Skip to content

Open-source speech recognition toolkit for training, inference, streaming ASR, VAD, punctuation, speaker diarization pipelines, and OpenAI-compatible/MCP serving.

Funasr1.0 (#1279)

zhifu gao3919d7454c07
diff --git a/.gitignore b/.gitignore
index 4023869b5..6bdfd5d06 100644
--- a/.gitignore
+++ b/.gitignore
@@ -23,3 +23,4 @@ samples
.ipynb_checkpoints
outputs*
emotion2vec*
+GPT-SoVITS*
diff --git a/funasr/models/paraformer/model.py b/funasr/models/paraformer/model.py
index 0c4f14aae..6e422ad75 100644
--- a/funasr/models/paraformer/model.py
+++ b/funasr/models/paraformer/model.py
@@ -528,9 +528,9 @@ class Paraformer(torch.nn.Module):
if tokenizer is not None:
# Change integer-ids to tokens
token = tokenizer.ids2tokens(token_int)
- text = tokenizer.tokens2text(token)
-
- text_postprocessed, _ = postprocess_utils.sentence_postprocess(token)
+ text_postprocessed = tokenizer.tokens2text(token)
+ if not hasattr(tokenizer, "bpemodel"):
+ text_postprocessed, _ = postprocess_utils.sentence_postprocess(token)
result_i = {"key": key[i], "text": text_postprocessed}
diff --git a/funasr/tokenizer/sentencepiece_tokenizer.py b/funasr/tokenizer/sentencepiece_tokenizer.py
index 9a65920d2..05ca3489a 100644
--- a/funasr/tokenizer/sentencepiece_tokenizer.py
+++ b/funasr/tokenizer/sentencepiece_tokenizer.py
@@ -5,12 +5,16 @@ from typing import Union
import sentencepiece as spm
-from funasr.tokenizer.abs_tokenizer import AbsTokenizer
-
-
-class SentencepiecesTokenizer(AbsTokenizer):
- def __init__(self, model: Union[Path, str]):
- self.model = str(model)
+from funasr.tokenizer.abs_tokenizer import BaseTokenizer
+from funasr.register import tables
+
[email protected]("tokenizer_classes", "SentencepiecesTokenizer")
+class SentencepiecesTokenizer(BaseTokenizer):
+ def __init__(self, bpemodel: Union[Path, str],
+ **kwargs
+ ):
+ super().__init__(**kwargs)
+ self.bpemodel = str(bpemodel)
# NOTE(kamo):
# Don't build SentencePieceProcessor in __init__()
# because it's not picklable and it may cause following error,
@@ -19,13 +23,13 @@ class SentencepiecesTokenizer(AbsTokenizer):
self.sp = None
def __repr__(self):
- return f'{self.__class__.__name__}(model="{self.model}")'
+ return f'{self.__class__.__name__}(model="{self.bpemodel}")'
def _build_sentence_piece_processor(self):
# Build SentencePieceProcessor lazily.
if self.sp is None:
self.sp = spm.SentencePieceProcessor()
- self.sp.load(self.model)
+ self.sp.load(self.bpemodel)
def text2tokens(self, line: str) -> List[str]:
self._build_sentence_piece_processor()
@@ -34,3 +38,11 @@ class SentencepiecesTokenizer(AbsTokenizer):
def tokens2text(self, tokens: Iterable[str]) -> str:
self._build_sentence_piece_processor()
return self.sp.DecodePieces(list(tokens))
+
+ def encode(self, line: str) -> List[int]:
+ self._build_sentence_piece_processor()
+ return self.sp.EncodeAsIds(line)
+
+ def decode(self, line: List[int]):
+ self._build_sentence_piece_processor()
+ return self.sp.DecodeIds(line)
\ No newline at end of file