Lava 0.3.0

[mgkwill]

Lava 0.3.0 includes bug fixes, updated documentation, improved error handling, refactoring of the Lava Runtime and support for sigma delta neuron enconding and decoding.

New Features and Improvements

• Added sigma delta neuron enconding and decoding support (PR #180, Issue #179)
• Implementation of ReadVar and ResetVar IO process (PR #156, Issue #155)
• Added Runtime handling of exceptions occuring in ProcessModels and the Runtime now returns exeception stack traces (PR #135, Issue #83)
• Virtual ports for reshaping and transposing (permuting) are now supported. (PR #187, Issue #185, PR #195, Issue #194)
• A Ternary-LIF neuron model was added to the process library. This new variant supports both positive and negative threshold for processing of signed signals (PR #151, Issue #150)
• Refactored runtime to reduce the number of channels used for communication(PR #157, Issue #86)
• Refactored Runtime to follow a state machine model and refactored ProcessModels to use command design pattern, implemented PAUSE and RUN CONTINOUS (PR #180, Issue #86, Issue #52)
• Refactored builder to its own package (PR #170, Issue #169)
• Refactored PyPorts implementation to fix incomplete PyPort hierarchy (PR #131, Issue #84)
• Added improvements to the MNIST tutorial (PR #147, Issue #146)
• A standardized template is now in use on new Pull Requests and Issues (PR #140)
• Support added for editable install (PR #93, Issue #19)
• Improved runtime documentation (PR #167)

Bug Fixes and Other Changes

• Fixed multiple Monitor related issues (PR #128, Issue #103, Issue #104, Issue #116, Issue #127)
• Fixed packaging issue regarding the dataloader for MNIST (PR #133)
• Fixed multiprocessing bug by checking process lineage before join (PR #177, Issue #176)
• Fixed priority of channel commands in model (PR #190, Issue #186)
• Fixed RefPort time step handling (PR #205, Issue #204)

Breaking Changes

• No breaking changes in this release

Known Issues

• No support for Intel Loihi
• CSP channels process communication, implemented with Python multiprocessing, needs improvement to reduce the overhead from inter-process communication to approach native execution speeds of similar implementations without CSP channel overhead
• Virtual ports for concatenation are not supported
• Joining and forking of virtual ports is not supported
• A Monitor process cannot monitor more than one Var/InPort of a process, as a result multi-var probing with a singular Monitor process is not supported
• Limited API documentation

What's Changed

• Fixing multiple small issues of the Monitor proc by @elvinhajizada in Fixing multiple small issues of the Monitor proc #128
• GitHub Issue/Pull request template by @mgkwill in GitHub Issue/Pull request template #140
• Fixing MNIST dataloader by @tihbe in Fixing MNIST dataloader #133
• Runtime error handling by @PhilippPlank in Runtime error handling #135
• Reduced the number of channels between service and process (fix typo Proces -> Process in Readme.md #1) by @ysingh7 in Reduced the number of channels between service and process (#1) #157
• TernaryLIF and refactoring of LIF to inherit from AbstractLIF by @srrisbud in TernaryLIF and refactoring of LIF to inherit from AbstractLIF #151
• Proc_params for communicating arbitrary object between process and process model by @bamsumit in Proc_params for communicating arbitrary object between process and process model #162
• Support editable install by @matham in Support editable install #93
• Implementation of ReadVar and ResetVar IO process and bugfixes for LIF, Dense and Conv processes by @bamsumit in Implementation of ReadVar and ResetVar IO process and bugfixes for LIF, Dense and Conv processes #156
• Refactor builder to module by @mgkwill in Refactor builder to module #170
• Use unittest ci by @mgkwill in Use unittest ci #173
• Improve mnist tutorial by @srrisbud in Improve mnist tutorial #147
• Multiproc bug by @mgkwill in Multiproc bug #177
• Refactoring py/ports by @PhilippPlank in Refactoring py/ports #131
• Adds runtime documentation by @joyeshmishra in Adds runtime documentation #167
• Implementation of Pause and Run Continuous with refactoring of Runtime by @ysingh7 in Implementation of Pause and Run Continuous with refactoring of Runtime #171
• Ref port debug by @PhilippPlank in Ref port debug #183
• Sigma delta neuron, encoding and decoding support by @bamsumit in Sigma delta neuron, encoding and decoding support #180
• Add NxSDKRuntimeService by @mgkwill in Add NxSDKRuntimeService #182
• Partial implementation of virtual ports for PyProcModels by @mathisrichter in Partial implementation of virtual ports for PyProcModels #187
• Remove old runtime_service.py by @mgkwill in Remove old runtime_service.py #192
• Fixing priority of channel commands in model by @PhilippPlank in Fixing priority of channel commands in model #190
• Virtual ports between RefPorts and VarPorts by @mathisrichter in Virtual ports between RefPorts and VarPorts #195
• RefPort's sometimes handled a time step late by @PhilippPlank in RefPort's sometimes handled a time step late #205
• Fixed reset timing offset by @bamsumit in Fixed reset timing offset #207
• Update README.md by @mgkwill in Update README.md #202
• Virtual ports no longer block Process discovery in compiler by @mathisrichter in Virtual ports no longer block Process discovery in compiler #211
• Remove pybuilder, Add poetry by @mgkwill in Remove pybuilder, Add poetry #215
• Added wait() to refvar unittests by @bamsumit in Added wait() to refvar unittests #220
• Update Install Instructions by @mgkwill in Update Install Instructions #218

Thanks to our Contributors

Intel Corporation: All contributing members of the Intel Neuromorphic Computing Lab

Open-source community: (Ismael Balafrej, Matt Einhorn)

New Contributors

• @tihbe made their first contribution in Fixing MNIST dataloader #133
• @ysingh7 made their first contribution in Reduced the number of channels between service and process (#1) #157
• @matham made their first contribution in Support editable install #93

Full Changelog: v0.2.0...v0.3.0

---

This discussion was created from the release Lava 0.3.0.