[BUG] loading model error

Describe the bug
I use 8 * A100-80GB GPUs to fine-tune a 72B model, but when it loads 16/37 safetensor, it always been shutdown for no reasons without any hints!

The error is as follows:

Loading checkpoint Shards"43%16/37 [19:49<25:57，74.18s/it] 
 Sending process4029695 closing signal SIGTERK
failad (edtcote: -5) lcalrank:5 

I use zero-3 and acclerate to launch my script. The command I use is accelerate launch --config_file xxx --gpu_ids 0,1,2,3,4,5,6,7 xxx.py.

The deepspeed file is as follows:

{
    "fp16": {
        "enabled": true,
        "loss_scale": 0,
        "loss_scale_window": 1000,
        "initial_scale_power": 16,
        "hysteresis": 2,
        "min_loss_scale": 1
    },
    "optimizer": {
        "type": "AdamW",
        "params": {
            "lr": "auto",
            "weight_decay": "auto"
        }
    },
    "scheduler": {
        "type": "WarmupDecayLR",
        "params": {
            "warmup_min_lr": "auto",
            "warmup_max_lr": "auto",
            "warmup_num_steps": "auto",
            "total_num_steps": "auto"
        }
    },
    "zero_optimization": {
        "stage": 3,
        "offload_optimizer": {
            "device": "cpu",
            "pin_memory": true
        },
        "offload_param": {
            "device": "cpu",
            "pin_memory": true
        },
        "overlap_comm": true,
        "contiguous_gradients": true,
        "reduce_bucket_size": "auto",
        "stage3_prefetch_bucket_size": "auto",
        "stage3_param_persistence_threshold": "auto",
        "sub_group_size": 1e9,
        "stage3_max_live_parameters": 1e9,
        "stage3_max_reuse_distance": 1e9,
        "stage3_gather_16bit_weights_on_model_save": "auto"
    },
    "gradient_accumulation_steps": 1,
    "gradient_clipping": "auto",
    "steps_per_print": 2000,
    "train_batch_size": "auto",
    "train_micro_batch_size_per_gpu": "auto",
    "wall_clock_breakdown": false
}

Expected behavior

I think it should load all of safetensors, what's more, during the loading process, I notice that gpu memory is still 0.

Screenshots

System info (please complete the following information):

• OS: Ubuntu 18.04
• GPU count and types: 8*A100-80GB
• Python version 3.12

Start by reproducing the reported command with the provided DeepSpeed ZeRO-3 configuration on 8 A100-80GB GPUs while loading the 72B model. Inspect the checkpoint-loading stage around shard 16/37 and collect the process termination and system logs. Done means the shutdown cause is identified and loading completes or the issue is narrowed to a documented failure condition.