deps(deps-dev): bump the development-dependencies group across 1 directory with 2 updates

[dependabot]

Updates the requirements on braintools and mypy to permit the latest version.
Updates braintools to 0.3.0

Release notes

Sourced from braintools's releases.

v0.3.0

This release completes the library-wide correctness audit campaign begun in 0.2.0. The two remaining major modules — cogtask and metric — received their first dedicated static audits, while optim, trainer, visualize, surrogate, and init underwent deeper second-pass re-audits whose findings were each verified against reference implementations (optax 0.2.6, torch.optim, PyTorch's NAdam) and numeric reproductions. Every fix is locked in behind a regression test suite. The changes are corrections to existing behavior — no intentional API breaks — but the breadth and the genuine numerical/algorithmic bugs corrected (double-applied SM3 momentum, a centered RMSprop that was a silent no-op, dropped Parallel branches in cogtask, distributed broadcast that failed on more than one device, and more) warrant the minor-version bump.

Highlights

• Audit campaign completed: cogtask and metric are now audited, and the five previously-audited modules were re-audited against their reference implementations, replacing plausible-but-unverified behavior with reference-checked correctness.
• Real algorithmic fixes: SM3 no longer applies momentum twice; RMSprop(centered=True) actually centers; Nadam honors momentum_decay; cogtask.Parallel no longer silently drops compound (>> / Repeat) branches; distributed broadcast() works on more than one device.
• Restored public API: cogtask.create_task and metric.L1Loss are now importable from their packages (both were documented but absent from __all__).
• Regression coverage: each audited module ships a dedicated regression test suite (visualize exercised source at 99%, surrogate at 100%).

Added

• braintools.cogtask.create_task is now exported from the package (#122).
• braintools.metric.L1Loss is now exported (documented but previously missing from __all__) (#121).

Fixed

braintools.optim (#120) — re-audit verified against optax 0.2.6 / torch.optim

• SM3 no longer applies first-moment momentum twice; momentum=0 now truly disables it (matches optax.sm3).
• RMSprop(centered=True) uses scale_by_stddev instead of silently falling back to scale_by_rms.
• Nadam implements PyTorch-style scheduled momentum so momentum_decay takes effect (verified against torch.optim.NAdam).
• LBFGS reads the live LR schedule each step instead of freezing the LR at construction; CosineAnnealingWarmRestarts.step(epoch=...) recomputes T_cur/T_i from the absolute epoch.
• Fail-fast validation of optimizer hyperparameters and scheduler enums;

... (truncated)

Changelog

Sourced from braintools's changelog.

Version 0.3.0 (2026-06-19)

This release completes the library-wide correctness audit campaign begun in 0.2.0. The two remaining major modules — cogtask and metric — received their first dedicated static audits, while optim, trainer, visualize, surrogate, and init underwent deeper second-pass re-audits whose findings were each verified against reference implementations (optax 0.2.6, torch.optim, PyTorch's NAdam) and numeric reproductions. Every fix is locked in behind a regression test suite. The changes are corrections to existing behavior — no intentional API breaks — but the breadth and the genuine numerical/algorithmic bugs corrected (double-applied SM3 momentum, a centered RMSprop that was a silent no-op, dropped Parallel branches in cogtask, distributed broadcast that failed on more than one device, and more) warrant the minor-version bump.

Highlights

• Audit campaign completed: cogtask and metric are now audited, and the five previously-audited modules were re-audited against their reference implementations, replacing plausible-but-unverified behavior with reference-checked correctness.
• Real algorithmic fixes: SM3 no longer applies momentum twice; RMSprop(centered=True) actually centers; Nadam honors momentum_decay; cogtask.Parallel no longer silently drops compound (>> / Repeat) branches; distributed broadcast() works on more than one device.
• Restored public API: cogtask.create_task and metric.L1Loss are now importable from their packages (both were documented but absent from __all__).
• Regression coverage: each audited module ships a dedicated regression test suite (visualize exercised source at 99%, surrogate at 100%).

Added

• braintools.cogtask.create_task is now exported from the package (#122).
• braintools.metric.L1Loss is now exported (documented but previously missing from __all__) (#121).

Fixed

braintools.optim (#120) — re-audit verified against optax 0.2.6 / torch.optim

• SM3 no longer applies first-moment momentum twice; momentum=0 now truly disables it (matches optax.sm3).
• RMSprop(centered=True) uses scale_by_stddev instead of silently falling back to scale_by_rms.
• Nadam implements PyTorch-style scheduled momentum so momentum_decay takes effect (verified against torch.optim.NAdam).
• LBFGS reads the live LR schedule each step instead of freezing the LR at construction; CosineAnnealingWarmRestarts.step(epoch=...) recomputes T_cur/T_i from the absolute epoch.

... (truncated)

Commits

• f37f3c0 release: 0.3.0 (#123)
• bc6da89 delete issues doc
• d6e86bf fix(cogtask): repair Parallel compound children, make_encoder scalar, and bro...
• 072c6cd fix(metric): correctness, brainunit, and edge-case fixes across braintools.me...
• 6a29629 fix(optim): re-audit fixes — SM3 double momentum, RMSprop centered, NAdam sch...
• e9d4ae8 fix(trainer): correct LR-monitor logging, save_top_k pruning, resume, and dis...
• 336bebc fix(visualize): resolve audit issues, add regression tests (>90% cov) (#118)
• 252d4ff fix(surrogate): make LogTailedRelu.surrogate_fun continuous, fix docs (#117)
• 1189c9d delete issues doc
• d92bc18 fix(init): correct TruncatedNormal array std, param() State handling & docs (...
• Additional commits viewable in compare view

Updates mypy to 2.3.0

Changelog

Sourced from mypy's changelog.

Mypy Release Notes

Next Release

Packaging changes

• No longer provide mypyc-accelerated wheels for macOS x86_64 [mypyc-wheels #119](mypyc/mypy_mypyc-wheels#119)

Mypy 2.3

We've just uploaded mypy 2.3.0 to the Python Package Index (PyPI). Mypy is a static type checker for Python. This release includes new features, performance improvements and bug fixes. You can install it as follows:

python3 -m pip install -U mypy

You can read the full documentation for this release on Read the Docs.

The Upcoming Switch to the New Native Parser

We are planning to enable the new native parser (--native-parser) by default soon. We recommend that you test the native parser in your projects and report any issues in the mypy issue tracker.

Mypyc Free-threading Memory Safety

Free-threaded Python builds that don't have the GIL require additional synchronization primitives or lock-free algorithms to ensure memory safety when there are race conditions (for example, when a thread reads a list item while another thread writes the same list item concurrently). This release greatly improves memory safety of free threading.

List operations are now memory-safe on free threaded Python builds, even in the presence of race conditions. This has some performance cost. For list-heavy workloads, using librt.vecs.vec instead of list is often significantly faster, but note that vec is not (and likely won't be) fully memory safe, and the user is expected to avoid race conditions. The newly introduced librt.threading.Lock helps with this. Using variable-length tuples can also be more efficient than lists, since tuples are immutable and don't require expensive synchronization to ensure memory safety.

Instance attribute access is also (mostly) memory safe now on free-threaded builds in the presence of race conditions. We are planning to fix the remaining unsafe cases in a future release.

Full list of changes:

• Make attribute access memory safe on free-threaded builds (Jukka Lehtosalo, PR 21705)
• Fix unsafe borrowing of instance attributes with free-threading (Jukka Lehtosalo, PR 21688)
• Make list get/set item more memory safe on free-threaded builds (Jukka Lehtosalo, PR 21683)
• Don't borrow list items on free-threaded builds (Jukka Lehtosalo, PR 21679)
• Make multiple assignment from list memory-safe on free-threaded builds (Jukka Lehtosalo, PR 21684)

... (truncated)

Commits

• 8aabf84 Drop +dev from version
• 4d8ad2a Update changelog for 2.3 release (#21728)
• 2c21546 [mypyc] Update documentation of race conditions under free threading (#21726)
• a9f62a3 [mypyc] Make attribute access memory safe on free-threaded builds (#21705)
• 0faa413 Use PYODIDE environment variable for Emscripten cross-compilation detection...
• 3d75cdb [mypyc] Borrow final attributes more aggressively (#21702)
• 24c237d [mypyc] Improve documentation of Final (#21713)
• b5be217 [mypyc] Update free threading Python compatibility docs (#21711)
• cbcb51a Narrow for frozendict membership check (#21709)
• af2bc0f Sync typeshed (#21707)
• Additional commits viewable in compare view

[dependabot]

Labels

The following labels could not be found: automerge:disabled. Please create it before Dependabot can add it to a pull request.

Please fix the above issues or remove invalid values from dependabot.yml.