Crate ri_esp_llm 

Re-exports

pub use compressed_attention::CompressedAttentionCache;

pub use int4::pack_i4_pair;

pub use int4::quantize_to_i4;

pub use int4::unpack_i4;

pub use sample::argmax;

pub use sample::Lcg;

Modules

compressed_attention

Compressed attention cache for ESP32 using per-dimension quantized scoring.

int4

kv_cache

rnn

sample