Assume in vectorisation of HMC bug.

[trappmartin]

We call setgid! only if assume is called for a single distribution during HMC sampling

Turing.jl/src/inference/hmc.jl

Lines 265 to 267 in 65371ce

	function assume(spl::Sampler{<:Hamiltonian}, dist::Distribution, vn::VarName, vi::VarInfo)
	Turing.DEBUG && @debug "assuming..."
	updategid!(vi, vn, spl)

But we do not do the same during vectorisation.