Update app.py

app.py

@@ -1,14 +1,30 @@
-from transformers import AutoTokenizer, AutoModelForCausalLM, AutoTokenizer
+
 import soundfile as sf
 import torch
-
 from datetime import datetime
 import random
 import time
+from datetime import datetime
+import whisper
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer, VitsModel
+import torch
 import numpy as np
 import os
 import argparse
 import gradio as gr
+from timeit import default_timer as timer
+import torch
+import numpy as np
+import pandas as pd
+import whisper
+
+
+# whisper_model = whisper.load_model("medium").to("cuda")
+tts_model = VitsModel.from_pretrained("facebook/mms-tts-pol")
+tts_model.to("cuda")
+print("TTS Loaded!")
+tokenizer_tss = AutoTokenizer.from_pretrained("facebook/mms-tts-pol")
 
 
 def save_to_txt(text_to_save):
@@ -21,7 +37,6 @@ def read_txt():
     return lines 
 
 
-
 ##### Chat z LLAMA ####
 ##### Chat z LLAMA ####
 ##### Chat z LLAMA ####
@@ -32,7 +47,7 @@ def _load_model_tokenizer():
     tokenizer = AutoTokenizer.from_pretrained(model_id,  trust_remote_code=True)
     model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto",trust_remote_code=True, fp16=True).eval()  
     return model, tokenizer
-model_llm, tokenizer_llm = _load_model_tokenizer()
+
 
 
 def postprocess(self, y):
@@ -82,7 +97,7 @@ def predict(_query, _chatbot, _task_history):
     _chatbot.append((_parse_text(_query), ""))
     full_response = ""
 
-    for response in model_llm.chat_stream(tokenizer_llm, _query, history=_task_history,system = "Jesteś assystentem AI. Odpowiadaj zawsze w języku poslkim"  ):
+    for response in model.chat_stream(tokenizer, _query, history=_task_history,system = "Jesteś assystentem AI. Odpowiadaj zawsze w języku poslkim"  ):
         _chatbot[-1] = (_parse_text(_query), _parse_text(response))
 
         yield _chatbot
@@ -106,8 +121,9 @@ def update_audio(text):
 
 def translate(audio):
     print("__Wysyłam nagranie do whisper!")
-    transcription = whisper_model.transcribe(audio, language="pl")
-    return transcription["text"]
+    # transcription = whisper_model.transcribe(audio, language="pl")
+    return "Co możesz powiedzieć o ING Banku Śląskim?"
+    # return transcription["text"]
 
 
 def predict(audio, _chatbot, _task_history):
@@ -118,7 +134,7 @@ def predict(audio, _chatbot, _task_history):
     _chatbot.append((_parse_text(_query), ""))
     full_response = "" 
 
-    for response in model_llm.chat_stream(tokenizer_llm,
+    for response in model.chat_stream(tokenizer,
                                       _query,
                                       history= _task_history,
                                       system = "Jesteś assystentem AI. Odpowiadaj zawsze w języku polskim. Odpowiadaj krótko."):
@@ -132,7 +148,6 @@ def predict(audio, _chatbot, _task_history):
     print("____full_response",full_response)
     audio_file = read_text(_parse_text(full_response))  # Generowanie audio
     return full_response
-    # return 'temp_file.wav'  # Zwrócenie ścieżki do pliku audio
 
 def regenerate(_chatbot, _task_history):
     if not _task_history:
@@ -158,18 +173,6 @@ with gr.Blocks() as chat_demo:
     submit_btn.click(predict, [query, chatbot, task_history], [chatbot], show_progress=True)
     submit_audio_btn.click(predict, [audio_upload, chatbot, task_history], [chatbot], show_progress=True).then(update_audio, chatbot, audio_output)
 
+chat_demo.queue().launch(share=True)
 
 
-
-
-##### Audio Gen ####
-##### Audio Gen ####
-##### Audio Gen ####
-
-
-##### Run Alll #######
-##### Run Alll #######
-##### Run Alll #######
-
-chat_demo.queue()
-chat_demo.launch()