| { | |
| "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": 0.44, | |
| "eval_loss": 0.3608391582965851, | |
| "eval_runtime": 2.9715, | |
| "eval_samples_per_second": 33.653, | |
| "eval_steps_per_second": 4.375, | |
| "step": 25 | |
| }, | |
| { | |
| "best_epoch": 0, | |
| "best_eval_accuracy": 0.44, | |
| "epoch": 1.0, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.57, | |
| "eval_loss": 0.2890351712703705, | |
| "eval_runtime": 3.0216, | |
| "eval_samples_per_second": 33.095, | |
| "eval_steps_per_second": 4.302, | |
| "step": 50 | |
| }, | |
| { | |
| "best_epoch": 1, | |
| "best_eval_accuracy": 0.57, | |
| "epoch": 2.0, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.58, | |
| "eval_loss": 0.29608333110809326, | |
| "eval_runtime": 3.0821, | |
| "eval_samples_per_second": 32.446, | |
| "eval_steps_per_second": 4.218, | |
| "step": 75 | |
| }, | |
| { | |
| "best_epoch": 2, | |
| "best_eval_accuracy": 0.58, | |
| "epoch": 3.0, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.65, | |
| "eval_loss": 0.28651857376098633, | |
| "eval_runtime": 3.1117, | |
| "eval_samples_per_second": 32.137, | |
| "eval_steps_per_second": 4.178, | |
| "step": 100 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 4.0, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.58, | |
| "eval_loss": 0.2900591790676117, | |
| "eval_runtime": 3.136, | |
| "eval_samples_per_second": 31.888, | |
| "eval_steps_per_second": 4.145, | |
| "step": 125 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 5.0, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.46, | |
| "eval_loss": 0.2933403253555298, | |
| "eval_runtime": 3.1448, | |
| "eval_samples_per_second": 31.798, | |
| "eval_steps_per_second": 4.134, | |
| "step": 150 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 6.0, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.64, | |
| "eval_loss": 0.32912155985832214, | |
| "eval_runtime": 3.1478, | |
| "eval_samples_per_second": 31.768, | |
| "eval_steps_per_second": 4.13, | |
| "step": 175 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 7.0, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.62, | |
| "eval_loss": 0.28637343645095825, | |
| "eval_runtime": 3.1555, | |
| "eval_samples_per_second": 31.69, | |
| "eval_steps_per_second": 4.12, | |
| "step": 200 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 8.0, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.42, | |
| "eval_loss": 0.29787373542785645, | |
| "eval_runtime": 3.1584, | |
| "eval_samples_per_second": 31.662, | |
| "eval_steps_per_second": 4.116, | |
| "step": 225 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 9.0, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_accuracy": 0.63, | |
| "eval_loss": 0.3034938871860504, | |
| "eval_runtime": 3.1614, | |
| "eval_samples_per_second": 31.632, | |
| "eval_steps_per_second": 4.112, | |
| "step": 250 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 10.0, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_accuracy": 0.59, | |
| "eval_loss": 0.29021456837654114, | |
| "eval_runtime": 3.1628, | |
| "eval_samples_per_second": 31.618, | |
| "eval_steps_per_second": 4.11, | |
| "step": 275 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 11.0, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "eval_accuracy": 0.5, | |
| "eval_loss": 0.2916863262653351, | |
| "eval_runtime": 3.1631, | |
| "eval_samples_per_second": 31.615, | |
| "eval_steps_per_second": 4.11, | |
| "step": 300 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 12.0, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "eval_accuracy": 0.44, | |
| "eval_loss": 0.29346537590026855, | |
| "eval_runtime": 3.1644, | |
| "eval_samples_per_second": 31.601, | |
| "eval_steps_per_second": 4.108, | |
| "step": 325 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 13.0, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "eval_accuracy": 0.44, | |
| "eval_loss": 0.3056786060333252, | |
| "eval_runtime": 3.1659, | |
| "eval_samples_per_second": 31.586, | |
| "eval_steps_per_second": 4.106, | |
| "step": 350 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 14.0, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "eval_accuracy": 0.45, | |
| "eval_loss": 0.2980119287967682, | |
| "eval_runtime": 3.1656, | |
| "eval_samples_per_second": 31.589, | |
| "eval_steps_per_second": 4.107, | |
| "step": 375 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 15.0, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "eval_accuracy": 0.47, | |
| "eval_loss": 0.2946815490722656, | |
| "eval_runtime": 3.166, | |
| "eval_samples_per_second": 31.585, | |
| "eval_steps_per_second": 4.106, | |
| "step": 400 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 16.0, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 17.0, | |
| "eval_accuracy": 0.5, | |
| "eval_loss": 0.29449501633644104, | |
| "eval_runtime": 3.1641, | |
| "eval_samples_per_second": 31.605, | |
| "eval_steps_per_second": 4.109, | |
| "step": 425 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 17.0, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "eval_accuracy": 0.49, | |
| "eval_loss": 0.2923860251903534, | |
| "eval_runtime": 3.1667, | |
| "eval_samples_per_second": 31.579, | |
| "eval_steps_per_second": 4.105, | |
| "step": 450 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 18.0, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 19.0, | |
| "eval_accuracy": 0.55, | |
| "eval_loss": 0.2921885848045349, | |
| "eval_runtime": 3.1694, | |
| "eval_samples_per_second": 31.552, | |
| "eval_steps_per_second": 4.102, | |
| "step": 475 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 19.0, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "learning_rate": 0.037500000000000006, | |
| "loss": 1.1902, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "eval_accuracy": 0.45, | |
| "eval_loss": 0.292321115732193, | |
| "eval_runtime": 3.1784, | |
| "eval_samples_per_second": 31.462, | |
| "eval_steps_per_second": 4.09, | |
| "step": 500 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 20.0, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 21.0, | |
| "eval_accuracy": 0.55, | |
| "eval_loss": 0.2864084541797638, | |
| "eval_runtime": 3.1692, | |
| "eval_samples_per_second": 31.554, | |
| "eval_steps_per_second": 4.102, | |
| "step": 525 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 21.0, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 22.0, | |
| "eval_accuracy": 0.42, | |
| "eval_loss": 0.2925306558609009, | |
| "eval_runtime": 3.1691, | |
| "eval_samples_per_second": 31.555, | |
| "eval_steps_per_second": 4.102, | |
| "step": 550 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 22.0, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 23.0, | |
| "eval_accuracy": 0.58, | |
| "eval_loss": 0.2909621000289917, | |
| "eval_runtime": 3.1683, | |
| "eval_samples_per_second": 31.563, | |
| "eval_steps_per_second": 4.103, | |
| "step": 575 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 23.0, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 24.0, | |
| "eval_accuracy": 0.58, | |
| "eval_loss": 0.28951704502105713, | |
| "eval_runtime": 3.1698, | |
| "eval_samples_per_second": 31.548, | |
| "eval_steps_per_second": 4.101, | |
| "step": 600 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 24.0, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 25.0, | |
| "eval_accuracy": 0.62, | |
| "eval_loss": 0.2918489873409271, | |
| "eval_runtime": 3.1538, | |
| "eval_samples_per_second": 31.708, | |
| "eval_steps_per_second": 4.122, | |
| "step": 625 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 25.0, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 26.0, | |
| "eval_accuracy": 0.42, | |
| "eval_loss": 0.2921474874019623, | |
| "eval_runtime": 3.1532, | |
| "eval_samples_per_second": 31.714, | |
| "eval_steps_per_second": 4.123, | |
| "step": 650 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 26.0, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 27.0, | |
| "eval_accuracy": 0.58, | |
| "eval_loss": 0.29183533787727356, | |
| "eval_runtime": 3.153, | |
| "eval_samples_per_second": 31.716, | |
| "eval_steps_per_second": 4.123, | |
| "step": 675 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 27.0, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 28.0, | |
| "eval_accuracy": 0.6, | |
| "eval_loss": 0.2910160720348358, | |
| "eval_runtime": 3.1555, | |
| "eval_samples_per_second": 31.691, | |
| "eval_steps_per_second": 4.12, | |
| "step": 700 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 28.0, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 29.0, | |
| "eval_accuracy": 0.57, | |
| "eval_loss": 0.29186198115348816, | |
| "eval_runtime": 3.1611, | |
| "eval_samples_per_second": 31.635, | |
| "eval_steps_per_second": 4.113, | |
| "step": 725 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 29.0, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 30.0, | |
| "eval_accuracy": 0.48, | |
| "eval_loss": 0.2920497953891754, | |
| "eval_runtime": 3.1584, | |
| "eval_samples_per_second": 31.662, | |
| "eval_steps_per_second": 4.116, | |
| "step": 750 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 30.0, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 31.0, | |
| "eval_accuracy": 0.41, | |
| "eval_loss": 0.2922162115573883, | |
| "eval_runtime": 3.1557, | |
| "eval_samples_per_second": 31.688, | |
| "eval_steps_per_second": 4.119, | |
| "step": 775 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 31.0, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 32.0, | |
| "eval_accuracy": 0.53, | |
| "eval_loss": 0.2919849157333374, | |
| "eval_runtime": 3.1524, | |
| "eval_samples_per_second": 31.722, | |
| "eval_steps_per_second": 4.124, | |
| "step": 800 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 32.0, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 33.0, | |
| "eval_accuracy": 0.51, | |
| "eval_loss": 0.29196542501449585, | |
| "eval_runtime": 3.1533, | |
| "eval_samples_per_second": 31.713, | |
| "eval_steps_per_second": 4.123, | |
| "step": 825 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 33.0, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 34.0, | |
| "eval_accuracy": 0.54, | |
| "eval_loss": 0.29190951585769653, | |
| "eval_runtime": 3.1549, | |
| "eval_samples_per_second": 31.697, | |
| "eval_steps_per_second": 4.121, | |
| "step": 850 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 34.0, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 35.0, | |
| "eval_accuracy": 0.52, | |
| "eval_loss": 0.2919704020023346, | |
| "eval_runtime": 3.1601, | |
| "eval_samples_per_second": 31.645, | |
| "eval_steps_per_second": 4.114, | |
| "step": 875 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 35.0, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 36.0, | |
| "eval_accuracy": 0.39, | |
| "eval_loss": 0.29210489988327026, | |
| "eval_runtime": 3.1561, | |
| "eval_samples_per_second": 31.684, | |
| "eval_steps_per_second": 4.119, | |
| "step": 900 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 36.0, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 37.0, | |
| "eval_accuracy": 0.53, | |
| "eval_loss": 0.2919726073741913, | |
| "eval_runtime": 3.155, | |
| "eval_samples_per_second": 31.696, | |
| "eval_steps_per_second": 4.12, | |
| "step": 925 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 37.0, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 38.0, | |
| "eval_accuracy": 0.49, | |
| "eval_loss": 0.2919931411743164, | |
| "eval_runtime": 3.1566, | |
| "eval_samples_per_second": 31.68, | |
| "eval_steps_per_second": 4.118, | |
| "step": 950 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 38.0, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 39.0, | |
| "eval_accuracy": 0.4, | |
| "eval_loss": 0.29224902391433716, | |
| "eval_runtime": 3.1538, | |
| "eval_samples_per_second": 31.708, | |
| "eval_steps_per_second": 4.122, | |
| "step": 975 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 39.0, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 40.0, | |
| "learning_rate": 0.025, | |
| "loss": 0.8276, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 40.0, | |
| "eval_accuracy": 0.58, | |
| "eval_loss": 0.2919129729270935, | |
| "eval_runtime": 3.1515, | |
| "eval_samples_per_second": 31.731, | |
| "eval_steps_per_second": 4.125, | |
| "step": 1000 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 40.0, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 41.0, | |
| "eval_accuracy": 0.62, | |
| "eval_loss": 0.2918228805065155, | |
| "eval_runtime": 3.156, | |
| "eval_samples_per_second": 31.685, | |
| "eval_steps_per_second": 4.119, | |
| "step": 1025 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 41.0, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 42.0, | |
| "eval_accuracy": 0.61, | |
| "eval_loss": 0.2918192744255066, | |
| "eval_runtime": 3.153, | |
| "eval_samples_per_second": 31.716, | |
| "eval_steps_per_second": 4.123, | |
| "step": 1050 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 42.0, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 43.0, | |
| "eval_accuracy": 0.42, | |
| "eval_loss": 0.2921864688396454, | |
| "eval_runtime": 3.1516, | |
| "eval_samples_per_second": 31.73, | |
| "eval_steps_per_second": 4.125, | |
| "step": 1075 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 43.0, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 44.0, | |
| "eval_accuracy": 0.43, | |
| "eval_loss": 0.2921425700187683, | |
| "eval_runtime": 3.1517, | |
| "eval_samples_per_second": 31.729, | |
| "eval_steps_per_second": 4.125, | |
| "step": 1100 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 44.0, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 45.0, | |
| "eval_accuracy": 0.42, | |
| "eval_loss": 0.29200559854507446, | |
| "eval_runtime": 3.1538, | |
| "eval_samples_per_second": 31.708, | |
| "eval_steps_per_second": 4.122, | |
| "step": 1125 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 45.0, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 46.0, | |
| "eval_accuracy": 0.42, | |
| "eval_loss": 0.29199379682540894, | |
| "eval_runtime": 3.1521, | |
| "eval_samples_per_second": 31.725, | |
| "eval_steps_per_second": 4.124, | |
| "step": 1150 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 46.0, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 47.0, | |
| "eval_accuracy": 0.35, | |
| "eval_loss": 0.29200825095176697, | |
| "eval_runtime": 3.1518, | |
| "eval_samples_per_second": 31.728, | |
| "eval_steps_per_second": 4.125, | |
| "step": 1175 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 47.0, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 48.0, | |
| "eval_accuracy": 0.54, | |
| "eval_loss": 0.291986346244812, | |
| "eval_runtime": 3.1552, | |
| "eval_samples_per_second": 31.693, | |
| "eval_steps_per_second": 4.12, | |
| "step": 1200 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 48.0, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 49.0, | |
| "eval_accuracy": 0.6, | |
| "eval_loss": 0.29196932911872864, | |
| "eval_runtime": 3.1536, | |
| "eval_samples_per_second": 31.709, | |
| "eval_steps_per_second": 4.122, | |
| "step": 1225 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 49.0, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 50.0, | |
| "eval_accuracy": 0.52, | |
| "eval_loss": 0.2919679880142212, | |
| "eval_runtime": 3.1556, | |
| "eval_samples_per_second": 31.69, | |
| "eval_steps_per_second": 4.12, | |
| "step": 1250 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 50.0, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 51.0, | |
| "eval_accuracy": 0.37, | |
| "eval_loss": 0.2920134961605072, | |
| "eval_runtime": 3.1538, | |
| "eval_samples_per_second": 31.707, | |
| "eval_steps_per_second": 4.122, | |
| "step": 1275 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 51.0, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 52.0, | |
| "eval_accuracy": 0.45, | |
| "eval_loss": 0.2920159101486206, | |
| "eval_runtime": 3.1519, | |
| "eval_samples_per_second": 31.727, | |
| "eval_steps_per_second": 4.125, | |
| "step": 1300 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 52.0, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 53.0, | |
| "eval_accuracy": 0.44, | |
| "eval_loss": 0.2919909358024597, | |
| "eval_runtime": 3.1539, | |
| "eval_samples_per_second": 31.707, | |
| "eval_steps_per_second": 4.122, | |
| "step": 1325 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 53.0, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 54.0, | |
| "eval_accuracy": 0.59, | |
| "eval_loss": 0.29197338223457336, | |
| "eval_runtime": 3.1515, | |
| "eval_samples_per_second": 31.731, | |
| "eval_steps_per_second": 4.125, | |
| "step": 1350 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 54.0, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 55.0, | |
| "eval_accuracy": 0.44, | |
| "eval_loss": 0.29198840260505676, | |
| "eval_runtime": 3.1535, | |
| "eval_samples_per_second": 31.71, | |
| "eval_steps_per_second": 4.122, | |
| "step": 1375 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 55.0, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 56.0, | |
| "eval_accuracy": 0.58, | |
| "eval_loss": 0.2919729948043823, | |
| "eval_runtime": 3.1547, | |
| "eval_samples_per_second": 31.699, | |
| "eval_steps_per_second": 4.121, | |
| "step": 1400 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 56.0, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 57.0, | |
| "eval_accuracy": 0.57, | |
| "eval_loss": 0.29197487235069275, | |
| "eval_runtime": 3.1506, | |
| "eval_samples_per_second": 31.739, | |
| "eval_steps_per_second": 4.126, | |
| "step": 1425 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 57.0, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 58.0, | |
| "eval_accuracy": 0.46, | |
| "eval_loss": 0.2920148968696594, | |
| "eval_runtime": 3.1554, | |
| "eval_samples_per_second": 31.692, | |
| "eval_steps_per_second": 4.12, | |
| "step": 1450 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 58.0, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 59.0, | |
| "eval_accuracy": 0.42, | |
| "eval_loss": 0.29200279712677, | |
| "eval_runtime": 3.1551, | |
| "eval_samples_per_second": 31.694, | |
| "eval_steps_per_second": 4.12, | |
| "step": 1475 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 59.0, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 60.0, | |
| "learning_rate": 0.0125, | |
| "loss": 0.6389, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 60.0, | |
| "eval_accuracy": 0.37, | |
| "eval_loss": 0.2919899523258209, | |
| "eval_runtime": 3.1542, | |
| "eval_samples_per_second": 31.704, | |
| "eval_steps_per_second": 4.121, | |
| "step": 1500 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 60.0, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 61.0, | |
| "eval_accuracy": 0.6, | |
| "eval_loss": 0.29192057251930237, | |
| "eval_runtime": 3.1549, | |
| "eval_samples_per_second": 31.697, | |
| "eval_steps_per_second": 4.121, | |
| "step": 1525 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 61.0, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 62.0, | |
| "eval_accuracy": 0.6, | |
| "eval_loss": 0.29190564155578613, | |
| "eval_runtime": 3.1549, | |
| "eval_samples_per_second": 31.696, | |
| "eval_steps_per_second": 4.121, | |
| "step": 1550 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 62.0, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 63.0, | |
| "eval_accuracy": 0.55, | |
| "eval_loss": 0.29197636246681213, | |
| "eval_runtime": 3.1562, | |
| "eval_samples_per_second": 31.684, | |
| "eval_steps_per_second": 4.119, | |
| "step": 1575 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 63.0, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 64.0, | |
| "eval_accuracy": 0.52, | |
| "eval_loss": 0.2919798195362091, | |
| "eval_runtime": 3.1523, | |
| "eval_samples_per_second": 31.723, | |
| "eval_steps_per_second": 4.124, | |
| "step": 1600 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 64.0, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 65.0, | |
| "eval_accuracy": 0.5, | |
| "eval_loss": 0.2919777035713196, | |
| "eval_runtime": 3.157, | |
| "eval_samples_per_second": 31.676, | |
| "eval_steps_per_second": 4.118, | |
| "step": 1625 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 65.0, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 66.0, | |
| "eval_accuracy": 0.36, | |
| "eval_loss": 0.2920242249965668, | |
| "eval_runtime": 3.1564, | |
| "eval_samples_per_second": 31.682, | |
| "eval_steps_per_second": 4.119, | |
| "step": 1650 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 66.0, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 67.0, | |
| "eval_accuracy": 0.58, | |
| "eval_loss": 0.29195553064346313, | |
| "eval_runtime": 3.1547, | |
| "eval_samples_per_second": 31.698, | |
| "eval_steps_per_second": 4.121, | |
| "step": 1675 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 67.0, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 68.0, | |
| "eval_accuracy": 0.38, | |
| "eval_loss": 0.29199790954589844, | |
| "eval_runtime": 3.1555, | |
| "eval_samples_per_second": 31.69, | |
| "eval_steps_per_second": 4.12, | |
| "step": 1700 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 68.0, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 69.0, | |
| "eval_accuracy": 0.58, | |
| "eval_loss": 0.2919797897338867, | |
| "eval_runtime": 3.1523, | |
| "eval_samples_per_second": 31.723, | |
| "eval_steps_per_second": 4.124, | |
| "step": 1725 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 69.0, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 70.0, | |
| "eval_accuracy": 0.53, | |
| "eval_loss": 0.2919650375843048, | |
| "eval_runtime": 3.1536, | |
| "eval_samples_per_second": 31.71, | |
| "eval_steps_per_second": 4.122, | |
| "step": 1750 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 70.0, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 71.0, | |
| "eval_accuracy": 0.37, | |
| "eval_loss": 0.2920193374156952, | |
| "eval_runtime": 3.1551, | |
| "eval_samples_per_second": 31.695, | |
| "eval_steps_per_second": 4.12, | |
| "step": 1775 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 71.0, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 72.0, | |
| "eval_accuracy": 0.39, | |
| "eval_loss": 0.2920474112033844, | |
| "eval_runtime": 3.1506, | |
| "eval_samples_per_second": 31.74, | |
| "eval_steps_per_second": 4.126, | |
| "step": 1800 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 72.0, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 73.0, | |
| "eval_accuracy": 0.36, | |
| "eval_loss": 0.29199421405792236, | |
| "eval_runtime": 3.1521, | |
| "eval_samples_per_second": 31.724, | |
| "eval_steps_per_second": 4.124, | |
| "step": 1825 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 73.0, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 74.0, | |
| "eval_accuracy": 0.43, | |
| "eval_loss": 0.29198718070983887, | |
| "eval_runtime": 3.1542, | |
| "eval_samples_per_second": 31.703, | |
| "eval_steps_per_second": 4.121, | |
| "step": 1850 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 74.0, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 75.0, | |
| "eval_accuracy": 0.38, | |
| "eval_loss": 0.29200947284698486, | |
| "eval_runtime": 3.1537, | |
| "eval_samples_per_second": 31.709, | |
| "eval_steps_per_second": 4.122, | |
| "step": 1875 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 75.0, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 76.0, | |
| "eval_accuracy": 0.43, | |
| "eval_loss": 0.29198354482650757, | |
| "eval_runtime": 3.1536, | |
| "eval_samples_per_second": 31.71, | |
| "eval_steps_per_second": 4.122, | |
| "step": 1900 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 76.0, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 77.0, | |
| "eval_accuracy": 0.37, | |
| "eval_loss": 0.2920118570327759, | |
| "eval_runtime": 3.152, | |
| "eval_samples_per_second": 31.726, | |
| "eval_steps_per_second": 4.124, | |
| "step": 1925 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 77.0, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 78.0, | |
| "eval_accuracy": 0.37, | |
| "eval_loss": 0.29201093316078186, | |
| "eval_runtime": 3.1508, | |
| "eval_samples_per_second": 31.738, | |
| "eval_steps_per_second": 4.126, | |
| "step": 1950 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 78.0, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 79.0, | |
| "eval_accuracy": 0.38, | |
| "eval_loss": 0.2919941246509552, | |
| "eval_runtime": 3.1558, | |
| "eval_samples_per_second": 31.688, | |
| "eval_steps_per_second": 4.119, | |
| "step": 1975 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 79.0, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 80.0, | |
| "learning_rate": 0.0, | |
| "loss": 0.5225, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 80.0, | |
| "eval_accuracy": 0.4, | |
| "eval_loss": 0.2919914722442627, | |
| "eval_runtime": 3.1525, | |
| "eval_samples_per_second": 31.721, | |
| "eval_steps_per_second": 4.124, | |
| "step": 2000 | |
| }, | |
| { | |
| "best_epoch": 3, | |
| "best_eval_accuracy": 0.65, | |
| "epoch": 80.0, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 80.0, | |
| "step": 2000, | |
| "total_flos": 2.9821702864896e+16, | |
| "train_loss": 0.7947869720458984, | |
| "train_runtime": 1692.4229, | |
| "train_samples_per_second": 18.908, | |
| "train_steps_per_second": 1.182 | |
| } | |
| ], | |
| "max_steps": 2000, | |
| "num_train_epochs": 80, | |
| "total_flos": 2.9821702864896e+16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |