probnmn.evaluators.joint_training_evaluator

class probnmn.evaluators.joint_training_evaluator.JointTrainingEvaluator(config: probnmn.config.Config, models: Dict[str, Type[torch.nn.modules.module.Module]], gpu_ids: List[int] = [0], cpu_workers: int = 0)[source]

Bases: probnmn.evaluators._evaluator._Evaluator

Performs evaluation for joint_training phase, using batches of evaluation examples from JointTrainingDataset.

Parameters
config: Config

A Config object with all the relevant configuration parameters.

models: Dict[str, Type[nn.Module]]

All the models which interact with each other for evaluation. This should come from JointTrainingTrainer.

gpu_ids: List[int], optional (default=[0])

List of GPU IDs to use or evaluation, [-1] - use CPU.

cpu_workers: int, optional (default = 0)

Number of CPU workers to use for fetching batch examples in dataloader.

Examples

To evaluate a pre-trained checkpoint:

>>> config = Config("config.yaml")  # PHASE must be "joint_training"
>>> trainer = JointTrainingTrainer(config, serialization_dir="/tmp")
>>> trainer.load_checkpoint("/path/to/joint_training_checkpoint.pth")
>>> evaluator = JointTrainingEvaluator(config, trainer.models)
>>> eval_metrics = evaluator.evaluate(num_batches=50)
_do_iteration(self, batch:Dict[str, Any]) → Dict[str, Any][source]

Perform one iteration, given a batch. Take a forward pass to accumulate metrics in ProgramGenerator and NeuralModulenetwork.

Parameters
batch: Dict[str, Any]

A batch of evaluation examples sampled from dataloader.

Returns
Dict[str, Any]

A dictionary containing model predictions and/or batch validation losses of ProgramGenerator and NeuralModuleNetwork. Nested dict structure:

{
    "program_generator": {"predictions", "loss"}
    "nmn": {"predictions", "loss"}
}