| { | |
| "best_metric": 0.951527777777778, | |
| "best_model_checkpoint": "/n/holyscratch01/economics/esilcock/syria_models/8_5e-06/checkpoint-740", | |
| "epoch": 17.61904761904762, | |
| "global_step": 740, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.48, | |
| "eval_accuracy": 0.18055555555555555, | |
| "eval_f1": 0.05162907268170425, | |
| "eval_loss": 2.1331899166107178, | |
| "eval_precision": 0.05357142857142857, | |
| "eval_recall": 0.08588235294117648, | |
| "eval_runtime": 2.9184, | |
| "eval_samples_per_second": 24.671, | |
| "eval_steps_per_second": 3.084, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "eval_accuracy": 0.25, | |
| "eval_f1": 0.10818399044205496, | |
| "eval_loss": 2.088547468185425, | |
| "eval_precision": 0.09869674185463659, | |
| "eval_recall": 0.12676470588235295, | |
| "eval_runtime": 2.6064, | |
| "eval_samples_per_second": 27.625, | |
| "eval_steps_per_second": 3.453, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "eval_accuracy": 0.2777777777777778, | |
| "eval_f1": 0.09715938834550213, | |
| "eval_loss": 1.9733208417892456, | |
| "eval_precision": 0.08442028985507247, | |
| "eval_recall": 0.13751633986928105, | |
| "eval_runtime": 2.5925, | |
| "eval_samples_per_second": 27.773, | |
| "eval_steps_per_second": 3.472, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "eval_accuracy": 0.5555555555555556, | |
| "eval_f1": 0.22415689149560114, | |
| "eval_loss": 1.8020447492599487, | |
| "eval_precision": 0.21013071895424837, | |
| "eval_recall": 0.26833333333333337, | |
| "eval_runtime": 2.5848, | |
| "eval_samples_per_second": 27.855, | |
| "eval_steps_per_second": 3.482, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "eval_accuracy": 0.5833333333333334, | |
| "eval_f1": 0.2961904761904762, | |
| "eval_loss": 1.4616996049880981, | |
| "eval_precision": 0.3290064102564102, | |
| "eval_recall": 0.335686274509804, | |
| "eval_runtime": 2.9293, | |
| "eval_samples_per_second": 24.579, | |
| "eval_steps_per_second": 3.072, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "eval_accuracy": 0.6527777777777778, | |
| "eval_f1": 0.43249639249639255, | |
| "eval_loss": 1.2280499935150146, | |
| "eval_precision": 0.4880341880341881, | |
| "eval_recall": 0.4904901960784313, | |
| "eval_runtime": 2.6424, | |
| "eval_samples_per_second": 27.248, | |
| "eval_steps_per_second": 3.406, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 3.33, | |
| "eval_accuracy": 0.75, | |
| "eval_f1": 0.6376634768740032, | |
| "eval_loss": 0.9447543025016785, | |
| "eval_precision": 0.7244047619047619, | |
| "eval_recall": 0.6264705882352941, | |
| "eval_runtime": 2.6104, | |
| "eval_samples_per_second": 27.582, | |
| "eval_steps_per_second": 3.448, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 3.81, | |
| "eval_accuracy": 0.8055555555555556, | |
| "eval_f1": 0.6916894699503395, | |
| "eval_loss": 0.7263244986534119, | |
| "eval_precision": 0.7515151515151516, | |
| "eval_recall": 0.6826797385620915, | |
| "eval_runtime": 2.6192, | |
| "eval_samples_per_second": 27.489, | |
| "eval_steps_per_second": 3.436, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 4.29, | |
| "eval_accuracy": 0.8611111111111112, | |
| "eval_f1": 0.780800727050727, | |
| "eval_loss": 0.6345593333244324, | |
| "eval_precision": 0.7710130718954249, | |
| "eval_recall": 0.826797385620915, | |
| "eval_runtime": 2.6126, | |
| "eval_samples_per_second": 27.559, | |
| "eval_steps_per_second": 3.445, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 4.76, | |
| "eval_accuracy": 0.875, | |
| "eval_f1": 0.8041340603840604, | |
| "eval_loss": 0.5365733504295349, | |
| "eval_precision": 0.8055159958720332, | |
| "eval_recall": 0.8326797385620915, | |
| "eval_runtime": 2.7985, | |
| "eval_samples_per_second": 25.728, | |
| "eval_steps_per_second": 3.216, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 5.24, | |
| "eval_accuracy": 0.875, | |
| "eval_f1": 0.8001352813852813, | |
| "eval_loss": 0.41519981622695923, | |
| "eval_precision": 0.800515995872033, | |
| "eval_recall": 0.8187908496732026, | |
| "eval_runtime": 2.6402, | |
| "eval_samples_per_second": 27.271, | |
| "eval_steps_per_second": 3.409, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 5.71, | |
| "eval_accuracy": 0.8472222222222222, | |
| "eval_f1": 0.774096712111418, | |
| "eval_loss": 0.41259926557540894, | |
| "eval_precision": 0.7647058823529411, | |
| "eval_recall": 0.8209150326797385, | |
| "eval_runtime": 2.6387, | |
| "eval_samples_per_second": 27.286, | |
| "eval_steps_per_second": 3.411, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 6.19, | |
| "eval_accuracy": 0.875, | |
| "eval_f1": 0.7898691749427044, | |
| "eval_loss": 0.34084904193878174, | |
| "eval_precision": 0.7705882352941177, | |
| "eval_recall": 0.8379084967320262, | |
| "eval_runtime": 2.7068, | |
| "eval_samples_per_second": 26.6, | |
| "eval_steps_per_second": 3.325, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 6.67, | |
| "eval_accuracy": 0.8611111111111112, | |
| "eval_f1": 0.7678326775480475, | |
| "eval_loss": 0.39358699321746826, | |
| "eval_precision": 0.7325162337662338, | |
| "eval_recall": 0.8477124183006536, | |
| "eval_runtime": 2.6003, | |
| "eval_samples_per_second": 27.689, | |
| "eval_steps_per_second": 3.461, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 7.14, | |
| "eval_accuracy": 0.8888888888888888, | |
| "eval_f1": 0.8779316203509753, | |
| "eval_loss": 0.2763499915599823, | |
| "eval_precision": 0.8507142857142858, | |
| "eval_recall": 0.9372549019607843, | |
| "eval_runtime": 2.6589, | |
| "eval_samples_per_second": 27.078, | |
| "eval_steps_per_second": 3.385, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 7.62, | |
| "eval_accuracy": 0.9305555555555556, | |
| "eval_f1": 0.9120108649520414, | |
| "eval_loss": 0.217352032661438, | |
| "eval_precision": 0.8800000000000001, | |
| "eval_recall": 0.9653594771241831, | |
| "eval_runtime": 2.6648, | |
| "eval_samples_per_second": 27.019, | |
| "eval_steps_per_second": 3.377, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 8.1, | |
| "eval_accuracy": 0.8888888888888888, | |
| "eval_f1": 0.8755410011444494, | |
| "eval_loss": 0.22454272210597992, | |
| "eval_precision": 0.8466666666666667, | |
| "eval_recall": 0.9424836601307189, | |
| "eval_runtime": 2.6406, | |
| "eval_samples_per_second": 27.267, | |
| "eval_steps_per_second": 3.408, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 8.57, | |
| "eval_accuracy": 0.9444444444444444, | |
| "eval_f1": 0.9410636043273615, | |
| "eval_loss": 0.19036133587360382, | |
| "eval_precision": 0.925, | |
| "eval_recall": 0.9712418300653596, | |
| "eval_runtime": 2.6164, | |
| "eval_samples_per_second": 27.519, | |
| "eval_steps_per_second": 3.44, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 9.05, | |
| "eval_accuracy": 0.9444444444444444, | |
| "eval_f1": 0.9299524932162504, | |
| "eval_loss": 0.20558035373687744, | |
| "eval_precision": 0.905, | |
| "eval_recall": 0.9712418300653596, | |
| "eval_runtime": 2.6164, | |
| "eval_samples_per_second": 27.519, | |
| "eval_steps_per_second": 3.44, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 9.52, | |
| "eval_accuracy": 0.9444444444444444, | |
| "eval_f1": 0.9299524932162504, | |
| "eval_loss": 0.20008066296577454, | |
| "eval_precision": 0.905, | |
| "eval_recall": 0.9712418300653596, | |
| "eval_runtime": 2.5909, | |
| "eval_samples_per_second": 27.79, | |
| "eval_steps_per_second": 3.474, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_accuracy": 0.9444444444444444, | |
| "eval_f1": 0.9410636043273615, | |
| "eval_loss": 0.17470227181911469, | |
| "eval_precision": 0.925, | |
| "eval_recall": 0.9712418300653596, | |
| "eval_runtime": 2.8957, | |
| "eval_samples_per_second": 24.864, | |
| "eval_steps_per_second": 3.108, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 10.48, | |
| "eval_accuracy": 0.9583333333333334, | |
| "eval_f1": 0.9476656268568033, | |
| "eval_loss": 0.11614257842302322, | |
| "eval_precision": 0.93, | |
| "eval_recall": 0.977124183006536, | |
| "eval_runtime": 2.6154, | |
| "eval_samples_per_second": 27.53, | |
| "eval_steps_per_second": 3.441, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 10.95, | |
| "eval_accuracy": 0.9305555555555556, | |
| "eval_f1": 0.9205122655122654, | |
| "eval_loss": 0.2461088001728058, | |
| "eval_precision": 0.8983333333333334, | |
| "eval_recall": 0.9601307189542485, | |
| "eval_runtime": 2.6479, | |
| "eval_samples_per_second": 27.191, | |
| "eval_steps_per_second": 3.399, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 11.43, | |
| "eval_accuracy": 0.9305555555555556, | |
| "eval_f1": 0.9205122655122654, | |
| "eval_loss": 0.20940905809402466, | |
| "eval_precision": 0.8983333333333334, | |
| "eval_recall": 0.9601307189542485, | |
| "eval_runtime": 2.5922, | |
| "eval_samples_per_second": 27.776, | |
| "eval_steps_per_second": 3.472, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 11.9, | |
| "learning_rate": 3.015873015873016e-06, | |
| "loss": 0.6157, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 11.9, | |
| "eval_accuracy": 0.9444444444444444, | |
| "eval_f1": 0.9299524932162504, | |
| "eval_loss": 0.19723846018314362, | |
| "eval_precision": 0.905, | |
| "eval_recall": 0.9712418300653596, | |
| "eval_runtime": 2.6516, | |
| "eval_samples_per_second": 27.154, | |
| "eval_steps_per_second": 3.394, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 12.38, | |
| "eval_accuracy": 0.9305555555555556, | |
| "eval_f1": 0.9120108649520414, | |
| "eval_loss": 0.20041564106941223, | |
| "eval_precision": 0.8800000000000001, | |
| "eval_recall": 0.9653594771241831, | |
| "eval_runtime": 2.6152, | |
| "eval_samples_per_second": 27.531, | |
| "eval_steps_per_second": 3.441, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 12.86, | |
| "eval_accuracy": 0.9305555555555556, | |
| "eval_f1": 0.9120108649520414, | |
| "eval_loss": 0.1812739223241806, | |
| "eval_precision": 0.8800000000000001, | |
| "eval_recall": 0.9653594771241831, | |
| "eval_runtime": 2.6055, | |
| "eval_samples_per_second": 27.634, | |
| "eval_steps_per_second": 3.454, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 13.33, | |
| "eval_accuracy": 0.9583333333333334, | |
| "eval_f1": 0.9476656268568033, | |
| "eval_loss": 0.16734400391578674, | |
| "eval_precision": 0.93, | |
| "eval_recall": 0.977124183006536, | |
| "eval_runtime": 2.6389, | |
| "eval_samples_per_second": 27.284, | |
| "eval_steps_per_second": 3.411, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 13.81, | |
| "eval_accuracy": 0.9444444444444444, | |
| "eval_f1": 0.9382070707070707, | |
| "eval_loss": 0.17178066074848175, | |
| "eval_precision": 0.9237500000000001, | |
| "eval_recall": 0.966013071895425, | |
| "eval_runtime": 2.7026, | |
| "eval_samples_per_second": 26.641, | |
| "eval_steps_per_second": 3.33, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 14.29, | |
| "eval_accuracy": 0.9444444444444444, | |
| "eval_f1": 0.9382070707070707, | |
| "eval_loss": 0.15890704095363617, | |
| "eval_precision": 0.9237500000000001, | |
| "eval_recall": 0.966013071895425, | |
| "eval_runtime": 2.605, | |
| "eval_samples_per_second": 27.639, | |
| "eval_steps_per_second": 3.455, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 14.76, | |
| "eval_accuracy": 0.9583333333333334, | |
| "eval_f1": 0.9476656268568033, | |
| "eval_loss": 0.131246879696846, | |
| "eval_precision": 0.93, | |
| "eval_recall": 0.977124183006536, | |
| "eval_runtime": 2.6046, | |
| "eval_samples_per_second": 27.643, | |
| "eval_steps_per_second": 3.455, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 15.24, | |
| "eval_accuracy": 0.9583333333333334, | |
| "eval_f1": 0.9476656268568033, | |
| "eval_loss": 0.1490662544965744, | |
| "eval_precision": 0.93, | |
| "eval_recall": 0.977124183006536, | |
| "eval_runtime": 2.5887, | |
| "eval_samples_per_second": 27.813, | |
| "eval_steps_per_second": 3.477, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 15.71, | |
| "eval_accuracy": 0.9444444444444444, | |
| "eval_f1": 0.9382070707070707, | |
| "eval_loss": 0.22067798674106598, | |
| "eval_precision": 0.9237500000000001, | |
| "eval_recall": 0.966013071895425, | |
| "eval_runtime": 2.6374, | |
| "eval_samples_per_second": 27.299, | |
| "eval_steps_per_second": 3.412, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 16.19, | |
| "eval_accuracy": 0.9583333333333334, | |
| "eval_f1": 0.9476656268568033, | |
| "eval_loss": 0.22119665145874023, | |
| "eval_precision": 0.93, | |
| "eval_recall": 0.977124183006536, | |
| "eval_runtime": 2.6042, | |
| "eval_samples_per_second": 27.647, | |
| "eval_steps_per_second": 3.456, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 16.67, | |
| "eval_accuracy": 0.9583333333333334, | |
| "eval_f1": 0.9476656268568033, | |
| "eval_loss": 0.2090146392583847, | |
| "eval_precision": 0.93, | |
| "eval_recall": 0.977124183006536, | |
| "eval_runtime": 2.6202, | |
| "eval_samples_per_second": 27.478, | |
| "eval_steps_per_second": 3.435, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 17.14, | |
| "eval_accuracy": 0.9583333333333334, | |
| "eval_f1": 0.9476656268568033, | |
| "eval_loss": 0.18439681828022003, | |
| "eval_precision": 0.93, | |
| "eval_recall": 0.977124183006536, | |
| "eval_runtime": 2.5985, | |
| "eval_samples_per_second": 27.708, | |
| "eval_steps_per_second": 3.464, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 17.62, | |
| "eval_accuracy": 0.9583333333333334, | |
| "eval_f1": 0.951527777777778, | |
| "eval_loss": 0.17655248939990997, | |
| "eval_precision": 0.9355555555555556, | |
| "eval_recall": 0.977124183006536, | |
| "eval_runtime": 2.893, | |
| "eval_samples_per_second": 24.888, | |
| "eval_steps_per_second": 3.111, | |
| "step": 740 | |
| } | |
| ], | |
| "max_steps": 1260, | |
| "num_train_epochs": 30, | |
| "total_flos": 5517182798069760.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |