---
dataset_info:
  features:
  - name: audio
    dtype: audio
  - name: transcript
    dtype: string
  - name: source
    dtype: string
  splits:
  - name: test
    num_bytes: 980296593.0
    num_examples: 9352
  - name: validation
    num_bytes: 899503003.0
    num_examples: 8686
  - name: train
    num_bytes: 78822585753.0
    num_examples: 627822
  download_size: 80661360008
  dataset_size: 80702385349.0
configs:
- config_name: default
  data_files:
  - split: test
    path: data/test-*
  - split: validation
    path: data/validation-*
  - split: train
    path: data/train-*
task_categories:
- automatic-speech-recognition
language:
- kk
pretty_name: 🎤🇰🇿 kazakh_speech_corpus_2 📚
size_categories:
- 100K<n<1M
---
# Kazakh_speech_dataset_2
This dataset contains [Kazakh_speech_dataset_2](https://huggingface.co/datasets/issai/Kazakh_Speech_Corpus_2) from ISSAI but in parquet format.
## Dataset info 
1. 645,860 Utterances
2. 1194 Hours in total
3. Sources in each split:
```json
test : {'tv_news', 'crowdsourced', 'radio', 'talkshow', 'parliament', 'tts', 'podcasts'}
train : {'tv_news', 'crowdsourced', 'radio', 'talkshow', 'parliament', 'tts', 'podcasts'}
validation : {'tv_news', 'crowdsourced', 'radio', 'talkshow', 'parliament', 'tts','podcasts'}
```


## Guides
### Load data 1
Replace the `export HF_HOME` with your HF_HOME path
```python
from datasets import load_dataset
# export HF_HOME="/data/vladimir_albrekht/hf_cache"
ds = load_dataset("SRP-base-model-training/kazakh_speech_corpus_2") # split ='test' or 'validation' or 'train'
```
### Load data 2 patially test/train/validation
`data_files = [f for f in files if f.startswith('')]` # to download something specific use this (data/test or data/train or data/validation)
1. **Step** download subset test/train/validation
   
```python
from huggingface_hub import hf_hub_download, login, HfApi
import os
import dotenv
dotenv.load_dotenv()
login(token=os.getenv("HF_TOKEN"))
repo_id = "SRP-base-model-training/kazakh_speech_corpus_2"
local_dir = "/data/vladimir_albrekht/asr/kazakh_speech_corpus_2"
def download_all_files(repo_id, local_dir):
    print(f"Downloading files from {repo_id}...")
    os.makedirs(local_dir, exist_ok=True)
    
    api = HfApi()
    files = api.list_repo_files(repo_id, repo_type="dataset")
    
    data_files = [f for f in files if f.startswith('')] # to download something specific use this (data/test or data/train or data/validation)
    
    for filename in data_files:
        try:
            hf_hub_download(
                repo_id=repo_id,
                filename=filename,
                repo_type="dataset",
                local_dir=local_dir
            )
            print(f"Downloaded {filename}")
        except Exception as e:
            print(f"Error downloading {filename}: {e}")
download_all_files(repo_id, local_dir)
```
2. **Step** load data with ds as usual.
```
from datasets import load_dataset
ds = load_dataset("/data/vladimir_albrekht/asr/kazakh_speech_corpus_2")
```