| { | |
| "best_metric": 0.464372456073761, | |
| "best_model_checkpoint": "./mistral/21-04-24-Weni-WeniGPT-Agents-Mistral-1.0.6-SFT-1.0.11-DPO_Experiment on DPO with other hyperparameters and best SFT model of WeniGPT-2_max_steps-366_batch_4_2024-04-21_ppid_9/checkpoint-90", | |
| "epoch": 1.4634146341463414, | |
| "eval_steps": 30, | |
| "global_step": 90, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.16260162601626016, | |
| "grad_norm": 11.074726104736328, | |
| "learning_rate": 3.6363636363636366e-06, | |
| "logits/chosen": -1.7639459371566772, | |
| "logits/rejected": -1.8396574258804321, | |
| "logps/chosen": -186.35623168945312, | |
| "logps/rejected": -242.7454376220703, | |
| "loss": 0.6923, | |
| "rewards/accuracies": 0.550000011920929, | |
| "rewards/chosen": 0.0025298120453953743, | |
| "rewards/margins": 0.002917728153988719, | |
| "rewards/rejected": -0.000387916574254632, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.3252032520325203, | |
| "grad_norm": 13.123202323913574, | |
| "learning_rate": 4.9154929577464795e-06, | |
| "logits/chosen": -1.8020210266113281, | |
| "logits/rejected": -1.827330231666565, | |
| "logps/chosen": -179.5525665283203, | |
| "logps/rejected": -159.48426818847656, | |
| "loss": 0.6821, | |
| "rewards/accuracies": 1.0, | |
| "rewards/chosen": 0.0269151721149683, | |
| "rewards/margins": 0.020867865532636642, | |
| "rewards/rejected": 0.00604730611667037, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.4878048780487805, | |
| "grad_norm": 14.401883125305176, | |
| "learning_rate": 4.774647887323944e-06, | |
| "logits/chosen": -1.7544463872909546, | |
| "logits/rejected": -1.8119304180145264, | |
| "logps/chosen": -131.93418884277344, | |
| "logps/rejected": -186.62869262695312, | |
| "loss": 0.6531, | |
| "rewards/accuracies": 0.949999988079071, | |
| "rewards/chosen": 0.0737195536494255, | |
| "rewards/margins": 0.08270301669836044, | |
| "rewards/rejected": -0.008983459323644638, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.4878048780487805, | |
| "eval_logits/chosen": -1.7289739847183228, | |
| "eval_logits/rejected": -1.7717361450195312, | |
| "eval_logps/chosen": -135.4998321533203, | |
| "eval_logps/rejected": -185.3151397705078, | |
| "eval_loss": 0.6393889784812927, | |
| "eval_rewards/accuracies": 0.7857142686843872, | |
| "eval_rewards/chosen": 0.12710317969322205, | |
| "eval_rewards/margins": 0.1346290558576584, | |
| "eval_rewards/rejected": -0.007525864988565445, | |
| "eval_runtime": 17.3728, | |
| "eval_samples_per_second": 1.612, | |
| "eval_steps_per_second": 0.806, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.6504065040650406, | |
| "grad_norm": 15.026944160461426, | |
| "learning_rate": 4.647887323943662e-06, | |
| "logits/chosen": -1.8521251678466797, | |
| "logits/rejected": -1.9095609188079834, | |
| "logps/chosen": -167.38348388671875, | |
| "logps/rejected": -215.0012969970703, | |
| "loss": 0.6343, | |
| "rewards/accuracies": 0.800000011920929, | |
| "rewards/chosen": 0.11476834118366241, | |
| "rewards/margins": 0.10640791803598404, | |
| "rewards/rejected": 0.008360415697097778, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.8130081300813008, | |
| "grad_norm": 15.622106552124023, | |
| "learning_rate": 4.507042253521127e-06, | |
| "logits/chosen": -1.7336807250976562, | |
| "logits/rejected": -1.7874727249145508, | |
| "logps/chosen": -151.20423889160156, | |
| "logps/rejected": -212.6553192138672, | |
| "loss": 0.5966, | |
| "rewards/accuracies": 0.949999988079071, | |
| "rewards/chosen": 0.16852515935897827, | |
| "rewards/margins": 0.19663922488689423, | |
| "rewards/rejected": -0.0281140748411417, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.975609756097561, | |
| "grad_norm": 15.992025375366211, | |
| "learning_rate": 4.3661971830985915e-06, | |
| "logits/chosen": -1.8077377080917358, | |
| "logits/rejected": -1.8177680969238281, | |
| "logps/chosen": -166.0524139404297, | |
| "logps/rejected": -166.77749633789062, | |
| "loss": 0.5732, | |
| "rewards/accuracies": 0.8999999761581421, | |
| "rewards/chosen": 0.28099626302719116, | |
| "rewards/margins": 0.25574439764022827, | |
| "rewards/rejected": 0.0252518467605114, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.975609756097561, | |
| "eval_logits/chosen": -1.734244465827942, | |
| "eval_logits/rejected": -1.7778174877166748, | |
| "eval_logps/chosen": -133.12557983398438, | |
| "eval_logps/rejected": -185.54177856445312, | |
| "eval_loss": 0.5475863218307495, | |
| "eval_rewards/accuracies": 0.7857142686843872, | |
| "eval_rewards/chosen": 0.36452820897102356, | |
| "eval_rewards/margins": 0.39471906423568726, | |
| "eval_rewards/rejected": -0.030190791934728622, | |
| "eval_runtime": 17.4075, | |
| "eval_samples_per_second": 1.609, | |
| "eval_steps_per_second": 0.804, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 1.1382113821138211, | |
| "grad_norm": 13.867135047912598, | |
| "learning_rate": 4.225352112676057e-06, | |
| "logits/chosen": -1.8027187585830688, | |
| "logits/rejected": -1.882067084312439, | |
| "logps/chosen": -174.7197723388672, | |
| "logps/rejected": -223.87844848632812, | |
| "loss": 0.5589, | |
| "rewards/accuracies": 0.949999988079071, | |
| "rewards/chosen": 0.2828216254711151, | |
| "rewards/margins": 0.361534059047699, | |
| "rewards/rejected": -0.07871242612600327, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 1.3008130081300813, | |
| "grad_norm": 14.660954475402832, | |
| "learning_rate": 4.0845070422535216e-06, | |
| "logits/chosen": -1.8306058645248413, | |
| "logits/rejected": -1.8653805255889893, | |
| "logps/chosen": -170.78506469726562, | |
| "logps/rejected": -210.1864013671875, | |
| "loss": 0.4755, | |
| "rewards/accuracies": 0.949999988079071, | |
| "rewards/chosen": 0.4126508831977844, | |
| "rewards/margins": 0.48294001817703247, | |
| "rewards/rejected": -0.07028912007808685, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 1.4634146341463414, | |
| "grad_norm": 15.84911060333252, | |
| "learning_rate": 3.943661971830986e-06, | |
| "logits/chosen": -1.7707087993621826, | |
| "logits/rejected": -1.81671142578125, | |
| "logps/chosen": -143.22067260742188, | |
| "logps/rejected": -205.97433471679688, | |
| "loss": 0.5252, | |
| "rewards/accuracies": 1.0, | |
| "rewards/chosen": 0.4095473885536194, | |
| "rewards/margins": 0.5323301553726196, | |
| "rewards/rejected": -0.12278278172016144, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 1.4634146341463414, | |
| "eval_logits/chosen": -1.74051034450531, | |
| "eval_logits/rejected": -1.7852736711502075, | |
| "eval_logps/chosen": -130.74990844726562, | |
| "eval_logps/rejected": -186.0896759033203, | |
| "eval_loss": 0.464372456073761, | |
| "eval_rewards/accuracies": 0.7857142686843872, | |
| "eval_rewards/chosen": 0.6020953059196472, | |
| "eval_rewards/margins": 0.6870760321617126, | |
| "eval_rewards/rejected": -0.08498072624206543, | |
| "eval_runtime": 17.3793, | |
| "eval_samples_per_second": 1.611, | |
| "eval_steps_per_second": 0.806, | |
| "step": 90 | |
| } | |
| ], | |
| "logging_steps": 10, | |
| "max_steps": 366, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 6, | |
| "save_steps": 90, | |
| "total_flos": 0.0, | |
| "train_batch_size": 1, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |