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
23
24
25
26

pub mod audio;
pub mod config;
pub mod detection;
pub mod encoder;
pub mod hybrid;
pub mod llama;
pub mod mamba;
pub mod multimodal;
pub mod registry;
pub mod registry_inference;
pub mod traits;
pub mod vision;

pub use config::{
    AttentionConfig, AudioConfig, HuggingFaceConfig, HybridConfig, ModelConfig, MoeConfig,
    RopeScalingConfig, SsmConfig, UniversalConfig, VisionConfig, load_config_auto,
    load_huggingface_config,
};
pub use encoder::{EmbeddingPipeline, Encoder, EncoderClient, EncoderConfig, Pooling};
pub use hybrid::HybridModel;
pub use llama::model::blocks::ExpertWeights;
pub use llama::{Llama, LlamaTp};
pub use mamba::{Mamba2, Mamba2Config, Mamba2Model, Mamba2Weights};
pub use multimodal::{ModelInput, MultimodalModel};
pub use registry::LoadedModel;
pub use traits::{Model, ModelClient};