| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 14.992481203007518, | |
| "global_step": 990, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.75, | |
| "eval_loss": 4.109964847564697, | |
| "eval_runtime": 6.6506, | |
| "eval_samples_per_second": 39.695, | |
| "eval_steps_per_second": 4.962, | |
| "eval_wer": 1.0, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "eval_loss": 3.392470121383667, | |
| "eval_runtime": 6.6261, | |
| "eval_samples_per_second": 39.842, | |
| "eval_steps_per_second": 4.98, | |
| "eval_wer": 1.0, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "eval_loss": 3.1406397819519043, | |
| "eval_runtime": 6.687, | |
| "eval_samples_per_second": 39.479, | |
| "eval_steps_per_second": 4.935, | |
| "eval_wer": 1.008097165991903, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "eval_loss": 3.0505788326263428, | |
| "eval_runtime": 6.6386, | |
| "eval_samples_per_second": 39.767, | |
| "eval_steps_per_second": 4.971, | |
| "eval_wer": 1.2834008097165992, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 3.78, | |
| "eval_loss": 2.8735897541046143, | |
| "eval_runtime": 6.5891, | |
| "eval_samples_per_second": 40.066, | |
| "eval_steps_per_second": 5.008, | |
| "eval_wer": 1.0323886639676114, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 4.54, | |
| "eval_loss": 2.7715253829956055, | |
| "eval_runtime": 6.6425, | |
| "eval_samples_per_second": 39.744, | |
| "eval_steps_per_second": 4.968, | |
| "eval_wer": 1.1437246963562753, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 5.3, | |
| "eval_loss": 2.5352094173431396, | |
| "eval_runtime": 6.6247, | |
| "eval_samples_per_second": 39.851, | |
| "eval_steps_per_second": 4.981, | |
| "eval_wer": 1.0101214574898785, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 6.06, | |
| "eval_loss": 2.0723843574523926, | |
| "eval_runtime": 6.6216, | |
| "eval_samples_per_second": 39.869, | |
| "eval_steps_per_second": 4.984, | |
| "eval_wer": 1.0263157894736843, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 6.81, | |
| "eval_loss": 1.4460173845291138, | |
| "eval_runtime": 6.6321, | |
| "eval_samples_per_second": 39.806, | |
| "eval_steps_per_second": 4.976, | |
| "eval_wer": 0.9696356275303644, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 7.57, | |
| "learning_rate": 0.00016584269662921346, | |
| "loss": 3.4597, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 7.57, | |
| "eval_loss": 1.1178003549575806, | |
| "eval_runtime": 6.6144, | |
| "eval_samples_per_second": 39.913, | |
| "eval_steps_per_second": 4.989, | |
| "eval_wer": 0.8461538461538461, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 8.33, | |
| "eval_loss": 0.950170636177063, | |
| "eval_runtime": 6.8415, | |
| "eval_samples_per_second": 38.588, | |
| "eval_steps_per_second": 4.823, | |
| "eval_wer": 0.8461538461538461, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 9.09, | |
| "eval_loss": 0.8674038052558899, | |
| "eval_runtime": 6.727, | |
| "eval_samples_per_second": 39.245, | |
| "eval_steps_per_second": 4.906, | |
| "eval_wer": 0.7894736842105263, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 9.84, | |
| "eval_loss": 0.7972543239593506, | |
| "eval_runtime": 6.6391, | |
| "eval_samples_per_second": 39.765, | |
| "eval_steps_per_second": 4.971, | |
| "eval_wer": 0.7692307692307693, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 10.6, | |
| "eval_loss": 0.7665630578994751, | |
| "eval_runtime": 6.8502, | |
| "eval_samples_per_second": 38.539, | |
| "eval_steps_per_second": 4.817, | |
| "eval_wer": 0.7753036437246964, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 11.36, | |
| "eval_loss": 0.7286367416381836, | |
| "eval_runtime": 8.3531, | |
| "eval_samples_per_second": 31.605, | |
| "eval_steps_per_second": 3.951, | |
| "eval_wer": 0.7530364372469636, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 12.12, | |
| "eval_loss": 0.7024735808372498, | |
| "eval_runtime": 6.7477, | |
| "eval_samples_per_second": 39.125, | |
| "eval_steps_per_second": 4.891, | |
| "eval_wer": 0.7307692307692307, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 12.87, | |
| "eval_loss": 0.6833031177520752, | |
| "eval_runtime": 6.8878, | |
| "eval_samples_per_second": 38.328, | |
| "eval_steps_per_second": 4.791, | |
| "eval_wer": 0.7348178137651822, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 13.63, | |
| "eval_loss": 0.669713020324707, | |
| "eval_runtime": 6.7425, | |
| "eval_samples_per_second": 39.154, | |
| "eval_steps_per_second": 4.894, | |
| "eval_wer": 0.7226720647773279, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 14.39, | |
| "eval_loss": 0.6649630069732666, | |
| "eval_runtime": 6.8935, | |
| "eval_samples_per_second": 38.297, | |
| "eval_steps_per_second": 4.787, | |
| "eval_wer": 0.7368421052631579, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 14.99, | |
| "step": 990, | |
| "total_flos": 1.166783970335839e+18, | |
| "train_loss": 2.406289642025726, | |
| "train_runtime": 1161.5479, | |
| "train_samples_per_second": 27.377, | |
| "train_steps_per_second": 0.852 | |
| } | |
| ], | |
| "max_steps": 990, | |
| "num_train_epochs": 15, | |
| "total_flos": 1.166783970335839e+18, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |