True. Also, you are probably doing it wrong by turning a matrix to matrix multiplication into a for loop (over rows). The optimal solution results in better pe

True. Also, you are probably doing it wrong by turning a matrix to matrix multiplication into a for loop (over rows). The optimal solution results in better performance sim = np.vstack(df.col) @ vec