| { | |
| "best_metric": 0.25109806656837463, | |
| "best_model_checkpoint": "./outputs/google/flan-t5-base_02112024_16_44_41/checkpoint-6165", | |
| "epoch": 6.0, | |
| "global_step": 12330, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 4.9513381995133826e-05, | |
| "loss": 0.4755, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 4.9026763990267643e-05, | |
| "loss": 0.356, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 4.854014598540147e-05, | |
| "loss": 0.3303, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 4.8053527980535284e-05, | |
| "loss": 0.3177, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_f1": 0.5368667407893281, | |
| "eval_loss": 0.28315970301628113, | |
| "eval_precision": 0.5924271713968363, | |
| "eval_recall": 0.5446244929870324, | |
| "eval_runtime": 62.0446, | |
| "eval_samples_per_second": 14.715, | |
| "eval_steps_per_second": 1.854, | |
| "step": 2055 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 4.75669099756691e-05, | |
| "loss": 0.282, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 4.708029197080292e-05, | |
| "loss": 0.2714, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 4.659367396593674e-05, | |
| "loss": 0.2748, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 4.610705596107056e-05, | |
| "loss": 0.2716, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_f1": 0.5570042238944544, | |
| "eval_loss": 0.2516656517982483, | |
| "eval_precision": 0.6152143750680749, | |
| "eval_recall": 0.5385389128346834, | |
| "eval_runtime": 58.438, | |
| "eval_samples_per_second": 15.623, | |
| "eval_steps_per_second": 1.968, | |
| "step": 4110 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 4.5620437956204383e-05, | |
| "loss": 0.2382, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 4.51338199513382e-05, | |
| "loss": 0.2368, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 4.4647201946472024e-05, | |
| "loss": 0.2459, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 4.416058394160584e-05, | |
| "loss": 0.2347, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_f1": 0.5757671862207844, | |
| "eval_loss": 0.25109806656837463, | |
| "eval_precision": 0.6050661809680828, | |
| "eval_recall": 0.556382019200333, | |
| "eval_runtime": 56.0866, | |
| "eval_samples_per_second": 16.278, | |
| "eval_steps_per_second": 2.05, | |
| "step": 6165 | |
| }, | |
| { | |
| "epoch": 3.16, | |
| "learning_rate": 4.3673965936739665e-05, | |
| "loss": 0.2143, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 4.318734793187348e-05, | |
| "loss": 0.2079, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 4.27007299270073e-05, | |
| "loss": 0.2033, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 3.89, | |
| "learning_rate": 4.221411192214112e-05, | |
| "loss": 0.2162, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_f1": 0.5610363812916264, | |
| "eval_loss": 0.26298487186431885, | |
| "eval_precision": 0.5794654348603949, | |
| "eval_recall": 0.5633207068606695, | |
| "eval_runtime": 63.3485, | |
| "eval_samples_per_second": 14.412, | |
| "eval_steps_per_second": 1.815, | |
| "step": 8220 | |
| }, | |
| { | |
| "epoch": 4.14, | |
| "learning_rate": 4.172749391727494e-05, | |
| "loss": 0.1905, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 4.38, | |
| "learning_rate": 4.124087591240876e-05, | |
| "loss": 0.1808, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 4.62, | |
| "learning_rate": 4.075425790754258e-05, | |
| "loss": 0.1763, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 4.87, | |
| "learning_rate": 4.02676399026764e-05, | |
| "loss": 0.1788, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_f1": 0.5664179866042858, | |
| "eval_loss": 0.2608945965766907, | |
| "eval_precision": 0.5989987635784213, | |
| "eval_recall": 0.5500245894998113, | |
| "eval_runtime": 58.3428, | |
| "eval_samples_per_second": 15.649, | |
| "eval_steps_per_second": 1.971, | |
| "step": 10275 | |
| }, | |
| { | |
| "epoch": 5.11, | |
| "learning_rate": 3.978102189781022e-05, | |
| "loss": 0.1782, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 5.35, | |
| "learning_rate": 3.929440389294404e-05, | |
| "loss": 0.1604, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 5.6, | |
| "learning_rate": 3.8807785888077864e-05, | |
| "loss": 0.1534, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 5.84, | |
| "learning_rate": 3.832116788321168e-05, | |
| "loss": 0.1568, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_f1": 0.5681745597369022, | |
| "eval_loss": 0.3027566373348236, | |
| "eval_precision": 0.5835724964223891, | |
| "eval_recall": 0.5637640012101204, | |
| "eval_runtime": 58.9519, | |
| "eval_samples_per_second": 15.487, | |
| "eval_steps_per_second": 1.951, | |
| "step": 12330 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "step": 12330, | |
| "total_flos": 1.3900294734603264e+16, | |
| "train_loss": 0.23737415118809163, | |
| "train_runtime": 2537.1213, | |
| "train_samples_per_second": 80.968, | |
| "train_steps_per_second": 20.249 | |
| } | |
| ], | |
| "max_steps": 51375, | |
| "num_train_epochs": 25, | |
| "total_flos": 1.3900294734603264e+16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |