Minimind/ds_config.json

{
    "train_batch_size": "auto",
    "train_micro_batch_size_per_gpu": "auto",
    "gradient_accumulation_steps": "auto",
    "gradient_clipping": "auto",
    "zero_optimization": {
        "stage": 2,
        "offload_optimizer": {
            "device": "cpu",
            "pin_memory": true,
            "buffer_count": 4,
            "fast_init": false
        },
        "offload_param": {
            "device": "cpu",
            "pin_memory": true,
            "buffer_count": 4,
            "buffer_size": 1e8,
            "max_in_cpu": 1e9
        },
        "allgather_partitions": true,
        "allgather_bucket_size": 2e8,
        "overlap_comm": true,
        "reduce_scatter": true,
        "reduce_bucket_size": 2e8,
        "contiguous_gradients": true,
        "cpu_offload": true
    },
    "fp16": {
        "enabled": false
    },
    "bf16": {
        "enabled": true
    },
    "aio": {
        "block_size": 1048576,
        "queue_depth": 8,
        "thread_count": 1,
        "single_submit": false,
        "overlap_events": true
    },
    "steps_per_print": 100,
    "wall_clock_breakdown": false,
    "memory_breakdown": false
}