aina-translator-ca-zh-onnx

aina-translator-ca-zh-onnx

ONNX export of projecte-aina/aina-translator-ca-zh, Projecte Aina's (Barcelona Supercomputing Center, Language Technologies Unit) Catalan → Chinese machine translation model, fine-tuned from facebook/m2m100_1.2B. All credit for training data, fine-tuning and evaluation goes to Projecte Aina — see the source model card for training details, BLEU/ChrF numbers, and paper reference.

Licence

apache-2.0, verbatim as declared on the source model card (projecte-aina/aina-translator-ca-zh). Same licence applies to this derived ONNX export.

Files

encoder_model.onnx(_data)                fp32 encoder
decoder_model.onnx(_data)                fp32 decoder (no cache)
decoder_with_past_model.onnx(_data)      fp32 decoder (with KV cache)
int8/encoder_model.onnx                  dynamic-quantized (uint8) encoder
int8/decoder_model.onnx                  dynamic-quantized (uint8) decoder
int8/decoder_with_past_model.onnx        dynamic-quantized (uint8) decoder w/ cache
sentencepiece.bpe.model, vocab.json, tokenizer_config.json, ...   tokenizer files (M2M100Tokenizer)

Export

Base architecture: M2M100ForConditionalGeneration (transformers model_type: m2m_100), fine-tuned by Projecte Aina from facebook/m2m100_1.2B.

optimum-cli export onnx \
  --model projecte-aina/aina-translator-ca-zh \
  --task text2text-generation-with-past \
  --no-post-process \
  aina-translator-ca-zh-onnx

--no-post-process is required: optimum's decoder-merge step OOMs on this model size on constrained hardware. As a result the ONNX export ships an un-merged decoder_model.onnx (no cache) and decoder_with_past_model.onnx (with cache) instead of a single decoder_model_merged.onnx.

int8 dynamic quantization (optimum.onnxruntime.ORTQuantizer, AVX2 config) was applied to each of the three graphs.

Target-language mechanism

Same family as aina-translator-zh-ca-onnx but the reverse direction. Unlike the zh→ca checkpoint, this one's tokenizer_config.json does carry explicit src_lang: "ca" / tgt_lang: "zh" fields, and they are already the tokenizer's default — no extra argument is required at inference time. There is still no forced_bos_token_id in generation_config.json; a plain tokenizer(text) + model.generate() call is sufficient and yields Chinese output directly, exactly as in the upstream model card's usage example.

Parity

8 held-out Catalan sentences, num_beams=4, max_new_tokens=64, compared against the original PyTorch model (transformers.AutoModelForSeq2SeqLM) with identical decoding settings.

Sample translations observed (ca → zh):

Usage

from transformers import AutoTokenizer
from optimum.onnxruntime import ORTModelForSeq2SeqLM

model_id = "TigreGotico/aina-translator-ca-zh-onnx"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = ORTModelForSeq2SeqLM.from_pretrained(model_id)  # fp32
# int8: ORTModelForSeq2SeqLM.from_pretrained(model_id, subfolder="int8")

text = "Benvingut al projecte Aina!"
ids = tokenizer(text, return_tensors="pt").input_ids
out = model.generate(ids, num_beams=4, max_new_tokens=64)
print(tokenizer.decode(out[0], skip_special_tokens=True))
# 欢迎来到Aina项目！

Attribution

All modeling and training work is by Projecte Aina (Language Technologies Unit, Barcelona Supercomputing Center) — this repository only republishes an ONNX conversion of their weights for offline/CPU inference. Source model: projecte-aina/aina-translator-ca-zh. Contact for the original model: [email protected].