ri-esp-llm 0.1.1

no_std embedded inference primitives: int4 packing, KV cache, RNN helpers, sampling, compressed attention
Documentation
1
2
3
4
5
6
7
8
9
10
11
#![no_std]

pub mod compressed_attention;
pub mod int4;
pub mod kv_cache;
pub mod rnn;
pub mod sample;

pub use compressed_attention::CompressedAttentionCache;
pub use int4::{pack_i4_pair, quantize_to_i4, unpack_i4};
pub use sample::{argmax, Lcg};