Factor Loadings using sklearn

Multiply each component by the square root of its corresponding eigenvalue:

pca.components_.T * np.sqrt(pca.explained_variance_)

This should produce your loading matrix.