Create Dockerfile

Dockerfile

@@ -0,0 +1,14 @@
+FROM ghcr.io/ggerganov/llama.cpp:server
+
+# HF Spaces 要求用 7860 端口
+EXPOSE 7860
+
+# 下载模型并启动 llama-server
+# 用 -hf 参数直接从 HF Hub 拉取，无需预先下载
+CMD ["--hf-repo", "bartowski/Qwen2.5-3B-Instruct-GGUF", \
+     "--hf-file", "Qwen2.5-3B-Instruct-Q4_K_M.gguf", \
+     "--host", "0.0.0.0", \
+     "--port", "7860", \
+     "--ctx-size", "4096", \
+     "--threads", "4", \
+     "--n-predict", "512"]