Mirroristas/whisper
2
0
Fork 0
mirror of https://github.com/openai/whisper.git synced 2025-11-24 14:35:57 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fix truncated words list when the replacement character is decoded (#1089)

Browse Source
This commit is contained in:
Guillaume Klein 2023-03-14 17:32:41 +01:00 committed by GitHub
parent ba88b8e1b3
commit 5f9ac653b7
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 21 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
tests/test_tokenizer.py
View File

@ -12,3 +12,13 @@ def test_tokenizer():
assert gpt2_tokenizer.decode(gpt2_tokens) == text assert gpt2_tokenizer.decode(gpt2_tokens) == text
assert multilingual_tokenizer.decode(multilingual_tokens) == text assert multilingual_tokenizer.decode(multilingual_tokens) == text
assert len(gpt2_tokens) > len(multilingual_tokens) assert len(gpt2_tokens) > len(multilingual_tokens)
def test_split_on_unicode():
multilingual_tokenizer = get_tokenizer(multilingual=True)
tokens = [8404, 871, 287, 6, 246, 526, 3210, 20378]
words, word_tokens = multilingual_tokenizer.split_tokens_on_unicode(tokens)
assert words == [" elle", " est", " l", "'", "<EFBFBD>", "é", "rit", "oire"]
assert word_tokens == [[8404], [871], [287], [6], [246], [526], [3210], [20378]]

12
whisper/tokenizer.py
View File

@ -279,17 +279,27 @@ class Tokenizer:
return self.split_tokens_on_spaces(tokens) return self.split_tokens_on_spaces(tokens)
def split_tokens_on_unicode(self, tokens: List[int]): def split_tokens_on_unicode(self, tokens: List[int]):
decoded_full = self.decode_with_timestamps(tokens)
replacement_char = "\ufffd"
words = [] words = []
word_tokens = [] word_tokens = []
current_tokens = [] current_tokens = []
unicode_offset = 0
for token in tokens: for token in tokens:
current_tokens.append(token) current_tokens.append(token)
decoded = self.decode_with_timestamps(current_tokens) decoded = self.decode_with_timestamps(current_tokens)
if "\ufffd" not in decoded:
if (
replacement_char not in decoded
or decoded_full[unicode_offset + decoded.index(replacement_char)]
== replacement_char
):
words.append(decoded) words.append(decoded)
word_tokens.append(current_tokens) word_tokens.append(current_tokens)
current_tokens = [] current_tokens = []
unicode_offset += len(decoded)
return words, word_tokens return words, word_tokens