def main(argv=None):
# unroll arguments of train
config_agent.init_FLAGS('train')
VARS['mode'] = 'train'
load_data_lst(data_lst)
FLAGS['input_queue']['capacity'] = 1
Multi_gpu_solver().start()
def main(argv=None):
# unroll arguments of eval
config_agent.init_FLAGS('eval')
VARS['mode'] = 'eval'
Eval_solver().start()
def main(argv=None):
# unroll arguments of train
config_agent.init_FLAGS('train')
Train_solver().start()
self.reader.close()
def read_into_hdf5(self):
max_steps = self.all_count // self.input_batch_size
print('total steps: %d' % max_steps)
records = ''
for step in xrange(max_steps):
hdf5_path = self.hdf5_path.replace('$', '%d' % step)
f = h5py.File(hdf5_path, 'w')
f.create_dataset('data', (self.input_batch_size, 3, 224, 224))
f.create_dataset('label', (self.input_batch_size, 1, 1, 1))
inputs = self.sess.run(self.inputs)
assert (inputs['X'].shape[0] == self.input_batch_size)
f['data'][:] = np.transpose(inputs['X'], [0, 3, 1, 2])[:]
f['label'][:] = inputs['Y'].reshape(
[self.input_batch_size, 1, 1, 1])[:]
f.close()
records += '%s\n' % hdf5_path
if step % 1 == 0:
print(step)
with open(self.txt_path, 'w') as f:
f.write(records)
if __name__ == '__main__':
config_agent.init_FLAGS('train')
VARS['mode'] = 'train'
Fake_reader().start()
def main(argv=None):
#unroll arguments of prediction
config_agent.init_FLAGS('eval')
Output_solver().start()
def main(argv=None):
# unroll arguments of prediction
config_agent.init_FLAGS('eval')
VARS['mode'] = 'test'
Test_solver().start()
def main(argv=None):
# unroll arguments of train
config_agent.init_FLAGS('train')
VARS['mode'] = 'train'
Multi_gpu_solver().start()