Skip to content

Reuse memory for ELBO calculation #75

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
Open

Reuse memory for ELBO calculation #75

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
b69aabf
in-place
cscherrer Jun 7, 2022
e00e7ce
in-place
cscherrer Jun 7, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 7 additions & 4 deletions src/elbo.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,18 +1,21 @@
function maximize_elbo(rng, logp, dists, ndraws, executor)
μ = dists[1].μ
N = length(μ)
u = Random.randn!(rng, similar(μ, N, ndraws))
EE = Core.Compiler.return_type(
elbo_and_samples, Tuple{typeof(rng),typeof(logp),eltype(dists),Int}
elbo_and_samples!, Tuple{typeof(rng),typeof(u),typeof(logp),eltype(dists),Int}
)
estimates = similar(dists, EE)
isempty(estimates) && return 0, estimates
Folds.map!(estimates, dists, executor) do dist
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Since u is being overwritten here, I think we cannot use Transducers, or different threads could be writing to it at the same time. This will probably just need to be made a simple map

return elbo_and_samples(rng, logp, dist, ndraws)
return elbo_and_samples!(rng, u, logp, dist, ndraws)
end
_, iteration_opt = _findmax(estimates |> Transducers.Map(est -> est.value))
return iteration_opt, estimates
end

function elbo_and_samples(rng, logp, dist, ndraws)
ϕ, logqϕ = rand_and_logpdf(rng, dist, ndraws)
function elbo_and_samples!(rng, u, logp, dist, ndraws)
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

draws will need to be removed from ELBOEstimate as well.

ϕ, logqϕ = rand_and_logpdf!(rng, u, dist, ndraws)
logpϕ = similar(logqϕ)
logpϕ .= logp.(eachcol(ϕ))
logr = logpϕ - logqϕ
Expand Down
4 changes: 2 additions & 2 deletions src/mvnormal.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,13 +19,13 @@ function fit_mvnormals(θs, ∇logpθs; kwargs...)
end

# faster than computing `logpdf` and `rand` independently
function rand_and_logpdf(rng, dist::Distributions.MvNormal, ndraws)
function rand_and_logpdf!(rng, u, dist::Distributions.MvNormal, ndraws)
μ = dist.μ
Σ = dist.Σ
N = length(μ)

# draw points
u = Random.randn!(rng, similar(μ, N, ndraws))
Random.randn!(rng, u)
unormsq = vec(sum(abs2, u; dims=1))
x = PDMats.unwhiten!(u, Σ, u)
x .+= μ
Expand Down