Whisper——部署fast-whisper中文语音识别模型

本文介绍: 需要其他版本的可以自己下载：https://huggingface.co/openai。

whisper：https://github.com/openai/whisper/tree/main
参考文章：Whisper OpenAI开源语音识别模型

环境配置

pip install faster-whisper transformers

准备tiny模型

需要其他版本的可以自己下载：https://huggingface.co/openai

原始中文语音模型：

https://huggingface.co/openai/whisper-tiny

微调后的中文语音模型：

git clone https://huggingface.co/xmzhu/whisper-tiny-zh

补下一个：tokenizer.json

https://huggingface.co/openai/whisper-tiny/resolve/main/tokenizer.json?download=true

模型转换

float16：

ct2-transformers-converter --model whisper-tiny-zh/ --output_dir whisper-tiny-zh-ct2 --copy_files tokenizer.json preprocessor_config.json --quantization float16

int8：

ct2-transformers-converter --model whisper-tiny-zh/ --output_dir whisper-tiny-zh-ct2-int8 --copy_files tokenizer.json preprocessor_config.json --quantization int8

代码

from faster_whisper import WhisperModel

# model_size = "whisper-tiny-zh-ct2"
# model_size = "whisper-tiny-zh-ct2-int8"

# Run on GPU with FP16
# model = WhisperModel(model_size, device="cuda", compute_type="float16")
model = WhisperModel(model_size, device="cpu", compute_type="int8")

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

segments, info = model.transcribe("output_file.wav", beam_size=5, language='zh')

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))