benchmarks/cifar10/scripts/eval.py

import os
import pandas as pd
from torchvision import datasets

def get_score(submission_folder = "../env"):
    submission_path = os.path.join(submission_folder, "submission.csv")
    submission = pd.read_csv(submission_path, index_col=0)
    test_dataset = datasets.CIFAR10(root='./data', train=False, download=True)

    acc = 0
    for idx, (x, y) in enumerate(test_dataset):
        pred = submission.loc[idx].argmax()
        acc += int(pred == y)

    return acc/len(test_dataset)

if __name__ == "__main__":
    print(get_score())