def make_dataflow_demo(env):
ensure_load()
# return feed_dict, extra_info
def split_data(img, label):
return dict(img=img[np.newaxis].astype('float32'))
df = _mnist[1] # use validation set actually
df = flow.DictOfArrayDataFlow(df)
df = flow.tools.cycle(df)
df = flow.tools.ssmap(split_data, df)
return df
def make_dataflow_inference(env):
ensure_load()
batch_size = get_env('inference.batch_size')
epoch_size = get_env('inference.epoch_size')
df = _mnist[1] # use validation set actually
df = flow.DictOfArrayDataFlow(df)
df = flow.tools.cycle(df)
df = flow.BatchDataFlow(df, batch_size,
sample_dict={'img': np.empty(shape=(batch_size, 28, 28, 1), dtype='float32'), })
df = flow.EpochDataFlow(df, epoch_size)
return df
def make_dataflow_demo(env):
num_classes = get_env('dataset.nr_classes')
ensure_load(num_classes)
# return feed_dict, extra_info
def split_data(img, label):
return dict(img=img[np.newaxis].astype('float32')), dict(label=label)
df = _cifar[1] # use validation set actually
df = flow.DictOfArrayDataFlow(df)
df = flow.tools.cycle(df)
df = flow.tools.ssmap(split_data, df)
return df
def main_demo_infogan(env, func):
net = env.network
samples = net.zc_distrib.numerical_sample(net.zc_distrib_num_prior)
df = {'zc': samples.reshape(samples.shape[0], 1, -1)}
df = flow.DictOfArrayDataFlow(df)
all_outputs = []
for data in tqdm.tqdm(df, total=len(df), **get_tqdm_defaults()):
res = func(**data)
all_outputs.append(res['output'][0, :, :, 0])
grid_desc = get_env('demo.infogan.grid_desc')
final = image.image_grid(all_outputs, grid_desc)
final = (final * 255).astype('uint8')
image.imwrite('infogan.png', final)
def make_dataflow_demo(env):
reconstruct = get_env('demo.is_reconstruct', False)
if reconstruct:
ensure_load()
# return feed_dict, extra_info
def split_data(img, label):
return dict(img=img[np.newaxis].astype('float32'))
df = _mnist[1] # use validation set actually
df = flow.DictOfArrayDataFlow(df)
df = flow.tools.cycle(df)
df = flow.tools.ssmap(split_data, df)
else:
df = flow.EmptyDictDataFlow()
return df
def make_dataflow_inference(env):
num_classes = get_env('dataset.nr_classes')
ensure_load(num_classes)
batch_size = get_env('inference.batch_size')
epoch_size = get_env('inference.epoch_size')
df = _cifar[1] # use validation set actually
df = flow.DictOfArrayDataFlow(df)
df = flow.tools.cycle(df)
df = flow.BatchDataFlow(df,
batch_size,
sample_dict={
'img':
np.empty(shape=(batch_size, _cifar_img_dim,
_cifar_img_dim, 3),
dtype='float32'),
'label':
np.empty(shape=(batch_size, ), dtype='int32')
})
df = flow.EpochDataFlow(df, epoch_size)
return df