from transformers import AutoTokenizer, AutoModelForSequenceClassification
from fastapi import APIRouter
from datetime import datetime
from datasets import load_dataset
from sklearn.metrics import accuracy_score
import torch
import numpy as np

from .utils.evaluation import TextEvaluationRequest
from .utils.emissions import tracker, clean_emissions_data, get_space_info

router = APIRouter()

DESCRIPTION = "FrugalDisinfoHunter Model"
ROUTE = "/text"

@router.post(ROUTE, tags=["Text Task"], description=DESCRIPTION)
async def evaluate_text(request: TextEvaluationRequest):
    """
    Evaluate text classification for climate disinformation detection.
    """
    # Get space info
    username, space_url = get_space_info()

    # Define the label mapping
    LABEL_MAPPING = {
        "0_not_relevant": 0,
        "1_not_happening": 1,
        "2_not_human": 2,
        "3_not_bad": 3,
        "4_solutions_harmful_unnecessary": 4,
        "5_science_unreliable": 5,
        "6_proponents_biased": 6,
        "7_fossil_fuels_needed": 7
    }

    # Load and prepare the dataset
    dataset = load_dataset(request.dataset_name)

    # Convert string labels to integers
    dataset = dataset.map(lambda x: {"label": LABEL_MAPPING[x["label"]]})

    # Split dataset
    train_test = dataset["train"].train_test_split(test_size=request.test_size, seed=request.test_seed)
    test_dataset = train_test["test"]
    
    # Start tracking emissions
    tracker.start()
    tracker.start_task("inference")

    try:
        # Model configuration
        model_name = "google/mobilebert-uncased"  # Base model
        local_weights = "model/model.pt"  # Path to our trained weights
        BATCH_SIZE = 32
        MAX_LENGTH = 256  # Increased from 128

        # Initialize tokenizer and model
        tokenizer = AutoTokenizer.from_pretrained(model_name)
        model = AutoModelForSequenceClassification.from_pretrained(
            model_name,
            num_labels=8,
            problem_type="single_label_classification"
        )
        
        # Load our trained weights
        try:
            state_dict = torch.load(local_weights, map_location='cpu')
            model.load_state_dict(state_dict)
        except Exception as e:
            print(f"Error loading weights: {e}")
            # Continue with base model if weights fail to load
            pass

        # Move model to appropriate device
        device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
        model = model.to(device)
        model.eval()  # Set to evaluation mode

        # Get test texts and process in batches
        test_texts = test_dataset["quote"]
        predictions = []

        # Process in batches
        for i in range(0, len(test_texts), BATCH_SIZE):
            # Clear CUDA cache if using GPU
            if torch.cuda.is_available():
                torch.cuda.empty_cache()

            batch_texts = test_texts[i:i + BATCH_SIZE]
            
            # Tokenize with padding and attention masks
            inputs = tokenizer(
                batch_texts,
                padding=True,
                truncation=True,
                max_length=MAX_LENGTH,
                return_tensors="pt"
            )
            
            # Move inputs to device
            inputs = {k: v.to(device) for k, v in inputs.items()}

            # Run inference with no gradient computation
            with torch.no_grad():
                outputs = model(**inputs)
                batch_preds = torch.argmax(outputs.logits, dim=1)
                predictions.extend(batch_preds.cpu().numpy())

        # Get true labels
        true_labels = test_dataset['label']

        # Stop tracking emissions
        emissions_data = tracker.stop_task()
        
        # Calculate accuracy
        accuracy = accuracy_score(true_labels, predictions)
        
        # Prepare results dictionary
        results = {
            "username": username,
            "space_url": space_url,
            "submission_timestamp": datetime.now().isoformat(),
            "model_description": DESCRIPTION,
            "accuracy": float(accuracy),
            "energy_consumed_wh": emissions_data.energy_consumed * 1000,
            "emissions_gco2eq": emissions_data.emissions * 1000,
            "emissions_data": clean_emissions_data(emissions_data),
            "api_route": ROUTE,
            "dataset_config": {
                "dataset_name": request.dataset_name,
                "test_size": request.test_size,
                "test_seed": request.test_seed
            }
        }
        
        return results
        
    except Exception as e:
        # Stop tracking in case of error
        tracker.stop_task()
        raise e