Update app.py

app.py

@@ -160,56 +160,27 @@ async def load_model():
         model = model.to(device)
         model.eval()
         
+        # Patch model config for compatibility with newer transformers
+        if hasattr(model, 'config'):
+            # Ensure _output_attentions exists (fixes AttributeError)
+            if not hasattr(model.config, '_output_attentions'):
+                setattr(model.config, '_output_attentions', False)
+                
+            # Ensure output_attentions property uses the attribute
+            if not hasattr(model.config, 'output_attentions'):
+                model.config.output_attentions = False
+                
+            # Ensure other common missing attributes
+            if not hasattr(model.config, 'return_dict'):
+                model.config.return_dict = True
+                
         logger.info("✅ PlasmidGPT model loaded successfully!")
         
     except Exception as e:
-        logger.error(f"Failed to load model: {str(e)}")
-        logger.error(f"Error type: {type(e).__name__}")
-        import traceback
-        logger.error(traceback.format_exc())
-        raise
-@app.get("/", response_model=HealthResponse)
-async def root():
-    """Health check endpoint."""
-    return HealthResponse(
-        status="healthy" if model is not None else "loading",
-        model_loaded=model is not None,
-        device=device,
-        model_name="lingxusb/PlasmidGPT"
-    )
-@app.get("/health", response_model=HealthResponse)
-async def health():
-    """Health check endpoint."""
-    return HealthResponse(
-        status="healthy" if model is not None else "loading",
-        model_loaded=model is not None,
-        device=device,
-        model_name="lingxusb/PlasmidGPT"
-    )
-@app.post("/generate", response_model=GenerationResponse)
-async def generate_sequences(request: GenerationRequest):
-    """
-    Generate DNA sequences using PlasmidGPT.
-    
-    Args:
-        request: Generation parameters
-        
-    Returns:
-        Generated sequences with metadata
-    """
-    if model is None or tokenizer is None:
-        raise HTTPException(
-            status_code=503,
-            detail="Model is still loading. Please wait and try again."
-        )
-    
-    try:
-        start_time = time.time()
-        
-        # Tokenize input using custom tokenizer
-        # Custom tokenizer uses encode() method (returns list, not tensor)
+# ... (skip to generate_sequences)
         encoded = tokenizer.encode(request.prompt)
         input_ids = torch.tensor([encoded.ids], dtype=torch.long).to(device)
+        attention_mask = torch.ones_like(input_ids).to(device)
         
         # Generate sequences using custom model
         # PlasmidGPT model has custom generate() method
@@ -232,7 +203,9 @@ async def generate_sequences(request: GenerationRequest):
                         "max_length": request.max_length,
                         "num_return_sequences": request.num_return_sequences,
                         "temperature": request.temperature,
-                        "do_sample": request.do_sample
+                        "do_sample": request.do_sample,
+                        "attention_mask": attention_mask,
+                        "pad_token_id": tokenizer.eos_token_id if hasattr(tokenizer, 'eos_token_id') else 50256
                     }
                     if generation_config:
                         gen_kwargs["generation_config"] = generation_config