vak.eval.frame_classification.eval_frame_classification_model¶
- vak.eval.frame_classification.eval_frame_classification_model(model_config: dict, dataset_config: dict, trainer_config: dict, checkpoint_path: str | Path, labelmap_path: str | Path, output_dir: str | Path, num_workers: int, frames_standardizer_path: str | Path | None = None, post_tfm_kwargs: dict | None = None) None [source]¶
Evaluate a trained model.
- Parameters:
model_config (dict) – Model configuration in a
dict
. Can be obtained by callingvak.config.ModelConfig.asdict()
.dataset_config (dict) – Dataset configuration in a
dict
. Can be obtained by callingvak.config.DatasetConfig.asdict()
.trainer_config (dict) – Configuration for
lightning.pytorch.Trainer
. Can be obtained by callingvak.config.TrainerConfig.asdict()
.checkpoint_path (str, pathlib.Path) – Path to directory with checkpoint files saved by Torch, to reload model
output_dir (str, pathlib.Path) – Path to location where .csv files with evaluation metrics should be saved.
labelmap_path (str, pathlib.Path) – Path to ‘labelmap.json’ file.
num_workers (int) – Number of processes to use for parallel loading of data. Argument to torch.DataLoader. Default is 2.
frames_standardizer_path (str, pathlib.Path) – Path to a saved
vak.transforms.FramesStandardizer
object used to standardize (normalize) frames. If frames were standardized during training and this is not provided, will give incorrect results. Default is None.post_tfm_kwargs (dict) – Keyword arguments to post-processing transform. If None, then no additional clean-up is applied when transforming labeled timebins to segments, the default behavior. The transform used is
vak.transforms.frame_labels.PostProcess`. Valid keyword argument names are 'majority_vote' and 'min_segment_dur', and should be appropriate values for those arguments: Boolean for ``majority_vote
, a float value formin_segment_dur
. See the docstring of the transform for more details on these arguments and how they work.
Notes
Note that unlike
predict()
, this function can modifylabelmap
so that metrics like edit distance are correctly computed, by converting any string labels inlabelmap
with multiple characters to (mock) single-character labels, withvak.labels.multi_char_labels_to_single_char()
.