0.3.1

New Features

• Added Branches and Dup components that generalize Parallel
• Added predicate-based termination criterion in Loop, enabling run-to-energy mode
• Even lazier auto-embedding QPU sampler
• Added parallel tempering workflow examples
• Importable workflows for Kerberos, ParallelTempering/HybridizedParallelTempering
  and SimplifiedQbsolv

Fixes

• Looser traits enforcing (run-time only, skip construction-time checks)
• RandomConstraintDecomposer edge cases
• RoofDualityComposite energy calculation

0.3.0

New Features

• Tabu/SimulatedAnnealing samplers now use input state samples to initialize
• Racing context propagated to children via runopts
• Added stoppable class decorator to simplify implementing thread-safe stop behaviour
• Added RoofDualifyDecomposer
• Exposed all SA parameters in SimulatedAnnealing sampler
• Exposed non-interruptible Tabu timeout in Kerberos

Fixes

• Longer default runtime in Tabu
• Better handling of larger problems by Kerberos

Changes

• endomorphic option removed from RacingBranches and ParallelBranches in favor of explicit Identity block
• sweeps renamed to num_sweeps in SA samplers

0.2.1

Changes since 0.2.0:

New components

• MergeSamples that vertically stacks SampleSets from multiple States
• SliceSamples that slices the set of samples (and trivially truncates)

Performance improvements

• Multi-threaded support in Tabu and Simulated Annealing samplers

General

• Added States.first property

Fixes

• Compatibility with dimod==0.8.8 (breaking change of SampleSet.samples() behaviour)

0.2.0

Changes since 0.1.4:

New components

• Reverse annealing sampler: ReverseAnnealingAutoEmbeddingSampler
• Best-state tracking block: TrackMin
• GreedyPathMerge sample composer that finds the best sample on a greedy path between two samples
• Identity runnable

Looping

• Loop refactored and generalized. Loops now come in two flavours: LoopUntilNoImprovement and LoopWhileNoImprovement.
• Loops are now infinite by default, but all accept max_iter, and max_time (a new terminating condition).
• Infinite loops are now interruptible/stoppable
• SimpleIterator deprecated.

Decomposition

• Structural decomposition added to EnergyImpactDecomposer. Two new modes of traversal are Breadth-First-Search and Priority-First-Search, both energy impact seeded.
• Unwinding context implicitly propagated to child runnables
• Synchronous Unwind

General

• Added deferred runtime parameters defined during Runnable construction
• Added State.from_(sub)problem factories
• Added SampleSet.hstack and SampleSet.vstack utility functions
• Added Dialectic Search example workflow

Fixes

• Blocks don't start after stop
• Performance counter access fails before the first run

0.1.4

Changes since 0.1.3:

• EnergyImpactDecomposer rewritten as a rolling, streaming, decomposer
• consistent use of size in all decomposers
• new blocks: Unwind and Reduce
• runnables are picklable
• new module hybrid.concurrency (with new process executor)
• profiling refactor: accumulating timers and integer counters
• misc bugfixes

0.1.3

Changes since 0.1.2:

• adaptive de-structuring of child samplers in QPU auto embedding sampler
• stop()/halt() interface change
• bugfixes, docs updates

0.1.2

Changes since v0.1.1:

• relaxed energy impact decomposition defaults (wider search)
• require latest dwave-tabu with bug fixes for trivial edge cases

0.1.1

Changes from v0.1.0:

• improved multi-state I/O traits
• new runnables: Map, Lambda
• code re-org (top-level imports, renamed some flow control runnables)
• better tests, docs, and CI

0.1.0

First public alpha release.