kf = KFold(n_splits=n_folds)
trainval_id_type_list = np.array(trainval_id_type_list)
for train_index, test_index in kf.split(trainval_id_type_list):
train_id_type_list, val_id_type_list = trainval_id_type_list[
train_index], trainval_id_type_list[test_index]
if len(val_fold_indices) > 0:
if val_fold_index not in val_fold_indices:
val_fold_index += 1
continue
val_fold_index += 1
print("\n\n ---- Validation fold index: ", val_fold_index, "/", n_folds)
print(datetime.now(), len(train_id_type_list), len(val_id_type_list))
assert len(to_set(train_id_type_list)
& to_set(val_id_type_list)) == 0, "WTF"
cnn = params['network'](lr=params['lr_kwargs']['lr'],
**params,
**params['network_kwargs'])
params['save_prefix'] = params['save_prefix_template'].format(
cnn_name=cnn.name, fold_index=val_fold_index - 1)
print("\n {} - Loaded {} model ...".format(datetime.now(), cnn.name))
if 'pretrained_model' in params:
load_pretrained_model(cnn, **params)
print("\n {} - Start training ...".format(datetime.now()))
h = train(cnn, train_id_type_list, val_id_type_list, **params)
if h is None:
np.random.seed(seed)
cache = DataCache(0) # !!! CHECK BEFORE LOAD TO FLOYD
class_index = 0
trainval_id_type_list = get_id_type_list_for_class(class_index)
n_other_samples = len(trainval_id_type_list)
class_indices = list(equalized_data_classes.keys())
class_indices.remove(class_index)
for index in class_indices:
id_type_list = np.array(get_id_type_list_for_class(index))
id_type_list = list(to_set(id_type_list) - to_set(trainval_id_type_list))
np.random.shuffle(id_type_list)
trainval_id_type_list.extend(id_type_list[:n_other_samples])
print(len(trainval_id_type_list), len(to_set(trainval_id_type_list)))
params = {
'seed': seed,
'xy_provider': image_class_labels_provider,
'network': get_densenet,
'network_kwargs': {
'depth': 121,
'input_shape': (256, 256, 3),
