boostr 0.1.0

ML framework built on numr - attention, quantization, model architectures
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22

//! CPU dequantization kernels
//!
//! Re-exports from format-specific modules:
//! - `dequant_simple` — Q4_0, Q4_1, Q5_0, Q5_1, Q8_0, Q8_1
//! - `dequant_k_quants` — Q2_K, Q3_K, Q4_K, Q5_K, Q6_K, Q8_K
//! - `dequant_iq1` — IQ1_S, IQ1_M
//! - `dequant_iq2` — IQ2_XXS, IQ2_XS, IQ2_S
//! - `dequant_iq3` — IQ3_XXS, IQ3_S
//! - `dequant_iq4` — IQ4_NL, IQ4_XS
//! - `dequant_tq` — TQ1_0, TQ2_0

pub use super::dequant_iq1::{dequant_iq1_m, dequant_iq1_s};
pub use super::dequant_iq2::{dequant_iq2_s, dequant_iq2_xs, dequant_iq2_xxs};
pub use super::dequant_iq3::{dequant_iq3_s, dequant_iq3_xxs};
pub use super::dequant_iq4::{dequant_iq4_nl, dequant_iq4_xs};
pub use super::dequant_k_quants::{
    dequant_q2k, dequant_q3k, dequant_q4k, dequant_q5k, dequant_q6k, dequant_q8k,
};
pub use super::dequant_simple::{
    dequant_q4_0, dequant_q4_1, dequant_q5_0, dequant_q5_1, dequant_q8_0, dequant_q8_1,
};
pub use super::dequant_tq::{dequant_tq1_0, dequant_tq2_0};