multinomial casts input to np.float64

np.random.multinomial seems to cast its second argument pvals to an array of dtype float64. Whilst this isn’t an issue in and of itself, I’ve come across an interesting scenario where I have an array of dtype float32 whose sum is 0.99999994, and when it gets cast to float64, its sum is now 1.0000000222053895, which causes np.random.multinomial to raise ValueError: sum(pvals[:-1]) > 1.0.

For a little context, I have an array of dtype float32 because it’s come off the GPU, where single precision calculations are a great deal faster.

[olfMombach]

I don’t really see how that fixes the problem

[leachim]

I happened to run into this issue today. Wondering whether there is any reason for casting the array to float64? This should probably be fixed in numpy itself.