const FLOAT_EPSILON: f64 = 1e-6;Expand description
Tolerance for comparing a published value against a resolved one.
Same value and reason as gglib_proxy::props’s. A GGUF FLOAT32 0.7
stringifies to "0.7" and parses back to f64 0.7, while gglib’s own
resolved f32 0.7 widens to 0.699999988079071. The ~1.2e-8 gap is an
artefact of the round trip, not a disagreement, and must not render as one.