mmalyska's picture
Update README.md (#1)
80043dd verified
metadata
license: apache-2.0
language:
  - pl
pipeline_tag: automatic-speech-recognition
tags:
  - audio
datasets:
  - Aspik101/distil-whisper-large-v3-pl
library_name: ctranslate2

Fine-tuned Polish Aspik101/distil-whisper-large-v3-pl model for CTranslate2

This repository contains the Aspik101/distil-whisper-large-v3-pl model converted to the CTranslate2 format.

Usage

from faster_whisper import WhisperModel
from huggingface_hub import snapshot_download

downloaded_model_path = snapshot_download(repo_id="mmalyska/distil-whisper-large-v3-pl-ct2")

# Run on GPU with FP16
model = WhisperModel(downloaded_model_path, device="cuda", compute_type="float16")
# or run on GPU with INT8
# model = WhisperModel(downloaded_model_path, device="cuda", compute_type="int8_float16")
# or run on CPU with INT8
# model = WhisperModel(downloaded_model_path, device="cpu", compute_type="int8")

segments, info = model.transcribe("./sample.wav", beam_size=1)

print("Detected language '%s' with probability %f" % (info.language, info.language_probability))

for segment in segments:
    print("[%.2fs -> %.2fs] %s" % (segment.start, segment.end, segment.text))

Conversion

The original model was converted with the following command:

ct2-transformers-converter --model Aspik101/distil-whisper-large-v3-pl --output_dir distil-whisper-large-v3-pl-ct2 --copy_files tokenizer.json preprocessor_config.json --quantization float16