yaml-models

Python package generating torch models from a yaml configuration file. Each type item in model list need to specify at minimum the default values needed by the torch.nn module. Arguments not specified will use the default values.

Example config:

model:
-   type: Linear
    in_features: 2
    out_features: 10
-   type: ReLU
-   type: Linear
    in_features: 10
    out_features: 1
-   type: Sigmoid

torch.nn.Linear needs the default arguments in_features and out_features.

Usage

pip install yaml-models

>>> from yaml_models.model import Model
>>> model = Model(config_path="./example_config/model.yaml")
>>> print(model)
Model(
  (layers): ModuleList(
    (0): Linear(in_features=2, out_features=10, bias=True)
    (1): ReLU()
    (2): Linear(in_features=10, out_features=1, bias=True)
    (3): Sigmoid()
  )
)