metadata

language:
  - zh
license: apache-2.0
tags:
  - whisper-event
  - generated_from_trainer
base_model: openai/whisper-small
datasets:
  - mozilla-foundation/common_voice_11_0
model-index:
  - name: Distil-Whisper Small zh-HK - Alvin
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: mozilla-foundation/common_voice_16_0 yue
          type: mozilla-foundation/common_voice_16_0
          config: yue
          split: test
          args: yue
        metrics:
          - name: Normalized CER
            type: cer
            value: 9.77

Distil-Whisper Small zh-HK - Alvin

This model is a distilled and fine-tuned version of openai/whisper-small on the Cantonese language. It achieves a 9.77 CER (without punctuations), 11.7 CER (with punctuations) on Common Voice 16.0

Training and evaluation data

For training,

CantoMap: Winterstein, Grégoire, Tang, Carmen and Lai, Regine (2020) "CantoMap: a Hong Kong Cantonese MapTask Corpus", in Proceedings of The 12th Language Resources and Evaluation Conference, Marseille: European Language Resources Association, p. 2899-2906.
Cantonse-ASR: Yu, Tiezheng, Frieske, Rita, Xu, Peng, Cahyawijaya, Samuel, Yiu, Cheuk Tung, Lovenia, Holy, Dai, Wenliang, Barezi, Elham, Chen, Qifeng, Ma, Xiaojuan, Shi, Bertram, Fung, Pascale (2022) "Automatic Speech Recognition Datasets in Cantonese: A Survey and New Dataset", 2022. Link: https://arxiv.org/pdf/2201.02419.pdf
Common Voice yue and zh-HK train sets

For evaluation, Common Voice 16.0 yue Test set is used.

Results

CER (lower is better): 0.117
GPU Inference with Fast Attention (example below): 0.039s/sample
- Note all GPU evaluations are done on RTX 3090 GPU
GPU Inference: s/sample
CPU Inference: 2.57s/sample
GPU VRAM: ~2 GB

Using the Model

import librosa

import torch
from transformers import WhisperForConditionalGeneration, WhisperProcessor

y, sr = librosa.load('audio.mp3', sr=16000)

MODEL_NAME = "alvanlii/distil-whisper-small-cantonese"

processor = WhisperProcessor.from_pretrained(MODEL_NAME)
model = WhisperForConditionalGeneration.from_pretrained(MODEL_NAME)

model.config.forced_decoder_ids = None
model.config.suppress_tokens = []
model.config.use_cache = False

processed_in = processor(y, sampling_rate=sr, return_tensors="pt")
gout = model.generate(
    input_features=processed_in.input_features, 
    output_scores=True, return_dict_in_generate=True
)
transcription = processor.batch_decode(gout.sequences, skip_special_tokens=True)[0]
print(transcription)

Alternatively, you can use huggingface pipelines

from transformers import pipeline
MODEL_NAME = "alvanlii/distil-whisper-small-cantonese" 
lang = "zh"
pipe = pipeline(
    task="automatic-speech-recognition",
    model=MODEL_NAME,
    chunk_length_s=30,
    device=device,
)
pipe.model.config.forced_decoder_ids = pipe.tokenizer.get_decoder_prompt_ids(language=lang, task="transcribe")
text = pipe(file)["text"]

Model Speedup

Just add attn_implementation="sdpa" for Flash Attention.

model = AutoModelForSpeechSeq2Seq.from_pretrained(
    "alvanlii/distil-whisper-small-cantonese",
    torch_dtype=torch_dtype,
    low_cpu_mem_usage=True,
    use_safetensors=True,
    attn_implementation="sdpa",
)

Using Flash Attention reduced the amount of time taken per sample from s to 0.039s.