Sync from upstream llama.cpp repository

examples/model-conversion/scripts/utils/run-embedding-server.sh

@@ -0,0 +1,22 @@
+#!/usr/bin/env bash
+
+set -e
+#
+# First try command line argument, then environment variable, then file
+CONVERTED_MODEL="${1:-"$CONVERTED_MODEL"}"
+
+# Final check if we have a model path
+if [ -z "$CONVERTED_MODEL" ]; then
+    echo "Error: Model path must be provided either as:" >&2
+    echo "  1. Command line argument" >&2
+    echo "  2. CONVERTED_MODEL environment variable" >&2
+    exit 1
+fi
+
+echo $CONVERTED_MODEL
+
+cmake --build ../../build --target llama-server
+
+../../build/bin/llama-server -m $CONVERTED_MODEL \
+    --embedding \
+    --pooling none