reverted changes

app.py

@@ -1,38 +1,14 @@
-import subprocess
-import os
-import sys
 import spaces
 from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
 import torch
-import numpy as np
-import soundfile as sf
-from huggingface_hub import hf_hub_download
 import gradio as gr
 from PIL import Image
+import os
 import random
 import re
-
-subprocess.run(["pip", "install", "-U", "scipy", "soundfile", "torch", "numpy", "gradio", "huggingface_hub", "Cython"])
-# Clone VITS if not already present
-if not os.path.exists("vits"):
-    subprocess.run(["git", "clone", "https://github.com/jaywalnut310/vits.git"])
-
-# Add VITS directory to the Python path
-sys.path.append("vits")
-
-monotonic_align_path = "vits/monotonic_align"
-if not os.path.exists(f"{monotonic_align_path}/monotonic_align/core.so"):
-    print("🔧 Compiling monotonic_align...")
-    subprocess.run(["python3", "setup.py", "build_ext", "--inplace"], cwd=monotonic_align_path, check=True)
-
-# Import VITS modules
-from vits.models import SynthesizerTrn
-from vits.utils import load_checkpoint
-from vits.text import text_to_sequence
-from vits.text.symbols import symbols  # Ensure symbols are imported for text processing
-
-# Clear unused GPU memory
-torch.cuda.empty_cache()
+import subprocess
+torch.cuda.empty_cache()  # Clears unused GPU memory
+torch.cuda.memory_allocated()  # Checks available GPU memory
 
 # Define the model name
 OUTPUT_DIR = "output"
@@ -40,14 +16,6 @@ os.makedirs(OUTPUT_DIR, exist_ok=True)
 
 model_name = "TheBloke/Amethyst-13B-Mistral-AWQ"
 
-# Load the TTS model
-model_path = hf_hub_download(repo_id="Lycoris53/Vits-TTS-Japanese-Only-Sakura-Miko", filename="G_SakuraMiko.pth")
-device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-
-# Initialize VITS model
-model = SynthesizerTrn(256, 512, 1024, 32, 5).to(device)
-load_checkpoint(model, model_path)  # Corrected argument order
-
 # Load the tokenizer
 tokenizer = AutoTokenizer.from_pretrained(model_name)
 
@@ -214,19 +182,6 @@ def remove_emojis(text):
 
 concerned_streak = 0  
 
-def generate_speech(text):
-    text_norm = text_to_sequence(text)
-    text_tensor = torch.LongTensor(text_norm).unsqueeze(0).to(device)
-    
-    with torch.no_grad():
-        audio = model.infer(text_tensor)
-    
-    # Save as WAV
-    wav_path = "output.wav"
-    sf.write(wav_path, np.array(audio.cpu().detach()), samplerate=22050)
-    
-    return wav_path
-
 @spaces.GPU
 def chat(input_text):
     global conversation_history, current_emotion, previous_emotion, concerned_streak
@@ -352,24 +307,13 @@ Rena:"""
     # ✅ Handle fallback if response is empty
     if not response.strip():
         response = "Hmm, I’m not sure how to respond to that. Can you try rephrasing?"
-    audio_file = generate_speech(response)
+
     # ✅ Add Rena's response to the conversation history
     conversation_history.append(f"Rena: {response}")
 
-    return response, avatar_image, audio_file
-
-def generate_speech(text):
-    text_norm = text_to_sequence(text)
-    text_tensor = torch.LongTensor(text_norm).unsqueeze(0).to(device)
-
-    with torch.no_grad():
-        audio = model.infer(text_tensor)
+    return response, avatar_image
 
-    # Save as WAV
-    wav_path = "output.wav"
-    sf.write(wav_path, np.array(audio.cpu().detach()), samplerate=22050)
 
-    return wav_path
 
 
 
@@ -397,14 +341,13 @@ with gr.Blocks(css=css) as interface:
     with gr.Row():
         user_input = gr.Textbox(label="Your Message", lines=2, interactive=True)
         rena_response = gr.Textbox(label="Rena's Response", lines=10, interactive=False)
-        tts_output = gr.Audio(label="Rena's Voice", interactive=False)  # Added TTS output
     
     # Add event to handle `Enter` key press
-    user_input.submit(chat, inputs=[user_input], outputs=[rena_response, avatar, tts_output])
+    user_input.submit(chat, inputs=[user_input], outputs=[rena_response, avatar])
 
     # Submit button (optional)
     submit_button = gr.Button("Submit")
-    submit_button.click(chat, inputs=[user_input], outputs=[rena_response, avatar, tts_output])
+    submit_button.click(chat, inputs=[user_input], outputs=[rena_response, avatar])
 
 # Launch the app
 interface.launch()

requirements.txt

@@ -8,11 +8,7 @@ bitsandbytes
 autoawq
 transformers>=4.37.0
 triton
-torchaudio
-numpy
-soundfile
-matplotlib
-scipy
+