Add rf and rf_iz neurons to lava #378
Conversation
|
I have been running into the following error when I try to run During handling of the above exception, another exception occurred: Traceback (most recent call last): Does this command only work on linux? |
…es the same method
|
It seems not to work with Windows. You can look at the linting issues here: https://github.com/lava-nc/lava/actions/runs/3204188629/jobs/5235186915 |
|
BTW @Michaeljurado24, can you email me, I think it's better we have a 1:1 meeting. |
…s. Fixed formatting
Sure thing. I could not find your email so I sent you a 'connect' request on LinkedIn with my email so we can collaborate more closely on this |
…st code for rf floating point process model
…ntly hanging after merge from main
|
I am not exactly sure why the RF neuron no decay test is failing for Ubuntu rn since i tried it on a Ubuntu machine and it passed... I have a suspicion it is caused by a slight floating point difference and will look into it. |
|
I saw that @bamsumit accepted the PR 4 days ago. |
An approval just means that the feature and the code quality is okay. All unit tests must still pass, before we can merge the PR. I restarted the unit tests. |
… unit test more robust
Thank you all as well for the help and for answering both my logistical and technical questions. |
* Added code for floating point rf and rf_iz neurons. Need to move toward fixed point * Fixed RF process. * Added fixed point for rf neurons. Floating point rf behavior looks weird * Added modified rf activation for floating point * fixed fixed_pt rf_iz neuron * moved location of scale threshold function to be in line with lif * moving scale_threshold back to abstract rf model class since rf_iz uses the same method * added state_exp. Removed most of the deepcopies. Removed some defaults. Fixed formatting * removed accidental print statement * Refactored rf and rf_iz class to look like sdn. Added comments and test code for rf floating point process model * Cleaned up test_resonator_process.py * Added fixed point unit tests for rf and rf_iz * Lava implementation of RF neurons matches lava-dl. RF unit test currently hanging after merge from main * Removed temporary testing files. Fixed liniting errors * Fixed typo * Clarified comments * increase specificity of unit tests * Added comments and cleaned up test code. Added additional fixed point test for rf neurons * Rf unit test fails to spike due to floating point error on some machines. Fixed the error * Increase magnitude of input to rf float no decay unit test to encourage spiking * Made floating point rf test possibly more robust to floating point errors * fixed linting errors * Removed useless abs() op in unit test. Added small enhacement to make unit test more robust * Slight change to test float no decay to test for periodic spiking * Add abs value to statement checking for proper spike periodicity for rf neurons * Simplify unit test and see if it passes run ci Co-authored-by: Jurado <[email protected]> Co-authored-by: PhilippPlank <[email protected]>
Issue Number:
Objective of pull request: Add rf and rf_iz neurons to lava
Pull request checklist
Your PR fulfills the following requirements:
flakeheaven lint src/lava tests/) and (bandit -r src/lava/.) pass locallypytest) passes locallyPull request type
Please check your PR type:
What is the current behavior?
-No support for rf neurons in lava
What is the new behavior?
-There should be a floating point and loihi bit accurate implementation of rf neurons in lava. This implementation should match the lava-dl rf neuron implementations.
Does this introduce a breaking change?
Supplemental information
This issue is a result of an ongoing discussion in which I ask whether or not rf neurons will be added to lava.
Please see proof of concept example here