{ "best_metric": null, "best_model_checkpoint": null, "epoch": 10.0, "eval_steps": 500, "global_step": 5750, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.87, "learning_rate": 4.646017699115045e-05, "loss": 1.0657, "step": 500 }, { "epoch": 1.0, "eval_accuracy_Flower": 0.0, "eval_accuracy_Fruit": 0.07758079781710635, "eval_accuracy_Leaf": 0.0, "eval_accuracy_Stem": 0.0012399750759916263, "eval_accuracy_void": NaN, "eval_iou_Flower": 0.0, "eval_iou_Fruit": 0.07758070866277463, "eval_iou_Leaf": 0.0, "eval_iou_Stem": 0.0012399750759916263, "eval_iou_void": 0.0, "eval_loss": 1.1178135871887207, "eval_mean_accuracy": 0.019705193223274492, "eval_mean_iou": 0.01576413674775325, "eval_overall_accuracy": 0.02611538600789552, "eval_runtime": 242.364, "eval_samples_per_second": 5.929, "eval_steps_per_second": 0.594, "step": 575 }, { "epoch": 1.74, "learning_rate": 4.20353982300885e-05, "loss": 0.5398, "step": 1000 }, { "epoch": 2.0, "eval_accuracy_Flower": 0.0, "eval_accuracy_Fruit": 0.17832909898177904, "eval_accuracy_Leaf": 4.357268097477316e-07, "eval_accuracy_Stem": 0.0020119175696714886, "eval_accuracy_void": NaN, "eval_iou_Flower": 0.0, "eval_iou_Fruit": 0.17826792711632508, "eval_iou_Leaf": 4.357268097477316e-07, "eval_iou_Stem": 0.0020119175696714886, "eval_iou_void": 0.0, "eval_loss": 1.0816317796707153, "eval_mean_accuracy": 0.045085363069565065, "eval_mean_iou": 0.03605605608256126, "eval_overall_accuracy": 0.059915823868905815, "eval_runtime": 235.0967, "eval_samples_per_second": 6.112, "eval_steps_per_second": 0.613, "step": 1150 }, { "epoch": 2.61, "learning_rate": 3.7610619469026545e-05, "loss": 0.4794, "step": 1500 }, { "epoch": 3.0, "eval_accuracy_Flower": 0.0, "eval_accuracy_Fruit": 0.05171248760607522, "eval_accuracy_Leaf": 0.0002904845398318211, "eval_accuracy_Stem": 0.0032168043504958487, "eval_accuracy_void": NaN, "eval_iou_Flower": 0.0, "eval_iou_Fruit": 0.051712190472556895, "eval_iou_Leaf": 0.0002904845398318211, "eval_iou_Stem": 0.0032168043504958487, "eval_iou_void": 0.0, "eval_loss": 1.2122489213943481, "eval_mean_accuracy": 0.013804944124100722, "eval_mean_iou": 0.011043895872576913, "eval_overall_accuracy": 0.01788594812989718, "eval_runtime": 232.8558, "eval_samples_per_second": 6.171, "eval_steps_per_second": 0.618, "step": 1725 }, { "epoch": 3.48, "learning_rate": 3.3185840707964604e-05, "loss": 0.446, "step": 2000 }, { "epoch": 4.0, "eval_accuracy_Flower": 0.0, "eval_accuracy_Fruit": 0.056192687434985046, "eval_accuracy_Leaf": 0.0, "eval_accuracy_Stem": 0.0006151770459164299, "eval_accuracy_void": NaN, "eval_iou_Flower": 0.0, "eval_iou_Fruit": 0.056192674519866245, "eval_iou_Leaf": 0.0, "eval_iou_Stem": 0.0006151770459164299, "eval_iou_void": 0.0, "eval_loss": 1.3138337135314941, "eval_mean_accuracy": 0.01420196612022537, "eval_mean_iou": 0.011361570313156535, "eval_overall_accuracy": 0.018877250329402743, "eval_runtime": 230.6703, "eval_samples_per_second": 6.23, "eval_steps_per_second": 0.624, "step": 2300 }, { "epoch": 4.35, "learning_rate": 2.8761061946902656e-05, "loss": 0.4422, "step": 2500 }, { "epoch": 5.0, "eval_accuracy_Flower": 0.0, "eval_accuracy_Fruit": 0.0012236486887147139, "eval_accuracy_Leaf": 0.0, "eval_accuracy_Stem": 0.0012965691004549593, "eval_accuracy_void": NaN, "eval_iou_Flower": 0.0, "eval_iou_Fruit": 0.0012236486887147139, "eval_iou_Leaf": 0.0, "eval_iou_Stem": 0.0012965691004549593, "eval_iou_void": 0.0, "eval_loss": 1.2360085248947144, "eval_mean_accuracy": 0.0006300544472924183, "eval_mean_iou": 0.0005040435578339347, "eval_overall_accuracy": 0.0005853571843389319, "eval_runtime": 238.302, "eval_samples_per_second": 6.03, "eval_steps_per_second": 0.604, "step": 2875 }, { "epoch": 5.22, "learning_rate": 2.433628318584071e-05, "loss": 0.4183, "step": 3000 }, { "epoch": 6.0, "eval_accuracy_Flower": 0.0, "eval_accuracy_Fruit": 0.08851296345923922, "eval_accuracy_Leaf": 0.0, "eval_accuracy_Stem": 0.0, "eval_accuracy_void": NaN, "eval_iou_Flower": 0.0, "eval_iou_Fruit": 0.08850915938902014, "eval_iou_Leaf": 0.0, "eval_iou_Stem": 0.0, "eval_iou_void": 0.0, "eval_loss": 1.3598012924194336, "eval_mean_accuracy": 0.022128240864809804, "eval_mean_iou": 0.01770183187780403, "eval_overall_accuracy": 0.029603238520330803, "eval_runtime": 239.5023, "eval_samples_per_second": 6.0, "eval_steps_per_second": 0.601, "step": 3450 }, { "epoch": 6.09, "learning_rate": 1.991150442477876e-05, "loss": 0.403, "step": 3500 }, { "epoch": 6.96, "learning_rate": 1.5486725663716813e-05, "loss": 0.3921, "step": 4000 }, { "epoch": 7.0, "eval_accuracy_Flower": 0.0, "eval_accuracy_Fruit": 0.13330944567611186, "eval_accuracy_Leaf": 4.909188723157776e-05, "eval_accuracy_Stem": 0.0, "eval_accuracy_void": NaN, "eval_iou_Flower": 0.0, "eval_iou_Fruit": 0.13328089591240883, "eval_iou_Leaf": 4.909188723157776e-05, "eval_iou_Stem": 0.0, "eval_iou_void": 0.0, "eval_loss": 1.252343773841858, "eval_mean_accuracy": 0.03333963439083586, "eval_mean_iou": 0.02666599755992808, "eval_overall_accuracy": 0.044611443129532434, "eval_runtime": 238.2111, "eval_samples_per_second": 6.032, "eval_steps_per_second": 0.605, "step": 4025 }, { "epoch": 7.83, "learning_rate": 1.1061946902654869e-05, "loss": 0.3743, "step": 4500 }, { "epoch": 8.0, "eval_accuracy_Flower": 0.0, "eval_accuracy_Fruit": 0.20021158733712072, "eval_accuracy_Leaf": 0.0, "eval_accuracy_Stem": 0.0004923680128309972, "eval_accuracy_void": NaN, "eval_iou_Flower": 0.0, "eval_iou_Fruit": 0.200070693842378, "eval_iou_Leaf": 0.0, "eval_iou_Stem": 0.0004923680128309972, "eval_iou_void": 0.0, "eval_loss": 1.3145976066589355, "eval_mean_accuracy": 0.05017598883748793, "eval_mean_iou": 0.0401126123710418, "eval_overall_accuracy": 0.067027817572415, "eval_runtime": 237.3143, "eval_samples_per_second": 6.055, "eval_steps_per_second": 0.607, "step": 4600 }, { "epoch": 8.7, "learning_rate": 6.6371681415929215e-06, "loss": 0.3695, "step": 5000 }, { "epoch": 9.0, "eval_accuracy_Flower": 0.0, "eval_accuracy_Fruit": 0.14719822603117225, "eval_accuracy_Leaf": 0.0, "eval_accuracy_Stem": 7.300629155769959e-05, "eval_accuracy_void": NaN, "eval_iou_Flower": 0.0, "eval_iou_Fruit": 0.1471454001898686, "eval_iou_Leaf": 0.0, "eval_iou_Stem": 7.300629155769959e-05, "eval_iou_void": 0.0, "eval_loss": 1.2872849702835083, "eval_mean_accuracy": 0.03681780808068249, "eval_mean_iou": 0.029443681296285263, "eval_overall_accuracy": 0.04924049232728562, "eval_runtime": 239.2613, "eval_samples_per_second": 6.006, "eval_steps_per_second": 0.602, "step": 5175 }, { "epoch": 9.57, "learning_rate": 2.2123893805309734e-06, "loss": 0.3796, "step": 5500 }, { "epoch": 10.0, "eval_accuracy_Flower": 0.0, "eval_accuracy_Fruit": 0.12991062583976454, "eval_accuracy_Leaf": 0.0, "eval_accuracy_Stem": 9.56439013430328e-05, "eval_accuracy_void": NaN, "eval_iou_Flower": 0.0, "eval_iou_Fruit": 0.12988737047880816, "eval_iou_Leaf": 0.0, "eval_iou_Stem": 9.56439013430328e-05, "eval_iou_void": 0.0, "eval_loss": 1.20041024684906, "eval_mean_accuracy": 0.03250156743527689, "eval_mean_iou": 0.02599660287603024, "eval_overall_accuracy": 0.0434617139888683, "eval_runtime": 240.3818, "eval_samples_per_second": 5.978, "eval_steps_per_second": 0.599, "step": 5750 }, { "epoch": 10.0, "step": 5750, "total_flos": 8.429846724478771e+18, "train_loss": 0.4772091090990149, "train_runtime": 7651.7496, "train_samples_per_second": 7.508, "train_steps_per_second": 0.751 } ], "logging_steps": 500, "max_steps": 5750, "num_input_tokens_seen": 0, "num_train_epochs": 10, "save_steps": 500, "total_flos": 8.429846724478771e+18, "train_batch_size": 10, "trial_name": null, "trial_params": null }