feat: serve

entrypoint.sh

@@ -2,4 +2,12 @@
 
 mkdir -p data
 
-HF_HUB_ENABLE_HF_TRANSFER=1 huggingface-cli download Qwen/Qwen2.5-1.5B-Instruct --local-dir ./data/Qwen2.5-1.5B-Instruct
+HF_MODEL_PATH=Qwen/Qwen2.5-1.5B-Instruct
+HF_MODEL_NAME=$(basename ${HF_MODEL_PATH})
+LOCAL_MODEL_PATH=./data/${HF_MODEL_NAME}
+
+# download model
+HF_HUB_ENABLE_HF_TRANSFER=1 huggingface-cli download ${HF_MODEL_PATH} --local-dir ${LOCAL_MODEL_PATH}
+
+# start vllm server
+${HOME}/.local/bin/vllm serve ${LOCAL_MODEL_PATH} --served-model-name ${HF_MODEL_NAME}