⬆️ bump braintrace from 0.1.2 to 0.2.0

[dependabot]

Bumps braintrace from 0.1.2 to 0.2.0.

Release notes

Sourced from braintrace's releases.

BrainTrace 0.2.0 is a major step forward. It adds a family of spiking neural network (SNN) online-learning algorithms, rewrites the eligibility-trace compiler around primitive-type dispatch, generalizes every ETP primitive to support multiple trainable inputs (fixing a silent bias-gradient drop), delivers substantial performance gains for D-RTRL and multi-step rollouts, and hardens the package with PEP 561 typing and a BPTT-oracle-backed test suite.

Highlights

New: SNN online-learning algorithms

Five SNN online-learning algorithms ship as flat ETraceVjpAlgorithm subclasses, all exported at the top level: EProp, OSTL (OSTLRecurrent / OSTLFeedforward), OTPE, OTTT, and OSTTP. A new _compute_learning_signal hook supports target-projection algorithms (OSTTP), and new trace helpers (PresynapticTrace, KappaFilter, FixedRandomFeedback) back them. The algorithms are cross-checked for regime equivalence and verified to decrease loss in integration smoke tests.

ETP compiler rewrite

The eligibility-trace compiler now dispatches on primitive-type identity rather than string-matching op/trace names, with structured, leveled diagnostics (DiagnosticKind, DiagnosticLevel, CompilationRecord) replacing ad-hoc warnings. New compile-time diagnostics surface previously silent issues — e.g. TRAINABLE_INVAR_NOT_PARAMSTATE flags a trainable input (such as a constant bias) that does not trace to a ParamState.

Multi-trainable-input ETP primitives (bias gradients)

Every ETP primitive was generalized from a single-weight assumption to an arbitrary named dict of trainable inputs, fixing a silent bias-gradient drop and a LoRA executor signature mismatch in one coherent refactor. All built-in primitives (elemwise, dense mm/mv, conv, sparse mm/mv, lora) now have first-class bias gradient support, each verified element-wise against a BPTT oracle. Layout-aware conv axis handling (NHWC/NCHW, OIHW/HWIO) and non-square dense weight broadcasting were also fixed.

Performance

• D-RTRL einsum fast path (fast_solve=True, default on): direct einsum kernels for mm/mv/elemwise replace nested vmap-of-vjp and per-step lax.cond overhead.
• Reduced-precision trace storage (trace_dtype, e.g. bf16/fp16) halves the dominant B*N^2 trace bandwidth while keeping Jacobians, learning signals, and final gradients in fp32.
• Multi-step trace fusion: the per-step eligibility-trace roll for exact algorithms is threaded into the forward scan, eliminating an O(T × Jacobian) HBM round-trip.

Typing, testing & packaging

• The package is now PEP 561 compliant (ships py.typed), with a pragmatic mypy config and CI type/packaging checks.
• A BPTT gradient oracle and a layered correctness suite (P2–P8) cover per-operator rules, public-API contracts, exact-class element-wise equivalence, approximate-class direction alignment, transform/integration invariance, and per-cell compiler relation guardrails.
• All public-API docstrings converted to NumPy-doc style with math, references, and runnable examples.

Deprecations

The entire v0.1.x class-based operator/parameter API is deprecated in favor of the new primitive-based ETP user-API. The legacy classes still work (as thin shims that route through the new primitives) but each emits a DeprecationWarning on first use and will be removed in a future release.

Deprecated (v0.1.x)	Use instead (v0.2.0)
MatMulOp	braintrace.matmul
ElemWiseOp	braintrace.element_wise
ConvOp	braintrace.conv
SpMatMulOp	braintrace.sparse_matmul
LoraOp	braintrace.lora_matmul
ETraceParam / ElemWiseParam	brainstate.ParamState + an ETP primitive function
NonTempParam	brainstate.ParamState + plain JAX ops

Breaking changes

1. OSTL factory removed — use OSTLRecurrent or OSTLFeedforward directly.
2. OTTT and OTPE require an explicit leak — no longer inferred from model.states(); both reject hidden groups with num_state > 1 at compile time.
3. Unit dependency change — brainunit replaced by saiunit.
4. ETPPrimitiveSpec removed — custom primitives register layout metadata via register_primitive keyword arguments (trainable_invars_fn, x_invar_index, y_outvar_index).

See the full changelog for the complete migration guide.

Version 0.1.3

What's Changed

• ⬆️ Bump styfle/cancel-workflow-action from 0.12.1 to 0.13.0 by @​dependabot[bot] in chaobrain/braintrace#56
• ⬆️ Bump actions/upload-artifact from 6 to 7 by @​dependabot[bot] in chaobrain/braintrace#57

... (truncated)

Changelog

Sourced from braintrace's changelog.

Version 0.2.0

This release is a major step for BrainTrace. It adds a family of spiking neural network (SNN) online-learning algorithms, rewrites the eligibility-trace compiler around primitive-type dispatch, generalizes every ETP primitive to support multiple trainable inputs (fixing a silent bias-gradient drop), delivers substantial performance gains for D-RTRL and multi-step rollouts, and hardens the package with PEP 561 typing and a BPTT-oracle-backed test suite.

Major Changes

New: SNN Online-Learning Algorithms

• Added five SNN online-learning algorithms as flat ETraceVjpAlgorithm subclasses: EProp, OSTL (OSTLRecurrent / OSTLFeedforward), OTPE, OTTT, and OSTTP. All are exported at the top level.
• Added a _compute_learning_signal hook to ETraceVjpAlgorithm to support target-projection algorithms (OSTTP) without disrupting the existing D-RTRL and pp-prop paths.
• Added supporting trace helpers: PresynapticTrace, KappaFilter, FixedRandomFeedback, and target-signal extraction utilities.
• Algorithms are cross-checked for regime equivalence and verified to decrease loss in integration smoke tests.

ETP Compiler Rewrite

• Rewrote the eligibility-trace compiler to dispatch on primitive-type identity rather than string-matching op or trace names, with structured, leveled diagnostics (DiagnosticKind, DiagnosticLevel, CompilationRecord) replacing ad-hoc warnings.
• Added compile-time diagnostics that surface previously silent issues — e.g. TRAINABLE_INVAR_NOT_PARAMSTATE flags a trainable input (such as a constant bias) that does not trace to a ParamState, so users can wrap it intentionally instead of silently losing its gradient.

Multi-Trainable-Input ETP Primitives (Bias Gradients)

• Generalized every ETP primitive from a single-"weight" assumption to an arbitrary named dict of trainable inputs. This fixes a silent bias-gradient drop and a LoRA executor signature mismatch in one coherent refactor.
• Migrated all built-in primitives (elemwise, dense mm/mv, conv, sparse mm/mv, and lora) to the dict-based rule API with first-class bias gradient support, each verified element-wise against a BPTT oracle.
• Fixed layout-aware axis handling in conv primitives (1D/2D, NHWC/NCHW, OIHW/HWIO kernel layouts) that previously corrupted gradients on non-default layouts, and fixed non-square dense weight broadcasting in _mm_yw_to_w.
• Eligibility traces are now stored as per-key dicts; the transitional legacy-array adapter has been fully removed.

Performance

... (truncated)

Commits

• d0627d4 ci: fix Publish.yml version check and merge CI workflows (#101)
• 651584b docs: comprehensively cover and restructure docs/apis (#100)
• 2aae3e0 docs: complete public-API docstrings + 0.2.0 changelog (#99)
• 799d59d feat: braintrace.compile() entry point + deprecate v0.1.x legacy shims (#98)
• 01250c3 test/p8 diagnostic exploration (#97)
• 317a4ee test: pin per-cell compiler relation guardrails tied to the cell registry (P7...
• 7486d19 test: verify L4 transform/integration invariance of online algorithms (P6) (#95)
• 8378794 test: verify approximate-class online gradients are direction-aligned with BP...
• a1e1e32 perf: fuse multi-step eligibility-trace roll into the forward scan (#93)
• 43f0825 test: prove exact-class online gradients match BPTT (P4 L3-A/B) (#92)
• Additional commits viewable in compare view

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.

---

Dependabot commands and options

You can trigger Dependabot actions by commenting on this PR:

• @dependabot rebase will rebase this PR
• @dependabot recreate will recreate this PR, overwriting any edits that have been made to it
• @dependabot show <dependency name> ignore conditions will show all of the ignore conditions of the specified dependency
• @dependabot ignore this major version will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)
• @dependabot ignore this minor version will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)
• @dependabot ignore this dependency will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)