Skip to content

Instantly share code, notes, and snippets.

@ravirahman
Last active May 12, 2019 17:55
Show Gist options
  • Star 0 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save ravirahman/076427e90b4fab8e384b244da74df65c to your computer and use it in GitHub Desktop.
Save ravirahman/076427e90b4fab8e384b244da74df65c to your computer and use it in GitHub Desktop.
This JSON object describes the hyperparameter search space, in the format required by Google Cloud Hypertune (https://cloud.google.com/ml-engine/reference/rest/v1/projects.jobs#HyperparameterSpec)
{
"goal": "MINIMIZE",
"params": [
{
"parameterName": "kernel-size-mel",
"minValue": 2,
"maxValue": 8,
"type": "INTEGER"
},
{
"parameterName": "kernel-size-time",
"minValue": 2,
"maxValue": 10,
"type": "INTEGER"
},
{
"parameterName": "pool-size-time",
"minValue": 1,
"maxValue": 10,
"type": "INTEGER"
},
{
"parameterName": "num-layers",
"minValue": 4,
"maxValue": 8,
"type": "INTEGER"
},
{
"parameterName": "num-units",
"type": "DISCRETE",
"discreteValues": [
256,
512
]
},
{
"parameterName": "momentum",
"minValue": 0.8,
"maxValue": 0.99999,
"type": "DOUBLE",
"scaleType": "UNIT_LOG_SCALE"
},
{
"parameterName": "lr",
"minValue": 0.000001,
"maxValue": 0.001,
"type": "DOUBLE",
"scaleType": "UNIT_LOG_SCALE"
},
{
"parameterName": "dropout-rate",
"minValue": 0.1,
"maxValue": 0.5,
"type": "DOUBLE",
"scaleType": "UNIT_LINEAR_SCALE"
},
{
"parameterName": "decay",
"minValue": 0.00001,
"maxValue": 0.1,
"type": "DOUBLE",
"scaleType": "UNIT_LOG_SCALE"
},
{
"parameterName": "l2-regularization",
"minValue": 0.0001,
"maxValue": 0.1,
"type": "DOUBLE",
"scaleType": "UNIT_LOG_SCALE"
}
],
"maxTrials": 50,
"maxParallelTrials": 3,
"hyperparameterMetricTag": "val_binary_crossentropy",
"enableTrialEarlyStopping": true
}
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment