import torch import gradio as gr from transformers import BertTokenizer, BertForSequenceClassification import matplotlib.pyplot as plt import numpy as np from wordcloud import WordCloud from collections import Counter import re import json import csv import tempfile from datetime import datetime import logging from functools import lru_cache, wraps from dataclasses import dataclass from typing import List, Dict, Optional, Tuple, Callable from contextlib import contextmanager import gc import pandas as pd @dataclass class Config: MAX_HISTORY_SIZE: int = 1000 BATCH_SIZE_LIMIT: int = 50 MAX_TEXT_LENGTH: int = 512 MIN_WORD_LENGTH: int = 2 CACHE_SIZE: int = 128 BATCH_PROCESSING_SIZE: int = 8 FIGURE_SIZE_SINGLE: Tuple[int, int] = (8, 5) FIGURE_SIZE_BATCH: Tuple[int, int] = (12, 8) WORDCLOUD_SIZE: Tuple[int, int] = (10, 5) THEMES = { 'default': {'pos': '#4ecdc4', 'neg': '#ff6b6b'}, 'ocean': {'pos': '#0077be', 'neg': '#ff6b35'}, 'forest': {'pos': '#228b22', 'neg': '#dc143c'}, 'sunset': {'pos': '#ff8c00', 'neg': '#8b0000'}, } STOP_WORDS = { 'the', 'a', 'an', 'and', 'or', 'but', 'in', 'on', 'at', 'to', 'for', 'of', 'with', 'by', 'is', 'are', 'was', 'were', 'be', 'been', 'have', 'has', 'had', 'will', 'would', 'could', 'should', } config = Config() logger = logging.getLogger(__name__) # --------------------------------------------------------------------------- # Decorators / Context Managers # --------------------------------------------------------------------------- def handle_errors(default_return=None): """Centralised error-handling decorator.""" def decorator(func: Callable) -> Callable: @wraps(func) def wrapper(*args, **kwargs): try: return func(*args, **kwargs) except Exception as e: logger.error(f"{func.__name__} failed: {e}") return default_return if default_return is not None else f"Error: {str(e)}" return wrapper return decorator @contextmanager def managed_figure(*args, **kwargs): """Context manager for matplotlib figures — prevents memory leaks.""" fig = plt.figure(*args, **kwargs) try: yield fig finally: plt.close(fig) gc.collect() class ThemeContext: def __init__(self, theme: str = 'default'): self.theme = theme self.colors = config.THEMES.get(theme, config.THEMES['default']) # --------------------------------------------------------------------------- # Model Manager (singleton, lazy-loaded) # --------------------------------------------------------------------------- class ModelManager: _instance = None _model = None _tokenizer = None _device = None def __new__(cls): if cls._instance is None: cls._instance = super().__new__(cls) return cls._instance @property def model(self): if self._model is None: self._load_model() return self._model @property def tokenizer(self): if self._tokenizer is None: self._load_model() return self._tokenizer @property def device(self): if self._device is None: self._device = torch.device("cuda" if torch.cuda.is_available() else "cpu") return self._device def _load_model(self): try: self._device = torch.device("cuda" if torch.cuda.is_available() else "cpu") self._tokenizer = BertTokenizer.from_pretrained("entropy25/sentimentanalysis") self._model = BertForSequenceClassification.from_pretrained( "entropy25/sentimentanalysis", output_attentions=True ) self._model.to(self._device) logger.info(f"Model loaded on {self._device}") except Exception as e: logger.error(f"Model loading failed: {e}") raise # --------------------------------------------------------------------------- # Text Processor # --------------------------------------------------------------------------- class TextProcessor: @staticmethod @lru_cache(maxsize=config.CACHE_SIZE) def clean_text(text: str) -> Tuple[str, ...]: words = re.findall(r'\b\w{3,}\b', text.lower()) return tuple(w for w in words if w not in config.STOP_WORDS) # --------------------------------------------------------------------------- # History Manager # --------------------------------------------------------------------------- class HistoryManager: def __init__(self): self._history: List[Dict] = [] def add(self, entry: Dict): self._history.append({**entry, 'timestamp': datetime.now().isoformat()}) if len(self._history) > config.MAX_HISTORY_SIZE: self._history = self._history[-config.MAX_HISTORY_SIZE:] def get_all(self) -> List[Dict]: return self._history.copy() def clear(self) -> int: count = len(self._history) self._history.clear() return count def size(self) -> int: return len(self._history) # --------------------------------------------------------------------------- # Sentiment Engine — attention-based keyword extraction (no LIME dependency) # --------------------------------------------------------------------------- class SentimentEngine: """Sentiment analysis with attention-based keyword extraction.""" def __init__(self): self.model_manager = ModelManager() # ------------------------------------------------------------------ # Prediction helpers # ------------------------------------------------------------------ def predict_proba(self, texts): if isinstance(texts, str): texts = [texts] inputs = self.model_manager.tokenizer( texts, return_tensors="pt", padding=True, truncation=True, max_length=config.MAX_TEXT_LENGTH, ).to(self.model_manager.device) with torch.no_grad(): outputs = self.model_manager.model(**inputs) probs = torch.nn.functional.softmax(outputs.logits, dim=-1).cpu().numpy() return probs # ------------------------------------------------------------------ # Keyword extraction via attention weights # ------------------------------------------------------------------ def extract_key_words_attention(self, text: str, top_k: int = 10) -> List[Tuple[str, float]]: """ Extract influential words by averaging attention weights from the last transformer layer across all heads, then mapping scores back to the original tokens. """ try: encoding = self.model_manager.tokenizer( text, return_tensors="pt", truncation=True, max_length=config.MAX_TEXT_LENGTH, ).to(self.model_manager.device) with torch.no_grad(): outputs = self.model_manager.model(**encoding, output_attentions=True) # attentions: tuple of (1, heads, seq, seq) per layer — use last layer last_layer_attn = outputs.attentions[-1][0] # (heads, seq, seq) mean_attn = last_layer_attn.mean(dim=0).cpu().numpy() # (seq, seq) # Aggregate attention received by each token (column sum) token_scores = mean_attn.sum(axis=0) tokens = self.model_manager.tokenizer.convert_ids_to_tokens( encoding['input_ids'][0].cpu().tolist() ) # Merge sub-word tokens and accumulate scores word_scores: Dict[str, float] = {} current_word = "" current_score = 0.0 for token, score in zip(tokens, token_scores): if token in ('[CLS]', '[SEP]', '[PAD]'): continue if token.startswith('##'): current_word += token[2:] current_score += float(score) else: if current_word and len(current_word) >= config.MIN_WORD_LENGTH: w = current_word.lower() if w not in config.STOP_WORDS: word_scores[w] = word_scores.get(w, 0) + current_score current_word = token current_score = float(score) # Flush the last word if current_word and len(current_word) >= config.MIN_WORD_LENGTH: w = current_word.lower() if w not in config.STOP_WORDS: word_scores[w] = word_scores.get(w, 0) + current_score # Normalise and rank if not word_scores: return [] max_score = max(word_scores.values()) normalised = [(w, s / max_score) for w, s in word_scores.items()] normalised.sort(key=lambda x: x[1], reverse=True) return normalised[:top_k] except Exception as e: logger.error(f"Attention-based keyword extraction failed: {e}") return [] # ------------------------------------------------------------------ # Heatmap # ------------------------------------------------------------------ def create_heatmap_html(self, text: str, word_scores: Dict[str, float]) -> str: words = text.split() html_parts = ['
'] max_score = max(abs(s) for s in word_scores.values()) if word_scores else 0 for word in words: clean_word = re.sub(r'[^\w]', '', word.lower()) score = word_scores.get(clean_word, 0) if score > 0 and max_score > 0: intensity = min(255, int(180 * score / max_score)) color = f"rgba(0, {intensity}, 0, 0.35)" else: color = "transparent" html_parts.append( f'{word} ' ) html_parts.append('
') return ''.join(html_parts) # ------------------------------------------------------------------ # Single analysis # ------------------------------------------------------------------ @handle_errors(default_return={ 'sentiment': 'Unknown', 'confidence': 0.0, 'key_words': [], 'heatmap_html': '', }) def analyze_single(self, text: str) -> Dict: if not text.strip(): raise ValueError("Empty text") probs = self.predict_proba([text])[0] sentiment = "Positive" if probs[1] > probs[0] else "Negative" key_words = self.extract_key_words_attention(text) word_scores_dict = dict(key_words) heatmap_html = self.create_heatmap_html(text, word_scores_dict) return { 'sentiment': sentiment, 'confidence': float(probs.max()), 'pos_prob': float(probs[1]), 'neg_prob': float(probs[0]), 'key_words': key_words, 'heatmap_html': heatmap_html, } # ------------------------------------------------------------------ # Batch analysis # ------------------------------------------------------------------ @handle_errors(default_return=[]) def analyze_batch(self, texts: List[str], progress_callback=None) -> List[Dict]: if len(texts) > config.BATCH_SIZE_LIMIT: texts = texts[:config.BATCH_SIZE_LIMIT] results = [] batch_size = config.BATCH_PROCESSING_SIZE for i in range(0, len(texts), batch_size): batch = texts[i:i + batch_size] if progress_callback: progress_callback((i + len(batch)) / len(texts)) inputs = self.model_manager.tokenizer( batch, return_tensors="pt", padding=True, truncation=True, max_length=config.MAX_TEXT_LENGTH, ).to(self.model_manager.device) with torch.no_grad(): outputs = self.model_manager.model(**inputs) probs = torch.nn.functional.softmax(outputs.logits, dim=-1).cpu().numpy() for text, prob in zip(batch, probs): sentiment = "Positive" if prob[1] > prob[0] else "Negative" key_words = self.extract_key_words_attention(text, top_k=5) results.append({ 'text': text[:50] + '...' if len(text) > 50 else text, 'full_text': text, 'sentiment': sentiment, 'confidence': float(prob.max()), 'pos_prob': float(prob[1]), 'neg_prob': float(prob[0]), 'key_words': key_words, }) return results # --------------------------------------------------------------------------- # Plot Factory # --------------------------------------------------------------------------- class PlotFactory: @staticmethod @handle_errors(default_return=None) def create_sentiment_bars(probs: np.ndarray, theme: ThemeContext) -> plt.Figure: with managed_figure(figsize=config.FIGURE_SIZE_SINGLE) as fig: ax = fig.add_subplot(111) labels = ["Negative", "Positive"] colors = [theme.colors['neg'], theme.colors['pos']] bars = ax.bar(labels, probs, color=colors, alpha=0.8) ax.set_title("Sentiment Probabilities", fontweight='bold') ax.set_ylabel("Probability") ax.set_ylim(0, 1) for bar, prob in zip(bars, probs): ax.text( bar.get_x() + bar.get_width() / 2., bar.get_height() + 0.02, f'{prob:.3f}', ha='center', va='bottom', fontweight='bold', ) fig.tight_layout() return fig @staticmethod @handle_errors(default_return=None) def create_confidence_gauge(confidence: float, sentiment: str, theme: ThemeContext) -> plt.Figure: with managed_figure(figsize=config.FIGURE_SIZE_SINGLE) as fig: ax = fig.add_subplot(111) theta = np.linspace(0, np.pi, 100) colors = [theme.colors['neg'] if i < 50 else theme.colors['pos'] for i in range(100)] for i in range(len(theta) - 1): ax.fill_between([theta[i], theta[i + 1]], [0, 0], [0.8, 0.8], color=colors[i], alpha=0.7) pos = np.pi * (0.5 + (0.4 if sentiment == 'Positive' else -0.4) * confidence) ax.plot([pos, pos], [0, 0.6], 'k-', linewidth=6) ax.plot(pos, 0.6, 'ko', markersize=10) ax.set_xlim(0, np.pi) ax.set_ylim(0, 1) ax.set_title(f'{sentiment} — Confidence: {confidence:.3f}', fontweight='bold') ax.set_xticks([0, np.pi / 2, np.pi]) ax.set_xticklabels(['Negative', 'Neutral', 'Positive']) ax.axis('off') fig.tight_layout() return fig @staticmethod @handle_errors(default_return=None) def create_keyword_chart( key_words: List[Tuple[str, float]], sentiment: str, theme: ThemeContext ) -> Optional[plt.Figure]: if not key_words: return None with managed_figure(figsize=config.FIGURE_SIZE_SINGLE) as fig: ax = fig.add_subplot(111) words = [w for w, _ in key_words] scores = [s for _, s in key_words] color = theme.colors['pos'] if sentiment == 'Positive' else theme.colors['neg'] bars = ax.barh(range(len(words)), scores, color=color, alpha=0.7) ax.set_yticks(range(len(words))) ax.set_yticklabels(words) ax.set_xlabel('Attention Weight') ax.set_title(f'Top Contributing Words ({sentiment})', fontweight='bold') for bar, score in zip(bars, scores): ax.text( bar.get_width() + 0.001, bar.get_y() + bar.get_height() / 2., f'{score:.3f}', ha='left', va='center', fontsize=9, ) ax.invert_yaxis() ax.grid(axis='x', alpha=0.3) fig.tight_layout() return fig @staticmethod @handle_errors(default_return=None) def create_wordcloud(text: str, sentiment: str, theme: ThemeContext) -> Optional[plt.Figure]: if len(text.split()) < 3: return None colormap = 'Greens' if sentiment == 'Positive' else 'Reds' wc = WordCloud( width=800, height=400, background_color='white', colormap=colormap, max_words=30, ).generate(text) with managed_figure(figsize=config.WORDCLOUD_SIZE) as fig: ax = fig.add_subplot(111) ax.imshow(wc, interpolation='bilinear') ax.axis('off') ax.set_title(f'{sentiment} Word Cloud', fontweight='bold') fig.tight_layout() return fig @staticmethod @handle_errors(default_return=None) def create_batch_analysis(results: List[Dict], theme: ThemeContext) -> plt.Figure: with managed_figure(figsize=config.FIGURE_SIZE_BATCH) as fig: gs = fig.add_gridspec(2, 2, hspace=0.3, wspace=0.3) # Sentiment distribution ax1 = fig.add_subplot(gs[0, 0]) sent_counts = Counter(r['sentiment'] for r in results) colors = [theme.colors['pos'], theme.colors['neg']] ax1.pie( sent_counts.values(), labels=sent_counts.keys(), autopct='%1.1f%%', colors=colors[:len(sent_counts)], ) ax1.set_title('Sentiment Distribution') # Confidence histogram ax2 = fig.add_subplot(gs[0, 1]) confs = [r['confidence'] for r in results] ax2.hist(confs, bins=8, alpha=0.7, color='skyblue', edgecolor='black') ax2.set_title('Confidence Distribution') ax2.set_xlabel('Confidence') # Sentiment over time ax3 = fig.add_subplot(gs[1, :]) pos_probs = [r['pos_prob'] for r in results] scatter_colors = [ theme.colors['pos'] if r['sentiment'] == 'Positive' else theme.colors['neg'] for r in results ] ax3.scatter(range(len(results)), pos_probs, c=scatter_colors, alpha=0.7, s=60) ax3.axhline(y=0.5, color='gray', linestyle='--', alpha=0.5) ax3.set_title('Sentiment Progression') ax3.set_xlabel('Review Index') ax3.set_ylabel('Positive Probability') return fig # --------------------------------------------------------------------------- # Data Handler # --------------------------------------------------------------------------- class DataHandler: @staticmethod @handle_errors(default_return=(None, "Export failed")) def export_data(data: List[Dict], format_type: str) -> Tuple[Optional[str], str]: if not data: return None, "No data to export" tmp = tempfile.NamedTemporaryFile( mode='w', delete=False, suffix=f'.{format_type}', encoding='utf-8' ) if format_type == 'csv': writer = csv.writer(tmp) writer.writerow([ 'Timestamp', 'Text', 'Sentiment', 'Confidence', 'Pos_Prob', 'Neg_Prob', 'Key_Words', ]) for entry in data: writer.writerow([ entry.get('timestamp', ''), entry.get('text', ''), entry.get('sentiment', ''), f"{entry.get('confidence', 0):.4f}", f"{entry.get('pos_prob', 0):.4f}", f"{entry.get('neg_prob', 0):.4f}", "|".join( f"{w}:{s:.3f}" for w, s in entry.get('key_words', []) ), ]) elif format_type == 'json': json.dump(data, tmp, indent=2, ensure_ascii=False) tmp.close() return tmp.name, f"Exported {len(data)} entries" @staticmethod @handle_errors(default_return="") def process_file(file) -> str: if not file: return "" file_path = file.name if file_path.endswith('.csv'): for encoding in ['utf-8', 'latin-1', 'cp1252', 'iso-8859-1']: try: df = pd.read_csv(file_path, encoding=encoding) text_columns = [] for col in df.columns: sample = df[col].dropna().head(10) if sample.empty: continue text_count = sum( 1 for v in sample if isinstance(v, str) and len(str(v).strip()) > 10 ) if text_count > len(sample) * 0.7: text_columns.append(col) col = text_columns[0] if text_columns else df.columns[0] reviews = [ r.strip() for r in df[col].dropna().astype(str).tolist() if len(r.strip()) > 10 and r.strip().lower() != 'nan' ] if reviews: logger.info(f"Read {len(reviews)} rows from CSV ({encoding})") return '\n'.join(reviews) except Exception as e: logger.warning(f"CSV read failed ({encoding}): {e}") # Fallback: plain text read try: with open(file_path, 'r', encoding='utf-8', errors='ignore') as f: lines = f.read().strip().split('\n') cleaned = [] for i, line in enumerate(lines): line = line.strip() if i == 0 and any(kw in line.lower() for kw in ('review', 'text', 'comment')): continue line = re.sub(r'^[",\']*|[",\']*$', '', line) if len(line) > 10: cleaned.append(line) if cleaned: return '\n'.join(cleaned) except Exception as e: logger.error(f"Fallback CSV read failed: {e}") return "Error: Could not read CSV file. Please check the file format and encoding." else: for encoding in ['utf-8', 'latin-1', 'cp1252']: try: with open(file_path, 'r', encoding=encoding) as f: content = f.read().strip() if content: return content except Exception: continue return "Error: Could not read text file. Please check the file encoding." # --------------------------------------------------------------------------- # Main Application # --------------------------------------------------------------------------- class SentimentApp: def __init__(self): self.engine = SentimentEngine() self.history = HistoryManager() self.data_handler = DataHandler() self.examples = [ ["While the film's visual effects were undeniably impressive, the story lacked emotional weight, and the pacing felt inconsistent throughout."], ["An extraordinary achievement in filmmaking — the direction was masterful, the script was sharp, and every performance added depth and realism."], ["Despite a promising start, the film quickly devolved into a series of clichés, with weak character development and an ending that felt rushed and unearned."], ["A beautifully crafted story with heartfelt moments and a soundtrack that perfectly captured the emotional tone of each scene."], ["The movie was far too long, with unnecessary subplots and dull dialogue that made it difficult to stay engaged until the end."], ] @handle_errors(default_return=("Please enter text", None, None, None, None, None)) def analyze_single(self, text: str, theme: str = 'default'): if not text.strip(): return "Please enter text", None, None, None, None, None result = self.engine.analyze_single(text) self.history.add({'text': text[:100], 'full_text': text, **result}) theme_ctx = ThemeContext(theme) probs = np.array([result['neg_prob'], result['pos_prob']]) prob_plot = PlotFactory.create_sentiment_bars(probs, theme_ctx) gauge_plot = PlotFactory.create_confidence_gauge(result['confidence'], result['sentiment'], theme_ctx) cloud_plot = PlotFactory.create_wordcloud(text, result['sentiment'], theme_ctx) keyword_plot = PlotFactory.create_keyword_chart(result['key_words'], result['sentiment'], theme_ctx) key_words_str = ", ".join(f"{w}({s:.3f})" for w, s in result['key_words'][:5]) result_text = ( f"Sentiment: {result['sentiment']} (Confidence: {result['confidence']:.3f})\n" f"Key Words: {key_words_str}" ) return result_text, prob_plot, gauge_plot, cloud_plot, keyword_plot, result['heatmap_html'] @handle_errors(default_return=None) def analyze_batch(self, reviews: str, progress=None): if not reviews.strip(): return None texts = [r.strip() for r in reviews.split('\n') if r.strip()] if len(texts) < 2: return None results = self.engine.analyze_batch(texts, progress) for result in results: self.history.add(result) return PlotFactory.create_batch_analysis(results, ThemeContext('default')) @handle_errors(default_return=(None, "No history available")) def plot_history(self, theme: str = 'default'): history = self.history.get_all() if len(history) < 2: return None, f"Need at least 2 analyses for trends. Current: {len(history)}" theme_ctx = ThemeContext(theme) with managed_figure(figsize=(12, 8)) as fig: gs = fig.add_gridspec(2, 1, hspace=0.3) indices = list(range(len(history))) pos_probs = [item['pos_prob'] for item in history] confs = [item['confidence'] for item in history] ax1 = fig.add_subplot(gs[0, 0]) colors = [theme_ctx.colors['pos'] if p > 0.5 else theme_ctx.colors['neg'] for p in pos_probs] ax1.scatter(indices, pos_probs, c=colors, alpha=0.7, s=60) ax1.plot(indices, pos_probs, alpha=0.5, linewidth=2) ax1.axhline(y=0.5, color='gray', linestyle='--', alpha=0.5) ax1.set_title('Sentiment History') ax1.set_ylabel('Positive Probability') ax1.grid(True, alpha=0.3) ax2 = fig.add_subplot(gs[1, 0]) ax2.bar(indices, confs, alpha=0.7, color='lightblue', edgecolor='navy') ax2.set_title('Confidence Over Time') ax2.set_xlabel('Analysis Number') ax2.set_ylabel('Confidence') ax2.grid(True, alpha=0.3) fig.tight_layout() return fig, f"History: {len(history)} analyses" # --------------------------------------------------------------------------- # Gradio Interface # --------------------------------------------------------------------------- def create_interface(): app = SentimentApp() with gr.Blocks(theme=gr.themes.Soft(), title="Movie Sentiment Analyzer") as demo: gr.Markdown("# 🎬 Movie Sentiment Analyzer") gr.Markdown("Sentiment analysis with attention-based keyword extraction and interactive visualizations.") with gr.Tab("Single Analysis"): with gr.Row(): with gr.Column(): text_input = gr.Textbox( label="Movie Review", placeholder="Enter your movie review...", lines=5, ) with gr.Row(): analyze_btn = gr.Button("Analyze", variant="primary") theme_selector = gr.Dropdown( choices=list(config.THEMES.keys()), value="default", label="Theme", ) gr.Examples(examples=app.examples, inputs=text_input) with gr.Column(): result_output = gr.Textbox(label="Result", lines=3) heatmap_output = gr.HTML(label="Word Importance Heatmap") with gr.Row(): prob_plot = gr.Plot(label="Probabilities") gauge_plot = gr.Plot(label="Confidence") with gr.Row(): wordcloud_plot = gr.Plot(label="Word Cloud") keyword_plot = gr.Plot(label="Key Contributing Words") with gr.Tab("Batch Analysis"): with gr.Row(): with gr.Column(): file_upload = gr.File(label="Upload File", file_types=[".csv", ".txt"]) batch_input = gr.Textbox(label="Reviews (one per line)", lines=8) with gr.Column(): load_btn = gr.Button("Load File") batch_btn = gr.Button("Analyze Batch", variant="primary") batch_plot = gr.Plot(label="Batch Results") with gr.Tab("History & Export"): with gr.Row(): refresh_btn = gr.Button("Refresh") clear_btn = gr.Button("Clear", variant="stop") status_btn = gr.Button("Status") with gr.Row(): csv_btn = gr.Button("Export CSV") json_btn = gr.Button("Export JSON") history_status = gr.Textbox(label="Status") history_plot = gr.Plot(label="History Trends") csv_file = gr.File(label="CSV Download", visible=True) json_file = gr.File(label="JSON Download", visible=True) # Event bindings analyze_btn.click( app.analyze_single, inputs=[text_input, theme_selector], outputs=[result_output, prob_plot, gauge_plot, wordcloud_plot, keyword_plot, heatmap_output], ) load_btn.click(app.data_handler.process_file, inputs=file_upload, outputs=batch_input) batch_btn.click(app.analyze_batch, inputs=batch_input, outputs=batch_plot) refresh_btn.click( lambda theme: app.plot_history(theme), inputs=theme_selector, outputs=[history_plot, history_status], ) clear_btn.click( lambda: f"Cleared {app.history.clear()} entries", outputs=history_status, ) status_btn.click( lambda: f"History: {app.history.size()} entries", outputs=history_status, ) csv_btn.click( lambda: app.data_handler.export_data(app.history.get_all(), 'csv'), outputs=[csv_file, history_status], ) json_btn.click( lambda: app.data_handler.export_data(app.history.get_all(), 'json'), outputs=[json_file, history_status], ) return demo # --------------------------------------------------------------------------- # Entry Point # --------------------------------------------------------------------------- if __name__ == "__main__": logging.basicConfig(level=logging.INFO) demo = create_interface() demo.launch(share=True)