disable cache for better compability

app.py

@@ -108,6 +108,7 @@ def load_pipeline(model_name):
                 trust_remote_code=True,
                 torch_dtype=dtype,
                 device_map="auto",
+                use_cache=False,      # ← disable past-key-value caching
                 token=access_token)
             PIPELINES[model_name] = pipe
             return pipe