Whisper small-dv model for CTranslate2
This repository contains the conversion of whisper-small-dv to the CTranslate2 model format. The model is a finetuned version of openai/whisper-small to Divehi language using the Common Voice 13 dataset This model can be used in CTranslate2 or projects based on CTranslate2 such as faster-whisper.
Example
from faster_whisper import WhisperModel
# load from local folder
# model = WhisperModel("whisper-small-dv-ct2")
# load from the hub
model = WhisperModel("davidggphy/whisper-small-dv-ct2")
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 davidggphy/whisper-small-dv --output_dir whisper-small-dv-ct2 --copy_files tokenizer.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.