Update README.md
Browse files
README.md
CHANGED
|
@@ -117,8 +117,10 @@ From my early tests:
|
|
| 117 |
- Much less GPU memory required
|
| 118 |
- It seems that performance is on par with the original
|
| 119 |
- It seems that this combination is faster than just using the CTranslate2 int8 quantization.
|
|
|
|
| 120 |
Quantization method TBA.
|
| 121 |
To use this model use the faster_whisper module as stated in [the original faster-whisper model](https://huggingface.co/Systran/faster-whisper-large-v3)
|
|
|
|
| 122 |
|
| 123 |
Any benchmark results are appreciated. I probably do not have time to do it myself.
|
| 124 |
|
|
|
|
| 117 |
- Much less GPU memory required
|
| 118 |
- It seems that performance is on par with the original
|
| 119 |
- It seems that this combination is faster than just using the CTranslate2 int8 quantization.
|
| 120 |
+
|
| 121 |
Quantization method TBA.
|
| 122 |
To use this model use the faster_whisper module as stated in [the original faster-whisper model](https://huggingface.co/Systran/faster-whisper-large-v3)
|
| 123 |
+
Or use [WhisperX](https://github.com/m-bain/whisperX), this is what I used for my small tests (do not forget to set dtype to int8).
|
| 124 |
|
| 125 |
Any benchmark results are appreciated. I probably do not have time to do it myself.
|
| 126 |
|