feat(模型加载): 添加sync_to_temp选项支持临时目录加载

添加sync_to_temp参数控制是否将模型文件复制到临时目录后再加载

Dockerfile

@@ -0,0 +1,29 @@
+FROM corex:4.3.8 as builder
+ADD . /app
+WORKDIR /app
+RUN sed -i 's/-x cuda/-x ivcore/g' /usr/local/share/cmake-3.25/Modules/Compiler/Clang-CUDA.cmake &&\
+    cmake -S . -B build \
+    -DGGML_CUDA=ON \
+    -DLLAMA_CURL=OFF \
+    -DCMAKE_CUDA_ARCHITECTURES=ivcore11 \
+    -DCMAKE_CUDA_FLAGS="-x ivcore -std=c++17" \
+    -DCMAKE_CUDA_STANDARD=17 \
+    -DCMAKE_CXX_STANDARD=17 \
+    -DGGML_CUDA_FA=OFF \
+    -DBUILD_SHARED_LIBS=OFF \
+    -DGGML_CUDA_FORCE_CUBLAS=ON \
+    -DGGML_CPU=ON &&\
+    cmake --build build --config Release -j \
+    --target llama-server llama-cli
+
+
+
+FROM ubuntu
+WORKDIR /app
+COPY --from=builder /usr/local/corex/ /usr/local/corex/
+COPY --from=builder /usr/local/openmpi/lib/ /usr/local/openmpi/lib/
+COPY --from=builder /app/build/bin/llama-server /app/llama-server
+COPY --from=builder /app/build/bin/llama-cli /app/llama-cli
+RUN apt update && apt install -y libgomp1
+ENV PATH=/usr/local/corex/bin:/usr/local/openmpi/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
+ENV LD_LIBRARY_PATH=/usr/local/corex/lib64:/usr/local/openmpi/lib:/usr/local/lib