rxn.onmt_utils.translator.Translator
- class rxn.onmt_utils.translator.Translator(opt)[source]
Bases:
objectWraps the OpenNMT translation functionality into a class.
- Parameters
opt (
Namespace) –
- __init__(opt)[source]
Should not be called directly as implementation may change; call the classmethods from_model_path or from_opt instead.
- Parameters
opt (
Namespace) – model options.
Methods
__init__(opt)Should not be called directly as implementation may change; call the classmethods from_model_path or from_opt instead.
from_model_path(model_path, **kwargs)Create a Translator instance from the model path(s).
from_opt(opt)Create a Translator instance from the opt arguments.
translate_multiple_with_scores(sentences[, ...])Translate multiple sentences.
translate_sentences(sentences)Translate multiple sentences.
translate_single(sentence)Translate one single sentence.
- classmethod from_model_path(model_path, **kwargs)[source]
Create a Translator instance from the model path(s).
- Parameters
model_path (
Union[str,Iterable[str]]) – path to the translation model file(s). If multiple are given, will be an ensemble model.kwargs (
Any) – Additional values to be parsed for instantiating the translator, such as n_best, beam_size, max_length, etc.
- Return type
- classmethod from_opt(opt)[source]
Create a Translator instance from the opt arguments.
- Parameters
opt (
Namespace) – model options.- Return type
- translate_multiple_with_scores(sentences, n_best=None)[source]
Translate multiple sentences.
- Parameters
sentences (
Iterable[str]) – Sentences to translate.n_best (
Optional[int], default:None) – if provided, will overwrite the number of predictions to make.
- Return type
Iterator[List[TranslationResult]]