Prediction of Sentinel images by SVM classifier

I’m looking for a way to reduce the time taken by SVM classifier (without changing the dimenisons of my training data or data on which prediction has to be done) to predict the raster image data.
I have used the below classifier to predict the data .

    from sklearn.svm import SVC
    import numpy as np
    clf = SVC(C=4, kernel = 'rbf',gamma= 0.25) 
    clf.fit(X, y) 
    
    bands_data = np.dstack(bands_data) #bands_data is the stack of raster images(there are around 11 files) data in a ndarray.
    bands_data =np.nan_to_num(bands_data)
    rows, cols, n_bands= bands_data.shape
    n_samples = rows * cols
    flat_pixels = bands_data.reshape((n_samples,n_bands))
    r = clf.predict(flat_pixels)