evaluate_rnn.py

def evaluate(sentence):
    try:
        embedded = encode(sentence)
    except KeyError as error:
        print("! the word", error, "is unknown")
        return
    if embedded.shape[0] > SENTENCE_LENGTH_LIMIT:
        print(f"! the sentence should not contain more than {SENTENCE_LENGTH_LIMIT} tokens")
        return
    padded = np.zeros((SENTENCE_LENGTH_LIMIT, 300))
    padded[SENTENCE_LENGTH_LIMIT - embedded.shape[0]:, :] = embedded
    original = model(torch.tensor(padded.reshape((-1, 1, 300))).float()).detach().numpy()[0, 0, 0]
    simulated = homomorphic_inferer.simulate(padded)[0]
    start = timeit.default_timer()
    actual = homomorphic_inferer.encrypt_and_run(keys, padded)[0]
    end = timeit.default_timer()
    if actual < 0.35:
        print("- the sentence was negative", end=' ')
    elif actual > 0.65:
        print("+ the sentence was positive", end=' ')
    else:
        print("~ the sentence was neutral", end=' ')
    print(
        f"("
        f"original: {original * 100:.2f}%, "
        f"simulated: {simulated * 100:.2f}%, "
        f"actual: {actual * 100:.2f}%, "
        f"difference: {np.abs(original - actual) * 100:.2f}%, "
        f"took: {end - start:.3f} seconds"
        f")"
    )