Gahow Wang
7821bd9c34
Add the batched-forward primitives. Linears/norms/elementwise/embedding/CE already act on flat [rows,dim], so they work unchanged on [B*S,dim]; only attention + RoPE need sequence awareness: - RoPE: kernel takes a `period` (= seq len) so position = row % period, i.e. per-sequence position on a flattened batch (period == tokens = single seq). - Fused batched causal attention: new `Tensor::attention`/`attention_backward` + ops node, running QKᵀ and PV as cublasSgemmStridedBatched over the B*nh (sequence,head) blocks (new sgemm_strided_batched binding) and a causal softmax kernel (scale + per-row causal mask inline) — the whole attention is 3 launches regardless of B*nh, no per-head/per-seq loop, no host round-trip. - transpose_4d12 ([B,S,nh,hd] <-> [B,nh,S,hd]) to lay out the batched heads. grad-checks: new batched-rope, transpose_4d12, batched-attention dQ/dK/dV all pass finite-diff (attn dK 1.5e-2, dQ 7.5e-3, dV 2.9e-4; rest tighter) alongside the existing 12. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
48 lines
1.7 KiB
Rust
48 lines
1.7 KiB
Rust
use std::env;
|
|
use std::path::Path;
|
|
use std::process::Command;
|
|
|
|
fn main() {
|
|
println!("cargo:rustc-check-cfg=cfg(no_cuda)");
|
|
println!("cargo:rerun-if-changed=../../csrc/");
|
|
|
|
let cuda_path = env::var("CUDA_HOME")
|
|
.or_else(|_| env::var("CUDA_PATH"))
|
|
.unwrap_or_else(|_| "/usr/local/cuda".to_string());
|
|
|
|
// Locally there is no nvcc / GPU. Detect that and skip the CUDA build so
|
|
// `cargo check`/`cargo build` of host-side Rust still works. The `no_cuda`
|
|
// cfg makes the FFI bindings + smoke test compile (but not run) without nvcc.
|
|
if !nvcc_available(&cuda_path) {
|
|
println!("cargo:warning=nvcc not found — skipping CUDA compilation (host-only build).");
|
|
println!("cargo:rustc-cfg=no_cuda");
|
|
return;
|
|
}
|
|
|
|
println!("cargo:rustc-link-search=native={cuda_path}/lib64");
|
|
println!("cargo:rustc-link-lib=dylib=cudart");
|
|
println!("cargo:rustc-link-lib=dylib=cuda");
|
|
// cuBLAS is used only as a correctness reference for the hand-written GEMM.
|
|
println!("cargo:rustc-link-lib=dylib=cublas");
|
|
|
|
cc::Build::new()
|
|
.cuda(true)
|
|
.cudart("shared")
|
|
.flag("-gencode=arch=compute_120,code=sm_120")
|
|
.file("../../csrc/test/vecadd.cu")
|
|
.file("../../csrc/ops/elementwise.cu")
|
|
.file("../../csrc/ops/gemm.cu")
|
|
.file("../../csrc/ops/nn.cu")
|
|
.file("../../csrc/ops/model.cu")
|
|
.file("../../csrc/ops/optim.cu")
|
|
.file("../../csrc/ops/attention.cu")
|
|
.compile("xtrain_cuda_kernels");
|
|
}
|
|
|
|
fn nvcc_available(cuda_path: &str) -> bool {
|
|
if Command::new("nvcc").arg("--version").output().is_ok() {
|
|
return true;
|
|
}
|
|
Path::new(&format!("{cuda_path}/bin/nvcc")).exists()
|
|
}
|