Sinabs

Latest version: v2.0.2

Safety actively analyzes 714815 Python packages for vulnerabilities to keep your Python projects secure.

Scan your dependencies

Page 2 of 6

1.2.4

------

* replace torch.inf with math.inf
* ignore data and cache directories
* added test
* Take averace across samples within batch when collecting synops
* Unit test for synops counter with batch size > 1
* added noconfirm
* added update keyring to gitlab-ci
* renamed test

1.2.3

------

* fix synops/s for Linear layers when no Squeeze layers are used in network
* update pre-commit config

1.2.2

------

* expand saved input in spiking layers for analyzer
* derive spiking layer stats from single saved output tensor and also save input
* Speck2fDevKit support added
* add n\_neurons to model\_stats
* Test added
* Create random indices tensor on the same device as the buffer
* update samna version requirements
* update ChangeLog
* add speck2f module devkit support and remove speck2f characterization board support

1.2.1

------

* distinguish between accumulated and mini batch stats for firing rates
* distinguish between accumulated and mini batch statistics in SNNAnalyzer for synops
* only compute connection\_map once in SynopsHook
* detach synops accumulation inbetween mini batches for SNNAnalyzer

1.2.0

------

* update release notes
* make sure deconvolve outputs same size as conv input in SNNAnalyzer
* Implementation a test that will create a single neuron on a single core, that receives a single event and outputs a single event
* edit visualizer.md
* Added the visualizer documentation to the index.rst
* Replacing the notebook which cannot spawn a samnagui window with a markdown file
* Visualizer example updated with the window scale parameter
* Window scale parameter is now passed to the constructor
* load model onto CPU and change warning message for JIT compilation
* exclude Samna visualiser generated file
* Samna version requirement incremented to cover the latest changes
* Notebook forgotten connect method added. Readout visualizer only supports .png images. Information regarding to this has been added to the documentation
* Example notebook with explanations added
* Minor bugfixes in readout layout and feature name automatic setting to numbers
* A fully working model
* (WIP) Apart from one issue on not getting any output from the chip it works
* make it possible to use IAF with from\_model
* (WIP) blocked by problems in samna. Turns out and were not tested in samna prior to release. Sys-Int team will be fixing it, then I can fix in samna. Then all the features, but plotting the readoutpins will function
* Tests updated to cover the values, so that we can confirm that the states indeed have not been changed but updated
* remove outdated references to SpkConverter and SNNSynopCounter
* Clarify scaling factor
* clarify in docstring why we're using transpose convolution and scaling factor for AvgPool layer in SNNAnalyzer
* add tutorial text about how to scale weights for conversion
* Clarify in-code documentation
* Add test for changing the trailing dim, which should not happen
* Add documentation to the mismatch handling method
* For non-squeeze layers in the case of input with different batch sizes are received, the states are going to be reshaped accordingly. The change will be randomly sampling states from created by samples in the existing batch repeatedly
* Test norm\_input for LIF
* Test norm\_input for ExpLeak
* added docstring description for SNNAnalyzer
* fix errors in using\_radout\_layer.ipynb
* merge conflicts
* update ChangeLog
* modify config making tests can cover different types of devkits
* add speck2f characterization board support
* Waiting for sys-int team to provide a JitDvsToViz template directly into samna itself. I implement it myself if it is not ready on monday. Anything related to readout has been removed, as well as the JitReadoutNode. This functionality already exists in under . The API that uses strings in the front end to create plots have been changed with one that uses boolean flags to create readout and power monitor plots
* fix errors in tests after init DVSLayer in any case
* make DVSLayer can be initnitialized in anycase
* (WIP) For readout we have been using something that has been out of performance, this should be replaced with a proper functionality. In order to support all the chips and cameras ToViz converter filters should be Jitted, for which there is no template in right now
* (WIP) Dynapcnn Visualizer initial implementation
* black formatted
* renamed test
* update ChangeLog
* update figure
* add a time.sleep statement after dynapcnn network reset-states to wait the config take effect
* modify speck2e test to deploy dummy network
* added test for killbit
* test updated to check for type
* added checks if the network is empty
* add spike count visualization notebook into tutorials.rst
* fix typo in the notebook of visualize spike count
* delete chip top level figure
* change the speck top level figure path
* change top level figure
* add notebooks into docs/source/tutorials.rst
* update the noteboob
* Update requirements.txt
* notebook and image for spike counts visualization
* modify the parent folder name identical to the notebook
* modify the image load path
* make spike\_threshold and min\_v\_mem a parameter
* fix test which seems to have relied on older version of PyTorch
* fix wrong calculation formula in remapping neruon index function
* use deconvolve module to compute connection map for accurate synops
* fix conflict when merge develop into 34-add-readout-layer-usage-instruction-notebook
* add samna graph plot for power monitor notebook
* add notebook for power monitoring
* update the readout-layer intro notebook
* add FAQ and known dev kit bugs
* rm change in tutorials.rst to avoid merge conflict
* add seaborn to docs requirements
* replace plotly image with seaborn
* update synops\_loss\_ann notebook
* use consistent American spelling of analyzer

1.1.6

------

* make min\_vem and spike\_thresholds parameters do that they're included in state\_dict
* Detach recordings before plotting
* Allow backpropagating through recorded states
* add figures for overview notebook
* add overview tutorial notebook for speck/dynapcnn
* removed all samna checks from imports
* imports fixed and type annotation for Tuple corrected
* add notebook about dvs input visualization
* added standardize\_device\_id before all IO methods
* add codecov badges
* pytest command fixed for coverage
* added path to installation for coverage
* try hard-coded path for coverage
* added conditional deplopment
* rerunning test in deploy stage
* custom stage added
* added dependenct to unitttests
* two stage coverage deployment
* changed command for acquiring code coverage
* coverage path fixed
* remove gallery from docs as not needed at the moment
* added galary template, updated config -> sinabs
* add coverage config file that specifies the folder to include in the coverage, excluding the test folder
* remove some whitespace in the gitlab ci config
* use codecov token so that hopefully CI is pushed via gitlab.com
* update gitlab CI to produce html coverage report for pages
* try using gitlab pages for coverage report
* adding test coverage job to CI pipeline according to Python template from gitlab.com
* remove gallery from docs as not needed at the moment
* add figures for the readout layer intro notebook
* add readout layer usage example notebook
* added galary template, updated config -> sinabs
* init tutorial file
* Fix to the issue
* Change last uint
* Remove comment
* Uint64 to int
* Check the timestamp and delay factor for negative values
* Casting to int after conversion
* Do not sleep while resetting the states when the function is called. Stop the input graph, while the neuron values are written. Sleep for a bit and then start the graph again
* Fix to the bug

Page 2 of 6

© 2025 Safety CLI Cybersecurity Inc. All Rights Reserved.