ds_zero2.json

{
    "zero_optimization": {
        "stage": 2,
        "offload_optimizer": {
            "device": "auto",
            "pin_memory": true
        },
        "allgather_partitions": true,
        "allgather_bucket_size": "auto",
        "overlap_comm": true,
        "reduce_scatter": true,
        "reduce_bucket_size": "auto",
        "contiguous_gradients": true,
        "round_robin_gradients": true
    },
    "fp16": {
        "enabled": false,
        "auto_cast": true,
        "loss_scale": "auto",
        "loss_scale_window": "auto",
        "hysteresis": "auto",
        "min_loss_scale": "auto"
    },
    "bf16":{
        "enabled":true
    },
    "gradient_accumulation_steps": "auto",
    "gradient_clipping": 1.0,
    "train_batch_size": "auto",
    "train_micro_batch_size_per_gpu": "auto",
    "steps_per_print": "auto",
    "wall_clock_breakdown": false,
    "eigenvalue_enabled": true
}