Lava Deep Learning 0.5.0

[mgkwill]

Lava Deep Learning v0.5.0 Release Notes

November 9, 2023

What's Changed

• Ensure clamping of delay values during network export and import by @bamsumit in Ensure clamping of delay values during network export and import #215
• Bump tornado from 6.3.2 to 6.3.3 by @dependabot in Bump tornado from 6.3.2 to 6.3.3 #228
• Bump cryptography from 41.0.2 to 41.0.3 by @dependabot in Bump cryptography from 41.0.2 to 41.0.3 #229
• Affine hdf5 export (cuba.Affine layer cannot be exported to hdf5 #221) by @ahenkes1 in Affine hdf5 export (#221) #222
• Added XOR-Regression tutorial. by @ahenkes1 in Added XOR-Regression tutorial. #227
• Spikemoid pr by @Michaeljurado42 in Spikemoid pr #231
• Bump gitpython from 3.1.32 to 3.1.34 by @dependabot in Bump gitpython from 3.1.32 to 3.1.34 #232
• Bump gitpython from 3.1.34 to 3.1.35 by @dependabot in Bump gitpython from 3.1.34 to 3.1.35 #234
• Bump cryptography from 41.0.3 to 41.0.4 by @dependabot in Bump cryptography from 41.0.3 to 41.0.4 #240
• Device parameter for Sigma Dendrite by @bamsumit in Device parameter for Sigma Dendrite #241
• Sparsity netx pr by @Michaeljurado42 in Sparsity netx pr #238
• Bump urllib3 from 1.26.16 to 1.26.17 by @dependabot in Bump urllib3 from 1.26.16 to 1.26.17 #245
• Bump pillow from 9.5.0 to 10.0.1 by @dependabot in Bump pillow from 9.5.0 to 10.0.1 #246
• Update pillow version in pyproject.toml by @PhilippPlank in Update pillow version in pyproject.toml #247
• Set user defined spike_exp level globally when creating netx network by @bamsumit in Set user defined spike_exp level globally when creating netx network #249
• Bump gitpython from 3.1.35 to 3.1.37 by @dependabot in Bump gitpython from 3.1.35 to 3.1.37 #251
• Dev/feature yolo by @bamsumit in Dev/feature yolo #243
• Bump torch requirements by @bamsumit in Bump torch requirements #250
• Updated readme by @bamsumit in Updated readme #254
• Reduce the file size of yolo notebook using mp4 export by @bamsumit in Reduce the file size of yolo notebook using mp4 export #256
• Bump urllib3 from 2.0.6 to 2.0.7 by @dependabot in Bump urllib3 from 2.0.6 to 2.0.7 #257
• YOLO-KP inference by @bamsumit in YOLO-KP inference #262
• Yolo kp part II by @bamsumit in Yolo kp part II #263
• Fix pypi publish in cd.yml by @mgkwill in Fix pypi publish in cd.yml #264

New Features and Improvements

• Lava-dl SLAYER now has extended support for training and inference of video object detection networks and the associated pre and post processing utilities used for object detection. The object detection module is available as lava.lib.dl.slayer.obd. The modules are described below:

  Module	Description
  obd.yolo_base	the foundational model for YOLO object detection training which can be used to build a variety of YOLO models
  obd.models	selected pre-trained YOLO SDNN models which can be fine-tuned for user-specific applications
  obd.dataset	object detection dataset library (will be progressively extended)
  obd.bbox.metrics	modules to evaluate object detection models
  obd.{bbox, dataset}.utils	utilities to manipulate bounding boxes and dataset processing including frame visualization and video export

  Extensive tutorials for

  • YOLO SDNN training for video object detection,
  • YOLO SDNN inference on GPU, and
  • YOLO SDNN inference on Lava and Loihi

  are also available.

  In addition, the lava-dl SLAYER tutorials now include XOR regression tutorial as a basic example to get started with lava-dl training.

  Finally, lava-dl SLAYER now supports SpikeMoid loss, the official implementation of the spike-based loss introduced in

  Jurado et. al., Spikemoid: Updated Spike-based Loss Methods for Classification.

  which enables more advanced tuning of SNNs for classification.

• Lava-dl NetX now supports users to configure inference of fully connected layers using sparse synapse instead of the default dense synapse. This allows the network to leverage the compression offered by sparse synapse if the fully connected weights are sparse enough. It is as simple as setting sparse_fc_layer=True when initializing a netx.hdf5.Network. netx.hdf5.Network also supports global control of spike exponent (the fraction portion of spike message) by setting spike_exp keyword. This allows users to control the network behavior in a more fine-grained manner and potentially avoid data overflow on Loihi hardware.

  In addition, lava-dl NetX now includes sequential modules netx.modules. These modules allow the creation of PyTorch style callable constructs whose behavior is described in the forward function. In addition, these sequential modules also allow the execution of non-critical, but expensive management between calls in a parallel thread so that the execution flow is not blocked.

  netx.modules.Quantize and netx.modules.Dequantize are now pre-built to allow for consistent quantization and dequantization to/from the fixed precision representation in the NetX network. Their usage can be seen in the YOLO SDNN inference on Lava and Loihi tutorial.

Bug Fixes and Other Changes

• Lava-dl SLAYER is now Torch 2.0 compatible allowing our users to use advanced Torch 2.0+ features.
• Fixes have been included that enable hdf5 export of affine block and proper handling of out-of-bound delays during hdf5 export in lava-dl SLAYER.

Breaking Changes

• No breaking changes in this release.

Known Issues

• No known issues in this release.

New Contributors

• @ahenkes1 made their first contribution in Affine hdf5 export (#221) #222

Full Changelog: v0.4.0...v0.5.0

---

This discussion was created from the release Lava Deep Learning 0.5.0.