An Improved Single Step Non-autoregressive Transformer for Automatic Speech Recognition