v1.0.7 Type checks and backward compatibility

Most importantly, this release supports default (None) arguments for the newly introduced v_reset parameter in neuron models.

What's Changed

• Update supported simulators docs by @benkroehs in #156
• Mention jaxsnn in NIR documentation by @benkroehs in #158
• fix: URLs to jaxsnn and BrainScaleS by @benkroehs in #160
• docs: Update README badges - Collaboration Network by @neural-loop in #161
• Fix docs regarding NIR support in Lava by @benkroehs in #163
• Added lava to NIR Oxford example and documentation by @alejandroggener in #165
• Adds an option to omit the type checking when reading a NIR graph by @Jegp in #167
• Add overview of supported NIR nodes per framework by @benkroehs in #159
• fix: Example for snntorch by @benkroehs in #170
• Fix GitHub Actions workflow for supported primitives by @benkroehs in #169
• Rename spike to threshold in primitives by @Jegp in #172
• feat(hxtorch): Mention hxtorch in documentation by @benkroehs in #171
• Permit optional v_reset values by @Jegp in #173
• Added some units to the primitives by @Jegp in #166

New Contributors

• @neural-loop made their first contribution in #161
• @alejandroggener made their first contribution in #165

Full Changelog: v1.0.6...v1.0.7

v1.0.6 Serialization and type checking improvements

This release fixes some deserialization errors, improves type checking in graphs (while cleaning up the code, nice!), and updates the dependencies.

Thanks a lot to @mrontio and @fabio-innatera 🎉

What's Changed

• Add reset potential for IF, LIF and CubaLIF by @benkroehs in #129
• Remove duplicated _forward_type_inference method from NIRGraph by @fabio-innatera in #139
• CI: Check code formatting by @fabio-innatera in #147
• pre-commit hook: bump docformatter to fix error by @fabio-innatera in #144
• Fix deserialization of legacy .nir files by @benkroehs in #143
• Prevent from_dict from modifying non-local scope by @mrontio in #148
• Black formatter version fix to 25.1.0 by @mrontio in #151
• Fix black to 25.1.0 in pre-commit too by @mrontio in #152
• Enhanced type inference and checking, called from NIRGraph.init by @fabio-innatera in #153

New Contributors

• @fabio-innatera made their first contribution in #139
• @mrontio made their first contribution in #148

Full Changelog: v1.0.5...v1.0.6

Type checking

What's Changed

• neuroir.org website by @Jegp in #104
• update snntorch example by @jeshraghian in #100
• 'int' object has no attribute 'item' on conv shapes calculation (issue 96) by @stevenabreu7 in #101
• add lava docs by @stevenabreu7 in #105
• Fix S2 SCNN notebook by @matjobst in #108
• Add rockpool example to docs by @bauerfe in #103
• Updated docs with snntorch to Norse example and dev docs by @biphasic in #109
• update spinnaker2 docs: add infer_types() to example by @matjobst in #110
• Add metadata to missing nodes by @bvogginger in #112
• README typo fixed in 03_rnn by @VitF in #113
• Renamed surrogate gradient. Added Identity by @Jegp in #114
• Simplified docs and added example page by @Jegp in #116
• Fixed dollarmath docs formatting by @Jegp in #119
• fix: Issues in examples by @benkroehs in #118
• Throw AttributeError for NIRNode constructor by @hanleyc01 in #122
• Feature check types by @Jegp in #124
• General docs improvements by @Jegp in #125
• Add CuBaLI neuron by @benkroehs in #126
• Bumped sphinx version by @Jegp in #128
• Update dependencies so tests run smoothly on initial clone by @katywarr in #127
• Feature web by @Jegp in #134

New Contributors

• @bauerfe made their first contribution in #103
• @VitF made their first contribution in #113
• @benkroehs made their first contribution in #118
• @hanleyc01 made their first contribution in #122
• @katywarr made their first contribution in #127

Full Changelog: v1.0.4...v1.0.5

v1.0.4 Reduction of size of releases

What's Changed

• Remove docs and paper data/code from release by @matjobst in #99

Full Changelog: v1.0.3...v1.0.4

v1.0.3 Metadata and code reorganization

What's Changed

• Created example usage notebook for Spyx. by @kmheckel in #77
• Reorganise ir and add auto serialisation by @alexfanqi in #78
• cleanup spinnaker2 scnn example by @matjobst in #80
• Add AvgPool2d by @bvogginger in #84
• fixed NIRNode import issue #85 by @matjobst in #86
• Added roadmap draft by @Jegp in #82
• Added metadata node to all NIRNodes by @Jegp in #79
• Paper folder reorganization by @stevenabreu7 in #88
• Updates for the paper: renaming + lava refactoring by @stevenabreu7 in #90
• Paper: plotting by @stevenabreu7 in #93
• Updated SpiNNaker2 example in docs by @matjobst in #94
• Update s2 paper notebooks by @matjobst in #95
• Paper: add CNN pretraining by @stevenabreu7 in #91
• Paper: refactor snnTorch experiments by @stevenabreu7 in #92
• Automatically read version number from git by @Jegp in #97

New Contributors

• @kmheckel made their first contribution in #77
• @alexfanqi made their first contribution in #78

Full Changelog: v1.0.1...v1.0.3

NIR v1.0.1

This minor release aligns and fixes licensing for NIR. It also adds a CITATION.cff file

NIR v1.0

This release marks the first stable version of NIR, used in the paper Neuromorphic Intermediate Representation: A Unified Instruction Set for Interoperable Brain-Inspired Computing

Thank you to everyone for the impressive amount of work and effort! <3
@stevenabreu7 @matjobst @bvogginger @VitF @sheiksadique @tcstewar @DylanMuir @biphasic @kmheckel @bauerfe

What's Changed

• 53 ir for sumpool2d by @sheiksadique in #54
• Upgrade SpiNNaker2 example to v0.2.0 by @bvogginger in #57
• implemented .inputs and .outputs properties for NIRGraphs by @matjobst in #30
• Specify spatial shapes in convolutions by @Jegp in #59
• Documentation update by @biphasic in #68
• Bug fixes: nir.Output type inference & CubaLIF.w_in read/write by @stevenabreu7 in #71
• Update README.md by @jeshraghian in #73
• Paper by @Jegp in #72

New Contributors

• @bvogginger made their first contribution in #57
• @biphasic made their first contribution in #68
• @jeshraghian made their first contribution in #73

Full Changelog: v0.2...v1.0

Input/output types and general improvements

This release features introduces input/output types (#50), additional examples (#32, #42), quality-of-life improvements (#43), licensing, and fixes.

What's Changed

• Doc/developers by @sheiksadique in #33
• Example/sinabs by @sheiksadique in #32
• fixes in sinabs example by @sheiksadique in #35
• Added version to IR nodes and fixed linting by @Jegp in #34
• formatting conflicts fixed by @sheiksadique in #38
• fixed dev_requirements.txt by @matjobst in #40
• edges as str when reading model from hdf5 file by @matjobst in #41
• added spinnaker2 example by @matjobst in #42
• meaningful node naming when using from_list() by @matjobst in #43
• Updated docs and frameworks by @Jegp in #45
• Add BSD-3 LICENSE by @stevenabreu7 in #46
• add input/output shape annotations for all nodes by @stevenabreu7 in #50
• Improved argument parsing for shapes by @Jegp in #51
• Improved docs on types and renamed input_shape to input_type by @Jegp in #52

New Contributors

• @sheiksadique made their first contribution in #33

Full Changelog: v0.1...v0.2

NIR v0.1

This release wraps up the work at Telluride, and we can now declare NIR in v0.1! 🥳 🎈

Concretely, since 0.0.6, we added

• String indices instead of numerical indices
• A flatten node
• Minor corrections in node specification

Recursive node definition and current-based LIF model

This release adds

• Recursive node definition where everything in NIR is a subclass of a NIRNode, including a NIRGraph
• A current-based LIF model (CubaLIF)
• Standardized renaming (weights -> weight)