diff --git a/crates/xserv-kernels/build.rs b/crates/xserv-kernels/build.rs index 819d7c8..a91024a 100644 --- a/crates/xserv-kernels/build.rs +++ b/crates/xserv-kernels/build.rs @@ -27,6 +27,7 @@ fn main() { .file("../../csrc/embedding/transpose.cu") .file("../../csrc/attention/flash_attention.cu") .file("../../csrc/attention/paged_attention.cu") + .file("../../csrc/quant/mxfp4.cu") .compile("xserv_kernels"); println!("cargo:rerun-if-changed=../../csrc/"); diff --git a/crates/xserv-kernels/src/lib.rs b/crates/xserv-kernels/src/lib.rs index 1af95b0..75ca199 100644 --- a/crates/xserv-kernels/src/lib.rs +++ b/crates/xserv-kernels/src/lib.rs @@ -4,6 +4,7 @@ pub mod dispatch; pub mod embedding; pub mod gemm; pub mod layernorm; +pub mod quant; pub mod rmsnorm; pub mod rope; pub mod softmax;