Upload app.py with huggingface_hub

app.py

@@ -269,7 +269,7 @@ def _llm_report(p_mean: float, per_model: list, net_saliency: dict | None = None
         from openai import OpenAI
         if _VLLM_URL:
             # Live AMD MI300X inference via vLLM
-            client = OpenAI(base_url=_VLLM_URL, api_key="not-required", timeout=60.0)
+            client = OpenAI(base_url=_VLLM_URL, api_key="not-required", timeout=5.0)
             model_id = _LLM_MODEL
         else:
             # Fallback: HF Inference API