README.md

metadata

library_name: mlx
license: apache-2.0
language:
  - km
pipeline_tag: automatic-speech-recognition
datasets:
  - seanghay/km-speech-corpus
  - seanghay/khmer_mwpt_speech
tags:
  - Khmer
  - mlx
base_model: openai-whisper-base
model-index:
  - name: whisper-base-khmer-mlx-fp32 by Kimang KHUN
    results:
      - task:
          type: automatic-speech-recognition
          name: Speech Recognition
        dataset:
          name: test split of "km_kh" in google/fleurs
          type: google/fleurs
        metrics:
          - type: wer
            value: 69.7%
            name: test

whisper-base-khmer-mlx-fp32

This model was converted to MLX format from openai-whisper-base, then fine-tined to Khmer language using three datasets:

• seanghay/khmer_mpwt_speech
• seanghay/km-speech-corpus
• train split of openslr/openslr SLR42

It achieves the following word error rate (wer) on 2 popular datasets:

• 69.7% on test split of google/fleurs km-kh

NOTE MLX format is usable for M-chip series of Apple.

Use with mlx

pip install mlx-whisper

Write a python script, example.py, as the following

import mlx_whisper

result = mlx_whisper.transcribe(
    SPEECH_FILE_NAME,
    path_or_hf_repo="mlx-community/whisper-base-khmer-mlx-fp32",
    fp16=False
)
print(result['text'])

Then execute this script example.py to see the result.

You can also use command line in terminal

mlx_whisper --model mlx-community/whisper-base-khmer-mlx-fp32 --task transcribe SPEECH_FILE_NAME --fp16 False