baseten train - Baseten

Package training code with baseten train init and baseten train push, run it as a job, and deploy the checkpoints the job syncs. Projects group jobs and share a cache.

init

baseten train init [OPTIONS]

Create a training project directory, either empty from the built-in template or from one of the published training examples. This command needs no credentials: it only writes files, and the examples are downloaded from a public repository.

Options

TEXT

Directory to create. Defaults to ./truss-train-init for an empty project, or the current directory when examples are named.

TEXT (repeatable)

Name of a published training example to download instead of the empty template. Repeatable.

TEXT

Filter JSON output with a jq expression; implies --output json (or jsonl for streamed commands)

BOOL

List the available example names and exit.

TEXT

default:"text"

Output formatOne of: text, json, jsonl, none

TEXT

Use a specific stored profile for this command, overriding BASETEN_PROFILE and the current profile

TEXT

Run this truss executable instead of fetching one with uv, e.g. a virtualenv’s bin/truss. A value with no path separator is looked up on PATH, so ‘--truss-executable truss’ runs the truss you installed. Defaults to BASETEN_TRUSS_EXECUTABLE.

TEXT

Version of truss to fetch and run with ‘uv tool run’, e.g. 0.18.26. Defaults to BASETEN_TRUSS_VERSION, or the latest release. Mutually exclusive with --truss-executable.

BOOL

Enable verbose logging

Examples

Scaffold an empty training project

baseten train init --dir ./my-training

List the available examples

baseten train init --list-examples

Download two examples into the current directory

baseten train init --example sft-lora --example grpo

Output

Text mode (--output text): A line per directory created, naming where it was written. With --list-examples, the available example names, one per line. JSON mode (--output json): payload type cmd.TrussDelegatedResult. Under --output json the text output goes to stderr and stdout is an empty object: this command reports nothing structured yet.

push

baseten train push [OPTIONS]

Create and start a training job, uploading the config file’s directory as the job’s code. The config is a Python file that defines the training project and its job, so it is executed rather than read. The flags below override what it declares. Returns once the job is created. Follow it with ‘baseten train job logs --job-id id --tail’.

Options

TEXT

Accelerator type and count, for example ‘H200:8’. Overrides the config. Supported hardware is documented at https://docs.baseten.co/training/concepts/basics.

TEXT

required

Python file defining the training project and its job.

TEXT

Command the job runs. Overrides the config.

TEXT

Start an interactive session on this trigger.One of: on-startup, on-failure, on-demand

TEXT

How long an interactive session stays up, as a duration (e.g. ‘90m’, ‘2h’). Rounded down to whole minutes; must be at least 1m.

TEXT

Name for the training job. Defaults to a generated name.

TEXT

Filter JSON output with a jq expression; implies --output json (or jsonl for streamed commands)

INTEGER

Number of compute nodes. Overrides the config.

TEXT

default:"text"

Output formatOne of: text, json, jsonl, none

INTEGER

Queue priority. Higher values are dequeued first.

TEXT

Use a specific stored profile for this command, overriding BASETEN_PROFILE and the current profile

BOOL

Run on interruptible spot capacity, overriding the config. Checkpointing your own progress is up to you.

TEXT

Team name or ID that owns the training project. Run ‘baseten org team list’ to see teams.

TEXT

BOOL

Do not forward this CLI’s credentials to truss, leaving it to resolve a remote from your trussrc. Old truss versions ignore forwarded credentials and use the trussrc regardless.

TEXT

Version of truss to fetch and run with ‘uv tool run’, e.g. 0.18.26. Defaults to BASETEN_TRUSS_VERSION, or the latest release. Mutually exclusive with --truss-executable.

BOOL

Enable verbose logging

Examples

Run a training job from a config file

baseten train push --config ./train.py

Run a named job on eight H200s over spot capacity

baseten train push --config ./train.py --job-name sweep-3 --accelerator H200:8 --spot

Run a job that opens an interactive session if it fails

baseten train push --config ./train.py --interactive on-failure --interactive-timeout 2h

Output

Text mode (--output text): A confirmation that the job was created, with its ID, the commands to follow it, and a link to it in the Baseten UI. JSON mode (--output json): payload type cmd.TrussDelegatedResult. Under --output json the text output goes to stderr and stdout is an empty object: this command reports nothing structured yet.

job

Submit training jobs, then read their logs and metrics, change their priority, and stop or recreate them.

list

baseten train job list [OPTIONS]

List training jobs, newest first. Lists every job in projects your teams can access. Pass --project to narrow to one project, or --status to narrow to particular job states.

Options

TEXT

default:"desc"

Sort order by creation time: ‘desc’ (newest first) or ‘asc’ (oldest first).One of: asc, desc

TEXT

Filter JSON output with a jq expression; implies --output json (or jsonl for streamed commands)

TEXT

default:"text"

Output formatOne of: text, json, jsonl, none

TEXT

Use a specific stored profile for this command, overriding BASETEN_PROFILE and the current profile

TEXT

Only list jobs in this training project, by name or ID.

TEXT (repeatable)

Only list jobs in these states: pending, created, deploying, deploy-failed, running, completed, failed, stopped, preempted. Repeatable.

BOOL

Enable verbose logging

Examples

List every training job you can see

baseten train job list

List the running jobs of one project

baseten train job list --project my-project --status running

List jobs oldest first

baseten train job list --direction asc

Filter output with `--jq`

Print just the job IDs

baseten train job list --jq '.training_jobs[].id'

Output

Text mode (--output text): Table with columns: ID, PROJECT, NAME, STATUS, INSTANCE TYPE, NODES, CREATED. When no jobs match, prints “No training jobs found.” to stderr. JSON mode (--output json): payload type managementapi.SearchTrainingJobsResponse.

describe

baseten train job describe [OPTIONS]

Describe a training job by ID, including its project, compute, and checkpoint sync state.

Options

TEXT

required

ID of the training job.

TEXT

Filter JSON output with a jq expression; implies --output json (or jsonl for streamed commands)

TEXT

default:"text"

Output formatOne of: text, json, jsonl, none

TEXT

Use a specific stored profile for this command, overriding BASETEN_PROFILE and the current profile

BOOL

Enable verbose logging

Examples

Describe a job

baseten train job describe --job-id p7qr9qv

Filter output with `--jq`

Print the job’s current status

baseten train job describe --job-id p7qr9qv --jq '.training_job.current_status'

Output

Text mode (--output text): A field-per-line summary of the job and its project. JSON mode (--output json): payload type managementapi.GetTrainingJobResponse.

logs

baseten train job logs [OPTIONS]

Fetch logs for a training job. Without --tail, returns the logs in a time window; with --tail, streams new logs until the job stops producing them or you interrupt with Ctrl-C.

Options

TEXT

End of the log time range. Accepts ISO 8601; values without a timezone designator are interpreted in the local timezone. Default is now. Window must be at most 7 days.

TEXT

required

ID of the training job.

TEXT

Filter JSON output with a jq expression; implies --output json (or jsonl for streamed commands)

INTEGER

default:"5000"

Maximum number of log lines to return, paging backward from the end of the window. Use 0 for no limit (every log line in the window). Not applicable with --tail.

TEXT

Only return logs at or above this severity level.One of: debug, info, warning, error

TEXT

default:"text"

Output formatOne of: text, json, jsonl, none

TEXT

Use a specific stored profile for this command, overriding BASETEN_PROFILE and the current profile

TEXT

Shortcut for fetching logs from a relative time ago until now. Accepts a duration (e.g. ‘30m’, ‘1h30m’) or ‘Nd’ (e.g. ‘3d’). Maximum ‘7d’. Mutually exclusive with --start and --end.

TEXT

Start of the log time range. Accepts ISO 8601 (e.g. ‘2026-05-14’, ‘2026-05-14T12:00:00’, ‘2026-05-14T12:00:00Z’). Values without a timezone designator are interpreted in the local timezone. Default is 30 minutes before the end. Window must be at most 7 days.

BOOL

Stream new logs as they arrive until the job stops producing them or you interrupt with Ctrl-C. Cannot be combined with the time-range or filter flags. For machine-readable streaming, prefer --output jsonl over --output json.

BOOL

Enable verbose logging

Examples

Fetch the last 30 minutes of logs

baseten train job logs --job-id p7qr9qv

Stream logs as they arrive

baseten train job logs --job-id p7qr9qv --tail

Fetch a day of error logs

baseten train job logs --job-id p7qr9qv --since 1d --min-level error

Filter output with `--jq`

Print just the log messages

baseten train job logs --job-id p7qr9qv --jq '.message'

Output

Text mode (--output text): One log line per row, timestamped. With --output jsonl, one JSON log record per line. JSON mode (--output json): payload type managementapi.GetLogsResponse.

metrics

baseten train job metrics [OPTIONS]

Report CPU, memory, GPU, and storage metrics for a training job. Each series is reported as its most recent sample. Pass --start and --end, or --since, to select the window the samples come from.

Options

TEXT

End of the sample window. Accepts ISO 8601; values without a timezone designator are interpreted in the local timezone. Default is now.

TEXT

required

ID of the training job.

TEXT

Filter JSON output with a jq expression; implies --output json (or jsonl for streamed commands)

TEXT

default:"text"

Output formatOne of: text, json, jsonl, none

TEXT

Use a specific stored profile for this command, overriding BASETEN_PROFILE and the current profile

TEXT

Shortcut for sampling from a relative time ago until now. Accepts a duration (e.g. ‘30m’, ‘1h30m’) or ‘Nd’ (e.g. ‘3d’). Mutually exclusive with --start and --end.

TEXT

Start of the sample window. Accepts ISO 8601; values without a timezone designator are interpreted in the local timezone.

BOOL

Enable verbose logging

Examples

Report a job’s latest metrics

baseten train job metrics --job-id p7qr9qv

Report metrics sampled over the last hour

baseten train job metrics --job-id p7qr9qv --since 1h

Filter output with `--jq`

Print the GPU utilization series

baseten train job metrics --job-id p7qr9qv --jq '.gpu_utilization'

Output

Text mode (--output text): Table with columns: METRIC, NODE, VALUE, MEASURED. JSON mode (--output json): payload type managementapi.GetTrainingJobMetricsResponse.

stop

baseten train job stop [OPTIONS]

Stop a running or queued training job. Stopping is not reversible: use ‘baseten train job recreate’ to run the same configuration again. Checkpoints already synced remain accessible.

Options

TEXT

required

ID of the training job.

TEXT

Filter JSON output with a jq expression; implies --output json (or jsonl for streamed commands)

TEXT

default:"text"

Output formatOne of: text, json, jsonl, none

TEXT

Use a specific stored profile for this command, overriding BASETEN_PROFILE and the current profile

BOOL

Skip the interactive confirmation prompt. Required when stdin is not a terminal.

BOOL

Enable verbose logging

Examples

Stop a job, confirming interactively

baseten train job stop --job-id p7qr9qv

Stop a job without confirmation, for scripts

baseten train job stop --job-id p7qr9qv --yes

Filter output with `--jq`

Print the stopped job’s status

baseten train job stop --job-id p7qr9qv --yes --jq '.training_job.current_status'

Output

Text mode (--output text): On success, prints the stopped job’s ID to stderr; no stdout output. JSON mode (--output json): payload type managementapi.StopTrainingJobResponse.

recreate

baseten train job recreate [OPTIONS]

Create a new training job from an existing job’s configuration. The original job is left as it is. The new job gets a new ID and starts from the beginning.

Options

TEXT

required

ID of the training job.

TEXT

Filter JSON output with a jq expression; implies --output json (or jsonl for streamed commands)

TEXT

default:"text"

Output formatOne of: text, json, jsonl, none

TEXT

Use a specific stored profile for this command, overriding BASETEN_PROFILE and the current profile

BOOL

Enable verbose logging

Examples

Rerun a job’s configuration

baseten train job recreate --job-id p7qr9qv

Filter output with `--jq`

Print just the new job’s ID

baseten train job recreate --job-id p7qr9qv --jq '.training_job.id'

Output

Text mode (--output text): On success, prints the new job’s ID to stderr; no stdout output. JSON mode (--output json): payload type managementapi.RecreateTrainingJobResponse.

update

baseten train job update [OPTIONS]

Update a queued training job’s priority. Higher values are dequeued first. Only jobs in the PENDING state can have their priority changed.

Options

TEXT

New capacity guarantee. ‘dedicated’ runs on on-demand capacity that is not preempted; ‘spot’ runs on interruptible capacity that may be preempted, and checkpointing your own progress is up to you.One of: dedicated, spot

TEXT

required

ID of the training job.

TEXT

Filter JSON output with a jq expression; implies --output json (or jsonl for streamed commands)

TEXT

default:"text"

Output formatOne of: text, json, jsonl, none

INTEGER

New queue priority. Higher values are dequeued first.

TEXT

Use a specific stored profile for this command, overriding BASETEN_PROFILE and the current profile

BOOL

Enable verbose logging

Examples

Raise a queued job’s priority

baseten train job update --job-id p7qr9qv --priority 10

Filter output with `--jq`

Print the job’s new priority

baseten train job update --job-id p7qr9qv --priority 10 --jq '.training_job.priority'

Output

Text mode (--output text): On success, prints the job’s ID and new priority to stderr; no stdout output. JSON mode (--output json): payload type managementapi.UpdateTrainingJobResponse.

download

baseten train job download [OPTIONS]

Download the code archive that was uploaded with a training job as an uncompressed tar. Exactly one of --out-file or --out-dir is required. --out-file writes the raw tar bytes; --out-dir extracts the tar into the directory. Use --overwrite to replace an existing file or write into a non-empty directory. This is the job’s input code, not its checkpoints; use ‘baseten train checkpoint files’ for those.

Options

TEXT

required

ID of the training job.

TEXT

Filter JSON output with a jq expression; implies --output json (or jsonl for streamed commands)

TEXT

Extract the artifact tar into this directory.Mutually exclusive with other flags in group download-out.

TEXT

Save the artifact as an uncompressed tar file at this path.Mutually exclusive with other flags in group download-out.

TEXT

default:"text"

Output formatOne of: text, json, jsonl, none

BOOL

Allow overwriting an existing file or non-empty directory.

TEXT

Use a specific stored profile for this command, overriding BASETEN_PROFILE and the current profile

BOOL

Enable verbose logging

Examples

Extract a job’s code into a directory

baseten train job download --job-id p7qr9qv --out-dir ./job-code

Save the archive as a tar file instead of extracting it

baseten train job download --job-id p7qr9qv --out-file job.tar

Filter output with `--jq`

Print just the destination path

baseten train job download --job-id p7qr9qv --out-file job.tar --jq '.out_file'

Output

Text mode (--output text): Writes the artifact to disk; prints progress and the final destination path to stderr; no stdout output. JSON mode (--output json): payload type cmd.TrainJobDownloadResult. On success, stdout is a JSON object naming the job, with either out_file or out_dir set to the path written.

job session

A running job can hold an interactive session on each replica. Describe those sessions.

describe

baseten train job session describe [OPTIONS]

Describe the interactive sessions of a training job, one per replica. Only replicas with a live session are listed, so a job whose session has not started yet reports none.

Options

TEXT

required

ID of the training job.

TEXT

Filter JSON output with a jq expression; implies --output json (or jsonl for streamed commands)

TEXT

default:"text"

Output formatOne of: text, json, jsonl, none

TEXT

Use a specific stored profile for this command, overriding BASETEN_PROFILE and the current profile

BOOL

Enable verbose logging

Examples

Show a job’s interactive sessions

baseten train job session describe --job-id p7qr9qv

Filter output with `--jq`

Print the first replica’s auth URL

baseten train job session describe --job-id p7qr9qv --jq '.auth_codes[0].auth_url'

Output

Text mode (--output text): Table with columns: REPLICA, SESSION, AUTH URL, EXPIRES. When no session is live, prints “No interactive sessions found.” to stderr. JSON mode (--output json): payload type managementapi.GetAuthCodesResponse.

capacity

Capacity is the number of GPUs of each type a team may use concurrently. Read the limits and raise them per team.

describe

baseten train capacity describe [OPTIONS]

Describe the organization’s training GPU limits and current usage, per GPU type. Per-team limits are listed below the organization totals when any are set.

Options

TEXT

Filter JSON output with a jq expression; implies --output json (or jsonl for streamed commands)

TEXT

default:"text"

Output formatOne of: text, json, jsonl, none

TEXT

Use a specific stored profile for this command, overriding BASETEN_PROFILE and the current profile

BOOL

Enable verbose logging

Examples

Describe training GPU capacity

baseten train capacity describe

Filter output with `--jq`

Print the organization’s H100 limit

baseten train capacity describe --jq '.gpu_capacities[] | select(.gpu_type == "H100") | .limit'

Output

Text mode (--output text): An organization table with columns: GPU TYPE, IN USE, LIMIT, BASELINE, then a per-team table with a leading TEAM column when teams have limits set. JSON mode (--output json): payload type managementapi.GetTrainingGpuCapacityResponse.

update

baseten train capacity update [OPTIONS]

Update the maximum concurrent GPUs of one type that a team may use. The limit is per GPU type, so raising H100 leaves other types alone.

Options

TEXT

required

GPU type the limit applies to (for example H100).

TEXT

Filter JSON output with a jq expression; implies --output json (or jsonl for streamed commands)

INTEGER

required

Maximum concurrent GPUs of this type the team may use.

TEXT

default:"text"

Output formatOne of: text, json, jsonl, none

TEXT

Use a specific stored profile for this command, overriding BASETEN_PROFILE and the current profile

TEXT

required

Team name or ID whose limit changes. Run ‘baseten org team list’ to see teams.

BOOL

Enable verbose logging

Examples

Raise a team’s H100 limit

baseten train capacity update --team research --gpu-type H100 --max-gpus 32

Filter output with `--jq`

Print the applied limit

baseten train capacity update --team research --gpu-type H100 --max-gpus 32 --jq '.team_gpu_capacity.limit'

Output

Text mode (--output text): On success, prints the team, GPU type, and new limit to stderr; no stdout output. JSON mode (--output json): payload type managementapi.PatchTeamTrainingGpuCapacityResponse.

checkpoint

Jobs that enable checkpointing sync their checkpoints. List them, download their files, and deploy LoRA checkpoints as a model served by vLLM.

list

baseten train checkpoint list [OPTIONS]

List the checkpoints a training job has synced, newest first. Checkpoints only appear for jobs that enabled checkpointing.

Options

TEXT

default:"desc"

Sort order by creation time: ‘desc’ (newest first) or ‘asc’ (oldest first).One of: asc, desc

TEXT

required

ID of the training job.

TEXT

Filter JSON output with a jq expression; implies --output json (or jsonl for streamed commands)

TEXT

default:"text"

Output formatOne of: text, json, jsonl, none

TEXT

Use a specific stored profile for this command, overriding BASETEN_PROFILE and the current profile

BOOL

Enable verbose logging

Examples

List a job’s checkpoints

baseten train checkpoint list --job-id p7qr9qv

List a job’s checkpoints oldest first

baseten train checkpoint list --job-id p7qr9qv --direction asc

Filter output with `--jq`

Print just the checkpoint IDs

baseten train checkpoint list --job-id p7qr9qv --jq '.checkpoints[].checkpoint_id'

Output

Text mode (--output text): Table with columns: ID, TYPE, BASE MODEL, SIZE, SYNC, CREATED. When the job has no checkpoints, prints “No checkpoints found.” to stderr. JSON mode (--output json): payload type managementapi.GetTrainingJobCheckpointsResponse.

files

baseten train checkpoint files [OPTIONS]

List presigned download URLs for the files of a training job’s checkpoints. The URLs are short-lived. For machine-readable streaming, prefer --output jsonl over --output json.

Options

TEXT

required

ID of the training job.

TEXT

Filter JSON output with a jq expression; implies --output json (or jsonl for streamed commands)

TEXT

default:"text"

Output formatOne of: text, json, jsonl, none

TEXT

Use a specific stored profile for this command, overriding BASETEN_PROFILE and the current profile

BOOL

Enable verbose logging

Examples

List every checkpoint file of a job

baseten train checkpoint files --job-id p7qr9qv

Save the URLs to a file for a download script

baseten train checkpoint files --job-id p7qr9qv --output json > checkpoint-urls.json

Filter output with `--jq`

Print just the download URLs

baseten train checkpoint files --job-id p7qr9qv --output jsonl --jq '.url'

Output

Text mode (--output text): Table with columns: NODE, NAME, SIZE, MODIFIED, URL. When there are no files, prints “No checkpoint files found.” to stderr. JSON mode (--output json): payload type managementapi.CheckpointFile.

deploy

baseten train checkpoint deploy [OPTIONS]

Deploy a training job’s LoRA checkpoints as a model served by vLLM. Without --config, the checkpoints and the model that serves them are chosen interactively. Pass --config for a repeatable deploy: it is a Python file defining a DeployCheckpointsConfig, so it is executed rather than read. The generated model config is written to disk either way, so a deploy can be re-run or edited by hand afterward.

Options

TEXT

Python file defining a DeployCheckpointsConfig: which checkpoints deploy and the model that serves them. The type is documented at https://docs.baseten.co/reference/sdk/training#deploycheckpointsconfig.

TEXT

Directory the generated model config is written to. Defaults to a directory under ./truss_configs.

BOOL

Write the generated model config without deploying anything.

TEXT

Training job whose checkpoints are deployed. Required unless --config names them.

TEXT

Filter JSON output with a jq expression; implies --output json (or jsonl for streamed commands)

TEXT

default:"text"

Output formatOne of: text, json, jsonl, none

TEXT

Use a specific stored profile for this command, overriding BASETEN_PROFILE and the current profile

TEXT

BOOL

Do not forward this CLI’s credentials to truss, leaving it to resolve a remote from your trussrc. Old truss versions ignore forwarded credentials and use the trussrc regardless.

TEXT

Version of truss to fetch and run with ‘uv tool run’, e.g. 0.18.26. Defaults to BASETEN_TRUSS_VERSION, or the latest release. Mutually exclusive with --truss-executable.

BOOL

Enable verbose logging

Examples

Deploy a job’s checkpoints, choosing them interactively

baseten train checkpoint deploy --job-id p7qr9qv

Deploy the checkpoints a config file names

baseten train checkpoint deploy --config ./deploy_checkpoints.py

Generate the model config without deploying

baseten train checkpoint deploy --job-id p7qr9qv --dry-run --config-out-dir ./generated

Output

Text mode (--output text): The new model and deployment IDs with links to the deployment’s logs, the path the model config was written to, and an example request naming the deployed checkpoints. With --dry-run, only the config path. JSON mode (--output json): payload type cmd.TrussDelegatedResult. Under --output json the text output goes to stderr and stdout is an empty object: this command reports nothing structured yet.

project

Projects group training jobs and share a cache.

list

baseten train project list [OPTIONS]

List the training projects your teams can access, newest first.

Options

TEXT

Filter JSON output with a jq expression; implies --output json (or jsonl for streamed commands)

TEXT

default:"text"

Output formatOne of: text, json, jsonl, none

TEXT

Use a specific stored profile for this command, overriding BASETEN_PROFILE and the current profile

BOOL

Enable verbose logging

Examples

List your training projects

baseten train project list

Filter output with `--jq`

Print just the project names

baseten train project list --jq '.training_projects[].name'

Output

Text mode (--output text): Table with columns: ID, NAME, TEAM, LATEST JOB, LATEST STATUS, CREATED. When there are no projects, prints “No training projects found.” to stderr. JSON mode (--output json): payload type managementapi.ListTrainingProjectsResponse.

project cache

describe

baseten train project cache describe [OPTIONS]

Describe the files in a training project’s shared cache, largest first. The cache is shared by every job in the project and persists between runs.

Options

TEXT

Filter JSON output with a jq expression; implies --output json (or jsonl for streamed commands)

TEXT

default:"text"

Output formatOne of: text, json, jsonl, none

TEXT

Use a specific stored profile for this command, overriding BASETEN_PROFILE and the current profile

TEXT

required

Training project to describe, by name or ID.

TEXT

default:"size"

Sort the listing by ‘size’ (largest first) or ‘path’ (alphabetical).One of: size, path

BOOL

Enable verbose logging

Examples

Describe a project’s cache

baseten train project cache describe --project my-project

Sort the cache listing by path

baseten train project cache describe --project my-project --sort path

Filter output with `--jq`

Print the cached file paths

baseten train project cache describe --project my-project --jq '.file_summaries[].path'

Output

Text mode (--output text): A total size line, then a table with columns: PATH, TYPE, SIZE, PERMISSIONS, MODIFIED. When the cache is empty, prints “Cache is empty.” to stderr. JSON mode (--output json): payload type managementapi.GetCacheSummaryResponse.

workstation

A workstation is a training job that holds GPUs for an interactive SSH session instead of running a script.

create

baseten train workstation create [OPTIONS]

Create an SSH workstation: a training job that holds GPUs for an interactive session instead of running a training script. Pass --node-count for a multi-node workstation, where each node has eight GPUs and SLURM is set up across them. Connect once the job is running.

Options

TEXT

default:"H100"

GPU accelerator type.

TEXT

Training job whose latest checkpoint the workstation starts from.

TEXT

Path inside the container where checkpoints are written.

INTEGER

Size of the checkpoint volume in GiB.

BOOL

Attach checkpoint storage to the workstation.

INTEGER

Number of GPUs on a single node, 1 to 8. Mutually exclusive with --node-count.

TEXT

Docker base image the workstation runs.

TEXT

Filter JSON output with a jq expression; implies --output json (or jsonl for streamed commands)

INTEGER

Number of eight-GPU nodes, 1 to 16. Mutually exclusive with --gpu-count.

TEXT

default:"slurm"

Multi-node orchestrator set up across the nodes.One of: slurm

TEXT

default:"text"

Output formatOne of: text, json, jsonl, none

TEXT

Use a specific stored profile for this command, overriding BASETEN_PROFILE and the current profile

TEXT

Training project that owns the workstation. Defaults to ‘workstation-accelerator’.

BOOL

Run on interruptible spot capacity.

TEXT

Team name or ID that owns the training project. Run ‘baseten org team list’ to see teams.

TEXT

BOOL

Do not forward this CLI’s credentials to truss, leaving it to resolve a remote from your trussrc. Old truss versions ignore forwarded credentials and use the trussrc regardless.

TEXT

Version of truss to fetch and run with ‘uv tool run’, e.g. 0.18.26. Defaults to BASETEN_TRUSS_VERSION, or the latest release. Mutually exclusive with --truss-executable.

BOOL

Enable verbose logging

Examples

Create a single-GPU workstation

baseten train workstation create

Create a four-node workstation with checkpoint storage

baseten train workstation create --node-count 4 --enable-checkpointing --checkpoint-volume-size 512

Output

Text mode (--output text): A confirmation that the workstation was created, with the ssh host of each node and the commands to view its logs or stop it. JSON mode (--output json): payload type cmd.TrussDelegatedResult. Under --output json the text output goes to stderr and stdout is an empty object: this command reports nothing structured yet.

​init

​Options

​Examples

​Output

​push

​Options

​Examples

​Output

​job

​list

​Options

​Examples

​Filter output with --jq

​Output

​describe

​Options

​Examples

​Filter output with --jq

​Output

​logs

​Options

​Examples

​Filter output with --jq

​Output

​metrics

​Options

​Examples

​Filter output with --jq

​Output

​stop

​Options

​Examples

​Filter output with --jq

​Output

​recreate

​Options

​Examples

​Filter output with --jq

​Output

​update

​Options

​Examples

​Filter output with --jq

​Output

​download

​Options

​Examples

​Filter output with --jq

​Output

​job session

​describe

​Options

​Examples

​Filter output with --jq

​Output

​capacity

​describe

​Options

​Examples

​Filter output with --jq

​Output

​update

​Options

​Examples

​Filter output with --jq

​Output

​checkpoint

​list

​Options

​Examples

​Filter output with --jq

​Output

​files

​Options

​Examples

​Filter output with --jq

​Output

​deploy

​Options

​Examples

init

Options

Examples

Output

push

Options

Examples

Output

job

list

Options

Examples

Filter output with `--jq`

Output

describe

Options

Examples

Filter output with `--jq`

Output

logs

Options

Examples

Filter output with `--jq`

Output

metrics

Options

Examples

Filter output with `--jq`

Output

stop

Options

Examples

Filter output with `--jq`

Output

recreate

Options

Examples

Filter output with `--jq`

Output

update

Options

Examples

Filter output with `--jq`

Output

download

Options

Examples

Filter output with `--jq`

Output

job session

describe

Options

Examples

Filter output with `--jq`

Output

capacity

describe

Options

Examples

Filter output with `--jq`

Output

update

Options

Examples

Filter output with `--jq`

Output

checkpoint

list

Options

Examples

Filter output with `--jq`

Output

files

Options

Examples

Filter output with `--jq`

Output

deploy

Options

Examples