tasks/text.py

from fastapi import APIRouter
from datetime import datetime
from datasets import load_dataset
from sklearn.metrics import accuracy_score
import random

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.
    
    Current Model: FrugalDisinfoHunter
    """
    # 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")

    #--------------------------------------------------------------------------------------------
    # YOUR MODEL INFERENCE CODE HERE
    # Update the code below to replace the random baseline by your model inference within the inference pass where the energy consumption and emissions are tracked.
    #--------------------------------------------------------------------------------------------   
    #--------------------------------------------------------------------------------------------
    # Load your model and tokenizer from Hugging Face or local path
    #--------------------------------------------------------------------------------------------

    from transformers import AutoTokenizer, AutoModelForSequenceClassification
    import torch

    # Load model from Hugging Face (adjust if you uploaded it there)
    model_name = "Zen0/FrugalDisinfoHunter"  # Replace with your model identifier if different
    tokenizer = AutoTokenizer.from_pretrained(model_name)
    model = AutoModelForSequenceClassification.from_pretrained(model_name)

    #--------------------------------------------------------------------------------------------
    # Load the dataset
    #--------------------------------------------------------------------------------------------

    # Assuming 'quotaclimat/frugalaichallenge-text-train' is the dataset you're working with
    dataset = load_dataset(request.dataset_name)

    # Access the test dataset (you can change this if you want to use a different split)
    test_dataset = dataset['test']  # Assuming you have a 'test' split available

    #--------------------------------------------------------------------------------------------
    # Tokenize the text data
    #--------------------------------------------------------------------------------------------

    # Tokenize the test data (the text field contains the quotes)
    test_texts = test_dataset["text"]  # The field 'text' contains the climate quotes

    inputs = tokenizer(test_texts, padding=True, truncation=True, return_tensors="pt", max_length=512)

    #--------------------------------------------------------------------------------------------
    # Inference
    #--------------------------------------------------------------------------------------------

    # Run inference on the dataset using the model
    with torch.no_grad():  # Disable gradient calculations
        outputs = model(**inputs)
        logits = outputs.logits

    # Get predictions from the logits (choose the class with the highest logit)
    predictions = torch.argmax(logits, dim=-1).cpu().numpy()  # Convert to numpy array for use

    #--------------------------------------------------------------------------------------------
    # YOUR MODEL INFERENCE STOPS HERE
    #--------------------------------------------------------------------------------------------   

    
    # 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