Upload folder using huggingface_hub

server.py

@@ -77,7 +77,7 @@ def run_inference(text: str) -> str:
     with torch.no_grad():
         outputs = model.generate(
             **inputs,
-            max_new_tokens=512,
+            max_new_tokens=256,
             temperature=0.7,
             do_sample=True,
             top_p=0.9,