hachimimt-30-web

hachimimt-30-web (ONNX, INT8, web build)

Browser-ready build of ngocdang83/HachimiMT-30-zh-vi, a MarianMT model for Chinese -> Vietnamese web-novel translation. 37M, nhẹ hơn HachimiMT-60.

All credit for training the base model goes to the original author (ngocdang83). This repo only adds:

• ONNX export with KV-cache (encoder_model_quantized.onnx + decoder_model_merged_quantized.onnx)
• Dynamic INT8 quantization (~4x smaller than fp32)
• A hand-built tokenizer.json fast-tokenizer (the original model only ships slow sentencepiece files, which transformers.js cannot load directly) — verified to produce byte-identical token ids to the original slow MarianTokenizer.

Usage (transformers.js)

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

const translator = await pipeline('translation', 'yennguyen45/hachimimt-30-web', { dtype: 'q8' });
const out = await translator('你确定要这么做吗？');
console.log(out[0].translation_text);

License

CC-BY-4.0, inherited from the base model. Please credit the original author (ngocdang83) when using this model.