get_transforms_paths

behavenet.data.utils.get_transforms_paths(data_type, hparams, sess_id, check_splits=True)[source]

Helper function for generating session-specific transforms and paths.

Parameters:

  • data_type (str) – ‘neural’ | ‘ae_latents’ | ‘arhmm_states’ | ‘neural_ae_predictions’ | ‘neural_arhmm_predictions’

  • hparams (dict) –

    • required keys for data_type=neural: ‘neural_type’, ‘neural_thresh’

    • required keys for data_type=ae_latents: ‘ae_experiment_name’, ‘ae_model_type’, ‘n_ae_latents’, ‘ae_version’ or ‘ae_latents_file’; this last option defines either the specific ae version (as ‘best’ or an int) or a path to a specific ae latents pickle file.

    • required keys for data_type=arhmm_states: ‘arhmm_experiment_name’, ‘n_arhmm_states’, ‘kappa’, ‘noise_type’, ‘n_ae_latents’, ‘arhmm_version’ or ‘arhmm_states_file’; this last option defines either the specific arhmm version (as ‘best’ or an int) or a path to a specific ae latents pickle file.

    • required keys for data_type=neural_ae_predictions: ‘neural_ae_experiment_name’, ‘neural_ae_model_type’, ‘neural_ae_version’ or ‘ae_predictions_file’ plus keys for neural and ae_latents data types.

    • required keys for data_type=neural_arhmm_predictions: ‘neural_arhmm_experiment_name’, ‘neural_arhmm_model_type’, ‘neural_arhmm_version’ or ‘arhmm_predictions_file’, plus keys for neural and arhmm_states data types.

  • sess_id (dict) – each list entry is a session-specific dict with keys ‘lab’, ‘expt’, ‘animal’, ‘session’

  • check_splits (bool, optional) – check data splits and data rng seed between hparams and loaded model outputs (e.g. latents)

Returns:

Return type:

tuple