| { | |
| "best_metric": 0.0015446161851286888, | |
| "best_model_checkpoint": "checkpoints/Qwen2.5-VL-7B-Instruct-GRPO-3DSRBench-small-2025-02-10-22-17-37/checkpoint-16", | |
| "epoch": 4.0, | |
| "eval_steps": 500, | |
| "global_step": 16, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "completion_length": 104.45833587646484, | |
| "epoch": 0.25, | |
| "grad_norm": 12.969447688340658, | |
| "kl": 0.0, | |
| "learning_rate": 9.374999999999999e-07, | |
| "loss": -0.0, | |
| "reward": 0.3333333432674408, | |
| "reward_std": 0.2357022613286972, | |
| "rewards/accuracy_reward": 0.3333333432674408, | |
| "rewards/format_reward": 0.0, | |
| "step": 1 | |
| }, | |
| { | |
| "completion_length": 58.5, | |
| "epoch": 0.5, | |
| "grad_norm": 469.390542705169, | |
| "kl": 0.005462646484375, | |
| "learning_rate": 8.75e-07, | |
| "loss": 0.0002, | |
| "reward": 0.4166666865348816, | |
| "reward_std": 0.33247750997543335, | |
| "rewards/accuracy_reward": 0.4166666865348816, | |
| "rewards/format_reward": 0.0, | |
| "step": 2 | |
| }, | |
| { | |
| "completion_length": 55.5, | |
| "epoch": 0.75, | |
| "grad_norm": 232.56452697554002, | |
| "kl": 0.053466796875, | |
| "learning_rate": 8.125e-07, | |
| "loss": 0.0021, | |
| "reward": 0.625, | |
| "reward_std": 0.48112308979034424, | |
| "rewards/accuracy_reward": 0.625, | |
| "rewards/format_reward": 0.0, | |
| "step": 3 | |
| }, | |
| { | |
| "completion_length": 91.0, | |
| "epoch": 1.0, | |
| "grad_norm": 193.90471690480015, | |
| "kl": 0.0693359375, | |
| "learning_rate": 7.5e-07, | |
| "loss": 0.003, | |
| "reward": 0.0, | |
| "reward_std": 0.0, | |
| "rewards/accuracy_reward": 0.0, | |
| "rewards/format_reward": 0.0, | |
| "step": 4 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_completion_length": 94.5999984741211, | |
| "eval_kl": 0.1806640625, | |
| "eval_loss": 0.006281447596848011, | |
| "eval_reward": 0.0, | |
| "eval_reward_std": 0.0, | |
| "eval_rewards/accuracy_reward": 0.0, | |
| "eval_rewards/format_reward": 0.0, | |
| "eval_runtime": 8.9021, | |
| "eval_samples_per_second": 0.562, | |
| "eval_steps_per_second": 0.112, | |
| "step": 4 | |
| }, | |
| { | |
| "completion_length": 59.54166793823242, | |
| "epoch": 1.25, | |
| "grad_norm": 11.754392724379205, | |
| "kl": 0.1416015625, | |
| "learning_rate": 6.875e-07, | |
| "loss": 0.0056, | |
| "reward": 0.5416666865348816, | |
| "reward_std": 0.17251639068126678, | |
| "rewards/accuracy_reward": 0.5416666865348816, | |
| "rewards/format_reward": 0.0, | |
| "step": 5 | |
| }, | |
| { | |
| "completion_length": 64.91667175292969, | |
| "epoch": 1.5, | |
| "grad_norm": 13.750858312191527, | |
| "kl": 0.2255859375, | |
| "learning_rate": 6.249999999999999e-07, | |
| "loss": 0.009, | |
| "reward": 0.3333333432674408, | |
| "reward_std": 0.0, | |
| "rewards/accuracy_reward": 0.3333333432674408, | |
| "rewards/format_reward": 0.0, | |
| "step": 6 | |
| }, | |
| { | |
| "completion_length": 75.125, | |
| "epoch": 1.75, | |
| "grad_norm": 15.657561566262425, | |
| "kl": 0.037109375, | |
| "learning_rate": 5.625e-07, | |
| "loss": 0.0015, | |
| "reward": 0.4583333432674408, | |
| "reward_std": 0.4082186818122864, | |
| "rewards/accuracy_reward": 0.4583333432674408, | |
| "rewards/format_reward": 0.0, | |
| "step": 7 | |
| }, | |
| { | |
| "completion_length": 73.0, | |
| "epoch": 2.0, | |
| "grad_norm": 54.0708979591708, | |
| "kl": 0.041259765625, | |
| "learning_rate": 5e-07, | |
| "loss": 0.0057, | |
| "reward": 0.0, | |
| "reward_std": 0.0, | |
| "rewards/accuracy_reward": 0.0, | |
| "rewards/format_reward": 0.0, | |
| "step": 8 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_completion_length": 87.5999984741211, | |
| "eval_kl": 0.09228515625, | |
| "eval_loss": 0.0028523791115731, | |
| "eval_reward": 0.0, | |
| "eval_reward_std": 0.0, | |
| "eval_rewards/accuracy_reward": 0.0, | |
| "eval_rewards/format_reward": 0.0, | |
| "eval_runtime": 9.426, | |
| "eval_samples_per_second": 0.53, | |
| "eval_steps_per_second": 0.106, | |
| "step": 8 | |
| }, | |
| { | |
| "completion_length": 39.375, | |
| "epoch": 2.25, | |
| "grad_norm": 448.1591916712212, | |
| "kl": 0.275390625, | |
| "learning_rate": 4.375e-07, | |
| "loss": 0.011, | |
| "reward": 0.5833333730697632, | |
| "reward_std": 0.4993361234664917, | |
| "rewards/accuracy_reward": 0.5833333730697632, | |
| "rewards/format_reward": 0.0, | |
| "step": 9 | |
| }, | |
| { | |
| "completion_length": 59.833335876464844, | |
| "epoch": 2.5, | |
| "grad_norm": 14.127720206827185, | |
| "kl": 0.0390625, | |
| "learning_rate": 3.75e-07, | |
| "loss": 0.0016, | |
| "reward": 0.875, | |
| "reward_std": 0.17251639068126678, | |
| "rewards/accuracy_reward": 0.875, | |
| "rewards/format_reward": 0.0, | |
| "step": 10 | |
| }, | |
| { | |
| "completion_length": 34.708335876464844, | |
| "epoch": 2.75, | |
| "grad_norm": 307.5804345526168, | |
| "kl": 0.06396484375, | |
| "learning_rate": 3.1249999999999997e-07, | |
| "loss": 0.0026, | |
| "reward": 0.2916666865348816, | |
| "reward_std": 0.2721545100212097, | |
| "rewards/accuracy_reward": 0.2916666865348816, | |
| "rewards/format_reward": 0.0, | |
| "step": 11 | |
| }, | |
| { | |
| "completion_length": 4.0, | |
| "epoch": 3.0, | |
| "grad_norm": 33.51591785639859, | |
| "kl": 0.1171875, | |
| "learning_rate": 2.5e-07, | |
| "loss": 0.0024, | |
| "reward": 1.0, | |
| "reward_std": 0.0, | |
| "rewards/accuracy_reward": 1.0, | |
| "rewards/format_reward": 0.0, | |
| "step": 12 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_completion_length": 82.20000457763672, | |
| "eval_kl": 0.07275390625, | |
| "eval_loss": 0.0028513565193861723, | |
| "eval_reward": 0.20000000298023224, | |
| "eval_reward_std": 0.46291008591651917, | |
| "eval_rewards/accuracy_reward": 0.20000000298023224, | |
| "eval_rewards/format_reward": 0.0, | |
| "eval_runtime": 9.0691, | |
| "eval_samples_per_second": 0.551, | |
| "eval_steps_per_second": 0.11, | |
| "step": 12 | |
| }, | |
| { | |
| "completion_length": 66.45833587646484, | |
| "epoch": 3.25, | |
| "grad_norm": 23.952974818627894, | |
| "kl": 0.0242919921875, | |
| "learning_rate": 1.875e-07, | |
| "loss": 0.001, | |
| "reward": 0.7916666865348816, | |
| "reward_std": 0.3268197476863861, | |
| "rewards/accuracy_reward": 0.7916666865348816, | |
| "rewards/format_reward": 0.0, | |
| "step": 13 | |
| }, | |
| { | |
| "completion_length": 82.04167175292969, | |
| "epoch": 3.5, | |
| "grad_norm": 6.6087277944458, | |
| "kl": 0.02783203125, | |
| "learning_rate": 1.25e-07, | |
| "loss": 0.0011, | |
| "reward": 0.5416666865348816, | |
| "reward_std": 0.17251639068126678, | |
| "rewards/accuracy_reward": 0.5416666865348816, | |
| "rewards/format_reward": 0.0, | |
| "step": 14 | |
| }, | |
| { | |
| "completion_length": 58.875, | |
| "epoch": 3.75, | |
| "grad_norm": 58.01807470412879, | |
| "kl": 0.041748046875, | |
| "learning_rate": 6.25e-08, | |
| "loss": 0.0017, | |
| "reward": 0.25, | |
| "reward_std": 0.4629100561141968, | |
| "rewards/accuracy_reward": 0.25, | |
| "rewards/format_reward": 0.0, | |
| "step": 15 | |
| }, | |
| { | |
| "completion_length": 23.0, | |
| "epoch": 4.0, | |
| "grad_norm": 6.755101305420566, | |
| "kl": 0.080078125, | |
| "learning_rate": 0.0, | |
| "loss": 0.0015, | |
| "reward": 0.0, | |
| "reward_std": 0.0, | |
| "rewards/accuracy_reward": 0.0, | |
| "rewards/format_reward": 0.0, | |
| "step": 16 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_completion_length": 78.0, | |
| "eval_kl": 0.05615234375, | |
| "eval_loss": 0.0015446161851286888, | |
| "eval_reward": 0.0, | |
| "eval_reward_std": 0.0, | |
| "eval_rewards/accuracy_reward": 0.0, | |
| "eval_rewards/format_reward": 0.0, | |
| "eval_runtime": 10.3374, | |
| "eval_samples_per_second": 0.484, | |
| "eval_steps_per_second": 0.097, | |
| "step": 16 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "step": 16, | |
| "total_flos": 0.0, | |
| "train_loss": 0.003118676513016283, | |
| "train_runtime": 613.0331, | |
| "train_samples_per_second": 0.065, | |
| "train_steps_per_second": 0.026 | |
| } | |
| ], | |
| "logging_steps": 1.0, | |
| "max_steps": 16, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 4, | |
| "save_steps": 500, | |
| "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": 1, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |