| { |
| "best_global_step": 36, |
| "best_metric": 0.9930555555555556, |
| "best_model_checkpoint": "/Users/lukehinds/dev/hedgehog-workspace/security-hardened-models/classifier-training/models/refusal_classifier_v2/checkpoints/checkpoint-36", |
| "epoch": 5.0, |
| "eval_steps": 500, |
| "global_step": 90, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 0.5555555555555556, |
| "grad_norm": 6.884082317352295, |
| "learning_rate": 3.6000000000000003e-06, |
| "loss": 1.291097640991211, |
| "step": 10 |
| }, |
| { |
| "epoch": 1.0, |
| "eval_avg_accuracy": 0.8472222222222222, |
| "eval_explanation_accuracy": 0.9444444444444444, |
| "eval_explanation_f1": 0.9375, |
| "eval_loss": 1.1282232999801636, |
| "eval_refusal_accuracy": 0.75, |
| "eval_refusal_f1": 0.8333333333333334, |
| "eval_runtime": 0.4555, |
| "eval_samples_per_second": 158.056, |
| "eval_steps_per_second": 10.976, |
| "step": 18 |
| }, |
| { |
| "epoch": 1.1111111111111112, |
| "grad_norm": 3.985938787460327, |
| "learning_rate": 7.600000000000001e-06, |
| "loss": 1.1933887481689454, |
| "step": 20 |
| }, |
| { |
| "epoch": 1.6666666666666665, |
| "grad_norm": 5.078998565673828, |
| "learning_rate": 1.16e-05, |
| "loss": 0.8807908058166504, |
| "step": 30 |
| }, |
| { |
| "epoch": 2.0, |
| "eval_avg_accuracy": 0.9930555555555556, |
| "eval_explanation_accuracy": 0.9861111111111112, |
| "eval_explanation_f1": 0.9836065573770492, |
| "eval_loss": 0.3898311257362366, |
| "eval_refusal_accuracy": 1.0, |
| "eval_refusal_f1": 1.0, |
| "eval_runtime": 0.1462, |
| "eval_samples_per_second": 492.56, |
| "eval_steps_per_second": 34.206, |
| "step": 36 |
| }, |
| { |
| "epoch": 2.2222222222222223, |
| "grad_norm": 3.4086709022521973, |
| "learning_rate": 1.5600000000000003e-05, |
| "loss": 0.4771848678588867, |
| "step": 40 |
| }, |
| { |
| "epoch": 2.7777777777777777, |
| "grad_norm": 0.6608096361160278, |
| "learning_rate": 1.9600000000000002e-05, |
| "loss": 0.14329642057418823, |
| "step": 50 |
| }, |
| { |
| "epoch": 3.0, |
| "eval_avg_accuracy": 0.9861111111111112, |
| "eval_explanation_accuracy": 0.9861111111111112, |
| "eval_explanation_f1": 0.9836065573770492, |
| "eval_loss": 0.09537098556756973, |
| "eval_refusal_accuracy": 0.9861111111111112, |
| "eval_refusal_f1": 0.9887640449438202, |
| "eval_runtime": 0.1442, |
| "eval_samples_per_second": 499.4, |
| "eval_steps_per_second": 34.681, |
| "step": 54 |
| }, |
| { |
| "epoch": 3.3333333333333335, |
| "grad_norm": 0.3193122148513794, |
| "learning_rate": 1.55e-05, |
| "loss": 0.03496098518371582, |
| "step": 60 |
| }, |
| { |
| "epoch": 3.888888888888889, |
| "grad_norm": 0.14034102857112885, |
| "learning_rate": 1.0500000000000001e-05, |
| "loss": 0.012411238998174668, |
| "step": 70 |
| }, |
| { |
| "epoch": 4.0, |
| "eval_avg_accuracy": 0.9861111111111112, |
| "eval_explanation_accuracy": 0.9861111111111112, |
| "eval_explanation_f1": 0.9836065573770492, |
| "eval_loss": 0.08972182869911194, |
| "eval_refusal_accuracy": 0.9861111111111112, |
| "eval_refusal_f1": 0.9887640449438202, |
| "eval_runtime": 0.1489, |
| "eval_samples_per_second": 483.475, |
| "eval_steps_per_second": 33.575, |
| "step": 72 |
| }, |
| { |
| "epoch": 4.444444444444445, |
| "grad_norm": 0.12835660576820374, |
| "learning_rate": 5.500000000000001e-06, |
| "loss": 0.008212362974882126, |
| "step": 80 |
| }, |
| { |
| "epoch": 5.0, |
| "grad_norm": 0.09251071512699127, |
| "learning_rate": 5.000000000000001e-07, |
| "loss": 0.007145626097917556, |
| "step": 90 |
| }, |
| { |
| "epoch": 5.0, |
| "eval_avg_accuracy": 0.9861111111111112, |
| "eval_explanation_accuracy": 0.9861111111111112, |
| "eval_explanation_f1": 0.9836065573770492, |
| "eval_loss": 0.10063999891281128, |
| "eval_refusal_accuracy": 0.9861111111111112, |
| "eval_refusal_f1": 0.9887640449438202, |
| "eval_runtime": 0.1262, |
| "eval_samples_per_second": 570.447, |
| "eval_steps_per_second": 39.614, |
| "step": 90 |
| } |
| ], |
| "logging_steps": 10, |
| "max_steps": 90, |
| "num_input_tokens_seen": 0, |
| "num_train_epochs": 5, |
| "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": 0.0, |
| "train_batch_size": 16, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
