Prediction configuration

Prediction configuration lives under the Predictor root object.

Predictor:
  Model:
    classpath: UNet.yml
    UNet:
      ...
  Dataset:
    ...
  outputs_dataset:
    ...
  train_name: SEG_BASELINE

Running it

From the directory that contains Prediction.yml and the Checkpoints/ folder written by training:

konfai PREDICTION -y --gpu 0 --config Prediction.yml \
  --models Checkpoints/SEG_BASELINE/<checkpoint>.pt

You can also pass multiple checkpoints:

konfai PREDICTION -y --gpu 0 --config Prediction.yml \
  --models ckpt_a.pt ckpt_b.pt ckpt_c.pt

When multiple checkpoints are provided, the predictor combines them using the combine strategy from the YAML, usually Mean or Median.

Top-level fields

Field	Type	Default in code	Required	Effect
Model	mapping	ModelLoader()	Yes	Selects the model class used for prediction.
Dataset	mapping	DataPrediction()	Yes	Defines inference data loading and test-time augmentation.
outputs_dataset	mapping	default output dataset	Yes in practice	Controls which outputs are written to disk and how.
combine	string	Mean	No	Reduces outputs across multiple checkpoints.
train_name	string	"name"	Yes in practice	Names the prediction run and output folder.
manual_seed	int or null	None	No	Optional seed.
gpu_checkpoints	list or null	None	No	Module placement optimization.
autocast	bool	false	No	Enables AMP during inference.
data_log	list or null	None	No	Optional TensorBoard logging.

Predictor.Model

Prediction uses the same classpath convention as training:

Model:
  classpath: Model:UNetpp5
  outputs_criterions: {}

In most prediction configs:

• you select the architecture

• you keep only the inference-relevant parameters

• you disable or simplify training-only criteria

Checkpoint loading is controlled by the CLI argument --models, not by the YAML file itself.

Predictor.Dataset

Prediction datasets are instantiated through DataPrediction.

Key fields:

Field	Type	Effect
dataset_filenames	list[str]	Input dataset sources.
groups_src	mapping	Input groups and preprocessing transforms.
augmentations	mapping	Test-time augmentation definitions.
Patch	mapping	Sliding-window or slice-wise inference setup.
subset	object	Restricts evaluated cases.
batch_size	int	Number of patches per inference batch.

Use Dataset.Patch when:

• the full input does not fit in memory

• you want slice-wise or sliding-window inference

• you need the same spatial strategy as training

outputs_dataset

outputs_dataset defines how selected model outputs become files on disk.

outputs_dataset:
  Head:Tanh:
    OutputDataset:
      name_class: OutSameAsGroupDataset
      group: sCT
      same_as_group: MR:MR
      reduction: Mean

Important nested fields:

Field	Effect
output key	Selects the model output to export.
name_class	Selects the output dataset implementation.
group	Output group name written to disk.
dataset_filename	Destination dataset path and format.
same_as_group	Geometry reference group for exported volumes.
before_reduction_transforms	Applied before combining ensemble or TTA outputs.
after_reduction_transforms	Applied after reduction.
final_transforms	Final transforms applied before writing.
reduction	Combines multiple predictions, usually Mean or Median.
patch_combine	Optional patch reassembly strategy.

Note

One Prediction.yml can be shared between different checkpoints as long as the exported output name stays consistent.

Examples

See:

• examples/Segmentation/Prediction.yml

• examples/Synthesis/Prediction.yml

Troubleshooting

• If geometry or intensity range is wrong, review the final transforms in outputs_dataset.

Next steps

• Evaluation configuration — to score the written predictions against ground truth.

• Datasets and groups — the shared dataset_filenames, groups_src, and subset conventions.

• Model graph and output naming — how the model output paths referenced by outputs_dataset are named.