Create start.sh

start.sh

@@ -0,0 +1,27 @@
+#!/bin/bash
+
+# Default to CODER if no variable is set
+# Options: CODER, VISION
+MODE="${MODEL_TYPE:-CODER}"
+
+echo "Selected Mode: $MODE"
+
+if [ "$MODE" = "VISION" ]; then
+    echo "Loading Vision Model..."
+    MODEL_FILE="/app/qwen2.5-vl-7b-instruct-q4_k_m.gguf"
+    # Lower context for Vision to save RAM
+    CONTEXT=4096
+else
+    echo "Loading Coder Model..."
+    MODEL_FILE="/app/qwen2.5-coder-14b-instruct-q4_k_m.gguf"
+    # Higher context for Coding
+    CONTEXT=8192
+fi
+
+# Run the server
+exec llama-server \
+    -m "$MODEL_FILE" \
+    --host 0.0.0.0 \
+    --port 7860 \
+    -c $CONTEXT \
+    --n-gpu-layers 0