Repository navigation
Releases
OptimaMT 1.1.0: technical report
Adds the technical report OptimaMT: Lighter OPUS-MT Translation Models for ONNX Runtime (v1.0, November 2026): how the models work, why ONNX Runtime's fast mode doubled their memory, how the rebuild removes the duplicate, and the full measurements on Linux, Windows and in the browser.
PDF: report/OptimaMT-Technical-Report.pdf
The models themselves are unchanged from 1.0.0: huggingface.co/OptimaMT.
OptimaMT 1.0.0
First release: 49 OPUS-MT translation pairs as ONNX models that use about half the memory in ONNX Runtime's fast mode, with identical translations.
Models: Hugging Face collection · one repository per pair, a drop-in folder for transformers.js and ONNX Runtime.
What changed
In the original int8 exports, the decoder builds its output layer in both branches of its If node, so ONNX Runtime's fast mode (session.disable_quant_qdq = 1) folds it to float twice. OptimaMT moves that subgraph in front of the branches, so it is folded once. No weight is changed, retrained or re-quantized.
Measured
| Linux | Windows | |
|---|---|---|
| Fast mode memory, original → OptimaMT (median, 49 pairs) | 525 → 281 MiB | 500 → 257 MiB |
| Defaults memory, original → OptimaMT | 277 → 217 MiB | 196 → 166 MiB |
| ms/step, defaults → fast mode (median, 5 pairs) | 17.3 → 6.9 | 21.3 → 9.1 |
| Speed change from the rebuild | none | none |
- Same translations: FLORES-200 devtest, 49 pairs × 2 settings × 1012 sentences: identical output and identical BLEU and chrF in 98 of 98 runs.
- Browser (transformers.js): fast mode 1288 → 1050 MiB after load (−18 %); no saving with the defaults.
- The 2.3–2.5× speed-up is ONNX Runtime's fast mode; OptimaMT removes its memory cost.
Full method and every number: REPORT.md · interactive results and a live demo: OptimaMT/results
Licences
Code: MIT. Models: CC BY 4.0, OPUS-MT by Helsinki-NLP, ONNX exports by Xenova, modified (see NOTICE.md).