| 96 | |
| 97 | |
| 98 | class Encoder_SP: |
| 99 | def __init__(self, model_path): |
| 100 | self.sp = spm.SentencePieceProcessor() |
| 101 | self.sp.Load(model_path) |
| 102 | |
| 103 | def encode(self, text): |
| 104 | """ |
| 105 | text="...." |
| 106 | """ |
| 107 | return self.sp.EncodeAsIds(text) |
| 108 | |
| 109 | def decode(self, tokens): |
| 110 | """ |
| 111 | tokens=[x1,x2,...] |
| 112 | """ |
| 113 | text = [int(token) for token in tokens] |
| 114 | # print(text) |
| 115 | return self.sp.DecodeIds(text) |
| 116 | |
| 117 | def tokenize(self, text): |
| 118 | return self.sp.EncodeAsPieces(text) |
| 119 | |
| 120 | def convert_tokens_to_ids(self, tokens): |
| 121 | return [self.sp.PieceToId(token) for token in tokens] |
| 122 | |
| 123 | def convert_token_to_id(self, token): |
| 124 | return self.sp.PieceToId(token) |
| 125 | |
| 126 | def convert_id_to_token(self, idx): |
| 127 | return self.sp.IdToPiece(idx) |
| 128 | |
| 129 | |
| 130 | def get_encoder(encoder_file, bpe_file): |