diff --git a/scripts/convert_gguf.sh b/scripts/convert_gguf.sh
new file mode 100644
index 0000000..c8b8a00
--- /dev/null
+++ b/scripts/convert_gguf.sh
@@ -0,0 +1,25 @@
+#!/bin/bash
+
+MODEL_NAME=$1
+echo $MODEL_NAME
+
+git lfs install
+
+export CUDA_VISIBLE_DEVICES=-1
+
+REPO_URL=https://huggingface.co/yuiseki/$MODEL_NAME
+echo $REPO_URL
+
+status_code=$(curl --write-out %{http_code} --silent --output /dev/null $REPO_URL)
+
+if [[ "$status_code" -ne 200 ]] ; then
+  echo "!!! Model: ${MODEL_NAME} NOT FOUUND !!!"
+  exit 1
+fi
+
+REPO_PATH=git@hf.co:yuiseki/$MODEL_NAME.git
+
+
+git clone $REPO_PATH
+python3 ~/llama.cpp/convert.py $MODEL_NAME
+~/llama.cpp/quantize $MODEL_NAME/ggml-model-f16.gguf $MODEL_NAME/$MODEL_NAME-Q4_K_M.gguf Q4_K_M