chore: decrease max len

model.py

@@ -123,7 +123,7 @@ class SALMONN(nn.Module):
         prompt,
         prompt_pattern="USER: <Speech><SpeechHere></Speech> {}\nASSISTANT:",
         device='cuda:0',
-        max_length=200,
+        max_length=150,
         num_beams=4,
         do_sample=True,
         min_length=1,