The thing that has kept me from Julia is the size of the ecosystem, but this was maybe 2 years ago now that I last tried it otherwise I thought the language seemed very nice. How are things doing for ML, NLP specifically if anyone knows?
I also feel that Knet, which is quite similar to PyTorch, is a bit underappreciated.
Still, with those two, the functionality that is currently implemented is significantly smaller than what Python offers.