OptimaMT ru → uk (Russian to Ukrainian)

Drop-in copy of Xenova/opus-mt-ru-uk at revision 6d8a744ed6ea with a lighter merged decoder: ONNX Runtime's fast mode (disable_quant_qdq = 1, 2.3–2.5× faster) at half the memory. Same weights, same tokens. Part of OptimaMT: all 49 pairs, named OptimaMT/<src>-<tgt>.

Used in production:

  • OpenCircle, a desktop screen translator for Windows and Linux, ships these models since v0.1.4.
Memory after load Linux Windows Identical tokens
disable_quant_qdq = 1 494 → 268 MiB (-46 %) 470 → 244 MiB (-48 %) yes
ONNX Runtime defaults 265 → 208 MiB (-21 %) 189 → 159 MiB (-16 %) yes

FLORES-200 devtest (1,012 sentences): the same translations as the original, BLEU 20.36, chrF 49.51 (greedy).

ONNX Runtime 1.31.1, CPU, 4 threads (Linux; Windows in a VM). Interactive results: OptimaMT/results · method: REPORT.md.

Use

import { pipeline } from "@huggingface/transformers";

const translator = await pipeline("translation", "OptimaMT/ru-uk", {
  dtype: "q8",
  session_options: { extra: { session: { disable_quant_qdq: "1" } } },
});
const [out] = await translator("Маяк стоит на берегу моря.");
console.log(out.translation_text);

Python (translate.py: onnxruntime, sentencepiece, numpy):

python translate.py OptimaMT/ru-uk "Маяк стоит на берегу моря."

Files

Only onnx/decoder_model_merged_quantized.onnx is rebuilt; every other file is the original, byte for byte. Vocabulary: 57,777 tokens.

Licence and credits

  • Weights: OPUS-MT by the University of Helsinki, CC BY 4.0.
  • ONNX export: Xenova/opus-mt-ru-uk at revision 6d8a744ed6ea.
  • Modified: the decoder's output projection was moved out of its two branches (graph change only). CC BY 4.0.
Downloads last month
33
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for OptimaMT/ru-uk

Quantized
(1)
this model

Space using OptimaMT/ru-uk 1

Collection including OptimaMT/ru-uk