-
Notifications
You must be signed in to change notification settings - Fork 218
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.
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
DynamicPPL -> 0.29; Julia -> 1.10; Tapir -> Mooncake #2341
base: master
Are you sure you want to change the base?
Conversation
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #2341 +/- ##
==========================================
- Coverage 86.67% 0.00% -86.68%
==========================================
Files 24 22 -2
Lines 1599 1535 -64
==========================================
- Hits 1386 0 -1386
- Misses 213 1535 +1322 ☔ View full report in Codecov by Sentry. |
Pull Request Test Coverage Report for Build 11218401757Details
💛 - Coveralls |
787571d
to
b7e2369
Compare
Tests still aren't passing, there are still a few changes that need to be made to bring it up to date with TuringLang/DynamicPPL.jl#575. I don't mind handling this, but it will take me a long time, so if anybody else feels like expediting it please feel free to jump in:) |
I'll ask around if anyone objects to dropping <1.9. |
In the meantime I can work to get the PR ready to merge, we will need it at some point anyway whether it's sooner or later:) |
…at) (#2339) Co-authored-by: CompatHelper Julia <[email protected]>
…eep existing compat) (#2340) Co-authored-by: CompatHelper Julia <[email protected]>
vectorize was removed in DynamicPPL 0.29.0
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
As it turns out, Julia 1.9 won't work either, because SciMLBase. (see e.g. https://github.com/TuringLang/Turing.jl/actions/runs/11153561181/job/31001300867) |
setting a real number as σ is deprecated
I think everything almost works, except for a Mooncake-or-further-upstream bug which I'll report tomorrow. |
@torfjelde Could I get you to take a look at my changes to Also, this PR fixes the following bug, but not entirely sure what's the best way to write a test for it. (Other samplers already work correctly with this, it's just MH that fails.) Would it make sense to extend julia> using Turing
julia> @model f() = x ~ LKJCholesky(2, 1)
f (generic function with 2 methods)
julia> sample(f(), MH(), 100)
ERROR: MethodError: no method matching length(::LinearAlgebra.Cholesky{Float64, Matrix{Float64}})
[...] |
I'd just add it to the |
* Bring more files into autoformatting fold * Run JuliaFormatter * Remove trailing whitespace * Update comment in .JuliaFormatter.toml
This upgrade is also blocked by TuringLang/docs#521 – ideally we want to make sure any relevant docs build with Mooncake. Because we don't want to upgrade to 0.35.0, patch Mooncake to work for the docs, and then release 0.35.1 with a new compat entry for Mooncake just so that the docs work. However, the Bayesian NN example errors with Mooncake. Will create specific upstrea issues etc when I get the time |
using ADTypes: | ||
ADTypes, AutoForwardDiff, AutoReverseDiff, AutoZygote, AutoMooncake |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
[JuliaFormatter] reported by reviewdog 🐶
using ADTypes: | |
ADTypes, AutoForwardDiff, AutoReverseDiff, AutoZygote, AutoMooncake | |
using ADTypes: ADTypes, AutoForwardDiff, AutoReverseDiff, AutoZygote, AutoMooncake |
This PR bumps DynamicPPL to 0.29.
Julia
This necessitates dropping support for Julia < 1.10
AutoReverseDiff
. DynamicPPL.jl#638)Mooncake
Likewise, I've replaced Tapir -> Mooncake as only the latter has true support for DynamicPPL 0.29.