def build_sp_pipeline(spm_file):
tokenizer = sentencepiece_tokenizer(spm_file)
vocab = PretrainedSPVocab(load_sp_model(spm_file))
# Insert token in vocab to match a pretrained vocab
vocab.insert_token('<pad>', 1)
pipeline = TextSequentialTransforms(tokenizer, vocab)
jit_pipeline = torch.jit.script(pipeline.to_ivalue())
print('jit sentencepiece pipeline success!')
return pipeline, pipeline.to_ivalue(), jit_pipeline
def build_sp_pipeline(spm_file):
tokenizer = PretrainedSPTokenizer(spm_file)
vocab = PretrainedSPVocab(spm_file)
# Insert token in vocab to match a pretrained vocab
vocab.insert_token('<pad>', 1)
pipeline = TextDataPipeline(tokenizer, vocab)
jit_pipeline = torch.jit.script(pipeline)
print('jit sentencepiece pipeline success!')
return pipeline, jit_pipeline
def build_sp_pipeline(spm_file):
tokenizer = PretrainedSPTokenizer(spm_file)
vocab = PretrainedSPVocab(spm_file)
# Insert token in vocab to match a pretrained vocab
vocab.insert_token('<pad>', 1)
pipeline = TextSequentialTransforms(tokenizer, vocab, ToLongTensor())
jit_pipeline = torch.jit.script(pipeline)
print('jit sentencepiece pipeline success!')
return pipeline, jit_pipeline
def build_sp_pipeline(args):
spm_file = args.spm_filename
if spm_file in PRETRAINED_SP_MODEL:
spm_file = download_from_url(PRETRAINED_SP_MODEL[spm_file])
tokenizer = sentencepiece_tokenizer(spm_file)
vocab = PretrainedSPVocab(load_sp_model(spm_file))
# Insert token in vocab to match a pretrained vocab
pipeline = TextSequentialTransforms(tokenizer, vocab)
jit_pipeline = torch.jit.script(pipeline)
print('jit sentencepiece pipeline success!')
return pipeline, pipeline, jit_pipeline