Class Model

examples/server/server.py:11229–12299

class Model:
    @dataclass(frozen=True)
    class LoraAdapter:
        path: str
        scale: float = 1.0

    def __init__(
        self,
        *,
        model_path: str,
        model_alias: Optional[str] = None,
        chat_template: Optional[str] = None,
        loras: Optional[List["Model.LoraAdapter"]] = None,
        n_gpu_layers: Optional[int] = None,
        split_mode: Optional[int] = None,
        main_gpu: Optional[int] = None,
        tensor_split: Optional[List[float]] = None,
        vocab_only: Optional[bool] = None,
        use_mmap: Optional[bool] = None,
        use_mlock: Optional[bool] = None,
        kv_overrides: Optional[Dict[str, Union[bool, int, float, str]]] = None,
        n_ctx: Optional[int],
        n_batch: Optional[int],
        n_ubatch: Optional[int] = None,
        n_seq_max: Optional[int],
        n_threads: Optional[int],
        n_threads_batch: Optional[int],
        rope_scaling_type: Optional[int] = None,
        pooling_type: Optional[int] = None,
        attention_type: Optional[int] = None,
        embedding: Optional[bool] = None,
        rope_freq_base: Optional[float] = None,
        rope_freq_scale: Optional[float] = None,
        yarn_ext_factor: Optional[float] = None,
        yarn_attn_factor: Optional[float] = None,
        yarn_beta_fast: Optional[float] = None,
        yarn_beta_slow: Optional[float] = None,
        yarn_orig_ctx: Optional[int] = None,
        offload_kqv: Optional[bool] = None,
        flash_attn: Optional[bool] = None,
        op_offload: Optional[bool] = None,
        swa_full: Optional[bool] = None,
        no_perf: Optional[bool] = None,
        type_k: Optional[int] = None,
        type_v: Optional[int] = None,
        kv_unified: bool = True,
        max_seq_len: Optional[int] = None,
        max_output_tokens: Optional[int] = None,
        draft_model: Optional[str] = None,
        draft_model_path: Optional[str] = None,
        draft_model_num_pred_tokens: int = 16,
        draft_model_max_ngram_size: int = 2,
        draft_model_top_k: int = 1,
        draft_model_p_min: float = 0.0,
        draft_model_max_batch_size: Optional[int] = None,
        draft_model_threads: Optional[int] = None,
        draft_model_threads_batch: Optional[int] = None,
        response_schema: Optional[Dict[str, Any]] = None,