Function differences with torchtext.data.functional.load_sp_model

torchtext.data.functional.load_sp_model

torchtext.data.functional.load_sp_model(
    spm
)

For more information, see torchtext.data.functional.load_sp_model.

mindspore.dataset.text.SentencePieceVocab

class mindspore.dataset.text.SentencePieceVocab

For more information, see mindspore.dataset.text.SentencePieceVocab.

Differences

PyTorch: Load a sentencepiece model for file. Input the path of the sentence fragment model and output the sentence fragment model.

MindSpore: SentencePiece object that is used to perform words segmentation. The input can be a dataset object or a glossary file.

Code Example

import mindspore.dataset as ds
from mindspore.dataset import text
from mindspore.dataset.text import SentencePieceModel, SPieceTokenizerOutType
from torchtext.data.functional import load_sp_model

# In MindSpore, return tokenizer from vocab object.
sentence_piece_vocab_file = "/path/to/test_sentencepiece/botchan.txt"

vocab = text.SentencePieceVocab.from_file([sentence_piece_vocab_file], 500, 0.9995,
                                          SentencePieceModel.WORD, {})
tokenizer = text.SentencePieceTokenizer(vocab, out_type=SPieceTokenizerOutType.STRING)
text_file_dataset_dir = "/path/to/testTokenizerData/sentencepiece_tokenizer.txt"
text_file_dataset = ds.TextFileDataset(dataset_files=text_file_dataset_dir)
text_file_dataset = text_file_dataset.map(operations=tokenizer)

for i in text_file_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
    ret = i["text"]
    for value in ret:
        print(value)
# Out:
# ▁I
# ▁saw
# ▁a
# ▁girl
# ▁with
# ▁a
# ▁telescope.

# In torch, return the sentencepiece model according to the input model path.
sp_model = load_sp_model("m_user.model")
sp_model = load_sp_model(open("m_user.model", 'rb'))