comments possible options

code/config.py

@@ -6,53 +6,53 @@
 freq_bands = 64
 
 # model parameters
-verbose = True
+verbose = True  # [True, False]
 n_filters = [32, 64, 128]
 pools_size = [(1, 10), (1, 5), (1, 5)]
 dropouts_rate = [0.3, 0.3, 0.3]
 
 ratio = 2
 
-reshape_method = 'global_avg'
+reshape_method = 'global_avg'  # ['global_avg', 'global_max', 'flatten']
 dense_layer = None
 dropouts_rate_cl = None
 
-split_freqs = False
+split_freqs = False  # [True, False]
 n_split_freqs = 3
 f_split_freqs = [64, 128]
 
 # callbacks
 # Early Stopping
-early_stopping = True
+early_stopping = True  # [True, False]
 monitor_es = None
 min_delta_es = None
 mode_es = None
 patience_es = None
 
 # Get Learning Rate
-get_lr_after_epoch = True
+get_lr_after_epoch = True  # [True, False]
 
 # Reduce LR OnPlateau
-lr_on_plateau = True
+lr_on_plateau = True  # [True, False]
 monitor_lr_on_plateau = None
 factor_lr_on_plateau = None
 patience_lr_on_plateau = None
 min_lr_on_plateau = None
 
 # Save models and csvs
-save_outputs = True
+save_outputs = True  # [True, False]
 outputs_path = '/repos/DCASE2021-Task1/outputs/'
 best_model_name = 'best.h5'
 last_model_name = 'last.h5'
 log_name = 'log.csv'
 
 # training hyperparamteres
-quick_test = True
-loss_type = 'focal_loss'
+quick_test = True  # [True, False]
+loss_type = 'focal_loss'  # ['focal_loss', 'categorical_loss']
 fl_alpha = 0.25  # needed if focal
 fl_gamma = 2  # needed if focal
 data_augmentation = 'mixup'
 epochs = 500
 mixup_alpha = 0.4
 batch_size = 32
-training_verbose = None
+training_verbose = None  # [None, 0, 1, 2]

code/tests.py

@@ -4,6 +4,7 @@
 import keras
 from utils import create_folder_time, moving_config_file_to_folder
 
+
 def check_reshape_variable(reshape_method):
     possible_options = ['global_avg', 'global_max', 'flatten']
 
@@ -61,7 +62,6 @@ def is_boolean(inp):
 
 
 def check_callbacks():
-
     if (config.early_stopping is not True and config.get_lr_after_epoch is not True
             and config.factor_lr_on_plateau and config.save_outputs is not True):
         return None