opus-mt-en-pt
Overview
This repository contains the English-to-Portuguese OPUS-MT transformer model from Helsinki-NLP/opus-mt-tc-big-en-pt.
Source
| Field | Value |
|---|---|
| Upstream model | Helsinki-NLP/opus-mt-tc-big-en-pt |
| Upstream source revision | 9f2863d807ecf91a374bdbecb8d01e402e90622e |
| Export tool/script | Weight mirror from upstream Marian safetensors (no conversion) |
| Quantization recipe | FP32 model.safetensors (Transformers Marian) |
Precision and Packaging
Export tooling, precision, and quantization are recorded in the Source table above. This packaging mirror does not publish independent parity benchmarks; validate on your target execution provider before production use.
Files
| File | Description |
|---|---|
model.safetensors | Model weights |
config.json | Marian model configuration |
generation_config.json | Text generation configuration |
source.spm | Source SentencePiece model |
target.spm | Target SentencePiece model |
vocab.json | Vocabulary |
tokenizer_config.json | Tokenizer configuration |
special_tokens_map.json | Special token mapping |
Intended Use
Use this model for English-to-Portuguese machine translation with Transformers-compatible Marian tooling.
Training Data
The upstream model card states that training data is taken from OPUS.
Evaluation
The upstream model card reports self-reported BLEU scores for English-to-Portuguese translation:
| Test Set | BLEU |
|---|---|
flores101-devtest | 50.4 |
tatoeba-test-v2021-08-07 | 49.6 |
Limitations
- Evaluation scores are inherited from the upstream model card and are not independently verified here.
- Translation quality can vary by domain and input style.
- This repository does not document additional fine-tuning or conversion beyond the upstream source model files.
License
CC-BY-4.0, matching the upstream source model metadata.