llama_cpp_sys_v3/
types.rs

1#![allow(non_camel_case_types)]
2
3// No imports needed
4
5pub type llama_pos = i32;
6pub type llama_token = i32;
7pub type llama_seq_id = i32;
8
9pub const LLAMA_TOKEN_NULL: llama_token = -1;
10
11// Opaque types
12#[repr(C)]
13pub struct llama_vocab {
14    _data: [u8; 0],
15    _marker: core::marker::PhantomData<(*mut u8, core::marker::PhantomPinned)>,
16}
17
18#[repr(C)]
19pub struct llama_model {
20    _data: [u8; 0],
21    _marker: core::marker::PhantomData<(*mut u8, core::marker::PhantomPinned)>,
22}
23
24#[repr(C)]
25pub struct llama_context {
26    _data: [u8; 0],
27    _marker: core::marker::PhantomData<(*mut u8, core::marker::PhantomPinned)>,
28}
29
30#[repr(C)]
31pub struct llama_sampler {
32    _data: [u8; 0],
33    _marker: core::marker::PhantomData<(*mut u8, core::marker::PhantomPinned)>,
34}
35
36#[repr(C)]
37#[derive(Debug, Copy, Clone)]
38pub struct llama_chat_message {
39    pub role: *const std::ffi::c_char,
40    pub content: *const std::ffi::c_char,
41}
42
43#[repr(C)]
44#[derive(Debug, Copy, Clone)]
45pub struct llama_batch {
46    pub n_tokens: i32,
47    pub token: *mut llama_token,
48    pub embd: *mut f32,
49    pub pos: *mut llama_pos,
50    pub n_seq_id: *mut i32,
51    pub seq_id: *mut *mut llama_seq_id,
52    pub logits: *mut i8, // bool in C
53}
54
55#[repr(C)]
56#[derive(Debug, Copy, Clone)]
57pub struct llama_model_params {
58    pub devices: *mut std::ffi::c_void,
59    pub tensor_buft_overrides: *const std::ffi::c_void,
60    pub n_gpu_layers: i32,
61    pub split_mode: i32,
62    pub main_gpu: i32,
63    pub tensor_split: *const f32,
64    pub progress_callback: Option<unsafe extern "C" fn(f32, *mut std::ffi::c_void) -> bool>,
65    pub progress_callback_user_data: *mut std::ffi::c_void,
66    pub kv_overrides: *const std::ffi::c_void,
67    pub vocab_only: bool,
68    pub use_mmap: bool,
69    pub use_direct_io: bool,
70    pub use_mlock: bool,
71    pub check_tensors: bool,
72    pub use_extra_bufts: bool,
73    pub no_host: bool,
74    pub no_alloc: bool,
75}
76
77#[repr(C)]
78#[derive(Debug, Copy, Clone)]
79pub struct llama_context_params {
80    pub n_ctx: u32,
81    pub n_batch: u32,
82    pub n_ubatch: u32,
83    pub n_seq_max: u32,
84    pub n_threads: i32,
85    pub n_threads_batch: i32,
86    pub rope_scaling_type: i32,
87    pub pooling_type: i32,
88    pub attention_type: i32,
89    pub flash_attn_type: i32,
90    pub rope_freq_base: f32,
91    pub rope_freq_scale: f32,
92    pub yarn_ext_factor: f32,
93    pub yarn_attn_factor: f32,
94    pub yarn_beta_fast: f32,
95    pub yarn_beta_slow: f32,
96    pub yarn_orig_ctx: u32,
97    pub defrag_thold: f32,
98    pub cb_eval: Option<unsafe extern "C" fn(*mut std::ffi::c_void, *mut std::ffi::c_void) -> bool>,
99    pub cb_eval_user_data: *mut std::ffi::c_void,
100    pub type_k: i32,
101    pub type_v: i32,
102    pub abort_callback: Option<unsafe extern "C" fn(*mut std::ffi::c_void) -> bool>,
103    pub abort_callback_data: *mut std::ffi::c_void,
104    pub embeddings: bool,
105    pub offload_kqv: bool,
106    pub no_perf: bool,
107    pub op_offload: bool,
108    pub swa_full: bool,
109    pub kv_unified: bool,
110    pub samplers: *mut std::ffi::c_void,
111    pub n_samplers: usize,
112}
113
114#[repr(C)]
115#[derive(Debug, Copy, Clone)]
116pub struct llama_sampler_chain_params {
117    pub no_perf: bool,
118}
llama_cpp_sys_v3/types.rs

llama_cpp_sys_v3/
types.rs
