---
license: apache-2.0
library_name: pytorch
pipeline_tag: audio-classification
tags:
  - hlm-micro
  - polynomial-hopfield
  - edge-ai
  - keyword-spotting
  - speech-commands
  - tinyml
  - real-data
datasets:
  - speech-commands-v2
---

# HLM-Micro Keyword Spotting v1 - Speech Commands

HLM-Micro Keyword Spotting v1 is a compact polynomial-Hopfield keyword spotter trained on the Google Speech Commands V2 12-class TinyML-style subset.

## Results

| Field | Value |
|---|---:|
| Parameters | 184,449 |
| Classes | 12 |
| Reported best validation accuracy | 92.09% |
| Final validation accuracy | 92.01% |
| Best checkpoint | epoch 6 of 8 |

Classes: `yes`, `no`, `up`, `down`, `left`, `right`, `on`, `off`, `stop`, `go`, `_silence_`, `_unknown_`.

This is a realistic small keyword-spotting result, not a state-of-the-art accuracy claim. The test split is intentionally not claimed here.

## Files

| File | Purpose |
|---|---|
| `model.pt` | Sanitized model-only PyTorch checkpoint |
| `config.json` | Public architecture, task, classes, and metric metadata |
| `metrics.jsonl` | Training/evaluation metrics from the local run |

## Intended Use

- Research on compact keyword spotting.
- TinyML and edge-audio experiments.
- Baseline for adding replayable audit metadata to edge classifiers.

## Limitations

- Metric reported here is validation accuracy, not final held-out test accuracy.
- Not optimized with the full augmentation/training recipe used by top keyword-spotting systems.
- Not a wake-word product or production speech interface.