Bump the all-julia-packages group across 2 directories with 2 updates

[dependabot]

Updates the requirements on Pluto and Mooncake to permit the latest version.
Updates Pluto to 1.0.1

Release notes

Sourced from Pluto's releases.

v1.0.1

Try this release in your browser! (Available 30 minutes after the release)

Learn more about Binder and Pluto at plutojl.org/en/docs/binder!

New features

• update static featured notebooks bundle to Pluto 1.0

Fixes

• audio recording problem (#3559)

Diff since v1.0.0

Commits

• a31607e Bundle Web Assets
• cc4652b Fix #3559
• ece8a0a Fix missing localization of t_split_this_cell_into_cells
• 8fd1adf Update finnish translations (#3561)
• b1a5414 Update featured sources with 1.0 export
• 699cda7 Remove deprecations (#3560)
• 2e11e99 Update Project.toml
• 3670f5a Settings panel (#3556)
• 0385559 Put Julia version in statefile
• 8e9ff50 Make --pluto-schema-types-color a brighter color in dark mode (#3558)
• Additional commits viewable in compare view

Updates Mooncake to 0.5.31

Release notes

Sourced from Mooncake's releases.

v0.5.31

Mooncake v0.5.31

Diff since v0.5.30

Merged pull requests:

• Guard codual_type/fcodual_type against unbound TypeVars (#1191) (#1192) (@​yebai)

Closed issues:

• fcodual_type / codual_type crash on UnionAll Type inputs with free TypeVars (#1191)

Changelog

Sourced from Mooncake's changelog.

0.5.26

• Add Config(empty_cache=true) to free internal caches before rebuilding rules.

config = Mooncake.Config(empty_cache=true)
cache = Mooncake.prepare_gradient_cache(sin, 1.0; config)

0.5.25

• Add nfwd: a new N-wide forward-mode implementation built around NDual, with Nfwd / NfwdMooncake internals and broad tests for scalar, array, and rule-building paths.
• Expand Mooncake's forward-mode interface and caching around nfwd, including prepared derivative/gradient cache improvements and lower-allocation hot paths for repeated calls.
• Route a broader scalar-math set through nfwd-backed direct primitive frule!! / rrule!! wrappers, reducing dependence on imported ChainRules rules for these cases.
• Move the ChainRules-backed matrix exp rule into MooncakeChainRulesExt, making ChainRules a weak dependency rather than a core dependency.
• Add precompile workloads, including complex scalar reverse/forward-mode paths for ComplexF64 and ComplexF32.
• Improve docs for nfwd, including usage examples, interface notes, and clarification of nfwd/public-interface overheads.

The friendly_tangents=true path previously converted every internal tangent to a value of the primal type via tangent_to_primal!!. This relied on _copy_output to pre-allocate a buffer and tangent_to_primal_internal!! to fill it on every call. Both steps proved problematic:

• _copy_output is best-effort and not guaranteed correct for all types — #1084 shows a recent silent failure
• The primal round-trip was wrong for types with shared storage (e.g. Symmetric, where one stored entry represents two logical positions), silently returning an incorrect gradient — #937

Default behaviour change

Before	After
default: value of primal type	default: raw Mooncake tangent
primal round-trip: always	primal round-trip: explicit opt-in
custom gradient: not possible	custom gradient: explicit opt-in

The raw-tangent default (friendly_tangents=false) is safer: it never silently drops or corrupts information and avoids unnecessary allocation. Under the default, arrays of IEEEFloat (or complex) elements have plain array tangents; callables with no captured differentiable state return NoTangent; and structs or closures with differentiable fields return a Mooncake.Tangent (immutable) or Mooncake.MutableTangent (mutable) wrapping a named tuple of their field tangents.

With friendly_tangents=true, structs (both immutable and mutable with the standard MutableTangent tangent type) and closures additionally unwrap to plain NamedTuples. Mutable structs with custom tangent types return raw tangent unchanged. Types whose raw tangent reflects internal implementation layout rather than user-visible structure — AbstractDict (hash-table internals), Symmetric, Hermitian, SymTridiagonal — require explicit gradient reconstruction and are opt-in, each with their own tests.

0.5.24

Add stop_gradient(x) to block gradient propagation, analogous to tf.stop_gradient in TensorFlow and jax.lax.stop_gradient in JAX.

julia> using Mooncake
julia> f(x) = x[1] * Mooncake.stop_gradient(x)[2]
f (generic function with 1 method)
julia> cache = Mooncake.prepare_gradient_cache(f, [3.0, 4.0]);
julia> , (, g) = Mooncake.value_and_gradient!!(cache, f, [3.0, 4.0]);
julia> g  # g[2] == 0.0: gradient through x[2] inside stop_gradient is blocked
2-element Vector{Float64}:
</tr></table>

... (truncated)

Commits

• 79e7c85 Guard codual_type/fcodual_type against unbound TypeVars (#1191) (#1192)
• 4407008 remove redundant resolve_unbound_globalrefs (#1180)
• b3a848f Bump compat for LogExpFunctions to 1 (#1186)
• 1baa6f1 Fix missing space in error message (#1187)
• 79dbe18 Pin JuliaFormatter to 2.4 in the format test env. (#1190)
• 8d345ef rule reuse consistency checks (#1172)
• 954d48b Update README.md
• 2381b60 max_norm_perturbation kwarg (#1173)
• 46bfb7c Note grant-funded institutions in contribution policy
• 386ef24 Reduce duplication in contribution policy
• Additional commits viewable in compare view

[github-actions]

TuringGLM.jl documentation for PR #187 is available at:
https://TuringLang.github.io/TuringGLM.jl/previews/PR187/