debug

app.py

@@ -12,20 +12,22 @@ asr = pipeline(
     feature_extractor=processor.feature_extractor, decoder=processor.decoder
 )
 def run_asr(audio):
-    sr, audio_array = audio
-    audio_array = audio_array.astype(np.float32)
-    if len(audio_array.shape) > 1:
-        if audio_array.shape[1] == 1:
-            audio_array = audio_array.squeeze()
-        elif audio_array.shape[1] == 2:
-            audio_array = to_mono(audio_array.T)
-        else:
-            raise ValueError("Audio with > 2 channels not supported")
-    if sr != 16_000:
-        audio_array = resample(audio_array, orig_sr=sr, target_sr=16_000)
-    res = asr(audio_array, chunk_length_s=20, stride_length_s=2)
-    
-    return res["text"]
+    try:
+        sr, audio_array = audio
+        audio_array = audio_array.astype(np.float32)
+        if len(audio_array.shape) > 1:
+            if audio_array.shape[1] == 1:
+                audio_array = audio_array.squeeze()
+            elif audio_array.shape[1] == 2:
+                audio_array = to_mono(audio_array.T)
+            else:
+                raise ValueError("Audio with > 2 channels not supported")
+        if sr != 16_000:
+            audio_array = resample(audio_array, orig_sr=sr, target_sr=16_000)
+        res = asr(audio_array, chunk_length_s=20, stride_length_s=2)["text"]
+    except Exception as e:
+        res = e
+    return res
 
 text_out = gr.outputs.Textbox(label="transcript")
 interface = gr.Interface(
@@ -38,4 +40,4 @@ interface = gr.Interface(
     flagging_options=["incorrect"]
 )
 
-interface.launch()
+interface.launch(debug=True)