Whisper large-v3 model for Kinyarwanda in CTranslate2 format

This repository contains the conversion of leophill/whisper-large-v3-sn-kinyarwanda to the CTranslate2 model format.

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

Example

from faster_whisper import WhisperModel
model = WhisperModel("leophill/whisper-large-v3-sn-kinyarwanda-ct2")
language = "sn"
beam_size = 5
best_of = 5
decode_options = dict(language=language, beam_size=beam_size, best_of = best_of, vad_filter=True, vad_parameters=dict(min_silence_duration_ms=500), word_timestamps=False)
audio_file = "audio.wav"
segments, info = model.transcribe(audio_file, **decode_options)
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 leophill/whisper-large-v3-sn-kinyarwanda --output_dir whisper-large-v3-sn-kinyarwanda-ct2 \
    --copy_files tokenizer.json preprocessor_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.

More information

For more information about the original model, please see its model card.