Adam Montgomerie
commited on
Commit
β’
bbf4996
1
Parent(s):
d6f17cc
Update README.md
Browse files
README.md
CHANGED
@@ -1,5 +1,5 @@
|
|
1 |
---
|
2 |
-
language:
|
3 |
datasets:
|
4 |
- common_voice
|
5 |
tags:
|
@@ -9,13 +9,13 @@ tags:
|
|
9 |
- xlsr-fine-tuning-week
|
10 |
license: apache-2.0
|
11 |
model-index:
|
12 |
-
- name: XLSR Wav2Vec2
|
13 |
results:
|
14 |
- task:
|
15 |
name: Speech Recognition
|
16 |
type: automatic-speech-recognition
|
17 |
dataset:
|
18 |
-
name: Common Voice
|
19 |
type: common_voice
|
20 |
args: {lang_id}
|
21 |
metrics:
|
@@ -24,9 +24,9 @@ model-index:
|
|
24 |
value: 43.48
|
25 |
---
|
26 |
|
27 |
-
# Wav2Vec2-Large-XLSR-53-
|
28 |
|
29 |
-
Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) in
|
30 |
When using this model, make sure that your speech input is sampled at 16kHz.
|
31 |
|
32 |
## Usage
|
@@ -67,7 +67,7 @@ print("Reference:", test_dataset["sentence"][:2])
|
|
67 |
|
68 |
## Evaluation
|
69 |
|
70 |
-
The model can be evaluated as follows on the
|
71 |
|
72 |
|
73 |
```python
|
@@ -84,7 +84,7 @@ processor = Wav2Vec2Processor.from_pretrained("iarfmoose/wav2vec2-large-xlsr-sor
|
|
84 |
model = Wav2Vec2ForCTC.from_pretrained("iarfmoose/wav2vec2-large-xlsr-sorbian")
|
85 |
model.to("cuda")
|
86 |
|
87 |
-
chars_to_ignore_regex = '[
|
88 |
resampler = torchaudio.transforms.Resample(48_000, 16_000)
|
89 |
|
90 |
def speech_file_to_array_fn(batch):
|
|
|
1 |
---
|
2 |
+
language: hsb
|
3 |
datasets:
|
4 |
- common_voice
|
5 |
tags:
|
|
|
9 |
- xlsr-fine-tuning-week
|
10 |
license: apache-2.0
|
11 |
model-index:
|
12 |
+
- name: XLSR Wav2Vec2 Sorbian by Adam Montgomerie
|
13 |
results:
|
14 |
- task:
|
15 |
name: Speech Recognition
|
16 |
type: automatic-speech-recognition
|
17 |
dataset:
|
18 |
+
name: Common Voice hsb
|
19 |
type: common_voice
|
20 |
args: {lang_id}
|
21 |
metrics:
|
|
|
24 |
value: 43.48
|
25 |
---
|
26 |
|
27 |
+
# Wav2Vec2-Large-XLSR-53-Sorbian
|
28 |
|
29 |
+
Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) in Sorbian using the [Common Voice](https://huggingface.co/datasets/common_voice)
|
30 |
When using this model, make sure that your speech input is sampled at 16kHz.
|
31 |
|
32 |
## Usage
|
|
|
67 |
|
68 |
## Evaluation
|
69 |
|
70 |
+
The model can be evaluated as follows on the Sorbian test data of Common Voice.
|
71 |
|
72 |
|
73 |
```python
|
|
|
84 |
model = Wav2Vec2ForCTC.from_pretrained("iarfmoose/wav2vec2-large-xlsr-sorbian")
|
85 |
model.to("cuda")
|
86 |
|
87 |
+
chars_to_ignore_regex = '[\\,\\?\\.\\!\\-\\;\\:\\"\\β\\%\\β\\β\\οΏ½\\β\\β\\Β¬\\β
]'
|
88 |
resampler = torchaudio.transforms.Resample(48_000, 16_000)
|
89 |
|
90 |
def speech_file_to_array_fn(batch):
|