{ "dtype": "int8_int4", "input_info": null, "optimum_version": "2.0.0", "quantization_config": { "dataset": "contextual", "default_config": null, "ignored_scope": null, "num_samples": null, "processor": null, "quantization_configs": { "lm_model": { "bits": 8, "dataset": null, "dtype": "int8", "fast_bias_correction": true, "ignored_scope": null, "model_type": "transformer", "num_samples": null, "overflow_fix": "disable", "processor": null, "smooth_quant_alpha": null, "sym": false, "tokenizer": null, "trust_remote_code": false }, "text_embeddings_model": { "all_layers": null, "backup_precision": null, "bits": 4, "dataset": null, "dtype": "int4", "gptq": null, "group_size": null, "ignored_scope": null, "lora_correction": null, "num_samples": null, "processor": null, "quant_method": "default", "ratio": 1.0, "scale_estimation": null, "sensitivity_metric": null, "statistics_path": null, "sym": false, "tokenizer": null, "trust_remote_code": false } }, "tokenizer": null, "trust_remote_code": true }, "save_onnx_model": false, "transformers_version": "4.52.4" }