From bcda0c9ee6a6e687e53ac933f3541dd5c5a1fe9b Mon Sep 17 00:00:00 2001 From: Yineng Zhang Date: Wed, 22 Jan 2025 20:33:13 +0800 Subject: [PATCH] sync the upstream updates of flashinfer (#3051) --- .github/workflows/pr-test-sgl-kernel.yml | 1 + sgl-kernel/3rdparty/flashinfer | 2 +- sgl-kernel/setup.py | 6 ++++++ 3 files changed, 8 insertions(+), 1 deletion(-) diff --git a/.github/workflows/pr-test-sgl-kernel.yml b/.github/workflows/pr-test-sgl-kernel.yml index 3d9802658..794a73f36 100644 --- a/.github/workflows/pr-test-sgl-kernel.yml +++ b/.github/workflows/pr-test-sgl-kernel.yml @@ -40,6 +40,7 @@ jobs: run: | pip3 install torch==2.5.1 && pip3 install pytest && pip3 install vllm pip3 uninstall sgl-kernel -y || true + find . -name index.lock -delete cd sgl-kernel git submodule deinit --all --force && git submodule sync --recursive && git submodule update --init --force --recursive pip3 install . diff --git a/sgl-kernel/3rdparty/flashinfer b/sgl-kernel/3rdparty/flashinfer index a0e99a3a8..4e8eb1879 160000 --- a/sgl-kernel/3rdparty/flashinfer +++ b/sgl-kernel/3rdparty/flashinfer @@ -1 +1 @@ -Subproject commit a0e99a3a820109763d9a757138a5cdf7bbcd1f85 +Subproject commit 4e8eb1879f9c3ba6d75511e5893183bf8f289a62 diff --git a/sgl-kernel/setup.py b/sgl-kernel/setup.py index 1197611d6..b9324c355 100644 --- a/sgl-kernel/setup.py +++ b/sgl-kernel/setup.py @@ -47,6 +47,7 @@ include_dirs = [ cutlass.resolve() / "tools" / "util" / "include", root / "src" / "sgl-kernel" / "csrc", flashinfer.resolve() / "include", + flashinfer.resolve() / "include" / "gemm", flashinfer.resolve() / "csrc", ] nvcc_flags = [ @@ -91,7 +92,12 @@ ext_modules = [ "src/sgl-kernel/csrc/sampling_scaling_penalties.cu", "src/sgl-kernel/csrc/sgl_kernel_ops.cu", "src/sgl-kernel/csrc/rotary_embedding.cu", + "3rdparty/flashinfer/csrc/activation.cu", + "3rdparty/flashinfer/csrc/bmm_fp8.cu", + "3rdparty/flashinfer/csrc/group_gemm.cu", + "3rdparty/flashinfer/csrc/group_gemm_sm90.cu", "3rdparty/flashinfer/csrc/norm.cu", + "3rdparty/flashinfer/csrc/sampling.cu", ], include_dirs=include_dirs, extra_compile_args={