I was working for Intel Client where we used to seamlessly integrate with Tensorflow and pytorch. It has got best optimization in terms of deep learning models across our hardware backends like CPUs, GPUs. The best thing is it can also compiles non-intel hardwares if utilized. We also get through some regular updates which is good. Review collected by and hosted on G2.com.

Nothing dislike as such. Need better documentation if you are unaware to use (basically for beginners). Non-intel architecture needs some broader support since you may need extra effort to integrate. Review collected by and hosted on G2.com.