Models

MONAI Interoperability

Meddlr’s configuration files can also be used to build models from monai <https://docs.monai.io/en/stable/>`.

First, find model_name, the class name of the model you want to build. The casing of the string must match the casing of the class name in MONAI.

Then identify the arguments that the model requires for intantiation. To find the argument names, look at the __init__ method of the model class.

MODEL:
    name: monai/<model_name>
    MONAI:
        <model_name>:
            <arg1_name>: <value1>
            <arg2_name>: <value2>
            ...

An example configuration for building a MONAI VNet model could look like:

MODEL:
    name: monai/VNet
    MONAI:
        VNet:
            in_channels: 1
            out_channels: 1
            spatial_dims: 2
            dropout_dim: 2
            dropout_prob: 0.2

Building the models from the config is identical to how we do it for built-in models:

from meddlr.config import get_cfg, CfgNode

cfg = get_cfg()

# If the config is loaded from a file
cfg.merge_from_file("config.yaml")

# If the config is built in python.
cfg = get_cfg()
cfg.MODEL.MONAI.VNet = CfgNode()
cfg.MODEL.MONAI.VNet.in_channels = 1
cfg.MODEL.MONAI.VNet.out_channels = 1
cfg.MODEL.MONAI.VNet.spatial_dims = 2
cfg.MODEL.MONAI.VNet.dropout_dim = 2
cfg.MODEL.MONAI.VNet.dropout_prob = 0.2

# The torch.nn.Module.
model = build_model(cfg)