update test

app.py

@@ -28,12 +28,11 @@ def pipe(file, return_timestamps=False):
     asr = pipeline(
         task="automatic-speech-recognition",
         model=MODEL_NAME,
-        chunk_length_s=30,
+        chunk_length_s=26,
         device=device,
         token=auth_token,
-        num_beams=5,
         torch_dtype=torch.float16,
-        model_kwargs={"attn_implementation": "flash_attention_2"} if FLASH_ATTENTION else {"attn_implementation": "sdpa"},
+        model_kwargs={"attn_implementation": "flash_attention_2", "num_beams": 5} if FLASH_ATTENTION else {"attn_implementation": "sdpa", "num_beams": 5},
     )
     asr.model.config.forced_decoder_ids = asr.tokenizer.get_decoder_prompt_ids(
         language=lang,