Create app.py

app.py

@@ -0,0 +1,132 @@
+import torch
+import transformers
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from PIL import Image
+import warnings
+import gradio as gr
+import os
+
+# Disable warnings for cleaner output
+transformers.logging.set_verbosity_error()
+transformers.logging.disable_progress_bar()
+warnings.filterwarnings('ignore')
+
+# Set device - will use CUDA if available, otherwise CPU
+device = 'cuda' if torch.cuda.is_available() else 'cpu'
+torch.set_default_device(device)
+
+# Model configuration
+model_name = 'qnguyen3/nanoLLaVA-1.5'
+
+print(f"Loading model {model_name} on {device}...")
+
+# Create model
+model = AutoModelForCausalLM.from_pretrained(
+    model_name,
+    torch_dtype=torch.float16,
+    device_map='auto',
+    trust_remote_code=True)
+tokenizer = AutoTokenizer.from_pretrained(
+    model_name,
+    trust_remote_code=True)
+
+print("Model loaded successfully!")
+
+def analyze_character(image_path, analysis_type):
+    """
+    Analyze a character image for dramaturgical insights
+    
+    Args:
+        image_path: Path to the character image
+        analysis_type: Type of character analysis to perform
+    
+    Returns:
+        str: The generated character analysis
+    """
+    # Load and process image
+    try:
+        image = Image.open(image_path).convert('RGB')
+        image_tensor = model.process_images([image], model.config).to(dtype=model.dtype)
+    except Exception as e:
+        return f"Error processing image: {str(e)}"
+    
+    # Create prompt based on analysis type
+    if analysis_type == "full_analysis":
+        prompt = ("Analyze this character as a dramaturg would. Describe their appearance, " 
+                 "potential personality traits, character archetype, suitable roles, and how they might " 
+                 "function within a dramatic narrative. Consider costume, posture, expression, and visual symbolism.")
+    elif analysis_type == "archetype":
+        prompt = ("Identify the potential character archetype(s) represented in this image. " 
+                 "Consider both classical archetypes (hero, mentor, trickster, etc.) and modern " 
+                 "interpretations. Explain your reasoning based on visual cues.")
+    elif analysis_type == "historical_context":
+        prompt = ("Analyze this character's appearance in terms of historical context. " 
+                 "Identify the likely time period, cultural influences, and how these elements " 
+                 "would influence the character's role in a dramatic work. Consider costume details, " 
+                 "props, and stylistic elements.")
+    else:
+        prompt = "Describe this character in detail for dramatic casting purposes."
+    
+    # Format input for the model using ChatML format
+    messages = [
+        {"role": "system", "content": "You are an expert dramaturg with deep knowledge of character analysis, theatrical traditions, and visual storytelling."},
+        {"role": "user", "content": f'<image>\n{prompt}'}
+    ]
+    
+    text = tokenizer.apply_chat_template(
+        messages,
+        tokenize=False,
+        add_generation_prompt=True
+    )
+    
+    # Split text around image placeholder
+    text_chunks = [tokenizer(chunk).input_ids for chunk in text.split('<image>')]
+    input_ids = torch.tensor(text_chunks[0] + [-200] + text_chunks[1], dtype=torch.long).unsqueeze(0)
+    
+    # Generate response
+    try:
+        output_ids = model.generate(
+            input_ids,
+            images=image_tensor,
+            max_new_tokens=1024,
+            temperature=0.7,
+            top_p=0.9,
+            use_cache=True)[0]
+        
+        response = tokenizer.decode(output_ids[input_ids.shape[1]:], skip_special_tokens=True).strip()
+        return response
+    except Exception as e:
+        return f"Error generating analysis: {str(e)}"
+
+# Create Gradio interface
+def create_ui():
+    with gr.Blocks(title="Dramaturg Character Analyzer") as demo:
+        gr.Markdown("# Dramaturg Character Analyzer")
+        gr.Markdown("Upload a character image to receive a dramaturgical analysis")
+        
+        with gr.Row():
+            with gr.Column():
+                input_image = gr.Image(type="filepath", label="Upload Character Image")
+                analysis_type = gr.Radio(
+                    ["full_analysis", "archetype", "historical_context", "basic_description"],
+                    label="Analysis Type",
+                    value="full_analysis"
+                )
+                analyze_btn = gr.Button("Analyze Character")
+            
+            with gr.Column():
+                output_text = gr.Textbox(label="Character Analysis", lines=20)
+        
+        analyze_btn.click(
+            fn=analyze_character,
+            inputs=[input_image, analysis_type],
+            outputs=output_text
+        )
+        
+    return demo
+
+# Main function
+if __name__ == "__main__":
+    demo = create_ui()
+    demo.launch(share=True)
+    print("Dramaturg Character Analyzer is now running!")