Open Source Toolkits for Speech Recognition

Has anyone here compared the deep learning capabilities of Kaldi to that of Tensorflow, Torch or the other DL frameworks?