performance enhancement with torch autograd disabled

app.py

@@ -9,9 +9,10 @@ lac = LAC(mode="seg")
 
 def make_request(chinese_text):
   tokenizer.src_lang="zh"
-  encoded_zh = tokenizer(chinese_text, return_tensors="pt")
-  generated_tokens = model.generate(**encoded_zh, forced_bos_token_id=tokenizer.get_lang_id("en"))
-  return tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)
+  with torch.no_grad():
+      encoded_zh = tokenizer(chinese_text, return_tensors="pt")
+      generated_tokens = model.generate(**encoded_zh, forced_bos_token_id=tokenizer.get_lang_id("en"))
+      return tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)
 
 def generatepinyin(input):
   pinyin_list = pinyin(input)