Ask HN: Which would you prefer for ML: Torch or Julia?

I know Torch is generally preferred for Deep Learning-type research (awesome nn library) but I'm wondering how the two compare in other ML areas (like more standard data science / data-wrangling).