| { | |
| "best_metric": 0.8784255981445312, | |
| "best_model_checkpoint": "../Modelos/mt0_QG_SQAC_SQuAD 17-11-22_10:03:48/checkpoint-1200", | |
| "epoch": 2.9976617303195634, | |
| "global_step": 1200, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.12, | |
| "eval_loss": 1.0145061016082764, | |
| "eval_runtime": 65.3999, | |
| "eval_samples_per_second": 190.122, | |
| "eval_steps_per_second": 11.896, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "eval_loss": 0.9660477638244629, | |
| "eval_runtime": 65.4002, | |
| "eval_samples_per_second": 190.122, | |
| "eval_steps_per_second": 11.896, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "eval_loss": 0.9502729773521423, | |
| "eval_runtime": 65.4029, | |
| "eval_samples_per_second": 190.114, | |
| "eval_steps_per_second": 11.895, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "eval_loss": 0.9393932819366455, | |
| "eval_runtime": 65.4026, | |
| "eval_samples_per_second": 190.115, | |
| "eval_steps_per_second": 11.896, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "eval_loss": 0.9244149923324585, | |
| "eval_runtime": 65.4024, | |
| "eval_samples_per_second": 190.115, | |
| "eval_steps_per_second": 11.896, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "eval_loss": 0.912540078163147, | |
| "eval_runtime": 65.4032, | |
| "eval_samples_per_second": 190.113, | |
| "eval_steps_per_second": 11.895, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "eval_loss": 0.9208225607872009, | |
| "eval_runtime": 65.4136, | |
| "eval_samples_per_second": 190.083, | |
| "eval_steps_per_second": 11.894, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_loss": 0.907939076423645, | |
| "eval_runtime": 65.3953, | |
| "eval_samples_per_second": 190.136, | |
| "eval_steps_per_second": 11.897, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "eval_loss": 0.903082013130188, | |
| "eval_runtime": 65.3937, | |
| "eval_samples_per_second": 190.141, | |
| "eval_steps_per_second": 11.897, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 0.001, | |
| "loss": 1.1491, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "eval_loss": 0.8966090083122253, | |
| "eval_runtime": 65.4073, | |
| "eval_samples_per_second": 190.101, | |
| "eval_steps_per_second": 11.895, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "eval_loss": 0.8984885215759277, | |
| "eval_runtime": 65.3845, | |
| "eval_samples_per_second": 190.167, | |
| "eval_steps_per_second": 11.899, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "eval_loss": 0.8956364393234253, | |
| "eval_runtime": 65.3948, | |
| "eval_samples_per_second": 190.137, | |
| "eval_steps_per_second": 11.897, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "eval_loss": 0.8861657381057739, | |
| "eval_runtime": 65.3936, | |
| "eval_samples_per_second": 190.141, | |
| "eval_steps_per_second": 11.897, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "eval_loss": 0.8837835788726807, | |
| "eval_runtime": 65.4055, | |
| "eval_samples_per_second": 190.106, | |
| "eval_steps_per_second": 11.895, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "eval_loss": 0.875281035900116, | |
| "eval_runtime": 65.3939, | |
| "eval_samples_per_second": 190.14, | |
| "eval_steps_per_second": 11.897, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_loss": 0.889029860496521, | |
| "eval_runtime": 65.4822, | |
| "eval_samples_per_second": 189.884, | |
| "eval_steps_per_second": 11.881, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "eval_loss": 0.9000933766365051, | |
| "eval_runtime": 65.3866, | |
| "eval_samples_per_second": 190.161, | |
| "eval_steps_per_second": 11.898, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "eval_loss": 0.897091805934906, | |
| "eval_runtime": 65.4252, | |
| "eval_samples_per_second": 190.049, | |
| "eval_steps_per_second": 11.891, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "eval_loss": 0.8837863206863403, | |
| "eval_runtime": 65.4877, | |
| "eval_samples_per_second": 189.868, | |
| "eval_steps_per_second": 11.88, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 0.001, | |
| "loss": 0.7877, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "eval_loss": 0.8839919567108154, | |
| "eval_runtime": 65.4142, | |
| "eval_samples_per_second": 190.081, | |
| "eval_steps_per_second": 11.893, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "eval_loss": 0.8844324350357056, | |
| "eval_runtime": 65.4264, | |
| "eval_samples_per_second": 190.046, | |
| "eval_steps_per_second": 11.891, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "eval_loss": 0.8891133666038513, | |
| "eval_runtime": 65.4136, | |
| "eval_samples_per_second": 190.083, | |
| "eval_steps_per_second": 11.894, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 2.87, | |
| "eval_loss": 0.8817307949066162, | |
| "eval_runtime": 65.4132, | |
| "eval_samples_per_second": 190.084, | |
| "eval_steps_per_second": 11.894, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_loss": 0.8784255981445312, | |
| "eval_runtime": 65.4323, | |
| "eval_samples_per_second": 190.029, | |
| "eval_steps_per_second": 11.89, | |
| "step": 1200 | |
| } | |
| ], | |
| "max_steps": 1200, | |
| "num_train_epochs": 3, | |
| "total_flos": 3.392579398501663e+17, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |