| { | |
| "best_metric": 0.7888103127479553, | |
| "best_model_checkpoint": "../experiments/attribute_minig_mslacerda/run-4/checkpoint-516", | |
| "epoch": 12.0, | |
| "eval_steps": 500, | |
| "global_step": 516, | |
| "is_hyper_param_search": true, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.017117117117117116, | |
| "eval_f1": 0.0037002775208140612, | |
| "eval_loss": 4.726041316986084, | |
| "eval_precision": 0.002952029520295203, | |
| "eval_recall": 0.004956629491945477, | |
| "eval_runtime": 0.3653, | |
| "eval_samples_per_second": 796.702, | |
| "eval_steps_per_second": 101.299, | |
| "step": 43 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.05345345345345345, | |
| "eval_f1": 0.005950770895320531, | |
| "eval_loss": 4.565603733062744, | |
| "eval_precision": 0.00528084493518963, | |
| "eval_recall": 0.006815365551425031, | |
| "eval_runtime": 0.3702, | |
| "eval_samples_per_second": 786.087, | |
| "eval_steps_per_second": 99.949, | |
| "step": 86 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.14594594594594595, | |
| "eval_f1": 0.022442244224422443, | |
| "eval_loss": 4.133880138397217, | |
| "eval_precision": 0.02401129943502825, | |
| "eval_recall": 0.021065675340768277, | |
| "eval_runtime": 0.3715, | |
| "eval_samples_per_second": 783.407, | |
| "eval_steps_per_second": 99.608, | |
| "step": 129 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.3012012012012012, | |
| "eval_f1": 0.08808290155440414, | |
| "eval_loss": 3.4629595279693604, | |
| "eval_precision": 0.109375, | |
| "eval_recall": 0.07372986369268897, | |
| "eval_runtime": 0.361, | |
| "eval_samples_per_second": 805.99, | |
| "eval_steps_per_second": 102.48, | |
| "step": 172 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.42372372372372374, | |
| "eval_f1": 0.1667268134247564, | |
| "eval_loss": 2.7417962551116943, | |
| "eval_precision": 0.1996542783059637, | |
| "eval_recall": 0.14312267657992564, | |
| "eval_runtime": 0.3592, | |
| "eval_samples_per_second": 810.024, | |
| "eval_steps_per_second": 102.993, | |
| "step": 215 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.5027027027027027, | |
| "eval_f1": 0.24229390681003585, | |
| "eval_loss": 2.2062222957611084, | |
| "eval_precision": 0.28741496598639454, | |
| "eval_recall": 0.2094175960346964, | |
| "eval_runtime": 0.3678, | |
| "eval_samples_per_second": 791.169, | |
| "eval_steps_per_second": 100.595, | |
| "step": 258 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.587987987987988, | |
| "eval_f1": 0.3353617308992563, | |
| "eval_loss": 1.7638441324234009, | |
| "eval_precision": 0.36904761904761907, | |
| "eval_recall": 0.3073110285006196, | |
| "eval_runtime": 0.3685, | |
| "eval_samples_per_second": 789.62, | |
| "eval_steps_per_second": 100.398, | |
| "step": 301 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.6618618618618619, | |
| "eval_f1": 0.4317065181788405, | |
| "eval_loss": 1.4482687711715698, | |
| "eval_precision": 0.45795691452397497, | |
| "eval_recall": 0.4083023543990087, | |
| "eval_runtime": 0.3688, | |
| "eval_samples_per_second": 789.152, | |
| "eval_steps_per_second": 100.339, | |
| "step": 344 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.7261261261261261, | |
| "eval_f1": 0.5071335927367056, | |
| "eval_loss": 1.2015855312347412, | |
| "eval_precision": 0.5319727891156463, | |
| "eval_recall": 0.48451053283767037, | |
| "eval_runtime": 0.3937, | |
| "eval_samples_per_second": 739.049, | |
| "eval_steps_per_second": 93.968, | |
| "step": 387 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_accuracy": 0.7624624624624624, | |
| "eval_f1": 0.5714285714285714, | |
| "eval_loss": 0.9903969168663025, | |
| "eval_precision": 0.5886990801576872, | |
| "eval_recall": 0.5551425030978935, | |
| "eval_runtime": 0.3687, | |
| "eval_samples_per_second": 789.274, | |
| "eval_steps_per_second": 100.354, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_accuracy": 0.8087087087087087, | |
| "eval_f1": 0.6433002481389578, | |
| "eval_loss": 0.8383856415748596, | |
| "eval_precision": 0.6440993788819875, | |
| "eval_recall": 0.6425030978934325, | |
| "eval_runtime": 0.3684, | |
| "eval_samples_per_second": 789.848, | |
| "eval_steps_per_second": 100.427, | |
| "step": 473 | |
| }, | |
| { | |
| "epoch": 11.627906976744185, | |
| "grad_norm": 2.290496826171875, | |
| "learning_rate": 6.52888491924384e-06, | |
| "loss": 2.5968, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "eval_accuracy": 0.818018018018018, | |
| "eval_f1": 0.6617375231053604, | |
| "eval_loss": 0.7888103127479553, | |
| "eval_precision": 0.6580882352941176, | |
| "eval_recall": 0.6654275092936803, | |
| "eval_runtime": 0.3634, | |
| "eval_samples_per_second": 800.826, | |
| "eval_steps_per_second": 101.823, | |
| "step": 516 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 516, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 12, | |
| "save_steps": 500, | |
| "stateful_callbacks": { | |
| "TrainerControl": { | |
| "args": { | |
| "should_epoch_stop": false, | |
| "should_evaluate": false, | |
| "should_log": false, | |
| "should_save": true, | |
| "should_training_stop": true | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 140567218624704.0, | |
| "train_batch_size": 32, | |
| "trial_name": null, | |
| "trial_params": { | |
| "learning_rate": 2.85638715216918e-05, | |
| "num_train_epochs": 12, | |
| "per_device_train_batch_size": 32, | |
| "warmup_steps": 446, | |
| "weight_decay": 0.2979496592454386 | |
| } | |
| } | |