| { | |
| "best_metric": 0.45846349000930786, | |
| "best_model_checkpoint": "./vit-plant-classification/checkpoint-280", | |
| "epoch": 3.5, | |
| "global_step": 280, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.00019375000000000002, | |
| "loss": 2.4669, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0001875, | |
| "loss": 2.0664, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00018125000000000001, | |
| "loss": 1.6293, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.000175, | |
| "loss": 1.3971, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "eval_accuracy": 0.7439252336448599, | |
| "eval_loss": 1.2291492223739624, | |
| "eval_runtime": 366.9713, | |
| "eval_samples_per_second": 1.458, | |
| "eval_steps_per_second": 0.183, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00016875, | |
| "loss": 1.2067, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.00016250000000000002, | |
| "loss": 1.1033, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 0.00015625, | |
| "loss": 0.9196, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 0.00015000000000000001, | |
| "loss": 0.7434, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.8280373831775701, | |
| "eval_loss": 0.7487128973007202, | |
| "eval_runtime": 364.1928, | |
| "eval_samples_per_second": 1.469, | |
| "eval_steps_per_second": 0.184, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 0.00014375, | |
| "loss": 0.6383, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 0.0001375, | |
| "loss": 0.6098, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 0.00013125000000000002, | |
| "loss": 0.5539, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 0.000125, | |
| "loss": 0.5029, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "eval_accuracy": 0.8355140186915888, | |
| "eval_loss": 0.608514666557312, | |
| "eval_runtime": 368.7757, | |
| "eval_samples_per_second": 1.451, | |
| "eval_steps_per_second": 0.182, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 0.00011875, | |
| "loss": 0.4423, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 0.00011250000000000001, | |
| "loss": 0.513, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 0.00010625000000000001, | |
| "loss": 0.4399, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 0.0001, | |
| "loss": 0.3822, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.8579439252336448, | |
| "eval_loss": 0.48495471477508545, | |
| "eval_runtime": 366.309, | |
| "eval_samples_per_second": 1.461, | |
| "eval_steps_per_second": 0.183, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 9.375e-05, | |
| "loss": 0.3066, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 8.75e-05, | |
| "loss": 0.3012, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 8.125000000000001e-05, | |
| "loss": 0.2988, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 7.500000000000001e-05, | |
| "loss": 0.2371, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "eval_accuracy": 0.8392523364485981, | |
| "eval_loss": 0.481742799282074, | |
| "eval_runtime": 367.6014, | |
| "eval_samples_per_second": 1.455, | |
| "eval_steps_per_second": 0.182, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 6.875e-05, | |
| "loss": 0.3036, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 6.25e-05, | |
| "loss": 0.2545, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 2.88, | |
| "learning_rate": 5.6250000000000005e-05, | |
| "loss": 0.2251, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 5e-05, | |
| "loss": 0.2403, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.8373831775700935, | |
| "eval_loss": 0.531984269618988, | |
| "eval_runtime": 367.6325, | |
| "eval_samples_per_second": 1.455, | |
| "eval_steps_per_second": 0.182, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 4.375e-05, | |
| "loss": 0.1588, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "learning_rate": 3.7500000000000003e-05, | |
| "loss": 0.1976, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 3.38, | |
| "learning_rate": 3.125e-05, | |
| "loss": 0.1927, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 3.5, | |
| "learning_rate": 2.5e-05, | |
| "loss": 0.1426, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 3.5, | |
| "eval_accuracy": 0.8560747663551402, | |
| "eval_loss": 0.45846349000930786, | |
| "eval_runtime": 367.7702, | |
| "eval_samples_per_second": 1.455, | |
| "eval_steps_per_second": 0.182, | |
| "step": 280 | |
| } | |
| ], | |
| "max_steps": 320, | |
| "num_train_epochs": 4, | |
| "total_flos": 3.467364497947607e+17, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |