generate binary dependencies based on GOARCH on macos (#459)

llm/llama.cpp/generate.go

+//go:build !darwin
+// +build !darwin
+
 package llm
 
 //go:generate git submodule init

llm/llama.cpp/generate_darwin_amd64.go

+package llm
+
+//go:generate git submodule init
+//go:generate git submodule update --force ggml
+//go:generate git -C ggml apply ../ggml_patch/0001-add-detokenize-endpoint.patch
+//go:generate git -C ggml apply ../ggml_patch/0002-34B-model-support.patch
+//go:generate cmake --fresh -S ggml -B ggml/build/cpu -DLLAMA_ACCELERATE=on -DLLAMA_K_QUANTS=on -DCMAKE_SYSTEM_PROCESSOR=x86_64 -DCMAKE_OSX_ARCHITECTURES=x86_64
+//go:generate cmake --build ggml/build/cpu --target server --config Release

llm/llama.cpp/generate_darwin.go → llm/llama.cpp/generate_darwin_arm64.go

-//go:build darwin
-// +build darwin
-
 package llm
 
 //go:generate git submodule init
 //go:generate git submodule update --force ggml
 //go:generate git -C ggml apply ../ggml_patch/0001-add-detokenize-endpoint.patch
 //go:generate git -C ggml apply ../ggml_patch/0002-34B-model-support.patch
-//go:generate cmake -S ggml -B ggml/build/gpu -DLLAMA_METAL=on -DLLAMA_ACCELERATE=on -DLLAMA_K_QUANTS=on
+//go:generate cmake -S ggml -B ggml/build/gpu -DLLAMA_METAL=on -DLLAMA_ACCELERATE=on -DLLAMA_K_QUANTS=on -DCMAKE_SYSTEM_PROCESSOR=arm64 -DCMAKE_OSX_ARCHITECTURES=arm64
 //go:generate cmake --build ggml/build/gpu --target server --config Release

scripts/build_darwin.sh

@@ -6,8 +6,10 @@ GO_LDFLAGS="-X github.com/jmorganca/ollama/version.Version=$VERSION"
 GO_LDFLAGS="$GO_LDFLAGS -X github.com/jmorganca/ollama/server.mode=release"
 
 # build universal binary
-CGO_ENABLED=1 GOARCH=arm64 go build -ldflags "$GO_LDFLAGS" -o dist/ollama-darwin-arm64
-CGO_ENABLED=1 GOARCH=amd64 go build -ldflags "$GO_LDFLAGS" -o dist/ollama-darwin-amd64
+GOARCH=arm64 go generate ./...
+GOARCH=arm64 go build -ldflags "$GO_LDFLAGS" -o dist/ollama-darwin-arm64
+GOARCH=amd64 go generate ./...
+GOARCH=amd64 go build -ldflags "$GO_LDFLAGS" -o dist/ollama-darwin-amd64
 lipo -create -output dist/ollama dist/ollama-darwin-arm64 dist/ollama-darwin-amd64
 rm dist/ollama-darwin-amd64 dist/ollama-darwin-arm64
 codesign --deep --force --options=runtime --sign "$APPLE_IDENTITY" --timestamp dist/ollama