Build whisper on Jetson Nano Orin (Cuda 12.6)

README.md

[DRAFT] 1- Update Nano

apt-get update && apt upgrade
apt-get install cmake

2- clone repo and go to source

git clone https://github.com/ggml-org/whisper.cpp.git
cd whisper.cpp

3- Build Check nvidia compiler

nvcc --version

If missing (check folder /usr/local/cuda-12.6/bin )
Install cuda toolkit

sudo apt install cuda-toolkit-12-6
nvcc --version

Build configuration & build

cmake -B build -DGGML_CUDA=ON -DCMAKE_BUILD_TYPE=Release -DCMAKE_CUDA_COMPILER=/usr/local/cuda-12/bin/nvcc
cmake --build build -- -j2

And wait a long long time (3hours)
4- Download model

cd models
download-ggml-model.sh base

5- Run Nota : bin is in build folder
Test with wav files

./build/bin/whisper-cli -m models/ggml-base.bin -f samples/jfk.wav

use -l option for language
Start server

./build/bin/whisper-server -m models/ggml-base.bin --port 8080

check examples folder for different use case