quantize_config.json

{
  "bits": 8,
  "group_size": -1,
  "desc_act": false,
  "lm_head": false,
  "method": "fp8",
  "quant_method": "fp8",
  "format": "float8_e4m3fn",
  "checkpoint_format": "float8_e4m3fn",
  "pack_dtype": "int32",
  "meta": {
    "quantizer": [
      "gptqmodel:7.3.4"
    ],
    "uri": "https://github.com/modelcloud/gptqmodel",
    "damp_percent": null,
    "damp_auto_increment": null,
    "static_groups": null,
    "true_sequential": true,
    "mse": null,
    "gptaq": null,
    "foem": null,
    "act_group_aware": null,
    "fallback": {
      "strategy": "rtn",
      "threshold": "0.5%",
      "smooth": null
    },
    "offload_to_disk": false,
    "offload_to_disk_path": null,
    "pack_impl": "cpu",
    "gc_mode": "interval",
    "wait_for_submodule_finalizers": false,
    "auto_forward_data_parallel": true,
    "dense_vram_strategy": "exclusive",
    "dense_vram_strategy_devices": null,
    "moe_vram_strategy": "exclusive",
    "moe_vram_strategy_devices": null
  },
  "weight_scale_method": "row",
  "weight_scale_semantics": "inverse"
}