Update app.py

app.py

@@ -17,15 +17,22 @@ feature_extractor = AutoFeatureExtractor.from_pretrained(
 )
 sampling_rate = feature_extractor.sampling_rate
 
-asr = pipeline(
-    "automatic-speech-recognition", model="anuragshas/wav2vec2-xls-r-1b-hi-with-lm"
-)
+asr = pipeline("automatic-speech-recognition", model="anuragshas/wav2vec2-xls-r-1b-hi-with-lm")
+
+
+
+model = AutoModelForSeq2SeqLM.from_pretrained('hackathon-pln-es/t5-small-spanish-nahuatl')
+tokenizer = AutoTokenizer.from_pretrained('hackathon-pln-es/t5-small-spanish-nahuatl')
 
 
 def predict_and_ctc_lm_decode(input_file):
     speech = load_and_fix_data(input_file, sampling_rate)
     transcribed_text = asr(speech, chunk_length_s=5, stride_length_s=1)
-    return transcribed_text["text"]
+    transcribed_text = transcribed_text["text"]
+    input_ids = tokenizer('translate Spanish to Nahuatl: ' + transcribed_text, return_tensors='pt').input_ids
+    outputs = model.generate(input_ids, max_length=512)
+    outputs = tokenizer.batch_decode(outputs, skip_special_tokens=True)[0]
+    return outputs
 
 
 gr.Interface(
@@ -35,9 +42,8 @@ gr.Interface(
     ],
     outputs=[gr.outputs.Textbox()],
     examples=[["example1.wav"]],
-    title="Hindi ASR using Wav2Vec2-1B with LM",
-    article="<p><center><img src='https://visitor-badge.glitch.me/badge?page_id=anuragshas/Hindi_ASR' alt='visitor badge'></center></p>",
-    description="Built during Robust Speech Event",
+    title="Spanish-Audio-Transcriptions-to-Nahuatl-Translation",
+    article="<p><center><img src='........e'></center></p>",
     layout="horizontal",
     theme="huggingface",
 ).launch(enable_queue=True, cache_examples=True)