nnabla-browser is a browser-based visualizer for Neural Network Libraries.
python >= 3.6
$ pip install nnabla-browser
Once you have installed nnabla-browser in your environment, you can launch server from anywhere as long as you use same python environment.
$ nnabla-browser --logdir /path/to/logdir --port PORT
For macOS users, you might have to set an environment variable to allow NNabla Browser to use multi-process as follows:
export OBJC_DISABLE_INITIALIZE_FORK_SAFETY=YES
Launch any browser application and just access to <server-address>:PORT/
For example, when you launch server on the same computer which browser can be used, you can access localhost:8888/
- pipenv
- npm
# install javascript dependencies
$ npm install
# install python dependencies (`--pre` is required to install black.)
$ pipenv install --pre
$ npm run build
# launch python server
$ pipenv run dev
# launch frontend server with hot-reloading
$ npm run serve
Then, access to localhost:8000
from your browser.
Using npm installed in your environment, you can create wheel as follows.
$ make nnabla-browser-wheel # npm is required
$ pip install ./dist/nnablabrowser-<version>-<env>.whl
We also provide docker image so that you don't have to set up environment on your server:
$ make bwd-nnabla-browser-wheel # `bwd` stands for build with docker
$ pip install ./dist/nnablabrowser-<version>-<env>.whl
Note that .whl file is created in ./dist for default setting.