Run a session

Usage:

nsml run Options

Description:

It sends the files in the current location to the NSML server to run the session. Current location’s main.py is recognized as an entry file and if it does not exist, an error will occur. You can change the entry file with the -e option.

Example:

nsml run -d mnist

Must specify the dataset name to use when you run the session.

Options:

-d, --dataset string

Select the dataset to use when the session runs.

-g, --gpus string_or_integer

Enter the number of gpu you want to assign. ex) nsml run -g 4

set option -g as 0 if you want to use only cpu for your session.

The detail rules for the GPU usage, check Policy.

ex) nsml run -g 2

-i, --interactive

the session is automatically deleted when you finish training the session.

-m, --memo string

can type simple memo to session

ex) nsml run -m ‘LSTM’

-e, --entry file

Sets the main entry file at the time the session is run.

ex) nsml run -e lstm.py

-a, --args string

Passes the arguments in the main entry file.

ex) nsml run -a “–batch 1000 –lr 0.02”

-c, --cpus string_or_integer

The number of cpu to use. ex) nsml run -c 5

--memory string

CPU Memory size required. (default ‘24G’)

--shm-size string

Required share memory size. (default ‘1G’)