feat: Upgrade to CUDA 12.5 (#2601)

Signed-off-by: Rene Leonhardt <65483435+reneleonhardt@users.noreply.github.com>

.github/workflows/image-pr.yml

@@ -46,7 +46,7 @@ jobs:
             makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "12"
-            cuda-minor-version: "1"
+            cuda-minor-version: "5"
             platforms: 'linux/amd64'
             tag-latest: 'false'
             tag-suffix: '-cublas-cuda12-ffmpeg'
@@ -119,7 +119,7 @@ jobs:
             makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "12"
-            cuda-minor-version: "1"
+            cuda-minor-version: "5"
             platforms: 'linux/amd64'
             tag-latest: 'false'
             tag-suffix: '-cublas-cuda12-ffmpeg-core'
@@ -127,4 +127,4 @@ jobs:
             image-type: 'core'
             runs-on: 'ubuntu-latest'
             base-image: "ubuntu:22.04"
-            makeflags: "--jobs=4 --output-sync=target"
+            makeflags: "--jobs=4 --output-sync=target"

.github/workflows/image.yml

@@ -64,7 +64,7 @@ jobs:
             makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "11"
-            cuda-minor-version: "7"
+            cuda-minor-version: "8"
             platforms: 'linux/amd64'
             tag-latest: 'false'
             tag-suffix: '-cublas-cuda11'
@@ -75,7 +75,7 @@ jobs:
             makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "12"
-            cuda-minor-version: "1"
+            cuda-minor-version: "5"
             platforms: 'linux/amd64'
             tag-latest: 'false'
             tag-suffix: '-cublas-cuda12'
@@ -86,7 +86,7 @@ jobs:
             makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "11"
-            cuda-minor-version: "7"
+            cuda-minor-version: "8"
             platforms: 'linux/amd64'
             tag-latest: 'auto'
             tag-suffix: '-cublas-cuda11-ffmpeg'
@@ -100,7 +100,7 @@ jobs:
             makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "12"
-            cuda-minor-version: "1"
+            cuda-minor-version: "5"
             platforms: 'linux/amd64'
             tag-latest: 'auto'
             tag-suffix: '-cublas-cuda12-ffmpeg'
@@ -232,7 +232,7 @@ jobs:
             grpc-base-image: "ubuntu:22.04"
             runs-on: 'arc-runner-set'
             makeflags: "--jobs=3 --output-sync=target"
-  
+
   core-image-build:
     uses: ./.github/workflows/image_build.yml
     with:
@@ -273,7 +273,7 @@ jobs:
             makeflags: "--jobs=4 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "11"
-            cuda-minor-version: "7"
+            cuda-minor-version: "8"
             platforms: 'linux/amd64'
             tag-latest: 'false'
             tag-suffix: '-cublas-cuda11-core'
@@ -284,7 +284,7 @@ jobs:
             makeflags: "--jobs=4 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "12"
-            cuda-minor-version: "1"
+            cuda-minor-version: "5"
             platforms: 'linux/amd64'
             tag-latest: 'false'
             tag-suffix: '-cublas-cuda12-core'
@@ -295,7 +295,7 @@ jobs:
             makeflags: "--jobs=4 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "11"
-            cuda-minor-version: "7"
+            cuda-minor-version: "8"
             platforms: 'linux/amd64'
             tag-latest: 'false'
             tag-suffix: '-cublas-cuda11-ffmpeg-core'
@@ -306,7 +306,7 @@ jobs:
             makeflags: "--jobs=4 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "12"
-            cuda-minor-version: "1"
+            cuda-minor-version: "5"
             platforms: 'linux/amd64'
             tag-latest: 'false'
             tag-suffix: '-cublas-cuda12-ffmpeg-core'

.github/workflows/image_build.yml

@@ -19,11 +19,11 @@ on:
         type: string
       cuda-major-version:
         description: 'CUDA major version'
-        default: "11"
+        default: "12"
         type: string
       cuda-minor-version:
         description: 'CUDA minor version'
-        default: "7"
+        default: "5"
         type: string
       platforms:
         description: 'Platforms'
@@ -324,7 +324,7 @@ jobs:
           docker pull quay.io/go-skynet/local-ai:${{ steps.meta_aio.outputs.version }}
           docker tag quay.io/go-skynet/local-ai:${{ steps.meta_aio.outputs.version }} quay.io/go-skynet/local-ai:${{ inputs.latest-image-aio }}
           docker push quay.io/go-skynet/local-ai:${{ inputs.latest-image-aio }}
-  
+
       - name: job summary
         run: |
           echo "Built image: ${{ steps.meta.outputs.labels }}" >> $GITHUB_STEP_SUMMARY

.github/workflows/release.yaml

@@ -40,7 +40,7 @@ jobs:
           sudo apt-get update
           sudo apt-get install -y cuda-cross-aarch64 cuda-nvcc-cross-aarch64-${CUDA_VERSION} libcublas-cross-aarch64-${CUDA_VERSION}
         env:
-          CUDA_VERSION: 12-4
+          CUDA_VERSION: 12-5
       - name: Cache grpc
         id: cache-grpc
         uses: actions/cache@v4
@@ -167,15 +167,15 @@ jobs:
           ROCM_VERSION: "6.1"
           AMDGPU_VERSION: "6.1"
         run: |
-            set -ex 
+            set -ex
 
             sudo apt-get update
-            sudo DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends ca-certificates curl libnuma-dev gnupg 
+            sudo DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends ca-certificates curl libnuma-dev gnupg
-            
+
-            curl -sL https://repo.radeon.com/rocm/rocm.gpg.key | sudo apt-key add - 
+            curl -sL https://repo.radeon.com/rocm/rocm.gpg.key | sudo apt-key add -
-              
+
             printf "deb [arch=amd64] https://repo.radeon.com/rocm/apt/$ROCM_VERSION/ jammy main" | sudo tee /etc/apt/sources.list.d/rocm.list
-            
+
             printf "deb [arch=amd64] https://repo.radeon.com/amdgpu/$AMDGPU_VERSION/ubuntu jammy main" | sudo tee /etc/apt/sources.list.d/amdgpu.list
             printf 'Package: *\nPin: release o=repo.radeon.com\nPin-Priority: 600' | sudo tee /etc/apt/preferences.d/rocm-pin-600
             sudo apt-get update
@@ -183,10 +183,10 @@ jobs:
             sudo DEBIAN_FRONTEND=noninteractive apt-get install -y \
                 hipblas-dev rocm-dev \
                 rocblas-dev
-          
+
             sudo apt-get clean
             sudo rm -rf /var/lib/apt/lists/*
-            sudo ldconfig 
+            sudo ldconfig
       - name: Cache grpc
         id: cache-grpc
         uses: actions/cache@v4
@@ -291,7 +291,7 @@ jobs:
           export C_INCLUDE_PATH=/usr/local/include
           export CPLUS_INCLUDE_PATH=/usr/local/include
           export PATH=$PATH:$GOPATH/bin
-          
+
           BACKEND_LIBS="$(ls /opt/homebrew/opt/grpc/lib/*.dylib /opt/homebrew/opt/re2/lib/*.dylib /opt/homebrew/opt/openssl@3/lib/*.dylib /opt/homebrew/opt/protobuf/lib/*.dylib /opt/homebrew/opt/abseil/lib/*.dylib | xargs)" GO_TAGS=p2p make dist
       - uses: actions/upload-artifact@v4
         with:
@@ -309,4 +309,4 @@ jobs:
         with:
           detached: true
           connect-timeout-seconds: 180
-          limit-access-to-actor: true
+          limit-access-to-actor: true

Dockerfile

@@ -33,7 +33,7 @@ RUN curl -L -s https://go.dev/dl/go${GO_VERSION}.linux-${TARGETARCH}.tar.gz | ta
 ENV PATH $PATH:/root/go/bin:/usr/local/go/bin
 
 # Install grpc compilers
-RUN go install google.golang.org/protobuf/cmd/protoc-gen-go@v1.34.1 && \
+RUN go install google.golang.org/protobuf/cmd/protoc-gen-go@v1.34.2 && \
     go install google.golang.org/grpc/cmd/protoc-gen-go-grpc@1958fcbe2ca8bd93af633f11e97d44e567e945af
 
 COPY --chmod=644 custom-ca-certs/* /usr/local/share/ca-certificates/
@@ -98,8 +98,8 @@ RUN pip install --user grpcio-tools
 FROM requirements-${IMAGE_TYPE} AS requirements-drivers
 
 ARG BUILD_TYPE
-ARG CUDA_MAJOR_VERSION=11
+ARG CUDA_MAJOR_VERSION=12
-ARG CUDA_MINOR_VERSION=8
+ARG CUDA_MINOR_VERSION=5
 
 ENV BUILD_TYPE=${BUILD_TYPE}
 
@@ -292,7 +292,7 @@ ENV REBUILD=false
 ENV HEALTHCHECK_ENDPOINT=http://localhost:8080/readyz
 ENV MAKEFLAGS=${MAKEFLAGS}
 
-ARG CUDA_MAJOR_VERSION=11
+ARG CUDA_MAJOR_VERSION=12
 ENV NVIDIA_DRIVER_CAPABILITIES=compute,utility
 ENV NVIDIA_REQUIRE_CUDA="cuda>=${CUDA_MAJOR_VERSION}.0"
 ENV NVIDIA_VISIBLE_DEVICES=all

Makefile

@@ -400,7 +400,7 @@ prepare-e2e:
 	mkdir -p $(TEST_DIR)
 	cp -rfv $(abspath ./tests/e2e-fixtures)/gpu.yaml $(TEST_DIR)/gpu.yaml
 	test -e $(TEST_DIR)/ggllm-test-model.bin || wget -q https://huggingface.co/TheBloke/CodeLlama-7B-Instruct-GGUF/resolve/main/codellama-7b-instruct.Q2_K.gguf -O $(TEST_DIR)/ggllm-test-model.bin
-	docker build --build-arg GRPC_BACKENDS="$(GRPC_BACKENDS)" --build-arg IMAGE_TYPE=core --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg CUDA_MAJOR_VERSION=11 --build-arg CUDA_MINOR_VERSION=7 --build-arg FFMPEG=true -t localai-tests .
+	docker build --build-arg GRPC_BACKENDS="$(GRPC_BACKENDS)" --build-arg IMAGE_TYPE=core --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg CUDA_MAJOR_VERSION=12 --build-arg CUDA_MINOR_VERSION=5 --build-arg FFMPEG=true -t localai-tests .
 
 run-e2e-image:
 	ls -liah $(abspath ./tests/e2e-fixtures)
@@ -810,6 +810,17 @@ docker:
 		--build-arg BUILD_TYPE=$(BUILD_TYPE) \
 		-t $(DOCKER_IMAGE) .
 
+docker-cuda11:
+	docker build \
+		--build-arg CUDA_MAJOR_VERSION=11 \
+		--build-arg CUDA_MINOR_VERSION=8 \
+		--build-arg BASE_IMAGE=$(BASE_IMAGE) \
+		--build-arg IMAGE_TYPE=$(IMAGE_TYPE) \
+		--build-arg GO_TAGS="$(GO_TAGS)" \
+		--build-arg MAKEFLAGS="$(DOCKER_MAKEFLAGS)" \
+		--build-arg BUILD_TYPE=$(BUILD_TYPE) \
+		-t $(DOCKER_IMAGE)-cuda11 .
+
 docker-aio:
 	@echo "Building AIO image with base $(BASE_IMAGE) as $(DOCKER_AIO_IMAGE)"
 	docker build \