Lauler commited on
Commit
a8a6f91
·
verified ·
1 Parent(s): ff380a3

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +5 -0
README.md CHANGED
@@ -23,6 +23,8 @@ The National Library of Sweden releases a new suite of Whisper models trained on
23
  | [large-v3](https://huggingface.co/KBLab/kb-whisper-large) | **KBLab** | **5.4** | **4.1** | **5.2** |
24
  | | OpenAI | 7.8 | 9.5 | 11.3 |
25
 
 
 
26
  ### Usage
27
 
28
  ```python
@@ -74,6 +76,8 @@ The default when loading our models through Hugging Face is **Stage 2**. We have
74
 
75
  ### Evaluation
76
 
 
 
77
  | Model size | | FLEURS | CommonVoice | NST |
78
  |------------|---------|--------|-------------|------|
79
  | [tiny](https://huggingface.co/KBLab/kb-whisper-tiny) | **KBLab** | **13.2** | **12.9** | **11.2** |
@@ -88,6 +92,7 @@ The default when loading our models through Hugging Face is **Stage 2**. We have
88
  | | OpenAI | 7.8 | 9.5 | 11.3 |
89
 
90
 
 
91
  | Model size | | FLEURS | CommonVoice | NST |
92
  |------------|---------|--------|-------------|------|
93
  | tiny | KBLab | **76.6** | **73.7** | **74.3** |
 
23
  | [large-v3](https://huggingface.co/KBLab/kb-whisper-large) | **KBLab** | **5.4** | **4.1** | **5.2** |
24
  | | OpenAI | 7.8 | 9.5 | 11.3 |
25
 
26
+ Table: **Word Error Rate (WER)** comparison between KBLab's Whisper models and the corresponding OpenAI versions.
27
+
28
  ### Usage
29
 
30
  ```python
 
76
 
77
  ### Evaluation
78
 
79
+
80
+ #### WER
81
  | Model size | | FLEURS | CommonVoice | NST |
82
  |------------|---------|--------|-------------|------|
83
  | [tiny](https://huggingface.co/KBLab/kb-whisper-tiny) | **KBLab** | **13.2** | **12.9** | **11.2** |
 
92
  | | OpenAI | 7.8 | 9.5 | 11.3 |
93
 
94
 
95
+ #### BLEU Score
96
  | Model size | | FLEURS | CommonVoice | NST |
97
  |------------|---------|--------|-------------|------|
98
  | tiny | KBLab | **76.6** | **73.7** | **74.3** |