OptimaMT id β†’ en (Indonesian to English)

Drop-in copy of Xenova/opus-mt-id-en at revision c38ef36c843b with a lighter merged decoder: ONNX Runtime's fast mode (disable_quant_qdq = 1, 2.3–2.5Γ— faster) at half the memory. Same weights, same tokens. Part of OptimaMT: all 51 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 = 1 472 β†’ 258 MiB (-45 %) 449 β†’ 235 MiB (-48 %) yes
ONNX Runtime defaults 254 β†’ 201 MiB (-21 %) 181 β†’ 155 MiB (-15 %) yes

FLORES-200 devtest (1,012 sentences): the same translations as the original, BLEU 32.89, chrF 60.75 (greedy).

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

Use

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

const translator = await pipeline("translation", "OptimaMT/id-en", {
  dtype: "q8",
  session_options: { extra: { session: { disable_quant_qdq: "1" } } },
});
const [out] = await translator("Mercusuar berdiri di pantai.");
console.log(out.translation_text);

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

python translate.py OptimaMT/id-en "Mercusuar berdiri di pantai."

Files

Only onnx/decoder_model_merged_quantized.onnx is rebuilt; every other file is the original, byte for byte. Vocabulary: 54,796 tokens.

Licence and credits

  • Weights: OPUS-MT by the University of Helsinki, CC BY 4.0.
  • ONNX export: Xenova/opus-mt-id-en at revision c38ef36c843b.
  • Modified: the decoder's output projection was moved out of its two branches (graph change only). CC BY 4.0.
Downloads last month
96
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support

Model tree for OptimaMT/id-en

Quantized
(1)
this model

Space using OptimaMT/id-en 1

Collection including OptimaMT/id-en