tract-cuda 0.23.4

Tiny, no-nonsense, self contained, TensorFlow and ONNX inference
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
mod conv;
mod flash_attn;
mod fused_axis_op;
mod gemm;
mod iff;
mod quant_q81;
pub use conv::{CudaConv, wire_cuda_conv};
pub use flash_attn::CudaFlashAttention;
pub use fused_axis_op::CudaFusedAxisOp;
pub use gemm::CudaGgmlGemm;
pub use iff::CudaIff;
pub use quant_q81::{CudaGgmlQuantQ81, GgmlQuantQ81Fact};