loop_train.py

for i, step_size in enumerate(step_sizes):
    
    X_used = X[used_samples, ...]
    Y_used = Y[used_samples, ...]

    X_unused = X[unused_samples, ...]
    Y_unused = Y[unused_samples, ...]

    model = get_model()

    model.fit(X_used, Y_used, epochs=45, verbose=1, batch_size=32)

    pred_ununsed = model.predict(X_unused).tolist()
    
    entr = [entropy(l) for l in pred_ununsed]
    
    threshold = sorted(entr, reverse=True)[step_size]

    pred_test = model.predict(X_test)
    pred_test = np.argmax(pred_test, axis=-1)

    f1 = f1_score(Y_test, pred_test, average="macro")

    acc = accuracy_score(Y_test, pred_test)

    results.append({"size": X_used.shape[0], "accuracy": acc, "f1": f1})

    step = [x for x, v in zip(unused_samples, entr) if v >= threshold]
    used_samples += step
    unused_samples = list(set(unused_samples) - set(step))