Update app.py

app.py

@@ -1,30 +1,23 @@
-from transformers import TTSForConditionalGeneration, AutoModelForSeq2SeqLM, AutoModelForCausalLM
-from datasets import load_dataset
+from transformers import pipeline
 from gradio import Interface
 
+# Create a pipeline for text-to-speech
+tts = pipeline("text-to-speech", model="facebook/s2s-wav2vec2-ft")
 
-tts_model = TTSForConditionalGeneration.from_pretrained("facebook/s2s-wav2vec2-ft")
-whisper_model = AutoModelForSeq2SeqLM.from_pretrained("openai/whisper-medium")
-gpt2_xl_model = AutoModelForCausalLM.from_pretrained("openai-community/gpt2-xl")
+# Create a pipeline for speech-to-text
+stt = pipeline("automatic-speech-recognition", model="openai/whisper-medium")
 
+# Create a pipeline for text generation
+chat = pipeline("text-generation", model="gpt-3.5-turbo")
 
 def voice_chat(user_voice):
+    user_text = stt(user_voice)["text"]
     messages = [{"role": "system", "content": "You are a kind helpful assistant."}]
-    user_message = whisper_model.generate(user_voice)["generated_text"]
-    messages.append({"role": "user", "content": user_message})
-    print(messages)
-
-    # Utilize gpt2-xl locally for chat completion
-    with gpt2_xl_model.no_grad():
-        input_ids = gpt2_xl_model.tokenizer(messages, return_tensors="pt").input_ids
-        outputs = gpt2_xl_model.generate(input_ids, max_length=100)
-        chat_response = gpt2_xl_model.tokenizer.decode(outputs[0], skip_special_tokens=True)
-
-    reply = chat_response
-    messages.append({"role": "assistant", "content": reply})
-    audio = tts_model.generate(text=reply, return_tensors="pt").tolist()[0][0]
-    return reply, audio
-
+    messages.append({"role": "user", "content": user_text})
+    chat_reply = chat(messages=messages, max_length=100, top_p=0.95, temperature=0.7)[0]["generated_text"]
+    messages.append({"role": "assistant", "content": chat_reply})
+    audio = tts(chat_reply)["audio"]
+    return chat_reply, audio
 
 text_reply = Interface.Textbox(label="ChatGPT Text")
 voice_reply = Interface.Audio(type="audio/wav")