-
Notifications
You must be signed in to change notification settings - Fork 1
Add Variables to Experiment
Jonny M edited this page Mar 10, 2018
·
2 revisions
Experiments require further setup before submitting for optimization. You will need to add variables to your experiment.
- name (str): The name of the variable. Must be unique.
- v_type (str): Type of variable. (int, float, or enum)
- v_range (list/tuple): A min and max range to test (only for int/float Variables)
- values (list): The list of possible enum values (only for enum Variables)
- step_size(int or float): the preferred step size for use with grid search
- Integer
- Float
- Enum
Integers and Floats require a range and step:
v_range=[v_min, v_max]
step_size
enum requires a set of values.
values=["v1", "v2", ..., "v3"]
iris_expt2.add_var(Variable("hidden_size", "int", v_range=[5, 20], step_size=1))
iris_expt2.add_var(Variable("lr", "float", v_range=[0.001, 0.1], step_size=.01))
iris_expt2.add_var(Variable("optimizer", "enum", values=["gd", "adam", "rms"]))