch": 32.51, "learning_rate": 5.142857142857143e-06, "loss": 1.3549, "step": 6280000 }, { "epoch": 32.52, "learning_rate": 5.139285714285715e-06, "loss": 1.3599, "step": 6280500 }, { "epoch": 32.52, "learning_rate": 5.135714285714286e-06, "loss": 1.3661, "step": 6281000 }, { "epoch": 32.52, "learning_rate": 5.1321428571428575e-06, "loss": 1.3626, "step": 6281500 }, { "epoch": 32.52, "learning_rate": 5.128571428571429e-06, "loss": 1.3637, "step": 6282000 }, { "epoch": 32.53, "learning_rate": 5.125e-06, "loss": 1.3552, "step": 6282500 }, { "epoch": 32.53, "learning_rate": 5.121428571428572e-06, "loss": 1.3507, "step": 6283000 }, { "epoch": 32.53, "learning_rate": 5.117857142857143e-06, "loss": 1.3626, "step": 6283500 }, { "epoch": 32.53, "learning_rate": 5.114285714285714e-06, "loss": 1.3535, "step": 6284000 }, { "epoch": 32.54, "learning_rate": 5.110714285714286e-06, "loss": 1.3533, "step": 6284500 }, { "epoch": 32.54, "learning_rate": 5.107142857142858e-06, "loss": 1.3493, "step": 6285000 }, { "epoch": 32.54, "learning_rate": 5.1035714285714294e-06, "loss": 1.3563, "step": 6285500 }, { "epoch": 32.55, "learning_rate": 5.1e-06, "loss": 1.3642, "step": 6286000 }, { "epoch": 32.55, "learning_rate": 5.096428571428571e-06, "loss": 1.3621, "step": 6286500 }, { "epoch": 32.55, "learning_rate": 5.092857142857143e-06, "loss": 1.3566, "step": 6287000 }, { "epoch": 32.55, "learning_rate": 5.0892857142857146e-06, "loss": 1.3671, "step": 6287500 }, { "epoch": 32.56, "learning_rate": 5.085714285714286e-06, "loss": 1.3525, "step": 6288000 }, { "epoch": 32.56, "learning_rate": 5.082142857142857e-06, "loss": 1.3572, "step": 6288500 }, { "epoch": 32.56, "learning_rate": 5.078571428571429e-06, "loss": 1.3567, "step": 6289000 }, { "epoch": 32.56, "learning_rate": 5.0750000000000005e-06, "loss": 1.3607, "step": 6289500 }, { "epoch": 32.57, "learning_rate": 5.071428571428571e-06, "loss": 1.3529, "step": 6290000 }, { "epoch": 32.57, "learning_rate": 5.067857142857143e-06, "loss": 1.3493, "step": 6290500 }, { "epoch": 32.57, "learning_rate": 5.064285714285715e-06, "loss": 1.353, "step": 6291000 }, { "epoch": 32.57, "learning_rate": 5.060714285714286e-06, "loss": 1.3665, "step": 6291500 }, { "epoch": 32.58, "learning_rate": 5.057142857142857e-06, "loss": 1.3639, "step": 6292000 }, { "epoch": 32.58, "learning_rate": 5.053571428571429e-06, "loss": 1.354, "step": 6292500 }, { "epoch": 32.58, "learning_rate": 5.050000000000001e-06, "loss": 1.3557, "step": 6293000 }, { "epoch": 32.59, "learning_rate": 5.046428571428572e-06, "loss": 1.3551, "step": 6293500 }, { "epoch": 32.59, "learning_rate": 5.042857142857143e-06, "loss": 1.3585, "step": 6294000 }, { "epoch": 32.59, "learning_rate": 5.039285714285714e-06, "loss": 1.3532, "step": 6294500 }, { "epoch": 32.59, "learning_rate": 5.035714285714286e-06, "loss": 1.3548, "step": 6295000 }, { "epoch": 32.6, "learning_rate": 5.032142857142858e-06, "loss": 1.3574, "step": 6295500 }, { "epoch": 32.6, "learning_rate": 5.028571428571429e-06, "loss": 1.3557, "step": 6296000 }, { "epoch": 32.6, "learning_rate": 5.025e-06, "loss": 1.3486, "step": 6296500 }, { "epoch": 32.6, "learning_rate": 5.021428571428571e-06, "loss": 1.3596, "step": 6297000 }, { "epoch": 32.61, "learning_rate": 5.017857142857143e-06, "loss": 1.3536, "step": 6297500 }, { "epoch": 32.61, "learning_rate": 5.0142857142857144e-06, "loss": 1.3608, "step": 6298000 }, { "epoch": 32.61, "learning_rate": 5.010714285714286e-06, "loss": 1.3511, "step": 6298500 }, { "epoch": 32.61, "learning_rate": 5.007142857142858e-06, "loss": 1.3544, "step": 6299000 }, { "epoch": 32.62, "learning_rate": 5.003571428571429e-06, "loss": 1.3499, "step": 6299500 }, { "epoch": 32.62, "learning_rate": 5e-06, "loss": 1.3597, "step": 6300000 }, { "epoch": 32.62, "learning_rate": 4.996428571428571e-06, "loss": 1.3526, "step": 6300500 }, { "epoch": 32.63, "learning_rate": 4.992857142857143e-06, "loss": 1.3599, "step": 6301000 }, { "epoch": 32.63, "learning_rate": 4.989285714285715e-06, "loss": 1.3556, "step": 6301500 }, { "epoch": 32.63, "learning_rate": 4.9857142857142855e-06, "loss": 1.3605, "step": 6302000 }, { "epoch": 32.63, "learning_rate": 4.982142857142857e-06, "loss": 1.353, "step": 6302500 }, { "epoch": 32.64, "learning_rate": 4.978571428571429e-06, "loss": 1.3593, "step": 6303000 }, { "epoch": 32.64, "learning_rate": 4.975000000000001e-06, "loss": 1.3521, "step": 6303500 }, { "epoch": 32.64, "learning_rate": 4.9714285714285715e-06, "loss": 1.3576, "step": 6304000 }, { "epoch": 32.64, "learning_rate": 4.967857142857143e-06, "loss": 1.3583, "step": 6304500 }, { "epoch": 32.65, "learning_rate": 4.964285714285714e-06, "loss": 1.3507, "step": 6305000 }, { "epoch": 32.65, "learning_rate": 4.960714285714286e-06, "loss": 1.3601, "step": 6305500 }, { "epoch": 32.65, "learning_rate": 4.9571428571428575e-06, "loss": 1.3651, "step": 6306000 }, { "epoch": 32.65, "learning_rate": 4.953571428571429e-06, "loss": 1.3567, "step": 6306500 }, { "epoch": 32.66, "learning_rate": 4.950000000000001e-06, "loss": 1.3542, "step": 6307000 }, { "epoch": 32.66, "learning_rate": 4.946428571428572e-06, "loss": 1.3477, "step": 6307500 }, { "epoch": 32.66, "learning_rate": 4.942857142857143e-06, "loss": 1.3575, "step": 6308000 }, { "epoch": 32.67, "learning_rate": 4.939285714285714e-06, "loss": 1.356, "step": 6308500 }, { "epoch": 32.67, "learning_rate": 4.935714285714286e-06, "loss": 1.3543, "step": 6309000 }, { "epoch": 32.67, "learning_rate": 4.932142857142858e-06, "loss": 1.3705, "step": 6309500 }, { "epoch": 32.67, "learning_rate": 4.9285714285714286e-06, "loss": 1.3601, "step": 6310000 }, { "epoch": 32.68, "learning_rate": 4.925e-06, "loss": 1.361, "step": 6310500 }, { "epoch": 32.68, "learning_rate": 4.921428571428572e-06, "loss": 1.358, "step": 6311000 }, { "epoch": 32.68, "learning_rate": 4.917857142857143e-06, "loss": 1.3435, "step": 6311500 }, { "epoch": 32.68, "learning_rate": 4.9142857142857145e-06, "loss": 1.3599, "step": 6312000 }, { "epoch": 32.69, "learning_rate": 4.910714285714286e-06, "loss": 1.3578, "step": 6312500 }, { "epoch": 32.69, "learning_rate": 4.907142857142857e-06, "loss": 1.3591, "step": 6313000 }, { "epoch": 32.69, "learning_rate": 4.903571428571429e-06, "loss": 1.349, "step": 6313500 }, { "epoch": 32.69, "learning_rate": 4.9000000000000005e-06, "loss": 1.3551, "step": 6314000 }, { "epoch": 32.7, "learning_rate": 4.896428571428572e-06, "loss": 1.3548, "step": 6314500 }, { "epoch": 32.7, "learning_rate": 4.892857142857143e-06, "loss": 1.356, "step": 6315000 }, { "epoch": 32.7, "learning_rate": 4.889285714285714e-06, "loss": 1.3496, "step": 6315500 }, { "epoch": 32.71, "learning_rate": 4.885714285714286e-06, "loss": 1.3513, "step": 6316000 }, { "epoch": 32.71, "learning_rate": 4.882142857142857e-06, "loss": 1.3637, "step": 6316500 }, { "epoch": 32.71, "learning_rate": 4.878571428571429e-06, "loss": 1.3574, "step": 6317000 }, { "epoch": 32.71, "learning_rate": 4.875000000000001e-06, "loss": 1.3649, "step": 6317500 }, { "epoch": 32.72, "learning_rate": 4.871428571428572e-06, "loss": 1.3613, "step": 6318000 }, { "epoch": 32.72, "learning_rate": 4.8678571428571425e-06, "loss": 1.3537, "step": 6318500 }, { "epoch": 32.72, "learning_rate": 4.864285714285714e-06, "loss": 1.3433, "step": 6319000 }, { "epoch": 32.72, "learning_rate": 4.860714285714286e-06, "loss": 1.3598, "step": 6319500 }, { "epoch": 32.73, "learning_rate": 4.857142857142858e-06, "loss": 1.3466, "step": 6320000 }, { "epoch": 32.73, "learning_rate": 4.8535714285714284e-06, "loss": 1.3581, "step": 6320500 }, { "epoch": 32.73, "learning_rate": 4.85e-06, "loss": 1.3598, "step": 6321000 }, { "epoch": 32.73, "learning_rate": 4.846428571428572e-06, "loss": 1.3629, "step": 6321500 }, { "epoch": 32.74, "learning_rate": 4.8428571428571436e-06, "loss": 1.3579, "step": 6322000 }, { "epoch": 32.74, "learning_rate": 4.839285714285714e-06, "loss": 1.3566, "step": 6322500 }, { "epoch": 32.74, "learning_rate": 4.835714285714286e-06, "loss": 1.3497, "step": 6323000 }, { "epoch": 32.74, "learning_rate": 4.832142857142857e-06, "loss": 1.3559, "step": 6323500 }, { "epoch": 32.75, "learning_rate": 4.828571428571429e-06, "loss": 1.351, "step": 6324000 }, { "epoch": 32.75, "learning_rate": 4.825e-06, "loss": 1.3627, "step": 6324500 }, { "epoch": 32.75, "learning_rate": 4.821428571428572e-06, "loss": 1.361, "step": 6325000 }, { "epoch": 32.76, "learning_rate": 4.817857142857144e-06, "loss": 1.3538, "step": 6325500 }, { "epoch": 32.76, "learning_rate": 4.814285714285714e-06, "loss": 1.353, "step": 6326000 }, { "epoch": 32.76, "learning_rate": 4.8107142857142855e-06, "loss": 1.3601, "step": 6326500 }, { "epoch": 32.76, "learning_rate": 4.807142857142857e-06, "loss": 1.3548, "step": 6327000 }, { "epoch": 32.77, "learning_rate": 4.803571428571429e-06, "loss": 1.3586, "step": 6327500 }, { "epoch": 32.77, "learning_rate": 4.800000000000001e-06, "loss": 1.3641, "step": 6328000 }, { "epoch": 32.77, "learning_rate": 4.7964285714285715e-06, "loss": 1.3516, "step": 6328500 }, { "epoch": 32.77, "learning_rate": 4.792857142857143e-06, "loss": 1.3541, "step": 6329000 }, { "epoch": 32.78, "learning_rate": 4.789285714285714e-06, "loss": 1.3588, "step": 6329500 }, { "epoch": 32.78, "learning_rate": 4.785714285714286e-06, "loss": 1.3653, "step": 6330000 }, { "epoch": 32.78, "learning_rate": 4.7821428571428575e-06, "loss": 1.3545, "step": 6330500 }, { "epoch": 32.78, "learning_rate": 4.778571428571429e-06, "loss": 1.3462, "step": 6331000 }, { "epoch": 32.79, "learning_rate": 4.775e-06, "loss": 1.3656, "step": 6331500 }, { "epoch": 32.79, "learning_rate": 4.771428571428572e-06, "loss": 1.3576, "step": 6332000 }, { "epoch": 32.79, "learning_rate": 4.7678571428571434e-06, "loss": 1.3536, "step": 6332500 }, { "epoch": 32.8, "learning_rate": 4.764285714285714e-06, "loss": 1.3519, "step": 6333000 }, { "epoch": 32.8, "learning_rate": 4.760714285714286e-06, "loss": 1.3589, "step": 6333500 }, { "epoch": 32.8, "learning_rate": 4.757142857142857e-06, "loss": 1.3592, "step": 6334000 }, { "epoch": 32.8, "learning_rate": 4.7535714285714286e-06, "loss": 1.3529, "step": 6334500 }, { "epoch": 32.81, "learning_rate": 4.75e-06, "loss": 1.3509, "step": 6335000 }, { "epoch": 32.81, "learning_rate": 4.746428571428572e-06, "loss": 1.3582, "step": 6335500 }, { "epoch": 32.81, "learning_rate": 4.742857142857144e-06, "loss": 1.3496, "step": 6336000 }, { "epoch": 32.81, "learning_rate": 4.7392857142857145e-06, "loss": 1.3634, "step": 6336500 }, { "epoch": 32.82, "learning_rate": 4.735714285714285e-06, "loss": 1.3565, "step": 6337000 }, { "epoch": 32.82, "learning_rate": 4.732142857142857e-06, "loss": 1.3562, "step": 6337500 }, { "epoch": 32.82, "learning_rate": 4.728571428571429e-06, "loss": 1.3527, "step": 6338000 }, { "epoch": 32.82, "learning_rate": 4.7250000000000005e-06, "loss": 1.3669, "step": 6338500 }, { "epoch": 32.83, "learning_rate": 4.721428571428572e-06, "loss": 1.3617, "step": 6339000 }, { "epoch": 32.83, "learning_rate": 4.717857142857143e-06, "loss": 1.3621, "step": 6339500 }, { "epoch": 32.83, "learning_rate": 4.714285714285715e-06, "loss": 1.3575, "step": 6340000 }, { "epoch": 32.84, "learning_rate": 4.710714285714286e-06, "loss": 1.3625, "step": 6340500 }, { "epoch": 32.84, "learning_rate": 4.707142857142857e-06, "loss": 1.3571, "step": 6341000 }, { "epoch": 32.84, "learning_rate": 4.703571428571429e-06, "loss": 1.351, "step": 6341500 }, { "epoch": 32.84, "learning_rate": 4.7e-06, "loss": 1.3519, "step": 6342000 }, { "epoch": 32.85, "learning_rate": 4.696428571428572e-06, "loss": 1.3555, "step": 6342500 }, { "epoch": 32.85, "learning_rate": 4.692857142857143e-06, "loss": 1.3512, "step": 6343000 }, { "epoch": 32.85, "learning_rate": 4.689285714285715e-06, "loss": 1.3582, "step": 6343500 }, { "epoch": 32.85, "learning_rate": 4.685714285714286e-06, "loss": 1.3517, "step": 6344000 }, { "epoch": 32.86, "learning_rate": 4.6821428571428576e-06, "loss": 1.3599, "step": 6344500 }, { "epoch": 32.86, "learning_rate": 4.6785714285714284e-06, "loss": 1.3501, "step": 6345000 }, { "epoch": 32.86, "learning_rate": 4.675e-06, "loss": 1.3506, "step": 6345500 }, { "epoch": 32.86, "learning_rate": 4.671428571428572e-06, "loss": 1.3596, "step": 6346000 }, { "epoch": 32.87, "learning_rate": 4.6678571428571435e-06, "loss": 1.3592, "step": 6346500 }, { "epoch": 32.87, "learning_rate": 4.664285714285714e-06, "loss": 1.3542, "step": 6347000 }, { "epoch": 32.87, "learning_rate": 4.660714285714285e-06, "loss": 1.3434, "step": 6347500 }, { "epoch": 32.88, "learning_rate": 4.657142857142857e-06, "loss": 1.3587, "step": 6348000 }, { "epoch": 32.88, "learning_rate": 4.653571428571429e-06, "loss": 1.3588, "step": 6348500 }, { "epoch": 32.88, "learning_rate": 4.65e-06, "loss": 1.3618, "step": 6349000 }, { "epoch": 32.88, "learning_rate": 4.646428571428572e-06, "loss": 1.3553, "step": 6349500 }, { "epoch": 32.89, "learning_rate": 4.642857142857143e-06, "loss": 1.354, "step": 6350000 }, { "epoch": 32.89, "learning_rate": 4.639285714285715e-06, "loss": 1.3556, "step": 6350500 }, { "epoch": 32.89, "learning_rate": 4.6357142857142855e-06, "loss": 1.3477, "step": 6351000 }, { "epoch": 32.89, "learning_rate": 4.632142857142857e-06, "loss": 1.3521, "step": 6351500 }, { "epoch": 32.9, "learning_rate": 4.628571428571429e-06, "loss": 1.3627, "step": 6352000 }, { "epoch": 32.9, "learning_rate": 4.625e-06, "loss": 1.3562, "step": 6352500 }, { "epoch": 32.9, "learning_rate": 4.6214285714285715e-06, "loss": 1.3485, "step": 6353000 }, { "epoch": 32.9, "learning_rate": 4.617857142857143e-06, "loss": 1.3539, "step": 6353500 }, { "epoch": 32.91, "learning_rate": 4.614285714285715e-06, "loss": 1.3575, "step": 6354000 }, { "epoch": 32.91, "learning_rate": 4.610714285714287e-06, "loss": 1.3445, "step": 6354500 }, { "epoch": 32.91, "learning_rate": 4.6071428571428574e-06, "loss": 1.3559, "step": 6355000 }, { "epoch": 32.92, "learning_rate": 4.603571428571428e-06, "loss": 1.3643, "step": 6355500 }, { "epoch": 32.92, "learning_rate": 4.6e-06, "loss": 1.3495, "step": 6356000 }, { "epoch": 32.92, "learning_rate": 4.596428571428572e-06, "loss": 1.3587, "step": 6356500 }, { "epoch": 32.92, "learning_rate": 4.592857142857143e-06, "loss": 1.3558, "step": 6357000 }, { "epoch": 32.93, "learning_rate": 4.589285714285715e-06, "loss": 1.3512, "step": 6357500 }, { "epoch": 32.93, "learning_rate": 4.585714285714286e-06, "loss": 1.3534, "step": 6358000 }, { "epoch": 32.93, "learning_rate": 4.582142857142857e-06, "loss": 1.3526, "step": 6358500 }, { "epoch": 32.93, "learning_rate": 4.5785714285714285e-06, "loss": 1.35, "step": 6359000 }, { "epoch": 32.94, "learning_rate": 4.575e-06, "loss": 1.3499, "step": 6359500 }, { "epoch": 32.94, "learning_rate": 4.571428571428572e-06, "loss": 1.3456, "step": 6360000 }, { "epoch": 32.94, "learning_rate": 4.567857142857143e-06, "loss": 1.3518, "step": 6360500 }, { "epoch": 32.94, "learning_rate": 4.5642857142857145e-06, "loss": 1.3638, "step": 6361000 }, { "epoch": 32.95, "learning_rate": 4.560714285714286e-06, "loss": 1.3576, "step": 6361500 }, { "epoch": 32.95, "learning_rate": 4.557142857142857e-06, "loss": 1.3516, "step": 6362000 }, { "epoch": 32.95, "learning_rate": 4.553571428571429e-06, "loss": 1.349, "step": 6362500 }, { "epoch": 32.96, "learning_rate": 4.5500000000000005e-06, "loss": 1.3647, "step": 6363000 }, { "epoch": 32.96, "learning_rate": 4.546428571428571e-06, "loss": 1.3587, "step": 6363500 }, { "epoch": 32.96, "learning_rate": 4.542857142857143e-06, "loss": 1.3484, "step": 6364000 }, { "epoch": 32.96, "learning_rate": 4.539285714285715e-06, "loss": 1.3569, "step": 6364500 }, { "epoch": 32.97, "learning_rate": 4.5357142857142865e-06, "loss": 1.3664, "step": 6365000 }, { "epoch": 32.97, "learning_rate": 4.532142857142857e-06, "loss": 1.359, "step": 6365500 }, { "epoch": 32.97, "learning_rate": 4.528571428571428e-06, "loss": 1.3572, "step": 6366000 }, { "epoch": 32.97, "learning_rate": 4.525e-06, "loss": 1.3527, "step": 6366500 }, { "epoch": 32.98, "learning_rate": 4.521428571428572e-06, "loss": 1.3598, "step": 6367000 }, { "epoch": 32.98, "learning_rate": 4.517857142857143e-06, "loss": 1.3485, "step": 6367500 }, { "epoch": 32.98, "learning_rate": 4.514285714285715e-06, "loss": 1.3519, "step": 6368000 }, { "epoch": 32.98, "learning_rate": 4.510714285714286e-06, "loss": 1.357, "step": 6368500 }, { "epoch": 32.99, "learning_rate": 4.5071428571428576e-06, "loss": 1.3599, "step": 6369000 }, { "epoch": 32.99, "learning_rate": 4.503571428571428e-06, "loss": 1.348, "step": 6369500 }, { "epoch": 32.99, "learning_rate": 4.5e-06, "loss": 1.3491, "step": 6370000 }, { "epoch": 33.0, "learning_rate": 4.496428571428572e-06, "loss": 1.3492, "step": 6370500 }, { "epoch": 33.0, "learning_rate": 4.492857142857143e-06, "loss": 1.3515, "step": 6371000 }, { "epoch": 33.0, "learning_rate": 4.489285714285714e-06, "loss": 1.3444, "step": 6371500 }, { "epoch": 33.0, "learning_rate": 4.485714285714286e-06, "loss": 1.341, "step": 6372000 }, { "epoch": 33.01, "learning_rate": 4.482142857142858e-06, "loss": 1.3545, "step": 6372500 }, { "epoch": 33.01, "learning_rate": 4.478571428571429e-06, "loss": 1.354, "step": 6373000 }, { "epoch": 33.01, "learning_rate": 4.475e-06, "loss": 1.3519, "step": 6373500 }, { "epoch": 33.01, "learning_rate": 4.471428571428571e-06, "loss": 1.3497, "step": 6374000 }, { "epoch": 33.02, "learning_rate": 4.467857142857143e-06, "loss": 1.3582, "step": 6374500 }, { "epoch": 33.02, "learning_rate": 4.464285714285715e-06, "loss": 1.3543, "step": 6375000 }, { "epoch": 33.02, "learning_rate": 4.460714285714286e-06, "loss": 1.3598, "step": 6375500 }, { "epoch": 33.02, "learning_rate": 4.457142857142858e-06, "loss": 1.3529, "step": 6376000 }, { "epoch": 33.03, "learning_rate": 4.453571428571429e-06, "loss": 1.3579, "step": 6376500 }, { "epoch": 33.03, "learning_rate": 4.45e-06, "loss": 1.3546, "step": 6377000 }, { "epoch": 33.03, "learning_rate": 4.4464285714285715e-06, "loss": 1.3493, "step": 6377500 }, { "epoch": 33.04, "learning_rate": 4.442857142857143e-06, "loss": 1.3532, "step": 6378000 }, { "epoch": 33.04, "learning_rate": 4.439285714285715e-06, "loss": 1.36, "step": 6378500 }, { "epoch": 33.04, "learning_rate": 4.435714285714286e-06, "loss": 1.3496, "step": 6379000 }, { "epoch": 33.04, "learning_rate": 4.432142857142857e-06, "loss": 1.3483, "step": 6379500 }, { "epoch": 33.05, "learning_rate": 4.428571428571428e-06, "loss": 1.3583, "step": 6380000 }, { "epoch": 33.05, "learning_rate": 4.425e-06, "loss": 1.3608, "step": 6380500 }, { "epoch": 33.05, "learning_rate": 4.421428571428572e-06, "loss": 1.35, "step": 6381000 }, { "epoch": 33.05, "learning_rate": 4.417857142857143e-06, "loss": 1.3455, "step": 6381500 }, { "epoch": 33.06, "learning_rate": 4.414285714285714e-06, "loss": 1.3611, "step": 6382000 }, { "epoch": 33.06, "learning_rate": 4.410714285714286e-06, "loss": 1.3514, "step": 6382500 }, { "epoch": 33.06, "learning_rate": 4.407142857142858e-06, "loss": 1.3518, "step": 6383000 }, { "epoch": 33.06, "learning_rate": 4.403571428571429e-06, "loss": 1.3469, "step": 6383500 }, { "epoch": 33.07, "learning_rate": 4.4e-06, "loss": 1.3495, "step": 6384000 }, { "epoch": 33.07, "learning_rate": 4.396428571428571e-06, "loss": 1.3507, "step": 6384500 }, { "epoch": 33.07, "learning_rate": 4.392857142857143e-06, "loss": 1.3502, "step": 6385000 }, { "epoch": 33.08, "learning_rate": 4.3892857142857145e-06, "loss": 1.3552, "step": 6385500 }, { "epoch": 33.08, "learning_rate": 4.385714285714286e-06, "loss": 1.3503, "step": 6386000 }, { "epoch": 33.08, "learning_rate": 4.382142857142858e-06, "loss": 1.3508, "step": 6386500 }, { "epoch": 33.08, "learning_rate": 4.378571428571429e-06, "loss": 1.3597, "step": 6387000 }, { "epoch": 33.09, "learning_rate": 4.375e-06, "loss": 1.3538, "step": 6387500 }, { "epoch": 33.09, "learning_rate": 4.371428571428571e-06, "loss": 1.3669, "step": 6388000 }, { "epoch": 33.09, "learning_rate": 4.367857142857143e-06, "loss": 1.3505, "step": 6388500 }, { "epoch": 33.09, "learning_rate": 4.364285714285715e-06, "loss": 1.3572, "step": 6389000 }, { "epoch": 33.1, "learning_rate": 4.3607142857142864e-06, "loss": 1.3574, "step": 6389500 }, { "epoch": 33.1, "learning_rate": 4.357142857142857e-06, "loss": 1.353, "step": 6390000 }, { "epoch": 33.1, "learning_rate": 4.353571428571429e-06, "loss": 1.3579, "step": 6390500 }, { "epoch": 33.1, "learning_rate": 4.35e-06, "loss": 1.3471, "step": 6391000 }, { "epoch": 33.11, "learning_rate": 4.3464285714285716e-06, "loss": 1.35, "step": 6391500 }, { "epoch": 33.11, "learning_rate": 4.342857142857143e-06, "loss": 1.3602, "step": 6392000 }, { "epoch": 33.11, "learning_rate": 4.339285714285714e-06, "loss": 1.3529, "step": 6392500 }, { "epoch": 33.12, "learning_rate": 4.335714285714286e-06, "loss": 1.3482, "step": 6393000 }, { "epoch": 33.12, "learning_rate": 4.3321428571428575e-06, "loss": 1.3497, "step": 6393500 }, { "epoch": 33.12, "learning_rate": 4.328571428571429e-06, "loss": 1.3464, "step": 6394000 }, { "epoch": 33.12, "learning_rate": 4.325e-06, "loss": 1.3464, "step": 6394500 }, { "epoch": 33.13, "learning_rate": 4.321428571428572e-06, "loss": 1.3482, "step": 6395000 }, { "epoch": 33.13, "learning_rate": 4.317857142857143e-06, "loss": 1.3497, "step": 6395500 }, { "epoch": 33.13, "learning_rate": 4.314285714285714e-06, "loss": 1.3488, "step": 6396000 }, { "epoch": 33.13, "learning_rate": 4.310714285714286e-06, "loss": 1.3439, "step": 6396500 }, { "epoch": 33.14, "learning_rate": 4.307142857142858e-06, "loss": 1.3435, "step": 6397000 }, { "epoch": 33.14, "learning_rate": 4.303571428571429e-06, "loss": 1.351, "step": 6397500 }, { "epoch": 33.14, "learning_rate": 4.2999999999999995e-06, "loss": 1.3508, "step": 6398000 }, { "epoch": 33.14, "learning_rate": 4.296428571428571e-06, "loss": 1.3573, "step": 6398500 }, { "epoch": 33.15, "learning_rate": 4.292857142857143e-06, "loss": 1.365, "step": 6399000 }, { "epoch": 33.15, "learning_rate": 4.289285714285715e-06, "loss": 1.3497, "step": 6399500 }, { "epoch": 33.15, "learning_rate": 4.285714285714286e-06, "loss": 1.3393, "step": 6400000 }, { "epoch": 33.15, "learning_rate": 4.282142857142857e-06, "loss": 1.3482, "step": 6400500 }, { "epoch": 33.16, "learning_rate": 4.278571428571429e-06, "loss": 1.3471, "step": 6401000 }, { "epoch": 33.16, "learning_rate": 4.2750000000000006e-06, "loss": 1.3507, "step": 6401500 }, { "epoch": 33.16, "learning_rate": 4.2714285714285714e-06, "loss": 1.354, "step": 6402000 }, { "epoch": 33.17, "learning_rate": 4.267857142857143e-06, "loss": 1.3436, "step": 6402500 }, { "epoch": 33.17, "learning_rate": 4.264285714285714e-06, "loss": 1.3486, "step": 6403000 }, { "epoch": 33.17, "learning_rate": 4.260714285714286e-06, "loss": 1.3608, "step": 6403500 }, { "epoch": 33.17, "learning_rate": 4.257142857142857e-06, "loss": 1.3474, "step": 6404000 }, { "epoch": 33.18, "learning_rate": 4.253571428571429e-06, "loss": 1.3599, "step": 6404500 }, { "epoch": 33.18, "learning_rate": 4.250000000000001e-06, "loss": 1.3418, "step": 6405000 }, { "epoch": 33.18, "learning_rate": 4.246428571428572e-06, "loss": 1.3516, "step": 6405500 }, { "epoch": 33.18, "learning_rate": 4.2428571428571425e-06, "loss": 1.361, "step": 6406000 }, { "epoch": 33.19, "learning_rate": 4.239285714285714e-06, "loss": 1.353, "step": 6406500 }, { "epoch": 33.19, "learning_rate": 4.235714285714286e-06, "loss": 1.3506, "step": 6407000 }, { "epoch": 33.19, "learning_rate": 4.232142857142858e-06, "loss": 1.357, "step": 6407500 }, { "epoch": 33.19, "learning_rate": 4.228571428571429e-06, "loss": 1.3477, "step": 6408000 }, { "epoch": 33.2, "learning_rate": 4.225e-06, "loss": 1.344, "step": 6408500 }, { "epoch": 33.2, "learning_rate": 4.221428571428571e-06, "loss": 1.3567, "step": 6409000 }, { "epoch": 33.2, "learning_rate": 4.217857142857143e-06, "loss": 1.3422, "step": 6409500 }, { "epoch": 33.21, "learning_rate": 4.2142857142857145e-06, "loss": 1.3319, "step": 6410000 }, { "epoch": 33.21, "learning_rate": 4.210714285714286e-06, "loss": 1.3486, "step": 6410500 }, { "epoch": 33.21, "learning_rate": 4.207142857142857e-06, "loss": 1.3346, "step": 6411000 }, { "epoch": 33.21, "learning_rate": 4.203571428571429e-06, "loss": 1.3463, "step": 6411500 }, { "epoch": 33.22, "learning_rate": 4.2000000000000004e-06, "loss": 1.3589, "step": 6412000 }, { "epoch": 33.22, "learning_rate": 4.196428571428571e-06, "loss": 1.3503, "step": 6412500 }, { "epoch": 33.22, "learning_rate": 4.192857142857143e-06, "loss": 1.3496, "step": 6413000 }, { "epoch": 33.22, "learning_rate": 4.189285714285715e-06, "loss": 1.3472, "step": 6413500 }, { "epoch": 33.23, "learning_rate": 4.1857142857142856e-06, "loss": 1.357, "step": 6414000 }, { "epoch": 33.23, "learning_rate": 4.182142857142857e-06, "loss": 1.3452, "step": 6414500 }, { "epoch": 33.23, "learning_rate": 4.178571428571429e-06, "loss": 1.3612, "step": 6415000 }, { "epoch": 33.23, "learning_rate": 4.175000000000001e-06, "loss": 1.3407, "step": 6415500 }, { "epoch": 33.24, "learning_rate": 4.1714285714285715e-06, "loss": 1.3444, "step": 6416000 }, { "epoch": 33.24, "learning_rate": 4.167857142857142e-06, "loss": 1.364, "step": 6416500 }, { "epoch": 33.24, "learning_rate": 4.164285714285714e-06, "loss": 1.3527, "step": 6417000 }, { "epoch": 33.25, "learning_rate": 4.160714285714286e-06, "loss": 1.3531, "step": 6417500 }, { "epoch": 33.25, "learning_rate": 4.1571428571428575e-06, "loss": 1.3515, "step": 6418000 }, { "epoch": 33.25, "learning_rate": 4.153571428571429e-06, "loss": 1.344, "step": 6418500 }, { "epoch": 33.25, "learning_rate": 4.15e-06, "loss": 1.3468, "step": 6419000 }, { "epoch": 33.26, "learning_rate": 4.146428571428572e-06, "loss": 1.3461, "step": 6419500 }, { "epoch": 33.26, "learning_rate": 4.142857142857143e-06, "loss": 1.3559, "step": 6420000 }, { "epoch": 33.26, "learning_rate": 4.139285714285714e-06, "loss": 1.355, "step": 6420500 }, { "epoch": 33.26, "learning_rate": 4.135714285714286e-06, "loss": 1.3559, "step": 6421000 }, { "epoch": 33.27, "learning_rate": 4.132142857142858e-06, "loss": 1.3513, "step": 6421500 }, { "epoch": 33.27, "learning_rate": 4.128571428571429e-06, "loss": 1.3454, "step": 6422000 }, { "epoch": 33.27, "learning_rate": 4.125e-06, "loss": 1.3404, "step": 6422500 }, { "epoch": 33.27, "learning_rate": 4.121428571428572e-06, "loss": 1.3396, "step": 6423000 }, { "epoch": 33.28, "learning_rate": 4.117857142857143e-06, "loss": 1.359, "step": 6423500 }, { "epoch": 33.28, "learning_rate": 4.114285714285715e-06, "loss": 1.3385, "step": 6424000 }, { "epoch": 33.28, "learning_rate": 4.1107142857142854e-06, "loss": 1.3505, "step": 6424500 }, { "epoch": 33.29, "learning_rate": 4.107142857142857e-06, "loss": 1.3397, "step": 6425000 }, { "epoch": 33.29, "learning_rate": 4.103571428571429e-06, "loss": 1.3493, "step": 6425500 }, { "epoch": 33.29, "learning_rate": 4.1000000000000006e-06, "loss": 1.3545, "step": 6426000 }, { "epoch": 33.29, "learning_rate": 4.096428571428572e-06, "loss": 1.3488, "step": 6426500 }, { "epoch": 33.3, "learning_rate": 4.092857142857143e-06, "loss": 1.3487, "step": 6427000 }, { "epoch": 33.3, "learning_rate": 4.089285714285714e-06, "loss": 1.3484, "step": 6427500 }, { "epoch": 33.3, "learning_rate": 4.085714285714286e-06, "loss": 1.3543, "step": 6428000 }, { "epoch": 33.3, "learning_rate": 4.082142857142857e-06, "loss": 1.3396, "step": 6428500 }, { "epoch": 33.31, "learning_rate": 4.078571428571429e-06, "loss": 1.3455, "step": 6429000 }, { "epoch": 33.31, "learning_rate": 4.075e-06, "loss": 1.3412, "step": 6429500 }, { "epoch": 33.31, "learning_rate": 4.071428571428572e-06, "loss": 1.3508, "step": 6430000 }, { "epoch": 33.31, "learning_rate": 4.067857142857143e-06, "loss": 1.3456, "step": 6430500 }, { "epoch": 33.32, "learning_rate": 4.064285714285714e-06, "loss": 1.3596, "step": 6431000 }, { "epoch": 33.32, "learning_rate": 4.060714285714286e-06, "loss": 1.3544, "step": 6431500 }, { "epoch": 33.32, "learning_rate": 4.057142857142858e-06, "loss": 1.3562, "step": 6432000 }, { "epoch": 33.33, "learning_rate": 4.0535714285714285e-06, "loss": 1.3482, "step": 6432500 }, { "epoch": 33.33, "learning_rate": 4.05e-06, "loss": 1.3463, "step": 6433000 }, { "epoch": 33.33, "learning_rate": 4.046428571428572e-06, "loss": 1.3506, "step": 6433500 }, { "epoch": 33.33, "learning_rate": 4.042857142857144e-06, "loss": 1.3533, "step": 6434000 }, { "epoch": 33.34, "learning_rate": 4.0392857142857145e-06, "loss": 1.3526, "step": 6434500 }, { "epoch": 33.34, "learning_rate": 4.035714285714285e-06, "loss": 1.3512, "step": 6435000 }, { "epoch": 33.34, "learning_rate": 4.032142857142857e-06, "loss": 1.349, "step": 6435500 }, { "epoch": 33.34, "learning_rate": 4.028571428571429e-06, "loss": 1.347, "step": 6436000 }, { "epoch": 33.35, "learning_rate": 4.0250000000000004e-06, "loss": 1.3439, "step": 6436500 }, { "epoch": 33.35, "learning_rate": 4.021428571428572e-06, "loss": 1.355, "step": 6437000 }, { "epoch": 33.35, "learning_rate": 4.017857142857143e-06, "loss": 1.3421, "step": 6437500 }, { "epoch": 33.35, "learning_rate": 4.014285714285714e-06, "loss": 1.3501, "step": 6438000 }, { "epoch": 33.36, "learning_rate": 4.0107142857142856e-06, "loss": 1.3472, "step": 6438500 }, { "epoch": 33.36, "learning_rate": 4.007142857142857e-06, "loss": 1.341, "step": 6439000 }, { "epoch": 33.36, "learning_rate": 4.003571428571429e-06, "loss": 1.3462, "step": 6439500 }, { "epoch": 33.37, "learning_rate": 4.000000000000001e-06, "loss": 1.3542, "step": 6440000 }, { "epoch": 33.37, "learning_rate": 3.9964285714285715e-06, "loss": 1.3459, "step": 6440500 }, { "epoch": 33.37, "learning_rate": 3.992857142857143e-06, "loss": 1.3484, "step": 6441000 }, { "epoch": 33.37, "learning_rate": 3.989285714285714e-06, "loss": 1.3465, "step": 6441500 } ], "max_steps": 7000000, "num_train_epochs": 38, "total_flos": 1.2194298841987817e+19, "trial_name": null, "trial_params": null }