Make colab functional for llama.cpp

- Download only Q4_K_M for GGUF repositories by default
- Use maximum n-gpu-layers by default

Colab-TextGen-GPU.ipynb

@@ -115,7 +115,7 @@
         "    output_folder = \"\"\n",
         "\n",
         "# Start the web UI\n",
-        "cmd = f\"python server.py --share\"\n",
+        "cmd = f\"python server.py --share --n-gpu-layers 128\"\n",
         "if output_folder != \"\":\n",
         "    cmd += f\" --model {output_folder}\"\n",
         "cmd += f\" {command_line_flags}\"\n",