@@ -27,29 +27,28 @@ jobs:
2727 - name : Install dependencies
2828 run : yarn install --frozen-lockfile
2929
30- - name : Get latest llama.cpp release tag
30+ - name : Download llama.cpp release binaries
3131 id : llama
3232 shell : bash
3333 run : |
34- TAG=$(curl -s https://api.github.com/repos/ggml-org/llama.cpp/releases/latest | jq -r '.tag_name')
34+ # Fetch all asset download URLs from the latest release
35+ RELEASE=$(curl -s https://api.github.com/repos/ggml-org/llama.cpp/releases/latest)
36+ TAG=$(echo "$RELEASE" | jq -r '.tag_name')
3537 echo "tag=$TAG" >> "$GITHUB_OUTPUT"
3638 echo "Latest llama.cpp release: $TAG"
3739
38- - name : Download llama.cpp CUDA binaries
39- shell : bash
40- run : |
41- TAG="${{ steps.llama.outputs.tag }}"
42- URL="https://github.com/ggml-org/llama.cpp/releases/download/${TAG}/llama-${TAG}-bin-win-cuda-13.1-x64.zip"
43- echo "Downloading $URL"
44- curl -L -o llama-cuda.zip "$URL"
40+ # List all assets so we can see exact names
41+ echo "$RELEASE" | jq -r '.assets[].name' | head -30
4542
46- - name : Download llama.cpp CUDA runtime
47- shell : bash
48- run : |
49- TAG="${{ steps.llama.outputs.tag }}"
50- URL="https://github.com/ggml-org/llama.cpp/releases/download/${TAG}/cudart-llama-bin-win-cuda-13.1-x64.zip"
51- echo "Downloading $URL"
52- curl -L -o llama-cudart.zip "$URL"
43+ # Find and download the CUDA binary
44+ CUDA_URL=$(echo "$RELEASE" | jq -r '.assets[] | select(.name | test("bin-win-cuda.*x64\\.zip$")) | select(.name | test("cudart") | not) | .browser_download_url')
45+ echo "CUDA binary URL: $CUDA_URL"
46+ curl -fL -o llama-cuda.zip "$CUDA_URL"
47+
48+ # Find and download the CUDA runtime
49+ CUDART_URL=$(echo "$RELEASE" | jq -r '.assets[] | select(.name | test("cudart.*cuda.*x64\\.zip$")) | .browser_download_url')
50+ echo "CUDART URL: $CUDART_URL"
51+ curl -fL -o llama-cudart.zip "$CUDART_URL"
5352
5453 - name : Extract and merge into assets/server
5554 shell : bash
0 commit comments