configs/ds_zero3.json

{
    "train_batch_size": 8,
    "gradient_accumulation_steps": 16,
    "steps_per_print": 10,
    "gradient_clipping": 1.0,
    "zero_optimization": {
        "stage": 3,
        "offload_optimizer": {
            "device": "cpu",
            "pin_memory": true
        },
        "offload_param": {
            "device": "cpu",
            "pin_memory": true
        },
        "overlap_comm": true,
        "contiguous_gradients": true,
        "sub_group_size": 1e6,
        "reduce_bucket_size": "auto",
        "stage3_prefetch_bucket_size": "auto",
        "stage3_param_persistence_threshold": "auto",
        "stage3_max_live_parameters": 1e9,
        "stage3_max_reuse_distance": 1e9,
        "stage3_gather_16bit_weights_on_model_save": true
    },
    "bf16": {
        "enabled": true
    },
    "fp16": {
        "enabled": false
    },
    "optimizer": {
        "type": "AdamW",
        "params": {
            "lr": 1e-5,
            "betas": [0.9, 0.95],
            "eps": 1e-8,
            "weight_decay": 0.01
        }
    },
    "scheduler": {
        "type": "WarmupLR",
        "params": {
            "warmup_min_lr": 0,
            "warmup_max_lr": 1e-5,
            "warmup_num_steps": 100
        }
    },
    "activation_checkpointing": {
        "partition_activations": true,
        "cpu_checkpointing": true,
        "contiguous_memory_optimization": true
    },
    "wall_clock_breakdown": false
}