# At the top of app.py, add debug printing
import os
import logging

# Configure logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

class Summarizer:
    def __init__(self):
        try:
            # Print current directory contents for debugging
            st.write("Current directory contents:")
            st.write(os.listdir('.'))
            
            # Base model
            self.base_model = AutoModelForSeq2SeqLM.from_pretrained(
                "GanjinZero/biobart-base",
                local_files_only=False  # Allow downloading base model
            )
            self.tokenizer = AutoTokenizer.from_pretrained(
                "GanjinZero/biobart-base",
                local_files_only=False
            )
            
            # Load adapter config from local file
            adapter_config_path = "./adapter_config.json"
            if not os.path.exists(adapter_config_path):
                st.error(f"adapter_config.json not found in {os.getcwd()}")
                raise FileNotFoundError("adapter_config.json not found")
                
            st.write(f"Loading adapter config from {adapter_config_path}")
            
            # Create LoRA config
            lora_config = LoraConfig(
                r=8,
                lora_alpha=16,
                lora_dropout=0.1,
                bias="none",
                task_type="SEQ_2_SEQ_LM",
                target_modules=["q_proj", "v_proj"],
                inference_mode=True
            )
            
            # Load base model for fine-tuning
            base_model_for_finetuned = AutoModelForSeq2SeqLM.from_pretrained(
                "GanjinZero/biobart-base",
                local_files_only=False
            )
            
            st.write("Loading fine-tuned model...")
            # Try to load the PEFT model from the current directory
            self.finetuned_model = PeftModel.from_pretrained(
                base_model_for_finetuned,
                ".",  # Current directory
                config=lora_config,
                torch_dtype=torch.float32,
                is_trainable=False,
                local_files_only=True
            )
            
            self.finetuned_model.eval()
            st.success("Models loaded successfully!")
            
        except Exception as e:
            st.error(f"Error loading models: {str(e)}")
            st.write("Debug info:")
            st.write(f"Current working directory: {os.getcwd()}")
            st.write(f"Directory contents: {os.listdir('.')}")
            if os.path.exists('adapter_config.json'):
                with open('adapter_config.json', 'r') as f:
                    st.write("adapter_config.json contents:", f.read())
            raise