1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357
//! A safe wrapper around `llama_context_params`.
use std::fmt::Debug;
use std::num::NonZeroU32;
/// A rusty wrapper around `rope_scaling_type`.
#[repr(i8)]
#[derive(Copy, Clone, Debug, PartialEq, Eq)]
pub enum RopeScalingType {
/// The scaling type is unspecified
Unspecified = -1,
/// No scaling
None = 0,
/// Linear scaling
Linear = 1,
/// Yarn scaling
Yarn = 2,
}
/// Create a `RopeScalingType` from a `c_int` - returns `RopeScalingType::ScalingUnspecified` if
/// the value is not recognized.
impl From<i32> for RopeScalingType {
fn from(value: i32) -> Self {
match value {
0 => Self::None,
1 => Self::Linear,
2 => Self::Yarn,
_ => Self::Unspecified,
}
}
}
/// Create a `c_int` from a `RopeScalingType`.
impl From<RopeScalingType> for i32 {
fn from(value: RopeScalingType) -> Self {
match value {
RopeScalingType::None => 0,
RopeScalingType::Linear => 1,
RopeScalingType::Yarn => 2,
RopeScalingType::Unspecified => -1,
}
}
}
/// A safe wrapper around `llama_context_params`.
///
/// Generally this should be created with [`Default::default()`] and then modified with `with_*` methods.
///
/// # Examples
///
/// ```rust
/// # use std::num::NonZeroU32;
/// use llama_cpp_2::context::params::LlamaContextParams;
///
///let ctx_params = LlamaContextParams::default()
/// .with_n_ctx(NonZeroU32::new(2048))
/// .with_seed(1234);
///
/// assert_eq!(ctx_params.seed(), 1234);
/// assert_eq!(ctx_params.n_ctx(), NonZeroU32::new(2048));
/// ```
#[derive(Debug, Clone)]
#[allow(
missing_docs,
clippy::struct_excessive_bools,
clippy::module_name_repetitions
)]
pub struct LlamaContextParams {
pub(crate) context_params: llama_cpp_sys_2::llama_context_params,
}
/// SAFETY: we do not currently allow setting or reading the pointers that cause this to not be automatically send or sync.
unsafe impl Send for LlamaContextParams {}
unsafe impl Sync for LlamaContextParams {}
impl LlamaContextParams {
/// Set the seed of the context
///
/// # Examples
///
/// ```rust
/// use llama_cpp_2::context::params::LlamaContextParams;
/// let params = LlamaContextParams::default();
/// let params = params.with_seed(1234);
/// assert_eq!(params.seed(), 1234);
/// ```
#[must_use]
pub fn with_seed(mut self, seed: u32) -> Self {
self.context_params.seed = seed;
self
}
/// Get the seed of the context
///
/// # Examples
///
/// ```rust
/// use llama_cpp_2::context::params::LlamaContextParams;
/// let params = LlamaContextParams::default()
/// .with_seed(1234);
/// assert_eq!(params.seed(), 1234);
/// ```
#[must_use]
pub fn seed(&self) -> u32 {
self.context_params.seed
}
/// Set the side of the context
///
/// # Examples
///
/// ```rust
/// # use std::num::NonZeroU32;
/// use llama_cpp_2::context::params::LlamaContextParams;
/// let params = LlamaContextParams::default();
/// let params = params.with_n_ctx(NonZeroU32::new(2048));
/// assert_eq!(params.n_ctx(), NonZeroU32::new(2048));
/// ```
#[must_use]
pub fn with_n_ctx(mut self, n_ctx: Option<NonZeroU32>) -> Self {
self.context_params.n_ctx = n_ctx.map_or(0, std::num::NonZeroU32::get);
self
}
/// Get the size of the context.
///
/// [`None`] if the context size is specified by the model and not the context.
///
/// # Examples
///
/// ```rust
/// let params = llama_cpp_2::context::params::LlamaContextParams::default();
/// assert_eq!(params.n_ctx(), std::num::NonZeroU32::new(512));
#[must_use]
pub fn n_ctx(&self) -> Option<NonZeroU32> {
NonZeroU32::new(self.context_params.n_ctx)
}
/// Set the `n_batch`
///
/// # Examples
///
/// ```rust
/// # use std::num::NonZeroU32;
/// use llama_cpp_2::context::params::LlamaContextParams;
/// let params = LlamaContextParams::default()
/// .with_n_batch(2048);
/// assert_eq!(params.n_batch(), 2048);
/// ```
#[must_use]
pub fn with_n_batch(mut self, n_batch: u32) -> Self {
self.context_params.n_batch = n_batch;
self
}
/// Get the `n_batch`
///
/// # Examples
///
/// ```rust
/// use llama_cpp_2::context::params::LlamaContextParams;
/// let params = LlamaContextParams::default();
/// assert_eq!(params.n_batch(), 2048);
/// ```
#[must_use]
pub fn n_batch(&self) -> u32 {
self.context_params.n_batch
}
/// Set the type of rope scaling.
///
/// # Examples
///
/// ```rust
/// use llama_cpp_2::context::params::{LlamaContextParams, RopeScalingType};
/// let params = LlamaContextParams::default()
/// .with_rope_scaling_type(RopeScalingType::Linear);
/// assert_eq!(params.rope_scaling_type(), RopeScalingType::Linear);
/// ```
#[must_use]
pub fn with_rope_scaling_type(mut self, rope_scaling_type: RopeScalingType) -> Self {
self.context_params.rope_scaling_type = i32::from(rope_scaling_type);
self
}
/// Get the type of rope scaling.
///
/// # Examples
///
/// ```rust
/// let params = llama_cpp_2::context::params::LlamaContextParams::default();
/// assert_eq!(params.rope_scaling_type(), llama_cpp_2::context::params::RopeScalingType::Unspecified);
/// ```
#[must_use]
pub fn rope_scaling_type(&self) -> RopeScalingType {
RopeScalingType::from(self.context_params.rope_scaling_type)
}
/// Set the rope frequency base.
///
/// # Examples
///
/// ```rust
/// use llama_cpp_2::context::params::LlamaContextParams;
/// let params = LlamaContextParams::default()
/// .with_rope_freq_base(0.5);
/// assert_eq!(params.rope_freq_base(), 0.5);
/// ```
#[must_use]
pub fn with_rope_freq_base(mut self, rope_freq_base: f32) -> Self {
self.context_params.rope_freq_base = rope_freq_base;
self
}
/// Get the rope frequency base.
///
/// # Examples
///
/// ```rust
/// let params = llama_cpp_2::context::params::LlamaContextParams::default();
/// assert_eq!(params.rope_freq_base(), 0.0);
/// ```
#[must_use]
pub fn rope_freq_base(&self) -> f32 {
self.context_params.rope_freq_base
}
/// Set the rope frequency scale.
///
/// # Examples
///
/// ```rust
/// use llama_cpp_2::context::params::LlamaContextParams;
/// let params = LlamaContextParams::default()
/// .with_rope_freq_scale(0.5);
/// assert_eq!(params.rope_freq_scale(), 0.5);
/// ```
#[must_use]
pub fn with_rope_freq_scale(mut self, rope_freq_scale: f32) -> Self {
self.context_params.rope_freq_scale = rope_freq_scale;
self
}
/// Get the rope frequency scale.
///
/// # Examples
///
/// ```rust
/// let params = llama_cpp_2::context::params::LlamaContextParams::default();
/// assert_eq!(params.rope_freq_scale(), 0.0);
/// ```
#[must_use]
pub fn rope_freq_scale(&self) -> f32 {
self.context_params.rope_freq_scale
}
/// Get the number of threads.
///
/// # Examples
///
/// ```rust
/// let params = llama_cpp_2::context::params::LlamaContextParams::default();
/// assert_eq!(params.n_threads(), 4);
/// ```
#[must_use]
pub fn n_threads(&self) -> u32 {
self.context_params.n_threads
}
/// Get the number of threads allocated for batches.
///
/// # Examples
///
/// ```rust
/// let params = llama_cpp_2::context::params::LlamaContextParams::default();
/// assert_eq!(params.n_threads_batch(), 4);
/// ```
#[must_use]
pub fn n_threads_batch(&self) -> u32 {
self.context_params.n_threads_batch
}
/// Set the number of threads.
///
/// # Examples
///
/// ```rust
/// use llama_cpp_2::context::params::LlamaContextParams;
/// let params = LlamaContextParams::default()
/// .with_n_threads(8);
/// assert_eq!(params.n_threads(), 8);
/// ```
#[must_use]
pub fn with_n_threads(mut self, n_threads: u32) -> Self {
self.context_params.n_threads = n_threads;
self
}
/// Set the number of threads allocated for batches.
///
/// # Examples
///
/// ```rust
/// use llama_cpp_2::context::params::LlamaContextParams;
/// let params = LlamaContextParams::default()
/// .with_n_threads_batch(8);
/// assert_eq!(params.n_threads_batch(), 8);
/// ```
#[must_use]
pub fn with_n_threads_batch(mut self, n_threads: u32) -> Self {
self.context_params.n_threads_batch = n_threads;
self
}
/// Check whether embeddings are enabled
///
/// # Examples
///
/// ```rust
/// let params = llama_cpp_2::context::params::LlamaContextParams::default();
/// assert!(!params.embeddings());
/// ```
#[must_use]
pub fn embeddings(&self) -> bool {
self.context_params.embeddings
}
/// Enable the use of embeddings
///
/// # Examples
///
/// ```rust
/// use llama_cpp_2::context::params::LlamaContextParams;
/// let params = LlamaContextParams::default()
/// .with_embeddings(true);
/// assert!(params.embeddings());
/// ```
#[must_use]
pub fn with_embeddings(mut self, embedding: bool) -> Self {
self.context_params.embeddings = embedding;
self
}
}
/// Default parameters for `LlamaContext`. (as defined in llama.cpp by `llama_context_default_params`)
/// ```
/// # use std::num::NonZeroU32;
/// use llama_cpp_2::context::params::{LlamaContextParams, RopeScalingType};
/// let params = LlamaContextParams::default();
/// assert_eq!(params.n_ctx(), NonZeroU32::new(512), "n_ctx should be 512");
/// assert_eq!(params.rope_scaling_type(), RopeScalingType::Unspecified);
/// ```
impl Default for LlamaContextParams {
fn default() -> Self {
let context_params = unsafe { llama_cpp_sys_2::llama_context_default_params() };
Self { context_params }
}
}