Add a demo model with matrix assume + individual indexing x[i][j] / tuple indexing x[i, j]

[penelopeysm]

See: #733 (comment)

When refactoring the demo models into a separate package, we should include one that has this as well.

[penelopeysm]

(There's demo_assume_matrix_dot_observe_matrix, but it doesn't test with indexing on the LHS of a tilde)

@model function demo_assume_matrix_dot_observe_matrix(
    x=transpose([1.5 2.0;]), ::Type{TV}=Array{Float64}
) where {TV}
    n = length(x)
    d = n ÷ 2
    s ~ reshape(product_distribution(fill(InverseGamma(2, 3), n)), d, 2)
    s_vec = vec(s)
    m ~ MvNormal(zeros(n), Diagonal(s_vec))

    # Dotted observe for `Matrix`.
    x .~ MvNormal(m, Diagonal(s_vec))

    return (; s=s, m=m, x=x, logp=getlogp(__varinfo__))
end

[yebai]

Thanks, @penelopeysm; a quick thought: it might be helpful t to have separate sets of models in DynamicPPL.TestModels

• for testing the DynamicPPL compiler
• for testing inference algorithm correctness, e.g. models with analytical solutions or known solutions stored as constants
• for testing autograd and log density performance, type stability, etc.

Each set of models can be retrieved by, e.g. DynamicPP.TestModels.models_for_compiler_tests / models_for_inference_correctness_tests, models_for_performance_autograd_tests.