ci: reuse llama.cpp distributions for windows cuda packaging (#2927)

* ci: reuse llama.cpp distributions for windows cuda packaging * fix * fix * update * update * update * update
TabbyML · Aug 21, 2024 · 735080a · 735080a
1 parent 12a950f
commit 735080a
Show file tree

Hide file tree

Showing 2 changed files with 62 additions and 3 deletions.
diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml
@@ -35,8 +35,8 @@ jobs:
           - x86_64-manylinux2014-vulkan
           - x86_64-windows-msvc
           - x86_64-windows-msvc-vulkan
-          - x86_64-windows-msvc-cuda117
-          - x86_64-windows-msvc-cuda122
+          # - x86_64-windows-msvc-cuda117
+          # - x86_64-windows-msvc-cuda122
         include:
           - os: macos-latest
             target: aarch64-apple-darwin
@@ -244,9 +244,42 @@ jobs:
           retention-days: 3
           name: tabby_${{ matrix.binary }}${{ matrix.ext }}
           path: tabby_${{ matrix.binary }}${{ matrix.ext }}
+
+  package-win-cuda:
+    runs-on: ubuntu-latest
+    env:
+      LLAMA_CPP_VERSION: b3571
+    needs: [release-binary]
+    permissions:
+      contents: write
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v3
+
+      - name: Download all artifacts
+        uses: actions/download-artifact@v3
+
+      - name: Display structure of downloaded files
+        run: ls -R
+
+      - name: Package CUDA 11.7
+        run: >
+          LLAMA_CPP_PLATFORM=cuda-cu11.7.1-x64 OUTPUT_NAME=tabby_x86_64-windows-msvc-cuda117 ./ci/package-win.sh
+
+      - name: Package CUDA 12.2
+        run: >
+          LLAMA_CPP_PLATFORM=cuda-cu12.2.0-x64 OUTPUT_NAME=tabby_x86_64-windows-msvc-cuda122 ./ci/package-win.sh
+
+      - name: Upload artifacts
+        uses: actions/upload-artifact@v3
+        with:
+          retention-days: 3
+          name: dist
+          path: dist/
+
 
   pre-release:
-    needs: [release-llama-server-binary, release-binary]
+    needs: [release-llama-server-binary, release-binary, package-win-cuda]
     runs-on: ubuntu-latest
     permissions:
       contents: write
@@ -291,6 +324,9 @@ jobs:
               fi
             done
           done
+      
+      - name: Display structure of created files
+        run: ls -R dist
 
       - uses: ncipollo/release-action@v1
         if: github.event_name == 'push'

diff --git a/ci/package-win.sh b/ci/package-win.sh
@@ -0,0 +1,23 @@
+#!/bin/sh
+
+# Input variables
+LLAMA_CPP_VERSION=${LLAMA_CPP_VERSION:-b3571}
+LLAMA_CPP_PLATFORM=${LLAMA_CPP_PLATFORM:-cuda-cu11.7.1-x64}
+OUTPUT_NAME=${OUTPUT_NAME:-tabby_x86_64-windows-msvc-cuda117}
+
+NAME=llama-${LLAMA_CPP_VERSION}-bin-win-${LLAMA_CPP_PLATFORM}
+ZIP_FILE=${NAME}.zip
+
+curl https://github.com/ggerganov/llama.cpp/releases/download/${LLAMA_CPP_VERSION}/${ZIP_FILE} -L -o ${ZIP_FILE}
+unzip ${ZIP_FILE} -d ${OUTPUT_NAME}
+
+pushd ${OUTPUT_NAME}
+rm $(ls *.exe | grep -v "llama-server")
+cp ../tabby_x86_64-windows-msvc.exe tabby.exe
+popd
+
+zip -r ${OUTPUT_NAME}.zip ${OUTPUT_NAME}
+rm -rf ${OUTPUT_NAME}
+
+mkdir -p dist
+mv ${OUTPUT_NAME}.zip dist/