A high-performance inference engine for AI models
#!/usr/bin/env bash
set -euo pipefail
PROJECT_DIR="$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")" && pwd)"
BUILD_DIR="$PROJECT_DIR/build/release-server"
BIN_DIR="$BUILD_DIR/bin"
PARALLEL_JOBS="$(getconf _NPROCESSORS_ONLN 2>/dev/null || printf '1\n')"
cmake \
--preset release \
-S "$PROJECT_DIR" \
-B "$BUILD_DIR" \
-DCMAKE_RUNTIME_OUTPUT_DIRECTORY="$BIN_DIR" \
-DLLAMA_BUILD_TOOLS=ON \
-DLLAMA_BUILD_SERVER=ON >&2
cmake --build "$BUILD_DIR" \
--config Release \
--target llama-server \
--parallel "$PARALLEL_JOBS" >&2
exec "$BIN_DIR/llama-server" "$@"