{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.9968652037617555,
  "eval_steps": 500,
  "global_step": 159,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.01,
      "learning_rate": 3.125e-08,
      "logits/chosen": -1.8967804908752441,
      "logits/rejected": -1.9837573766708374,
      "logps/chosen": -526.0550537109375,
      "logps/rejected": -649.3060302734375,
      "loss": 0.6931,
      "rewards/accuracies": 0.0,
      "rewards/chosen": 0.0,
      "rewards/margins": 0.0,
      "rewards/rejected": 0.0,
      "step": 1
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.1249999999999997e-07,
      "logits/chosen": -1.896860122680664,
      "logits/rejected": -1.8309091329574585,
      "logps/chosen": -482.1170349121094,
      "logps/rejected": -534.401123046875,
      "loss": 0.6896,
      "rewards/accuracies": 0.4722222089767456,
      "rewards/chosen": -0.026020465418696404,
      "rewards/margins": 0.004157062619924545,
      "rewards/rejected": -0.030177529901266098,
      "step": 10
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.990353313429303e-07,
      "logits/chosen": -1.636963129043579,
      "logits/rejected": -1.5652328729629517,
      "logps/chosen": -518.24560546875,
      "logps/rejected": -692.5880126953125,
      "loss": 0.6268,
      "rewards/accuracies": 0.65625,
      "rewards/chosen": -0.7620652914047241,
      "rewards/margins": 0.2627033591270447,
      "rewards/rejected": -1.024768590927124,
      "step": 20
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.882681251368548e-07,
      "logits/chosen": -0.6123544573783875,
      "logits/rejected": -0.5012580752372742,
      "logps/chosen": -642.9691162109375,
      "logps/rejected": -757.7621459960938,
      "loss": 0.6561,
      "rewards/accuracies": 0.643750011920929,
      "rewards/chosen": -1.682318091392517,
      "rewards/margins": 0.44499659538269043,
      "rewards/rejected": -2.127314567565918,
      "step": 30
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.6604720940421207e-07,
      "logits/chosen": -0.06902433931827545,
      "logits/rejected": 0.01894857920706272,
      "logps/chosen": -600.1622314453125,
      "logps/rejected": -754.4539794921875,
      "loss": 0.6058,
      "rewards/accuracies": 0.643750011920929,
      "rewards/chosen": -1.2441052198410034,
      "rewards/margins": 0.34002283215522766,
      "rewards/rejected": -1.5841280221939087,
      "step": 40
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.3344075855595097e-07,
      "logits/chosen": -0.14471003413200378,
      "logits/rejected": 0.360783189535141,
      "logps/chosen": -613.9149780273438,
      "logps/rejected": -760.083740234375,
      "loss": 0.5816,
      "rewards/accuracies": 0.65625,
      "rewards/chosen": -1.1845529079437256,
      "rewards/margins": 0.4947661757469177,
      "rewards/rejected": -1.6793190240859985,
      "step": 50
    },
    {
      "epoch": 0.38,
      "learning_rate": 3.920161866827889e-07,
      "logits/chosen": 0.5329059362411499,
      "logits/rejected": 0.6899431943893433,
      "logps/chosen": -675.7152099609375,
      "logps/rejected": -822.51025390625,
      "loss": 0.5791,
      "rewards/accuracies": 0.71875,
      "rewards/chosen": -1.8414958715438843,
      "rewards/margins": 0.6188977956771851,
      "rewards/rejected": -2.4603934288024902,
      "step": 60
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.4376480090239047e-07,
      "logits/chosen": 0.21469660103321075,
      "logits/rejected": 0.6365100145339966,
      "logps/chosen": -613.69921875,
      "logps/rejected": -791.498046875,
      "loss": 0.5683,
      "rewards/accuracies": 0.731249988079071,
      "rewards/chosen": -1.4487526416778564,
      "rewards/margins": 0.7974711656570435,
      "rewards/rejected": -2.2462239265441895,
      "step": 70
    },
    {
      "epoch": 0.5,
      "learning_rate": 2.910060778827554e-07,
      "logits/chosen": 0.050884656608104706,
      "logits/rejected": 0.49703994393348694,
      "logps/chosen": -631.5872192382812,
      "logps/rejected": -790.6751708984375,
      "loss": 0.5308,
      "rewards/accuracies": 0.71875,
      "rewards/chosen": -1.3860604763031006,
      "rewards/margins": 0.5967694520950317,
      "rewards/rejected": -1.9828300476074219,
      "step": 80
    },
    {
      "epoch": 0.56,
      "learning_rate": 2.3627616503391812e-07,
      "logits/chosen": 0.4010600447654724,
      "logits/rejected": 0.7713971138000488,
      "logps/chosen": -658.26513671875,
      "logps/rejected": -814.2492065429688,
      "loss": 0.5641,
      "rewards/accuracies": 0.737500011920929,
      "rewards/chosen": -1.5723729133605957,
      "rewards/margins": 0.7238161563873291,
      "rewards/rejected": -2.296189308166504,
      "step": 90
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.8220596619089573e-07,
      "logits/chosen": 0.13108967244625092,
      "logits/rejected": 0.49605846405029297,
      "logps/chosen": -634.7674560546875,
      "logps/rejected": -771.0775146484375,
      "loss": 0.547,
      "rewards/accuracies": 0.731249988079071,
      "rewards/chosen": -1.3438690900802612,
      "rewards/margins": 0.6646407246589661,
      "rewards/rejected": -2.008509874343872,
      "step": 100
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.3139467229135998e-07,
      "logits/chosen": 0.21024885773658752,
      "logits/rejected": 0.666106641292572,
      "logps/chosen": -694.7786254882812,
      "logps/rejected": -792.2867431640625,
      "loss": 0.5662,
      "rewards/accuracies": 0.6812499761581421,
      "rewards/chosen": -1.5787041187286377,
      "rewards/margins": 0.609796404838562,
      "rewards/rejected": -2.18850040435791,
      "step": 110
    },
    {
      "epoch": 0.75,
      "learning_rate": 8.628481651367875e-08,
      "logits/chosen": 0.2621000409126282,
      "logits/rejected": 0.6290411949157715,
      "logps/chosen": -641.9815673828125,
      "logps/rejected": -823.8590087890625,
      "loss": 0.531,
      "rewards/accuracies": 0.71875,
      "rewards/chosen": -1.6526798009872437,
      "rewards/margins": 0.5855975151062012,
      "rewards/rejected": -2.2382771968841553,
      "step": 120
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.904486005914027e-08,
      "logits/chosen": 0.3799557089805603,
      "logits/rejected": 0.6229809522628784,
      "logps/chosen": -664.0032348632812,
      "logps/rejected": -802.6885986328125,
      "loss": 0.5458,
      "rewards/accuracies": 0.65625,
      "rewards/chosen": -1.6382343769073486,
      "rewards/margins": 0.5937812924385071,
      "rewards/rejected": -2.232015609741211,
      "step": 130
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.1464952759020856e-08,
      "logits/chosen": 0.492957204580307,
      "logits/rejected": 0.9118458032608032,
      "logps/chosen": -631.5418090820312,
      "logps/rejected": -813.458251953125,
      "loss": 0.5444,
      "rewards/accuracies": 0.6625000238418579,
      "rewards/chosen": -1.540797472000122,
      "rewards/margins": 0.7339878082275391,
      "rewards/rejected": -2.274785280227661,
      "step": 140
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.8708793644441086e-09,
      "logits/chosen": -0.018977392464876175,
      "logits/rejected": 0.5174835920333862,
      "logps/chosen": -643.7967529296875,
      "logps/rejected": -808.516357421875,
      "loss": 0.5419,
      "rewards/accuracies": 0.793749988079071,
      "rewards/chosen": -1.3859628438949585,
      "rewards/margins": 0.8056131601333618,
      "rewards/rejected": -2.1915760040283203,
      "step": 150
    },
    {
      "epoch": 1.0,
      "step": 159,
      "total_flos": 0.0,
      "train_loss": 0.5745523470752644,
      "train_runtime": 2655.032,
      "train_samples_per_second": 7.675,
      "train_steps_per_second": 0.06
    }
  ],
  "logging_steps": 10,
  "max_steps": 159,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 100,
  "total_flos": 0.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}