Merge remote-tracking branch 'xinhao/merged_bert' into bert_fix2

.dockerignore

@@ -11,3 +11,11 @@ python/flexflow/core/legion_cffi_header.py
 *.pb.h
 *.o
 *.a
+
+# Ignore inference assets
+/inference/weights/*
+/inference/tokenizer/*
+/inference/prompt/*
+/inference/output/*
+
+/tests/inference/python_test_configs/*.json

.github/PULL_REQUEST_TEMPLATE.md

@@ -10,6 +10,3 @@ Linked Issues:
 Issues closed by this PR:
 - Closes #
 
-**Before merging:**
-
-- [ ] Did you update the [flexflow-third-party](https://github.com/flexflow/flexflow-third-party) repo, if modifying any of the Cmake files, the build configs, or the submodules?

.github/workflows/build-skip.yml

@@ -3,6 +3,7 @@ on:
   pull_request:
     paths-ignore:
       - "include/**"
+      - "inference/**"
       - "cmake/**"
       - "config/**"
       - "deps/**"

.github/workflows/build.yml

@@ -3,6 +3,7 @@ on:
   pull_request:
     paths:
       - "include/**"
+      - "inference/**"
       - "cmake/**"
       - "config/**"
       - "deps/**"
@@ -15,6 +16,7 @@ on:
       - "master"
     paths:
       - "include/**"
+      - "inference/**"
       - "cmake/**"
       - "config/**"
       - "deps/**"
@@ -38,6 +40,8 @@ jobs:
       matrix:
         gpu_backend: ["cuda", "hip_rocm"]
       fail-fast: false
+    env:
+      FF_GPU_BACKEND: ${{ matrix.gpu_backend }}
     steps:
       - name: Checkout Git Repository
         uses: actions/checkout@v3
@@ -48,39 +52,49 @@ jobs:
         run: .github/workflows/helpers/free_space_on_runner.sh
 
       - name: Install CUDA
-        uses: Jimver/[email protected]
+        uses: Jimver/[email protected]
+        if: ${{ matrix.gpu_backend == 'cuda' }}
         id: cuda-toolkit
         with:
-          cuda: "11.8.0"
+          cuda: "12.1.1"
           # Disable caching of the CUDA binaries, since it does not give us any significant performance improvement
           use-github-cache: "false"
+          log-file-suffix: 'cmake_${{matrix.gpu_backend}}.txt'
 
       - name: Install system dependencies
-        run: FF_GPU_BACKEND=${{ matrix.gpu_backend }} .github/workflows/helpers/install_dependencies.sh
+        run: .github/workflows/helpers/install_dependencies.sh
 
       - name: Install conda and FlexFlow dependencies
         uses: conda-incubator/setup-miniconda@v2
         with:
           activate-environment: flexflow
-          environment-file: conda/environment.yml
+          environment-file: conda/flexflow.yml
           auto-activate-base: false
 
       - name: Build FlexFlow
         run: |
           export CUDNN_DIR="$CUDA_PATH"
           export CUDA_DIR="$CUDA_PATH"
           export FF_HOME=$(pwd)
-          export FF_GPU_BACKEND=${{ matrix.gpu_backend }}
           export FF_CUDA_ARCH=70
+          export FF_HIP_ARCH=gfx1100,gfx1036
+          export hip_version=5.6
+          export FF_BUILD_ALL_INFERENCE_EXAMPLES=ON
+
+          if [[ "${FF_GPU_BACKEND}" == "cuda" ]]; then
+            export FF_BUILD_ALL_EXAMPLES=ON
+            export FF_BUILD_UNIT_TESTS=ON
+          else 
+            export FF_BUILD_ALL_EXAMPLES=OFF
+            export FF_BUILD_UNIT_TESTS=OFF
+          fi
+
           cores_available=$(nproc --all)
           n_build_cores=$(( cores_available -1 ))
           if (( $n_build_cores < 1 )) ; then n_build_cores=1 ; fi
           mkdir build
           cd build
-          if [[ "${FF_GPU_BACKEND}" == "cuda" ]]; then
-            export FF_BUILD_ALL_EXAMPLES=ON 
-            export FF_BUILD_UNIT_TESTS=ON
-          fi
+          
           ../config/config.linux
           make -j $n_build_cores
 
@@ -89,35 +103,44 @@ jobs:
           export CUDNN_DIR="$CUDA_PATH"
           export CUDA_DIR="$CUDA_PATH"
           export FF_HOME=$(pwd)
-          export FF_GPU_BACKEND=${{ matrix.gpu_backend }}
           export FF_CUDA_ARCH=70
-          cd build
+          export FF_HIP_ARCH=gfx1100,gfx1036
+          export hip_version=5.6
+          export FF_BUILD_ALL_INFERENCE_EXAMPLES=ON
+          
           if [[ "${FF_GPU_BACKEND}" == "cuda" ]]; then
-            export FF_BUILD_ALL_EXAMPLES=ON 
+            export FF_BUILD_ALL_EXAMPLES=ON
             export FF_BUILD_UNIT_TESTS=ON
+          else 
+            export FF_BUILD_ALL_EXAMPLES=OFF
+            export FF_BUILD_UNIT_TESTS=OFF
           fi
+
+          cd build
           ../config/config.linux
           sudo make install
           sudo ldconfig
 
-      - name: Check availability of Python flexflow.core module
-        if: ${{ matrix.gpu_backend == 'cuda' }}
-        run: |
-          export LD_LIBRARY_PATH="$CUDA_PATH/lib64/stubs:$LD_LIBRARY_PATH"
-          sudo ln -s "$CUDA_PATH/lib64/stubs/libcuda.so" "$CUDA_PATH/lib64/stubs/libcuda.so.1"
-          export CPU_ONLY_TEST=1
-          python -c "import flexflow.core; exit()"
-
       - name: Run C++ unit tests
         if: ${{ matrix.gpu_backend == 'cuda' }}
         run: |
           export CUDNN_DIR="$CUDA_PATH"
           export CUDA_DIR="$CUDA_PATH"
           export LD_LIBRARY_PATH="$CUDA_PATH/lib64/stubs:$LD_LIBRARY_PATH"
           export FF_HOME=$(pwd)
+          sudo ln -s "$CUDA_PATH/lib64/stubs/libcuda.so" "$CUDA_PATH/lib64/stubs/libcuda.so.1"
           cd build
           ./tests/unit/unit-test
 
+      - name: Check availability of flexflow modules in Python
+        run: |
+          if [[ "${FF_GPU_BACKEND}" == "cuda" ]]; then
+            export LD_LIBRARY_PATH="$CUDA_PATH/lib64/stubs:$LD_LIBRARY_PATH"
+          fi
+          # Remove build folder to check that the installed version can run independently of the build files
+          rm -rf build
+          python -c "import flexflow.core; import flexflow.serve as ff; exit()"
+
   makefile-build:
     name: Build FlexFlow with the Makefile
     runs-on: ubuntu-20.04
@@ -134,11 +157,12 @@ jobs:
         run: .github/workflows/helpers/free_space_on_runner.sh
 
       - name: Install CUDA
-        uses: Jimver/[email protected].11
+        uses: Jimver/[email protected].16
         id: cuda-toolkit
         with:
-          cuda: "11.8.0"
+          cuda: "12.1.1"
           use-github-cache: "false"
+          log-file-suffix: 'makefile_${{matrix.gpu_backend}}.txt'
 
       - name: Install system dependencies
         run: .github/workflows/helpers/install_dependencies.sh
@@ -147,7 +171,7 @@ jobs:
         uses: conda-incubator/setup-miniconda@v2
         with:
           activate-environment: flexflow
-          environment-file: conda/environment.yml
+          environment-file: conda/flexflow.yml
           auto-activate-base: false
 
       - name: Build FlexFlow
@@ -163,5 +187,4 @@ jobs:
 
           cd python
           make -j $n_build_cores
-          export CPU_ONLY_TEST=1
           python -c 'import flexflow.core'

.github/workflows/clang-format-check.yml

@@ -10,7 +10,7 @@ jobs:
           - check: "src"
             exclude: '\.proto$'
           - check: "include"
-          - check: "nmt"
+          - check: "inference"
           - check: "python"
           - check: "scripts"
           - check: "tests"

.github/workflows/docker-build-skip.yml

@@ -13,27 +13,22 @@ concurrency:
   cancel-in-progress: true
 
 jobs:
-  docker-build:
-    name: Build and Install FlexFlow in a Docker Container
-    runs-on: ubuntu-20.04
+  docker-build-rocm:
+    name: Build and Install FlexFlow in a Docker Container (ROCm backend)
+    runs-on: ubuntu-latest
     strategy:
       matrix:
-        gpu_backend: ["cuda", "hip_rocm"]
-        cuda_version: ["11.1", "11.2", "11.3", "11.5", "11.6", "11.7", "11.8"]
-        # The CUDA version doesn't matter when building for hip_rocm, so we just pick one arbitrarily (11.8) to avoid building for hip_rocm once per number of CUDA version supported
-        exclude:
-          - gpu_backend: "hip_rocm"
-            cuda_version: "11.1"
-          - gpu_backend: "hip_rocm"
-            cuda_version: "11.2"
-          - gpu_backend: "hip_rocm"
-            cuda_version: "11.3"
-          - gpu_backend: "hip_rocm"
-            cuda_version: "11.5"
-          - gpu_backend: "hip_rocm"
-            cuda_version: "11.6"
-          - gpu_backend: "hip_rocm"
-            cuda_version: "11.7"
+        hip_version: ["5.3", "5.4", "5.5", "5.6"]
+      fail-fast: false
+    steps:
+      - run: 'echo "No docker-build required"'
+
+  docker-build-cuda:
+    name: Build and Install FlexFlow in a Docker Container (CUDA backend)
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        cuda_version: ["11.1", "11.6", "11.7", "11.8", "12.0", "12.1", "12.2"]
       fail-fast: false
     steps:
       - run: 'echo "No docker-build required"'