Skip to content

Latest commit

 

History

History
57 lines (35 loc) · 2.34 KB

DOCKER_README.md

File metadata and controls

57 lines (35 loc) · 2.34 KB
Raw

Cloud setup (Docker)

Instructions to setup environment and agent on a machine (e.g. cloud).

The agent image can be pulled with:

$ docker pull kaikun213/my-neural-net

The environment image can be pulled with the command:

$ docker pull kaikun213/my-world-of-bits

Afterwards simply start a docker container daemon with in shell mode:

$ docker run --privileged --rm -it -d -e DOCKER_NET_HOST=172.17.0.1 -v /var/run/docker.sock:/var/run/docker.sock -v /home/kaikun/my_neural_net/src/:/usr/local/universe/my_neural_net/src --name "dynamic_NewSensoryLayer" kaikun213/my-neural-net:0.1 bash

Where -v /home/kaikun/my_neural_net/src/:/usr/local/universe/my_neural_net/src is mounting the server directory. This is not necessarily needed. The process is not yet optimized for running multiple experiments/containers in parallel.

And cd into the src folder: $ cd /usr/local/universe/my_neural_net/src

Here the python script can be run: $ python myExample.py

It will start the agent which launches a docker container with the environment. To exit the docker daemon without stopping it press STR+P and STR+Q. To re-attach just write docker attach CONTAINER-ID. It can be stopped with docker stop CONTAINER-ID or writing exit in the container bash.

All docker images can be seen with: docker images

And all currently running docker containers including the VNC ports to connect to the environment with: docker ps -a

Connecting via VNC to observe an experiment simply with some software (e.g. Remmina Remote Desktop Client Ubuntu): HOSTNAME:PORT The password to connect is openai.

Copy or modify

Files from a running container (e.g. results) can be copied with:

docker cp CONTAINER-ID:/usr/local/universe/my_neural_net/src/results ./results

Example how new files can be added to the docker image (copy them, or if remove original first if replacement):

docker cp $MY_NEURAL_NET/src CONTAINER-ID:/usr/local/universe/my_neural_net/src

And after leave the running container with STR-P and STR-Q to commit the changes to the image:

docker commit -m "COMMIT_MESSAGE" CONTAINER-ID my-neural-net:0.1

Or to rebuild the image: docker build -t my-neural-net ..

To publish it online: Tag it docker tag my-neural-net:0.1 $DOCKER_ID/my-neural-net:0.1 and push docker push DOCKER_ID/my-neural-net:0.1