| { |
| "best_metric": 0.9343232438606511, |
| "best_model_checkpoint": ".//debugged_tr_imst_ses_udpipe_8_0.01_0.00005_10_04-23-24_03-58/checkpoint-4512", |
| "epoch": 10.0, |
| "global_step": 5640, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 1.0, |
| "learning_rate": 4.787775891341257e-05, |
| "loss": 1.4678, |
| "step": 564 |
| }, |
| { |
| "epoch": 1.0, |
| "eval_accuracy": 0.8510861948142957, |
| "eval_f1": 0.7813040969417195, |
| "eval_loss": 0.5929591655731201, |
| "eval_precision": 0.788309268747089, |
| "eval_recall": 0.774422328986502, |
| "eval_runtime": 2.5513, |
| "eval_samples_per_second": 221.062, |
| "eval_steps_per_second": 27.829, |
| "step": 564 |
| }, |
| { |
| "epoch": 2.0, |
| "learning_rate": 4.255800792303339e-05, |
| "loss": 0.4848, |
| "step": 1128 |
| }, |
| { |
| "epoch": 2.0, |
| "eval_accuracy": 0.8962859145059565, |
| "eval_f1": 0.8482901078723893, |
| "eval_loss": 0.39337363839149475, |
| "eval_precision": 0.8510246373474557, |
| "eval_recall": 0.8455730954015099, |
| "eval_runtime": 2.5796, |
| "eval_samples_per_second": 218.638, |
| "eval_steps_per_second": 27.524, |
| "step": 1128 |
| }, |
| { |
| "epoch": 3.0, |
| "learning_rate": 3.723825693265422e-05, |
| "loss": 0.2904, |
| "step": 1692 |
| }, |
| { |
| "epoch": 3.0, |
| "eval_accuracy": 0.9171338472319551, |
| "eval_f1": 0.8771331058020478, |
| "eval_loss": 0.32598426938056946, |
| "eval_precision": 0.8723693143245078, |
| "eval_recall": 0.8819492107069321, |
| "eval_runtime": 3.0713, |
| "eval_samples_per_second": 183.637, |
| "eval_steps_per_second": 23.117, |
| "step": 1692 |
| }, |
| { |
| "epoch": 4.0, |
| "learning_rate": 3.191850594227504e-05, |
| "loss": 0.1991, |
| "step": 2256 |
| }, |
| { |
| "epoch": 4.0, |
| "eval_accuracy": 0.9309740714786265, |
| "eval_f1": 0.899566309061858, |
| "eval_loss": 0.2912697494029999, |
| "eval_precision": 0.8975176497381007, |
| "eval_recall": 0.9016243422557767, |
| "eval_runtime": 2.5827, |
| "eval_samples_per_second": 218.373, |
| "eval_steps_per_second": 27.49, |
| "step": 2256 |
| }, |
| { |
| "epoch": 5.0, |
| "learning_rate": 2.659875495189587e-05, |
| "loss": 0.1269, |
| "step": 2820 |
| }, |
| { |
| "epoch": 5.0, |
| "eval_accuracy": 0.9407848633496847, |
| "eval_f1": 0.9136592051164916, |
| "eval_loss": 0.25637173652648926, |
| "eval_precision": 0.9122006841505131, |
| "eval_recall": 0.9151223976206818, |
| "eval_runtime": 2.6138, |
| "eval_samples_per_second": 215.78, |
| "eval_steps_per_second": 27.164, |
| "step": 2820 |
| }, |
| { |
| "epoch": 6.0, |
| "learning_rate": 2.1279003961516696e-05, |
| "loss": 0.0847, |
| "step": 3384 |
| }, |
| { |
| "epoch": 6.0, |
| "eval_accuracy": 0.9455150665732306, |
| "eval_f1": 0.9186232049236379, |
| "eval_loss": 0.2580364942550659, |
| "eval_precision": 0.9152850329320917, |
| "eval_recall": 0.9219858156028369, |
| "eval_runtime": 2.5405, |
| "eval_samples_per_second": 222.002, |
| "eval_steps_per_second": 27.947, |
| "step": 3384 |
| }, |
| { |
| "epoch": 7.0, |
| "learning_rate": 1.595925297113752e-05, |
| "loss": 0.0568, |
| "step": 3948 |
| }, |
| { |
| "epoch": 7.0, |
| "eval_accuracy": 0.9544498948843728, |
| "eval_f1": 0.9332264345435803, |
| "eval_loss": 0.24547669291496277, |
| "eval_precision": 0.9344036697247706, |
| "eval_recall": 0.9320521619766644, |
| "eval_runtime": 2.5538, |
| "eval_samples_per_second": 220.85, |
| "eval_steps_per_second": 27.802, |
| "step": 3948 |
| }, |
| { |
| "epoch": 8.0, |
| "learning_rate": 1.0639501980758348e-05, |
| "loss": 0.036, |
| "step": 4512 |
| }, |
| { |
| "epoch": 8.0, |
| "eval_accuracy": 0.9560266292922215, |
| "eval_f1": 0.9343232438606511, |
| "eval_loss": 0.24511215090751648, |
| "eval_precision": 0.9329379562043796, |
| "eval_recall": 0.9357126515671471, |
| "eval_runtime": 2.5419, |
| "eval_samples_per_second": 221.88, |
| "eval_steps_per_second": 27.932, |
| "step": 4512 |
| }, |
| { |
| "epoch": 9.0, |
| "learning_rate": 5.319750990379174e-06, |
| "loss": 0.0197, |
| "step": 5076 |
| }, |
| { |
| "epoch": 9.0, |
| "eval_accuracy": 0.9562018220042047, |
| "eval_f1": 0.934293223631585, |
| "eval_loss": 0.2513558268547058, |
| "eval_precision": 0.9333333333333333, |
| "eval_recall": 0.9352550903683368, |
| "eval_runtime": 2.5448, |
| "eval_samples_per_second": 221.627, |
| "eval_steps_per_second": 27.9, |
| "step": 5076 |
| }, |
| { |
| "epoch": 10.0, |
| "learning_rate": 0.0, |
| "loss": 0.0181, |
| "step": 5640 |
| }, |
| { |
| "epoch": 10.0, |
| "eval_accuracy": 0.9555010511562719, |
| "eval_f1": 0.932449422791176, |
| "eval_loss": 0.24302078783512115, |
| "eval_precision": 0.9317039744175423, |
| "eval_recall": 0.9331960649736902, |
| "eval_runtime": 3.4427, |
| "eval_samples_per_second": 163.824, |
| "eval_steps_per_second": 20.623, |
| "step": 5640 |
| }, |
| { |
| "epoch": 10.0, |
| "step": 5640, |
| "total_flos": 2809390960654344.0, |
| "train_loss": 0.2784407759389133, |
| "train_runtime": 2041.3688, |
| "train_samples_per_second": 22.088, |
| "train_steps_per_second": 2.763 |
| } |
| ], |
| "max_steps": 5640, |
| "num_train_epochs": 10, |
| "total_flos": 2809390960654344.0, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
