Update app.py

app.py

@@ -40,13 +40,14 @@ def transcribe(audio):
         device=device,
     )
     
-    # Create an attention mask
-    attention_mask = np.ones(audio_np.shape, dtype=np.int64)
+    # prediction = pipe(audio_np)["text"]
+
+    ds = load_dataset("hf-internal-testing/librispeech_asr_dummy", "clean", split="validation")
+    sample = ds[0]["audio"]
     
-    # Use EncoderDecoderCache for past_key_values
-    past_key_values = EncoderDecoderCache.from_legacy_cache(None)
+    prediction = pipe(sample.copy(), batch_size=8)["text"]
+    print(prediction)
     
-    prediction = pipe(audio_np, attention_mask=attention_mask, past_key_values=past_key_values)["text"]
     return prediction
     
 gradio_app = gr.Interface(