@@ -137,78 +137,6 @@ jobs:
137137 path : cudart-vulkan-llama-${{ steps.tag.outputs.name }}-bin-ubuntu-x64.zip
138138 name : cudart-vulkan-llama-bin-ubuntu-x64.zip
139139
140- windows-latest-cmake-cuda :
141- runs-on : windows-2019
142-
143- strategy :
144- matrix :
145- cuda : ["12.2.0", "11.7.1"]
146- build : ["cuda"]
147-
148- steps :
149- - name : Clone
150- id : checkout
151- uses : actions/checkout@v4
152- with :
153- fetch-depth : 0
154-
155- - name : Install CUDA toolkit
156- id : cuda-toolkit
157- 158- with :
159- cuda : ${{ matrix.cuda }}
160- method : " network"
161- sub-packages : ' ["nvcc", "cudart", "cublas", "cublas_dev", "thrust", "visual_studio_integration"]'
162-
163- - name : Build
164- id : cmake_build
165- run : |
166- mkdir build
167- cd build
168- cmake .. -DGGML_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DGGML_CUDA=ON -DBUILD_SHARED_LIBS=ON -DGGML_RPC=ON
169- cmake --build . --config Release -j $((${env:NUMBER_OF_PROCESSORS} - 1)) -t ggml
170- cmake --build . --config Release -j ${env:NUMBER_OF_PROCESSORS}
171-
172- - name : Determine tag name
173- id : tag
174- shell : bash
175- run : |
176- BUILD_NUMBER="$(git rev-list --count HEAD)"
177- SHORT_HASH="$(git rev-parse --short=7 HEAD)"
178- if [[ "${{ env.BRANCH_NAME }}" == "master" ]]; then
179- echo "name=b${BUILD_NUMBER}" >> $GITHUB_OUTPUT
180- else
181- SAFE_NAME=$(echo "${{ env.BRANCH_NAME }}" | tr '/' '-')
182- echo "name=${SAFE_NAME}-b${BUILD_NUMBER}-${SHORT_HASH}" >> $GITHUB_OUTPUT
183- fi
184-
185- - name : Pack artifacts
186- id : pack_artifacts
187- if : ${{ ( github.event_name == 'push' && github.ref == 'refs/heads/master' ) || github.event.inputs.create_release == 'true' }}
188- run : |
189- 7z a llama-${{ steps.tag.outputs.name }}-bin-win-${{ matrix.build }}-cu${{ matrix.cuda }}-x64.zip .\build\bin\Release\*
190-
191- - name : Upload artifacts
192- if : ${{ ( github.event_name == 'push' && github.ref == 'refs/heads/master' ) || github.event.inputs.create_release == 'true' }}
193- uses : actions/upload-artifact@v4
194- with :
195- path : llama-${{ steps.tag.outputs.name }}-bin-win-${{ matrix.build }}-cu${{ matrix.cuda }}-x64.zip
196- name : llama-bin-win-cu${{ matrix.cuda }}-x64.zip
197-
198- - name : Copy and pack Cuda runtime
199- run : |
200- echo "Cuda install location: ${{steps.cuda-toolkit.outputs.CUDA_PATH}}"
201- $dst='.\build\bin\cudart\'
202- robocopy "${{steps.cuda-toolkit.outputs.CUDA_PATH}}\bin" $dst cudart64_*.dll cublas64_*.dll cublasLt64_*.dll
203- 7z a cudart-llama-bin-win-cu${{ matrix.cuda }}-x64.zip $dst\*
204-
205- - name : Upload Cuda runtime
206- if : ${{ ( github.event_name == 'push' && github.ref == 'refs/heads/master' ) || github.event.inputs.create_release == 'true' }}
207- uses : actions/upload-artifact@v4
208- with :
209- path : cudart-llama-bin-win-cu${{ matrix.cuda }}-x64.zip
210- name : cudart-llama-bin-win-cu${{ matrix.cuda }}-x64.zip
211-
212140 # TODO: build with GGML_NO_METAL because test-backend-ops fail on "Apple Paravirtual device" and I don't know
213141 # how to debug it.
214142 # ref: https://github.com/ggerganov/llama.cpp/actions/runs/7131777249/job/19420981052#step:5:1124
0 commit comments