heimoshuiyu/faster-whisper
1
0
Fork 0
Code Issues Pull Requests 1 Packages Projects Releases Wiki Activity

Raise a more explicit error message for English-only models

Browse Source
This commit is contained in:
Guillaume Klein
2023-02-13 18:26:45 +01:00
parent 47a62ab975
commit 3dc44f7bb5
1 changed files with 14 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
faster_whisper/transcribe.py
View File

@@ -1,4 +1,5 @@
import collections import collections
import os
import zlib import zlib
import ctranslate2 import ctranslate2
@@ -66,6 +67,19 @@ class WhisperModel:
intra_threads=cpu_threads, intra_threads=cpu_threads,
) )
with open(os.path.join(model_path, "vocabulary.txt")) as vocab_file:
vocab_size = sum(1 for _ in vocab_file)
is_multilingual = vocab_size == 51865
if not is_multilingual:
raise NotImplementedError(
"English-only models are currently not supported. "
"The underlying CTranslate2 implementation makes some assumptions about "
"the prompt format that are not compatible with English-only models. "
"This will be improved in a future version. "
"Please use a multilingual model for now."
)
self.feature_extractor = FeatureExtractor() self.feature_extractor = FeatureExtractor()
self.tokenizer = tokenizers.Tokenizer.from_pretrained("openai/whisper-tiny") self.tokenizer = tokenizers.Tokenizer.from_pretrained("openai/whisper-tiny")
self.eot_id = self.tokenizer.token_to_id("<|endoftext|>") self.eot_id = self.tokenizer.token_to_id("<|endoftext|>")