{
  "train_batch_size": "auto",
  "train_micro_batch_size_per_gpu": "auto",
  "gradient_accumulation_steps": "auto",
  "zero_optimization": {
    "stage": 2,
    "offload_optimizer": {
      "device": "none"
    },
    "offload_param": {
      "device": "none"
    },
    "overlap_comm": false,
    "contiguous_gradients": false,
    "reduce_bucket_size": 2e8,
    "allgather_bucket_size": 2e8
  }
}