Create app.py

app.py

@@ -0,0 +1,70 @@
+from transformers import pipeline, T5ForConditionalGeneration, T5Tokenizer
+import torch
+import gradio as gr
+
+# Проверка доступности GPU
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
+# Инициализация модели для распознавания речи (ASR)
+asr_pipeline = pipeline("automatic-speech-recognition", model="openai/whisper-medium", device=0 if device.type == "cuda" else -1)
+
+# Инициализация модели для суммаризации
+summarizer = pipeline("summarization", model="facebook/bart-large-cnn", device=0 if device.type == "cuda" else -1)
+
+# Инициализация модели для перевода
+model_translation = T5ForConditionalGeneration.from_pretrained('utrobinmv/t5_translate_en_ru_zh_small_1024')
+model_translation.to(device)
+tokenizer_translation = T5Tokenizer.from_pretrained('utrobinmv/t5_translate_en_ru_zh_small_1024')
+
+def transcribe_audio(audio_file):
+    # Преобразование аудио в текст
+    result = asr_pipeline(audio_file)
+    return result["text"]
+
+def summarize(text, max_length=300, min_length=150):
+    # Суммаризация текста
+    result = summarizer(text, max_length=max_length, min_length=min_length, do_sample=False)
+    return result[0]['summary_text']
+
+def translate(text):
+    # Перевод текста на русский
+    prefix = 'translate to ru: '
+    src_text = prefix + text
+
+    input_ids = tokenizer_translation(src_text, return_tensors="pt")
+    generated_tokens = model_translation.generate(**input_ids.to(device))
+
+    result = tokenizer_translation.batch_decode(generated_tokens, skip_special_tokens=True)
+    return result[0]
+
+def process_audio(audio_file, language):
+    if not audio_file:
+        return "No input provided."
+    
+    # Преобразование аудио в текст
+    transcribed_text = transcribe_audio(audio_file)
+    print(f"Transcribed Text: {transcribed_text}")
+    
+    # Суммаризация текста
+    summary = summarize(transcribed_text)
+    print(f"Summary: {summary}")
+    
+    # Перевод, если выбран русский язык
+    if language == "rus":
+        return translate(summary)
+    return summary
+
+# Создание Gradio интерфейса
+with gr.Blocks() as demo:
+    gr.Markdown("# Audio Summarizer and Translator")
+    gr.Markdown("Upload an audio file to summarize and translate (if needed).")
+
+    language = gr.Radio(choices=["rus", "eng"], label="Output Language", value="rus")
+    audio_input = gr.Audio(label="Upload Audio File", type="filepath")
+    text_output = gr.Textbox(label="Processed Text", lines=10)
+    process_button = gr.Button("Process Audio")
+
+    process_button.click(process_audio, inputs=[audio_input, language], outputs=text_output)
+
+# Запуск приложения
+demo.launch(debug=True)