加入int4下linear的向量化访存，提高推理速度。 #103
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| name: Action Build | |
| on: [push] | |
| jobs: | |
| build: | |
| runs-on: ubuntu-latest | |
| steps: | |
| - name: Set up JDK | |
| uses: actions/setup-java@v1 | |
| with: | |
| java-version: '11' | |
| - name: Checkout code | |
| uses: actions/checkout@v2 | |
| - name: Build with arm64-v8a | |
| run: | | |
| wget https://dl.google.com/android/repository/android-ndk-r21d-linux-x86_64.zip | |
| unzip android-ndk-r21d-linux-x86_64.zip | |
| export NDK=$GITHUB_WORKSPACE/android-ndk-r21d | |
| mkdir build-android | |
| cd build-android | |
| #ls ${NDK}/build/cmake/android.toolchain.cmake | |
| cmake -DCMAKE_MAKE_PROGRAM=/usr/bin/make -DCMAKE_CXX_COMPILER=/usr/bin/g++ -DCMAKE_TOOLCHAIN_FILE=${NDK}/build/cmake/android.toolchain.cmake -DANDROID_ABI=arm64-v8a -DANDROID_PLATFORM=android-23 -DCMAKE_CXX_FLAGS=-march=armv8.2a+dotprod .. | |
| make -j -B | |
| cp main fastllm-main-android | |
| - name: Build with x86 | |
| run: | | |
| mkdir build-x86 | |
| cd build-x86 | |
| cmake .. -DUSE_CUDA=OFF | |
| make -j | |
| cp main fastllm-main-x86_64 | |
| - name: Export and Upload Artifact | |
| uses: actions/upload-artifact@v2 | |
| with: | |
| name: Output | |
| path: | | |
| build-android/fastllm-main-android | |
| build-x86/fastllm-main-x86_64 |