Update app.py

app.py

@@ -74,13 +74,17 @@ def run_on_cpu(func):
 
 @run_on_cpu
 def process_input(audio_file=None):
+    full_message = ""
     for partial_message in process_audio(audio_file):
-        yield partial_message
+        full_message = partial_message  # Always use the latest partial message
+    return full_message
 
 @run_on_cpu
 def process_transcribe_input(audio_file=None):
+    full_message = ""
     for partial_message in process_audio(audio_file, transcript=True):
-        yield partial_message
+        full_message = partial_message  # Always use the latest partial message
+    return full_message
 
 class StopOnTokens(StoppingCriteria):
     def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor, **kwargs) -> bool:
@@ -142,9 +146,9 @@ examples.extend(good_examples)
 examples.extend(bad_examples)
 
 with gr.Blocks() as iface:
-    gr.Markdown("# Llama3.1-S: checkpoint Aug 19, 2024")
+    gr.Markdown("# Llama3.2s Mini: checkpoint September 26, 2024")
     gr.Markdown("Enter text to convert to audio, then submit the audio to generate text or Upload Audio")
-    gr.Markdown("Powered by [Homebrew Ltd](https://homebrew.ltd/) | [Read our blog post](https://homebrew.ltd/blog/llama3-just-got-ears)")
+    gr.Markdown("Inspired by [Homebrew Ltd](https://homebrew.ltd/) | [Read our blog post](https://homebrew.ltd/blog/llama3-just-got-ears)")
 
     with gr.Row():
         input_type = gr.Radio(["text", "audio"], label="Input Type", value="audio")