product

Polyaxon v1.13: Easy hyperparameter tuning from CLI

Users can now use the cli to pass hyperparaeter directlu from the CI without creating a YAML file.

Starting hyperparameter tuning operations from the CLI

In Polyaxon v1.13, the CLI allows to create hyperparameter tuning operations without creating a YAML file. The run command has a new argument -HP/--hparams that allows to pass multiple values or a distribution to sample from for the input:

polyaxon run ... -HP param_name='range:0:10:1'

This is equivalent to the following YAML file:

matrix:
  kind: grid
  params:
    param_name:
      kind: range
      value: '0:10:1'
  concurrency: 1

Or:

polyaxon run ... -HP param_name='choice:[val1,val2,val3]'

This is equivalent to the following YAML file:

matrix:
  kind: grid
  params:
    param_name:
      kind: choice
      value: [val1, val2, val3]
  concurrency: 1

You should notice that the CLI uses the following format: kind:value to pass hyperparameters,

Starting hyperparameter tuning with different algorithms or concurency

In the previous example, an important aspect to notice is that matrix is of kind grid search and it runs the operations sequentially by setting the concurrency to 1.

Users can configure those options via CLI as well by passing the following extra arguments --matrix-kind, --matrix-concurrency, and --matrix-num-runs:

polyaxon run ... -p param_name1="some-value" -HP param_name2='range:0:10:1' -HP param_name3='uniform:0:1' --matrix-kind=random --matrix-concurrency=5 --matrix-num-runs=15

This is equivalent to the following YAML file:

params:
 param_name1: "some-value"
matrix:
  kind: random
  concurrency: 5
  numRuns: 15
  params:
    param_name2:
      kind: range
      value: [0,10,1]
    param_name2:
      kind: uniform
      value: [0,1]

Learn More about Polyaxon

This blog post just goes over a couple of features that we shipped in v1.13. To learn more about all the features, fixes, and enhancements, please visit the release notes and the short term roadmap.

Polyaxon continues to grow quickly and keeps improving and providing the simplest machine learning abstraction. We hope that these updates will improve your workflows and increase your productivity, and again, thank you for your continued feedback and support.

Subscribe to Polyaxon Blog

Get the latest posts delivered right to your inbox