def __init__(self, config_file):
self.config_data = load_config_data(config_file)
self.train_class = TrainClassifier(self.config_data['subset_config'])
self.search_algo = self.get_search_algo(
self.config_data['search_algo'], self.config_data['space'],
self.config_data['metric'], self.config_data['mode'])
# save subset method, to be used in log dir name
self.subset_method = self.train_class.configdata['dss_strategy'][
'type']
class TagsClassifier:
@classmethod
def __init__(self):
self.algorithm = NaiveBayesClassifier
self.trainer = TrainClassifier()
self.classifier = None
def classify(self, text):
# if not previously trained then train else classifier = redis().get('nosy:tagsClassifier')
self.train()
features = self.trainer.feature_extractor(text.split())
result = self.classifier.classify(features)
# print self.classifier.show_most_informative_features()
return result
@classmethod
def train(cls):
training_set = cls.trainer.train() # TrainClassifier().train()
cls.classifier = cls.algorithm.train(training_set) # NaiveBayesClassifier.train(training_set)
class TagsClassifier():
@classmethod
def __init__(self):
self.algorithm = NaiveBayesClassifier
self.trainer = TrainClassifier()
self.classifier = None
def classify(self, text):
# if not previously trained then train else classifier = redis().get('nosy:tagsClassifier')
self.train()
features = self.trainer.feature_extractor(text.split())
result = self.classifier.classify(features)
#print self.classifier.show_most_informative_features()
return result
@classmethod
def train(cls):
training_set = cls.trainer.train() #TrainClassifier().train()
cls.classifier = cls.algorithm.train(
training_set) #NaiveBayesClassifier.train(training_set)
def __init__(self):
self.algorithm = NaiveBayesClassifier
self.trainer = TrainClassifier()
self.classifier = None
args=parser.parse_args()
n_classes = 10
n_epochs = 200
pre = Preprocessing('digits')
pre.load_data(filename='train.csv', name='train')
X_df = pre.get(name='train').drop(columns=['0'])
y_df = pre.get(name='train')['0']
dtype = torch.float
device = torch.device("cpu")
model_name = 'logreg_digits'
model = LogReg(model_name, 256, n_classes)
learning_rate = 0.0001
batch_size = 32
train_classifier = TrainClassifier(model, X_df, y_df)
trained_model , optimizer, criterion, loss_hist, loss_val_hist, best_param = train_classifier.run_train(n_epochs = n_epochs, lr=learning_rate, batch_size=batch_size)
pre.save_results(loss_hist, loss_val_hist, f'{model_name}')
trained_model.load_state_dict(state_dict=best_param)
trained_model.eval()
if args.s_model:
m_exporter = ModelExporter('digits')
m_exporter.save_nn_model(trained_model, optimizer, 0, n_classes, n_epochs, trained_model.get_args())
args=parser.parse_args()
pre = Preprocessing('IMDB')
n_classes = 2
n_features = int(args.n_feat)
n_epochs = 100
pre.load_data(filename=f'training_data_{n_features}.csv', name='training_data')
X_df = pre.get(name='training_data').drop(columns=['target'])
y_df = pre.get(name='training_data')['target']
model = LogReg('log_reg', n_features, n_classes)
train_classifier = TrainClassifier(model, X_df, y_df)
trained_model, optimizer, criterion, loss_hist, loss_validate_hist = train_classifier.run_train(n_epochs = n_epochs)
pre.save_results(loss_hist, loss_validate_hist, f'log_reg_{100}')
m_exporter = ModelExporter('IMDB')
m_exporter.save_nn_model(trained_model, optimizer, n_features, n_classes, n_epochs)
##teeeeeest part
pre.load_data(filename=f'test_data_{n_features}.csv', name='test_data')
X_test_df = pre.get(name='test_data').drop(columns=['target'])
y_test_df = pre.get(name='test_data')['target']
dtype = torch.float
device = torch.device("cpu")
from train import TrainClassifier config_file = "configs/config_gradmatch_cifar10.py" classifier = TrainClassifier(config_file) classifier.configdata['dss_strategy']['select_every'] = 1 classifier.train()
class HyperParamTuning:
def __init__(self, config_file):
self.config_data = load_config_data(config_file)
self.train_class = TrainClassifier(self.config_data['subset_config'])
self.search_algo = self.get_search_algo(
self.config_data['search_algo'], self.config_data['space'],
self.config_data['metric'], self.config_data['mode'])
# save subset method, to be used in log dir name
self.subset_method = self.train_class.configdata['dss_strategy'][
'type']
def param_tune(self, config):
#update parameters in config dict
new_config = self.update_parameters(self.train_class.configdata,
config)
self.train_class.configdata = new_config
# turn on reporting to ray every time
self.train_class.configdata['report_tune'] = True
self.train_class.train()
def start_eval(self):
analysis = tune.run(self.param_tune,
num_samples=self.config_data['num_evals'],
config=self.config_data['space'],
search_alg=self.search_algo,
resources_per_trial={'gpu': 1},
local_dir=self.config_data['log_dir'] +
self.subset_method + '/',
log_to_file=True,
name=self.config_data['name'],
resume=self.config_data['resume'])
best_config = analysis.get_best_config(
metric=self.config_data['metric'], mode=self.config_data['mode'])
print("Best Config: ", best_config)
if self.config_data['final_train']:
self.final_train(best_config)
def get_search_algo(self, method, space, metric, mode):
if method == "hyperopt" or method == "TPE":
search = HyperOptSearch(space, metric="mean_accuracy", mode="max")
return search
def final_train(self, best_params):
# change strategy to Full (i.e use whole dataset)
# update (optimized) parameters
new_config = self.update_parameters(self.train_class.configdata,
best_params)
self.train_class.configdata = new_config
self.train_class.configdata['dss_strategy']['type'] = 'Full'
self.train_class.train()
def update_parameters(self, config, new_config):
# a generic function to update parameters
if 'learning_rate' in new_config:
config['optimizer']['lr'] = new_config['learning_rate']
if 'optimizer' in new_config:
config['optimizer']['type'] = new_config['optimizer']
if 'trn_batch_size' in new_config:
config['dataloader']['batch_size'] = new_config['trn_batch_size']
return config
from train import TrainClassifier config_file = "configs/config_gradmatch_cifar10.py" classifier = TrainClassifier(config_file) classifier.configdata['dss_strategy']['select_every'] = 1 classifier.configdata['model']['architecture'] = 'MobileNet2' classifier.configdata['optimizer']['weight_decay'] = 4e-5 classifier.train()
def __init__(self):
self.algorithm = NaiveBayesClassifier
self.trainer = TrainClassifier()
self.classifier = None
