| { |
| "best_metric": null, |
| "best_model_checkpoint": null, |
| "epoch": 0.9874476987447699, |
| "eval_steps": 500, |
| "global_step": 59, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 0.016736401673640166, |
| "grad_norm": 6.245392524863121, |
| "learning_rate": 8.333333333333333e-08, |
| "logits/chosen": -2.849759578704834, |
| "logits/rejected": -2.870671272277832, |
| "logps/chosen": -163.49493408203125, |
| "logps/pi_response": -170.3665313720703, |
| "logps/ref_response": -170.3665313720703, |
| "logps/rejected": -174.6565399169922, |
| "loss": 0.6916, |
| "rewards/accuracies": 0.0, |
| "rewards/chosen": 0.0, |
| "rewards/margins": 0.0, |
| "rewards/rejected": 0.0, |
| "step": 1 |
| }, |
| { |
| "epoch": 0.16736401673640167, |
| "grad_norm": 6.434100603220207, |
| "learning_rate": 4.930057285201027e-07, |
| "logits/chosen": -2.7605881690979004, |
| "logits/rejected": -2.7289905548095703, |
| "logps/chosen": -150.26988220214844, |
| "logps/pi_response": -155.58465576171875, |
| "logps/ref_response": -154.30699157714844, |
| "logps/rejected": -151.04278564453125, |
| "loss": 0.6918, |
| "rewards/accuracies": 0.4791666567325592, |
| "rewards/chosen": -0.012262338772416115, |
| "rewards/margins": 0.00022517387697007507, |
| "rewards/rejected": -0.012487512081861496, |
| "step": 10 |
| }, |
| { |
| "epoch": 0.33472803347280333, |
| "grad_norm": 6.340696277824271, |
| "learning_rate": 4.187457503795526e-07, |
| "logits/chosen": -2.763792037963867, |
| "logits/rejected": -2.7616288661956787, |
| "logps/chosen": -142.66026306152344, |
| "logps/pi_response": -144.825439453125, |
| "logps/ref_response": -143.23764038085938, |
| "logps/rejected": -143.56515502929688, |
| "loss": 0.6916, |
| "rewards/accuracies": 0.4312500059604645, |
| "rewards/chosen": -0.016828009858727455, |
| "rewards/margins": -0.002716110087931156, |
| "rewards/rejected": -0.014111898839473724, |
| "step": 20 |
| }, |
| { |
| "epoch": 0.502092050209205, |
| "grad_norm": 6.473140966519031, |
| "learning_rate": 2.8691164100062034e-07, |
| "logits/chosen": -2.6851718425750732, |
| "logits/rejected": -2.686378002166748, |
| "logps/chosen": -134.32498168945312, |
| "logps/pi_response": -140.74411010742188, |
| "logps/ref_response": -138.3269805908203, |
| "logps/rejected": -144.6517791748047, |
| "loss": 0.6906, |
| "rewards/accuracies": 0.5874999761581421, |
| "rewards/chosen": -0.017753150314092636, |
| "rewards/margins": 0.006084781605750322, |
| "rewards/rejected": -0.02383793331682682, |
| "step": 30 |
| }, |
| { |
| "epoch": 0.6694560669456067, |
| "grad_norm": 6.900347328427309, |
| "learning_rate": 1.4248369943086995e-07, |
| "logits/chosen": -2.731713056564331, |
| "logits/rejected": -2.7183914184570312, |
| "logps/chosen": -149.15603637695312, |
| "logps/pi_response": -149.87930297851562, |
| "logps/ref_response": -146.56552124023438, |
| "logps/rejected": -147.74722290039062, |
| "loss": 0.6898, |
| "rewards/accuracies": 0.606249988079071, |
| "rewards/chosen": -0.031016364693641663, |
| "rewards/margins": 0.007326220162212849, |
| "rewards/rejected": -0.038342587649822235, |
| "step": 40 |
| }, |
| { |
| "epoch": 0.8368200836820083, |
| "grad_norm": 7.3643241935272705, |
| "learning_rate": 3.473909705816111e-08, |
| "logits/chosen": -2.7875192165374756, |
| "logits/rejected": -2.7989630699157715, |
| "logps/chosen": -161.47055053710938, |
| "logps/pi_response": -166.01963806152344, |
| "logps/ref_response": -158.890625, |
| "logps/rejected": -164.82586669921875, |
| "loss": 0.6902, |
| "rewards/accuracies": 0.45625001192092896, |
| "rewards/chosen": -0.07169784605503082, |
| "rewards/margins": -0.002618619007989764, |
| "rewards/rejected": -0.06907922029495239, |
| "step": 50 |
| }, |
| { |
| "epoch": 0.9874476987447699, |
| "step": 59, |
| "total_flos": 0.0, |
| "train_loss": 0.6904925431235361, |
| "train_runtime": 1293.8247, |
| "train_samples_per_second": 11.812, |
| "train_steps_per_second": 0.046 |
| } |
| ], |
| "logging_steps": 10, |
| "max_steps": 59, |
| "num_input_tokens_seen": 0, |
| "num_train_epochs": 1, |
| "save_steps": 100, |
| "total_flos": 0.0, |
| "train_batch_size": 4, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
