Installation

Prerequisites

  • Python 3.8 | 3.9 | 3.10

  • PyTorch 1.13.0 (recommended) (compatible with: 1.11.x - 1.13.x)

  • TensorFlow 2.8.0 (recommended) (compatible with: 2.3.x - 2.8.x)

Install with PyPI (stable)

pip install netspresso

Install with GitHub

To install with editable mode,

git clone https://github.com/nota-netspresso/pynetspresso.git
cd pynetspresso
pip install -e .

Install with Docker

Please clone this repository and refer to Dockerfile and docker-compose.yml.

Docker with docker-compose

For the latest information, please check docker-compose.yml.

# run command
export TAG=v$(cat netspresso/VERSION) && \
docker compose run --service-ports --name netspresso-dev netspresso bash

Docker image build

If you run with docker run command, follow the image build and run command in the below:

# build an image
docker build -t netspresso:v$(cat netspresso/VERSION) .

# docker run command
docker run -it --ipc=host\
  --gpus='"device=0,1,2,3"'\  # your GPU id(s)
  -v /PATH/TO/DATA:/DATA/PATH/IN/CONTAINER\
  -v /PATH/TO/CHECKPOINT:/CHECKPOINT/PATH/IN/CONTAINER\
  -p 50001:50001\
  -p 50002:50002\
  -p 50003:50003\
  -p 50004:50004\
  --name netspresso-dev netspresso:v$(cat netspresso/VERSION)