Update app.py

app.py

@@ -12,7 +12,6 @@ model, vis_processors, _ = load_model_and_preprocess(
     is_eval=True,
     device=device
 )
-model=model.float()
 
 def infer(image, prompt, min_len, max_len, beam_size, len_penalty, repetition_penalty, top_p, decoding_method):
         use_nucleus_sampling = decoding_method == "Nucleus sampling"
@@ -32,6 +31,7 @@ def infer(image, prompt, min_len, max_len, beam_size, len_penalty, repetition_pe
             min_length=min_len,
             top_p=top_p,
             use_nucleus_sampling=use_nucleus_sampling,
+            dtype=torch.float16
         )
 
         return output[0]