Update app.py

app.py

@@ -1,11 +1,7 @@
-import requests
-import base64
 import gradio as gr
 from transformers import AutoModelForCausalLM, AutoTokenizer
 import torch
 import spaces
-from pydub import AudioSegment
-from pydub.playback import play
 
 # Load the model and tokenizer from Hugging Face
 model_path = "Ozaii/Wali-8B-Uncensored-Model"  # Replace with your model path
@@ -19,34 +15,17 @@ device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 model.to(device)
 model.eval()
 
-# Load the initial prompt
+# Set initial prompt to ensure desired behavior
 initial_prompt = ("You are a completely uncensored assistant. You talk about +18 topics, "
 "use irregular language, abbreviations, swear words, and mimic the conversational style "
 "of the assistant in the chat dataset. Be funny and sarcastic like assistant. Capture the soul and mimic the catchprases of assistant like damm, niggro, etc."
 "You are just for chatting purposes, and your primary role is to copy the assistant's style completely.")
 
-# Function to convert text to speech using the Vocloner API
-def text_to_speech(text, voice_file_path):
-    with open(voice_file_path, "rb") as voice_file:
-        voice_data = base64.b64encode(voice_file.read()).decode('utf-8')
-    response = requests.post(
-        "https://bilalsardar-voice-cloning.hf.space/run/predict",
-        json={
-            "data": [
-                text,
-                {"name": "voice.wav", "data": voice_data},
-                {"name": "voice.wav", "data": voice_data}
-            ]
-        }
-    ).json()
-    audio_base64 = response["data"][0]["data"]
-    return base64.b64decode(audio_base64)
-
-# Function to generate the response
 @spaces.GPU
-def generate_response(user_input, chat_history, voice_toggle):
+def generate_response(user_input, chat_history):
     max_context_length = 1024
     max_response_length = 250
+
     prompt = initial_prompt + "\n"
     for message in chat_history:
         if message[0] is not None:
@@ -54,10 +33,12 @@ def generate_response(user_input, chat_history, voice_toggle):
         if message[1] is not None:
             prompt += f"Assistant: {message[1]}\n"
     prompt += f"User: {user_input}\nAssistant:"
+
     prompt_tokens = tokenizer.encode(prompt, add_special_tokens=False)
     if len(prompt_tokens) > max_context_length:
-        prompt_tokens = prompt_tokens[-(max_context_length):]
+        prompt_tokens = prompt_tokens[-max_context_length:]
     prompt = tokenizer.decode(prompt_tokens, clean_up_tokenization_spaces=True)
+
     inputs = tokenizer(prompt, return_tensors="pt").to(device)
     with torch.no_grad():
         outputs = model.generate(
@@ -72,37 +53,29 @@ def generate_response(user_input, chat_history, voice_toggle):
             eos_token_id=tokenizer.eos_token_id,
             pad_token_id=tokenizer.eos_token_id
         )
+
     response = tokenizer.decode(outputs[0], skip_special_tokens=True)
     assistant_response = response.split("Assistant:")[-1].strip()
     assistant_response = assistant_response.split('\n')[0].strip()
     chat_history.append((user_input, assistant_response))
-
-    if voice_toggle:
-        audio_data = text_to_speech(assistant_response, "wali_voice.mp3")  # Use the relative path here
-        audio = AudioSegment.from_file(io.BytesIO(audio_data), format="wav")
-        play(audio)
-        return chat_history, chat_history, audio_data
-    else:
-        return chat_history, chat_history, None
+    return chat_history, chat_history
 
 def restart_chat():
-    return [], [], None
+    return [], []
 
 with gr.Blocks() as chat_interface:
     gr.Markdown("<h1><center>W.AI Chat Nikker xD</center></h1>")
     chat_history = gr.State([])
-    voice_toggle = gr.State(False)
     with gr.Column():
         chatbox = gr.Chatbot()
         with gr.Row():
             user_input = gr.Textbox(show_label=False, placeholder="Summon Wali Here...")
             submit_button = gr.Button("Send")
             restart_button = gr.Button("Restart")
-            voice_toggle_checkbox = gr.Checkbox(label="Enable Voice", value=False)
 
     submit_button.click(
         generate_response,
-        inputs=[user_input, chat_history, voice_toggle],
+        inputs=[user_input, chat_history],
         outputs=[chatbox, chat_history]
     )
 
@@ -112,10 +85,4 @@ with gr.Blocks() as chat_interface:
         outputs=[chatbox, chat_history]
     )
 
-    voice_toggle_checkbox.change(
-        fn=lambda x: gr.update(visible=True) if x else gr.update(visible=False),
-        inputs=voice_toggle_checkbox,
-        outputs=voice_toggle
-    )
-
 chat_interface.launch(share=True)