spine.utils.inference

Module with helper functions to run inference on a model configuration.

Functions

get_inference_cfg(cfg[, file_keys, ...])

Turns a training configuration into an inference configuration.
spine.utils.inference.get_inference_cfg(cfg, file_keys=None, weight_path=None, batch_size=None, num_workers=None, cpu=False)[source]

Turns a training configuration into an inference configuration.

This script does the following: - Turn train to False - Set sequential sampling - Load the specified validation file_keys, if requested - Load the specified set of weight_path, if requested - Reset the batch_size to a different value, if requested - Sets num_workers to a different value, if requested - Make the model run in CPU mode, if requested

Parameters:

  • cfg (Union[str, dict]) – Configuration file or Path to the configuration file

  • file_keys (str, optional) – Path to the dataset to use for inference

  • weight_path (str, optional) – Path to the weigths to use for inference

  • batch_size (int, optional) – Number of data samples per batch

  • num_workers (int, optional) – Number of workers that load data

  • cpu (bool, default False) – Whether or not to execute the inference on CPU

Returns:

Dictionary of parameters to initialize handlers

Return type:

dict