The Pythonic API

Hypster’s API is designed to be intuitive and expressive, allowing you to use familiar Python constructs in your configuration functions.

Conditional Statements for Dependent Variables

You can use conditional statements to define dependent variables:

from hypster import HP, config


@config
def conditional_config(hp: HP):
    model_type = hp.select(["CNN", "RNN", "Transformer"], default="CNN")

    if model_type == "CNN":
        num_layers = hp.select([3, 5, 7], default=5)
    elif model_type == "RNN":
        cell_type = hp.select(["LSTM", "GRU"], default="LSTM")
    else:  # Transformer
        num_heads = hp.select([4, 8, 16], default=8)

---------------------------------------------------------------------------
ModuleNotFoundError                       Traceback (most recent call last)
Cell In[1], line 1
----> 1 from hypster import HP, config
      4 @config
      5 def conditional_config(hp: HP):
      6     model_type = hp.select(["CNN", "RNN", "Transformer"], default="CNN")

ModuleNotFoundError: No module named 'hypster'

Loops

You can use loops to define repetitive configurations:

from hypster import HP, config


@config
def loop_config(hp: HP):
    num_layers = hp.select([3, 5, 7], default=5)
    layer_sizes = []

    for i in range(num_layers):
        layer_sizes.append(hp.select([32, 64, 128], default=64, name=f"layer_{i}_size"))

Changing Options Conditionally

You can dynamically change the options based on other selections:

from hypster import HP, config


@config
def dynamic_options_config(hp: HP):
    dataset_size = hp.select(["small", "medium", "large"], default="medium")

    if dataset_size == "small":
        model_options = ["simple_cnn", "small_rnn"]
    elif dataset_size == "medium":
        model_options = ["resnet", "lstm"]
    else:
        model_options = ["transformer", "large_cnn"]

    model = hp.select(model_options)

Summary

By allowing pythonic configuration spaces you can:

• Use conditional statements to define dependent variables

• Utilize loops for repetitive configurations

• Dynamically change options based on other selections

• And much more! :)