Dockerfile | Description |
---|---|
Dockerfile | Default docker file. Nothing special. |
Dockerfile.cpu_noavx | Building PyTorch from sourcecode for NoGPU and NoAVX. |
I have not tested in special hardware environments. If you have any problems, please post them in the issue. Also, I am not familiar with Docker. If you are familiar with Docker, do it your way.
docker build -t nunif Dockerfiles
When specifying a docfile (For example Dockerfiles/Dockerfile.cpu_noavx
)
docker build -t nunif -f Dockerfiles/Dockerfile.cpu_noavx Dockerfiles
docker run --gpus all -p 8812:8812 --rm nunif python3 -m waifu2x.web --port 8812 --bind-addr 0.0.0.0 --no-size-limit
For CPU only mode (use --gpu -1
option)
docker run -p 8812:8812 --rm nunif python3 -m waifu2x.web --port 8812 --bind-addr 0.0.0.0 --no-size-limit --gpu -1
For CLI commands, it is required to access the input and output directories from the docker container side to the host side. So, mount the drive on the host side and use it.
docker run --gpus all -v path_to_input_dir:/input_dir -v path_to_output_dir:/output_dir --rm nunif python3 -m waifu2x.cli -m scale -i /input_dir -o /output_dir
Note that directory paths for -v
option must be specified as absolute paths.
docker run --gpus all -v `pwd`/waifu2x/docs/images:/input_dir -v /tmp/output:/output_dir --rm nunif python3 -m waifu2x.cli -m scale -i /input_dir -o /output_dir
See also waifu2x Command Line Interface for example commands.