diff --git a/opensoundscape/localization.py b/opensoundscape/localization.py
index 6481fec8..5e9e4436 100644
--- a/opensoundscape/localization.py
+++ b/opensoundscape/localization.py
@@ -65,19 +65,47 @@ def __init__(
                 - Estimates the tdoas using cross_correlation if not already estimated.
                 - Estimates the location of the event using the tdoas and receiver locations.
                 - Returns the location estimate as a tuple of cartesian coordinates (x,y) or (x,y,z)
+
+        Editable Attributes:
+            These are parameters that can be set before calling estimate_location()
+            min_n_receivers: minimum number of receivers that must detect an event for it to be localized
+            cc_threshold: threshold for cross correlation
+            cc_filter: filter for generalized cross correlation, see
+                opensoundscape.signal_processing.gcc()
+            max_delay: only delays in +/- this range (seconds) will be considered for possible delay
+                (see opensoundscape.signal_processing.tdoa());
+            bandpass_range: bandpass audio to [low, high] frequencies in Hz before cross correlation
+            speed_of_sound: speed of sound in meters per second
+
+        Static Attributes:
+            receiver_files: list of audio files, one for each receiver
+            receiver_locations: list of [x,y] or [x,y,z] positions of each receiver in meters
+            start_timestamp: start time of detection as datetime.datetime
+            duration: length in seconds of the event
+            class_name: name of detection's class
+        Computed Attributes:
+            tdoas: time delay at each receiver (computed by _estimate_delays())
+            cc_maxs: max of cross correlation for each time delay (computed by _estimate_delays())
+            location_estimate: spatial position estimate (computed by estimate_location())
+            distance_residuals: distance residuals in meters (computed by estimate_location())
+            receivers_used_for_localization: list of receivers used for localization during estimate_location()
+            residual_rms: root mean square of distance residuals (computed by estimate_location())
         """
+        # editable attributes
+        self.min_n_receivers = min_n_receivers
+        self.cc_threshold = cc_threshold
+        self.cc_filter = cc_filter
+        self.max_delay = max_delay
+        self.bandpass_range = bandpass_range
+        self.speed_of_sound = speed_of_sound
+
+        # static attributes
         self.receiver_files = receiver_files
         self.receiver_locations = np.array(receiver_locations)
-        self.max_delay = max_delay
-        self.receiver_start_time_offsets = receiver_start_time_offsets
         self.start_timestamp = start_timestamp
-        self.min_n_receivers = min_n_receivers
         self.duration = duration
         self.class_name = class_name
-        self.bandpass_range = bandpass_range
-        self.cc_threshold = cc_threshold
-        self.cc_filter = cc_filter
-        self.speed_of_sound = speed_of_sound
+        self.receiver_start_time_offsets = receiver_start_time_offsets
 
         # Verify that max_delay is not longer than the duration of the audio and raise a value error if it is
         if self.max_delay >= self.duration:
@@ -85,7 +113,7 @@ def __init__(
                 f"max_delay ({self.max_delay}) is longer than duration ({self.duration}) of audio clips."
             )
 
-        # Initialize attributes
+        # computed attributes
         self.tdoas = None  # time delay at each receiver
         self.cc_maxs = None  # max of cross correlation for each time delay
         self.location_estimate = None  # cartesian location estimate in meters
diff --git a/opensoundscape/ml/cnn.py b/opensoundscape/ml/cnn.py
index 725135be..4c994153 100644
--- a/opensoundscape/ml/cnn.py
+++ b/opensoundscape/ml/cnn.py
@@ -847,14 +847,56 @@ def __init__(self, *args, **kwargs):
 
         Args:
             see SpectrogramModule for arguments
+
+        Methods:
+            predict: generate predictions across a set of audio files or a dataframe defining audio files and start/end clip times
+            train: fit the machine learning model using training data and evaluate with validation data
+            save: save the model to a file
+            load: load the model from a file
+            embed: generate embeddings for a set of audio files
+            generate_embeddings: generate embeddings for a set of audio files
+            generate_cams: generate gradient activation maps for a set of audio files
+            eval: evaluate performance by applying self.torch_metrics to predictions and labels
+            run_validation: test accuracy by running inference on a validation set and computing metrics
+            change_classes: change the classes that the model predicts
+            freeze_feature_extractor: freeze all layers except the classifier
+            freeze_layers_except: freeze all parameters of a model, optionally exluding some layers
+            train_dataloader: create dataloader for training
+            predict_dataloader: create dataloader for inference (predict/validate/test)
+            save_weights: save just the self.network state dict to a file
+            load_weights: load just the self.network state dict from a file
+
+        Editable Attributes & Properties:
+            single_target: (bool) if True, predict only class with max score
+            device: (torch.device or str) device to use for training and inference
+            preprocessor: object defining preprocessing and augmentation operations, e.g. SpectrogramPreprocessor
+            network: pytorch model object, e.g. Resnet18
+            loss_fn: callable object to use for calculating loss during training, e.g. BCEWithLogitsLoss_hot()
+            optimizer_params: (dict) with "class" and "kwargs" keys for class.__init__(**kwargs)
+            lr_scheduler_params: (dict) with "class" and "kwargs" for class.__init__(**kwargs)
+            use_amp: (bool) if True, uses automatic mixed precision for training
+            wandb_logging: (dict) settings for logging to Weights and Biases
+            score_metric: (str) name of the metric for overall evaluation - one of the keys in self.torch_metrics
+            log_file: (str) path to save output to a text file
+            logging_level: (int) amt of logging to log file. 0 for nothing, 1,2,3 for increasing logged info
+            verbose: (int) amt of logging to stdout. 0 for nothing, 1,2,3 for increasing printed output
+
+        Other attributes:
+            torch_metrics: dictionary of torchmetrics name:object pairs to use for calculating metrics
+                - override _init_torch_metrics() method in a subclass rather than modifying directly
+            classes: list of class names
+                - set via __init__() or change_classes(), rather than modifying directly
         """
         super().__init__(*args, **kwargs)
 
-        self.log_file = None  # specify a path to save output to a text file
-        self.logging_level = 1  # 0 for nothing, 1,2,3 for increasing logged info
-        self.verbose = 1  # 0 for nothing, 1,2,3 for increasing printed output
-        self.scheduler = None  # learning rate scheduler
-        self.optimizer = None  # optimizer during training
+        self.log_file = None
+        """specify a path to save output to a text file"""
+        self.logging_level = 1
+        """amount of logging to self.log_file. 0 for nothing, 1,2,3 for increasing logged info"""
+        self.verbose = 1
+        """amount of logging to stdout. 0 for nothing, 1,2,3 for increasing printed output"""
+        self.scheduler = None  # learning rate scheduler, initialized during configure_optimizers() call
+        self.optimizer = None  # optimizer during training , initialized during configure_optimizers() call
         self.current_epoch = 0
         """track number of trained epochs"""