veri rewards + veri jobs

Training a model on Veri is a two-step CLI flow: register a reward function with veri rewards upload, then submit a job with veri jobs create (or veri run). Once submitted, manage the job through completion with list / get / cancel / download.

For a script-driven flow, you can also use veri train <file.py> — it reads a @training_job(...)-decorated function from the file, bundles the reward, resolves the dataset, and submits the job in one call. See veri run for the full decorator + TOML reference.

veri rewards does not have a server-side DELETE today. The CLI’s delete subcommand removes the reward from your local config, not from the control plane. Reward functions are content-addressed, so re-uploading the same file returns the same id.

Reward functions

Command	Purpose
`upload <path.py>`	Register a reward function. Auto-runs AST validation.
`list`	List registered reward functions.
`get <id>`	Show one reward function.
`delete <id>`	Delete a reward function.

Upload

veri rewards upload ./reward.py --name math-reward

upload runs an AST-only structural check before uploading:

File parses as valid Python.
The file defines at least one function.

The check deliberately stops there — verifying specific signatures (TRL vs DPO vs custom) requires actually importing user code, which the CLI refuses to do in its own process. If your reward function imports modules missing from the training runtime, the failure surfaces in the training logs once the job starts. Skip validation explicitly when needed:

veri rewards upload ./reward.py --skip-validation

Flag	Purpose
`--name`	Display name (default: filename stem).
`--format`	Reward format (`trl` by default).
`--skip-validation`	Don’t run the AST check.

List, get, delete

veri rewards list
veri rewards get rf_abc
veri rewards delete rf_abc

veri rewards test (run the reward against a dataset) is not in Phase-1. It depends on dataset row iteration, which is owed by the SDK first.

Training jobs

Command	Purpose
`create <config>` or inline	Submit a training job.
`list`	List jobs.
`get <id>`	Show one job.
`cancel <id>`	Cancel a running job.
`download <id>`	Download the completed checkpoint.

Submit from a config

veri jobs create configs/train.toml
# or, via the hero verb:
veri run configs/train.toml

The config must declare kind = "train". See veri run for the full schema. create returns the job ID. Pipe-friendly form:

JOB_ID=$(veri jobs create configs/train.toml --quiet)

Submit inline (no config file)

veri jobs create \
  --base-model "Qwen/Qwen3-4B" \
  --dataset ds_abc \
  --reward rf_abc \
  --gpu-type A100-80GB \
  --gpu-count 1

Inline form requires --base-model AND --dataset. Pass --reward for grpo; omit for sft_video_gen.

Override config values

veri jobs create configs/train.toml \
  --set model.base=Qwen/Qwen3-8B \
  --set method.learning_rate=2e-6 \
  --set resources.gpu_count=4

The convenience flags --base-model, --gpu-type, --gpu-count, --dataset, --reward expand to --set internally. See Override syntax for type coercion and conflict detection.

List, get, cancel

veri jobs list                              # most recent 20
veri jobs list --status running             # filter
veri jobs list --method grpo
veri jobs get job_abc
veri jobs cancel job_abc

list supports --limit, --after, --status, --method. All commands accept --format json.

Watch progress

There is no --follow log tailer in Phase-1 (deferred until the SDK exposes streaming). For now:

# Poll status in a script:
while [ "$(veri jobs get $JOB_ID --format json | jq -r .status)" = "running" ]; do
  sleep 30
done

# Or open the dashboard for live logs:
veri jobs get $JOB_ID --format json | jq -r .dashboard_url | xargs open

Download a completed checkpoint

veri jobs download job_abc --output-dir ./checkpoints

The download URL is only populated after the job reaches completed. Trying to download an in-flight job exits 1 with a clear message.

Exit codes

Code	Meaning
`0`	Success.
`1`	User error (missing flag, bad config, job not downloadable).
`2`	API error.
`4`	Auth error.

What’s next

Deploy this model

Serve the checkpoint with an OpenAI-compatible endpoint.

Evaluate it

Score it on a held-out dataset.

The run verb

Config schemas and --set overrides.

Get started

Training

Deployments

Evaluations

CLI

Volumes

Reward functions

Upload

List, get, delete

Training jobs

Submit from a config

Submit inline (no config file)

Override config values

List, get, cancel

Watch progress

Download a completed checkpoint

Exit codes

What’s next

Deploy this model

Evaluate it

The run verb

​Reward functions

​Upload

​List, get, delete

​Training jobs

​Submit from a config

​Submit inline (no config file)

​Override config values

​List, get, cancel

​Watch progress

​Download a completed checkpoint

​Exit codes

​What’s next

Deploy this model

Evaluate it

The run verb

Reward functions

Upload

List, get, delete

Training jobs

Submit from a config

Submit inline (no config file)

Override config values

List, get, cancel

Watch progress

Download a completed checkpoint

Exit codes

What’s next