Interface LoadModelParams

interface LoadModelParams {
    cache_idle_slots?: boolean;
    cache_type_k?: "f32" | "f16" | "q8_0" | "q5_1" | "q5_0" | "q4_1" | "q4_0";
    cache_type_v?: "f32" | "f16" | "q8_0" | "q5_1" | "q5_0" | "q4_1" | "q4_0";
    chat_template?: string;
    cont_batching?: boolean;
    ctx_shift?: boolean;
    default_template_kwargs?: Record<string, any>;
    embeddings?: boolean;
    flash_attn?: boolean;
    image_max_tokens?: number;
    image_min_tokens?: number;
    jinja?: boolean;
    kv_overrides?: Record<string, string>;
    log_level?: LogLevel;
    lora_adapters?: { path: string; scale?: number }[];
    lora_init_without_apply?: boolean;
    mmproj_offload?: boolean;
    n_batch?: number;
    n_cache_reuse?: number;
    n_ctx?: number;
    n_gpu_layers?: number;
    n_keep?: number;
    n_threads?: number;
    no_kv_offload?: boolean;
    offload_kqv?: boolean;
    pooling_type?:
        | "LLAMA_POOLING_TYPE_UNSPECIFIED"
        | "LLAMA_POOLING_TYPE_NONE"
        | "LLAMA_POOLING_TYPE_MEAN"
        | "LLAMA_POOLING_TYPE_CLS"
        | "unspecified"
        | "none"
        | "mean"
        | "cls"
        | "last"
        | "rank";
    prefill_assistant?: boolean;
    reasoning?: boolean;
    reasoning_budget_message?: string;
    reasoning_budget_tokens?: number;
    reasoning_format?: "none"
    | "deepseek-legacy"
    | "deepseek";
    rope_freq_base?: number;
    rope_freq_scale?: number;
    rope_scaling_type?:
        | "LLAMA_ROPE_SCALING_TYPE_UNSPECIFIED"
        | "LLAMA_ROPE_SCALING_TYPE_NONE"
        | "LLAMA_ROPE_SCALING_TYPE_LINEAR"
        | "LLAMA_ROPE_SCALING_TYPE_YARN";
    seed?: number;
    skip_chat_parsing?: boolean;
    spec_draft_model?: string;
    spec_draft_n_max?: number;
    spec_draft_n_min?: number;
    spec_draft_ngl?: number;
    spec_draft_p_min?: number;
    spec_draft_threads?: number;
    spec_draft_threads_batch?: number;
    swa_full?: boolean;
    warmup?: boolean;
    yarn_attn_factor?: number;
    yarn_beta_fast?: number;
    yarn_beta_slow?: number;
    yarn_ext_factor?: number;
    yarn_orig_ctx?: number;
}

Properties

cache_idle_slots?: boolean
cache_type_k?: "f32" | "f16" | "q8_0" | "q5_1" | "q5_0" | "q4_1" | "q4_0"
cache_type_v?: "f32" | "f16" | "q8_0" | "q5_1" | "q5_0" | "q4_1" | "q4_0"
chat_template?: string
cont_batching?: boolean
ctx_shift?: boolean
default_template_kwargs?: Record<string, any>
embeddings?: boolean
flash_attn?: boolean
image_max_tokens?: number
image_min_tokens?: number
jinja?: boolean
kv_overrides?: Record<string, string>
log_level?: LogLevel
lora_adapters?: { path: string; scale?: number }[]
lora_init_without_apply?: boolean
mmproj_offload?: boolean
n_batch?: number
n_cache_reuse?: number
n_ctx?: number
n_gpu_layers?: number
n_keep?: number
n_threads?: number
no_kv_offload?: boolean
offload_kqv?: boolean
pooling_type?:
    | "LLAMA_POOLING_TYPE_UNSPECIFIED"
    | "LLAMA_POOLING_TYPE_NONE"
    | "LLAMA_POOLING_TYPE_MEAN"
    | "LLAMA_POOLING_TYPE_CLS"
    | "unspecified"
    | "none"
    | "mean"
    | "cls"
    | "last"
    | "rank"
prefill_assistant?: boolean
reasoning?: boolean
reasoning_budget_message?: string
reasoning_budget_tokens?: number
reasoning_format?: "none" | "deepseek-legacy" | "deepseek"
rope_freq_base?: number
rope_freq_scale?: number
rope_scaling_type?:
    | "LLAMA_ROPE_SCALING_TYPE_UNSPECIFIED"
    | "LLAMA_ROPE_SCALING_TYPE_NONE"
    | "LLAMA_ROPE_SCALING_TYPE_LINEAR"
    | "LLAMA_ROPE_SCALING_TYPE_YARN"
seed?: number
skip_chat_parsing?: boolean
spec_draft_model?: string
spec_draft_n_max?: number
spec_draft_n_min?: number
spec_draft_ngl?: number
spec_draft_p_min?: number
spec_draft_threads?: number
spec_draft_threads_batch?: number
swa_full?: boolean
warmup?: boolean
yarn_attn_factor?: number
yarn_beta_fast?: number
yarn_beta_slow?: number
yarn_ext_factor?: number
yarn_orig_ctx?: number