Update README with code snippets and scripts

#2
by reach-vb HF staff - opened
Files changed (1) hide show
  1. README.md +30 -0
README.md CHANGED
@@ -39,6 +39,7 @@ source_datasets:
39
  - extended|other-common-voice
40
  task_categories:
41
  - automatic-speech-recognition
 
42
  task_ids: []
43
  paperswithcode_id: null
44
  pretty_name: CoVoST 2
@@ -916,6 +917,7 @@ dataset_info:
916
  - [Dataset Summary](#dataset-summary)
917
  - [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
918
  - [Languages](#languages)
 
919
  - [Dataset Structure](#dataset-structure)
920
  - [Data Instances](#data-instances)
921
  - [Data Fields](#data-fields)
@@ -957,6 +959,34 @@ crowdsourced voice recordings. There are 2,900 hours of speech represented in th
957
 
958
  The dataset contains the audio, transcriptions, and translations in the following languages, French, German, Dutch, Russian, Spanish, Italian, Turkish, Persian, Swedish, Mongolian, Chinese, Welsh, Catalan, Slovenian, Estonian, Indonesian, Arabic, Tamil, Portuguese, Latvian, and Japanese.
959
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
960
  ## Dataset Structure
961
 
962
  ### Data Instances
 
39
  - extended|other-common-voice
40
  task_categories:
41
  - automatic-speech-recognition
42
+ - audio-to-audio
43
  task_ids: []
44
  paperswithcode_id: null
45
  pretty_name: CoVoST 2
 
917
  - [Dataset Summary](#dataset-summary)
918
  - [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
919
  - [Languages](#languages)
920
+ - [How to use](#how-to-use)
921
  - [Dataset Structure](#dataset-structure)
922
  - [Data Instances](#data-instances)
923
  - [Data Fields](#data-fields)
 
959
 
960
  The dataset contains the audio, transcriptions, and translations in the following languages, French, German, Dutch, Russian, Spanish, Italian, Turkish, Persian, Swedish, Mongolian, Chinese, Welsh, Catalan, Slovenian, Estonian, Indonesian, Arabic, Tamil, Portuguese, Latvian, and Japanese.
961
 
962
+ ## How to use
963
+ The `datasets` library allows you to load and pre-process your dataset in pure Python, at scale. The dataset can be downloaded and prepared in one call to your local drive by using the `load_dataset` function.
964
+
965
+ For example, to download the English-Turkish config, simply specify the corresponding language config name (i.e., "en_tr" for English to Turkish):
966
+ ```python
967
+ from datasets import load_dataset
968
+
969
+ covost2 = load_dataset("covost2", "en_tr", data_dir="<path/to/manual/data>, split="train")
970
+ ```
971
+ Note: For a successful load, you'd first need to download the Common Voice 4.0 `en` split from https://commonvoice.mozilla.org/en/datasets and unpack it with <code>tar xvzf file_name.tar</code>. Upon successful download pass the location of the CV 4.0 dataset in the `data_dir` argument of the CoVoST2 `load_dataset` call.
972
+
973
+ *Bonus*: create a [PyTorch dataloader](https://huggingface.co/docs/datasets/use_with_pytorch) directly with your own datasets.
974
+
975
+ ```python
976
+ from datasets import load_dataset
977
+ from torch.utils.data.sampler import BatchSampler, RandomSampler
978
+
979
+ covost2 = load_dataset("covost2", "en_tr", data_dir="<path/to/manual/data>, split="train")
980
+ batch_sampler = BatchSampler(RandomSampler(covost2), batch_size=32, drop_last=False)
981
+ dataloader = DataLoader(covost2, batch_sampler=batch_sampler)
982
+ ```
983
+
984
+ To find out more about loading and preparing audio datasets, head over to [hf.co/blog/audio-datasets](https://huggingface.co/blog/audio-datasets).
985
+
986
+ ### Example scripts
987
+
988
+ Train your own CTC or Seq2Seq Speech Translation models on CoVoST2 with `transformers` - [here](https://github.com/huggingface/transformers/tree/main/examples/pytorch/speech-recognition).
989
+
990
  ## Dataset Structure
991
 
992
  ### Data Instances