The discrete probability distribution in the digitRecognition benchmark does not sum to one

[dbasgoeze]

I'm not sure if this is intentional, but the probability mass function of the following distribution sums to 1.000004 , not 1.

I discovered this while attempting to port the benchmark to another framework, where such behavior is not allowed.

If this is intentional, what is the semantic of this?

dice/benchmarks/baselines/digitRecognition.dice

Line 1 in ed86716

let y = discrete(0.098717, 0.11237, 0.0993, 0.10218, 0.097367, 0.09035, 0.098633, 0.10442, 0.097517, 0.09915) in