def load_token2id(folder: str) -> pc.Token2Id:
dictionary: pd.DataFrame = smart_load(jj(folder, 'dictionary.zip'),
feather_pipe=pu.set_index,
columns='token_id')
token2id: pc.Token2Id = pc.Token2Id(
data={t: i
for (t, i) in zip(dictionary.token, dictionary.index)})
return token2id
def __post_init__(self):
if isinstance(self.corpus, (pc.VectorizedCorpus, pc.TokenizedCorpus)):
if not isinstance(self.token2id, pc.Token2Id):
self.token2id = pc.Token2Id(data=self.corpus.token2id)
self.document_index = self.corpus.document_index
if isinstance(self.token2id, dict):
self.token2id = pc.Token2Id(data=self.token2id)
self.update_token_counts()
self.vectorizer_args = {
**DEFAULT_VECTORIZE_PARAMS,
**(self.vectorizer_args or {})
}
def create_train_corpus() -> tm.TrainingCorpus:
corpus: TranströmerCorpus = TranströmerCorpus()
sparse_corpus, vocabulary = convert.TranslateCorpus().translate(corpus, id2token=None)
tc: tm.TrainingCorpus = tm.TrainingCorpus(
corpus=sparse_corpus,
document_index=corpus.document_index,
token2id=pc.Token2Id(vocabulary.token2id),
)
return tc
def test_id2token2token2id():
assert pc.id2token2token2id({1: 'a', 2: 'b'}) == {'a': 1, 'b': 2}
assert pc.id2token2token2id(pc.Token2Id({
1: 'a',
2: 'b'
})) == {
'a': 1,
'b': 2
}
def load(folder: str) -> TrainingCorpus:
"""Loads an training corpus from pickled file."""
"""Load from vectorized corpus if exists"""
if pc.VectorizedCorpus.dump_exists(tag='train', folder=folder):
corpus: pc.VectorizedCorpus = pc.VectorizedCorpus.load(
tag='train', folder=folder)
return TrainingCorpus(
corpus=corpus,
document_index=corpus.document_index,
token2id=pc.Token2Id(data=corpus.token2id),
corpus_options=utility.read_json(jj(folder,
CORPUS_OPTIONS_FILENAME),
default={}),
vectorizer_args=utility.read_json(jj(folder,
VECTORIZER_ARGS_FILENAME),
default={}),
)
return None
def token2id(self) -> pc.Token2Id:
return pc.Token2Id(data={
t: i
for t, i in zip(self.vocabulary.token, self.vocabulary.token_id)
})
def instream_to_corpus(self, id2token: Mapping[int, str] | None) -> tm.TrainingCorpus:
content_type: ContentType = self.resolved_prior_out_content_type()
if self.train_corpus_folder:
if tm.TrainingCorpus.exists(self.train_corpus_folder):
logger.info(
f"using existing corpus in folder {self.train_corpus_folder} for target mode {self.target_mode}"
)
corpus: tm.TrainingCorpus = tm.TrainingCorpus.load(self.train_corpus_folder)
return corpus
tags: List[str] = pc.VectorizedCorpus.find_tags(self.train_corpus_folder)
if len(tags) == 0:
raise ValueError(f"no train or predict input corpus found in {self.train_corpus_folder}")
if len(tags) > 1:
raise ValueError(f"multiple corpus found in folder {self.train_corpus_folder}")
logger.info(
f"using corpus tagged {tags[0]} in folder {self.train_corpus_folder} for target mode {self.target_mode}"
)
vectorized_corpus: pc.VectorizedCorpus = pc.VectorizedCorpus.load(
folder=self.train_corpus_folder, tag=tags[0]
)
corpus: tm.TrainingCorpus = tm.TrainingCorpus(corpus=vectorized_corpus)
return corpus
if content_type == ContentType.VECTORIZED_CORPUS:
logger.info("creating sparse corpus out of input stream...")
payload: DocumentPayload = next(self.prior.outstream())
vectorized_corpus: pc.VectorizedCorpus = payload.content
vectorize_opts: pc.VectorizeOpts = payload.recall('vectorize_opts')
if id2token is not None:
"""We must consolidate the vocabularies"""
logger.info("translating vocabulary to training model's vocabulary...")
vectorized_corpus.translate_to_vocab(id2token, inplace=True)
corpus: tm.TrainingCorpus = tm.TrainingCorpus(
corpus=vectorized_corpus,
corpus_options={},
vectorizer_args={} if vectorize_opts is None else vectorize_opts.props,
)
logger.info("training corpus created!")
return corpus
if content_type == ContentType.TOKENS:
token2id: pc.Token2Id = (
pc.Token2Id(pc.id2token2token2id(id2token)) if id2token is not None else self.pipeline.payload.token2id
)
corpus: tm.TrainingCorpus = tm.TrainingCorpus(
corpus=self.prior.filename_content_stream(),
document_index=self.document_index,
token2id=token2id,
corpus_options={},
)
return corpus
raise ValueError("unable to resolve input corpus")
def token2id(self) -> pc.Token2Id:
return pc.Token2Id(data=self.term2id)
