Update app.py

app.py

@@ -28,7 +28,15 @@ def predict(text):
     # generate and end generate if <|endoftext|> is not in text
     gen_tokens = model.generate(
       tokens, do_sample=True, temperature=0.8, max_new_tokens=64, top_k=50, top_p=0.8,
-      no_repeat_ngram_size=3, repetition_penalty=1.2, bad_words_ids=[[11066]], eos_token_id=tokenizer.eos_token_id
+      no_repeat_ngram_size=3, repetition_penalty=1.2,
+      bad_words_ids=[
+        tokenizer.encode('...'), 
+        tokenizer.encode('....'),  
+        tokenizer.encode('(중략)'),
+        tokenizer.encode('http')
+      ],
+      eos_token_id=tokenizer.eos_token_id,
+      pad_token_id=tokenizer.pad_token_id
     )
     generated = tokenizer.batch_decode(gen_tokens)[0]
     return generated