Fixed transformations

[penelopeysm]

Closes #1249. I'll write more here later, but you can see the changelog for a good overview of this PR.

Things to do

• Generalise LinkedVecTransformAccumulator to something that's more like FixedTransformAccumulator
• Add a convenience function for 'static'-fying an LDF (?)
• Add tests for new behaviour
• Add dev docs on FixedTransform

Possibly in a separate PR:

• Look at the Turing VI interface to see what it needs from bijector(model), and whether that can be removed (though, note also VI interface is too low-level Turing.jl#2783)

---

Some scripts to use to benchmark:

using DynamicPPL, Distributions, LogDensityProblems, Random, Chairmarks, ForwardDiff, ADTypes, LinearAlgebra

# @model function esc(J, y, sigma)
#     mu ~ Normal(0, 5)
#     tau ~ truncated(Cauchy(0, 5); lower=0)
#     theta ~ MvNormal(fill(mu, J), tau^2 * I)
#     for i in 1:J
#         y[i] ~ Normal(theta[i], sigma[i])
#     end
# end
# J = 8
# y = [28, 8, -3, 7, -1, 1, 18, 12]
# sigma = [15, 10, 16, 11, 9, 11, 10, 18]
# m = esc(J, y, sigma)

@model function f()
    x ~ product_distribution([Beta(2,2), Uniform(4, 5), Normal()])
end
m = f()

ldf1 = LogDensityFunction(m, getlogjoint_internal, LinkAll(); adtype=AutoForwardDiff(), fix_transforms=true);
p = rand(Xoshiro(468), ldf1);
@b LogDensityProblems.logdensity(ldf1 ,p)
@b LogDensityProblems.logdensity_and_gradient(ldf1 ,p)

ldf2 = LogDensityFunction(m, getlogjoint_internal, LinkAll(); adtype=AutoForwardDiff());
p = rand(Xoshiro(468), ldf2);
@b LogDensityProblems.logdensity(ldf2 ,p)
@b LogDensityProblems.logdensity_and_gradient(ldf2 ,p)

[github-actions]

Benchmark Report

• this PR's head: 328514d3adde9e4ee8ed5caab5f3e5ae821d87df
• base branch: e01fbd60d5a54a3b314b5bc5ea6654cefcba34a0

Computer Information

Julia Version 1.11.9
Commit 53a02c0720c (2026-02-06 00:27 UTC)
Build Info:
  Official https://julialang.org/ release
Platform Info:
  OS: Linux (x86_64-linux-gnu)
  CPU: 4 × AMD EPYC 7763 64-Core Processor
  WORD_SIZE: 64
  LLVM: libLLVM-16.0.6 (ORCJIT, znver3)
Threads: 1 default, 0 interactive, 1 GC (on 4 virtual cores)

Benchmark Results

┌───────────────────────┬───────┬─────────────┬────────┬───────────────────────────────┬────────────────────────────┬─────────────────────────────────┐
│                       │       │             │        │       t(eval) / t(ref)        │     t(grad) / t(eval)      │        t(grad) / t(ref)         │
│                       │       │             │        │ ─────────┬──────────┬──────── │ ───────┬─────────┬──────── │ ──────────┬───────────┬──────── │
│                 Model │   Dim │  AD Backend │ Linked │     base │  this PR │ speedup │   base │ this PR │ speedup │      base │   this PR │ speedup │
├───────────────────────┼───────┼─────────────┼────────┼──────────┼──────────┼─────────┼────────┼─────────┼─────────┼───────────┼───────────┼─────────┤
│               Dynamic │    10 │    mooncake │   true │   290.25 │   281.65 │    1.03 │   7.67 │    8.07 │    0.95 │   2227.42 │   2274.15 │    0.98 │
│                   LDA │    12 │ reversediff │   true │  2502.18 │  2615.12 │    0.96 │   2.18 │    2.10 │    1.04 │   5463.79 │   5504.42 │    0.99 │
│   Loop univariate 10k │ 10000 │    mooncake │   true │ 31061.43 │ 31536.61 │    0.98 │   6.38 │    6.72 │    0.95 │ 198132.80 │ 211866.61 │    0.94 │
├───────────────────────┼───────┼─────────────┼────────┼──────────┼──────────┼─────────┼────────┼─────────┼─────────┼───────────┼───────────┼─────────┤
│    Loop univariate 1k │  1000 │    mooncake │   true │  3167.57 │  3245.90 │    0.98 │   6.21 │    6.50 │    0.96 │  19685.67 │  21106.43 │    0.93 │
│      Multivariate 10k │ 10000 │    mooncake │   true │ 33426.98 │ 33122.24 │    1.01 │   9.93 │   14.85 │    0.67 │ 332031.62 │ 491917.53 │    0.67 │
│       Multivariate 1k │  1000 │    mooncake │   true │  3434.58 │  3589.19 │    0.96 │   9.25 │    8.96 │    1.03 │  31761.88 │  32149.59 │    0.99 │
├───────────────────────┼───────┼─────────────┼────────┼──────────┼──────────┼─────────┼────────┼─────────┼─────────┼───────────┼───────────┼─────────┤
│ Simple assume observe │     1 │ forwarddiff │  false │     0.87 │     0.88 │    1.00 │  10.35 │   10.44 │    0.99 │      9.05 │      9.17 │    0.99 │
│           Smorgasbord │   201 │ forwarddiff │  false │   952.99 │   975.98 │    0.98 │ 105.86 │   72.25 │    1.47 │ 100879.30 │  70511.64 │    1.43 │
│           Smorgasbord │   201 │      enzyme │   true │  1291.10 │  1281.81 │    1.01 │   4.88 │    4.93 │    0.99 │   6300.88 │   6320.38 │    1.00 │
├───────────────────────┼───────┼─────────────┼────────┼──────────┼──────────┼─────────┼────────┼─────────┼─────────┼───────────┼───────────┼─────────┤
│           Smorgasbord │   201 │ forwarddiff │   true │  1712.82 │  1314.32 │    1.30 │  52.92 │   69.17 │    0.77 │  90647.11 │  90905.39 │    1.00 │
│           Smorgasbord │   201 │    mooncake │   true │  1299.54 │  1438.22 │    0.90 │   4.59 │    4.26 │    1.08 │   5968.04 │   6133.70 │    0.97 │
│           Smorgasbord │   201 │ reversediff │   true │  1310.46 │  1293.32 │    1.01 │ 125.06 │  128.45 │    0.97 │ 163883.77 │ 166123.35 │    0.99 │
├───────────────────────┼───────┼─────────────┼────────┼──────────┼──────────┼─────────┼────────┼─────────┼─────────┼───────────┼───────────┼─────────┤
│              Submodel │     1 │    mooncake │   true │     0.87 │     0.88 │    1.00 │  26.87 │   26.31 │    1.02 │     23.47 │     23.08 │    1.02 │
└───────────────────────┴───────┴─────────────┴────────┴──────────┴──────────┴─────────┴────────┴─────────┴─────────┴───────────┴───────────┴─────────┘

[codecov]

Codecov Report

❌ Patch coverage is 75.87940% with 48 lines in your changes missing coverage. Please review.
✅ Project coverage is 78.64%. Comparing base (e01fbd6) to head (328514d).

Files with missing lines	Patch %	Lines
src/transformed_values.jl	52.54%	28 Missing ⚠️
src/varinfo.jl	86.53%	7 Missing ⚠️
src/accumulators/vector_values.jl	63.63%	4 Missing ⚠️
src/contexts/default.jl	40.00%	3 Missing ⚠️
src/abstract_varinfo.jl	50.00%	2 Missing ⚠️
src/accumulators/fixed_transforms.jl	85.71%	2 Missing ⚠️
src/contexts/init.jl	88.88%	2 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff              @@
##           breaking    #1327      +/-   ##
============================================
+ Coverage     78.19%   78.64%   +0.44%     
============================================
  Files            50       50              
  Lines          3582     3521      -61     
============================================
- Hits           2801     2769      -32     
+ Misses          781      752      -29     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[github-actions]

DynamicPPL.jl documentation for PR #1327 is available at:
https://TuringLang.github.io/DynamicPPL.jl/previews/PR1327/

[penelopeysm]

@sunxd3 would you be able to take a look? Here's the TLDR and my overall thoughts on this:

• src/transformed_values.jl -- There used to be UntransformedValue, VectorValue and LinkedVectorValue <: AbstractTransformedValue -- these are renamed to TransformedValue(val, NoTransform()), TransformedValue(val, Unlink()), and TransformedValue(val, DynamicLink()). A lot of the changes in this PR are just boilerplate resulting from this slightly different type hierarchy. Better names are also welcome, although renaming some of this might be breaking at this point.

• The new type hierarchy lets us extend it to TransformedValue(val, FixedTransform(f)). The new types are breaking, and we could avoid this by instead making a new subtype, FixedTransformedValue(val, f). I'm still a bit uncertain about which way to go on this: I went with the overhaul, but I don't actually think that it's necessary, and part of me thinks that it would just be better to not break the old API. If you agree, I'll try to change it back.

• The remaining question would be whether the general interface makes sense to you (we discussed this on Monday). I think the docs are the best place to read about the interface https://turinglang.org/DynamicPPL.jl/previews/PR1327/transforms_fixed/. The implementation of the interface is in src/accumulators/fixed_transforms.jl and src/logdensityfunction.jl, but that's all quite standard boilerplate, nothing particularly interesting.

[sunxd3]

I think the type hierarchy changes all makes sense. Although it is very breaking, even though I don't imagine a lot of people depend on the transformation interface, it might still worth giving some deprecation buffer (we can talk about this in the meeting).

The general interface also make sense to me.

Multivariate 10k has some performance regression on Mooncake, I wonder why.

[sunxd3]

an idea: this function can be run n times and compare the if the results are the same

[sunxd3]

do we need to remove these too?

[sunxd3]

need update?

Suggested change

	`RangeAndTransform`.
	`TransformedValue`.

?

[sunxd3]

Suggested change

	Finally, the `fix_transform` keyword argument allows you to specify whether the transforms
	Finally, the `fix_transforms` keyword argument allows you to specify whether the transforms

[sunxd3]

The biggest break from this might be the removal of vi[@varname]syntax. It just existed for quite a long time