Prepare a session environment¶
Before running session, you need to set the execution environment of code as below.
- If you can install with pip
Write the python library in setup.py or requirements.txt as shown below.
from distutils.core import setup setup( name='ladder_networks', version='1.0', install_requires=[ 'matplotlib', 'tqdm', 'pillow' ] )
- If you can not install with pip
You can set the execution environment by setting the docker environment. You can write the name of the image of the docker by attaching
'#nsml:'
to the top of setup.py or requirements.txt.
setup.py
#nsml: nsml/default_ml:cuda9 from distutils.core import setup setup( name='ladder_networks', version='1.0', install_requires=[ 'matplotlib', 'tqdm', 'pillow' ] )
requirements.txt
#nsml: nsml/default_ml:cuda9 matplotlib tqdm pillow
Note
If you have both setup.py and requirements.txt, install setup.py first and install requirements.txt.
Note
If you do not list the docker environment, use nsml/default_ml:latest. (latest version may change. )
- If you need to install a package with
apt-get
If you create a
nsml_package.txt
file and specify the required library, it will be installed.
nsml_package.txt
wget vimNote
Installed as apt-get -y update && apt-get -y install [package]. If you need more commands to install it, we recommend that you configure a new docker environment.
NSML ignore¶
When you run session with nsml run , the files that you do not need will be also uploaded and take a long time. If you list files that is not related to session execution in
.nsmlignore
such as session.gitignore
format, you can prevent unneeded files from uploading.
.nsmlignore
images/ models/ data/