anycloud job

Run a finite container Job. Aliases: anycloud submit, anycloud run.

anycloud job [image] [-- command...] [options]

Use -- before a container command that has its own flags. Omitting the image in an interactive terminal opens the GHCR image picker. In CI or another non-interactive shell, pass the image as the first argument.

Options

Option	Description
-i, --id ID	Custom deployment ID
-e, --env KEY=VALUE	Environment value; repeatable. -e KEY reads the current shell
--env-file FILE	Load values from a dotenv file
--secret NAME	Inject a named Secret; repeatable
--cluster ID-OR-NAME	Run on a Ready Kubernetes cluster
--credentials NAME	Select one saved compute credential
--region REGION	Pin one compute region
--vm-type TYPE	Exact VM; repeat for an ordered fallback pool
--gpu-type TYPE	GPU model and optional count; repeat for an ordered fallback pool
--spot	Use spot/preemptible capacity
--disk-size GB	Root disk capacity
--disk-tier TIER	medium, high, or ultra
--zone ZONE	Availability zone
--input-bucket NAME	Read-only input at /mnt/input
--output-bucket NAME	Writable output at /mnt/output
--checkpoint-bucket NAME	Existing read-write checkpoint at /mnt/checkpoint
--input-storage-credentials NAME	Cross-account input identity
--input-storage-region REGION	Cross-account input Bucket region
--output-storage-credentials NAME	Cross-account output identity
--output-storage-region REGION	Cross-account output Bucket region
--checkpoint-storage-credentials NAME	Cross-account checkpoint identity
--checkpoint-storage-region REGION	Cross-account checkpoint Bucket region
--gpus VALUE	Docker GPU exposure: all or a count
--shm-size SIZE	Shared memory, such as 8g
--memory LIMIT	Container memory limit
--cpus COUNT	Container CPU limit
--runtime RUNTIME	Docker runtime
--ipc MODE	host, private, or shareable
--ulimit VALUE	Docker ulimit; repeatable
--bind BIND...	Host bind mounts

--gpu-type and --vm-type are mutually exclusive. --credentials may appear once. If omitted for VM-backed compute, the CLI selects the sole saved credential or shows a picker in an interactive terminal. Multiple credentials in non-interactive use require --credentials. Local compute is credentialless. Kubernetes Jobs instead require --cluster; cluster selection is mutually exclusive with compute credential, region, VM/GPU type, spot, disk, and zone placement.

An explicit --checkpoint-bucket works with spot or on-demand Jobs, must already exist, and is always retained. Without that flag, a spot Job uses its automatic deployment-named checkpoint bucket and deletes it when the Job finishes.

AWS, Azure, and GCP Jobs use the selected compute identity for automatic spot checkpoints and do not need explicit checkpoint storage flags. Lambda, Vast, Local, and Kubernetes jobs do not have a compatible native bucket identity, so a checkpoint bucket requires explicit checkpoint storage credentials and region. Scoped cross-identity storage currently requires AWS credentials.

Examples

anycloud job ghcr.io/acme/train:latest \
  --credentials my-aws \
  --gpu-type h100:8 \
  --spot \
  --gpus all \
  -- python train.py --epochs 10

anycloud job ghcr.io/acme/app:latest \
  --credentials my-aws \
  --vm-type t3.medium

anycloud job ghcr.io/acme/batch:latest \
  --cluster production-k8s \
  --shm-size 8g \
  --gpus 1

Kubernetes Jobs currently translate --gpus and --shm-size. CPU and memory limits are not yet supported for Kubernetes and fail at submission. See Clusters for cluster creation and placement.

GITHUB_TOKEN supplies GitHub authentication in non-interactive use. ANYCLOUD_CREDENTIALS_NAME is the credentials-name fallback when the flag is absent.

For field precedence and constraints, see Configuration.