| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 1.9905213270142181, | |
| "eval_steps": 100, | |
| "global_step": 210, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.01, | |
| "grad_norm": 177.89952669184834, | |
| "learning_rate": 5e-09, | |
| "logits/chosen": 129.0, | |
| "logits/rejected": 125.5, | |
| "logps/chosen": -428.0, | |
| "logps/rejected": -470.0, | |
| "loss": 0.6914, | |
| "rewards/accuracies": 0.0, | |
| "rewards/chosen": 0.0, | |
| "rewards/margins": 0.0, | |
| "rewards/rejected": 0.0, | |
| "step": 1 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "grad_norm": 173.49613288568952, | |
| "learning_rate": 5e-08, | |
| "logits/chosen": 125.5, | |
| "logits/rejected": 132.0, | |
| "logps/chosen": -414.0, | |
| "logps/rejected": -450.0, | |
| "loss": 0.7165, | |
| "rewards/accuracies": 0.3888888955116272, | |
| "rewards/chosen": -0.00750732421875, | |
| "rewards/margins": -0.0037841796875, | |
| "rewards/rejected": -0.003875732421875, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "grad_norm": 181.136506215671, | |
| "learning_rate": 1e-07, | |
| "logits/chosen": 136.0, | |
| "logits/rejected": 136.0, | |
| "logps/chosen": -400.0, | |
| "logps/rejected": -430.0, | |
| "loss": 0.7186, | |
| "rewards/accuracies": 0.45625001192092896, | |
| "rewards/chosen": -0.0147705078125, | |
| "rewards/margins": -0.026123046875, | |
| "rewards/rejected": 0.01153564453125, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "grad_norm": 158.92632659059979, | |
| "learning_rate": 1.5e-07, | |
| "logits/chosen": 130.0, | |
| "logits/rejected": 130.0, | |
| "logps/chosen": -374.0, | |
| "logps/rejected": -390.0, | |
| "loss": 0.6979, | |
| "rewards/accuracies": 0.5625, | |
| "rewards/chosen": 0.08251953125, | |
| "rewards/margins": 0.08154296875, | |
| "rewards/rejected": 0.0013427734375, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "grad_norm": 172.98309546766845, | |
| "learning_rate": 2e-07, | |
| "logits/chosen": 115.0, | |
| "logits/rejected": 128.0, | |
| "logps/chosen": -338.0, | |
| "logps/rejected": -430.0, | |
| "loss": 0.6703, | |
| "rewards/accuracies": 0.6187499761581421, | |
| "rewards/chosen": 0.236328125, | |
| "rewards/margins": 0.1875, | |
| "rewards/rejected": 0.049072265625, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "grad_norm": 171.55866229471363, | |
| "learning_rate": 2.5e-07, | |
| "logits/chosen": 124.5, | |
| "logits/rejected": 124.5, | |
| "logps/chosen": -370.0, | |
| "logps/rejected": -388.0, | |
| "loss": 0.6286, | |
| "rewards/accuracies": 0.6312500238418579, | |
| "rewards/chosen": 0.2294921875, | |
| "rewards/margins": 0.318359375, | |
| "rewards/rejected": -0.08935546875, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "grad_norm": 175.84854471984127, | |
| "learning_rate": 3e-07, | |
| "logits/chosen": 116.0, | |
| "logits/rejected": 124.0, | |
| "logps/chosen": -342.0, | |
| "logps/rejected": -416.0, | |
| "loss": 0.5816, | |
| "rewards/accuracies": 0.706250011920929, | |
| "rewards/chosen": -0.035400390625, | |
| "rewards/margins": 0.640625, | |
| "rewards/rejected": -0.67578125, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "grad_norm": 134.077271259303, | |
| "learning_rate": 3.5e-07, | |
| "logits/chosen": 121.5, | |
| "logits/rejected": 117.5, | |
| "logps/chosen": -386.0, | |
| "logps/rejected": -394.0, | |
| "loss": 0.5647, | |
| "rewards/accuracies": 0.699999988079071, | |
| "rewards/chosen": -0.1162109375, | |
| "rewards/margins": 0.77734375, | |
| "rewards/rejected": -0.89453125, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "grad_norm": 145.7372885946958, | |
| "learning_rate": 4e-07, | |
| "logits/chosen": 117.0, | |
| "logits/rejected": 117.0, | |
| "logps/chosen": -342.0, | |
| "logps/rejected": -408.0, | |
| "loss": 0.5371, | |
| "rewards/accuracies": 0.71875, | |
| "rewards/chosen": -0.4765625, | |
| "rewards/margins": 0.91796875, | |
| "rewards/rejected": -1.3984375, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "grad_norm": 156.14756793621152, | |
| "learning_rate": 4.5e-07, | |
| "logits/chosen": 132.0, | |
| "logits/rejected": 134.0, | |
| "logps/chosen": -410.0, | |
| "logps/rejected": -444.0, | |
| "loss": 0.4753, | |
| "rewards/accuracies": 0.71875, | |
| "rewards/chosen": -0.77734375, | |
| "rewards/margins": 1.0234375, | |
| "rewards/rejected": -1.8046875, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "grad_norm": 142.7704392560794, | |
| "learning_rate": 5e-07, | |
| "logits/chosen": 120.0, | |
| "logits/rejected": 129.0, | |
| "logps/chosen": -390.0, | |
| "logps/rejected": -490.0, | |
| "loss": 0.5155, | |
| "rewards/accuracies": 0.7749999761581421, | |
| "rewards/chosen": -0.94140625, | |
| "rewards/margins": 1.3359375, | |
| "rewards/rejected": -2.28125, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "eval_logits/chosen": 95.0, | |
| "eval_logits/rejected": 97.5, | |
| "eval_logps/chosen": -378.0, | |
| "eval_logps/rejected": -456.0, | |
| "eval_loss": 0.500314474105835, | |
| "eval_rewards/accuracies": 0.686170220375061, | |
| "eval_rewards/chosen": -1.2421875, | |
| "eval_rewards/margins": 1.265625, | |
| "eval_rewards/rejected": -2.515625, | |
| "eval_runtime": 135.3377, | |
| "eval_samples_per_second": 5.542, | |
| "eval_steps_per_second": 0.347, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "grad_norm": 70.76245987345193, | |
| "learning_rate": 4.898732434036243e-07, | |
| "logits/chosen": 123.5, | |
| "logits/rejected": 118.5, | |
| "logps/chosen": -406.0, | |
| "logps/rejected": -460.0, | |
| "loss": 0.4082, | |
| "rewards/accuracies": 0.831250011920929, | |
| "rewards/chosen": -1.1328125, | |
| "rewards/margins": 1.7265625, | |
| "rewards/rejected": -2.859375, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "grad_norm": 83.18318775000992, | |
| "learning_rate": 4.603133832077953e-07, | |
| "logits/chosen": 117.5, | |
| "logits/rejected": 126.0, | |
| "logps/chosen": -362.0, | |
| "logps/rejected": -480.0, | |
| "loss": 0.2535, | |
| "rewards/accuracies": 0.862500011920929, | |
| "rewards/chosen": -0.546875, | |
| "rewards/margins": 2.828125, | |
| "rewards/rejected": -3.375, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "grad_norm": 80.23145875736671, | |
| "learning_rate": 4.137151834863213e-07, | |
| "logits/chosen": 111.0, | |
| "logits/rejected": 111.5, | |
| "logps/chosen": -400.0, | |
| "logps/rejected": -470.0, | |
| "loss": 0.2203, | |
| "rewards/accuracies": 0.8999999761581421, | |
| "rewards/chosen": -1.125, | |
| "rewards/margins": 2.75, | |
| "rewards/rejected": -3.875, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "grad_norm": 92.00162024415927, | |
| "learning_rate": 3.5385375325047163e-07, | |
| "logits/chosen": 118.0, | |
| "logits/rejected": 125.5, | |
| "logps/chosen": -398.0, | |
| "logps/rejected": -512.0, | |
| "loss": 0.2065, | |
| "rewards/accuracies": 0.9125000238418579, | |
| "rewards/chosen": -1.8359375, | |
| "rewards/margins": 2.84375, | |
| "rewards/rejected": -4.6875, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "grad_norm": 68.98201787598855, | |
| "learning_rate": 2.8557870956832133e-07, | |
| "logits/chosen": 107.0, | |
| "logits/rejected": 110.0, | |
| "logps/chosen": -394.0, | |
| "logps/rejected": -494.0, | |
| "loss": 0.1678, | |
| "rewards/accuracies": 0.9312499761581421, | |
| "rewards/chosen": -1.703125, | |
| "rewards/margins": 3.21875, | |
| "rewards/rejected": -4.9375, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "grad_norm": 80.45634124880881, | |
| "learning_rate": 2.1442129043167873e-07, | |
| "logits/chosen": 105.5, | |
| "logits/rejected": 125.0, | |
| "logps/chosen": -394.0, | |
| "logps/rejected": -506.0, | |
| "loss": 0.194, | |
| "rewards/accuracies": 0.918749988079071, | |
| "rewards/chosen": -1.28125, | |
| "rewards/margins": 3.671875, | |
| "rewards/rejected": -4.9375, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "grad_norm": 82.98758335477153, | |
| "learning_rate": 1.461462467495284e-07, | |
| "logits/chosen": 115.0, | |
| "logits/rejected": 116.0, | |
| "logps/chosen": -406.0, | |
| "logps/rejected": -472.0, | |
| "loss": 0.169, | |
| "rewards/accuracies": 0.949999988079071, | |
| "rewards/chosen": -0.95703125, | |
| "rewards/margins": 3.234375, | |
| "rewards/rejected": -4.1875, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "grad_norm": 87.6903501592833, | |
| "learning_rate": 8.628481651367875e-08, | |
| "logits/chosen": 111.0, | |
| "logits/rejected": 123.0, | |
| "logps/chosen": -406.0, | |
| "logps/rejected": -532.0, | |
| "loss": 0.1641, | |
| "rewards/accuracies": 0.956250011920929, | |
| "rewards/chosen": -0.73828125, | |
| "rewards/margins": 3.28125, | |
| "rewards/rejected": -4.0, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "grad_norm": 65.93189701960165, | |
| "learning_rate": 3.968661679220467e-08, | |
| "logits/chosen": 118.5, | |
| "logits/rejected": 119.5, | |
| "logps/chosen": -414.0, | |
| "logps/rejected": -510.0, | |
| "loss": 0.1756, | |
| "rewards/accuracies": 0.9312499761581421, | |
| "rewards/chosen": -0.8203125, | |
| "rewards/margins": 3.5, | |
| "rewards/rejected": -4.3125, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "grad_norm": 82.38035369049618, | |
| "learning_rate": 1.0126756596375685e-08, | |
| "logits/chosen": 100.5, | |
| "logits/rejected": 108.5, | |
| "logps/chosen": -384.0, | |
| "logps/rejected": -462.0, | |
| "loss": 0.1902, | |
| "rewards/accuracies": 0.9312499761581421, | |
| "rewards/chosen": -1.2890625, | |
| "rewards/margins": 3.0, | |
| "rewards/rejected": -4.28125, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "eval_logits/chosen": 87.0, | |
| "eval_logits/rejected": 89.5, | |
| "eval_logps/chosen": -388.0, | |
| "eval_logps/rejected": -476.0, | |
| "eval_loss": 0.44982096552848816, | |
| "eval_rewards/accuracies": 0.7393617033958435, | |
| "eval_rewards/chosen": -1.796875, | |
| "eval_rewards/margins": 1.71875, | |
| "eval_rewards/rejected": -3.515625, | |
| "eval_runtime": 135.9196, | |
| "eval_samples_per_second": 5.518, | |
| "eval_steps_per_second": 0.346, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "grad_norm": 54.35126708805902, | |
| "learning_rate": 0.0, | |
| "logits/chosen": 100.5, | |
| "logits/rejected": 102.5, | |
| "logps/chosen": -390.0, | |
| "logps/rejected": -444.0, | |
| "loss": 0.2165, | |
| "rewards/accuracies": 0.9437500238418579, | |
| "rewards/chosen": -1.109375, | |
| "rewards/margins": 3.078125, | |
| "rewards/rejected": -4.1875, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "step": 210, | |
| "total_flos": 0.0, | |
| "train_loss": 0.4032986215182713, | |
| "train_runtime": 4950.8357, | |
| "train_samples_per_second": 2.727, | |
| "train_steps_per_second": 0.042 | |
| } | |
| ], | |
| "logging_steps": 10, | |
| "max_steps": 210, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 2, | |
| "save_steps": 500, | |
| "total_flos": 0.0, | |
| "train_batch_size": 2, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |