| { | |
| "best_global_step": 1646, | |
| "best_metric": 0.7345098851242857, | |
| "best_model_checkpoint": "./final_lora_model\\checkpoint-1646", | |
| "epoch": 2.0, | |
| "eval_steps": 500, | |
| "global_step": 1646, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.12150668286755771, | |
| "grad_norm": 1.2585123777389526, | |
| "learning_rate": 1.65e-05, | |
| "loss": 1.0976, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.24301336573511542, | |
| "grad_norm": 1.4245030879974365, | |
| "learning_rate": 3.316666666666667e-05, | |
| "loss": 1.0778, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.3645200486026732, | |
| "grad_norm": 3.1913540363311768, | |
| "learning_rate": 4.9833333333333336e-05, | |
| "loss": 0.9574, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.48602673147023084, | |
| "grad_norm": 3.6243133544921875, | |
| "learning_rate": 4.771784232365146e-05, | |
| "loss": 0.7936, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.6075334143377886, | |
| "grad_norm": 4.164339065551758, | |
| "learning_rate": 4.541263254956201e-05, | |
| "loss": 0.7113, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.7290400972053463, | |
| "grad_norm": 4.441893100738525, | |
| "learning_rate": 4.3107422775472575e-05, | |
| "loss": 0.7106, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.850546780072904, | |
| "grad_norm": 4.538736820220947, | |
| "learning_rate": 4.0802213001383126e-05, | |
| "loss": 0.6494, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.9720534629404617, | |
| "grad_norm": 2.1902823448181152, | |
| "learning_rate": 3.849700322729369e-05, | |
| "loss": 0.6784, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.7184397163120567, | |
| "eval_f1": 0.7185922027631994, | |
| "eval_loss": 0.6663119792938232, | |
| "eval_precision": 0.7189266108548652, | |
| "eval_recall": 0.7184397163120567, | |
| "eval_runtime": 39.2565, | |
| "eval_samples_per_second": 35.918, | |
| "eval_steps_per_second": 2.267, | |
| "step": 823 | |
| }, | |
| { | |
| "epoch": 1.0935601458080195, | |
| "grad_norm": 2.8516221046447754, | |
| "learning_rate": 3.619179345320424e-05, | |
| "loss": 0.6113, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 1.2150668286755772, | |
| "grad_norm": 4.570059299468994, | |
| "learning_rate": 3.38865836791148e-05, | |
| "loss": 0.6011, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 1.336573511543135, | |
| "grad_norm": 4.810871601104736, | |
| "learning_rate": 3.158137390502536e-05, | |
| "loss": 0.6534, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 1.4580801944106927, | |
| "grad_norm": 2.5861403942108154, | |
| "learning_rate": 2.927616413093592e-05, | |
| "loss": 0.6319, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 1.5795868772782504, | |
| "grad_norm": 4.017718315124512, | |
| "learning_rate": 2.6970954356846474e-05, | |
| "loss": 0.6399, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 1.7010935601458081, | |
| "grad_norm": 6.661707878112793, | |
| "learning_rate": 2.466574458275703e-05, | |
| "loss": 0.603, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 1.8226002430133659, | |
| "grad_norm": 2.221937417984009, | |
| "learning_rate": 2.236053480866759e-05, | |
| "loss": 0.6289, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 1.9441069258809236, | |
| "grad_norm": 6.443672180175781, | |
| "learning_rate": 2.0055325034578147e-05, | |
| "loss": 0.6351, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.7361702127659574, | |
| "eval_f1": 0.7345098851242857, | |
| "eval_loss": 0.6419773101806641, | |
| "eval_precision": 0.735769114060402, | |
| "eval_recall": 0.7361702127659574, | |
| "eval_runtime": 38.5936, | |
| "eval_samples_per_second": 36.535, | |
| "eval_steps_per_second": 2.306, | |
| "step": 1646 | |
| } | |
| ], | |
| "logging_steps": 100, | |
| "max_steps": 2469, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 3, | |
| "save_steps": 500, | |
| "stateful_callbacks": { | |
| "TrainerControl": { | |
| "args": { | |
| "should_epoch_stop": false, | |
| "should_evaluate": false, | |
| "should_log": false, | |
| "should_save": true, | |
| "should_training_stop": false | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 291629305344780.0, | |
| "train_batch_size": 8, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |