job
Job operations

neuro job [OPTIONS] COMMAND [ARGS]...
Job operations.
Commands:
Usage
Description
attach
Attach terminal to a job
browse
Opens a job's URL in a web browser
Increase job life span
exec
Execute command in a running job
Generate command that will rerun given job
kill
Kill job(s)
logs
Print the logs for a job
ls
List all jobs
Forward port(s) of a job
run
Run a job
save
Save job's state to an image
status
Display status of a job
top
Display GPU/CPU/Memory usage

Attach terminal to a job

neuro job attach [OPTIONS] JOB
Attach terminal to a job
Attach local standard input, output, and error streams to a running job.

Name
Description
--help
Show this message and exit.
--port-forward LOCAL_PORT:REMOTE_RORT
Forward port(s) of a running job to local port(s) (use multiple times for forwarding several ports)

Opens a job's URL in a web browser

neuro job browse [OPTIONS] JOB
Opens a job's URL in a web browser.

Name
Description
--help
Show this message and exit.

Increase job life span

neuro job bump-life-span [OPTIONS] JOB TIMEDELTA
Increase job life span

Name
Description
--help
Show this message and exit.

Execute command in a running job

neuro job exec [OPTIONS] JOB -- CMD...
Execute command in a running job.

# Provides a shell to the container:
$ neuro exec my-job -- /bin/bash
# Executes a single command in the container and returns the control:
$ neuro exec --no-tty my-job -- ls -l

Name
Description
--help
Show this message and exit.
-t, --tty / -T, --no-tty
Allocate a TTY, can be useful for interactive jobs. By default is on if the command is executed from a terminal, non-tty mode is used if executed from a script.

Generate command that will rerun given job

neuro job generate-run-command [OPTIONS] JOB
Generate command that will rerun given job.

# You can use the following to directly re-execute it:
$ eval $(neuro job generate-run-command <job-id>)

Name
Description
--help
Show this message and exit.

Kill job(s)

neuro job kill [OPTIONS] JOBS...
Kill job(s).

Name
Description
--help
Show this message and exit.

Print the logs for a job

neuro job logs [OPTIONS] JOB
Print the logs for a job.

Name
Description
--help
Show this message and exit.
--since DATE_OR_TIMEDELTA
Only return logs after a specific date (including). Use value of format '1d2h3m4s' to specify moment in past relatively to current time.
--timestamps
Include timestamps on each line in the log output.

List all jobs

neuro job ls [OPTIONS]
List all jobs.

$ neuro ps -a
$ neuro ps -a --owner=user-1 --owner=user-2
$ neuro ps --name my-experiments-v1 -s failed -s succeeded
$ neuro ps --description=my favourite job
$ neuro ps -s failed -s succeeded -q
$ neuro ps -t tag1 -t tag2

Name
Description
--help
Show this message and exit.
-a, --all
Show all jobs regardless the status.
--cluster CLUSTER
Show jobs on a specified cluster (the current cluster by default).
-d, --description DESCRIPTION
Filter out jobs by description (exact match).
--distinct
Show only first job if names are same.
--format COLUMNS
Output table format, see "neuro help ps-format" for more info about the format specification. The default can be changed using the job.ps-format configuration variable documented in "neuro help user-config"
--full-uri
Output full image URI.
-n, --name NAME
Filter out jobs by name.
-o, --owner TEXT
Filter out jobs by owner (multiple option). Supports ME option to filter by the current user.
--recent-first / --recent-last
Show newer jobs first or last
--since DATE_OR_TIMEDELTA
Show jobs created after a specific date (including). Use value of format '1d2h3m4s' to specify moment in past relatively to current time.
-s, --status [pending | suspended | running | succeeded | failed | cancelled]
Filter out jobs by status (multiple option).
-t, --tag TAG
Filter out jobs by tag (multiple option)
--until DATE_OR_TIMEDELTA
Show jobs created before a specific date (including). Use value of format '1d2h3m4s' to specify moment in past relatively to current time.
-w, --wide
Do not cut long lines for terminal width.

Forward port(s) of a job

neuro job port-forward [OPTIONS] JOB LOCAL_PORT:REMOTE_RORT...
Forward port(s) of a job.
Forwards port(s) of a running job to local port(s).

# Forward local port 2080 to port 80 of job's container.
# You can use http://localhost:2080 in browser to access job's served http
$ neuro job port-forward my-fastai-job 2080:80
# Forward local port 2222 to job's port 22
# Then copy all data from container's folder '/data' to current folder
# (please run second command in other terminal)
$ neuro job port-forward my-job-with-ssh-server 2222:22
$ rsync -avxzhe ssh -p 2222 [email protected]:/data .
# Forward few ports at once
$ neuro job port-forward my-job 2080:80 2222:22 2000:100

Name
Description
--help
Show this message and exit.

Run a job

neuro job run [OPTIONS] IMAGE [-- CMD...]
Run a job
IMAGE docker image name to run in a job.
CMD list will be passed as arguments to the executed job's image.

# Starts a container pytorch:latest on a machine with smaller GPU resources
# (see exact values in `neuro config show`) and with two volumes mounted:
# storage:/<home-directory> --> /var/storage/home (in read-write mode),
# storage:/neuromation/public --> /var/storage/neuromation (in read-only mode).
$ neuro run --preset=gpu-small --volume=storage::/var/storage/home:rw \
$ --volume=storage:/neuromation/public:/var/storage/home:ro pytorch:latest
# Starts a container using the custom image my-ubuntu:latest stored in neuro
# registry, run /script.sh and pass arg1 and arg2 as its arguments:
$ neuro run -s cpu-small --entrypoint=/script.sh image:my-ubuntu:latest -- arg1 arg2

Name
Description
--help
Show this message and exit.
--browse
Open a job's URL in a web browser
--cluster CLUSTER
Run job in a specified cluster
-d, --description DESC
Optional job description in free format
--detach
Don't attach to job logs and don't wait for exit code
--entrypoint TEXT
Executable entrypoint in the container (note that it overwrites ENTRYPOINT and CMD instructions of the docker image)
-e, --env VAR=VAL
Set environment variable in container. Use multiple options to define more than one variable. See neuro help secrets for information about passing secrets as environment variables.
--env-file PATH
File with environment variables to pass
-x, --extshm / -X, --no-extshm
Request extended '/dev/shm' space [default: x]
--http-auth / --no-http-auth
Enable HTTP authentication for forwarded HTTP port [default: True]
--http-port PORT
Enable HTTP port forwarding to container [default: 80]
--life-span TIMEDELTA
Optional job run-time limit in the format '1d2h3m4s' (some parts may be missing). Set '0' to disable. Default value '1d' can be changed in the user config.
-n, --name NAME
Optional job name
--org ORG
Run job in a specified org
--pass-config / --no-pass-config
Upload neuro config to the job [default: no-pass-config]
--port-forward LOCAL_PORT:REMOTE_RORT
Forward port(s) of a running job to local port(s) (use multiple times for forwarding several ports)
-s, --preset PRESET
Predefined resource configuration (to see available values, run neuro config show)
--priority [low | normal | high]
Priority used to specify job's start order. Jobs with higher priority will start before ones with lower priority. Priority should be supported by cluster.
--privileged
Run job in privileged mode, if it is supported by cluster.
--restart [never | on-failure | always]
Restart policy to apply when a job exits [default: never]
--schedule-timeout TIMEDELTA
Optional job schedule timeout in the format '3m4s' (some parts may be missing).
--share USER
Share job write permissions to user or role.
--tag TAG
Optional job tag, multiple values allowed
-t, --tty / -T, --no-tty
Allocate a TTY, can be useful for interactive jobs. By default is on if the command is executed from a terminal, non-tty mode is used if executed from a script.
-v, --volume MOUNT
Mounts directory from vault into container. Use multiple options to mount more than one volume. See neuro help secrets for information about passing secrets as mounted files.
--wait-for-seat / --no-wait-for-seat
Wait for total running jobs quota [default: no-wait-for-seat]
--wait-start / --no-wait-start
Wait for a job start or failure [default: wait-start]
-w, --workdir TEXT
Working directory inside the container

Save job's state to an image

neuro job save [OPTIONS] JOB IMAGE
Save job's state to an image.

$ neuro job save job-id image:ubuntu-patched
$ neuro job save my-favourite-job image:ubuntu-patched:v1
$ neuro job save my-favourite-job image://bob/ubuntu-patched

Name
Description
--help
Show this message and exit.

Display status of a job

neuro job status [OPTIONS] JOB
Display status of a job.

Name
Description
--help
Show this message and exit.
--full-uri
Output full URI.

Display GPU/CPU/Memory usage

neuro job top [OPTIONS] [JOBS]...
Display GPU/CPU/Memory usage.

$ neuro top
$ neuro top job-1 job-2
$ neuro top --owner=user-1 --owner=user-2
$ neuro top --name my-experiments-v1
$ neuro top -t tag1 -t tag2

Name
Description
--help
Show this message and exit.
--cluster CLUSTER
Show jobs on a specified cluster (the current cluster by default).
-d, --description DESCRIPTION
Filter out jobs by description (exact match).
--format COLUMNS
Output table format, see "neuro help top-format" for more info about the format specification. The default can be changed using the job.top-format configuration variable documented in "neuro help user-config"
--full-uri
Output full image URI.
-n, --name NAME
Filter out jobs by name.
-o, --owner TEXT
Filter out jobs by owner (multiple option). Supports ME option to filter by the current user. Specify ALL to show jobs of all users.
--since DATE_OR_TIMEDELTA
Show jobs created after a specific date (including). Use value of format '1d2h3m4s' to specify moment in past relatively to current time.
--sort COLUMNS
Sort rows by specified column. Add "-" prefix to revert the sorting order. Multiple columns can be specified (comma separated). [default: cpu]
-t, --tag TAG
Filter out jobs by tag (multiple option)
--timeout FLOAT
Maximum allowed time for executing the command, 0 for no timeout [default: 0]
--until DATE_OR_TIMEDELTA
Show jobs created before a specific date (including). Use value of format '1d2h3m4s' to specify moment in past relatively to current time.
Copy link
On this page
Usage
attach
browse
bump-life-span
exec
generate-run-command
kill
logs
ls
port-forward
run
save
status
top