Update app.py

app.py

@@ -49,6 +49,14 @@ llm_model = transformers.AutoModelForCausalLM.from_pretrained(
     trust_remote_code=True, 
 )
 
+generation_args = {
+    "max_new_tokens": 500,
+    "return_full_text": True,
+    "temperature": 0.0,
+    "do_sample": False,
+}
+
+
 tokenizer = transformers.AutoTokenizer.from_pretrained("microsoft/Phi-3-mini-128k-instruct")
 
 llm_pipe = transformers.pipeline(
@@ -238,7 +246,7 @@ def on_src_or_tgt_lang_change(src_lang_value, tgt_lang_value, pnc_value):
 
 def main(audio_filepath, src_lang, tgt_lang, pnc):
     translated = transcribe(audio_filepath, src_lang, tgt_lang, pnc)
-    answer = llm_pipe(translated)
+    answer = llm_pipe(translated, **generation_args)
     return answer