Update app.py

app.py

@@ -17,6 +17,7 @@ print(f"CUDA version: {torch.version.cuda}")
 print(f"Python version: {platform.python_version()}")
 print(f"Pytorch version: {torch.__version__}")
 print(f"Gradio version: {gr. __version__}")
+print(f"HFhub version: {huggingface_hub.__version__}")
 
 
 """
@@ -70,7 +71,7 @@ def respond(message, history: list[tuple[str, str]], model, system_message, max_
     print(model)
     model_name = choose_model(model)
 
-    client = InferenceClient(model_name, token=os.getenv('deepseekv2'), provider="replicate")
+    client = InferenceClient(model_name, provider="sambanova", token=os.getenv('deepseekv2'), )
     
     messages = [{"role": "system", "content": system_message}]