Update app.py

updated order of execution

app.py

@@ -1,9 +1,24 @@
 import os
-import openai
 import torch
 from transformers import pipeline
 
-# The LLM Model
+# Loading the TTS and Vocoder ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+from transformers import SpeechT5Processor, SpeechT5ForTextToSpeech, SpeechT5HifiGan
+from datasets import load_dataset
+
+processor = SpeechT5Processor.from_pretrained("microsoft/speecht5_tts")
+model_default = SpeechT5ForTextToSpeech.from_pretrained("microsoft/speecht5_tts")
+vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan")
+
+# sending the model to device
+model_default.to(device)
+vocoder.to(device)
+
+# Loading speaker embedings
+embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
+speaker_embeddings = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze(0)
+
+# The LLM Model ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 from huggingface_hub import HfFolder
 from openai import OpenAI
 
@@ -12,41 +27,12 @@ if api_key is None:
     raise ValueError("API_KEY is not set in the environment variables.")
 print("API key successfully loaded.")
 
-
 # Initialize OpenAI client for Hugging Face Inference Endpoint
 client = OpenAI(
 	base_url="https://f2iozzwigntrzkve.us-east-1.aws.endpoints.huggingface.cloud/v1/",
 	api_key=api_key
 )
 
-
-# def generate_llm_response(text, model_id="ccibeekeoc42/Llama3.1-8b-base-SFT-2024-11-09"):
-#     """Generates LLM response for given text with streaming support"""
-#     full_response = []
-
-#     # Create streaming response
-#     chat_completion = client.chat.completions.create(
-# 	      model="tgi",
-# 	      messages=[
-#            {"role": "system", "content": "You are a BRIEF AND DIRECT assistant. A part of a speech pipeline so keep your responces short, fluent, and straight to the point. Avoid markdown in responses"},
-#            {"role": "user", "content": text}
-#            ],
-# 	      top_p=None,
-# 	      temperature=None,
-# 	      max_tokens=75,
-# 	      stream=True,
-# 	      seed=None,
-# 	      stop=None,
-# 	      frequency_penalty=None,
-# 	      presence_penalty=None
-#       )
-#     # Collect streamed response chunks
-#     for chunk in chat_completion:
-#         if chunk.choices[0].delta.content:
-#             full_response.append(chunk.choices[0].delta.content)
-
-#     return "".join(full_response)
-
 def generate_llm_response(text, model_id="ccibeekeoc42/Llama3.1-8b-base-SFT-2024-11-09"):
     full_response = []
     try:
@@ -79,36 +65,17 @@ def generate_llm_response(text, model_id="ccibeekeoc42/Llama3.1-8b-base-SFT-2024
 generate_llm_response("Explain Deep Learning in Igbo")
 
 
-# Loading the ST Model (Whisper)
+# Loading the ST Model (Whisper) ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 device = "cuda:0" if torch.cuda.is_available() else "cpu"
 pipe = pipeline("automatic-speech-recognition", model="okezieowen/whisper-small-multilingual-naija-11-03-2024", device=device)
 
-# Loading the TTS and Vocoder
-from transformers import SpeechT5Processor, SpeechT5ForTextToSpeech, SpeechT5HifiGan
-from datasets import load_dataset
-
-processor = SpeechT5Processor.from_pretrained("microsoft/speecht5_tts")
-
-model_default = SpeechT5ForTextToSpeech.from_pretrained("microsoft/speecht5_tts")
-model = SpeechT5ForTextToSpeech.from_pretrained("ccibeekeoc42/speecht5_finetuned_naija_ig_yo_2025-01-20_O2")
-vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan")
-
-# sending the model to device
-model_default.to(device)
-model.to(device)
-vocoder.to(device)
-
-# Loading speaker embedings
-embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
-speaker_embeddings = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze(0)
-
 # Take audio and return translated text
 def transcribe(audio):
     outputs = pipe(audio, max_new_tokens=256, generate_kwargs={"task": "transcribe"})
     return outputs["text"]
 
 
-# Helper Functions to Cleanup LLM Texts
+# Helper Functions to Cleanup LLM Texts ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 # Replacement rules
 import re
 # Language-specific replacements