NOTE: the manually build version with more features is now moved to guoq/imxnet-manual.
MXNet is a famous open source deep learning framework. Setting up the environment to run MXNet is a little bit complicated. Thanks to Kaixhin's images, we can use MXNet with Docker without such pain.
To further satisfiy the lazy (in fact, I spell this word T-A-L-E-N-T) guys who want to play with MXNet more cosily, for example, with kitematic
UI controls or with jupyter
(F/K/A ipython notebook
) to work in a browser rather than a terminal, as myself, I build this repository.
This repository is based on kaixhin/mxnet and I will be looking at the CUDA
versions soon.
- Interactive with
ipython
. - Directly access the amazing deep learning tools right within a browser, with
jupyter
. - Compatible with
kitematic
where you can run and maintain the container in a snap. Open your
JUST OPEN YOUR BROWSER AND SEE EVERYTHING READY!
Currently, I have just pushed the develop
tag.
-
with
kitematic
If you are using
kitematic
, select this tag and create a container than everything is OK. -
with a command line
If you are using command line, just pull with this tag:
docker pull guoq/imxnet:develop
and run it with exposed ports published. And the workspace binded to a host directory:
bash
docker run -d -P -v $(pwd):/root/workspace --name=imxnet guoq/imxnet:develop
Use `port` to find the port of your `jupyter` service:
bash
docker port imxnet
```
If you want to specify the port on the host to be use, the default port for `jupyter` is `8888`. The image use `tini` to inisialize the container and defaultly run `jupyter` by CMD:
```bash
jupyter notebook --ip=0.0.0.0 --port=8888 --no-browser
```
Override the command in run
to specify yours.
There is a terminal in `jupyter` where you can also interactive the container.
To stop the service, just stop the container:
```bash
docker stop imxnet ```
- Add
tmux
to allow actually working inside the container. - Add an entrypoint script to switch
inside mode
andoutside mode
. - The entrypoint script should also tackle with the
libdc1394
error message. - Add
your favoritezsh
+ohmyzsh
+vim
. - Add different
CUDA
versions.