---
language:
- ms
dataset_info:
  features:
  - name: id
    dtype: int64
  - name: prompt
    dtype: string
  - name: answer
    dtype: string
  - name: year
    dtype: string
  - name: subject
    dtype: string
  - name: subject_eng
    dtype: string
  - name: category
    dtype: string
  - name: level
    dtype: string
  - name: options
    sequence: string
  - name: num_options
    dtype: int64
  - name: key
    dtype: string
  - name: audio_filename
    dtype:
      audio:
        sampling_rate: 16000
  splits:
  - name: train
    num_bytes: 856803711.808
    num_examples: 14389
  download_size: 859589581
  dataset_size: 856803711.808
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

# Speech Instructions MalayMMLU

We filter [MalayMMLU](https://github.com/UMxYTL-AI-Labs/MalayMMLU) that suitable to convert to voice prompt and generate synthetic voice using [mesolitica/Malaysian-F5-TTS-v2](https://huggingface.co/mesolitica/Malaysian-F5-TTS-v2) also verified with [Force Alignment](https://github.com/MahmoudAshraf97/ctc-forced-aligner) to make sure the pronunciations almost correct.