Skip to main content
Manage your ML training jobs with the tp job command group.

​
tp job init

Create a new job configuration file. 
tp job init
Creates a {name}.tp.toml file in the current working directory Example: 
$ tp job init
Enter a name for your job:
> experiment1
experiment1.tp.toml created
Configure it, then run `tp job push experiment1.tp.toml c-xxx`

​
tp job push

Submit a job to run on TensorPool. 
tp job push <tp_config_path> <cluster_id> [--teardown]
Arguments:
  • tp_config_path - Path to your {name}.tp.toml configuration file
  • cluster_id - The ID of the cluster to run the job on (e.g., c-xxx)
Options:
  • --teardown - Automatically destroy the cluster after the job completes or is canceled
Requirements:
  • Valid TOML configuration file with commands, outputs, and ignore fields
  • An existing cluster in RUNNING state
Example: 
tp job push train.tp.toml c-abc123

# Auto-destroy cluster when the job finishes
tp job push train.tp.toml c-abc123 --teardown

​
tp job list

List all your jobs. 
tp job list [--org]
Aliases: tp job ls Options:
  • --org, --organization - List all jobs in your organization (not just yours)
Example: 
tp job list
tp job list --org

​
tp job info

Get detailed information about a specific job. 
tp job info <job_id>
Arguments:
  • job_id - The ID of the job to inspect
Example: 
tp job info job_abc123

​
tp job listen

Stream real-time logs from a running job. 
tp job listen <job_id>
Arguments:
  • job_id - The ID of the job to monitor
Behavior:
  • Continuously streams stdout/stderr from the job
  • Updates in real-time as the job runs
  • Press Ctrl+C to stop listening (job continues running)
Example: 
tp job listen job_abc123

​
tp job pull

Download output files from a completed job. 
tp job pull <job_id> [-i <private_key>] [--force]
Arguments:
  • job_id - The ID of the job to pull outputs from
Options:
  • -i, --private-key - Path to an SSH private key. Not required if SSH keys are already saved on your account (see tp me sshkey)
  • --force - Overwrite existing local files without prompting
Behavior:
  • Downloads files specified in the outputs configuration
  • By default, skips files that already exist locally
Example: 
# Download outputs (skip existing files)
tp job pull job_abc123

# Force overwrite existing files
tp job pull job_abc123 --force

# Specify a private key explicitly
tp job pull job_abc123 -i ~/.ssh/my_key

​
tp job cancel

Cancel a running or queued job. 
tp job cancel <job_id> [--wait] [--no-input]
Arguments:
  • job_id - The ID of the job to cancel
Options:
  • --wait - Wait for the job to finish canceling before returning
  • --no-input - Skip confirmation prompt and cancel immediately
Behavior:
  • Prompts for confirmation before canceling (unless --no-input is used)
  • By default, the CLI returns immediately while cancellation runs in the background. Use --wait to block until the job is fully canceled.
  • The cluster is preserved after cancellation unless --teardown was set during tp job push
Example: 
# Cancel with confirmation
tp job cancel job_abc123

# Cancel without confirmation and wait for it to complete
tp job cancel job_abc123 --no-input --wait

​
tp job delete

Delete a terminal (completed, canceled, or failed) job. 
tp job delete <job_id> [--no-input]
Arguments:
  • job_id - The ID of the job to delete
Options:
  • --no-input - Skip confirmation prompt and delete immediately
Behavior:
  • Only jobs in a terminal state (completed, canceled, or failed) can be deleted
  • Prompts for confirmation before deleting (unless --no-input is used)
Example: 
# Delete with confirmation
tp job delete job_abc123

# Delete without confirmation
tp job delete job_abc123 --no-input
Some job operations use SSH authentication. Manage your SSH keys with:

​
Next Steps