Update app.py

app.py

@@ -246,8 +246,8 @@ class Translators:
     def rosetta(self):
         model = AutoModelForCausalLM.from_pretrained(
             self.model_name,
-            dtype=torch.float32, # bfloat16
-            low_cpu_mem_usage=True,
+            dtype=torch.bfloat16, # float32 slow
+            low_cpu_mem_usage=False, # True
             device_map="auto")
         tokenizer = AutoTokenizer.from_pretrained(self.model_name)
         context = {"tone": "Formal and precise"}        
@@ -257,7 +257,6 @@ class Translators:
             {"role": "user", "content": self.input_text},
         ]     
         prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
-        print(prompt)
         inputs = tokenizer(prompt, return_tensors="pt").to(self.device)
         input_length = inputs["input_ids"].shape[1]       
         model.eval()