smash_config.json

{
    "batcher": null,
    "cacher": null,
    "compiler": null,
    "factorizer": null,
    "pruner": null,
    "quantizer": "llm_int8",
    "llm_int8_compute_dtype": "bfloat16",
    "llm_int8_double_quant": false,
    "llm_int8_enable_fp32_cpu_offload": false,
    "llm_int8_has_fp16_weight": false,
    "llm_int8_quant_type": "fp4",
    "llm_int8_threshold": 6.0,
    "llm_int8_weight_bits": 8,
    "batch_size": 1,
    "device": "cuda",
    "device_map": null,
    "save_fns": [],
    "load_fns": [
        "transformers"
    ],
    "reapply_after_load": {
        "factorizer": null,
        "pruner": null,
        "quantizer": null,
        "cacher": null,
        "compiler": null,
        "batcher": null
    }
}