Update app.py

app.py

@@ -684,6 +684,107 @@ def convert_model_to_phoenix(model_url, use_hierarchical=True, gpu_type="L40S"):
         return None, f"❌ Conversion failed: {str(e)}"
 
 
+def generate_text_phoenix(
+    model_url, use_hierarchical, convert_attention, 
+    prompt, max_new_tokens, temperature
+):
+    """PHOENIX로 텍스트 생성"""
+    try:
+        if not convert_attention or not model_url.strip():
+            return "⚠️ Enable 'Attention Replace' and provide model URL", ""
+        
+        # 1. 모델 변환
+        model_info, msg = convert_model_to_phoenix(model_url, use_hierarchical, "L40S")
+        
+        if model_info is None:
+            return msg, ""
+        
+        model = model_info['model']
+        
+        # 2. Tokenizer 로드
+        try:
+            tokenizer = AutoTokenizer.from_pretrained(model_url, trust_remote_code=True)
+        except Exception as e:
+            return f"❌ Tokenizer load failed: {e}", ""
+        
+        # 3. 입력 토크나이즈
+        inputs = tokenizer(prompt, return_tensors="pt").to(DEVICE)
+        input_ids = inputs["input_ids"]
+        
+        print(f"\n📝 Generating text...")
+        print(f"   Prompt: {prompt}")
+        print(f"   Input tokens: {input_ids.shape[1]}")
+        print(f"   Max new tokens: {max_new_tokens}")
+        
+        # 4. 생성
+        start_time = time.time()
+        generated_ids = []
+        
+        with torch.no_grad():
+            for _ in range(max_new_tokens):
+                # Forward pass
+                outputs = model(input_ids=input_ids)
+                logits = outputs.logits[:, -1, :]
+                
+                # Temperature sampling
+                if temperature > 0:
+                    probs = F.softmax(logits / temperature, dim=-1)
+                    next_token = torch.multinomial(probs, num_samples=1)
+                else:
+                    next_token = logits.argmax(dim=-1, keepdim=True)
+                
+                # Append
+                generated_ids.append(next_token.item())
+                input_ids = torch.cat([input_ids, next_token], dim=1)
+                
+                # Stop at EOS
+                if next_token.item() == tokenizer.eos_token_id:
+                    break
+        
+        elapsed = time.time() - start_time
+        
+        # 5. 디코드
+        generated_text = tokenizer.decode(generated_ids, skip_special_tokens=True)
+        full_text = prompt + generated_text
+        
+        # 6. 결과
+        output_md = f"""
+## 📝 Generated Text
+
+**Prompt**: {prompt}
+
+---
+
+**Generated**:
+
+{generated_text}
+
+---
+
+**Full Text**:
+
+{full_text}
+"""
+        
+        stats_md = f"""
+## 📊 Generation Statistics
+
+- **Input tokens**: {input_ids.shape[1] - len(generated_ids)}
+- **Generated tokens**: {len(generated_ids)}
+- **Total tokens**: {input_ids.shape[1]}
+- **Time**: {elapsed:.2f}s
+- **Speed**: {len(generated_ids) / elapsed:.1f} tokens/s
+- **Temperature**: {temperature}
+- **Model**: PHOENIX Retention (O(n))
+"""
+        
+        return output_md, stats_md
+        
+    except Exception as e:
+        import traceback
+        return f"❌ Generation failed:\n```\n{traceback.format_exc()}\n```", ""
+
+
 def run_phoenix_experiment(model_url, use_hierarchical, convert_attention, sequence_length, gpu_type):
     """Run PHOENIX experiment"""
     try:
@@ -871,6 +972,42 @@ with gr.Blocks(
                             [convert_url, convert_hierarchical, convert_gpu],
                             [gr.State(), convert_output])
         
+        with gr.Tab("💬 Text Generation (NEW!)"):
+            gr.Markdown("""
+            ### PHOENIX 텍스트 생성
+            
+            변환된 모델로 실제 텍스트를 생성합니다.
+            """)
+            
+            with gr.Row():
+                with gr.Column(scale=1):
+                    gen_model_url = gr.Textbox(label="🔗 Model URL", value=DEFAULT_MODEL)
+                    gen_hierarchical = gr.Checkbox(value=True, label="Hierarchical")
+                    gen_convert = gr.Checkbox(value=True, label="Enable Conversion")
+                    
+                    gen_prompt = gr.Textbox(
+                        label="📝 Input Prompt",
+                        placeholder="Enter your prompt here...",
+                        lines=3,
+                        value="The future of AI is"
+                    )
+                    
+                    gen_max_tokens = gr.Slider(16, 256, 64, step=16, label="Max New Tokens")
+                    gen_temperature = gr.Slider(0.1, 2.0, 0.7, step=0.1, label="Temperature")
+                    
+                    gen_btn = gr.Button("🚀 Generate Text", variant="primary")
+                
+                with gr.Column(scale=2):
+                    gen_output = gr.Markdown(label="Generated Text")
+                    gen_stats = gr.Markdown(label="Statistics")
+            
+            gen_btn.click(
+                fn=generate_text_phoenix,
+                inputs=[gen_model_url, gen_hierarchical, gen_convert, gen_prompt, 
+                       gen_max_tokens, gen_temperature],
+                outputs=[gen_output, gen_stats]
+            )
+        
         with gr.Tab("🧪 Experiment"):
             with gr.Row():
                 with gr.Column(scale=1):