Update app.py

app.py

@@ -3,8 +3,13 @@ import gradio as gr
 import json
 from transformers import pipeline
 
-# Load the translation pipeline
-text_translator = pipeline("translation", model="facebook/nllb-200-distilled-600M", torch_dtype=torch.bfloat16)
+# Load the translation pipeline with eager attention implementation
+text_translator = pipeline(
+    "translation", 
+    model="facebook/nllb-200-distilled-600M", 
+    torch_dtype=torch.bfloat16, 
+    attn_implementation="eager"
+)
 
 # Load the JSON data for language codes
 with open('language.json', 'r') as file:
@@ -40,24 +45,30 @@ def translate_text(text, destination_language):
     if dest_code is None:
         return f"Error: Could not find FLORES code for language {destination_language}"
     
-    translation = text_translator(text, src_lang="eng_Latn", tgt_lang=dest_code)
-    return translation[0]["translation_text"]
+    try:
+        translation = text_translator(text, src_lang="eng_Latn", tgt_lang=dest_code)
+        return translation[0]["translation_text"]
+    except Exception as e:
+        return f"Error during translation: {str(e)}"
 
 # Initialize the speech-to-text pipeline (Whisper model)
-# Use a pipeline as a high-level helper
-speech_to_text = pipeline("text-to-audio", model="facebook/musicgen-small")
+# Using the appropriate Whisper model for automatic speech recognition
+speech_to_text = pipeline("automatic-speech-recognition", model="openai/whisper-small")
 
 # Function to transcribe audio to text
 def transcribe_audio(audio_file, destination_language):
-    transcription_result = speech_to_text(audio_file)
-    print(f"Transcription result: {transcription_result}")  # Print the whole response to inspect
-    if "text" in transcription_result:
-        transcription = transcription_result["text"]
-    else:
-        return "Error: Unable to transcribe audio."
-    
-    return translate_text(transcription, destination_language)
-
+    try:
+        transcription_result = speech_to_text(audio_file)
+        print(f"Transcription result: {transcription_result}")  # Print the whole response to inspect
+        if "text" in transcription_result:
+            transcription = transcription_result["text"]
+        else:
+            return "Error: Unable to transcribe audio."
+        
+        # Translate the transcribed text
+        return translate_text(transcription, destination_language)
+    except Exception as e:
+        return f"Error during transcription: {str(e)}"
 
 # Gradio interface
 with gr.Blocks() as demo: