Llama3_1Base-L24M-8x/lm_config.json

{
    "device": "cuda",
    "seed": 42,
    "dtype": "torch.bfloat16",
    "model_name": "meta-llama/Meta-Llama-3.1-8B",
    "use_flash_attn": false,
    "cache_dir": null,
    "d_model": 4096,
    "local_files_only": false
}