{ "best_metric": 3.0083158016204834, "best_model_checkpoint": "/home/p318482/babyLM_controlled/models_trained/de_mlm/wikipedia_13/checkpoint-84000", "epoch": 61.56101135947233, "eval_steps": 2000, "global_step": 84000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 1.4657383657017222, "eval_loss": 8.087221145629883, "eval_runtime": 2.215, "eval_samples_per_second": 1035.232, "eval_steps_per_second": 65.012, "step": 2000 }, { "epoch": 2.9314767314034444, "grad_norm": 0.9454871416091919, "learning_rate": 1e-05, "loss": 8.1326, "step": 4000 }, { "epoch": 2.9314767314034444, "eval_loss": 7.397463798522949, "eval_runtime": 2.2812, "eval_samples_per_second": 1005.158, "eval_steps_per_second": 63.124, "step": 4000 }, { "epoch": 4.397215097105167, "eval_loss": 7.2718353271484375, "eval_runtime": 2.3459, "eval_samples_per_second": 977.446, "eval_steps_per_second": 61.383, "step": 6000 }, { "epoch": 5.862953462806889, "grad_norm": 1.4085547924041748, "learning_rate": 2e-05, "loss": 7.2846, "step": 8000 }, { "epoch": 5.862953462806889, "eval_loss": 7.183506965637207, "eval_runtime": 2.2787, "eval_samples_per_second": 1006.271, "eval_steps_per_second": 63.194, "step": 8000 }, { "epoch": 7.328691828508611, "eval_loss": 7.09923791885376, "eval_runtime": 2.1833, "eval_samples_per_second": 1050.26, "eval_steps_per_second": 65.956, "step": 10000 }, { "epoch": 8.794430194210333, "grad_norm": 1.4775667190551758, "learning_rate": 2.9995e-05, "loss": 7.1078, "step": 12000 }, { "epoch": 8.794430194210333, "eval_loss": 7.033052444458008, "eval_runtime": 2.2436, "eval_samples_per_second": 1022.003, "eval_steps_per_second": 64.182, "step": 12000 }, { "epoch": 10.260168559912056, "eval_loss": 6.949419975280762, "eval_runtime": 2.3034, "eval_samples_per_second": 995.501, "eval_steps_per_second": 62.517, "step": 14000 }, { "epoch": 11.725906925613778, "grad_norm": 2.364133358001709, "learning_rate": 3.999e-05, "loss": 6.942, "step": 16000 }, { "epoch": 11.725906925613778, "eval_loss": 6.889880180358887, "eval_runtime": 2.3275, "eval_samples_per_second": 985.162, "eval_steps_per_second": 61.868, "step": 16000 }, { "epoch": 13.191645291315501, "eval_loss": 6.782171726226807, "eval_runtime": 2.2331, "eval_samples_per_second": 1026.812, "eval_steps_per_second": 64.484, "step": 18000 }, { "epoch": 14.657383657017222, "grad_norm": 2.5763492584228516, "learning_rate": 4.9985e-05, "loss": 6.7676, "step": 20000 }, { "epoch": 14.657383657017222, "eval_loss": 6.718458652496338, "eval_runtime": 2.2092, "eval_samples_per_second": 1037.91, "eval_steps_per_second": 65.181, "step": 20000 }, { "epoch": 16.123122022718945, "eval_loss": 6.653624057769775, "eval_runtime": 2.33, "eval_samples_per_second": 984.135, "eval_steps_per_second": 61.804, "step": 22000 }, { "epoch": 17.588860388420667, "grad_norm": 4.001818656921387, "learning_rate": 5.9980000000000005e-05, "loss": 6.5959, "step": 24000 }, { "epoch": 17.588860388420667, "eval_loss": 6.543092727661133, "eval_runtime": 2.3092, "eval_samples_per_second": 992.981, "eval_steps_per_second": 62.359, "step": 24000 }, { "epoch": 19.054598754122388, "eval_loss": 6.392500400543213, "eval_runtime": 2.4925, "eval_samples_per_second": 919.956, "eval_steps_per_second": 57.773, "step": 26000 }, { "epoch": 20.520337119824113, "grad_norm": 4.359928607940674, "learning_rate": 6.99775e-05, "loss": 6.3624, "step": 28000 }, { "epoch": 20.520337119824113, "eval_loss": 6.211884021759033, "eval_runtime": 2.3499, "eval_samples_per_second": 975.797, "eval_steps_per_second": 61.28, "step": 28000 }, { "epoch": 21.986075485525834, "eval_loss": 5.952591896057129, "eval_runtime": 2.3365, "eval_samples_per_second": 981.37, "eval_steps_per_second": 61.63, "step": 30000 }, { "epoch": 23.451813851227556, "grad_norm": 5.47481632232666, "learning_rate": 7.997250000000001e-05, "loss": 5.9309, "step": 32000 }, { "epoch": 23.451813851227556, "eval_loss": 5.4161834716796875, "eval_runtime": 2.2165, "eval_samples_per_second": 1034.515, "eval_steps_per_second": 64.967, "step": 32000 }, { "epoch": 24.917552216929277, "eval_loss": 5.0255045890808105, "eval_runtime": 2.3492, "eval_samples_per_second": 976.095, "eval_steps_per_second": 61.299, "step": 34000 }, { "epoch": 26.383290582631002, "grad_norm": 5.85042667388916, "learning_rate": 8.996750000000001e-05, "loss": 5.0575, "step": 36000 }, { "epoch": 26.383290582631002, "eval_loss": 4.76804256439209, "eval_runtime": 2.2482, "eval_samples_per_second": 1019.914, "eval_steps_per_second": 64.05, "step": 36000 }, { "epoch": 27.849028948332723, "eval_loss": 4.502030372619629, "eval_runtime": 2.3066, "eval_samples_per_second": 994.084, "eval_steps_per_second": 62.428, "step": 38000 }, { "epoch": 29.314767314034444, "grad_norm": 4.6283698081970215, "learning_rate": 9.99625e-05, "loss": 4.5282, "step": 40000 }, { "epoch": 29.314767314034444, "eval_loss": 4.321413993835449, "eval_runtime": 2.2564, "eval_samples_per_second": 1016.223, "eval_steps_per_second": 63.819, "step": 40000 }, { "epoch": 30.780505679736166, "eval_loss": 4.131194114685059, "eval_runtime": 2.27, "eval_samples_per_second": 1010.131, "eval_steps_per_second": 63.436, "step": 42000 }, { "epoch": 32.24624404543789, "grad_norm": 4.427154064178467, "learning_rate": 9.336166666666667e-05, "loss": 4.1335, "step": 44000 }, { "epoch": 32.24624404543789, "eval_loss": 3.9708411693573, "eval_runtime": 2.1314, "eval_samples_per_second": 1075.829, "eval_steps_per_second": 67.562, "step": 44000 }, { "epoch": 33.71198241113961, "eval_loss": 3.861647605895996, "eval_runtime": 2.2216, "eval_samples_per_second": 1032.141, "eval_steps_per_second": 64.818, "step": 46000 }, { "epoch": 35.17772077684133, "grad_norm": 4.964674949645996, "learning_rate": 8.669666666666667e-05, "loss": 3.8339, "step": 48000 }, { "epoch": 35.17772077684133, "eval_loss": 3.763975143432617, "eval_runtime": 2.2678, "eval_samples_per_second": 1011.117, "eval_steps_per_second": 63.498, "step": 48000 }, { "epoch": 36.64345914254306, "eval_loss": 3.707411527633667, "eval_runtime": 2.1845, "eval_samples_per_second": 1049.673, "eval_steps_per_second": 65.919, "step": 50000 }, { "epoch": 38.109197508244776, "grad_norm": 5.166478633880615, "learning_rate": 8.003333333333333e-05, "loss": 3.6042, "step": 52000 }, { "epoch": 38.109197508244776, "eval_loss": 3.636035680770874, "eval_runtime": 2.279, "eval_samples_per_second": 1006.125, "eval_steps_per_second": 63.184, "step": 52000 }, { "epoch": 39.5749358739465, "eval_loss": 3.520251750946045, "eval_runtime": 2.3313, "eval_samples_per_second": 983.562, "eval_steps_per_second": 61.768, "step": 54000 }, { "epoch": 41.040674239648226, "grad_norm": 5.001477241516113, "learning_rate": 7.337000000000001e-05, "loss": 3.4291, "step": 56000 }, { "epoch": 41.040674239648226, "eval_loss": 3.4423604011535645, "eval_runtime": 2.4163, "eval_samples_per_second": 948.983, "eval_steps_per_second": 59.596, "step": 56000 }, { "epoch": 42.506412605349944, "eval_loss": 3.427623987197876, "eval_runtime": 2.1948, "eval_samples_per_second": 1044.73, "eval_steps_per_second": 65.609, "step": 58000 }, { "epoch": 43.97215097105167, "grad_norm": 5.074405670166016, "learning_rate": 6.670666666666668e-05, "loss": 3.286, "step": 60000 }, { "epoch": 43.97215097105167, "eval_loss": 3.3796520233154297, "eval_runtime": 2.2145, "eval_samples_per_second": 1035.457, "eval_steps_per_second": 65.027, "step": 60000 }, { "epoch": 45.437889336753386, "eval_loss": 3.327679395675659, "eval_runtime": 2.4352, "eval_samples_per_second": 941.608, "eval_steps_per_second": 59.133, "step": 62000 }, { "epoch": 46.90362770245511, "grad_norm": 4.627829551696777, "learning_rate": 6.004333333333334e-05, "loss": 3.1748, "step": 64000 }, { "epoch": 46.90362770245511, "eval_loss": 3.292200803756714, "eval_runtime": 2.358, "eval_samples_per_second": 972.453, "eval_steps_per_second": 61.07, "step": 64000 }, { "epoch": 48.369366068156836, "eval_loss": 3.2361395359039307, "eval_runtime": 2.2526, "eval_samples_per_second": 1017.92, "eval_steps_per_second": 63.925, "step": 66000 }, { "epoch": 49.835104433858554, "grad_norm": 4.962168216705322, "learning_rate": 5.338000000000001e-05, "loss": 3.0842, "step": 68000 }, { "epoch": 49.835104433858554, "eval_loss": 3.204303741455078, "eval_runtime": 2.2836, "eval_samples_per_second": 1004.135, "eval_steps_per_second": 63.059, "step": 68000 }, { "epoch": 51.30084279956028, "eval_loss": 3.187018871307373, "eval_runtime": 2.2063, "eval_samples_per_second": 1039.279, "eval_steps_per_second": 65.267, "step": 70000 }, { "epoch": 52.766581165262004, "grad_norm": 5.1038713455200195, "learning_rate": 4.671666666666667e-05, "loss": 3.0082, "step": 72000 }, { "epoch": 52.766581165262004, "eval_loss": 3.148749589920044, "eval_runtime": 2.1633, "eval_samples_per_second": 1059.95, "eval_steps_per_second": 66.565, "step": 72000 }, { "epoch": 54.23231953096372, "eval_loss": 3.1257400512695312, "eval_runtime": 2.188, "eval_samples_per_second": 1048.007, "eval_steps_per_second": 65.815, "step": 74000 }, { "epoch": 55.698057896665446, "grad_norm": 4.805927753448486, "learning_rate": 4.005333333333334e-05, "loss": 2.9483, "step": 76000 }, { "epoch": 55.698057896665446, "eval_loss": 3.1000759601593018, "eval_runtime": 2.3388, "eval_samples_per_second": 980.424, "eval_steps_per_second": 61.57, "step": 76000 }, { "epoch": 57.163796262367164, "eval_loss": 3.069366216659546, "eval_runtime": 2.2242, "eval_samples_per_second": 1030.923, "eval_steps_per_second": 64.742, "step": 78000 }, { "epoch": 58.62953462806889, "grad_norm": 5.11604642868042, "learning_rate": 3.339e-05, "loss": 2.8885, "step": 80000 }, { "epoch": 58.62953462806889, "eval_loss": 3.0604772567749023, "eval_runtime": 2.4526, "eval_samples_per_second": 934.925, "eval_steps_per_second": 58.713, "step": 80000 }, { "epoch": 60.095272993770614, "eval_loss": 3.0568008422851562, "eval_runtime": 2.3025, "eval_samples_per_second": 995.86, "eval_steps_per_second": 62.54, "step": 82000 }, { "epoch": 61.56101135947233, "grad_norm": 5.1547369956970215, "learning_rate": 2.6725e-05, "loss": 2.8416, "step": 84000 }, { "epoch": 61.56101135947233, "eval_loss": 3.0083158016204834, "eval_runtime": 2.1781, "eval_samples_per_second": 1052.731, "eval_steps_per_second": 66.111, "step": 84000 } ], "logging_steps": 4000, "max_steps": 100000, "num_input_tokens_seen": 0, "num_train_epochs": 74, "save_steps": 4000, "stateful_callbacks": { "TrainerControl": { "args": { "should_epoch_stop": false, "should_evaluate": false, "should_log": false, "should_save": true, "should_training_stop": false }, "attributes": {} } }, "total_flos": 2.619911333884723e+16, "train_batch_size": 16, "trial_name": null, "trial_params": null }