| { | |
| "best_global_step": null, | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 4.0, | |
| "eval_steps": 500, | |
| "global_step": 20, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "completion_length": 355.3125, | |
| "epoch": 0.42105263157894735, | |
| "grad_norm": 102.69873809814453, | |
| "kl": 0.0, | |
| "learning_rate": 5e-07, | |
| "loss": -0.0, | |
| "reward": 0.21049223048612475, | |
| "reward_std": 0.883908929725294, | |
| "rewards/concensus_correctness_reward_func": 0.0, | |
| "rewards/consensus_reward_func": 0.0, | |
| "rewards/cumulative_reward_2": 0.0, | |
| "rewards/final_correctness_reward_func": 0.0, | |
| "rewards/question_recreation_reward_func": 0.3366484845755622, | |
| "rewards/soft_format_reward_func": 0.0, | |
| "rewards/strict_format_reward_func": 0.0, | |
| "rewards/xmlcount_reward_func": -0.12615624256432056, | |
| "step": 2 | |
| }, | |
| { | |
| "completion_length": 282.46875, | |
| "epoch": 0.8421052631578947, | |
| "grad_norm": 11.320378303527832, | |
| "kl": 0.0014440583036048338, | |
| "learning_rate": 4.864543104251586e-07, | |
| "loss": 0.0, | |
| "reward": 0.9027491768356413, | |
| "reward_std": 0.6032452237559482, | |
| "rewards/concensus_correctness_reward_func": 0.059812501072883606, | |
| "rewards/consensus_reward_func": 0.0625, | |
| "rewards/cumulative_reward_2": 0.0, | |
| "rewards/final_correctness_reward_func": 0.1875, | |
| "rewards/question_recreation_reward_func": 0.36512421316001564, | |
| "rewards/soft_format_reward_func": 0.0, | |
| "rewards/strict_format_reward_func": 0.0, | |
| "rewards/xmlcount_reward_func": 0.22781249205581844, | |
| "step": 4 | |
| }, | |
| { | |
| "completion_length": 455.60714285714283, | |
| "epoch": 1.2105263157894737, | |
| "grad_norm": 5.977652549743652, | |
| "kl": 0.0017597835332188489, | |
| "learning_rate": 4.472851273490984e-07, | |
| "loss": 0.0, | |
| "reward": 0.9831949581525156, | |
| "reward_std": 0.6607641908007541, | |
| "rewards/concensus_correctness_reward_func": 0.13250000136239187, | |
| "rewards/consensus_reward_func": 0.07142857142857142, | |
| "rewards/cumulative_reward_2": 0.0, | |
| "rewards/final_correctness_reward_func": 0.14285714285714285, | |
| "rewards/question_recreation_reward_func": 0.4053377916238138, | |
| "rewards/soft_format_reward_func": 0.0, | |
| "rewards/strict_format_reward_func": 0.0, | |
| "rewards/xmlcount_reward_func": 0.2310714295932225, | |
| "step": 6 | |
| }, | |
| { | |
| "completion_length": 250.28125, | |
| "epoch": 1.631578947368421, | |
| "grad_norm": 25.159584045410156, | |
| "kl": 0.00793385949509684, | |
| "learning_rate": 3.867370395306068e-07, | |
| "loss": 0.0, | |
| "reward": 0.5738056732807308, | |
| "reward_std": 0.6877566411567386, | |
| "rewards/concensus_correctness_reward_func": 0.05999999865889549, | |
| "rewards/consensus_reward_func": 0.0, | |
| "rewards/cumulative_reward_2": 0.0, | |
| "rewards/final_correctness_reward_func": 0.0, | |
| "rewards/question_recreation_reward_func": 0.2658369077835232, | |
| "rewards/soft_format_reward_func": 0.0, | |
| "rewards/strict_format_reward_func": 0.0, | |
| "rewards/xmlcount_reward_func": 0.24796875566244125, | |
| "step": 8 | |
| }, | |
| { | |
| "completion_length": 346.5357142857143, | |
| "epoch": 2.0, | |
| "grad_norm": 5.285414218902588, | |
| "kl": 0.0035223962206925663, | |
| "learning_rate": 3.1137137178519977e-07, | |
| "loss": 0.0, | |
| "reward": 0.7206419948488474, | |
| "reward_std": 0.6630100615854774, | |
| "rewards/concensus_correctness_reward_func": 0.0, | |
| "rewards/consensus_reward_func": 0.0, | |
| "rewards/cumulative_reward_2": 0.0, | |
| "rewards/final_correctness_reward_func": 0.14285714285714285, | |
| "rewards/question_recreation_reward_func": 0.3452491302575384, | |
| "rewards/soft_format_reward_func": 0.0, | |
| "rewards/strict_format_reward_func": 0.0, | |
| "rewards/xmlcount_reward_func": 0.23253571029220307, | |
| "step": 10 | |
| }, | |
| { | |
| "completion_length": 312.03125, | |
| "epoch": 2.4210526315789473, | |
| "grad_norm": 5.405126571655273, | |
| "kl": 0.0024529736692784354, | |
| "learning_rate": 2.2935516363191693e-07, | |
| "loss": 0.0, | |
| "reward": 0.7590260014403611, | |
| "reward_std": 0.7920715521322563, | |
| "rewards/concensus_correctness_reward_func": 0.0, | |
| "rewards/consensus_reward_func": 0.0, | |
| "rewards/cumulative_reward_2": 0.0, | |
| "rewards/final_correctness_reward_func": 0.3125, | |
| "rewards/question_recreation_reward_func": 0.29855725495144725, | |
| "rewards/soft_format_reward_func": 0.0, | |
| "rewards/strict_format_reward_func": 0.0, | |
| "rewards/xmlcount_reward_func": 0.14796874905005097, | |
| "step": 12 | |
| }, | |
| { | |
| "completion_length": 470.5625, | |
| "epoch": 2.8421052631578947, | |
| "grad_norm": 3.935940980911255, | |
| "kl": 0.006017757339577656, | |
| "learning_rate": 1.4957614383675767e-07, | |
| "loss": 0.0, | |
| "reward": 0.764203691855073, | |
| "reward_std": 0.511967592057772, | |
| "rewards/concensus_correctness_reward_func": 0.05999999865889549, | |
| "rewards/consensus_reward_func": 0.0, | |
| "rewards/cumulative_reward_2": 0.0, | |
| "rewards/final_correctness_reward_func": 0.0625, | |
| "rewards/question_recreation_reward_func": 0.3350786907831207, | |
| "rewards/soft_format_reward_func": 0.0, | |
| "rewards/strict_format_reward_func": 0.0, | |
| "rewards/xmlcount_reward_func": 0.3066250067204237, | |
| "step": 14 | |
| }, | |
| { | |
| "completion_length": 378.0, | |
| "epoch": 3.2105263157894735, | |
| "grad_norm": 22.409475326538086, | |
| "kl": 0.005201808779799778, | |
| "learning_rate": 8.067960709356478e-08, | |
| "loss": 0.0, | |
| "reward": 1.00338282798683, | |
| "reward_std": 0.779172311399764, | |
| "rewards/concensus_correctness_reward_func": 0.06835714408329555, | |
| "rewards/consensus_reward_func": 0.07142857142857142, | |
| "rewards/cumulative_reward_2": 0.0, | |
| "rewards/final_correctness_reward_func": 0.0, | |
| "rewards/question_recreation_reward_func": 0.44459709249037715, | |
| "rewards/soft_format_reward_func": 0.0, | |
| "rewards/strict_format_reward_func": 0.0, | |
| "rewards/xmlcount_reward_func": 0.41899999869721277, | |
| "step": 16 | |
| }, | |
| { | |
| "completion_length": 357.3125, | |
| "epoch": 3.6315789473684212, | |
| "grad_norm": 5.217537879943848, | |
| "kl": 0.005285432001983281, | |
| "learning_rate": 3.013156219837776e-08, | |
| "loss": 0.0, | |
| "reward": 1.409838660620153, | |
| "reward_std": 0.8962057730532251, | |
| "rewards/concensus_correctness_reward_func": 0.23587500303983688, | |
| "rewards/consensus_reward_func": 0.125, | |
| "rewards/cumulative_reward_2": 0.0, | |
| "rewards/final_correctness_reward_func": 0.25, | |
| "rewards/question_recreation_reward_func": 0.4597761657787487, | |
| "rewards/soft_format_reward_func": 0.0, | |
| "rewards/strict_format_reward_func": 0.0, | |
| "rewards/xmlcount_reward_func": 0.33918749913573265, | |
| "step": 18 | |
| }, | |
| { | |
| "completion_length": 447.57142857142856, | |
| "epoch": 4.0, | |
| "grad_norm": 6.929130554199219, | |
| "kl": 0.019091870286501944, | |
| "learning_rate": 3.4096741493194193e-09, | |
| "loss": 0.0, | |
| "reward": 0.5597098918099489, | |
| "reward_std": 0.7218386048584112, | |
| "rewards/concensus_correctness_reward_func": 0.0, | |
| "rewards/consensus_reward_func": 0.07142857142857142, | |
| "rewards/cumulative_reward_2": 0.0, | |
| "rewards/final_correctness_reward_func": 0.0, | |
| "rewards/question_recreation_reward_func": 0.34185273446408765, | |
| "rewards/soft_format_reward_func": 0.0, | |
| "rewards/strict_format_reward_func": 0.0, | |
| "rewards/xmlcount_reward_func": 0.1464285653616701, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "step": 20, | |
| "total_flos": 0.0, | |
| "train_loss": 4.90974634885788e-06, | |
| "train_runtime": 2432.4537, | |
| "train_samples_per_second": 0.132, | |
| "train_steps_per_second": 0.008 | |
| } | |
| ], | |
| "logging_steps": 2, | |
| "max_steps": 20, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 5, | |
| "save_steps": 25, | |
| "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": 2, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |