HPO CLI arguments

HyperOptimizer injects optimized parameters as command-line arguments. The default prefix is --hpo-.

Naming

Use lowercase kebab-case slugs in the dashboard.

--hpo-learning-rate=0.003
--hpo-batch-size=64
--hpo-timeframe=5m

The hpo prefix keeps optimized parameters separate from your normal application flags.

Flag modes

Mode	Example	When to use
hpo (default)	--hpo-learning-rate=0.003	Recommended contract for new workloads
plain	--learning-rate=0.003	Existing entrypoints that already expect unprefixed flags

You can also place values with {{slug}} placeholders in a command template. See CLI parameters.

Parser behavior

Many parsers convert dashes to underscores in code.

parser.add_argument("--hpo-learning-rate", type=float)
args = parser.parse_args()

print(args.hpo_learning_rate)

Examples

Argument	Type	Description
--hpo-lookback-window	whole number (int)	Strategy or model window length.
--hpo-risk-multiplier	continuous (double)	Numeric multiplier used in scoring or sizing logic.
--hpo-timeframe	fixed choice (categorical)	Discrete option such as 1m, 5m, or 1h.
--hpo-use-filter	true/false	Feature flag represented as categorical true/false.

Note

Without a command template, the platform appends --hpo-<slug>=<value> (equals form) after any fixed args.