UserWarning: The value of the smallest subnormal for <class 'numpy.float32'> type is zero

[Gwohyanne]

It seems that the library is using the gcc/clang option -ffast-math somewhere, which causes the warning UserWarning: The value of the smallest subnormal for <class 'numpy.float32'> type is zero to appear sometimes. See this article for more info.
Might be worth looking into because messing with subnormals means changing floating point behavior and that's no good for precise maths. Could cause difficult-to-retrace issues in some circumstances.