Whisper small model for CTranslate2

This repository contains the conversion of fine-tunining Tiny version of Whisper to the CTranslate2 model format.

This model can be used in CTranslate2 or projects based on CTranslate2 such as faster-whisper.

Install faster-whisper

pip install git+https://github.com/guillaumekln/faster-whisper.git

Example

from faster_whisper import WhisperModel

model = WhisperModel("bl4dylion/faster-whisper-tiny-belarusian")

segments, info = model.transcribe("audio.mp3")
for segment in segments:
    print("[%.2fs -> %.2fs] %s" % (segment.start, segment.end, segment.text))

Conversion details

The original model was converted with the following command:

ct2-transformers-converter --model <model_path> --output_dir faster-whisper-tiny-belarusian \
    --copy_files tokenizer_config.json --quantization float16

Note that the model weights are saved in FP16. This type can be changed when the model is loaded using the compute_type option in CTranslate2.