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.naive_bayes import MultinomialNB
from sklearn.svm import SVC
from sklearn.metrics import accuracy_score

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

# Define the router for text tasks
router = APIRouter()

DESCRIPTION_NAIVE_BAYES = "Naive Bayes Text Classifier"
DESCRIPTION_SVM = "SVM Text Classifier with TF-IDF"

# Naive Bayes Endpoint
@router.post("/text", tags=["Text Task"], description=DESCRIPTION_NAIVE_BAYES)
async def evaluate_text(request: TextEvaluationRequest):
    """
    Evaluate text classification using Naive Bayes.
    """
    username, space_url = get_space_info()

    # 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 dataset
    dataset = load_dataset(request.dataset_name)
    dataset = dataset.map(lambda x: {"label": LABEL_MAPPING[x["label"]]})

    # Train-Test Split
    train_test = dataset["train"].train_test_split(test_size=request.test_size, seed=request.test_seed)
    train_texts = [x["text"] for x in train_test["train"]]
    train_labels = [x["label"] for x in train_test["train"]]
    test_texts = [x["text"] for x in train_test["test"]]
    test_labels = [x["label"] for x in train_test["test"]]

    # TF-IDF Vectorization
    vectorizer = TfidfVectorizer(max_features=5000)
    train_vectors = vectorizer.fit_transform(train_texts)
    test_vectors = vectorizer.transform(test_texts)

    # Train Naive Bayes Classifier
    model = MultinomialNB()
    model.fit(train_vectors, train_labels)

    # Track emissions
    tracker.start()
    tracker.start_task("inference")
    predictions = model.predict(test_vectors)
    emissions_data = tracker.stop_task()

    # Calculate Accuracy
    accuracy = accuracy_score(test_labels, predictions)

    return {
        "username": username,
        "space_url": space_url,
        "submission_timestamp": datetime.now().isoformat(),
        "model_description": DESCRIPTION_NAIVE_BAYES,
        "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": "/text",
        "dataset_config": {
            "dataset_name": request.dataset_name,
            "test_size": request.test_size,
            "test_seed": request.test_seed
        }
    }

# SVM Endpoint
@router.post("/text_svm", tags=["Text Task"], description=DESCRIPTION_SVM)
async def evaluate_text_svm(request: TextEvaluationRequest):
    """
    Evaluate text classification using SVM.
    """
    username, space_url = get_space_info()

    # 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 dataset
    dataset = load_dataset(request.dataset_name)
    dataset = dataset.map(lambda x: {"label": LABEL_MAPPING[x["label"]]})

    # Train-Test Split
    train_test = dataset["train"].train_test_split(test_size=request.test_size, seed=request.test_seed)
    train_texts = [x["text"] for x in train_test["train"]]
    train_labels = [x["label"] for x in train_test["train"]]
    test_texts = [x["text"] for x in train_test["test"]]
    test_labels = [x["label"] for x in train_test["test"]]

    # TF-IDF Vectorization
    vectorizer = TfidfVectorizer(max_features=5000)
    train_vectors = vectorizer.fit_transform(train_texts)
    test_vectors = vectorizer.transform(test_texts)

    # Train SVM Classifier
    model = SVC(kernel="linear", probability=True)
    model.fit(train_vectors, train_labels)

    # Track emissions
    tracker.start()
    tracker.start_task("inference")
    predictions = model.predict(test_vectors)
    emissions_data = tracker.stop_task()

    # Calculate Accuracy
    accuracy = accuracy_score(test_labels, predictions)

    return {
        "username": username,
        "space_url": space_url,
        "submission_timestamp": datetime.now().isoformat(),
        "model_description": DESCRIPTION_SVM,
        "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": "/text_svm",
        "dataset_config": {
            "dataset_name": request.dataset_name,
            "test_size": request.test_size,
            "test_seed": request.test_seed
        }
    }