docker

How to use

dependences

• docker
• docker-compose

1. Build the image (to do only the first time)

• create a folder for yours datasets:

  mkdir ~/datasets
  mkdir ~/scripts

• open the terminal on the docker folder (where Dockerfile is placed)

• build with command (it can take a lot of time):

  docker build --no-cache=true -t ml-image .

2. Run the image, Jupiter automatically starts

• open the terminal on the docker folder (were docker-compose is placed)

• run the container with the command:

  docker-compose up

• open Jupiter on browser (psw: ml):

  localhost:8888

3. Connect on a new TTY (usefull to run some scripts outside Jupiter)

docker exec -it [container-id] bash

scripts folder

cd ../scripts

Tensorboard

• If you need to use tensorboard run :

  docker exec -it machine_learning_tf tensorboard --logdir tf_logs/

• open it on browser (URL):

  localhost:6006

  Clean up

  • see all the images:

  docker images -a

  • when you modify the Dockerfile is good to clean the dangling images:

  docker images --filter "dangling=true"
  docker rmi $(docker images -q --filter "dangling=true")

other commands

docker build --no-cache=true -t ml-image-gpu -f Dockerfile.gpu .

docker-compose -f docker-compose.yml up

nvidia-docker run \
--rm \
--device /dev/nvidia0:/dev/nvidia0 \
--device /dev/nvidiactl:/dev/nvidiactl \
--device /dev/nvidia-uvm:/dev/nvidia-uvm \
-p 8888:8888 \
-v ~/ML_fileLocali/notebooks:/notebooks/samples \
-v ~/ML_fileLocali/datasets:/datasets \
-v ~/machine_learning/notebooks:/notebooks \
-v ~/machine_learning/scripts:/scripts \
ml-final-8-6