SentencePiece Python Wrapper

Python wrapper for SentencePiece. This API will offer the encoding, decoding and training of Sentencepiece.

Build and Install SentencePiece

For Linux (x64/i686), macOS, and Windows(win32/x64) environment, you can simply use pip command to install SentencePiece python module.

% pip install sentencepiece

To build and install the Python wrapper from source, please install SentencePiece C++ and try the following commands:

% python setup.py build
% sudo python setup.py install

If you don’t have write permission to the global site-packages directory or don’t want to install into it, please try:

% python setup.py install --user

Usage

See this google colab page to run sentencepiece interactively. (Note: this sample is written in old interface.)

Segmentation

% python
>>> import sentencepiece as spm
>>> sp = spm.SentencePieceProcessor(model_file='test/test_model.model')
>>> sp.encode('This is a test')
[284, 47, 11, 4, 15, 400]
>>> sp.encode(['This is a test', 'Hello world'], out_type=int)
[[284, 47, 11, 4, 15, 400], [151, 88, 21, 887]]
>>> sp.encode('This is a test', out_type=str)
['▁This', '▁is', '▁a', '▁', 't', 'est']
>>> sp.encode(['This is a test', 'Hello world'], out_type=str)
[['▁This', '▁is', '▁a', '▁', 't', 'est'], ['▁He', 'll', 'o', '▁world']]
>>> for _ in range(10):
...     sp.encode('This is a test', out_type=str, enable_sampling=True, alpha=0.1, nbest_size=-1)
... 
['▁', 'This', '▁', 'is', '▁a', '▁', 't', 'e', 'st']
['▁T', 'h', 'i', 's', '▁is', '▁a', '▁', 'te', 's', 't']
['▁T', 'h', 'is', '▁', 'is', '▁', 'a', '▁', 't', 'est']
['▁', 'This', '▁is', '▁', 'a', '▁', 't', 'e', 'st']
['▁', 'This', '▁', 'is', '▁', 'a', '▁', 't', 'e', 's', 't']
['▁This', '▁is', '▁a', '▁', 'te', 's', 't']
['▁This', '▁is', '▁', 'a', '▁', 't', 'e', 'st']
['▁', 'T', 'h', 'is', '▁', 'is', '▁', 'a', '▁', 'te', 'st']
['▁', 'This', '▁', 'i', 's', '▁a', '▁', 't', 'e', 'st']
['▁This', '▁', 'is', '▁a', '▁', 't', 'est']
>>> sp.decode([284, 47, 11, 4, 15, 400])
'This is a test'
>>> sp.decode([[284, 47, 11, 4, 15, 400], [151, 88, 21, 887]])
['This is a test', 'Hello world']
>>> sp.decode(['▁', 'This', '▁', 'is', '▁a', '▁', 't', 'e', 'st'])
'This is a test'
>>> sp.decode([['▁This', '▁is', '▁a', '▁', 't', 'est'], ['▁He', 'll', 'o', '▁world']])
['This is a test', 'Hello world']
>>> sp.get_piece_size()
1000
>>> sp.id_to_piece(2)
'</s>'
>>> sp.id_to_piece([2, 3, 4])
['</s>', '\r', '▁']
>>> sp.piece_to_id('<s>')
1
>>> sp.piece_to_id(['</s>', '\r', '▁'])
[2, 3, 4]
>>> len(sp)
1000
>>> sp['</s>']
2

Model Training

Training is performed by passing parameters of spm_train to SentencePieceTrainer.train() function.

>>> import sentencepiece as spm
>>> spm.SentencePieceTrainer.train(input='test/botchan.txt', model_prefix='m', vocab_size=1000, user_defined_symbols=['foo', 'bar'])
sentencepiece_trainer.cc(73) LOG(INFO) Starts training with : 
trainer_spec {
  input: test/botchan.txt
  .. snip
unigram_model_trainer.cc(500) LOG(INFO) EM sub_iter=1 size=1188 obj=10.2839 num_tokens=32182 num_tokens/piece=27.0892
unigram_model_trainer.cc(500) LOG(INFO) EM sub_iter=0 size=1100 obj=10.4269 num_tokens=33001 num_tokens/piece=30.0009
unigram_model_trainer.cc(500) LOG(INFO) EM sub_iter=1 size=1100 obj=10.4069 num_tokens=33002 num_tokens/piece=30.0018
trainer_interface.cc(595) LOG(INFO) Saving model: m.model
trainer_interface.cc(619) LOG(INFO) Saving vocabs: m.vocab
>>>

Training without local filesystem

Sentencepiece trainer can receive any iterable object to feed training sentences. You can also pass a file object (instance with write() method) to emit the output model to any devices. These features are useful to run sentencepiece on environment that have limited access to the local file system (e.g., Google colab.)

import urllib.request
import io
import sentencepiece as spm

# Loads model from URL as iterator and stores the model to BytesIO.
model = io.BytesIO()
with urllib.request.urlopen(
    'https://raw.githubusercontent.com/google/sentencepiece/master/data/botchan.txt'
) as response:
  spm.SentencePieceTrainer.train(
      sentence_iterator=response, model_writer=model, vocab_size=1000)

# Serialize the model as file.
# with open('out.model', 'wb') as f:
#   f.write(model.getvalue())

# Directly load the model from serialized model.
sp = spm.SentencePieceProcessor(model_proto=model.getvalue())
print(sp.encode('this is test'))

Segmentation (old interface)

% python
>>> import sentencepiece as spm
>>> sp = spm.SentencePieceProcessor()
>>> sp.Load("test/test_model.model")
True
>>> sp.EncodeAsPieces("This is a test")
['\xe2\x96\x81This', '\xe2\x96\x81is', '\xe2\x96\x81a', '\xe2\x96\x81', 't', 'est']
>>> sp.EncodeAsIds("This is a test")
[284, 47, 11, 4, 15, 400]
>>> sp.DecodePieces(['\xe2\x96\x81This', '\xe2\x96\x81is', '\xe2\x96\x81a', '\xe2\x96\x81', 't', 'est'])
'This is a test'
>>> sp.NBestEncodeAsPieces("This is a test", 5)
[['\xe2\x96\x81This', '\xe2\x96\x81is', '\xe2\x96\x81a', '\xe2\x96\x81', 't', 'est'], ['\xe2\x96\x81This', '\xe2\x96\x81is', '\xe2\x96\x81a', '\xe2\x96\x81', 'te', 'st'], ['\xe2\x96\x81This', '\xe2\x96\x81is', '\xe2\x96\x81a', '\xe2\x96\x81', 'te', 's', 't'], ['\xe2\x96\x81This', '\xe2\x96\x81is', '\xe2\x96\x81a', '\xe2\x96\x81', 't', 'e', 'st'], ['\xe2\x96\x81This', '\xe2\x96\x81is', '\xe2\x96\x81a', '\xe2\x96\x81', 't', 'es', 't']]
>>> for x in range(10):
...     sp.SampleEncodeAsPieces("This is a test", -1, 0.1)
...
['\xe2\x96\x81', 'T', 'h', 'i', 's', '\xe2\x96\x81', 'is', '\xe2\x96\x81a', '\xe2\x96\x81', 't', 'e', 's', 't']
['\xe2\x96\x81T', 'h', 'is', '\xe2\x96\x81is', '\xe2\x96\x81', 'a', '\xe2\x96\x81', 't', 'est']
['\xe2\x96\x81This', '\xe2\x96\x81is', '\xe2\x96\x81', 'a', '\xe2\x96\x81', 't', 'e', 'st']
['\xe2\x96\x81This', '\xe2\x96\x81is', '\xe2\x96\x81a', '\xe2\x96\x81', 't', 'e', 'st']
['\xe2\x96\x81This', '\xe2\x96\x81is', '\xe2\x96\x81a', '\xe2\x96\x81', 't', 'e', 's', 't']
['\xe2\x96\x81T', 'h', 'is', '\xe2\x96\x81', 'i', 's', '\xe2\x96\x81a', '\xe2\x96\x81', 'te', 's', 't']
['\xe2\x96\x81This', '\xe2\x96\x81', 'is', '\xe2\x96\x81a', '\xe2\x96\x81', 'te', 's', 't']
['\xe2\x96\x81This', '\xe2\x96\x81', 'i', 's', '\xe2\x96\x81a', '\xe2\x96\x81', 't', 'e', 'st']
['\xe2\x96\x81This', '\xe2\x96\x81', 'is', '\xe2\x96\x81', 'a', '\xe2\x96\x81', 't', 'e', 'st']
['\xe2\x96\x81This', '\xe2\x96\x81', 'i', 's', '\xe2\x96\x81', 'a', '\xe2\x96\x81', 'te', 's', 't']
>>> sp.DecodeIds([284, 47, 11, 4, 15, 400])
'This is a test'
>>> sp.GetPieceSize()
1000
>>> sp.IdToPiece(2)
'</s>'
>>> sp.PieceToId('</s>')
2
>>> len(sp)
1000
>>> sp['</s>']
2

Model Training (old interface)

Training is performed by passing parameters of spm_train to SentencePieceTrainer.Train() function.

>>> import sentencepiece as spm
>>> spm.SentencePieceTrainer.Train('--input=test/botchan.txt --model_prefix=m --vocab_size=1000')
unigram_model_trainer.cc(494) LOG(INFO) Starts training with : 
input: "test/botchan.txt"
model_prefix: "m"
model_type: UNIGRAM
..snip..
unigram_model_trainer.cc(529) LOG(INFO) EM sub_iter=0 size=1239 obj=10.4055 num_tokens=36256 num_tokens/piece=29.2623
unigram_model_trainer.cc(529) LOG(INFO) EM sub_iter=1 size=1239 obj=10.3187 num_tokens=36256 num_tokens/piece=29.2623
unigram_model_trainer.cc(529) LOG(INFO) EM sub_iter=0 size=1100 obj=10.5285 num_tokens=37633 num_tokens/piece=34.2118
unigram_model_trainer.cc(529) LOG(INFO) EM sub_iter=1 size=1100 obj=10.4973 num_tokens=37630 num_tokens/piece=34.2091
trainer_interface.cc(284) LOG(INFO) Saving model: m.model
trainer_interface.cc(293) LOG(INFO) Saving vocabs: m.vocab
>>>

Python2/3 String/Unicode compatibility

Sentencepiece python wrapper accepts both Unicode string and legacy byte string. The output string type is determined by the input string type. The output type of IdToPiece/DecodeIds methods is str, but note that it is a legacy byte string in Python2 and Unicode string in Python3 respectively.

• Python2:

>>> sp.EncodeAsPieces('吾輩は猫である')
['\xe2\x96\x81', '\xe5\x90\xbe', '\xe8\xbc\xa9', '\xe3\x81\xaf', '\xe7\x8c\xab', '\xe3\x81\xa7\xe3\x81\x82\xe3\x82\x8b']
>>> sp.EncodeAsPieces(u'吾輩は猫である')
[u'\u2581', u'\u543e', u'\u8f29', u'\u306f', u'\u732b', u'\u3067\u3042\u308b']
>>> sp.EncodeAsPieces(u'吾輩は猫である'.encode('utf-8'))
['\xe2\x96\x81', '\xe5\x90\xbe', '\xe8\xbc\xa9', '\xe3\x81\xaf', '\xe7\x8c\xab', '\xe3\x81\xa7\xe3\x81\x82\xe3\x82\x8b']
>>> sp.IdToPiece(10)
'\xe3\x81\xab'
>>> type(sp.IdToPiece(10))
<type 'str'>

• Python3:

>>> sp.EncodeAsPieces('吾輩は猫である')
['▁', '吾', '輩', 'は', '猫', 'である']
>>> sp.EncodeAsPieces('吾輩は猫である'.encode('utf-8'))
[b'\xe2\x96\x81', b'\xe5\x90\xbe', b'\xe8\xbc\xa9', b'\xe3\x81\xaf', b'\xe7\x8c\xab', b'\xe3\x81\xa7\xe3\x81\x82\xe3\x82\x8b']
>>>
>>> sp.IdToPiece(10)
'に'
>>> type(sp.IdToPiece(10))
<class 'str'>