Merge branch 'master' into feat-general_basic_communication

.github/workflows/canary.yml

@@ -55,7 +55,7 @@ jobs:
       - name: Checkout Oneflow-Inc/oneflow
         if: ${{ github.event.inputs.oneflow-ref == '' }}
         uses: actions/checkout@v2
-      - uses: Oneflow-Inc/get-oneflow@single-matrix-for-efficiency
+      - uses: Oneflow-Inc/get-oneflow@support-iree-ci
         name: Build manylinux
         id: build-cuda
         with:

.github/workflows/on_merge.yml

@@ -15,6 +15,6 @@ jobs:
     if: github.event.pull_request.merged == true
     runs-on: ubuntu-latest
     steps:
-      - uses: Oneflow-Inc/get-oneflow/update-benchmark-history@single-matrix-for-efficiency
+      - uses: Oneflow-Inc/get-oneflow/update-benchmark-history@support-iree-ci
         name: Update benchmark history
         timeout-minutes: 10

.github/workflows/release.yml

@@ -33,7 +33,7 @@ jobs:
         with:
           ref: ${{ github.event.pull_request.head.sha }}
           repository: ${{github.event.pull_request.head.repo.full_name}}
-      - uses: Oneflow-Inc/get-oneflow/cache-complete/matrix/build@single-matrix-for-efficiency
+      - uses: Oneflow-Inc/get-oneflow/cache-complete/matrix/build@support-iree-ci
         name: find cache
         id: find-cache
         timeout-minutes: 5
@@ -45,7 +45,7 @@ jobs:
             release
           oneflow-src: ${{ env.ONEFLOW_SRC }}
           entries: |
-            cu115
+            cu116
             cu112
             cu102
             cpu
@@ -74,7 +74,7 @@ jobs:
           python3 -m pip install -U pip setuptools wheel --user
           python3 -m pip install oss2  --user
       - uses: actions/checkout@v2
-      - uses: Oneflow-Inc/get-oneflow@single-matrix-for-efficiency
+      - uses: Oneflow-Inc/get-oneflow@support-iree-ci
         name: Build ${{ matrix.entry }}
         if: ${{ matrix.entry !='cpu' }}
         with:
@@ -98,7 +98,7 @@ jobs:
             3.8
             3.9
             3.10
-      - uses: Oneflow-Inc/get-oneflow@single-matrix-for-efficiency
+      - uses: Oneflow-Inc/get-oneflow@support-iree-ci
         name: Build ${{ matrix.entry }}
         if: ${{ matrix.entry =='cpu' }}
         with:

.github/workflows/simple.yml

@@ -245,7 +245,7 @@ jobs:
           repository: Oneflow-Inc/conda-env
           ref: 30a7f00eb48ee9009d85a848e720823e5054c66b
           path: conda-env
-      - uses: Oneflow-Inc/get-oneflow@single-matrix-for-efficiency
+      - uses: Oneflow-Inc/get-oneflow@support-iree-ci
         name: Build with gcc7
         if: ${{ matrix.build-type == 'gcc7'}}
         with:
@@ -254,7 +254,7 @@ jobs:
           oneflow-build-env: conda
           conda-env-file: conda-env/dev/gcc7/environment-v2.yml
           conda-env-name: oneflow-dev-gcc7-v2
-      - uses: Oneflow-Inc/get-oneflow@single-matrix-for-efficiency
+      - uses: Oneflow-Inc/get-oneflow@support-iree-ci
         name: Build with clang10
         if: ${{ matrix.build-type == 'clang10'}}
         with:

.github/workflows/test.yml

@@ -16,6 +16,8 @@ env:
   FLOW_VISION_COMMIT: ca8ebc663b58667cf8cd1b6ef0c861522780b7bb
   LIBAI_SRC: libai
   LIBAI_COMMIT: 7d31d9781e5f2d559dc0820f599e0bed798488ca
+  ONEFLOW_IREE_SRC: oneflow_iree
+  ONEFLOW_IREE_COMMIT: 4322cbad2545877b1664aa8e0f17a17f6b5f687c
   TEST_WITH_TORCH_IMG_TAG: registry.cn-beijing.aliyuncs.com/oneflow/test-with-pytorch-1.10.0-cuda11.3-cudnn8-runtime:afaf913e02a4ba02db92260daee22f99121cef62
   MLIR_DOCKER_ARGS: "-e ONEFLOW_MLIR_ENABLE_ROUND_TRIP=1 -e ONEFLOW_MLIR_PREFER_NHWC=0 -e ONEFLOW_MLIR_ENABLE_INFERENCE_OPTIMIZATION=1"
 
@@ -25,7 +27,7 @@ jobs:
     runs-on: ubuntu-latest
     if: github.event.pull_request.draft == false && github.base_ref == 'master' && contains(github.event.pull_request.requested_reviewers.*.login, 'oneflow-ci-bot')
     steps:
-      - uses: Oneflow-Inc/get-oneflow/priority-pr@single-matrix-for-efficiency
+      - uses: Oneflow-Inc/get-oneflow/priority-pr@support-iree-ci
         name: Check priority PR closed
         id: save-cache
         timeout-minutes: 5
@@ -159,7 +161,7 @@ jobs:
           fi
           echo "is_secrets_accessible=1" >> $GITHUB_ENV
       - name: Wait for GPU slot
-        uses: Oneflow-Inc/get-oneflow/wait-for-gpu@single-matrix-for-efficiency
+        uses: Oneflow-Inc/get-oneflow/wait-for-gpu@support-iree-ci
         if: env.is_secrets_accessible == '1'
         timeout-minutes: 90
         continue-on-error: true
@@ -183,7 +185,7 @@ jobs:
         with:
           ref: ${{ github.event.pull_request.head.sha }}
           repository: ${{github.event.pull_request.head.repo.full_name}}
-      - uses: Oneflow-Inc/get-oneflow/cache-complete/matrix/build@single-matrix-for-efficiency
+      - uses: Oneflow-Inc/get-oneflow/cache-complete/matrix/build@support-iree-ci
         name: find cache
         id: find-cache
         timeout-minutes: 5
@@ -230,7 +232,7 @@ jobs:
         with:
           ref: ${{ github.event.pull_request.head.sha }}
           repository: ${{github.event.pull_request.head.repo.full_name}}
-      - uses: Oneflow-Inc/get-oneflow/cache-complete@single-matrix-for-efficiency
+      - uses: Oneflow-Inc/get-oneflow/cache-complete@support-iree-ci
         name: Save cache if successful
         id: save-cache
         timeout-minutes: 5
@@ -244,13 +246,14 @@ jobs:
         run: |
           echo "::error file=test.yml,line=204,col=10::steps.save-cache.outputs.cache-hit != matrix.cache-hit"
           exit 1
-      - uses: Oneflow-Inc/get-oneflow@single-matrix-for-efficiency
+      - uses: Oneflow-Inc/get-oneflow@support-iree-ci
         name: Build manylinux ${{ matrix.entry }}
         id: build-cpu
         if: ${{ matrix.entry =='cpu' && !matrix.cache-hit }}
         with:
           cmake-init-cache: ${{ env.ONEFLOW_SRC }}/cmake/caches/ci/cpu.cmake
           build-script: ${{ env.ONEFLOW_SRC }}/ci/manylinux/build.sh
+          run-lit: true
           oneflow-src: ${{ env.ONEFLOW_SRC }}
           oneflow-build-env: manylinux
           wheelhouse-dir: ${{ env.WHEELHOUSE_DIR }}
@@ -265,7 +268,7 @@ jobs:
           python-versions: |
             3.6
             3.7
-      - uses: Oneflow-Inc/get-oneflow@single-matrix-for-efficiency
+      - uses: Oneflow-Inc/get-oneflow@support-iree-ci
         name: Build manylinux ${{ matrix.entry }}
         id: build-cuda
         if: ${{ matrix.entry =='cu102' && !matrix.cache-hit }}
@@ -285,7 +288,7 @@ jobs:
           clean-ccache: ${{ contains(github.event.pull_request.labels.*.name, 'need-clean-ccache') }}
           python-versions: |
             3.7
-      - uses: Oneflow-Inc/get-oneflow@single-matrix-for-efficiency
+      - uses: Oneflow-Inc/get-oneflow@support-iree-ci
         name: Build ${{ matrix.entry }}
         if: ${{ matrix.entry == 'llvm13' && !matrix.cache-hit }}
         with:
@@ -324,7 +327,7 @@ jobs:
             })
       - name: Upload packed liboneflow
         if: ${{ !fromJson(matrix.cache-hit) && matrix.entry != 'llvm13' && matrix.entry != 'cu102_xla' }}
-        uses: Oneflow-Inc/get-oneflow/digest/upload@single-matrix-for-efficiency
+        uses: Oneflow-Inc/get-oneflow/digest/upload@support-iree-ci
         timeout-minutes: 10
         with:
           digest: ${{ steps.save-cache.outputs.build-digest }}
@@ -335,7 +338,7 @@ jobs:
           dst-dir: cpack
       - name: Upload whl
         if: ${{ !fromJson(matrix.cache-hit) && matrix.entry != 'llvm13' && matrix.entry != 'cu102_xla' }}
-        uses: Oneflow-Inc/get-oneflow/digest/upload@single-matrix-for-efficiency
+        uses: Oneflow-Inc/get-oneflow/digest/upload@support-iree-ci
         timeout-minutes: 10
         with:
           digest: ${{ steps.save-cache.outputs.build-digest }}
@@ -360,7 +363,7 @@ jobs:
         with:
           ref: ${{ github.event.pull_request.head.sha }}
           repository: ${{github.event.pull_request.head.repo.full_name}}
-      - uses: Oneflow-Inc/get-oneflow/cache-complete/matrix/test@single-matrix-for-efficiency
+      - uses: Oneflow-Inc/get-oneflow/cache-complete/matrix/test@support-iree-ci
         name: find cache
         id: find-cache
         timeout-minutes: 5
@@ -391,7 +394,7 @@ jobs:
         with:
           ref: ${{ github.event.pull_request.head.sha }}
           repository: ${{github.event.pull_request.head.repo.full_name}}
-      - uses: Oneflow-Inc/get-oneflow/cache-complete/matrix/test@single-matrix-for-efficiency
+      - uses: Oneflow-Inc/get-oneflow/cache-complete/matrix/test@support-iree-ci
         name: find cache
         id: find-cache
         timeout-minutes: 5
@@ -455,12 +458,20 @@ jobs:
           # please use a commit here
           ref: ${{ env.LIBAI_COMMIT}}
           path: ${{ env.LIBAI_SRC}}
+      - name: Checkout Oneflow-Inc/oneflow_iree
+        if: ${{ !fromJson(matrix.cache-hit) && contains(matrix.runs-on, 'self-hosted') }}
+        uses: actions/checkout@v2
+        with:
+          repository: Oneflow-Inc/oneflow_iree
+          # please use a commit here
+          ref: ${{ env.ONEFLOW_IREE_COMMIT}}
+          path: ${{ env.ONEFLOW_IREE_SRC}}
       - name: Remove container
         timeout-minutes: 45
         if: ${{ contains(matrix.runs-on, 'self-hosted') }}
         run: |
           docker rm -f ${{ env.TEST_CONTAINER_NAME }} || true
-      - uses: Oneflow-Inc/get-oneflow/cache-complete@single-matrix-for-efficiency
+      - uses: Oneflow-Inc/get-oneflow/cache-complete@support-iree-ci
         name: Save cache if successful
         id: save-cache
         timeout-minutes: 5
@@ -476,7 +487,7 @@ jobs:
           exit 1
       - name: Download wheel and packed liboneflow
         if: ${{ !fromJson(matrix.cache-hit) && contains(matrix.runs-on, 'self-hosted') }}
-        uses: Oneflow-Inc/get-oneflow/digest/download@single-matrix-for-efficiency
+        uses: Oneflow-Inc/get-oneflow/digest/download@support-iree-ci
         id: download-digest
         timeout-minutes: 10
         with:
@@ -486,7 +497,7 @@ jobs:
           ssh-tank-path: ${{ env.SSH_TANK_PATH }}
       - name: Get primary node
         if: ${{ !fromJson(matrix.cache-hit) && contains(matrix.runs-on, 'self-hosted') }}
-        uses: Oneflow-Inc/get-oneflow/master-address@single-matrix-for-efficiency
+        uses: Oneflow-Inc/get-oneflow/master-address@support-iree-ci
         id: get-primary-node
         with:
           rank: ${{ matrix.rank }}
@@ -559,6 +570,7 @@ jobs:
           docker exec ${TEST_CONTAINER_NAME} python3 -m pip install -e ${{ env.FLOW_VISION_SRC}}
           docker exec ${TEST_CONTAINER_NAME} python3 -m pip install pybind11 --user
           docker exec ${TEST_CONTAINER_NAME} python3 -m pip install -e ${{ env.LIBAI_SRC}}
+          docker exec ${TEST_CONTAINER_NAME} python3 -m pip install -e ${{ env.ONEFLOW_IREE_SRC}}
       - name: Module API test (distributed)
         timeout-minutes: 90
         if: ${{ !fromJson(matrix.cache-hit) && matrix.test-type == 'module' && matrix.device == 'cuda' && fromJson(matrix.is-distributed) }}
@@ -648,12 +660,20 @@ jobs:
           # please use a commit here
           ref: ${{ env.LIBAI_COMMIT}}
           path: ${{ env.LIBAI_SRC}}
+      - name: Checkout Oneflow-Inc/oneflow_iree
+        if: ${{ !fromJson(matrix.cache-hit) && contains(matrix.runs-on, 'self-hosted') }}
+        uses: actions/checkout@v2
+        with:
+          repository: Oneflow-Inc/oneflow_iree
+          # please use a commit here
+          ref: ${{ env.ONEFLOW_IREE_COMMIT}}
+          path: ${{ env.ONEFLOW_IREE_SRC}}
       - name: Remove container
         timeout-minutes: 45
         if: ${{ contains(matrix.runs-on, 'self-hosted') }}
         run: |
           docker rm -f ${{ env.TEST_CONTAINER_NAME }} || true
-      - uses: Oneflow-Inc/get-oneflow/cache-complete@single-matrix-for-efficiency
+      - uses: Oneflow-Inc/get-oneflow/cache-complete@support-iree-ci
         name: Save cache if successful
         id: save-cache
         timeout-minutes: 5
@@ -669,7 +689,7 @@ jobs:
           exit 1
       - name: Download wheel and packed liboneflow
         if: ${{ !fromJson(matrix.cache-hit) && contains(matrix.runs-on, 'self-hosted') }}
-        uses: Oneflow-Inc/get-oneflow/digest/download@single-matrix-for-efficiency
+        uses: Oneflow-Inc/get-oneflow/digest/download@support-iree-ci
         id: download-digest
         timeout-minutes: 10
         with:
@@ -781,6 +801,7 @@ jobs:
           docker exec ${TEST_CONTAINER_NAME} python3 -m pip install -e ${{ env.FLOW_VISION_SRC}}
           docker exec ${TEST_CONTAINER_NAME} python3 -m pip install pybind11 --user
           docker exec ${TEST_CONTAINER_NAME} python3 -m pip install -e ${{ env.LIBAI_SRC}}
+          docker exec ${TEST_CONTAINER_NAME} python3 -m pip install -e ${{ env.ONEFLOW_IREE_SRC}}
       - name: Run OneFlow doctor
         if: ${{ !fromJson(matrix.cache-hit) && contains(matrix.runs-on, 'self-hosted') }}
         run: |
@@ -865,7 +886,7 @@ jobs:
               body: "<details>\n <summary>Speed stats:</summary>\n\n ``` \n${{ steps.speed.outputs.stats }}\n ``` \n\n</details>".replace(/\\n/g, '\n')
             })
       - name: Module API test
-        timeout-minutes: 45
+        timeout-minutes: 60
         if: ${{ !fromJson(matrix.cache-hit) && matrix.test-type == 'module' && !fromJson(matrix.is-distributed) }}
         run: |
           docker exec -e ONEFLOW_TEST_DIR=$PWD/python/oneflow/test/modules ${{ env.TEST_CONTAINER_NAME }} bash ci/test/generic_test_multi_client.sh
@@ -883,6 +904,11 @@ jobs:
           docker exec -e ONEFLOW_TEST_DEVICE_NUM=4 -w $PWD/${{ env.LIBAI_SRC }} ${{ env.TEST_CONTAINER_NAME }} python3 -m oneflow.distributed.launch --nproc_per_node 4 -m unittest -f tests/models/test_gpt.py
           docker exec -e ONEFLOW_TEST_DEVICE_NUM=4 -w $PWD/${{ env.LIBAI_SRC }} ${{ env.TEST_CONTAINER_NAME }} python3 -m oneflow.distributed.launch --nproc_per_node 4 -m unittest -f tests/models/test_t5.py
           docker exec -e ONEFLOW_TEST_DEVICE_NUM=4 -w $PWD/${{ env.LIBAI_SRC }} ${{ env.TEST_CONTAINER_NAME }} python3 -m oneflow.distributed.launch --nproc_per_node 4 -m unittest -f tests/models/test_vit.py
+      - name: oneflow_iree test
+        timeout-minutes: 45
+        if: ${{ !fromJson(matrix.cache-hit) && matrix.test-type == 'misc' }}
+        run: |
+          docker exec -w $PWD/${{ env.ONEFLOW_IREE_SRC }} ${{ env.TEST_CONTAINER_NAME }} python3 -m pytest examples
       - name: Expensive tests (models, cases require exclusive access to GPU)
         timeout-minutes: 45
         if: ${{ !fromJson(matrix.cache-hit) && (matrix.test-type == 'speed-test' || (matrix.test-type == 'misc' && matrix.device == 'cpu')) && !fromJson(matrix.is-distributed) }}
@@ -908,7 +934,7 @@ jobs:
       - name: Benchmark Test
         timeout-minutes: 100
         if: ${{ !fromJson(matrix.cache-hit) && matrix.test-type == 'benchmark' && matrix.device == 'cuda' }}
-        uses: Oneflow-Inc/get-oneflow/pytest-benchmark@single-matrix-for-efficiency
+        uses: Oneflow-Inc/get-oneflow/pytest-benchmark@support-iree-ci
         with:
           collect-path: ${{ env.FLOW_VISION_SRC }}/benchmark
           container-name: ${{ env.TEST_CONTAINER_NAME }}
@@ -961,7 +987,7 @@ jobs:
           ref: ${{ github.event.pull_request.head.sha }}
           repository: ${{github.event.pull_request.head.repo.full_name}}
           fetch-depth: 0
-      - uses: Oneflow-Inc/get-oneflow/cache-complete@single-matrix-for-efficiency
+      - uses: Oneflow-Inc/get-oneflow/cache-complete@support-iree-ci
         name: Save cache if successful
         id: save-cache
         timeout-minutes: 5

ci/manylinux/build-gcc7.sh

@@ -31,6 +31,11 @@ cmake -S ${ONEFLOW_CI_SRC_DIR} -C ${ONEFLOW_CI_CMAKE_INIT_CACHE} -DPython3_EXECU
 # cmake build
 cd ${ONEFLOW_CI_BUILD_DIR}
 cmake --build . --parallel ${ONEFLOW_CI_BUILD_PARALLEL}
+if [ ! -z "$ONEFLOW_CI_BUILD_RUN_LIT" ]; then
+    ${ONEFLOW_CI_PYTHON_EXE} -m pip install -i https://mirrors.aliyun.com/pypi/simple --user flowvision==0.1.0
+    export PATH=$PATH:$(dirname $ONEFLOW_CI_PYTHON_EXE)
+    cmake --build . -t c1
+fi
 
 # build pip
 cd ${ONEFLOW_CI_SRC_DIR}

ci/manylinux/build.sh

@@ -27,6 +27,11 @@ cmake -S ${ONEFLOW_CI_SRC_DIR} -C ${ONEFLOW_CI_CMAKE_INIT_CACHE} -DPython3_EXECU
 # cmake build
 cd ${ONEFLOW_CI_BUILD_DIR}
 cmake --build . --parallel ${ONEFLOW_CI_BUILD_PARALLEL}
+if [ ! -z "$ONEFLOW_CI_BUILD_RUN_LIT" ]; then
+    ${ONEFLOW_CI_PYTHON_EXE} -m pip install -i https://mirrors.aliyun.com/pypi/simple --user flowvision==0.1.0
+    export PATH=$PATH:$(dirname $ONEFLOW_CI_PYTHON_EXE)
+    cmake --build . -t c1
+fi
 
 # build pip
 cd ${ONEFLOW_CI_SRC_DIR}

cmake/caches/cn/fast/mlir-cpu.cmake

@@ -0,0 +1,24 @@
+set(BUILD_SHARED_LIBS YES CACHE BOOL "")
+# uncomment only if you know what you are doing
+# set(CMAKE_LINK_DEPENDS_NO_SHARED YES CACHE BOOL "")
+set(BUILD_CUDA NO CACHE BOOL "")
+set(BUILD_GIT_VERSION NO CACHE BOOL "")
+set(TREAT_WARNINGS_AS_ERRORS YES CACHE BOOL "")
+set(BUILD_HWLOC NO CACHE BOOL "")
+set(BUILD_TESTING OFF CACHE BOOL "")
+set(WITH_MLIR YES CACHE BOOL "")
+set(WITH_MLIR_CUDA_CODEGEN NO CACHE BOOL "")
+set(THIRD_PARTY_MIRROR aliyun CACHE STRING "")
+set(PIP_INDEX_MIRROR "https://pypi.tuna.tsinghua.edu.cn/simple" CACHE STRING "")
+set(CMAKE_BUILD_TYPE RelWithDebInfo CACHE STRING "")
+set(CMAKE_GENERATOR Ninja CACHE STRING "")
+set(CMAKE_C_COMPILER_LAUNCHER ccache CACHE STRING "")
+set(CMAKE_CXX_COMPILER_LAUNCHER ccache CACHE STRING "")
+set(CMAKE_INTERPROCEDURAL_OPTIMIZATION OFF CACHE BOOL "")
+set(CMAKE_EXE_LINKER_FLAGS_INIT "-fuse-ld=lld" CACHE STRING "")
+set(CMAKE_MODULE_LINKER_FLAGS_INIT "-fuse-ld=lld" CACHE STRING "")
+set(CMAKE_SHARED_LINKER_FLAGS_INIT "-fuse-ld=lld" CACHE STRING "")
+set(CPU_THREADING_RUNTIME SEQ CACHE STRING
+                                    "when using lld with TBB enabled, there will be linkage error")
+set(BUILD_HWLOC OFF CACHE BOOL "")
+set(WITH_ONEDNN OFF CACHE BOOL "")