| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 1.8777365254128793, | |
| "global_step": 11000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.00025, | |
| "loss": 1.6044, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0005, | |
| "loss": 0.957, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "eval_loss": 0.7217180132865906, | |
| "eval_phone_accuracy": 0.7656547891757681, | |
| "eval_runtime": 9.9001, | |
| "eval_samples_per_second": 20.202, | |
| "eval_steps_per_second": 1.717, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0004766703994027622, | |
| "loss": 0.8156, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0004533407988055245, | |
| "loss": 0.7327, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "eval_loss": 0.5681599974632263, | |
| "eval_phone_accuracy": 0.8123568272172383, | |
| "eval_runtime": 9.9031, | |
| "eval_samples_per_second": 20.196, | |
| "eval_steps_per_second": 1.717, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00043001119820828664, | |
| "loss": 0.6781, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0004066815976110489, | |
| "loss": 0.6426, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "eval_loss": 0.4999035894870758, | |
| "eval_phone_accuracy": 0.8317913778579135, | |
| "eval_runtime": 9.8493, | |
| "eval_samples_per_second": 20.306, | |
| "eval_steps_per_second": 1.726, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.0003833519970138111, | |
| "loss": 0.6148, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.0003600223964165734, | |
| "loss": 0.5924, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "eval_loss": 0.4667314887046814, | |
| "eval_phone_accuracy": 0.8421219869056944, | |
| "eval_runtime": 10.1485, | |
| "eval_samples_per_second": 19.707, | |
| "eval_steps_per_second": 1.675, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.0003366927958193356, | |
| "loss": 0.5743, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 0.0003133631952220978, | |
| "loss": 0.559, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "eval_loss": 0.43349266052246094, | |
| "eval_phone_accuracy": 0.8515954066476509, | |
| "eval_runtime": 10.2876, | |
| "eval_samples_per_second": 19.441, | |
| "eval_steps_per_second": 1.652, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 0.00029003359462486, | |
| "loss": 0.5442, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 0.00026670399402762223, | |
| "loss": 0.5294, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "eval_loss": 0.4208233654499054, | |
| "eval_phone_accuracy": 0.8561843252590042, | |
| "eval_runtime": 10.0364, | |
| "eval_samples_per_second": 19.928, | |
| "eval_steps_per_second": 1.694, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 0.00024337439343038447, | |
| "loss": 0.5128, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 0.0002200447928331467, | |
| "loss": 0.5037, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "eval_loss": 0.40133553743362427, | |
| "eval_phone_accuracy": 0.8622585460295878, | |
| "eval_runtime": 10.294, | |
| "eval_samples_per_second": 19.429, | |
| "eval_steps_per_second": 1.651, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 0.00019671519223590892, | |
| "loss": 0.497, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 0.00017338559163867115, | |
| "loss": 0.4918, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "eval_loss": 0.3905256986618042, | |
| "eval_phone_accuracy": 0.8658424840754918, | |
| "eval_runtime": 10.3057, | |
| "eval_samples_per_second": 19.407, | |
| "eval_steps_per_second": 1.65, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 0.0001500559910414334, | |
| "loss": 0.4832, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 0.0001267263904441956, | |
| "loss": 0.475, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "eval_loss": 0.37418654561042786, | |
| "eval_phone_accuracy": 0.8692195143579209, | |
| "eval_runtime": 10.1391, | |
| "eval_samples_per_second": 19.726, | |
| "eval_steps_per_second": 1.677, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 0.00010339678984695783, | |
| "loss": 0.4684, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 8.006718924972005e-05, | |
| "loss": 0.4647, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "eval_loss": 0.36982452869415283, | |
| "eval_phone_accuracy": 0.8727960628408435, | |
| "eval_runtime": 9.9379, | |
| "eval_samples_per_second": 20.125, | |
| "eval_steps_per_second": 1.711, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 5.678424785367675e-05, | |
| "loss": 0.4579, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 3.345464725643897e-05, | |
| "loss": 0.4556, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "eval_loss": 0.3634908199310303, | |
| "eval_phone_accuracy": 0.875249397750617, | |
| "eval_runtime": 10.1013, | |
| "eval_samples_per_second": 19.799, | |
| "eval_steps_per_second": 1.683, | |
| "step": 11000 | |
| } | |
| ], | |
| "max_steps": 11716, | |
| "num_train_epochs": 2, | |
| "total_flos": 1.8523977459584336e+19, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |