{
  "results": {
    "coqa": {
      "f1": 0.6882976860781418,
      "f1_stderr": 0.016322647326969194,
      "em": 0.5386666666666665,
      "em_stderr": 0.01995482540089559
    },
    "drop": {
      "em": 0.02569211409395973,
      "em_stderr": 0.0016202710827118362,
      "f1": 0.09853712248322138,
      "f1_stderr": 0.0021424507419289577
    },
    "race": {
      "acc": 0.36555023923444974,
      "acc_stderr": 0.014904654247182307
    }
  },
  "versions": {
    "coqa": 1,
    "race": 1,
    "drop": 1
  },
  "config": {
    "model": "hf-causal-experimental",
    "model_args": "pretrained=bigscience/bloom-7b1,use_accelerate=True",
    "num_fewshot": 0,
    "batch_size": "auto",
    "device": "cuda:0",
    "no_cache": true,
    "limit": null,
    "bootstrap_iters": 100000,
    "description_dict": {}
  }
}