trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.9874476987447699,
  "eval_steps": 500,
  "global_step": 59,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.016736401673640166,
      "grad_norm": 6.257743265264478,
      "learning_rate": 8.333333333333333e-08,
      "logits/chosen": -2.790410041809082,
      "logits/rejected": -2.7718005180358887,
      "logps/chosen": -148.21420288085938,
      "logps/pi_response": -141.92811584472656,
      "logps/ref_response": -141.92811584472656,
      "logps/rejected": -140.83279418945312,
      "loss": 0.6933,
      "rewards/accuracies": 0.0,
      "rewards/chosen": 0.0,
      "rewards/margins": 0.0,
      "rewards/rejected": 0.0,
      "step": 1
    },
    {
      "epoch": 0.16736401673640167,
      "grad_norm": 6.350719994933479,
      "learning_rate": 4.930057285201027e-07,
      "logits/chosen": -2.7450828552246094,
      "logits/rejected": -2.746872901916504,
      "logps/chosen": -144.24224853515625,
      "logps/pi_response": -146.00637817382812,
      "logps/ref_response": -144.05511474609375,
      "logps/rejected": -148.64894104003906,
      "loss": 0.6932,
      "rewards/accuracies": 0.4583333432674408,
      "rewards/chosen": -0.018231533467769623,
      "rewards/margins": 0.0018064328469336033,
      "rewards/rejected": -0.02003796584904194,
      "step": 10
    },
    {
      "epoch": 0.33472803347280333,
      "grad_norm": 6.211060932541832,
      "learning_rate": 4.187457503795526e-07,
      "logits/chosen": -2.7610819339752197,
      "logits/rejected": -2.7687811851501465,
      "logps/chosen": -148.03451538085938,
      "logps/pi_response": -146.87643432617188,
      "logps/ref_response": -142.35841369628906,
      "logps/rejected": -148.18875122070312,
      "loss": 0.6933,
      "rewards/accuracies": 0.484375,
      "rewards/chosen": -0.04550832509994507,
      "rewards/margins": -0.001607399433851242,
      "rewards/rejected": -0.043900929391384125,
      "step": 20
    },
    {
      "epoch": 0.502092050209205,
      "grad_norm": 6.192229523864876,
      "learning_rate": 2.8691164100062034e-07,
      "logits/chosen": -2.685403347015381,
      "logits/rejected": -2.6824450492858887,
      "logps/chosen": -142.15756225585938,
      "logps/pi_response": -141.01730346679688,
      "logps/ref_response": -141.1864776611328,
      "logps/rejected": -140.42800903320312,
      "loss": 0.6929,
      "rewards/accuracies": 0.5062500238418579,
      "rewards/chosen": 0.004916164558380842,
      "rewards/margins": 0.0020810612477362156,
      "rewards/rejected": 0.0028351028449833393,
      "step": 30
    },
    {
      "epoch": 0.6694560669456067,
      "grad_norm": 7.508024203763283,
      "learning_rate": 1.4248369943086995e-07,
      "logits/chosen": -2.698911190032959,
      "logits/rejected": -2.6966421604156494,
      "logps/chosen": -137.83132934570312,
      "logps/pi_response": -141.4623260498047,
      "logps/ref_response": -143.17706298828125,
      "logps/rejected": -141.35003662109375,
      "loss": 0.6916,
      "rewards/accuracies": 0.534375011920929,
      "rewards/chosen": 0.018742507323622704,
      "rewards/margins": 0.004439900163561106,
      "rewards/rejected": 0.01430260855704546,
      "step": 40
    },
    {
      "epoch": 0.8368200836820083,
      "grad_norm": 7.1415412733846,
      "learning_rate": 3.473909705816111e-08,
      "logits/chosen": -2.7535297870635986,
      "logits/rejected": -2.7535128593444824,
      "logps/chosen": -147.931396484375,
      "logps/pi_response": -148.46963500976562,
      "logps/ref_response": -147.69802856445312,
      "logps/rejected": -146.8707275390625,
      "loss": 0.6915,
      "rewards/accuracies": 0.5375000238418579,
      "rewards/chosen": -0.006227564066648483,
      "rewards/margins": 0.007055486552417278,
      "rewards/rejected": -0.013283051550388336,
      "step": 50
    },
    {
      "epoch": 0.9874476987447699,
      "step": 59,
      "total_flos": 0.0,
      "train_loss": 0.6926169082269831,
      "train_runtime": 2613.3667,
      "train_samples_per_second": 5.848,
      "train_steps_per_second": 0.023
    }
  ],
  "logging_steps": 10,
  "max_steps": 59,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 100,
  "total_flos": 0.0,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}