Views
No views yet
| File | Size | Format | Description |
|---|---|---|---|
ggml-whisper-large-v3-turbo-zh-TW.bin | 1.6GB | F16 | Full precision, best quality |
ggml-whisper-large-v3-turbo-zh-TW-q8_0.bin | 874MB | Q8_0 | Quantized, recommended for mobile devices |
1# Clone whisper.cpp
2git clone https://github.com/ggml-org/whisper.cpp
3cd whisper.cpp
4
5# Build
6cmake -B build && cmake --build build --config Release
7
8# Download model
9wget https://huggingface.co/leaker/whisper-large-v3-turbo-zh-TW-ggml/resolve/main/ggml-whisper-large-v3-turbo-zh-TW-q8_0.bin
10
11# Run inference
12./build/bin/whisper-cli -m ggml-whisper-large-v3-turbo-zh-TW-q8_0.bin -f audio.wav -l zhconvert-h5-to-ggml.py script from whisper.cpp:1python3 ./models/convert-h5-to-ggml.py \
2 /path/to/JacobLinCool/whisper-large-v3-turbo-common_voice_19_0-zh-TW \
3 /path/to/openai/whisper \
4 /output/path
5
6# Quantize to Q8_0
7./build/bin/whisper-quantize input.bin output-q8_0.bin q8_0