Allow gradient on LDF to be called with e.g. views

[penelopeysm]

I benchmarked locally and didn't find any performance regression on the normal Vector use case (since convert is just a no-op).

[github-actions]

Benchmark Report

• this PR's head: 05db8c21a0dd069950f4e856d29416fd4b1ad0d9
• base branch: e1cb3991cb353efa4ab7f3e2d1284fa52086a0ce

Computer Information

Julia Version 1.11.7
Commit f2b3dbda30a (2025-09-08 12:10 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 │           VarInfo │ Linked │      base │  this PR │ speedup │   base │ this PR │ speedup │      base │   this PR │ speedup │
├───────────────────────┼───────┼─────────────┼───────────────────┼────────┼───────────┼──────────┼─────────┼────────┼─────────┼─────────┼───────────┼───────────┼─────────┤
│               Dynamic │    10 │    mooncake │             typed │   true │    375.80 │   335.93 │    1.12 │   9.79 │   10.86 │    0.90 │   3680.29 │   3647.41 │    1.01 │
│                   LDA │    12 │ reversediff │             typed │   true │   6030.88 │  2379.45 │    2.53 │   2.97 │    5.03 │    0.59 │  17930.79 │  11974.56 │    1.50 │
│   Loop univariate 10k │ 10000 │    mooncake │             typed │   true │ 103206.12 │ 96340.67 │    1.07 │   4.02 │    4.44 │    0.91 │ 415069.62 │ 428039.88 │    0.97 │
├───────────────────────┼───────┼─────────────┼───────────────────┼────────┼───────────┼──────────┼─────────┼────────┼─────────┼─────────┼───────────┼───────────┼─────────┤
│    Loop univariate 1k │  1000 │    mooncake │             typed │   true │   7940.71 │  7091.99 │    1.12 │   4.82 │    5.02 │    0.96 │  38310.40 │  35582.89 │    1.08 │
│      Multivariate 10k │ 10000 │    mooncake │             typed │   true │  32194.26 │ 30255.81 │    1.06 │  10.22 │   10.10 │    1.01 │ 329181.01 │ 305574.83 │    1.08 │
│       Multivariate 1k │  1000 │    mooncake │             typed │   true │   3646.27 │  3337.53 │    1.09 │   9.27 │    9.33 │    0.99 │  33811.65 │  31130.68 │    1.09 │
├───────────────────────┼───────┼─────────────┼───────────────────┼────────┼───────────┼──────────┼─────────┼────────┼─────────┼─────────┼───────────┼───────────┼─────────┤
│ Simple assume observe │     1 │ forwarddiff │             typed │  false │      2.73 │     2.46 │    1.11 │   3.88 │    3.96 │    0.98 │     10.60 │      9.73 │    1.09 │
│           Smorgasbord │   201 │ forwarddiff │             typed │  false │   1225.18 │  1118.82 │    1.10 │ 121.27 │   62.63 │    1.94 │ 148575.01 │  70066.99 │    2.12 │
│           Smorgasbord │   201 │ forwarddiff │       simple_dict │   true │       err │      err │     err │    err │     err │     err │       err │       err │     err │
├───────────────────────┼───────┼─────────────┼───────────────────┼────────┼───────────┼──────────┼─────────┼────────┼─────────┼─────────┼───────────┼───────────┼─────────┤
│           Smorgasbord │   201 │ forwarddiff │ simple_namedtuple │   true │       err │      err │     err │    err │     err │     err │       err │       err │     err │
│           Smorgasbord │   201 │      enzyme │             typed │   true │   1699.11 │  1550.81 │    1.10 │   5.82 │    6.86 │    0.85 │   9887.09 │  10633.58 │    0.93 │
│           Smorgasbord │   201 │    mooncake │             typed │   true │   1708.60 │  1552.98 │    1.10 │   5.32 │    5.27 │    1.01 │   9085.87 │   8190.34 │    1.11 │
├───────────────────────┼───────┼─────────────┼───────────────────┼────────┼───────────┼──────────┼─────────┼────────┼─────────┼─────────┼───────────┼───────────┼─────────┤
│           Smorgasbord │   201 │ reversediff │             typed │   true │   1731.62 │  1560.20 │    1.11 │  87.66 │   88.16 │    0.99 │ 151798.80 │ 137549.11 │    1.10 │
│           Smorgasbord │   201 │ forwarddiff │      typed_vector │   true │   1724.42 │  1550.09 │    1.11 │  55.67 │   56.27 │    0.99 │  95996.31 │  87223.74 │    1.10 │
│           Smorgasbord │   201 │ forwarddiff │           untyped │   true │   1689.61 │  1542.15 │    1.10 │  56.48 │  118.41 │    0.48 │  95428.83 │ 182601.85 │    0.52 │
├───────────────────────┼───────┼─────────────┼───────────────────┼────────┼───────────┼──────────┼─────────┼────────┼─────────┼─────────┼───────────┼───────────┼─────────┤
│           Smorgasbord │   201 │ forwarddiff │    untyped_vector │   true │   1686.37 │  1539.98 │    1.10 │  56.51 │   56.28 │    1.00 │  95290.17 │  86665.75 │    1.10 │
│              Submodel │     1 │    mooncake │             typed │   true │      7.16 │     6.53 │    1.10 │   5.18 │    5.21 │    0.99 │     37.12 │     34.03 │    1.09 │
└───────────────────────┴───────┴─────────────┴───────────────────┴────────┴───────────┴──────────┴─────────┴────────┴─────────┴─────────┴───────────┴───────────┴─────────┘

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 80.01%. Comparing base (e1cb399) to head (05db8c2).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1163      +/-   ##
==========================================
+ Coverage   79.99%   80.01%   +0.01%     
==========================================
  Files          41       41              
  Lines        3874     3877       +3     
==========================================
+ Hits         3099     3102       +3     
  Misses        775      775              

☔ 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.

[mhauru]

Code looks good to me if tests pass.