Update README.md
Browse files
README.md
CHANGED
|
@@ -27,6 +27,7 @@ Wav2Vec2-Large-XLSR-Welsh
|
|
| 27 |
Fine-tuned facebook/wav2vec2-large-xlsr-53 on the Welsh Common Voice dataset.
|
| 28 |
|
| 29 |
When using this model, make sure that your speech input is sampled at 16kHz.
|
|
|
|
| 30 |
|
| 31 |
Usage
|
| 32 |
The model can be used directly (without a language model) as follows:
|
|
@@ -80,7 +81,7 @@ model = Wav2Vec2ForCTC.from_pretrained("Srulikbdd/Wav2vec2-large-xlsr-welsh")
|
|
| 80 |
|
| 81 |
model.to("cuda")
|
| 82 |
|
| 83 |
-
chars_to_ignore_regex = '[
|
| 84 |
|
| 85 |
resampler = torchaudio.transforms.Resample(48_000, 16_000)
|
| 86 |
|
|
|
|
| 27 |
Fine-tuned facebook/wav2vec2-large-xlsr-53 on the Welsh Common Voice dataset.
|
| 28 |
|
| 29 |
When using this model, make sure that your speech input is sampled at 16kHz.
|
| 30 |
+
Test Result: 29.4%
|
| 31 |
|
| 32 |
Usage
|
| 33 |
The model can be used directly (without a language model) as follows:
|
|
|
|
| 81 |
|
| 82 |
model.to("cuda")
|
| 83 |
|
| 84 |
+
chars_to_ignore_regex = '[\\\\,\\\\?\\\\.\\\\!\\\\-\\\\u2013\\\\u2014\\\\;\\\\:\\\\"\\\\\\\\%\\\\\\\\\\\\]'
|
| 85 |
|
| 86 |
resampler = torchaudio.transforms.Resample(48_000, 16_000)
|
| 87 |
|