| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 80.0, | |
| "global_step": 2000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.62, | |
| "eval_loss": 0.5568868517875671, | |
| "eval_runtime": 2.9587, | |
| "eval_samples_per_second": 33.798, | |
| "eval_steps_per_second": 4.394, | |
| "step": 25 | |
| }, | |
| { | |
| "best_epoch": 0, | |
| "best_eval_accuracy": 0.62, | |
| "epoch": 1.0, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.4, | |
| "eval_loss": 0.3271646797657013, | |
| "eval_runtime": 3.0122, | |
| "eval_samples_per_second": 33.199, | |
| "eval_steps_per_second": 4.316, | |
| "step": 50 | |
| }, | |
| { | |
| "best_epoch": 0, | |
| "best_eval_accuracy": 0.62, | |
| "epoch": 2.0, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.48, | |
| "eval_loss": 0.29993605613708496, | |
| "eval_runtime": 3.0698, | |
| "eval_samples_per_second": 32.575, | |
| "eval_steps_per_second": 4.235, | |
| "step": 75 | |
| }, | |
| { | |
| "best_epoch": 0, | |
| "best_eval_accuracy": 0.62, | |
| "epoch": 3.0, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.58, | |
| "eval_loss": 0.3037445843219757, | |
| "eval_runtime": 3.1074, | |
| "eval_samples_per_second": 32.181, | |
| "eval_steps_per_second": 4.184, | |
| "step": 100 | |
| }, | |
| { | |
| "best_epoch": 0, | |
| "best_eval_accuracy": 0.62, | |
| "epoch": 4.0, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.39, | |
| "eval_loss": 0.30919986963272095, | |
| "eval_runtime": 3.1329, | |
| "eval_samples_per_second": 31.919, | |
| "eval_steps_per_second": 4.15, | |
| "step": 125 | |
| }, | |
| { | |
| "best_epoch": 0, | |
| "best_eval_accuracy": 0.62, | |
| "epoch": 5.0, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.37, | |
| "eval_loss": 0.3146528899669647, | |
| "eval_runtime": 3.1332, | |
| "eval_samples_per_second": 31.916, | |
| "eval_steps_per_second": 4.149, | |
| "step": 150 | |
| }, | |
| { | |
| "best_epoch": 0, | |
| "best_eval_accuracy": 0.62, | |
| "epoch": 6.0, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.61, | |
| "eval_loss": 0.28715208172798157, | |
| "eval_runtime": 3.142, | |
| "eval_samples_per_second": 31.827, | |
| "eval_steps_per_second": 4.138, | |
| "step": 175 | |
| }, | |
| { | |
| "best_epoch": 0, | |
| "best_eval_accuracy": 0.62, | |
| "epoch": 7.0, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.68, | |
| "eval_loss": 0.289668470621109, | |
| "eval_runtime": 3.1475, | |
| "eval_samples_per_second": 31.771, | |
| "eval_steps_per_second": 4.13, | |
| "step": 200 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 8.0, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.41, | |
| "eval_loss": 0.2950332760810852, | |
| "eval_runtime": 3.1537, | |
| "eval_samples_per_second": 31.709, | |
| "eval_steps_per_second": 4.122, | |
| "step": 225 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 9.0, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_accuracy": 0.63, | |
| "eval_loss": 0.27791741490364075, | |
| "eval_runtime": 3.1543, | |
| "eval_samples_per_second": 31.703, | |
| "eval_steps_per_second": 4.121, | |
| "step": 250 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 10.0, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_accuracy": 0.41, | |
| "eval_loss": 0.2977082133293152, | |
| "eval_runtime": 3.1598, | |
| "eval_samples_per_second": 31.647, | |
| "eval_steps_per_second": 4.114, | |
| "step": 275 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 11.0, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "eval_accuracy": 0.59, | |
| "eval_loss": 0.2908729612827301, | |
| "eval_runtime": 3.1621, | |
| "eval_samples_per_second": 31.624, | |
| "eval_steps_per_second": 4.111, | |
| "step": 300 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 12.0, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "eval_accuracy": 0.49, | |
| "eval_loss": 0.2939791679382324, | |
| "eval_runtime": 3.1669, | |
| "eval_samples_per_second": 31.577, | |
| "eval_steps_per_second": 4.105, | |
| "step": 325 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 13.0, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "eval_accuracy": 0.49, | |
| "eval_loss": 0.2929081618785858, | |
| "eval_runtime": 3.1666, | |
| "eval_samples_per_second": 31.579, | |
| "eval_steps_per_second": 4.105, | |
| "step": 350 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 14.0, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "eval_accuracy": 0.49, | |
| "eval_loss": 0.29478296637535095, | |
| "eval_runtime": 3.1574, | |
| "eval_samples_per_second": 31.671, | |
| "eval_steps_per_second": 4.117, | |
| "step": 375 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 15.0, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "eval_accuracy": 0.57, | |
| "eval_loss": 0.293470561504364, | |
| "eval_runtime": 3.1585, | |
| "eval_samples_per_second": 31.66, | |
| "eval_steps_per_second": 4.116, | |
| "step": 400 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 16.0, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 17.0, | |
| "eval_accuracy": 0.43, | |
| "eval_loss": 0.29490238428115845, | |
| "eval_runtime": 3.1586, | |
| "eval_samples_per_second": 31.659, | |
| "eval_steps_per_second": 4.116, | |
| "step": 425 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 17.0, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "eval_accuracy": 0.59, | |
| "eval_loss": 0.29253777861595154, | |
| "eval_runtime": 3.1606, | |
| "eval_samples_per_second": 31.64, | |
| "eval_steps_per_second": 4.113, | |
| "step": 450 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 18.0, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 19.0, | |
| "eval_accuracy": 0.57, | |
| "eval_loss": 0.2926887273788452, | |
| "eval_runtime": 3.1587, | |
| "eval_samples_per_second": 31.658, | |
| "eval_steps_per_second": 4.116, | |
| "step": 475 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 19.0, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "learning_rate": 0.037500000000000006, | |
| "loss": 1.2287, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "eval_accuracy": 0.58, | |
| "eval_loss": 0.29341670870780945, | |
| "eval_runtime": 3.162, | |
| "eval_samples_per_second": 31.625, | |
| "eval_steps_per_second": 4.111, | |
| "step": 500 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 20.0, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 21.0, | |
| "eval_accuracy": 0.44, | |
| "eval_loss": 0.2946934700012207, | |
| "eval_runtime": 3.164, | |
| "eval_samples_per_second": 31.605, | |
| "eval_steps_per_second": 4.109, | |
| "step": 525 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 21.0, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 22.0, | |
| "eval_accuracy": 0.6, | |
| "eval_loss": 0.29339364171028137, | |
| "eval_runtime": 3.1614, | |
| "eval_samples_per_second": 31.631, | |
| "eval_steps_per_second": 4.112, | |
| "step": 550 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 22.0, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 23.0, | |
| "eval_accuracy": 0.6, | |
| "eval_loss": 0.29298409819602966, | |
| "eval_runtime": 3.1646, | |
| "eval_samples_per_second": 31.6, | |
| "eval_steps_per_second": 4.108, | |
| "step": 575 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 23.0, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 24.0, | |
| "eval_accuracy": 0.4, | |
| "eval_loss": 0.2943931818008423, | |
| "eval_runtime": 3.1654, | |
| "eval_samples_per_second": 31.592, | |
| "eval_steps_per_second": 4.107, | |
| "step": 600 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 24.0, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 25.0, | |
| "eval_accuracy": 0.39, | |
| "eval_loss": 0.2969507575035095, | |
| "eval_runtime": 3.165, | |
| "eval_samples_per_second": 31.595, | |
| "eval_steps_per_second": 4.107, | |
| "step": 625 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 25.0, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 26.0, | |
| "eval_accuracy": 0.39, | |
| "eval_loss": 0.2949477434158325, | |
| "eval_runtime": 3.1641, | |
| "eval_samples_per_second": 31.605, | |
| "eval_steps_per_second": 4.109, | |
| "step": 650 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 26.0, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 27.0, | |
| "eval_accuracy": 0.43, | |
| "eval_loss": 0.29417097568511963, | |
| "eval_runtime": 3.1651, | |
| "eval_samples_per_second": 31.595, | |
| "eval_steps_per_second": 4.107, | |
| "step": 675 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 27.0, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 28.0, | |
| "eval_accuracy": 0.43, | |
| "eval_loss": 0.29403185844421387, | |
| "eval_runtime": 3.1625, | |
| "eval_samples_per_second": 31.62, | |
| "eval_steps_per_second": 4.111, | |
| "step": 700 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 28.0, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 29.0, | |
| "eval_accuracy": 0.58, | |
| "eval_loss": 0.29325199127197266, | |
| "eval_runtime": 3.164, | |
| "eval_samples_per_second": 31.606, | |
| "eval_steps_per_second": 4.109, | |
| "step": 725 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 29.0, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 30.0, | |
| "eval_accuracy": 0.62, | |
| "eval_loss": 0.29301509261131287, | |
| "eval_runtime": 3.1644, | |
| "eval_samples_per_second": 31.601, | |
| "eval_steps_per_second": 4.108, | |
| "step": 750 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 30.0, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 31.0, | |
| "eval_accuracy": 0.6, | |
| "eval_loss": 0.29338109493255615, | |
| "eval_runtime": 3.1656, | |
| "eval_samples_per_second": 31.589, | |
| "eval_steps_per_second": 4.107, | |
| "step": 775 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 31.0, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 32.0, | |
| "eval_accuracy": 0.57, | |
| "eval_loss": 0.29339686036109924, | |
| "eval_runtime": 3.1683, | |
| "eval_samples_per_second": 31.563, | |
| "eval_steps_per_second": 4.103, | |
| "step": 800 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 32.0, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 33.0, | |
| "eval_accuracy": 0.54, | |
| "eval_loss": 0.2932019829750061, | |
| "eval_runtime": 3.1687, | |
| "eval_samples_per_second": 31.558, | |
| "eval_steps_per_second": 4.103, | |
| "step": 825 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 33.0, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 34.0, | |
| "eval_accuracy": 0.54, | |
| "eval_loss": 0.292117714881897, | |
| "eval_runtime": 3.1676, | |
| "eval_samples_per_second": 31.57, | |
| "eval_steps_per_second": 4.104, | |
| "step": 850 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 34.0, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 35.0, | |
| "eval_accuracy": 0.44, | |
| "eval_loss": 0.29495733976364136, | |
| "eval_runtime": 3.1673, | |
| "eval_samples_per_second": 31.572, | |
| "eval_steps_per_second": 4.104, | |
| "step": 875 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 35.0, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 36.0, | |
| "eval_accuracy": 0.41, | |
| "eval_loss": 0.2943909764289856, | |
| "eval_runtime": 3.1671, | |
| "eval_samples_per_second": 31.575, | |
| "eval_steps_per_second": 4.105, | |
| "step": 900 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 36.0, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 37.0, | |
| "eval_accuracy": 0.43, | |
| "eval_loss": 0.2940847873687744, | |
| "eval_runtime": 3.1669, | |
| "eval_samples_per_second": 31.577, | |
| "eval_steps_per_second": 4.105, | |
| "step": 925 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 37.0, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 38.0, | |
| "eval_accuracy": 0.55, | |
| "eval_loss": 0.2930077910423279, | |
| "eval_runtime": 3.1669, | |
| "eval_samples_per_second": 31.577, | |
| "eval_steps_per_second": 4.105, | |
| "step": 950 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 38.0, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 39.0, | |
| "eval_accuracy": 0.57, | |
| "eval_loss": 0.2932360768318176, | |
| "eval_runtime": 3.1781, | |
| "eval_samples_per_second": 31.465, | |
| "eval_steps_per_second": 4.09, | |
| "step": 975 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 39.0, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 40.0, | |
| "learning_rate": 0.025, | |
| "loss": 0.8805, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 40.0, | |
| "eval_accuracy": 0.57, | |
| "eval_loss": 0.2923229932785034, | |
| "eval_runtime": 3.1744, | |
| "eval_samples_per_second": 31.502, | |
| "eval_steps_per_second": 4.095, | |
| "step": 1000 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 40.0, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 41.0, | |
| "eval_accuracy": 0.61, | |
| "eval_loss": 0.29315364360809326, | |
| "eval_runtime": 3.1688, | |
| "eval_samples_per_second": 31.558, | |
| "eval_steps_per_second": 4.103, | |
| "step": 1025 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 41.0, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 42.0, | |
| "eval_accuracy": 0.46, | |
| "eval_loss": 0.29364487528800964, | |
| "eval_runtime": 3.1666, | |
| "eval_samples_per_second": 31.58, | |
| "eval_steps_per_second": 4.105, | |
| "step": 1050 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 42.0, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 43.0, | |
| "eval_accuracy": 0.55, | |
| "eval_loss": 0.29236218333244324, | |
| "eval_runtime": 3.1736, | |
| "eval_samples_per_second": 31.51, | |
| "eval_steps_per_second": 4.096, | |
| "step": 1075 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 43.0, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 44.0, | |
| "eval_accuracy": 0.44, | |
| "eval_loss": 0.2936778664588928, | |
| "eval_runtime": 3.1653, | |
| "eval_samples_per_second": 31.593, | |
| "eval_steps_per_second": 4.107, | |
| "step": 1100 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 44.0, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 45.0, | |
| "eval_accuracy": 0.55, | |
| "eval_loss": 0.29269474744796753, | |
| "eval_runtime": 3.164, | |
| "eval_samples_per_second": 31.605, | |
| "eval_steps_per_second": 4.109, | |
| "step": 1125 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 45.0, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 46.0, | |
| "eval_accuracy": 0.56, | |
| "eval_loss": 0.292283296585083, | |
| "eval_runtime": 3.1643, | |
| "eval_samples_per_second": 31.602, | |
| "eval_steps_per_second": 4.108, | |
| "step": 1150 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 46.0, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 47.0, | |
| "eval_accuracy": 0.6, | |
| "eval_loss": 0.2930302023887634, | |
| "eval_runtime": 3.1636, | |
| "eval_samples_per_second": 31.609, | |
| "eval_steps_per_second": 4.109, | |
| "step": 1175 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 47.0, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 48.0, | |
| "eval_accuracy": 0.43, | |
| "eval_loss": 0.293615460395813, | |
| "eval_runtime": 3.1649, | |
| "eval_samples_per_second": 31.597, | |
| "eval_steps_per_second": 4.108, | |
| "step": 1200 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 48.0, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 49.0, | |
| "eval_accuracy": 0.56, | |
| "eval_loss": 0.2934630811214447, | |
| "eval_runtime": 3.1667, | |
| "eval_samples_per_second": 31.579, | |
| "eval_steps_per_second": 4.105, | |
| "step": 1225 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 49.0, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 50.0, | |
| "eval_accuracy": 0.46, | |
| "eval_loss": 0.29368773102760315, | |
| "eval_runtime": 3.164, | |
| "eval_samples_per_second": 31.605, | |
| "eval_steps_per_second": 4.109, | |
| "step": 1250 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 50.0, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 51.0, | |
| "eval_accuracy": 0.59, | |
| "eval_loss": 0.2928531765937805, | |
| "eval_runtime": 3.1606, | |
| "eval_samples_per_second": 31.64, | |
| "eval_steps_per_second": 4.113, | |
| "step": 1275 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 51.0, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 52.0, | |
| "eval_accuracy": 0.55, | |
| "eval_loss": 0.29315054416656494, | |
| "eval_runtime": 3.161, | |
| "eval_samples_per_second": 31.635, | |
| "eval_steps_per_second": 4.113, | |
| "step": 1300 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 52.0, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 53.0, | |
| "eval_accuracy": 0.48, | |
| "eval_loss": 0.2939896881580353, | |
| "eval_runtime": 3.1656, | |
| "eval_samples_per_second": 31.589, | |
| "eval_steps_per_second": 4.107, | |
| "step": 1325 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 53.0, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 54.0, | |
| "eval_accuracy": 0.53, | |
| "eval_loss": 0.29334965348243713, | |
| "eval_runtime": 3.1612, | |
| "eval_samples_per_second": 31.634, | |
| "eval_steps_per_second": 4.112, | |
| "step": 1350 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 54.0, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 55.0, | |
| "eval_accuracy": 0.55, | |
| "eval_loss": 0.29343104362487793, | |
| "eval_runtime": 3.1613, | |
| "eval_samples_per_second": 31.633, | |
| "eval_steps_per_second": 4.112, | |
| "step": 1375 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 55.0, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 56.0, | |
| "eval_accuracy": 0.49, | |
| "eval_loss": 0.29364249110221863, | |
| "eval_runtime": 3.164, | |
| "eval_samples_per_second": 31.606, | |
| "eval_steps_per_second": 4.109, | |
| "step": 1400 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 56.0, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 57.0, | |
| "eval_accuracy": 0.59, | |
| "eval_loss": 0.2927549481391907, | |
| "eval_runtime": 3.1588, | |
| "eval_samples_per_second": 31.658, | |
| "eval_steps_per_second": 4.116, | |
| "step": 1425 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 57.0, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 58.0, | |
| "eval_accuracy": 0.53, | |
| "eval_loss": 0.2926596701145172, | |
| "eval_runtime": 3.1589, | |
| "eval_samples_per_second": 31.656, | |
| "eval_steps_per_second": 4.115, | |
| "step": 1450 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 58.0, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 59.0, | |
| "eval_accuracy": 0.6, | |
| "eval_loss": 0.2930474579334259, | |
| "eval_runtime": 3.1575, | |
| "eval_samples_per_second": 31.671, | |
| "eval_steps_per_second": 4.117, | |
| "step": 1475 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 59.0, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 60.0, | |
| "learning_rate": 0.0125, | |
| "loss": 0.6612, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 60.0, | |
| "eval_accuracy": 0.47, | |
| "eval_loss": 0.29359233379364014, | |
| "eval_runtime": 3.1591, | |
| "eval_samples_per_second": 31.655, | |
| "eval_steps_per_second": 4.115, | |
| "step": 1500 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 60.0, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 61.0, | |
| "eval_accuracy": 0.53, | |
| "eval_loss": 0.29333430528640747, | |
| "eval_runtime": 3.1611, | |
| "eval_samples_per_second": 31.635, | |
| "eval_steps_per_second": 4.113, | |
| "step": 1525 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 61.0, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 62.0, | |
| "eval_accuracy": 0.62, | |
| "eval_loss": 0.2932239770889282, | |
| "eval_runtime": 3.1594, | |
| "eval_samples_per_second": 31.651, | |
| "eval_steps_per_second": 4.115, | |
| "step": 1550 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 62.0, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 63.0, | |
| "eval_accuracy": 0.41, | |
| "eval_loss": 0.2936784029006958, | |
| "eval_runtime": 3.1617, | |
| "eval_samples_per_second": 31.628, | |
| "eval_steps_per_second": 4.112, | |
| "step": 1575 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 63.0, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 64.0, | |
| "eval_accuracy": 0.54, | |
| "eval_loss": 0.2932013273239136, | |
| "eval_runtime": 3.1572, | |
| "eval_samples_per_second": 31.674, | |
| "eval_steps_per_second": 4.118, | |
| "step": 1600 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 64.0, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 65.0, | |
| "eval_accuracy": 0.42, | |
| "eval_loss": 0.29397743940353394, | |
| "eval_runtime": 3.1602, | |
| "eval_samples_per_second": 31.644, | |
| "eval_steps_per_second": 4.114, | |
| "step": 1625 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 65.0, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 66.0, | |
| "eval_accuracy": 0.56, | |
| "eval_loss": 0.293092280626297, | |
| "eval_runtime": 3.1597, | |
| "eval_samples_per_second": 31.649, | |
| "eval_steps_per_second": 4.114, | |
| "step": 1650 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 66.0, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 67.0, | |
| "eval_accuracy": 0.36, | |
| "eval_loss": 0.29367607831954956, | |
| "eval_runtime": 3.1629, | |
| "eval_samples_per_second": 31.616, | |
| "eval_steps_per_second": 4.11, | |
| "step": 1675 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 67.0, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 68.0, | |
| "eval_accuracy": 0.63, | |
| "eval_loss": 0.2929648458957672, | |
| "eval_runtime": 3.1586, | |
| "eval_samples_per_second": 31.659, | |
| "eval_steps_per_second": 4.116, | |
| "step": 1700 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 68.0, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 69.0, | |
| "eval_accuracy": 0.63, | |
| "eval_loss": 0.29336318373680115, | |
| "eval_runtime": 3.158, | |
| "eval_samples_per_second": 31.665, | |
| "eval_steps_per_second": 4.116, | |
| "step": 1725 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 69.0, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 70.0, | |
| "eval_accuracy": 0.36, | |
| "eval_loss": 0.29374971985816956, | |
| "eval_runtime": 3.1604, | |
| "eval_samples_per_second": 31.641, | |
| "eval_steps_per_second": 4.113, | |
| "step": 1750 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 70.0, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 71.0, | |
| "eval_accuracy": 0.63, | |
| "eval_loss": 0.29300999641418457, | |
| "eval_runtime": 3.1587, | |
| "eval_samples_per_second": 31.659, | |
| "eval_steps_per_second": 4.116, | |
| "step": 1775 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 71.0, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 72.0, | |
| "eval_accuracy": 0.63, | |
| "eval_loss": 0.2931566536426544, | |
| "eval_runtime": 3.1592, | |
| "eval_samples_per_second": 31.654, | |
| "eval_steps_per_second": 4.115, | |
| "step": 1800 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 72.0, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 73.0, | |
| "eval_accuracy": 0.61, | |
| "eval_loss": 0.29304105043411255, | |
| "eval_runtime": 3.1567, | |
| "eval_samples_per_second": 31.679, | |
| "eval_steps_per_second": 4.118, | |
| "step": 1825 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 73.0, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 74.0, | |
| "eval_accuracy": 0.53, | |
| "eval_loss": 0.2932125926017761, | |
| "eval_runtime": 3.1604, | |
| "eval_samples_per_second": 31.642, | |
| "eval_steps_per_second": 4.113, | |
| "step": 1850 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 74.0, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 75.0, | |
| "eval_accuracy": 0.58, | |
| "eval_loss": 0.29323533177375793, | |
| "eval_runtime": 3.1581, | |
| "eval_samples_per_second": 31.664, | |
| "eval_steps_per_second": 4.116, | |
| "step": 1875 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 75.0, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 76.0, | |
| "eval_accuracy": 0.53, | |
| "eval_loss": 0.2934551239013672, | |
| "eval_runtime": 3.1621, | |
| "eval_samples_per_second": 31.625, | |
| "eval_steps_per_second": 4.111, | |
| "step": 1900 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 76.0, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 77.0, | |
| "eval_accuracy": 0.62, | |
| "eval_loss": 0.2931278347969055, | |
| "eval_runtime": 3.1589, | |
| "eval_samples_per_second": 31.656, | |
| "eval_steps_per_second": 4.115, | |
| "step": 1925 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 77.0, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 78.0, | |
| "eval_accuracy": 0.54, | |
| "eval_loss": 0.2932642102241516, | |
| "eval_runtime": 3.158, | |
| "eval_samples_per_second": 31.666, | |
| "eval_steps_per_second": 4.117, | |
| "step": 1950 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 78.0, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 79.0, | |
| "eval_accuracy": 0.61, | |
| "eval_loss": 0.29319581389427185, | |
| "eval_runtime": 3.1582, | |
| "eval_samples_per_second": 31.664, | |
| "eval_steps_per_second": 4.116, | |
| "step": 1975 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 79.0, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 80.0, | |
| "learning_rate": 0.0, | |
| "loss": 0.5295, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 80.0, | |
| "eval_accuracy": 0.6, | |
| "eval_loss": 0.29315412044525146, | |
| "eval_runtime": 3.16, | |
| "eval_samples_per_second": 31.646, | |
| "eval_steps_per_second": 4.114, | |
| "step": 2000 | |
| }, | |
| { | |
| "best_epoch": 7, | |
| "best_eval_accuracy": 0.68, | |
| "epoch": 80.0, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 80.0, | |
| "step": 2000, | |
| "total_flos": 2.9821702864896e+16, | |
| "train_loss": 0.8249960479736328, | |
| "train_runtime": 1692.8932, | |
| "train_samples_per_second": 18.903, | |
| "train_steps_per_second": 1.181 | |
| } | |
| ], | |
| "max_steps": 2000, | |
| "num_train_epochs": 80, | |
| "total_flos": 2.9821702864896e+16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |