quickmt-ko-en, int8 CTranslate2

The quickmt/quickmt-ko-en translation model (CC BY 4.0, by quickmt) re-saved with int8 weights for CTranslate2, as used by the Yaku Translate app. Output on the checked sentences is character for character the same as the float16 original. Conversion: per-row int8 quantization of every 2-D weight (same scheme as CTranslate2's ModelSpec quantization), 26.10.2026. Files: model.bin, config.json, source_vocabulary.json, target_vocabulary.json, src.spm.model, tgt.spm.model.

Downloads last month
17
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for value-productions/quickmt-ko-en-int8

Finetuned
(1)
this model