settings.py

import torch
seed = 97

settings = dict()
settings['bohb_results_dir'] = "./BOHB_results/"
settings['es_results_dir'] = './ES_results'
settings['data_dir'] = './data'
settings['loss_dict'] = {'cross_entropy': torch.nn.CrossEntropyLoss
            }#'mse': torch.nn.MSELoss}
settings['opti_dict'] = {'adam': torch.optim.Adam,
            'adad': torch.optim.Adadelta,
            'sgd': torch.optim.SGD}
settings['run_dir'] = './stored_models/'
settings['init_channels'] = 8
settings['batch_size'] = 64
settings['seed'] = seed

def get(variable):
    if variable in settings.keys():
        return settings[variable]
    else:
        raise KeyError

darts_args=dict()
darts_args['data'] ='./data'    #'location of the data corpus'   
darts_args['set'] ='cifar10'    #'location of the data corpus'   
darts_args['batch_size'] =32    #'batch size'   
darts_args['learning_rate']=0.1    #'init learning rate'   
darts_args['learning_rate_min']=0.001    #'min learning rate'   
darts_args['momentum']=0.9    #'momentum'   
darts_args['weight_decay']=3e-4    #'weight decay'   
darts_args['report_freq']=50    #'report frequency'   
darts_args['gpu']='cuda:0'    #'gpu device id'   
darts_args['epochs']= 20    #'num of training epochs'    
darts_args['init_channels']=8    #'num of init channels'   
darts_args['layers']= 4    #'total number of layers'   
darts_args['model_path']= 'saved_models'    #'path to save the model'   
darts_args['cutout']= False    #'use cutout'   
darts_args['cutout_length']= 16    #'cutout length'   
darts_args['drop_path_prob']= 0.3    #'drop path probability'   
darts_args['save']= 'EXP'    #'experiment name'   
darts_args['seed']=seed   #'random seed'   
darts_args['grad_clip']=5    #'gradient clipping'   
darts_args['train_portion']=0.5    #'portion of training data'   
darts_args['unrolled']=False    #'use one-step unrolled validation loss'   
darts_args['arch_learning_rate']=6e-4    #'learning rate for arch encoding'   
darts_args['arch_weight_decay']=1e-3    #'weight decay for arch encoding'   


main_args = dict()
main_args['seed']      =seed      # help='random seed',type=int)
main_args['batch_size']=64      # help='Batch size',type=int)
main_args['data_dir']  ='./data'# help='Directory in which the data is stored (can be downloaded)')
main_args['optimizer'] ='adam'  # help='Which optimizer to use during training',choices=list(get('opti_dict').keys()),type=str)
main_args['exp_no']    =5       # help='Experiment number',type=str)
main_args['verbose']   ='INFO'  # choices=['INFO', 'DEBUG'],help='verbosity'


def get_main_args():
    return main_args

def get_darts_args():
    return darts_args