Environment variables

You can configure your code to accept external parameters at runtime, overriding default any values in your code. These parameters, called Experiment Environment Variables, are automatically captured and tracked in Gradient.

Code sample:

train_epochs=int(os.environ.get('TRAIN_EPOCHS', 40)),
epochs_between_evals=int(os.environ.get('EPOCHS_EVAL', 100)),
batch_size=int(os.environ.get('BATCH_SIZE', 100)),

Adding parameters to Experiment:

The syntax for these parameters in the CLI is JSON.

gradient experiments ... --experimentEnv "{\"EPOCHS_EVAL\":5,\"TRAIN_EPOCHS\":2,\"MAX_STEPS\":5,\"EVAL_SECS\":10}"

These parameters will be injected into the Experiment at runtime and captured in Gradient.

Injecting Secrets:

Secrets are currently limited to experiments run on private clusters

Secrets can be injected into variables using the following syntax:

gradient experiments ... --experimentEnv '{"MY_SECRET":"secret:<secret_name>"}'