SNN Model Checkpoints for Keyword Recognition on Loihi 2

Balázs Mészáros provides a 534.9 MB dataset of trained Spiking Neural Network model checkpoints for the paper 'A Complete Pipeline for deploying SNNs with Synaptic Delays on Loihi 2'. The checkpoints are in numpy format and were trained on the Spiking Heidelberg Digits and Spiking Speech Commands datasets using mlGeNN 2.3.0. The dataset was last updated on 2026-05-07.

Use Cases

• Deploying pre-trained SNN models for keyword recognition based on the provided checkpoints.
• Benchmarking SNN performance and energy efficiency on neuromorphic hardware based on the described pipeline.
• Comparing SNN architectures with and without learned synaptic delays based on the different model configurations.
• Transferring trained SNN models to other Python-based frameworks based on the numpy format.

Strengths

• Checkpoints are provided in numpy format for easy use in any Python-based SNN framework.
• Model configurations include variations with and without learned synaptic delays for comparative analysis.
• Dataset size is 534.9 MB, indicating a substantial collection of model parameters.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Associated with the paper 'A Complete Pipeline for deploying SNNs with Synaptic Delays on Loihi 2' by Balázs Mészáros.

Collection Method

Models were trained using mlGeNN 2.3.0 on the Spiking Heidelberg Digits and Spiking Speech Commands datasets.

Freshness

Last updated 2026-05-07 10:50:22; freshness should be verified.