Commit
·
f746dcb
1
Parent(s):
f5bf8d0
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,54 @@
|
|
| 1 |
---
|
| 2 |
license: cc-by-4.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: cc-by-4.0
|
| 3 |
+
language:
|
| 4 |
+
- is
|
| 5 |
+
datasets:
|
| 6 |
+
- language-and-voice-lab/samromur_asr
|
| 7 |
+
- language-and-voice-lab/samromur_children
|
| 8 |
+
- language-and-voice-lab/malromur_asr
|
| 9 |
+
- language-and-voice-lab/althingi_asr
|
| 10 |
+
tags:
|
| 11 |
+
- audio
|
| 12 |
+
- automatic-speech-recognition
|
| 13 |
+
- icelandic
|
| 14 |
+
- whisper
|
| 15 |
+
- whisper-large
|
| 16 |
+
- iceland
|
| 17 |
+
- reykjavik
|
| 18 |
+
- samromur
|
| 19 |
+
- faster-whisper
|
| 20 |
---
|
| 21 |
+
# whisper-large-icelandic-30k-steps-1000h-ct2
|
| 22 |
+
|
| 23 |
+
This is a faster-whisper version of [language-and-voice-lab/whisper-large-icelandic-30k-steps-1000h](https://huggingface.co/language-and-voice-lab/whisper-large-icelandic-30k-steps-1000h).
|
| 24 |
+
|
| 25 |
+
The model was created as such like described in [faster-whisper](https://github.com/guillaumekln/faster-whisper/tree/master):
|
| 26 |
+
|
| 27 |
+
```bash
|
| 28 |
+
ct2-transformers-converter --model language-and-voice-lab/whisper-large-icelandic-30k-steps-1000h \
|
| 29 |
+
--output_dir whisper-large-icelandic-30k-steps-1000h-ct2 \
|
| 30 |
+
--quantization float16
|
| 31 |
+
```
|
| 32 |
+
|
| 33 |
+
# Usage
|
| 34 |
+
|
| 35 |
+
```python
|
| 36 |
+
from faster_whisper import WhisperModel
|
| 37 |
+
|
| 38 |
+
model_size = "whisper-large-icelandic-30k-steps-1000h-ct2"
|
| 39 |
+
|
| 40 |
+
# Run on GPU with FP16
|
| 41 |
+
model = WhisperModel(model_size, device="cuda", compute_type="float16")
|
| 42 |
+
|
| 43 |
+
# or run on GPU with INT8
|
| 44 |
+
# model = WhisperModel(model_size, device="cuda", compute_type="int8_float16")
|
| 45 |
+
# or run on CPU with INT8
|
| 46 |
+
# model = WhisperModel(model_size, device="cpu", compute_type="int8")
|
| 47 |
+
|
| 48 |
+
segments, info = model.transcribe("audio.mp3", beam_size=5)
|
| 49 |
+
|
| 50 |
+
print("Detected language '%s' with probability %f" % (info.language, info.language_probability))
|
| 51 |
+
|
| 52 |
+
for segment in segments:
|
| 53 |
+
print("[%.2fs -> %.2fs] %s" % (segment.start, segment.end, segment.text))
|
| 54 |
+
```
|