Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,49 @@
|
|
1 |
-
---
|
2 |
-
license: cc-by-4.0
|
3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: cc-by-4.0
|
3 |
+
task_categories:
|
4 |
+
- automatic-speech-recognition
|
5 |
+
language:
|
6 |
+
- en
|
7 |
+
size_categories:
|
8 |
+
- 100K<n<1M
|
9 |
+
---
|
10 |
+
|
11 |
+
|
12 |
+
|
13 |
+
|
14 |
+
|
15 |
+
# How to Use
|
16 |
+
|
17 |
+
- Almost the same with [Librispeech](https://huggingface.co/datasets/openslr/librispeech_asr) dataset module since i refered to the [source code](https://huggingface.co/datasets/openslr/librispeech_asr/blob/main/librispeech_asr.py).
|
18 |
+
|
19 |
+
|
20 |
+
```python
|
21 |
+
from datasets import load_dataset
|
22 |
+
from pprint import pprint
|
23 |
+
from IPython.display import display, Audio
|
24 |
+
import aiohttp
|
25 |
+
import os
|
26 |
+
|
27 |
+
# set huggingface cache directory for the extracted raw files and huggingface-cli token
|
28 |
+
os.environ['HF_HOME'] = "/data/to/download"
|
29 |
+
!export HF_HOME="/data/to/download"
|
30 |
+
|
31 |
+
|
32 |
+
# download dataset
|
33 |
+
# if already have librispeech_asr in the cache_dir it will use the same audio files.
|
34 |
+
libripc = load_dataset("yoom618/librispeech_pc",
|
35 |
+
"all", # all, clean, other
|
36 |
+
cache_dir="/data/to/download",
|
37 |
+
trust_remote_code=True,
|
38 |
+
# storage_options={'client_kwargs': {'timeout': aiohttp.ClientTimeout(total=7200)}}, # add if you need to increase the timeout
|
39 |
+
|
40 |
+
|
41 |
+
# check dataset info
|
42 |
+
print(libripc)
|
43 |
+
|
44 |
+
display(Audio(libripc['train.clean.100'][0]['audio']['array'],
|
45 |
+
rate=libripc['train.clean.100'][0]['audio']['sample_rate']),
|
46 |
+
autoplay=False)
|
47 |
+
pprint(libripc['train.clean.100'][0]['audio'])
|
48 |
+
|
49 |
+
```
|