| { |
| "best_global_step": null, |
| "best_metric": null, |
| "best_model_checkpoint": null, |
| "epoch": 5.0, |
| "eval_steps": 4240, |
| "global_step": 4240, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 0.589622641509434, |
| "grad_norm": 10.394728660583496, |
| "learning_rate": 1.9606402519023878e-05, |
| "loss": 0.6713, |
| "step": 500 |
| }, |
| { |
| "epoch": 1.179245283018868, |
| "grad_norm": 5.316000461578369, |
| "learning_rate": 1.69824193125164e-05, |
| "loss": 0.1576, |
| "step": 1000 |
| }, |
| { |
| "epoch": 1.7688679245283019, |
| "grad_norm": 5.894363880157471, |
| "learning_rate": 1.4358436106008921e-05, |
| "loss": 0.0938, |
| "step": 1500 |
| }, |
| { |
| "epoch": 2.358490566037736, |
| "grad_norm": 1.9153144359588623, |
| "learning_rate": 1.1734452899501446e-05, |
| "loss": 0.0618, |
| "step": 2000 |
| }, |
| { |
| "epoch": 2.94811320754717, |
| "grad_norm": 3.0426125526428223, |
| "learning_rate": 9.110469692993965e-06, |
| "loss": 0.0439, |
| "step": 2500 |
| }, |
| { |
| "epoch": 3.5377358490566038, |
| "grad_norm": 1.4173264503479004, |
| "learning_rate": 6.486486486486487e-06, |
| "loss": 0.0285, |
| "step": 3000 |
| }, |
| { |
| "epoch": 4.127358490566038, |
| "grad_norm": 1.4994899034500122, |
| "learning_rate": 3.862503279979009e-06, |
| "loss": 0.0254, |
| "step": 3500 |
| }, |
| { |
| "epoch": 4.716981132075472, |
| "grad_norm": 0.4605902433395386, |
| "learning_rate": 1.2385200734715299e-06, |
| "loss": 0.018, |
| "step": 4000 |
| }, |
| { |
| "epoch": 5.0, |
| "eval_runtime": 520.8291, |
| "eval_samples_per_second": 0.0, |
| "eval_steps_per_second": 0.0, |
| "eval_validation_ir_eval_cosine_accuracy@1": 0.6158521600943659, |
| "eval_validation_ir_eval_cosine_accuracy@10": 0.9647438522911581, |
| "eval_validation_ir_eval_cosine_accuracy@3": 0.8605809400547191, |
| "eval_validation_ir_eval_cosine_accuracy@5": 0.920690050623372, |
| "eval_validation_ir_eval_cosine_map@100": 0.7488402597891997, |
| "eval_validation_ir_eval_cosine_mrr@10": 0.7469129741926596, |
| "eval_validation_ir_eval_cosine_ndcg@10": 0.8007440911740579, |
| "eval_validation_ir_eval_cosine_precision@1": 0.6158521600943659, |
| "eval_validation_ir_eval_cosine_precision@10": 0.09647438522911582, |
| "eval_validation_ir_eval_cosine_precision@3": 0.28686031335157297, |
| "eval_validation_ir_eval_cosine_precision@5": 0.18413801012467443, |
| "eval_validation_ir_eval_cosine_recall@1": 0.6158521600943659, |
| "eval_validation_ir_eval_cosine_recall@10": 0.9647438522911581, |
| "eval_validation_ir_eval_cosine_recall@3": 0.8605809400547191, |
| "eval_validation_ir_eval_cosine_recall@5": 0.920690050623372, |
| "step": 4240 |
| } |
| ], |
| "logging_steps": 500, |
| "max_steps": 4240, |
| "num_input_tokens_seen": 0, |
| "num_train_epochs": 5, |
| "save_steps": 4240, |
| "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": 72, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
