Merge branch 'main' into cc_perf_fixes

Author: dhansen-nvidia

.github/workflows/label_issue.yml

@@ -15,10 +15,9 @@ jobs:
       - name: Checkout private action repository
         uses: actions/checkout@v4
         with:
-          repository: poweiw/goggles_action
+          repository: NVIDIA/goggles_action
           path: ./.github/actions/goggles_action # local path to store the action
-          token: ${{ secrets.GOGGLES_ACTION_REPO_TOKEN}} # token to access poweiw/goggles_action
-          ref: v1.2.1
+          ref: v1.3.0
 
       - name: AI Label Issue
         uses: ./.github/actions/goggles_action/actions/llm_label

.gitmodules

@@ -1,35 +0,0 @@
-[submodule "3rdparty/cutlass"]
-	path = 3rdparty/cutlass
-	url = https://github.com/NVIDIA/cutlass.git
-[submodule "3rdparty/json"]
-	path = 3rdparty/json
-	url = https://github.com/nlohmann/json.git
-[submodule "3rdparty/cxxopts"]
-	path = 3rdparty/cxxopts
-	url = https://github.com/jarro2783/cxxopts
-	branch = v3.1.1
-[submodule "3rdparty/NVTX"]
-	path = 3rdparty/NVTX
-	url = https://github.com/NVIDIA/NVTX.git
-[submodule "3rdparty/ucxx"]
-	path = 3rdparty/ucxx
-	url = https://github.com/rapidsai/ucxx.git
-[submodule "3rdparty/pybind11"]
-	path = 3rdparty/pybind11
-	url = https://github.com/pybind/pybind11.git
-[submodule "3rdparty/xgrammar"]
-	path = 3rdparty/xgrammar
-	url = https://github.com/mlc-ai/xgrammar.git
-[submodule "3rdparty/nanobind"]
-	path = 3rdparty/nanobind
-	url = https://github.com/wjakob/nanobind
-[submodule "3rdparty/cppzmq"]
-	path = 3rdparty/cppzmq
-	url = https://github.com/zeromq/cppzmq.git
-[submodule "3rdparty/DeepGEMM"]
-	path = 3rdparty/DeepGEMM
-	url = https://github.com/ruoqianguo/DeepGEMM.git
-	branch = swapab_sm100
-[submodule "3rdparty/flash-mla"]
-	path = 3rdparty/flash-mla
-	url = https://github.com/deepseek-ai/FlashMLA.git

.pre-commit-config.yaml

@@ -1059,7 +1059,6 @@ common-files: &common_files |
         tests/unittest/_torch/thop/parallel/test_logits_bitmask_op.py |
         tests/unittest/_torch/thop/parallel/test_mamba_conv1d_op.py |
         tests/unittest/_torch/thop/parallel/test_mamba2_chunk_ss_update.py |
-        tests/unittest/_torch/thop/parallel/test_moe.py |
         tests/unittest/_torch/thop/parallel/test_noaux_tc.py |
         tests/unittest/_torch/thop/parallel/test_scaled_mm.py |
         tests/unittest/_torch/thop/parallel/test_selective_scan_op.py |
@@ -1071,6 +1070,7 @@ common-files: &common_files |
         tests/unittest/_torch/thop/parallel/test_weight_only_quant_gemm.py |
         tests/unittest/_torch/thop/parallel/test_weight_only_quant_linear.py |
         tests/unittest/_torch/thop/serial/test_moe_alltoall.py |
+        tests/unittest/_torch/thop/serial/test_moe.py |
         tests/unittest/api_stability/api_stability_core.py |
         tests/unittest/api_stability/test_llm_api.py |
         tests/unittest/bindings/binding_test_utils.py |
@@ -1460,6 +1460,11 @@ repos:
         entry: ./scripts/format_test_list.py
         language: script
         files: tests/integration/test_lists/.*\.txt$
+    -   id: waive list check
+        name: Checks for duplicated test items in waives.txt
+        entry: ./scripts/check_test_list.py --check-duplicate-waives
+        language: script
+        pass_filenames: false
     -   id: DCO check
         name: Checks the commit message for a developer certificate of origin signature
         entry: ./scripts/dco_check.py

3rdparty/CMakeLists.txt

@@ -0,0 +1,114 @@
+include(ExternalProject)
+include(FetchContent)
+
+if(DEFINED ENV{GITHUB_MIRROR} AND NOT "$ENV{GITHUB_MIRROR}" STREQUAL "")
+  set(github_base_url "$ENV{GITHUB_MIRROR}")
+else()
+  set(github_base_url "https://github.com")
+endif()
+
+FetchContent_Declare(
+  cppzmq
+  GIT_REPOSITORY https://github.com/zeromq/cppzmq
+  GIT_TAG v4.10.0 # c94c20743ed7d4aa37835a5c46567ab0790d4acc
+  GIT_SHALLOW TRUE
+  # NOTE: TensorRT-LLM only uses the headers
+  SOURCE_SUBDIR
+  dont-add-this-project-with-add-subdirectory)
+
+FetchContent_Declare(
+  cutlass
+  GIT_REPOSITORY https://github.com/NVIDIA/cutlass
+  GIT_TAG v4.2.1 # f3fde58372d33e9a5650ba7b80fc48b3b49d40c8
+  GIT_SHALLOW TRUE
+  SOURCE_SUBDIR
+  dont-add-this-project-with-add-subdirectory)
+
+FetchContent_Declare(
+  cxxopts
+  GIT_REPOSITORY https://github.com/jarro2783/cxxopts
+  GIT_TAG v3.1.1 # eb787304d67ec22f7c3a184ee8b4c481d04357fd
+  GIT_SHALLOW TRUE)
+
+set(deep_ep_commit 5be51b228a7c82dbdb213ea58e77bffd12b38af8)
+set_property(GLOBAL PROPERTY DEEP_EP_COMMIT "${deep_ep_commit}")
+FetchContent_Declare(
+  deep_ep_download
+  URL ${github_base_url}/deepseek-ai/DeepEP/archive/${deep_ep_commit}.tar.gz)
+
+FetchContent_Declare(
+  deepgemm
+  GIT_REPOSITORY https://github.com/ruoqianguo/DeepGEMM
+  GIT_TAG 9fa5965e265e27995f539e0dd73a06351a8a9eaf
+  SOURCE_SUBDIR
+  dont-add-this-project-with-add-subdirectory)
+
+FetchContent_Declare(
+  eigen
+  GIT_REPOSITORY https://github.com/libeigen/eigen
+  GIT_TAG 3.4.0
+  GIT_SHALLOW TRUE)
+
+FetchContent_Declare(
+  flashmla
+  GIT_REPOSITORY https://github.com/deepseek-ai/FlashMLA.git
+  GIT_TAG 1408756a88e52a25196b759eaf8db89d2b51b5a1
+  SOURCE_SUBDIR
+  dont-add-this-project-with-add-subdirectory)
+
+FetchContent_Declare(
+  googlebenchmark
+  GIT_REPOSITORY https://github.com/google/benchmark
+  GIT_TAG v1.8.3
+  GIT_SHALLOW TRUE)
+
+FetchContent_Declare(
+  googletest
+  GIT_REPOSITORY https://github.com/google/googletest
+  GIT_TAG v1.15.2
+  GIT_SHALLOW TRUE)
+
+FetchContent_Declare(
+  json
+  GIT_REPOSITORY https://github.com/nlohmann/json
+  GIT_TAG v3.12.0 # 55f93686c01528224f448c19128836e7df245f72
+  GIT_SHALLOW TRUE
+  SOURCE_SUBDIR
+  dont-add-this-project-with-add-subdirectory)
+
+FetchContent_Declare(
+  nanobind
+  GIT_REPOSITORY https://github.com/wjakob/nanobind
+  GIT_TAG a0ed2587f1089ef7657e2ed49ad6756b01c74e9f)
+
+FetchContent_Declare(
+  nvtx
+  GIT_REPOSITORY https://github.com/NVIDIA/NVTX
+  GIT_TAG v3.1.0-c-cpp # a1ceb0677f67371ed29a2b1c022794f077db5fe7
+  GIT_SHALLOW TRUE
+  # NOTE: TensorRT-LLM only uses the headers
+  SOURCE_SUBDIR
+  dont-add-this-project-with-add-subdirectory)
+
+FetchContent_Declare(
+  pybind11
+  GIT_REPOSITORY https://github.com/pybind/pybind11
+  GIT_TAG f99ffd7e03001810a3e722bf48ad1a9e08415d7d)
+
+FetchContent_Declare(
+  ucxx
+  GIT_REPOSITORY https://github.com/rapidsai/ucxx
+  GIT_TAG 16eaa57c8d98c8ef54d666a2d2b11e76cfa565f5
+  # NOTE: See the notes in cpp/CMakeList.txt where this project is build at
+  # configure time and then included via find_package
+  SOURCE_SUBDIR
+  dont-add-this-project-with-add-subdirectory)
+
+FetchContent_Declare(
+  xgrammar
+  GIT_REPOSITORY https://github.com/mlc-ai/xgrammar
+  GIT_TAG v0.1.25 # e4e816f5f0fe39f5b1601a17a4552307fa3b70ff
+  GIT_SHALLOW TRUE
+  # NOTE: TensorRT-LLM only uses the headers
+  SOURCE_SUBDIR
+  dont-add-this-project-with-add-subdirectory)