test_norm = test.astype('float32')
# normalize to range 0-1
train_norm = train_norm / 255.0
test_norm = test_norm / 255.0
# return normalized images
return train_norm, test_norm
# prepare pixel data
trainX, testX = prep_pixels(trainX, testX)
hp = HP()
hp.save_path = 'saved_runs'
hp.description = "syclop micro feature learning runs"
hp.this_run_name = 'micro_{}'.format(run_index)
deploy_logs()
#%%
############################### Get Trained Teacher ##########################3
path = '/home/orram/Documents/GitHub/imagewalker/teacher_student/'
path = '/home/labs/ahissarlab/orra/imagewalker/teacher_student/'
def train_model(path, trainX, trainY):
def net():
input = keras.layers.Input(shape=(32, 32, 3))
#Define CNN
x = keras.layers.Conv2D(32, (3, 3),
activation='relu',
padding='same',
action='store_false')
parser.set_defaults(eval_mode=False,
decode_from_dvs=False,
test_mode=False,
rising_beta_schedule=True,
decoder_ignore_position=False,
curriculum_enable=True,
conv_fe=False,
acceleration_mode=False)
config = parser.parse_args()
config = vars(config)
hp.upadte_from_dict(config)
hp.this_run_name = sys.argv[
0] + '_noname_' + hp.run_name_suffix + '_' + lsbjob + '_' + str(
int(time.time()))
#define model
n_timesteps = hp.steps_per_episode
##
deploy_logs()
##
if hp.decoder_arch == 'multicore_201':
decoder = rnn_model_multicore_201(
n_cores=hp.decoder_n_cores,
lr=hp.decoder_learning_rate,
ignore_input_B=hp.decoder_ignore_position,
dropout=hp.decoder_dropout,
rnn_type=hp.decoder_rnn_type,