albertvillanova HF staff commited on
Commit
ddff094
1 Parent(s): dd3f886

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +2 -2
README.md CHANGED
@@ -4,8 +4,6 @@ tags:
4
  - zenodo
5
  ---
6
 
7
- <div class="course-tip course-tip-orange bg-gradient-to-br dark:bg-gradient-to-r before:border-orange-500 dark:before:border-orange-800 from-orange-50 dark:from-gray-900 to-white dark:to-gray-950 border border-orange-50 text-orange-700 dark:text-gray-400"><p><b>Deprecated:</b> Not every model supports a fast tokenizer. Take a look at this <a href="index#supported-frameworks">table</a> to check if a model has fast tokenizer support.</p></div>
8
-
9
  # Dataset Card for MultiLingual LibriSpeech
10
 
11
  ## Table of Contents
@@ -41,6 +39,8 @@ tags:
41
 
42
  ### Dataset Summary
43
 
 
 
44
  Multilingual LibriSpeech (MLS) dataset is a large multilingual corpus suitable for speech research. The dataset is derived from read audiobooks from LibriVox and consists of 8 languages - English, German, Dutch, Spanish, French, Italian, Portuguese, Polish.
45
 
46
  ### Supported Tasks and Leaderboards
4
  - zenodo
5
  ---
6
 
 
 
7
  # Dataset Card for MultiLingual LibriSpeech
8
 
9
  ## Table of Contents
39
 
40
  ### Dataset Summary
41
 
42
+ <div class="course-tip course-tip-orange bg-gradient-to-br dark:bg-gradient-to-r before:border-orange-500 dark:before:border-orange-800 from-orange-50 dark:from-gray-900 to-white dark:to-gray-950 border border-orange-50 text-orange-700 dark:text-gray-400"><p><b>Deprecated:</b> Not every model supports a fast tokenizer. Take a look at this <a href="index#supported-frameworks">table</a> to check if a model has fast tokenizer support.</p></div>
43
+
44
  Multilingual LibriSpeech (MLS) dataset is a large multilingual corpus suitable for speech research. The dataset is derived from read audiobooks from LibriVox and consists of 8 languages - English, German, Dutch, Spanish, French, Italian, Portuguese, Polish.
45
 
46
  ### Supported Tasks and Leaderboards