ohms-adaptq 0.9.8

Revolutionary Super-APQ: Zero-Cost Universal LLM Quantization Engine
Documentation 
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

pub mod quantization;
pub mod manifest;
pub mod verification;
pub mod super_apq_min;
pub mod universal_loader;
pub mod quantizer;
pub mod model_fetcher;

pub use quantization::*;
pub use manifest::*;
pub use verification::*;
pub use super_apq_min::{SuperAPQ, SuperAPQConfig, SuperQuantizedModel};
pub use universal_loader::{UniversalModel, UniversalLoader, load_any_model, find_model};
pub use quantizer::{Quantizer, QuantizedModel, QuantizationConfig, QuantizationMethod};
pub use model_fetcher::{ModelFetcher, ModelSource, parse_model_source};

pub type Result<T> = std::result::Result<T, Box<dyn std::error::Error + Send + Sync>>;

#[derive(Debug, Clone)]
pub struct ApqConfig {
    pub target_bits: u8,
    pub calibration_size: usize,
    pub chunk_size: usize,
    pub fidelity_threshold: f32,
    pub seed: Option<u64>,
}

impl Default for ApqConfig {
    fn default() -> Self {
        Self {
            target_bits: 4,
            calibration_size: 512,
            chunk_size: 2 * 1024 * 1024, // 2 MiB
            fidelity_threshold: 0.1,
            seed: Some(42),
        }
    }
}