docker

InstantDL with Docker

Here you can find the information of using InstantDL with docker.

Build a Docker container

first you need to download the repository. After that, being in the main folder, simply run

sudo docker/bin/docker_build_context.sh
sudo docker build --tag=instantdl:0.0.1 build

Prepration the data folder

For runnig the docker, first you need to create the data folder as exaplined in the examples. In case, you would like to pretrained weights as well, first create a folder called logs under the data folder and simply put the .hdf5 file in the folders. After that, set all the parameters according to the documentation EXCEPT path and pretrained_weights, and save the parameters as config.json in the data folder.

IMPORTANT NOTE: For the config file, in the json file, these parameters data and pretrained_weights should BE ALWAYS:

{
    ...
	"path": "/data/",
	"pretrained_weights": "/data/logs/pretrained_weights_Lung_SemanticSegmentation.hdf5",
    ...
}

The reason for this is that the folder will be mounted to the docker with the address /data/

Data Folder Structure

This is a folder structure of the data folder:

PATH_TO_DATA
    ├── train                    
    │   ├── image
    │   │    ├── 000003-num1.png
    │   │    ├── .
    │   │    └── 059994-num1.png     
    │   └── groundtruth  
    │        └── groundtruth.csv
    │
    ├── test                    
    │  ├── image
    │  │    ├── 000002-num1.png
    │  │    ├── .
    │  │    └── 009994-num1.png     
    │  └── groundtruth  
    │       └── groundtruth.csv
    ├── logs                    
    │  └── weights.hdf5
    |
    |
    └── config.json

Run Docker container

Runnig the code with GPU

docker run --gpus all -v PATH_TO_DATA/:/data -it instantdl:0.0.1 /bin/bash

Starting the docker containter terminal

docker run --gpus all -v PATH_TO_DATA/:/data -it instantdl:0.0.1 /bin/bash

In case you do not have GPU, you can omit the --gpus all.

GPU support

We are using the docker from tensorflow. To Check if a GPU is available:

lspci | grep -i nvidia

Verify your nvidia-docker installation:

sudo docker run --gpus all --rm nvidia/cuda nvidia-smi

Cleanup

Remove all containers and images:

docker system prune -a

TODO

• add cuda and gpu
• add how to create data