| { | |
| "best_metric": 3.1014273166656494, | |
| "best_model_checkpoint": "../Modelos/bloomz_AEx_SQUAD 16-11-22_14:58:43/checkpoint-1000", | |
| "epoch": 1.0787327266599258, | |
| "global_step": 1000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.05, | |
| "eval_loss": 5.220151424407959, | |
| "eval_runtime": 62.5146, | |
| "eval_samples_per_second": 109.095, | |
| "eval_steps_per_second": 27.274, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "eval_loss": 4.333649158477783, | |
| "eval_runtime": 62.5129, | |
| "eval_samples_per_second": 109.097, | |
| "eval_steps_per_second": 27.274, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "eval_loss": 3.8862874507904053, | |
| "eval_runtime": 62.5283, | |
| "eval_samples_per_second": 109.071, | |
| "eval_steps_per_second": 27.268, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "eval_loss": 3.6269617080688477, | |
| "eval_runtime": 62.5217, | |
| "eval_samples_per_second": 109.082, | |
| "eval_steps_per_second": 27.271, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "eval_loss": 3.50032639503479, | |
| "eval_runtime": 62.5086, | |
| "eval_samples_per_second": 109.105, | |
| "eval_steps_per_second": 27.276, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "eval_loss": 3.3939645290374756, | |
| "eval_runtime": 62.5177, | |
| "eval_samples_per_second": 109.089, | |
| "eval_steps_per_second": 27.272, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "eval_loss": 3.3134140968322754, | |
| "eval_runtime": 62.5154, | |
| "eval_samples_per_second": 109.093, | |
| "eval_steps_per_second": 27.273, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "eval_loss": 3.228942394256592, | |
| "eval_runtime": 62.5039, | |
| "eval_samples_per_second": 109.113, | |
| "eval_steps_per_second": 27.278, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "eval_loss": 3.2036657333374023, | |
| "eval_runtime": 62.5162, | |
| "eval_samples_per_second": 109.092, | |
| "eval_steps_per_second": 27.273, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.108234448040274e-05, | |
| "loss": 3.8523, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "eval_loss": 3.1620962619781494, | |
| "eval_runtime": 62.5012, | |
| "eval_samples_per_second": 109.118, | |
| "eval_steps_per_second": 27.279, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "eval_loss": 3.128020763397217, | |
| "eval_runtime": 62.5236, | |
| "eval_samples_per_second": 109.079, | |
| "eval_steps_per_second": 27.27, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "eval_loss": 3.0899899005889893, | |
| "eval_runtime": 62.5266, | |
| "eval_samples_per_second": 109.074, | |
| "eval_steps_per_second": 27.268, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "eval_loss": 3.073817729949951, | |
| "eval_runtime": 62.5093, | |
| "eval_samples_per_second": 109.104, | |
| "eval_steps_per_second": 27.276, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "eval_loss": 3.0438177585601807, | |
| "eval_runtime": 62.5327, | |
| "eval_samples_per_second": 109.063, | |
| "eval_steps_per_second": 27.266, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "eval_loss": 3.0443129539489746, | |
| "eval_runtime": 62.5242, | |
| "eval_samples_per_second": 109.078, | |
| "eval_steps_per_second": 27.269, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "eval_loss": 3.044921636581421, | |
| "eval_runtime": 62.5248, | |
| "eval_samples_per_second": 109.077, | |
| "eval_steps_per_second": 27.269, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "eval_loss": 3.0547332763671875, | |
| "eval_runtime": 62.5038, | |
| "eval_samples_per_second": 109.113, | |
| "eval_steps_per_second": 27.278, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "eval_loss": 3.041879415512085, | |
| "eval_runtime": 62.5198, | |
| "eval_samples_per_second": 109.086, | |
| "eval_steps_per_second": 27.271, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "eval_loss": 3.113288640975952, | |
| "eval_runtime": 62.5071, | |
| "eval_samples_per_second": 109.108, | |
| "eval_steps_per_second": 27.277, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 3.209277238403452e-05, | |
| "loss": 2.4796, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "eval_loss": 3.1014273166656494, | |
| "eval_runtime": 62.5073, | |
| "eval_samples_per_second": 109.107, | |
| "eval_steps_per_second": 27.277, | |
| "step": 1000 | |
| } | |
| ], | |
| "max_steps": 2781, | |
| "num_train_epochs": 3, | |
| "total_flos": 4.377761810061722e+16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |