Update app.py

app.py

@@ -1,174 +1,72 @@
 # ruff: noqa: E402
-# Correct import order for Hugging Face Spaces
+
+# 1. Force spaces to initialize first
 try:
-    import spaces  # Must be first import!
+    import spaces  # Must be first!
     USING_SPACES = True
 except ImportError:
     USING_SPACES = False
 
-# Suppress warnings
+# 2. Suppress all non-critical warnings
 import warnings
-warnings.filterwarnings("ignore", message="Can't initialize NVML")
-warnings.filterwarnings("ignore", category=UserWarning, module="thinc")
+warnings.filterwarnings("ignore", category=UserWarning)
+warnings.filterwarnings("ignore", message=".*GLIBCXX.*")
+warnings.filterwarnings("ignore", module="thinc")
 
-# Configure device
-if not USING_SPACES:
-    import torch
-    from torch.cuda import is_available as cuda_is_available
-    device = "cuda" if cuda_is_available() else "cpu"
-else:
-    device = "cpu"
+# 3. Configure device before any other imports
+import os
+os.environ["CUDA_VISIBLE_DEVICES"] = ""  # Force CPU-only
 
-# Rest of imports
+# 4. Import remaining packages
 import json
 import re
 import tempfile
 import hashlib
 import spacy
-from collections import defaultdict
 from pathlib import Path
-from importlib.resources import files
 from pydub import AudioSegment, silence
 from spacy.matcher import DependencyMatcher
-from typing import List, Dict
 
-import click
 import gradio as gr
 import numpy as np
 import soundfile as sf
 import torchaudio
-from cached_path import cached_path
-from transformers import AutoModelForCausalLM, AutoTokenizer
 
-# ========== CORE NLP COMPONENTS ========== #
+# 5. Load spaCy model
 nlp = spacy.load("en_core_web_trf")
 
+# 6. Simplified Dialogue Parser
 class DialogueParser:
     def __init__(self):
         self.matcher = DependencyMatcher(nlp.vocab)
-        self._add_patterns()
-        
-    def _add_patterns(self):
-        said_pattern = [
-            {"RIGHT_ID": "verb", "RIGHT_ATTRS": {"POS": "VERB", "LEMMA": {"IN": ["say", "ask", "reply"]}}},
+        pattern = [
+            {"RIGHT_ID": "verb", "RIGHT_ATTRS": {"POS": "VERB", "LEMMA": {"IN": ["say", "ask", "reply"]}},
             {"LEFT_ID": "verb", "REL_OP": ">", "RIGHT_ID": "speaker", "RIGHT_ATTRS": {"DEP": "nsubj"}}
         ]
-        self.matcher.add("SAID_CONSTRUCTION", [said_pattern])
-
-    def parse(self, text: str) -> List[Dict]:
-        doc = nlp(text)
-        matches = self.matcher(doc)
-        dialogues = []
-        
-        for match_id, token_ids in matches:
-            if nlp.vocab.strings[match_id] == "SAID_CONSTRUCTION":
-                speaker = doc[token_ids[1]].text
-                quote = self._extract_quote(doc[token_ids[0]])
-                dialogues.append({"speaker": speaker, "text": quote})
-        
-        return dialogues
-
-    def _extract_quote(self, verb_token):
-        return " ".join([t.text for t in verb_token.children if t.dep_ == "ccomp"])
-
-# ========== FIXED GRADIO UI ========== #
-class VoiceBank:
-    def __init__(self):
-        self.voices = {}
-        self.speaker_map = {}
-
-    def add_voice(self, speaker_name: str, audio_path: str):
-        voice_id = hashlib.md5(Path(audio_path).read_bytes()).hexdigest()
-        self.voices[voice_id] = self._preprocess_audio(audio_path)
-
-    def _preprocess_audio(self, audio_path: str):
-        audio, sr = torchaudio.load(audio_path)
-        if sr != 24000:
-            resampler = torchaudio.transforms.Resample(sr, 24000)
-            audio = resampler(audio)
-        return audio.numpy()
-
-class AudiobookGenerator:
-    def __init__(self, voice_bank: VoiceBank):
-        self.voice_bank = voice_bank
-        self.sample_rate = 24000
-
-    def generate_audiobook(self, text: str, progress=gr.Progress()):
-        parser = DialogueParser()
-        dialogues = parser.parse(text)
-        audio_segments = []
-        
-        for dialogue in progress.tqdm(dialogues):
-            audio = self._generate_segment(dialogue["text"])
-            audio_segments.append(audio)
-        
-        return (self.sample_rate, np.concatenate(audio_segments))
-
-    def _generate_segment(self, text: str):
-        # Implement your TTS logic here
-        return np.random.rand(16000)  # Placeholder
+        self.matcher.add("DIALOGUE", [pattern])
 
-voice_bank = VoiceBank()
-audiobook_gen = AudiobookGenerator(voice_bank)
+    def parse(self, text):
+        return [{"speaker": "Narrator", "text": text}]  # Simplified for demo
 
-# ========== CORRECTED GRADIO UI SECTION ========== #
-with gr.Blocks() as app_audiobook:
+# 7. Gradio UI
+with gr.Blocks() as app:
     gr.Markdown("# Audiobook Generator")
     
     with gr.Row():
-        txt_input = gr.File(label="Upload Text File", file_types=[".txt"])
-        analyze_btn = gr.Button("Analyze Text", variant="secondary")
-    
-    detected_speakers = gr.State()
-    with gr.Column(visible=False) as speaker_ui:
-        gr.Markdown("## Assign Voices to Speakers")
-        speaker_container = gr.Row()
+        txt_input = gr.File(label="Upload Text", file_types=[".txt"])
+        generate_btn = gr.Button("Generate", variant="primary")
     
-    generate_btn = gr.Button("Generate Audiobook", variant="primary", visible=False)
-    audio_output = gr.Audio(label="Generated Audiobook", autoplay=True)
+    audio_output = gr.Audio(label="Result")
 
-    def analyze_text(file):
-        with open(file.name, "r") as f:
-            text = f.read()
-        
-        parser = DialogueParser()
-        dialogues = parser.parse(text)
-        speakers = list({d["speaker"] for d in dialogues if d["speaker"] != "Narrator"})
-        
-        components = []
-        for speaker in speakers:
-            components.append(gr.Audio(label=f"Voice for {speaker}", type="filepath"))
-        
-        return [
-            gr.update(visible=True),  # speaker_ui visibility
-            gr.Row.update(components=components),  # Update speaker_container
-            gr.update(visible=True),  # generate_btn visibility
-            speakers  # Store in detected_speakers state
-        ]
+    def generate(file):
+        return (24000, np.random.rand(16000))  # Replace with TTS
 
-    analyze_btn.click(
-        analyze_text,
-        inputs=txt_input,
-        outputs=[speaker_ui, speaker_container, generate_btn, detected_speakers]
-    )
-    
     generate_btn.click(
-        lambda text: (24000, np.random.rand(16000)),  # Replace with actual TTS
+        generate,
         inputs=txt_input,
         outputs=audio_output
     )
-# ========== APP LAUNCH ========== #
-with gr.Blocks() as app:
-    gr.TabbedInterface(
-        [app_audiobook],
-        ["Audiobook"]
-    )
-
-@click.command()
-@click.option("--port", "-p", default=7860)
-@click.option("--host", "-H", default="0.0.0.0")
-def main(port, host):
-    app.queue().launch(server_port=port, server_name=host)
 
+# 8. Hugging Face Spaces entry point
 if __name__ == "__main__":
-    main()
+    app.launch(server_name="0.0.0.0", server_port=7860)