| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 3.5169988276670576, | |
| "global_step": 3000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 5e-05, | |
| "loss": 0.5593, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 5e-05, | |
| "loss": 0.3226, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 5e-05, | |
| "loss": 0.2389, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "eval_loss": 0.22972644865512848, | |
| "eval_mean_iou": 0.5079903689346261, | |
| "eval_runtime": 224.7647, | |
| "eval_samples_per_second": 0.948, | |
| "eval_steps_per_second": 0.948, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 5e-05, | |
| "loss": 0.2187, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 5e-05, | |
| "loss": 0.1961, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 5e-05, | |
| "loss": 0.1873, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "eval_loss": 0.18838702142238617, | |
| "eval_mean_iou": 0.6565456095704835, | |
| "eval_runtime": 223.134, | |
| "eval_samples_per_second": 0.955, | |
| "eval_steps_per_second": 0.955, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 5e-05, | |
| "loss": 0.172, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 5e-05, | |
| "loss": 0.2013, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 5e-05, | |
| "loss": 0.1549, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "eval_loss": 0.1733293980360031, | |
| "eval_mean_iou": 0.6822287052114181, | |
| "eval_runtime": 226.0674, | |
| "eval_samples_per_second": 0.942, | |
| "eval_steps_per_second": 0.942, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 5e-05, | |
| "loss": 0.1401, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 5e-05, | |
| "loss": 0.1365, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 5e-05, | |
| "loss": 0.151, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "eval_loss": 0.17005586624145508, | |
| "eval_mean_iou": 0.6851204106483545, | |
| "eval_runtime": 217.995, | |
| "eval_samples_per_second": 0.977, | |
| "eval_steps_per_second": 0.977, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 5e-05, | |
| "loss": 0.1461, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 5e-05, | |
| "loss": 0.1385, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 5e-05, | |
| "loss": 0.1737, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "eval_loss": 0.16983915865421295, | |
| "eval_mean_iou": 0.6635217840293423, | |
| "eval_runtime": 232.534, | |
| "eval_samples_per_second": 0.916, | |
| "eval_steps_per_second": 0.916, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 5e-05, | |
| "loss": 0.1277, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 5e-05, | |
| "loss": 0.1366, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_loss": 0.13025489449501038, | |
| "eval_mean_iou": 0.7721716611246382, | |
| "eval_runtime": 227.8304, | |
| "eval_samples_per_second": 0.935, | |
| "eval_steps_per_second": 0.935, | |
| "step": 1706 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 5e-05, | |
| "loss": 0.1235, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 5e-05, | |
| "loss": 0.1202, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 5e-05, | |
| "loss": 0.1199, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 5e-05, | |
| "loss": 0.1254, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 5e-05, | |
| "loss": 0.1214, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 5e-05, | |
| "loss": 0.1228, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 5e-05, | |
| "loss": 0.123, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 5e-05, | |
| "loss": 0.1113, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_loss": 0.1386958658695221, | |
| "eval_mean_iou": 0.7781228705976458, | |
| "eval_runtime": 228.3799, | |
| "eval_samples_per_second": 0.933, | |
| "eval_steps_per_second": 0.933, | |
| "step": 2559 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 5e-05, | |
| "loss": 0.1162, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 5e-05, | |
| "loss": 0.0971, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 3.28, | |
| "learning_rate": 5e-05, | |
| "loss": 0.097, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 3.4, | |
| "learning_rate": 5e-05, | |
| "loss": 0.091, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 3.52, | |
| "learning_rate": 5e-05, | |
| "loss": 0.0887, | |
| "step": 3000 | |
| } | |
| ], | |
| "max_steps": 170600, | |
| "num_train_epochs": 200, | |
| "total_flos": 6.68503279927296e+17, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |