diff --git a/.github/workflows/nightly-test-musa.yml b/.github/workflows/nightly-test-musa.yml index c4e43c377..eafb7f23b 100644 --- a/.github/workflows/nightly-test-musa.yml +++ b/.github/workflows/nightly-test-musa.yml @@ -74,7 +74,6 @@ jobs: if: steps.gate.outputs.run_job == 'true' timeout-minutes: 30 run: | - pytest sgl-kernel/tests/test_dsv3_router_gemm.py pytest sgl-kernel/tests/test_per_token_quant_fp8.py pytest sgl-kernel/tests/speculative/test_eagle_utils.py pytest sgl-kernel/tests/speculative/test_ngram_utils.py diff --git a/.github/workflows/pr-test-musa.yml b/.github/workflows/pr-test-musa.yml index 07fb54ddd..ec55ffa32 100644 --- a/.github/workflows/pr-test-musa.yml +++ b/.github/workflows/pr-test-musa.yml @@ -217,7 +217,6 @@ jobs: - name: Run sgl-kernel test timeout-minutes: 20 run: | - pytest sgl-kernel/tests/test_dsv3_router_gemm.py pytest sgl-kernel/tests/test_per_token_quant_fp8.py pytest sgl-kernel/tests/speculative/test_eagle_utils.py pytest sgl-kernel/tests/speculative/test_ngram_utils.py diff --git a/3rdparty/amd/wheel/sglang/pyproject.toml b/3rdparty/amd/wheel/sglang/pyproject.toml index 8ae54899c..9dc3dd71d 100644 --- a/3rdparty/amd/wheel/sglang/pyproject.toml +++ b/3rdparty/amd/wheel/sglang/pyproject.toml @@ -124,7 +124,7 @@ srt_musa = [ "sglang[runtime_common]", "torch", "torch_musa", - "torchada>=0.1.68", + "torchada>=0.1.74", "mthreads-ml-py", "mate>=0.2.0", "deep-gemm>=0.1.3", diff --git a/python/pyproject_other.toml b/python/pyproject_other.toml index 34f2179ce..1fd75b814 100755 --- a/python/pyproject_other.toml +++ b/python/pyproject_other.toml @@ -126,7 +126,7 @@ srt_musa = [ "sglang[runtime_common]", "torch", "torch_musa", - "torchada>=0.1.68", + "torchada>=0.1.74", ] diffusion_musa = [ diff --git a/sgl-kernel/csrc/common_extension_musa.cc b/sgl-kernel/csrc/common_extension_musa.cc index 1b17ad3ab..52b139def 100644 --- a/sgl-kernel/csrc/common_extension_musa.cc +++ b/sgl-kernel/csrc/common_extension_musa.cc @@ -100,9 +100,6 @@ TORCH_LIBRARY_EXPAND(sgl_kernel, m) { m.def("dsv3_fused_a_gemm(Tensor! output, Tensor mat_a, Tensor mat_b) -> ()"); m.impl("dsv3_fused_a_gemm", torch::kMUSA, &dsv3_fused_a_gemm); - m.def("dsv3_router_gemm(Tensor! output, Tensor mat_a, Tensor mat_b) -> ()"); - m.impl("dsv3_router_gemm", torch::kMUSA, &dsv3_router_gemm); - /* * From csrc/moe */ diff --git a/sgl-kernel/csrc/musa/top_k_top_p_sampling.mu b/sgl-kernel/csrc/musa/top_k_top_p_sampling.mu index 4e149086f..feb927c60 100644 --- a/sgl-kernel/csrc/musa/top_k_top_p_sampling.mu +++ b/sgl-kernel/csrc/musa/top_k_top_p_sampling.mu @@ -21,7 +21,7 @@ #include #include "torch_musa/csrc/aten/musa/UnpackRaw.muh" -#include +#include #include #include "musa.h" diff --git a/sgl-kernel/pyproject_musa.toml b/sgl-kernel/pyproject_musa.toml index a8d3b8cda..d7f4538ad 100644 --- a/sgl-kernel/pyproject_musa.toml +++ b/sgl-kernel/pyproject_musa.toml @@ -3,7 +3,7 @@ requires = [ "setuptools>=75.0", "scikit-build-core>=0.10", "torch", - "torchada>=0.1.68", + "torchada>=0.1.74", "wheel", ] build-backend = "setuptools.build_meta"