Update app.py

app.py

@@ -84,7 +84,7 @@ def predict(message, history):
     """
     batch = tokenizer(prompt, return_tensors='pt')
     with torch.cuda.amp.autocast():
-        output_tokens = model.generate(**batch, max_new_tokens=100)
+        output_tokens = model.generate(**batch, max_new_tokens=200)
     return tokenizer.decode(output_tokens[0], skip_special_tokens=True).replace(prompt,"")