Build cuda image

Signed-off-by: Liora Milbaum <[email protected]>

.github/workflows/model_servers.yaml

@@ -32,12 +32,6 @@ jobs:
             directory: llamacpp_python
             platforms: linux/amd64,linux/arm64
             no_gpu: 1
-          #- image_name: llamacpp_python_vulkan
-          #  model: mistral
-          #  flavor: vulkan
-          #  directory: llamacpp_python
-          #  platforms: linux/arm64
-          #  vulkan: 1
           - image_name: llamacpp_python_cuda
             model: mistral
             flavor: cuda
@@ -100,13 +94,6 @@ jobs:
         if: ${{ matrix.no_gpu }}
         run: make test REGISTRY=${{ env.REGISTRY }} IMAGE_NAME=${{ env.REGISTRY_ORG }}/${{ matrix.image_name}}:latest
 
-      # - name: Run cuda test # we dont have cuda tests
-      #   working-directory: ./model_servers/${{ matrix.directory }}/
-      #   if: ${{ matrix.cuda }}
-      #   run: make test-cuda
-      #   env:
-      #     IMAGE_NAME: ${{ matrix.image_name }}
-
       - name: Login to Container Registry
         if: github.event_name == 'push' && github.ref == 'refs/heads/main'
         uses: redhat-actions/[email protected]
@@ -116,7 +103,6 @@ jobs:
           password: ${{ secrets.GITHUB_TOKEN }}
 
       - name: Push image
-        id: push_image
         if: github.event_name == 'push' && github.ref == 'refs/heads/main'
         uses: redhat-actions/[email protected]
         with:

model_servers/llamacpp_python/Makefile

@@ -22,7 +22,23 @@ build-cuda:
 build-vulkan:
 	podman build --squash-all -t $(VULKAN_IMAGE) . -f vulkan/Containerfile
 
-.PHONY: download-model-mistral # default model
-download-model-mistral:
+.PHONY: download-model-tiny-llama
+download-model-tiny-llama:
+	curl -H "Cache-Control: no-cache" -s -S -L -f $(TINY_LLAMA_MODEL_URL) -z $(RELATIVE_MODELS_PATH)/$(LLAMA_MODEL_NAME) -o $(RELATIVE_MODELS_PATH)/$(LLAMA_MODEL_NAME).tmp && mv -f $(RELATIVE_MODELS_PATH)/$(LLAMA_MODEL_NAME).tmp $(RELATIVE_MODELS_PATH)/$(LLAMA_MODEL_NAME) 2>/dev/null || rm -f $(RELATIVE_MODELS_PATH)/$(LLAMA_MODEL_NAME).tmp $(RELATIVE_MODELS_PATH)/$(LLAMA_MODEL_NAME)
+
+.PHONY: download-model-llama
+download-model-llama:
+	curl -H "Cache-Control: no-cache" -s -S -L -f $(LLAMA_MODEL_URL) -z $(RELATIVE_MODELS_PATH)/$(TINY_LLAMA_MODEL_NAME) -o $(RELATIVE_MODELS_PATH)/$(TINY_LLAMA_MODEL_NAME).tmp && mv -f $(RELATIVE_MODELS_PATH)/$(TINY_LLAMA_MODEL_NAME).tmp $(RELATIVE_MODELS_PATH)/$(TINY_LLAMA_MODEL_NAME) 2>/dev/null || rm -f $(RELATIVE_MODELS_PATH)/$(TINY_LLAMA_MODEL_NAME).tmp $(RELATIVE_MODELS_PATH)/$(TINY_LLAMA_MODEL_NAME)
+
+.PHONY: mistral
+mistral:
+	curl -H "Cache-Control: no-cache" -s -S -L -f $(MISTRAL_MODEL_URL) -z $(RELATIVE_MODELS_PATH)/$(MISTRAL_MODEL_NAME) -o $(RELATIVE_MODELS_PATH)/$(MISTRAL_MODEL_NAME).tmp && mv -f $(RELATIVE_MODELS_PATH)/$(MISTRAL_MODEL_NAME).tmp $(RELATIVE_MODELS_PATH)/$(MISTRAL_MODEL_NAME) 2>/dev/null || rm -f $(RELATIVE_MODELS_PATH)/$(MISTRAL_MODEL_NAME).tmp $(RELATIVE_MODELS_PATH)/$(MISTRAL_MODEL_NAME)
+
+.PHONY: install
+install:
+	pip install -r tests/requirements.txt
+
+.PHONY: run
+run:
 	cd ../../models && \
 	make MODEL_NAME=mistral-7b-instruct-v0.1.Q4_K_M.gguf MODEL_URL=https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.1-GGUF/resolve/main/mistral-7b-instruct-v0.1.Q4_K_M.gguf -f  Makefile download-model

model_servers/llamacpp_python/cuda/Containerfile

@@ -1,8 +1,7 @@
 FROM quay.io/opendatahub/workbench-images:cuda-ubi9-python-3.9-20231206
 WORKDIR /locallm
 COPY src .
-RUN pip install --upgrade pip
 ENV CMAKE_ARGS="-DLLAMA_CUBLAS=on"
 ENV FORCE_CMAKE=1
-RUN pip install --no-cache-dir --upgrade -r /locallm/requirements.txt
-ENTRYPOINT [ "sh", "run.sh" ]
+RUN pip install --no-cache-dir -r ./requirements.txt
+ENTRYPOINT [ "sh", "run.sh" ]

model_servers/llamacpp_python/tests/conftest.py

@@ -32,7 +32,7 @@
         PORT = 8001
 
 MS = pytest_container.Container(
-        url=f"containers-storage:{REGISTRY}/{IMAGE_NAME}",
+        url=f"containers-storage:{os.environ['REGISTRY']}/containers/{os.environ['IMAGE_NAME']}:{os.environ['IMAGE_TAG']}",
         volume_mounts=[
             pytest_container.container.BindMount(
                 container_path="{MODEL_PATH}/{MODEL_NAME}".format(MODEL_PATH=MODEL_PATH, MODEL_NAME=MODEL_NAME),