CHORE: Specify xoscar version (#457)

setup.cfg

@@ -24,7 +24,7 @@ zip_safe = False
 include_package_data = True
 packages = find:
 install_requires =
-    xoscar
+    xoscar>=0.1.2
     xorbits
     gradio>=3.39.0
     click

xinference/model/llm/ggml/llamacpp.py

@@ -184,7 +184,15 @@ def _convert_ggml_to_gguf(self, model_path: str) -> str:
 
     def load(self):
         try:
+            import llama_cpp
             from llama_cpp import Llama
+
+            if llama_cpp.__version__ < "0.2.0":
+                raise ValueError(
+                    "The llama_cpp version must be greater than 0.2.0. "
+                    "Please upgrade your version via `pip install -U llama_cpp` or refer to "
+                    "https://github.com/abetlen/llama-cpp-python#installation-with-openblas--cublas--clblast--metal."
+                )
         except ImportError:
             error_message = "Failed to import module 'llama_cpp'"
             installation_guide = [