Crate oxbitnet 

oxbitnet

Run BitNet b1.58 ternary LLMs with wgpu.

use oxbitnet::BitNet;
use futures::StreamExt;

let mut bitnet = BitNet::load("model.gguf", Default::default()).await?;

{
    let mut stream = bitnet.generate("Hello!", Default::default());
    while let Some(token) = stream.next().await {
        print!("{token}");
    }
}

bitnet.dispose();

Re-exports

pub use error::BitNetError;

pub use error::Result;

pub use model::config::Activation;

pub use model::config::ModelConfig;

pub use model::loader::LoadOptions;

pub use model::loader::LoadProgress;

pub use tokenizer::ChatMessage;

pub use tokenizer::Tokenizer;

Modules

error

gpu

model

nn

sampling

tokenizer

Structs

BitNet

High-level API for BitNet inference on wgpu.

GenerateOptions

Generation options.