Struct LlamaParams 

pub struct LlamaParams {
    pub ctx_size: Option<u32>,
    pub n_gpu_layers: Option<i32>,
    pub threads: Option<u32>,
    pub batch_size: Option<u32>,
    pub parallel: Option<u32>,
    pub seed: Option<i64>,
    pub flash_attn: Option<bool>,
    pub mlock: Option<bool>,
    pub mmap: Option<bool>,
    pub cache_type_k: Option<String>,
    pub cache_type_v: Option<String>,
    pub enable_thinking: Option<bool>,
}

Per-model llama-server load parameters (the “Load” tab, à la LM Studio). Absent fields fall back to the server’s own defaults (no flag passed).

Fields

ctx_size: Option<u32>

Context window size (--ctx-size).

n_gpu_layers: Option<i32>

Layers offloaded to the GPU (--n-gpu-layers); 0 = CPU only.

threads: Option<u32>

CPU threads (--threads).

batch_size: Option<u32>

Logical batch size for prompt eval (--batch-size).

parallel: Option<u32>

Parallel sequences / max concurrent predictions (--parallel).

seed: Option<i64>

RNG seed (--seed); omit for random.

flash_attn: Option<bool>

Enable Flash Attention (--flash-attn).

mlock: Option<bool>

Lock the model in RAM (--mlock) - “keep model in memory”.

mmap: Option<bool>

Memory-map the model file (--mmap); false passes --no-mmap.

cache_type_k: Option<String>

KV cache K type, e.g. f16, q8_0, q4_0 (--cache-type-k).

cache_type_v: Option<String>

KV cache V type (--cache-type-v).

enable_thinking: Option<bool>

For reasoning models: when Some(false), disable “thinking” by passing --reasoning-budget 0 (much faster, lower memory). Only meaningful for models that support it; None/Some(true) leaves the default on.

Trait Implementations

impl Clone for LlamaParams

fn clone(&self) -> LlamaParams

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for LlamaParams

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for LlamaParams

fn default() -> LlamaParams

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for LlamaParams

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for LlamaParams

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.