tasks/text.py

from fastapi import APIRouter
from datetime import datetime
from datasets import load_dataset
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import GridSearchCV
from sklearn.metrics import accuracy_score
from sklearn.pipeline import Pipeline

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

router = APIRouter()

DESCRIPTION = "TF-IDF + Logistic Regression"
ROUTE = "/text"

@router.post(ROUTE, tags=["Text Task"], description=DESCRIPTION)
async def evaluate_text(request: TextEvaluationRequest):
    """
    Evaluate text classification for climate disinformation detection using TF-IDF and Logistic Regression.
    """
    # 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 into training and testing sets
    train_data = dataset["train"]
    test_data = dataset["test"]

    train_texts, train_labels = train_data["text"], train_data["label"]
    test_texts, test_labels = test_data["text"], test_data["label"]

    # Start tracking emissions
    tracker.start()
    tracker.start_task("inference")

    # Define the pipeline with TF-IDF and Logistic Regression
    pipeline = Pipeline([
        ('tfidf', TfidfVectorizer(max_features=10000, ngram_range=(1, 2), stop_words="english")),
        ('clf', LogisticRegression(max_iter=1000, random_state=42))
    ])

    # Set up GridSearchCV for hyperparameter tuning
    param_grid = {
        'tfidf__max_features': [5000, 10000, 15000],
        'tfidf__ngram_range': [(1, 1), (1, 2)],
        'clf__C': [0.1, 1, 10]  # Regularization strength
    }

    grid_search = GridSearchCV(pipeline, param_grid, cv=3, scoring='accuracy', verbose=2)
    grid_search.fit(train_texts, train_labels)

    # Get best estimator from GridSearch
    best_model = grid_search.best_estimator_

    # Model Inference
    predictions = best_model.predict(test_texts)

    # Stop tracking emissions
    emissions_data = tracker.stop_task()

    # Calculate accuracy
    accuracy = accuracy_score(test_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": len(test_data),
        },
        "best_params": grid_search.best_params_
    }

    return results