Published June 3, 2026
| Version v0.3.0
Software
Open
Atome LM: a tiny ternary language model for microcontroller deployment
Authors/Creators
Description
Initial public release. Routed-ternary tiny language model with bit-exact Python <-> C99 inference engine for microcontroller-class deployment. Apache-2.0. Honest scope: wins at ~60K params (6.31 vs 8.12 ppl, -22%), loses to vanilla FP32 at 944K; MCU parity is QEMU Cortex-M3 only. This release is archived to Zenodo for a citable DOI.
Notes
If you use this software, please cite it as below.
Files
TilelliLab/atome-lm-v0.3.0.zip
Files
(7.8 MB)
| Name | Size | Download all |
|---|---|---|
|
md5:5b731dc2b7a1dedfd691d941c6fc806b
|
7.8 MB | Preview Download |
Additional details
Related works
- Is supplement to
- Software: https://github.com/TilelliLab/atome-lm/tree/v0.3.0 (URL)
Software
- Repository URL
- https://github.com/TilelliLab/atome-lm