---
language:
- zh
license: cc-by-nc-sa-4.0
library_name: transformers
tags:
- audio
- automatic-speech-recognition
widget:
- example_title: Model Introduction 
  src: https://huggingface.co./andybi7676/cool-whisper-hf/resolve/main/sample1.weba
pipeline_tag: automatic-speech-recognition
---

# Cool-Whisper

### Leave No Knowledge Behind During Knowledge Distillation: Towards Practical and Effective Knowledge Distillation for Code-Switching ASR Using Realistic Data
<span style="font-size: 0.95em;">Liang-Hsuan Tseng, Zih-Ching Chen, Wei-Shun Chang, Cheng-Kuang Lee, Tsung-Ren Huang, Hung-yi Lee</span>

[![arXiv](https://img.shields.io/badge/arXiv-Paper-color.svg)](https://arxiv.org/abs/2407.10603) [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1ZikUWKch78Jv3Yw7LtUKUn4wMrFCx6lD?usp=sharing)

> ⚠️ Due to privacy and security concerns, this model will be temporarily taken offline. We are sorry for the inconvenience.

> ⚠️ 因為隱私安全疑慮，本模型將暫時下架。非常抱歉造成大家困擾。

## Introduction

* Cool-whisper is a distilled version of Whisper, mainly focused on **Mandarin-English** code-switching ASR for people in Taiwan. 
* We use 60,000 hours of **unlabeled** audio to train the model.
* Practically, we utilize *knowledge* not only from the large model (Whisper-large-v2) but also from the small model (Whisper-base).

## Basic Usage

``` python
import torch
from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline
from datasets import load_dataset

device = f"cuda" if torch.cuda.is_available() else "cpu"
torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32

model_id = "andybi7676/cool-whisper-hf"

model = AutoModelForSpeechSeq2Seq.from_pretrained(
    model_id, torch_dtype=torch_dtype, use_safetensors=True
)
processor = AutoProcessor.from_pretrained(model_id)

pipe = pipeline(
    "automatic-speech-recognition",
    model=model,
    tokenizer=processor.tokenizer,
    feature_extractor=processor.feature_extractor,
    max_new_tokens=256,
    return_timestamps=True,
    torch_dtype=torch_dtype,
    device=device,
)

dataset = load_dataset("andybi7676/ntuml2021_long", "default", split="test")
sample = dataset[0]["audio"]
# or your own audio path
# sample = "/your/path/to/audio.wav"

result = pipe(sample)
print("Basic Result: ")
print(result["text"])
# result with timestamps
print("\nResult with timestamps: ")
for chunk in result['chunks']:
  print(chunk)
```

## Faster-Whisper Support

[Faster-Whisper](https://github.com/SYSTRAN/faster-whisper) is a commonly used tool to accelerate the transcription generation speed based on [CTranslate2](https://github.com/OpenNMT/CTranslate2/). 
We also deploy our model in the form of CTranslate2 to allow using it in faster-whisper. 
Please visit [cool-whisper](https://huggingface.co./andybi7676/cool-whisper) for more details.