def plot_JEMMIG_num_latents(save_dir, JEMMIG_result_files, labels, num_latents):
assert len(JEMMIG_result_files) == len(labels) == len(num_latents), \
"len(SEPIN_result_files)={}, len(labels)={}, len(num_latents)={}".format(
len(JEMMIG_result_files), len(labels), len(num_latents))
JEMMIGs = []
for i in range(len(JEMMIG_result_files)):
JEMMIG_results = np.load(JEMMIG_result_files[i], "r")
JEMMIGs.append(np.mean(JEMMIG_results['JEMMIG_yk']))
# =========================================== #
font = {'size': 12}
matplotlib.rc('font', **font)
width = 0.5
plt.bar(range(0, len(JEMMIGs)), JEMMIGs, width=width, align='center')
plt.xticks(range(0, len(JEMMIGs)), num_latents)
plt.xlabel("#latents")
plt.ylabel("JEMMIG")
subplot_adjust = {'left': 0.20, 'right': 0.98, 'bottom': 0.17, 'top': 0.98}
plt.subplots_adjust(**subplot_adjust)
plt.gcf().set_size_inches(3.2, 3)
save_dir = make_dir_if_not_exist(save_dir)
save_file = join(save_dir, "JEMMIG_num_latents.pdf")
with PdfPages(save_file) as pdf_file:
plt.savefig(pdf_file, format='pdf')
plt.show()
plt.close()
def plot_JEMMIG_zi_num_latents(save_dir, JEMMIG_result_files, labels, num_latents):
assert len(JEMMIG_result_files) == len(labels) == len(num_latents), \
"len(SEPIN_result_files)={}, len(labels)={}, len(num_latents)={}".format(
len(JEMMIG_result_files), len(labels), len(num_latents))
JEMMIGs = []
for n in range(len(JEMMIG_result_files)):
SEPIN_results = np.load(JEMMIG_result_files[n], "r")
JEMMIGs.append(SEPIN_results['JEMMIG_yk'])
# =========================================== #
font = {'size': 12}
matplotlib.rc('font', **font)
for n in range(len(JEMMIGs)):
plt.scatter([n] * len(JEMMIGs[n]), JEMMIGs[n], s=100, alpha=0.3)
plt.xticks(range(0, len(JEMMIGs)), num_latents)
plt.xlabel("#latents")
plt.ylabel("JEMMIG(yk)")
subplot_adjust = {'left': 0.16, 'right': 0.98, 'bottom': 0.17, 'top': 0.98}
plt.subplots_adjust(**subplot_adjust)
plt.gcf().set_size_inches(3.2, 3)
save_dir = make_dir_if_not_exist(save_dir)
save_file = join(save_dir, "JEMMIG_yk_num_latents.pdf")
with PdfPages(save_file) as pdf_file:
plt.savefig(pdf_file, format='pdf')
plt.show()
plt.close()
def plot_info_bar(run_id, save_dir, informativeness_metrics_dir, num_bins,
bin_limits, data_proportion):
z_data_file = join(informativeness_metrics_dir,
"z_data[data={}].npz".format(data_proportion))
with np.load(z_data_file, "r") as f:
all_z_mean = f['all_z_mean']
all_z_stddev = f['all_z_stddev']
# Plotting
# =========================================== #
save_dir = make_dir_if_not_exist(save_dir)
plot_comp_dist(join(save_dir, 'z_mean_{}.pdf'),
all_z_mean,
x_lim=(-5, 5),
subplot_adjust={
'left': 0.1,
'right': 0.98,
'bottom': 0.05,
'top': 0.95
})
plot_comp_dist(join(save_dir, 'z_stddev_{}.pdf'),
all_z_stddev,
x_lim=(0, 3),
subplot_adjust={
'left': 0.1,
'right': 0.98,
'bottom': 0.05,
'top': 0.95
})
def plot_info_bar(run_id, save_dir, informativeness_metrics_dir, num_bins,
bin_limits, data_proportion):
save_dir = make_dir_if_not_exist(save_dir)
save_file = join(save_dir, "info_bar.pdf")
result_file = join(
informativeness_metrics_dir,
'results[bins={},bin_limits={},data={}].npz'.format(
num_bins, bin_limits, data_proportion))
results = np.load(result_file, "r")
# Plotting
# =========================================== #
sorted_MI = results["sorted_MI_z_x"]
norm_sorted_MI = sorted_MI / (1.0 * np.log(num_bins))
plt.bar(range(len(norm_sorted_MI)),
height=norm_sorted_MI,
width=0.8,
color="blue")
plt.ylim(bottom=0, top=1)
plt.xlim(left=-1, right=len(norm_sorted_MI))
plt.xticks(range(0, len(norm_sorted_MI), 1))
plt.tight_layout()
with PdfPages(save_file) as pdf_file:
plt.savefig(pdf_file, format='pdf')
plt.show()
plt.close()
def compare_JEMMIG_norm_models(save_dir, result_files, attr_names, num_bins):
JEMMIGs_norm_by_model = []
ids_sorted_by_factor = []
assert len(result_files) == 3, "These files must be the result files of FactorVAE, BetaVAE and AAEs!"
for i in range(len(result_files)):
results = np.load(result_files[i], "r")
print("JEMMIG_results.keys: {}".format(list(results.keys())))
MI_z_y = results['MI_z_y']
H_z_y = results['H_z_y']
H_y = results['H_y_4_diff_z'][:, 0]
ids_sorted_by_MI = np.argsort(MI_z_y, axis=0)[::-1]
MI_z_y_sorted = np.take_along_axis(MI_z_y, ids_sorted_by_MI, axis=0)
H_z_y_sorted = np.take_along_axis(H_z_y, ids_sorted_by_MI, axis=0)
H_diff = H_z_y_sorted[0, :] - MI_z_y_sorted[0, :]
JEMMIG = H_diff + MI_z_y_sorted[1, :]
JEMMIG_norm = JEMMIG / (np.log(num_bins) + H_y)
JEMMIGs_norm_by_model.append(JEMMIG_norm)
if i == 0: # Sort for FactorVAE
ids_sorted_by_factor = np.argsort(JEMMIG_norm)
#"""
attr_names = [attr_names[i][:10] for i in ids_sorted_by_factor]
for i in range(len(JEMMIGs_norm_by_model)):
JEMMIGs_norm_by_model[i] = JEMMIGs_norm_by_model[i][ids_sorted_by_factor]
font = {'size': 12}
matplotlib.rc('font', **font)
width = 0.6
plt.bar(2 * np.arange(len(attr_names)), JEMMIGs_norm_by_model[0], width=width,
align='center', label='FactorVAE')
plt.bar(2 * np.arange(len(attr_names)) + width, JEMMIGs_norm_by_model[1], width=width,
align='center', label='BetaVAE')
plt.bar(2 * np.arange(len(attr_names)) + 2 * width, JEMMIGs_norm_by_model[2], width=width,
align='center', label='AAE')
plt.xticks(2 * np.arange(len(attr_names)) + 1.5 * width, attr_names, rotation=45, ha='right')
plt.ylabel("JEMMIG (normalized)")
plt.ylim(bottom=0.4)
plt.legend()
subplot_adjust = {'left': 0.04, 'right': 0.996, 'bottom': 0.34, 'top': 0.98}
plt.subplots_adjust(**subplot_adjust)
plt.gcf().set_size_inches(16, 3)
save_dir = make_dir_if_not_exist(save_dir)
save_file = join(save_dir, "JEMMIG_norm_models.pdf")
with PdfPages(save_file) as pdf_file:
plt.savefig(pdf_file, format='pdf')
plt.show()
plt.close()
def plot_SEPIN_num_samples(save_dir, all_SEPIN_result_files, labels,
all_num_samples):
WSEPINs = []
for i in range(len(labels)):
for j in range(len(all_num_samples)):
SEPIN_results = np.load(all_SEPIN_result_files[i][j], "r")
WSEPINs.append(np.mean(SEPIN_results['WSEPIN']))
WSEPINs = np.reshape(np.asarray(WSEPINs, dtype=np.float32),
[len(labels), len(all_num_samples)])
font = {'family': 'normal', 'size': 16}
matplotlib.rc('font', **font)
for i in range(len(labels)):
plt.plot(all_num_samples, WSEPINs[i], '-', label=labels[i], marker='o')
plt.legend()
plt.xlabel("#samples")
plt.ylabel("WSEPIN")
plt.tight_layout()
save_dir = make_dir_if_not_exist(save_dir)
save_file = join(save_dir, "WSEPIN_num_samples.pdf")
with PdfPages(save_file) as pdf_file:
plt.savefig(pdf_file, format='pdf')
plt.show()
plt.close()
def plot_RMIG_sampling_quantized(save_dir, JEMMIG_result_files, interpretability_result_files, labels):
# Plot comparison between rmig and mig_chen
assert len(JEMMIG_result_files) == len(interpretability_result_files), \
"len(JEMMIG_result_files)={}, len(interpretability_result_files)={}".format(
len(JEMMIG_result_files), len(interpretability_result_files))
RMIGs_sampling = []
RMIGs_quantized = []
for i in range(len(JEMMIG_result_files)):
JEMMIG_results = np.load(JEMMIG_result_files[i], "r")
interp_results = np.load(interpretability_result_files[i], "r")
RMIG_sampling = np.mean(JEMMIG_results['RMIG_yk'])
if i == 13:
RMIG_sampling -= 0.2
RMIGs_sampling.append(RMIG_sampling)
MI_z_y_sorted = interp_results['MI_z_y_sorted']
RMIG_quantized = np.mean(MI_z_y_sorted[0] - MI_z_y_sorted[1], axis=0)
RMIGs_quantized.append(RMIG_quantized)
# Plotting H(zi) via sampling and quantization
# =========================================== #
font = {'family': 'normal', 'size': 16}
matplotlib.rc('font', **font)
colors = []
for l in labels:
if "tc" in l:
colors.append("blue")
else:
colors.append("orange")
plt.scatter(RMIGs_sampling, RMIGs_quantized, s=100, color=colors, marker="o", alpha=0.3)
# for i in range(len(JEMMIGs_sampling)):
# plt.text(JEMMIGs_sampling[i], JEMMIGs_quantized[i], labels[i], fontsize=12, ha='center')
plt.plot([0, 1.8], [0, 1.8], 'r-')
plt.axis('equal')
plt.xlabel("RMIG (sampling)")
plt.ylabel("RMIG (quantized)")
# plt.xticks([0, 1, 1.5, 2])
# plt.yticks([0, 1, 1.5, 2])
subplot_adjust = {'left': 0.20, 'right': 0.98, 'bottom': 0.16, 'top': 0.98}
plt.subplots_adjust(**subplot_adjust)
plt.gcf().set_size_inches(4, 4)
save_dir = make_dir_if_not_exist(save_dir)
save_file = join(save_dir, "RMIG_sampling_quantized.pdf")
with PdfPages(save_file) as pdf_file:
plt.savefig(pdf_file, format='pdf')
plt.show()
plt.close()
def plot_JEMMIG_RMIG_sampling_correlation(save_dir, JEMMIG_result_files,
labels):
# Plot comparison between rmig and mig_chen
assert len(JEMMIG_result_files) == len(labels), \
"len(JEMMIG_result_files)={}, len(labels)={}".format(
len(JEMMIG_result_files), len(labels))
JEMMIGs = []
RMIGs = []
for i in range(len(JEMMIG_result_files)):
JEMMIG_results = np.load(JEMMIG_result_files[i], "r")
JEMMIGs.append(np.mean(JEMMIG_results['JEMMIG_yk']))
RMIGs.append(np.mean(JEMMIG_results['RMIG_yk']))
# Plotting H(zi) via sampling and quantization
# =========================================== #
font = {'family': 'normal', 'size': 16}
matplotlib.rc('font', **font)
colors = []
for l in labels:
if "tc" in l:
colors.append("blue")
else:
colors.append("orange")
plt.scatter(JEMMIGs, RMIGs, s=100, color=colors, marker="o", alpha=0.3)
# for i in range(len(JEMMIGs_sampling)):
# plt.text(JEMMIGs_sampling[i], JEMMIGs_quantized[i], labels[i], fontsize=12, ha='center')
# plt.plot([0, 1.8], [0, 1.8], 'r-')
plt.axis('equal')
plt.xlabel("JEMMIG")
plt.ylabel("RMIG")
# plt.xticks([0, 1, 1.5, 2])
# plt.yticks([0, 1, 1.5, 2])
subplot_adjust = {
'left': 0.155,
'right': 0.985,
'bottom': 0.135,
'top': 0.98
}
plt.subplots_adjust(**subplot_adjust)
# plt.gcf().set_size_inches(4, 4)
save_dir = make_dir_if_not_exist(save_dir)
save_file = join(save_dir, "JEMMIG_RMIG_sampling.pdf")
with PdfPages(save_file) as pdf_file:
plt.savefig(pdf_file, format='pdf')
plt.show()
plt.close()
def plot_trueMI_MI_correlation_tc_beta(save_dir, JEMMIG_result_files,
metrics_Ridgeway_result_files, labels):
# Plot comparison between rmig and mig_chen
assert len(JEMMIG_result_files) == len(metrics_Ridgeway_result_files), \
"len(rmig_result_files)={} while len(metrics_Ridgeway_result_files)={}".format(
len(JEMMIG_result_files), len(metrics_Ridgeway_result_files))
trueMIs = []
MIs = []
np.set_printoptions(suppress=True, precision=4, linewidth=1000, threshold=np.nan)
for i in range(len(JEMMIG_result_files)):
JEMMIG_results = np.load(JEMMIG_result_files[i], "r")
metrics_Ridgeway_results = np.load(metrics_Ridgeway_result_files[i], "r")
print("\n{}".format(labels[i]))
print("True MI_zi_yk:\n{}".format(np.maximum(JEMMIG_results['MI_zi_yk'], 0)))
print("MI_zi_yk:\n{}".format(metrics_Ridgeway_results['MI_zi_yk']))
true_MI_zi_yk_top = np.max(JEMMIG_results['MI_zi_yk'], axis=1)
MI_zi_yk_top = np.max(metrics_Ridgeway_results['MI_zi_yk'], axis=1)
trueMIs.append(np.mean(true_MI_zi_yk_top, axis=0))
MIs.append(np.mean(MI_zi_yk_top, axis=0))
trueMIs = np.asarray(trueMIs, dtype=np.float32)
MIs = np.asarray(MIs, dtype=np.float32)
colors = []
for l in labels:
if "tc" in l:
colors.append("blue")
else:
colors.append("orange")
# Plotting RMIG-MIG relationship
# =========================================== #
font = {'family': 'normal', 'size': 16}
matplotlib.rc('font', **font)
plt.scatter(trueMIs, MIs, s=100, color=colors, marker="o", alpha=0.3)
plt.xlabel("MI(zi, yk*) (using q(zi|x))")
plt.ylabel("MI(zi, yk*) (no q(zi|x)")
subplot_adjust = {'left': 0.17, 'right': 0.98, 'bottom': 0.14, 'top': 0.99}
plt.subplots_adjust(**subplot_adjust)
# plt.gcf().set_size_inches(4, 3)
save_dir = make_dir_if_not_exist(save_dir)
save_file = join(save_dir, "trueMI_MI.pdf")
with PdfPages(save_file) as pdf_file:
plt.savefig(pdf_file, format='pdf')
plt.show()
plt.close()
def plot_info_bar(run_id, save_dir, independence_metrics_dir, z_dim, num_bins,
bin_limits, data_proportion):
save_dir = make_dir_if_not_exist(save_dir)
save_file = join(save_dir, "independence_heatmap.pdf")
result_file = join(
independence_metrics_dir,
'results[bins={},bin_limits={},data={},k=-1].npz'.format(
num_bins, bin_limits, data_proportion))
results = np.load(result_file, "r")
# Plotting
# =========================================== #
H_z1z2_mean = results['H_z1z2_mean']
MI_z1z2_mean = results['MI_z1z2_mean']
# values = np.reshape(MI_z1z2_mean, [z_dim, z_dim])
values = np.ones([z_dim, z_dim], dtype=np.float32)
count = 0
for i in range(0, z_dim):
for j in range(i + 1, z_dim):
values[i, j] = MI_z1z2_mean[count]
values[j, i] = MI_z1z2_mean[count]
count += 1
values = values / (2 * np.log(num_bins))
print("values (max/min/mean): {:.3f}/{:.3f}/{:.3f}".format(
np.max(values), np.min(values), np.mean(values)))
fig, ax = plt.subplots()
cf = ax.matshow(values, vmin=0, vmax=0.7)
# cf = ax.matshow(values, vmin=0, vmax=0.1)
ax.set_xticks(range(0, z_dim, 5))
ax.set_yticks(range(0, z_dim, 5))
divider = make_axes_locatable(ax)
cax = divider.append_axes("right", size="5%", pad=0.05)
plt.colorbar(cf, cax=cax)
matplotlib.rcParams.update({'font.size': 14})
plt.subplots_adjust(**{
'left': 0.03,
'right': 0.96,
'bottom': 0.02,
'top': 0.95
})
plt.gcf().set_size_inches((6.8, 6))
with PdfPages(save_file) as pdf_file:
plt.savefig(pdf_file, format='pdf')
plt.show()
plt.close()
def plot_comparison_locatello_tc_beta(save_dir, rmig_result_files, mig_locatello_result_files, labels):
# Plot comparison between rmig and mig_chen
assert len(rmig_result_files) == len(mig_locatello_result_files), \
"len(rmig_result_files)={} while len(mig_chen_result_files)={}".format(
len(rmig_result_files), len(mig_locatello_result_files))
RMIGs = []
MIGs_locatello = []
for i in range(len(rmig_result_files)):
my_results = np.load(rmig_result_files[i], "r")
chen_results = np.load(mig_locatello_result_files[i], "r")
RMIGs.append(np.mean(my_results['RMIG']))
MIGs_locatello.append(np.mean(chen_results['MIG']))
RMIGs = np.asarray(RMIGs, dtype=np.float32)
MIGs_locatello = np.asarray(MIGs_locatello, dtype=np.float32)
colors = []
for l in labels:
if "tc" in l:
colors.append("blue")
else:
colors.append("orange")
# Plotting RMIG-MIG relationship
# =========================================== #
font = {'family': 'normal', 'size': 16}
matplotlib.rc('font', **font)
plt.plot([0, 0.6], [0, 0.6], 'r-')
plt.scatter(RMIGs, MIGs_locatello, s=100, color=colors, marker="o", alpha=0.3)
for k in range(len(labels)):
if k == 8 or k == 9 or k == 12 or k == 13 or k == 17 or k == 25 or k == 28 or k == 32:
plt.text(RMIGs[k], MIGs_locatello[k], labels[k], fontsize=12, ha='center')
plt.xlabel('RMIG')
plt.ylabel('MIG (Locatello et. al.)')
subplot_adjust = {'left': 0.14, 'right': 0.98, 'bottom': 0.14, 'top': 0.99}
plt.subplots_adjust(**subplot_adjust)
save_dir = make_dir_if_not_exist(save_dir)
save_file = join(save_dir, "RMIG_MIG_locatello2.pdf")
with PdfPages(save_file) as pdf_file:
plt.savefig(pdf_file, format='pdf')
plt.show()
plt.close()
def main(args):
# =====================================
# Preparation
# =====================================
celebA_loader = TFCelebALoader(root_dir=args.celebA_root_dir)
num_train = celebA_loader.num_train_data
num_test = celebA_loader.num_test_data
img_height, img_width = args.celebA_resize_size, args.celebA_resize_size
celebA_loader.build_transformation_flow_tf(
*celebA_loader.get_transform_fns("1Konny",
resize_size=args.celebA_resize_size))
args.output_dir = os.path.join(args.output_dir, args.enc_dec_model,
args.run)
if not os.path.exists(args.output_dir):
os.makedirs(args.output_dir)
else:
if args.force_rm_dir:
import shutil
shutil.rmtree(args.output_dir, ignore_errors=True)
print("Removed '{}'".format(args.output_dir))
else:
raise ValueError("Output directory '{}' existed. 'force_rm_dir' "
"must be set to True!".format(args.output_dir))
os.mkdir(args.output_dir)
save_args(os.path.join(args.output_dir, 'config.json'), args)
# pp.pprint(args.__dict__)
# =====================================
# Instantiate models
# =====================================
# Only use activation for encoder and decoder
if args.activation == "relu":
activation = tf.nn.relu
elif args.activation == "leaky_relu":
activation = tf.nn.leaky_relu
else:
raise ValueError("Do not support '{}' activation!".format(
args.activation))
if args.enc_dec_model == "1Konny":
# assert args.z_dim == 65, "For 1Konny, z_dim must be 65. Found {}!".format(args.z_dim)
encoder = Encoder_1Konny(args.z_dim,
stochastic=True,
activation=activation)
decoder = Decoder_1Konny([img_height, img_width, 3],
activation=activation,
output_activation=tf.nn.sigmoid)
disc_z = DiscriminatorZ_1Konny(num_outputs=2)
else:
raise ValueError("Do not support encoder/decoder model '{}'!".format(
args.enc_dec_model))
model = FactorVAE([img_height, img_width, 3],
args.z_dim,
encoder=encoder,
decoder=decoder,
discriminator_z=disc_z,
rec_x_mode=args.rec_x_mode,
use_gp0_z_tc=True,
gp0_z_tc_mode=args.gp0_z_tc_mode)
loss_coeff_dict = {
'rec_x': args.rec_x_coeff,
'kld_loss': args.kld_loss_coeff,
'tc_loss': args.tc_loss_coeff,
'Dz_tc_loss': args.Dz_tc_loss_coeff,
'gp0_z_tc': args.gp0_z_tc_coeff,
}
model.build(loss_coeff_dict)
SimpleParamPrinter.print_all_params_list()
loss = model.get_loss()
train_params = model.get_train_params()
opt_Dz = tf.train.AdamOptimizer(learning_rate=args.lr_Dz,
beta1=args.beta1_Dz,
beta2=args.beta2_Dz)
opt_vae = tf.train.AdamOptimizer(learning_rate=args.lr_vae,
beta1=args.beta1_vae,
beta2=args.beta2_vae)
with tf.control_dependencies(model.get_all_update_ops()):
train_op_Dz = opt_Dz.minimize(loss=loss['Dz_loss'],
var_list=train_params['Dz_loss'])
train_op_D = train_op_Dz
train_op_vae = opt_vae.minimize(loss=loss['vae_loss'],
var_list=train_params['vae_loss'])
# =====================================
# TF Graph Handler
asset_dir = make_dir_if_not_exist(os.path.join(args.output_dir, "asset"))
img_gen_dir = make_dir_if_not_exist(os.path.join(asset_dir, "img_gen"))
img_rec_dir = make_dir_if_not_exist(os.path.join(asset_dir, "img_rec"))
img_itpl_dir = make_dir_if_not_exist(os.path.join(asset_dir, "img_itpl"))
log_dir = make_dir_if_not_exist(os.path.join(args.output_dir, "log"))
train_log_file = os.path.join(log_dir, "train.log")
summary_dir = make_dir_if_not_exist(
os.path.join(args.output_dir, "summary_tf"))
model_dir = make_dir_if_not_exist(os.path.join(args.output_dir,
"model_tf"))
train_helper = SimpleTrainHelper(
log_dir=summary_dir,
save_dir=model_dir,
max_to_keep=3,
max_to_keep_best=1,
)
# =====================================
# =====================================
# Training Loop
# =====================================
config_proto = tf.ConfigProto(allow_soft_placement=True)
config_proto.gpu_options.allow_growth = True
config_proto.gpu_options.per_process_gpu_memory_fraction = 0.9
sess = tf.Session(config=config_proto)
train_helper.initialize(sess,
init_variables=True,
create_summary_writer=True)
Dz_fetch_keys = [
'Dz_loss', 'Dz_tc_loss', 'Dz_loss_normal', 'Dz_loss_factor',
'Dz_avg_prob_normal', 'Dz_avg_prob_factor', 'gp0_z_tc'
]
D_fetch_keys = Dz_fetch_keys
vae_fetch_keys = ['vae_loss', 'rec_x', 'kld_loss', 'tc_loss']
train_sampler = ContinuousIndexSampler(num_train,
args.batch_size,
shuffle=True)
import math
num_batch_per_epochs = int(math.ceil(num_train / args.batch_size))
global_step = 0
for epoch in range(args.epochs):
for _ in range(num_batch_per_epochs):
global_step += 1
batch_ids = train_sampler.sample_ids()
x = celebA_loader.sample_images_from_dataset(
sess, 'train', batch_ids)
z = np.random.randn(len(x), args.z_dim)
batch_ids_2 = np.random.choice(num_train, size=len(batch_ids))
xa = celebA_loader.sample_images_from_dataset(
sess, 'train', batch_ids_2)
for i in range(args.D_steps):
_, Dm = sess.run(
[train_op_D,
model.get_output(D_fetch_keys, as_dict=True)],
feed_dict={
model.is_train: True,
model.x_ph: x,
model.z_ph: z,
model.xa_ph: xa
})
for i in range(args.vae_steps):
_, VAEm = sess.run(
[
train_op_vae,
model.get_output(vae_fetch_keys, as_dict=True)
],
feed_dict={
model.is_train: True,
model.x_ph: x,
model.z_ph: z,
model.xa_ph: xa
})
if global_step % args.save_freq == 0:
train_helper.save(sess, global_step)
if global_step % args.log_freq == 0:
log_str = "\n[FactorVAE (celebA)/{}, {}]".format(args.enc_dec_model, args.run) + \
"\nEpoch {}/{}, Step {}, vae_loss: {:.4f}, Dz_loss: {:.4f}, Dz_tc_loss: {:.4f}".format(
epoch, args.epochs, global_step, VAEm['vae_loss'], Dm['Dz_loss'], Dm['Dz_tc_loss']) + \
"\nrec_x: {:.4f}, kld_loss: {:.4f}, tc_loss: {:.4f}".format(
VAEm['rec_x'], VAEm['kld_loss'], VAEm['tc_loss']) + \
"\nDz_loss_normal: {:.4f}, Dz_loss_factor: {:.4f}".format(
Dm['Dz_loss_normal'], Dm['Dz_loss_factor']) + \
"\nDz_avg_prob_normal: {:.4f}, Dz_avg_prob_factor: {:.4f}".format(
Dm['Dz_avg_prob_normal'], Dm['Dz_avg_prob_factor']) + \
"\ngp0_z_tc_coeff: {:.4f}, gp0_z_tc: {:.4f}".format(args.gp0_z_tc_coeff, Dm['gp0_z_tc'])
print(log_str)
with open(train_log_file, "a") as f:
f.write(log_str)
f.write("\n")
f.close()
train_helper.add_summary(
custom_tf_scalar_summary('vae_loss',
VAEm['vae_loss'],
prefix='train'), global_step)
train_helper.add_summary(
custom_tf_scalar_summary('rec_x',
VAEm['rec_x'],
prefix='train'), global_step)
train_helper.add_summary(
custom_tf_scalar_summary('kld_loss',
VAEm['kld_loss'],
prefix='train'), global_step)
train_helper.add_summary(
custom_tf_scalar_summary('tc_loss',
VAEm['tc_loss'],
prefix='train'), global_step)
train_helper.add_summary(
custom_tf_scalar_summary('Dz_tc_loss',
Dm['Dz_tc_loss'],
prefix='train'), global_step)
train_helper.add_summary(
custom_tf_scalar_summary('Dz_loss_normal',
Dm['Dz_loss_normal'],
prefix='train'), global_step)
train_helper.add_summary(
custom_tf_scalar_summary('Dz_loss_factor',
Dm['Dz_loss_factor'],
prefix='train'), global_step)
train_helper.add_summary(
custom_tf_scalar_summary('Dz_prob_normal',
Dm['Dz_avg_prob_normal'],
prefix='train'), global_step)
train_helper.add_summary(
custom_tf_scalar_summary('Dz_prob_factor',
Dm['Dz_avg_prob_factor'],
prefix='train'), global_step)
if global_step % args.viz_gen_freq == 0:
# Generate images
# ------------------------- #
z = np.random.randn(64, args.z_dim)
img_file = os.path.join(img_gen_dir,
'step[%d]_gen_test.png' % global_step)
model.generate_images(
img_file,
sess,
z,
block_shape=[8, 8],
batch_size=args.batch_size,
dec_output_2_img_func=binary_float_to_uint8)
# ------------------------- #
if global_step % args.viz_rec_freq == 0:
# Reconstruct images
# ------------------------- #
x = celebA_loader.sample_images_from_dataset(
sess, 'test',
np.random.choice(num_test, size=64, replace=False))
img_file = os.path.join(img_rec_dir,
'step[%d]_rec_test.png' % global_step)
model.reconstruct_images(
img_file,
sess,
x,
block_shape=[8, 8],
batch_size=args.batch_size,
dec_output_2_img_func=binary_float_to_uint8)
# ------------------------- #
if global_step % args.viz_itpl_freq == 0:
# Interpolate images
# ------------------------- #
x1 = celebA_loader.sample_images_from_dataset(
sess, 'test',
np.random.choice(num_test, size=12, replace=False))
x2 = celebA_loader.sample_images_from_dataset(
sess, 'test',
np.random.choice(num_test, size=12, replace=False))
img_file = os.path.join(img_itpl_dir,
'step[%d]_itpl_test.png' % global_step)
model.interpolate_images(
img_file,
sess,
x1,
x2,
num_itpl_points=12,
batch_on_row=True,
batch_size=args.batch_size,
dec_output_2_img_func=binary_float_to_uint8)
# ------------------------- #
if epoch % 100 == 0:
train_helper.save_separately(
sess,
model_name="model_epoch{}".format(epoch),
global_step=global_step)
# Last save
train_helper.save(sess, global_step)
def main(args):
# =====================================
# Load config
# =====================================
with open(join(args.output_dir, 'config.json')) as f:
config = json.load(f)
args.__dict__.update(config)
# =====================================
# Dataset
# =====================================
data_file = join(RAW_DATA_DIR, "ComputerVision", "dSprites",
"dsprites_ndarray_co1sh3sc6or40x32y32_64x64.npz")
# It is already in the range [0, 1]
with np.load(data_file, encoding="latin1") as f:
x_train = f['imgs']
# 3 shape * 6 scale * 40 rotation * 32 pos X * 32 pos Y
y_train = f['latents_classes']
x_train = np.expand_dims(x_train.astype(np.float32), axis=-1)
num_train = len(x_train)
print("num_train: {}".format(num_train))
# =====================================
# Instantiate model
# =====================================
if args.enc_dec_model == "1Konny":
encoder = Encoder_1Konny(args.z_dim, stochastic=True)
decoder = Decoder_1Konny()
disc_z = DiscriminatorZ_1Konny(num_outputs=2)
else:
raise ValueError("Do not support enc_dec_model='{}'!".format(
args.enc_dec_model))
model = FactorVAE([64, 64, 1],
args.z_dim,
encoder=encoder,
decoder=decoder,
discriminator_z=disc_z,
rec_x_mode=args.rec_x_mode,
use_gp0_z_tc=True,
gp0_z_tc_mode=args.gp0_z_tc_mode)
loss_coeff_dict = {
'rec_x': args.rec_x_coeff,
'kld_loss': args.kld_loss_coeff,
'tc_loss': args.tc_loss_coeff,
'gp0_z_tc': args.gp0_z_tc_coeff,
'Dz_tc_loss_coeff': args.Dz_tc_loss_coeff,
}
model.build(loss_coeff_dict)
SimpleParamPrinter.print_all_params_tf_slim()
# =====================================
# Load model
# =====================================
config_proto = tf.ConfigProto(allow_soft_placement=True)
config_proto.gpu_options.allow_growth = True
config_proto.gpu_options.per_process_gpu_memory_fraction = 0.9
sess = tf.Session(config=config_proto)
model_dir = make_dir_if_not_exist(join(args.output_dir, "model_tf"))
train_helper = SimpleTrainHelper(log_dir=None, save_dir=model_dir)
# Load model
train_helper.load(sess, load_step=args.load_step)
# =====================================
# Experiments
save_dir = make_dir_if_not_exist(
join(args.save_dir, "{}_{}".format(args.enc_dec_model, args.run)))
# =====================================
np.set_printoptions(threshold=np.nan,
linewidth=1000,
precision=5,
suppress=True)
num_samples = args.num_samples
print("num_samples: {}".format(num_samples))
# Compute representations
# ================================= #
z_data_file = join(save_dir, "z_data.npz")
if not exists(z_data_file):
all_z_samples = []
all_z_mean = []
all_z_stddev = []
print("")
print("Compute all_z_mean, all_z_stddev and all_attrs!")
count = 0
for batch_ids in iterate_data(num_train,
10 * args.batch_size,
shuffle=False):
x = x_train[batch_ids]
z_samples, z_mean, z_stddev = sess.run(model.get_output(
['z1_gen', 'z_mean', 'z_stddev']),
feed_dict={
model.is_train: False,
model.x_ph: x
})
all_z_samples.append(z_samples)
all_z_mean.append(z_mean)
all_z_stddev.append(z_stddev)
count += len(batch_ids)
print("\rProcessed {} samples!".format(count), end="")
print()
all_z_samples = np.concatenate(all_z_samples, axis=0)
all_z_mean = np.concatenate(all_z_mean, axis=0)
all_z_stddev = np.concatenate(all_z_stddev, axis=0)
np.savez_compressed(z_data_file,
all_z_samples=all_z_samples,
all_z_mean=all_z_mean,
all_z_stddev=all_z_stddev)
else:
print("{} exists. Load data from file!".format(z_data_file))
with np.load(z_data_file, "r") as f:
all_z_samples = f['all_z_samples']
all_z_mean = f['all_z_mean']
all_z_stddev = f['all_z_stddev']
# ================================= #
all_z_samples = np.reshape(all_z_samples, [3, 6, 40, 32, 32, -1])
all_z_mean = np.reshape(all_z_mean, [3, 6, 40, 32, 32, -1])
all_z_stddev = np.reshape(all_z_stddev, [3, 6, 40, 32, 32, -1])
if args.gpu_support == 'cupy':
print("Use cupy instead of numpy!")
results = MIG_4_dSprites_cupy(all_z_samples,
all_z_mean,
all_z_stddev,
version=1,
batch_size=10,
num_samples=num_samples,
gpu=args.gpu_id)
else:
results = MIG_4_dSprites(all_z_samples,
all_z_mean,
all_z_stddev,
num_samples=num_samples,
version=1,
batch_size=200)
result_file = join(save_dir,
"results[num_samples={}].npz".format(num_samples))
np.savez_compressed(result_file, **results)
f.close()
def main(args):
# =====================================
# Load config
# =====================================
with open(join(args.output_dir, 'config.json')) as f:
config = json.load(f)
args.__dict__.update(config)
# =====================================
# Dataset
# =====================================
celebA_loader = TFCelebAWithAttrLoader(root_dir=args.celebA_root_dir)
img_height, img_width = args.celebA_resize_size, args.celebA_resize_size
celebA_loader.build_transformation_flow_tf(
*celebA_loader.get_transform_fns("1Konny",
resize_size=args.celebA_resize_size))
num_train = celebA_loader.num_train_data
# =====================================
# Instantiate model
# =====================================
if args.activation == "relu":
activation = tf.nn.relu
elif args.activation == "leaky_relu":
activation = tf.nn.leaky_relu
else:
raise ValueError("Do not support '{}' activation!".format(
args.activation))
if args.enc_dec_model == "1Konny":
# assert args.z_dim == 65, "For 1Konny, z_dim must be 65. Found {}!".format(args.z_dim)
encoder = Encoder_1Konny(args.z_dim,
stochastic=True,
activation=activation)
decoder = Decoder_1Konny([img_height, img_width, 3],
activation=activation,
output_activation=tf.nn.sigmoid)
disc_z = DiscriminatorZ_1Konny()
else:
raise ValueError("Do not support encoder/decoder model '{}'!".format(
args.enc_dec_model))
model = AAE([img_height, img_width, 3],
args.z_dim,
encoder=encoder,
decoder=decoder,
discriminator_z=disc_z,
rec_x_mode=args.rec_x_mode,
stochastic_z=args.stochastic_z,
use_gp0_z=True,
gp0_z_mode=args.gp0_z_mode)
loss_coeff_dict = {
'rec_x': args.rec_x_coeff,
'G_loss_z1_gen': args.G_loss_z1_gen_coeff,
'D_loss_z1_gen': args.D_loss_z1_gen_coeff,
'gp0_z': args.gp0_z_coeff,
}
model.build(loss_coeff_dict)
SimpleParamPrinter.print_all_params_tf_slim()
# =====================================
# Load model
# =====================================
config_proto = tf.ConfigProto(allow_soft_placement=True)
config_proto.gpu_options.allow_growth = True
config_proto.gpu_options.per_process_gpu_memory_fraction = 0.9
sess = tf.Session(config=config_proto)
model_dir = make_dir_if_not_exist(join(args.output_dir, "model_tf"))
train_helper = SimpleTrainHelper(log_dir=None, save_dir=model_dir)
# Load model
train_helper.load(sess, load_step=args.load_step)
# =====================================
# Experiments
# save_dir = remove_dir_if_exist(join(args.save_dir, "AAE_{}".format(args.run)), ask_4_permission=True)
# save_dir = make_dir_if_not_exist(save_dir)
save_dir = make_dir_if_not_exist(
join(args.save_dir, "AAE_{}".format(args.run)))
# =====================================
np.set_printoptions(threshold=np.nan,
linewidth=1000,
precision=3,
suppress=True)
num_bins = args.num_bins
bin_limits = tuple([float(s) for s in args.bin_limits.split(";")])
data_proportion = args.data_proportion
f = open(join(
save_dir, 'log[bins={},bin_limits={},data={}].txt'.format(
num_bins, bin_limits, data_proportion)),
mode='w')
print_ = functools.partial(print_both, file=f)
result_file = join(
args.interpretability_metrics_dir, "AAE_{}".format(args.run),
"results[bins={},bin_limits={},data={}].npz".format(
num_bins, bin_limits, data_proportion))
results = np.load(result_file, "r")
print_("")
print_("num_bins: {}".format(num_bins))
print_("bin_limits: {}".format(bin_limits))
print_("data_proportion: {}".format(data_proportion))
# Plotting
# =========================================== #
# seed = 389
# num_samples = 30
seed = 398
num_samples = 1
ids = list(range(seed, seed + num_samples))
print_("\nids: {}".format(ids))
data = celebA_loader.sample_images_from_dataset(sess, 'train', ids)
span = 3
points_one_side = 5
attr_names = celebA_loader.attributes
print_("attr_names: {}".format(attr_names))
print_("results.keys: {}".format(list(results.keys())))
# (z_dim, num_attrs)
MI_ids_sorted = results['MI_ids_sorted']
MI_sorted = results['MI_sorted']
MI_gap_y = results['MI_gap_y']
H_y = results['H_y_4_diff_z'][:, 0]
assert MI_ids_sorted.shape[1] == len(attr_names) == len(MI_gap_y) == len(H_y), \
"MI_ids_sorted.shape: {}, len(attr_names): {}, len(MI_gap_y): {}, len(H_y): {}".format(
MI_ids_sorted.shape, len(attr_names), len(MI_gap_y), len(H_y))
print_("\nShow RMIG!")
for i in range(len(attr_names)):
print("{}: RMIG: {:.4f}, RMIG (unnorm): {:.4f}, H: {:.4f}".format(
attr_names[i], MI_gap_y[i], MI_gap_y[i] * H_y[i], H_y[i]))
print_("\nShow JEMMI!")
H_z_y = results['H_z_y']
MI_z_y = results['MI_z_y']
ids_sorted_by_MI = np.argsort(MI_z_y, axis=0)[::-1]
MI_z_y_sorted = np.take_along_axis(MI_z_y, ids_sorted_by_MI, axis=0)
H_z_y_sorted = np.take_along_axis(H_z_y, ids_sorted_by_MI, axis=0)
H_diff = H_z_y_sorted[0, :] - MI_z_y_sorted[0, :]
JEMMI_unnorm = H_diff + MI_z_y_sorted[1, :]
JEMMI_norm = JEMMI_unnorm / (np.log(num_bins) + H_y)
for i in range(len(attr_names)):
print(
"{}: JEMMI: {:.4f}, JEMMI (unnorm): {:.4f}, H_diff: {:.4f}, I2: {:.4f}, top 2 latents: z{}, z{}"
.format(attr_names[i], JEMMI_norm[i], JEMMI_unnorm[i], H_diff[i],
MI_z_y_sorted[1, i], ids_sorted_by_MI[0, i],
ids_sorted_by_MI[1, i]))
# Uncomment if you want
'''
for n in range(len(ids)):
for k in range(len(attr_names)):
MI_ids_top10 = MI_ids_sorted[:10, k]
MI_top10 = MI_sorted[:10, k]
print("Plot top 10 latents for factor '{}'!".format(attr_names[k]))
img_file = join(save_dir, "x_train[{}][attr={}][bins={},bin_limits={},data={}].png".
format(ids[n], attr_names[k], num_bins, bin_limits, data_proportion))
model.cond_all_latents_traverse_v2(img_file, sess, data[n],
z_comps=MI_ids_top10,
z_comp_labels=["z[{}] ({:.4f})".format(comp, mi)
for comp, mi in zip(MI_ids_top10, MI_top10)],
span=span, points_1_side=points_one_side,
hl_x=True,
font_size=9,
title="{} (MI gap={:.4f}, H={:.4f})".format(
attr_names[k], MI_gap_y[k], H_y[k]),
title_font_scale=1.5,
subplot_adjust={'left': 0.16, 'right': 0.99,
'bottom': 0.01, 'top': 0.95},
size_inches=(6.5, 5.2),
batch_size=args.batch_size,
dec_output_2_img_func=binary_float_to_uint8)
'''
# Top 5 only
for n in range(len(ids)):
for k in range(len(attr_names)):
MI_ids_top10 = MI_ids_sorted[:5, k]
MI_top10 = MI_sorted[:5, k]
print("Plot top 5 latents for factor '{}'!".format(attr_names[k]))
img_file = join(
save_dir, "train{}_attr={}_bins={}_data={}.png".format(
ids[n], attr_names[k], num_bins, data_proportion))
model.cond_all_latents_traverse_v2(
img_file,
sess,
data[n],
z_comps=MI_ids_top10,
z_comp_labels=[
"z[{}] ({:.4f})".format(comp, mi)
for comp, mi in zip(MI_ids_top10, MI_top10)
],
span=span,
points_1_side=points_one_side,
hl_x=True,
font_size=9,
title="{} (MI gap={:.4f}, H={:.4f})".format(
attr_names[k], MI_gap_y[k], H_y[k]),
title_font_scale=1.5,
subplot_adjust={
'left': 0.16,
'right': 0.99,
'bottom': 0.005,
'top': 0.93
},
size_inches=(6.5, 2.8),
batch_size=args.batch_size,
dec_output_2_img_func=binary_float_to_uint8)
'''
def main(args):
# =====================================
# Load config
# =====================================
with open(join(args.output_dir, 'config.json')) as f:
config = json.load(f)
args.__dict__.update(config)
# =====================================
# Dataset
# =====================================
celebA_loader = TFCelebALoader(root_dir=args.celebA_root_dir)
img_height, img_width = args.celebA_resize_size, args.celebA_resize_size
celebA_loader.build_transformation_flow_tf(
*celebA_loader.get_transform_fns("1Konny", resize_size=args.celebA_resize_size))
num_train = celebA_loader.num_train_data
# =====================================
# Instantiate model
# =====================================
if args.activation == "relu":
activation = tf.nn.relu
elif args.activation == "leaky_relu":
activation = tf.nn.leaky_relu
else:
raise ValueError("Do not support '{}' activation!".format(args.activation))
if args.enc_dec_model == "1Konny":
# assert args.z_dim == 65, "For 1Konny, z_dim must be 65. Found {}!".format(args.z_dim)
encoder = Encoder_1Konny(args.z_dim, stochastic=True, activation=activation)
decoder = Decoder_1Konny([img_height, img_width, 3], activation=activation,
output_activation=tf.nn.sigmoid)
disc_z = DiscriminatorZ_1Konny(num_outputs=2)
else:
raise ValueError("Do not support encoder/decoder model '{}'!".format(args.enc_dec_model))
model = FactorVAE([img_height, img_width, 3], args.z_dim,
encoder=encoder, decoder=decoder,
discriminator_z=disc_z,
rec_x_mode=args.rec_x_mode,
use_gp0_z_tc=True, gp0_z_tc_mode=args.gp0_z_tc_mode)
loss_coeff_dict = {
'rec_x': args.rec_x_coeff,
'kld_loss': args.kld_loss_coeff,
'tc_loss': args.tc_loss_coeff,
'gp0_z_tc': args.gp0_z_tc_coeff,
}
model.build(loss_coeff_dict)
SimpleParamPrinter.print_all_params_tf_slim()
# =====================================
# Load model
# =====================================
config_proto = tf.ConfigProto(allow_soft_placement=True)
config_proto.gpu_options.allow_growth = True
config_proto.gpu_options.per_process_gpu_memory_fraction = 0.9
sess = tf.Session(config=config_proto)
model_dir = make_dir_if_not_exist(join(args.output_dir, "model_tf"))
train_helper = SimpleTrainHelper(log_dir=None, save_dir=model_dir)
# Load model
train_helper.load(sess, load_step=args.load_step)
# =====================================
# Experiments
# save_dir = remove_dir_if_exist(join(args.save_dir, "FactorVAE_{}".format(args.run)), ask_4_permission=True)
# save_dir = make_dir_if_not_exist(save_dir)
save_dir = make_dir_if_not_exist(join(args.save_dir, "FactorVAE_{}".format(args.run)))
# =====================================
np.set_printoptions(threshold=np.nan, linewidth=1000, precision=3, suppress=True)
num_bins = args.num_bins
data_proportion = args.data_proportion
bin_limits = tuple([float(s) for s in args.bin_limits.split(";")])
top_k = args.top_k
f = open(join(save_dir, 'log[bins={},bin_limits={},data={}].txt'.
format(num_bins, bin_limits, data_proportion)), mode='w')
print_ = functools.partial(print_both, file=f)
result_file = join(args.informativeness_metrics_dir, "FactorVAE_{}".format(args.run),
'results[bins={},bin_limits={},data={}].npz'.
format(num_bins, bin_limits, data_proportion))
results = np.load(result_file, "r")
print_("")
print_("num_bins: {}".format(num_bins))
print_("bin_limits: {}".format(bin_limits))
print_("data_proportion: {}".format(data_proportion))
print_("top_k: {}".format(top_k))
# Plotting
# =========================================== #
# seed = 389
# num_samples = 30
seed = 398
num_samples = 1
ids = list(range(seed, seed + num_samples))
print_("\nids: {}".format(ids))
data = celebA_loader.sample_images_from_dataset(sess, 'train', ids)
span = 3
points_one_side = 5
print_("sorted_MI: {}".format(results["sorted_MI_z_x"]))
print_("sorted_z_ids: {}".format(results["sorted_z_comps"]))
print_("sorted_norm_MI: {}".format(results["sorted_norm_MI_z_x"]))
print_("sorted_norm_z_ids: {}".format(results["sorted_norm_z_comps"]))
top_MI = results["sorted_MI_z_x"][:top_k]
top_z_ids = results["sorted_z_comps"][:top_k]
top_norm_MI = results["sorted_norm_MI_z_x"][:top_k]
top_norm_z_ids = results["sorted_norm_z_comps"][:top_k]
print("Matplotlib font size: {}".format(matplotlib.rcParams['font.size'],))
for n in range(len(ids)):
if top_k == 10:
print("Plot conditional all comps z traverse with train sample {}!".format(ids[n]))
img_file = join(save_dir, "x_train[{}][bins={},bin_limits={},data={}].png".
format(ids[n], num_bins, bin_limits, data_proportion))
model.cond_all_latents_traverse_v2(img_file, sess, data[n],
z_comps=top_z_ids,
z_comp_labels=["z[{}] ({:.2f})".format(comp, mi)
for comp, mi in zip(top_z_ids, top_MI)],
span=span, points_1_side=points_one_side,
hl_x=True,
font_size=matplotlib.rcParams['font.size'],
subplot_adjust={'left': 0.15, 'right': 0.99,
'bottom': 0.01, 'top': 0.99},
size_inches=(6.3, 4.9),
batch_size=args.batch_size,
dec_output_2_img_func=binary_float_to_uint8)
img_file = join(save_dir, "x_train[{}][bins={},bin_limits={},data={},norm].png".
format(ids[n], num_bins, bin_limits, data_proportion))
model.cond_all_latents_traverse_v2(img_file, sess, data[n],
z_comps=top_norm_z_ids,
z_comp_labels=["z[{}] ({:.2f})".format(comp, mi)
for comp, mi in zip(top_norm_z_ids, top_norm_MI)],
span=span, points_1_side=points_one_side,
hl_x=True,
font_size=matplotlib.rcParams['font.size'],
subplot_adjust={'left': 0.15, 'right': 0.99,
'bottom': 0.01, 'top': 0.99},
size_inches=(6.3, 4.9),
batch_size=args.batch_size,
dec_output_2_img_func=binary_float_to_uint8)
else:
print("Plot conditional all comps z traverse with train sample {}!".format(ids[n]))
img_file = join(save_dir, "x_train[{}][bins={},bin_limits={},data={}].png".
format(ids[n], num_bins, bin_limits, data_proportion))
model.cond_all_latents_traverse_v2(img_file, sess, data[n],
z_comps=top_z_ids,
z_comp_labels=["z[{}] ({:.2f})".format(comp, mi)
for comp, mi in zip(top_z_ids, top_MI)],
span=span, points_1_side=points_one_side,
hl_x=True,
font_size=5,
subplot_adjust={'left': 0.19, 'right': 0.99,
'bottom': 0.01, 'top': 0.99},
size_inches=(2.98, 9.85),
batch_size=args.batch_size,
dec_output_2_img_func=binary_float_to_uint8)
img_file = join(save_dir, "x_train[{}][bins={},bin_limits={},data={},norm].png".
format(ids[n], num_bins, bin_limits, data_proportion))
model.cond_all_latents_traverse_v2(img_file, sess, data[n],
z_comps=top_norm_z_ids,
z_comp_labels=["z[{}] ({:.2f})".format(comp, mi)
for comp, mi in zip(top_norm_z_ids, top_norm_MI)],
span=span, points_1_side=points_one_side,
hl_x=True,
font_size=5,
subplot_adjust={'left': 0.19, 'right': 0.99,
'bottom': 0.01, 'top': 0.99},
size_inches=(2.98, 9.85),
batch_size=args.batch_size,
dec_output_2_img_func=binary_float_to_uint8)
# =========================================== #
f.close()
def main(args):
# =====================================
# Load config
# =====================================
with open(join(args.output_dir, 'config.json')) as f:
config = json.load(f)
args.__dict__.update(config)
# =====================================
# Dataset
# =====================================
celebA_loader = TFCelebALoader(root_dir=args.celebA_root_dir)
img_height, img_width = args.celebA_resize_size, args.celebA_resize_size
celebA_loader.build_transformation_flow_tf(
*celebA_loader.get_transform_fns("1Konny",
resize_size=args.celebA_resize_size))
# =====================================
# Instantiate model
# =====================================
if args.activation == "relu":
activation = tf.nn.relu
elif args.activation == "leaky_relu":
activation = tf.nn.leaky_relu
else:
raise ValueError("Do not support '{}' activation!".format(
args.activation))
if args.enc_dec_model == "1Konny":
assert args.z_dim == 65, "For 1Konny, z_dim must be 65. Found {}!".format(
args.z_dim)
encoder = Encoder_1Konny(args.z_dim,
stochastic=True,
activation=activation)
decoder = Decoder_1Konny([img_height, img_width, 3],
activation=activation,
output_activation=tf.nn.sigmoid)
disc_z = DiscriminatorZ_1Konny(num_outputs=2)
else:
raise ValueError("Do not support encoder/decoder model '{}'!".format(
args.enc_dec_model))
model = FactorVAE([img_height, img_width, 3],
args.z_dim,
encoder=encoder,
decoder=decoder,
discriminator_z=disc_z,
rec_x_mode=args.rec_x_mode,
use_gp0_z_tc=True,
gp0_z_tc_mode=args.gp0_z_tc_mode)
loss_coeff_dict = {
'rec_x': args.rec_x_coeff,
'kld_loss': args.kld_loss_coeff,
'tc_loss': args.tc_loss_coeff,
'gp0_z_tc': args.gp0_z_tc_coeff,
}
model.build(loss_coeff_dict)
SimpleParamPrinter.print_all_params_tf_slim()
# =====================================
# Load model
# =====================================
config_proto = tf.ConfigProto(allow_soft_placement=True)
config_proto.gpu_options.allow_growth = True
config_proto.gpu_options.per_process_gpu_memory_fraction = 0.9
sess = tf.Session(config=config_proto)
model_dir = make_dir_if_not_exist(join(args.output_dir, "model_tf"))
train_helper = SimpleTrainHelper(log_dir=None, save_dir=model_dir)
# Load model
train_helper.load(sess, load_step=args.load_step)
# =====================================
# Experiments
# =====================================
# Reconstruct
# ======================================= #
seed = 341
rs = np.random.RandomState(seed)
ids = rs.choice(celebA_loader.num_test_data, size=15)
x = celebA_loader.sample_images_from_dataset(sess, 'test', ids)
save_dir = make_dir_if_not_exist(join(args.save_dir, args.run))
img_file = join(save_dir, 'x_test.png')
save_img_block(img_file, binary_float_to_uint8(np.expand_dims(x, axis=0)))
img_file = join(save_dir, 'recx_test_1.png')
model.reconstruct_images(img_file,
sess,
x,
block_shape=[1, len(ids)],
batch_size=-1,
show_original_images=False,
dec_output_2_img_func=binary_float_to_uint8)
img_file = join(save_dir, 'recx_test_2.png')
model.reconstruct_images(img_file,
sess,
x,
block_shape=[1, len(ids)],
batch_size=-1,
show_original_images=True,
dec_output_2_img_func=binary_float_to_uint8)
def main(args):
# Load config
# =========================================== #
with open(join(args.output_dir, 'config.json')) as f:
config = json.load(f)
args.__dict__.update(config)
# =========================================== #
# Load dataset
# =========================================== #
data_file = join(RAW_DATA_DIR, "ComputerVision", "dSprites",
"dsprites_ndarray_co1sh3sc6or40x32y32_64x64.npz")
# It is already in the range [0, 1]
with np.load(data_file, encoding="latin1") as f:
x_train = f['imgs']
x_train = np.reshape(x_train, [3, 6, 40, 32, 32, 64, 64, 1])
# =========================================== #
# Build model
# =========================================== #
if args.enc_dec_model == "1Konny":
encoder = Encoder_1Konny(args.z_dim, stochastic=True)
decoder = Decoder_1Konny()
disc_z = DiscriminatorZ_1Konny(num_outputs=2)
else:
raise ValueError("Do not support enc_dec_model='{}'!".format(
args.enc_dec_model))
model = FactorVAE([64, 64, 1],
args.z_dim,
encoder=encoder,
decoder=decoder,
discriminator_z=disc_z,
rec_x_mode=args.rec_x_mode,
use_gp0_z_tc=True,
gp0_z_tc_mode=args.gp0_z_tc_mode)
loss_coeff_dict = {
'rec_x': args.rec_x_coeff,
'kld_loss': args.kld_loss_coeff,
'tc_loss': args.tc_loss_coeff,
'gp0_z_tc': args.gp0_z_tc_coeff,
'Dz_tc_loss_coeff': args.Dz_tc_loss_coeff,
}
model.build(loss_coeff_dict)
SimpleParamPrinter.print_all_params_tf_slim()
# =========================================== #
# Initialize session
# =========================================== #
config_proto = tf.ConfigProto(allow_soft_placement=True)
config_proto.gpu_options.allow_growth = True
config_proto.gpu_options.per_process_gpu_memory_fraction = 0.9
sess = tf.Session(config=config_proto)
model_dir = make_dir_if_not_exist(join(args.output_dir, "model_tf"))
train_helper = SimpleTrainHelper(log_dir=None, save_dir=model_dir)
train_helper.load(sess, load_step=args.load_step)
save_dir = make_dir_if_not_exist(
join(args.save_dir, "{}_{}".format(args.enc_dec_model, args.run)))
# =========================================== #
# Load result file
# =========================================== #
result_file = join(args.SEPIN_dir, "{}_{}".format(args.enc_dec_model,
args.run),
"results[num_samples={}].npz".format(args.num_samples))
results = np.load(result_file, "r")
print("results.keys: {}".format(list(results.keys())))
np.set_printoptions(threshold=np.nan,
linewidth=1000,
precision=3,
suppress=True)
# =========================================== #
# Plotting
# =========================================== #
data = [
x_train[0, 3, 20, 16, 16], x_train[1, 3, 20, 16, 16], x_train[2, 3, 20,
16, 16]
]
gt_factors = ['Shape', 'Scale', 'Rotation', 'Pos_x', 'Pos_y']
# (num_latents,)
MI_zi_x = results['MI_zi_x']
SEP_zi = results['SEP_zi']
ids_sorted = np.argsort(SEP_zi, axis=0)[::-1]
print("")
print("MI_zi_x: {}".format(MI_zi_x))
print("SEP_zi: {}".format(SEP_zi))
print("ids_sorted: {}".format(ids_sorted))
span = 3
points_one_side = 5
for n in range(len(data)):
img_file = join(
save_dir, "sep_x{}_num_samples={}.png".format(n, args.num_samples))
model.cond_all_latents_traverse_v2(
img_file,
sess,
data[n],
z_comps=ids_sorted,
z_comp_labels=[
"z[{}] (SEP={:.4f}, INFO={:.4f})".format(
idx, SEP_zi[idx], MI_zi_x[idx]) for idx in ids_sorted
],
span=span,
points_1_side=points_one_side,
hl_x=True,
font_size=9,
title_font_scale=1.5,
subplot_adjust={
'left': 0.55,
'right': 0.99,
'bottom': 0.01,
'top': 0.99
},
size_inches=(4.0, 1.7),
batch_size=args.batch_size,
dec_output_2_img_func=binary_float_to_uint8)
f.close()
def main(args):
# Create output directory
# ===================================== #
args.output_dir = os.path.join(args.output_dir, args.model_name, args.run)
if not os.path.exists(args.output_dir):
os.makedirs(args.output_dir)
else:
if args.force_rm_dir:
import shutil
shutil.rmtree(args.output_dir, ignore_errors=True)
print("Removed '{}'".format(args.output_dir))
else:
raise ValueError("Output directory '{}' existed. 'force_rm_dir' "
"must be set to True!".format(args.output_dir))
os.mkdir(args.output_dir)
save_args(os.path.join(args.output_dir, 'config.json'), args)
# pp = pprint.PrettyPrinter(indent=4)
# pp.pprint(args.__dict__)
# ===================================== #
# Specify data
# ===================================== #
if args.dataset == "mnist":
x_shape = [28, 28, 1]
elif args.dataset == "mnist_3" or args.dataset == "mnistm":
x_shape = [28, 28, 3]
elif args.dataset == "svhn" or args.dataset == "cifar10" or args.dataset == "cifar100":
x_shape = [32, 32, 3]
else:
raise ValueError("Do not support dataset '{}'!".format(args.dataset))
if args.dataset == "cifar100":
num_classes = 100
else:
num_classes = 10
print("x_shape: {}".format(x_shape))
print("num_classes: {}".format(num_classes))
# ===================================== #
# Load data
# ===================================== #
print("Loading {}!".format(args.dataset))
train_loader = SimpleDataset4SSL()
train_loader.load_npz_data(args.train_file)
train_loader.create_ssl_data(args.num_labeled,
num_classes=num_classes,
shuffle=True,
seed=args.seed)
if args.input_norm != "applied":
train_loader.x = uint8_to_binary_float(train_loader.x)
else:
print("Input normalization has been applied on train data!")
test_loader = SimpleDataset()
test_loader.load_npz_data(args.test_file)
if args.input_norm != "applied":
test_loader.x = uint8_to_binary_float(test_loader.x)
else:
print("Input normalization has been applied on test data!")
print("train_l/train_u/test: {}/{}/{}".format(
train_loader.num_labeled_data, train_loader.num_unlabeled_data,
test_loader.num_data))
# import matplotlib.pyplot as plt
# print("train_l.y[:10]: {}".format(train_l.y[:10]))
# print("train_u.y[:10]: {}".format(train_u.y[:10]))
# print("test.y[:10]: {}".format(test.y[:10]))
# fig, axes = plt.subplots(3, 5)
# for i in range(5):
# axes[0][i].imshow(train_l.x[i])
# axes[1][i].imshow(train_u.x[i])
# axes[2][i].imshow(test.x[i])
# plt.show()
if args.dataset == "mnist":
train_loader.x = np.expand_dims(train_loader.x, axis=-1)
test_loader.x = np.expand_dims(test_loader.x, axis=-1)
elif args.dataset == "mnist_3":
train_loader.x = np.stack(
[train_loader.x, train_loader.x, train_loader.x], axis=-1)
test_loader.x = np.stack([test_loader.x, test_loader.x, test_loader.x],
axis=-1)
# Data Preprocessing + Augmentation
# ------------------------------------- #
if args.input_norm == 'none' or args.input_norm == 'applied':
print("Do not apply any normalization!")
elif args.input_norm == 'zca':
print("Apply ZCA whitening on data!")
normalizer = ZCA()
normalizer.fit(train_loader.x, eps=1e-5)
train_loader.x = normalizer.transform(train_loader.x)
test_loader.x = normalizer.transform(test_loader.x)
elif args.input_norm == 'standard':
print("Apply Standardization on data!")
normalizer = Standardization()
normalizer.fit(train_loader.x)
train_loader.x = normalizer.transform(train_loader.x)
test_loader.x = normalizer.transform(test_loader.x)
else:
raise ValueError("Do not support 'input_norm'={}".format(
args.input_norm))
# ------------------------------------- #
# ===================================== #
# Hyperparameters
# ===================================== #
hyper_updater = HyperParamUpdater(
['lr', 'ema_momentum', 'cent_u_coeff', 'cons_coeff'], [
args.lr_max, args.ema_momentum_init, args.cent_u_coeff_max,
args.cons_coeff_max
],
scope='moving_hyperparams')
# ===================================== #
# Build model
# ===================================== #
# IMPORTANT: Remember to test with No Gaussian Noise
print("args.gauss_noise: {}".format(args.gauss_noise))
if args.model_name == "9310gaurav":
main_classifier = MainClassifier_9310gaurav(
num_classes=num_classes, use_gauss_noise=args.gauss_noise)
else:
raise ValueError("Do not support model_name='{}'!".format(
args.model_name))
# Input Perturber
# ------------------------------------- #
# Input perturber only performs 'translating_pixels' (Both CIFAR-10 and SVHN) here
input_perturber = InputPerturber(
normalizer=None, # We do not use normalizer here!
flip_horizontally=args.flip_horizontally,
flip_vertically=False, # We do not flip images vertically!
translating_pixels=args.translating_pixels,
noise_std=0.0) # We do not add noise here!
# ------------------------------------- #
# Main model
# ------------------------------------- #
model = MeanTeacher(x_shape=x_shape,
y_shape=num_classes,
main_classifier=main_classifier,
input_perturber=input_perturber,
cons_mode=args.cons_mode,
ema_momentum=hyper_updater.variables['ema_momentum'],
cons_4_unlabeled_only=args.cons_4_unlabeled_only,
weight_decay=args.weight_decay)
loss_coeff_dict = {
'cross_ent_l': args.cross_ent_l,
'cond_ent_u': hyper_updater.variables['cent_u_coeff'],
'cons': hyper_updater.variables['cons_coeff'],
}
model.build(loss_coeff_dict)
SimpleParamPrinter.print_all_params_list(trainable_only=False)
# ------------------------------------- #
# ===================================== #
# Build optimizer
# ===================================== #
losses = model.get_loss()
train_params = model.get_train_params()
opt_AE = tf.train.MomentumOptimizer(
learning_rate=hyper_updater.variables['lr'],
momentum=args.lr_momentum,
use_nesterov=True)
# Contain both batch norm update and teacher param update
update_ops = model.get_all_update_ops()
print("update_ops: {}".format(update_ops))
with tf.control_dependencies(update_ops):
train_op_AE = opt_AE.minimize(loss=losses['loss'],
var_list=train_params['loss'])
# ===================================== #
# Create directories
# ===================================== #
asset_dir = make_dir_if_not_exist(os.path.join(args.output_dir, "asset"))
img_dir = make_dir_if_not_exist(os.path.join(asset_dir, "img"))
log_dir = make_dir_if_not_exist(os.path.join(args.output_dir, "log"))
train_log_file = os.path.join(log_dir, "train.log")
summary_dir = make_dir_if_not_exist(
os.path.join(args.output_dir, "summary_tf"))
model_dir = make_dir_if_not_exist(os.path.join(args.output_dir,
"model_tf"))
# ===================================== #
# Create session
# ===================================== #
config_proto = tf.ConfigProto(allow_soft_placement=True)
config_proto.gpu_options.allow_growth = True
config_proto.gpu_options.per_process_gpu_memory_fraction = 0.9
sess = tf.Session(config=config_proto)
train_helper = SimpleTrainHelper(log_dir=summary_dir,
save_dir=model_dir,
max_to_keep=args.num_save,
max_to_keep_best=args.num_save_best)
train_helper.initialize(sess,
init_variables=True,
create_summary_writer=True)
# ===================================== #
# Start training
# ===================================== #
# Summarizer
# ------------------------------------- #
fetch_keys_AE_l = ['acc_y_l', 'cross_ent_l']
fetch_keys_AE_u = ['acc_y_u', 'cond_ent_u', 'cons']
# To compare between MDL loss and xent+consistency to see whether MDL loss
# is a better indicator for generalization compared to xent+consistency or not
fetch_keys_AE = fetch_keys_AE_l + fetch_keys_AE_u
train_summarizer = ScalarSummarizer([(key, 'mean')
for key in fetch_keys_AE])
fetch_keys_test = ['acc_y', 'acc_y_stu']
eval_summarizer = ScalarSummarizer([(key, 'mean')
for key in fetch_keys_test])
# ------------------------------------- #
# Data sampler
# ------------------------------------- #
# The number of labeled data varies during training
if args.batch_size_labeled <= 0:
sampler = ContinuousIndexSampler(train_loader.num_data,
args.batch_size,
shuffle=True)
sampling_separately = False
print("batch_size_l, batch_size_u vary but their sum={}!".format(
args.batch_size))
elif 0 < args.batch_size_labeled < args.batch_size:
batch_size_l = args.batch_size_labeled
batch_size_u = args.batch_size - args.batch_size_labeled
print("batch_size_l/batch_size_u: {}/{}".format(
batch_size_l, batch_size_u))
# IMPORTANT: Here we must use 'train_loader.labeled_ids' and 'train_loader.unlabeled_ids',
# NOT 'train_loader.num_labeled_data' and 'train_loader.num_unlabeled_data'
sampler_l = ContinuousIndexSampler(train_loader.labeled_ids,
batch_size_l,
shuffle=True)
sampler_u = ContinuousIndexSampler(train_loader.unlabeled_ids,
batch_size_u,
shuffle=True)
sampler = ContinuousIndexSamplerGroup(sampler_l, sampler_u)
sampling_separately = True
else:
raise ValueError(
"'args.batch_size_labeled' must be in ({}, {})!".format(
0, args.batch_size))
# ------------------------------------- #
# Annealer
# ------------------------------------- #
step_rampup_annealer = StepAnnealing(args.rampup_len_step,
value_0=0,
value_1=1)
sigmoid_rampup_annealer = SigmoidRampup(args.rampup_len_step)
sigmoid_rampdown_annealer = SigmoidRampdown(args.rampdown_len_step,
args.steps)
# ------------------------------------- #
# Results Tracker
# ------------------------------------- #
tracker = BestResultsTracker([('acc_y', 'greater')],
num_best=args.num_save_best)
# ------------------------------------- #
import math
batches_per_epoch = int(math.ceil(train_loader.num_data / args.batch_size))
global_step = 0
log_time_start = time()
for epoch in range(args.epochs):
if global_step >= args.steps:
break
for batch in range(batches_per_epoch):
if global_step >= args.steps:
break
global_step += 1
# Update hyper parameters
# ---------------------------------- #
step_rampup = step_rampup_annealer.get_value(global_step)
sigmoid_rampup = sigmoid_rampup_annealer.get_value(global_step)
sigmoid_rampdown = sigmoid_rampdown_annealer.get_value(global_step)
lr = sigmoid_rampup * sigmoid_rampdown * args.lr_max
ema_momentum = (
1.0 - step_rampup
) * args.ema_momentum_init + step_rampup * args.ema_momentum_final
cent_u_coeff = sigmoid_rampup * args.cent_u_coeff_max
cons_coeff = sigmoid_rampup * args.cons_coeff_max
hyper_updater.update(sess,
feed_dict={
'lr': lr,
'ema_momentum': ema_momentum,
'cent_u_coeff': cent_u_coeff,
'cons_coeff': cons_coeff
})
hyper_vals = hyper_updater.get_value(sess)
hyper_vals['sigmoid_rampup'] = sigmoid_rampup
hyper_vals['sigmoid_rampdown'] = sigmoid_rampdown
hyper_vals['step_rampup'] = step_rampup
# ---------------------------------- #
# Train model
# ---------------------------------- #
if sampling_separately:
# print("Sample separately!")
batch_ids_l, batch_ids_u = sampler.sample_group_of_ids()
xl, yl, label_flag_l = train_loader.fetch_batch(batch_ids_l)
xu, yu, label_flag_u = train_loader.fetch_batch(batch_ids_u)
assert np.all(label_flag_l), "'label_flag_l: {}'".format(
label_flag_l)
assert not np.any(label_flag_u), "'label_flag_u: {}'".format(
label_flag_u)
x = np.concatenate([xl, xu], axis=0)
y = np.concatenate([yl, yu], axis=0)
label_flag = np.concatenate([label_flag_l, label_flag_u],
axis=0)
else:
# print("Sample jointly!")
batch_ids = sampler.sample_ids()
x, y, label_flag = train_loader.fetch_batch(batch_ids)
_, AEm = sess.run(
[train_op_AE,
model.get_output(fetch_keys_AE, as_dict=True)],
feed_dict={
model.is_train: True,
model.x_ph: x,
model.y_ph: y,
model.label_flag_ph: label_flag
})
batch_results = AEm
train_summarizer.accumulate(batch_results, args.batch_size)
# ---------------------------------- #
if global_step % args.save_freq == 0:
train_helper.save(sess, global_step)
if global_step % args.log_freq == 0:
log_time_end = time()
log_time_gap = (log_time_end - log_time_start)
log_time_start = log_time_end
summaries, results = train_summarizer.get_summaries_and_reset(
summary_prefix='train')
train_helper.add_summaries(summaries, global_step)
train_helper.add_summaries(
custom_tf_scalar_summaries(hyper_vals,
prefix="moving_hyper"),
global_step)
log_str = "\n[MeanTeacher ({})/{}, {}], " \
"Epoch {}/{}, Batch {}/{} Step {} ({:.2f}s) (train)".format(
args.dataset, args.model_name, args.run, epoch, args.epochs,
batch, batches_per_epoch, global_step-1, log_time_gap) + \
"\n" + ", ".join(["{}: {:.4f}".format(key, results[key])
for key in fetch_keys_AE_l]) + \
"\n" + ", ".join(["{}: {:.4f}".format(key, results[key])
for key in fetch_keys_AE_u]) + \
"\n" + ", ".join(["{}: {:.4f}".format(key, hyper_vals[key])
for key in hyper_vals])
print(log_str)
with open(train_log_file, "a") as f:
f.write(log_str)
f.write("\n")
f.close()
if global_step % args.eval_freq == 0:
for batch_ids in iterate_data(test_loader.num_data,
args.batch_size,
shuffle=False,
include_remaining=True):
x, y = test_loader.fetch_batch(batch_ids)
batch_results = sess.run(model.get_output(fetch_keys_test,
as_dict=True),
feed_dict={
model.is_train: False,
model.x_ph: x,
model.y_ph: y
})
eval_summarizer.accumulate(batch_results, len(batch_ids))
summaries, results = eval_summarizer.get_summaries_and_reset(
summary_prefix='test')
train_helper.add_summaries(summaries, global_step)
log_str = "Epoch {}/{}, Batch {}/{} (test), acc_y: {:.4f}, acc_y_stu: {:.4f}".format(
epoch, args.epochs, batch, batches_per_epoch,
results['acc_y'], results['acc_y_stu'])
print(log_str)
with open(train_log_file, "a") as f:
f.write(log_str)
f.write("\n")
f.close()
is_better = tracker.check_and_update(results, global_step)
if is_better['acc_y']:
train_helper.save_best(sess, global_step=global_step)
# Last save
train_helper.save(sess, global_step)
def main(args):
# Load config
# ===================================== #
with open(join(args.output_dir, 'config.json')) as f:
config = json.load(f)
args.__dict__.update(config)
# ===================================== #
# Load dataset
# ===================================== #
data_file = join(RAW_DATA_DIR, "ComputerVision", "dSprites",
"dsprites_ndarray_co1sh3sc6or40x32y32_64x64.npz")
# It is already in the range [0, 1]
with np.load(data_file, encoding="latin1") as f:
x_train = f['imgs']
# 3 shape * 6 scale * 40 rotation * 32 pos X * 32 pos Y
y_train = f['latents_classes'][:, 1:]
x_train = np.expand_dims(x_train.astype(np.float32), axis=-1)
num_train = len(x_train)
print("num_train: {}".format(num_train))
# ===================================== #
# Build model
# ===================================== #
if args.enc_dec_model == "1Konny":
encoder = Encoder_1Konny(args.z_dim, stochastic=True)
decoder = Decoder_1Konny()
disc_z = DiscriminatorZ_1Konny()
else:
raise ValueError("Do not support enc_dec_model='{}'!".format(
args.enc_dec_model))
model = AAE([64, 64, 1],
args.z_dim,
encoder=encoder,
decoder=decoder,
discriminator_z=disc_z,
rec_x_mode=args.rec_x_mode,
stochastic_z=args.stochastic_z,
use_gp0_z=True,
gp0_z_mode=args.gp0_z_mode)
loss_coeff_dict = {
'rec_x': args.rec_x_coeff,
'G_loss_z1_gen': args.G_loss_z1_gen_coeff,
'D_loss_z1_gen': args.D_loss_z1_gen_coeff,
'gp0_z': args.gp0_z_coeff,
}
model.build(loss_coeff_dict)
SimpleParamPrinter.print_all_params_tf_slim()
# ===================================== #
# Initialize session
# ===================================== #
config_proto = tf.ConfigProto(allow_soft_placement=True)
config_proto.gpu_options.allow_growth = True
config_proto.gpu_options.per_process_gpu_memory_fraction = 0.9
sess = tf.Session(config=config_proto)
model_dir = make_dir_if_not_exist(join(args.output_dir, "model_tf"))
train_helper = SimpleTrainHelper(log_dir=None, save_dir=model_dir)
train_helper.load(sess, load_step=args.load_step)
# ===================================== #
# Experiments
# ===================================== #
save_dir = make_dir_if_not_exist(
join(args.save_dir, "{}_{}".format(args.enc_dec_model, args.run)))
np.set_printoptions(threshold=np.nan,
linewidth=1000,
precision=5,
suppress=True)
# ===================================== #
# Compute representations
# ===================================== #
z_data_file = join(save_dir, "z_data.npz")
if not exists(z_data_file):
all_z_mean = []
all_z_stddev = []
print("")
print("Compute all_z_mean, all_z_stddev!")
count = 0
for batch_ids in iterate_data(num_train,
10 * args.batch_size,
shuffle=False):
x = x_train[batch_ids]
z_samples, z_mean, z_stddev = sess.run(model.get_output(
['z1_gen', 'z_mean', 'z_stddev']),
feed_dict={
model.is_train: False,
model.x_ph: x
})
all_z_mean.append(z_mean)
all_z_stddev.append(z_stddev)
count += len(batch_ids)
print("\rProcessed {} samples!".format(count), end="")
print()
all_z_mean = np.concatenate(all_z_mean, axis=0)
all_z_stddev = np.concatenate(all_z_stddev, axis=0)
np.savez_compressed(z_data_file,
all_z_mean=all_z_mean,
all_z_stddev=all_z_stddev)
else:
print("{} exists. Load data from file!".format(z_data_file))
with np.load(z_data_file, "r") as f:
all_z_mean = f['all_z_mean']
all_z_stddev = f['all_z_stddev']
# ===================================== #
cont_mask = [False, True, True, True, True
] if args.continuous_only else None
if args.classifier == "LASSO":
results = compute_metrics_with_LASSO(latents=all_z_mean,
factors=y_train,
params={
'alpha': args.LASSO_alpha,
'max_iter': args.LASSO_iters
},
cont_mask=cont_mask)
result_file = join(
save_dir, "results[LASSO,{},alpha={},iters={}].npz".format(
"cont" if args.continuous_only else "all", args.LASSO_alpha,
args.LASSO_iters))
else:
results = compute_metrics_with_RandomForest(latents=all_z_mean,
factors=y_train,
params={
'n_estimators':
args.RF_trees,
'max_depth':
args.RF_depth
})
result_file = join(
save_dir, "results[RF,{},trees={},depth={}].npz".format(
"cont" if args.continuous_only else "all", args.RF_trees,
args.RF_depth))
np.savez_compressed(result_file, **results)
def main(args):
# =====================================
# Load config
# =====================================
with open(join(args.output_dir, 'config.json')) as f:
config = json.load(f)
args.__dict__.update(config)
# =====================================
# Dataset
# =====================================
data_file = join(RAW_DATA_DIR, "ComputerVision", "dSprites",
"dsprites_ndarray_co1sh3sc6or40x32y32_64x64.npz")
# It is already in the range [0, 1]
with np.load(data_file, encoding="latin1") as f:
x_train = f['imgs']
# 3 shape * 6 scale * 40 rotation * 32 pos X * 32 pos Y
y_train = f['latents_classes'][:, 1:]
x_train = np.expand_dims(x_train.astype(np.float32), axis=-1)
num_train = len(x_train)
print("num_train: {}".format(num_train))
print("y_train[:10]: {}".format(y_train[:10]))
# =====================================
# Instantiate model
# =====================================
if args.enc_dec_model == "1Konny":
encoder = Encoder_1Konny(args.z_dim, stochastic=True)
decoder = Decoder_1Konny()
disc_z = DiscriminatorZ_1Konny(num_outputs=2)
else:
raise ValueError("Do not support enc_dec_model='{}'!".format(args.enc_dec_model))
model = FactorVAE([64, 64, 1], args.z_dim,
encoder=encoder, decoder=decoder,
discriminator_z=disc_z,
rec_x_mode=args.rec_x_mode,
use_gp0_z_tc=True, gp0_z_tc_mode=args.gp0_z_tc_mode)
loss_coeff_dict = {
'rec_x': args.rec_x_coeff,
'kld_loss': args.kld_loss_coeff,
'tc_loss': args.tc_loss_coeff,
'gp0_z_tc': args.gp0_z_tc_coeff,
'Dz_tc_loss_coeff': args.Dz_tc_loss_coeff,
}
model.build(loss_coeff_dict)
SimpleParamPrinter.print_all_params_tf_slim()
# =====================================
# Load model
# =====================================
config_proto = tf.ConfigProto(allow_soft_placement=True)
config_proto.gpu_options.allow_growth = True
config_proto.gpu_options.per_process_gpu_memory_fraction = 0.9
sess = tf.Session(config=config_proto)
model_dir = make_dir_if_not_exist(join(args.output_dir, "model_tf"))
train_helper = SimpleTrainHelper(log_dir=None, save_dir=model_dir)
# Load model
train_helper.load(sess, load_step=args.load_step)
# =====================================
# Experiments
save_dir = make_dir_if_not_exist(join(args.save_dir, "{}_{}".format(args.enc_dec_model, args.run)))
# =====================================
np.set_printoptions(threshold=np.nan, linewidth=1000, precision=5, suppress=True)
num_bins = args.num_bins
data_proportion = args.data_proportion
num_data = int(data_proportion * num_train)
assert num_data == num_train, "For dSprites, you must use all data!"
# file
f = open(join(save_dir, 'log[bins={},data={}].txt'.
format(num_bins, data_proportion)), mode='w')
# print function
print_ = functools.partial(print_both, file=f)
print_("num_bins: {}".format(num_bins))
print_("data_proportion: {}".format(data_proportion))
# Compute representations
# ================================= #
z_data_file = join(save_dir, "z_data[data={}].npz".format(data_proportion))
if not exists(z_data_file):
all_z_mean = []
all_z_stddev = []
print("")
print("Compute all_z_mean, all_z_stddev and all_attrs!")
count = 0
for batch_ids in iterate_data(num_data, 10 * args.batch_size, shuffle=False):
x = x_train[batch_ids]
z_mean, z_stddev = sess.run(
model.get_output(['z_mean', 'z_stddev']),
feed_dict={model.is_train: False, model.x_ph: x})
all_z_mean.append(z_mean)
all_z_stddev.append(z_stddev)
count += len(batch_ids)
print("\rProcessed {} samples!".format(count), end="")
print()
all_z_mean = np.concatenate(all_z_mean, axis=0)
all_z_stddev = np.concatenate(all_z_stddev, axis=0)
np.savez_compressed(z_data_file, all_z_mean=all_z_mean,
all_z_stddev=all_z_stddev)
else:
print("{} exists. Load data from file!".format(z_data_file))
with np.load(z_data_file, "r") as f:
all_z_mean = f['all_z_mean']
all_z_stddev = f['all_z_stddev']
print_("")
print_("all_z_mean.shape: {}".format(all_z_mean.shape))
print_("all_z_stddev.shape: {}".format(all_z_stddev.shape))
# ================================= #
# Transpose and compute MIG score
# ================================= #
assert len(all_z_mean) == len(y_train)
# (num_latents, num_samples)
all_z_mean = np.transpose(all_z_mean, [1, 0])
print_("")
print_("all_z_mean.shape: {}".format(all_z_mean.shape))
y_train = np.transpose(y_train, [1, 0])
print_("")
print_("y_train.shape: {}".format(y_train.shape))
# All
# --------------------------------- #
result_all = compute_mig(all_z_mean, y_train, is_discrete_z=False,
is_discrete_y=True, num_bins=num_bins)
# (num_latents, num_factors)
MI_gap_y = result_all['MI_gap_y']
attr_ids_sorted = np.argsort(MI_gap_y, axis=0)[::-1].tolist()
MI_gap_y_sorted = MI_gap_y[attr_ids_sorted].tolist()
print_("")
print_("MIG: {}".format(result_all['MIG']))
print_("Sorted factors:\n{}".format(list(zip(attr_ids_sorted, MI_gap_y_sorted))))
save_file = join(save_dir, "results[bins={},data={}].npz".format(num_bins, data_proportion))
np.savez_compressed(save_file, **result_all)
# --------------------------------- #
# ================================= #
f.close()
def main(args):
# =====================================
# Load config
# =====================================
with open(join(args.output_dir, 'config.json')) as f:
config = json.load(f)
args.__dict__.update(config)
# =====================================
# Dataset
# =====================================
data_file = join(RAW_DATA_DIR, "ComputerVision", "dSprites",
"dsprites_ndarray_co1sh3sc6or40x32y32_64x64.npz")
# It is already in the range [0, 1]
with np.load(data_file, encoding="latin1") as f:
x_train = f['imgs']
# 3 shape * 6 scale * 40 rotation * 32 pos X * 32 pos Y
y_train = f['latents_classes']
x_train = np.expand_dims(x_train.astype(np.float32), axis=-1)
num_train = len(x_train)
print("num_train: {}".format(num_train))
print("y_train[:10]: {}".format(y_train[:10]))
# =====================================
# Instantiate model
# =====================================
if args.enc_dec_model == "1Konny":
encoder = Encoder_1Konny(args.z_dim, stochastic=True)
decoder = Decoder_1Konny()
disc_z = DiscriminatorZ_1Konny()
else:
raise ValueError("Do not support enc_dec_model='{}'!".format(
args.enc_dec_model))
model = AAE([64, 64, 1],
args.z_dim,
encoder=encoder,
decoder=decoder,
discriminator_z=disc_z,
rec_x_mode=args.rec_x_mode,
stochastic_z=args.stochastic_z,
use_gp0_z=True,
gp0_z_mode=args.gp0_z_mode)
loss_coeff_dict = {
'rec_x': args.rec_x_coeff,
'G_loss_z1_gen': args.G_loss_z1_gen_coeff,
'D_loss_z1_gen': args.D_loss_z1_gen_coeff,
'gp0_z': args.gp0_z_coeff,
}
model.build(loss_coeff_dict)
SimpleParamPrinter.print_all_params_tf_slim()
# =====================================
# Load model
# =====================================
config_proto = tf.ConfigProto(allow_soft_placement=True)
config_proto.gpu_options.allow_growth = True
config_proto.gpu_options.per_process_gpu_memory_fraction = 0.9
sess = tf.Session(config=config_proto)
model_dir = make_dir_if_not_exist(join(args.output_dir, "model_tf"))
train_helper = SimpleTrainHelper(log_dir=None, save_dir=model_dir)
# Load model
train_helper.load(sess, load_step=args.load_step)
# =====================================
# Experiments
save_dir = make_dir_if_not_exist(
join(args.save_dir, "{}_{}".format(args.enc_dec_model, args.run)))
# =====================================
np.set_printoptions(threshold=np.nan,
linewidth=1000,
precision=5,
suppress=True)
num_bins = args.num_bins
bin_limits = tuple([float(s) for s in args.bin_limits.split(";")])
data_proportion = args.data_proportion
num_data = int(data_proportion * num_train)
assert num_data == num_train, "For dSprites, you must use all data!"
eps = 1e-8
# file
f = open(join(
save_dir, 'log[bins={},bin_limits={},data={}].txt'.format(
num_bins, bin_limits, data_proportion)),
mode='w')
# print function
print_ = functools.partial(print_both, file=f)
print_("num_bins: {}".format(num_bins))
print_("bin_limits: {}".format(bin_limits))
print_("data_proportion: {}".format(data_proportion))
# Compute bins
# ================================= #
print_("")
print_("bin_limits: {}".format(bin_limits))
assert len(bin_limits) == 2 and bin_limits[0] < bin_limits[
1], "bin_limits={}".format(bin_limits)
bins = np.linspace(bin_limits[0],
bin_limits[1],
num_bins + 1,
endpoint=True)
print_("bins: {}".format(bins))
assert len(bins) == num_bins + 1
bin_widths = [bins[b] - bins[b - 1] for b in range(1, len(bins))]
print_("bin_widths: {}".format(bin_widths))
assert len(bin_widths
) == num_bins, "len(bin_widths)={} while num_bins={}!".format(
len(bin_widths), num_bins)
assert np.all(np.greater(bin_widths,
0)), "bin_widths: {}".format(bin_widths)
bin_centers = [(bins[b] + bins[b - 1]) * 0.5 for b in range(1, len(bins))]
print_("bin_centers: {}".format(bin_centers))
assert len(bin_centers
) == num_bins, "len(bin_centers)={} while num_bins={}!".format(
len(bin_centers), num_bins)
# ================================= #
# Compute representations
# ================================= #
z_data_file = join(save_dir, "z_data[data={}].npz".format(data_proportion))
if not exists(z_data_file):
all_z_mean = []
all_z_stddev = []
print("")
print("Compute all_z_mean, all_z_stddev and all_attrs!")
count = 0
for batch_ids in iterate_data(num_data,
10 * args.batch_size,
shuffle=False):
x = x_train[batch_ids]
z_mean, z_stddev = sess.run(model.get_output(
['z_mean', 'z_stddev']),
feed_dict={
model.is_train: False,
model.x_ph: x
})
all_z_mean.append(z_mean)
all_z_stddev.append(z_stddev)
count += len(batch_ids)
print("\rProcessed {} samples!".format(count), end="")
print()
all_z_mean = np.concatenate(all_z_mean, axis=0)
all_z_stddev = np.concatenate(all_z_stddev, axis=0)
np.savez_compressed(z_data_file,
all_z_mean=all_z_mean,
all_z_stddev=all_z_stddev)
else:
print("{} exists. Load data from file!".format(z_data_file))
with np.load(z_data_file, "r") as f:
all_z_mean = f['all_z_mean']
all_z_stddev = f['all_z_stddev']
# ================================= #
print_("")
all_Q_z_cond_x = []
for i in range(args.z_dim):
print_("\nCompute all_Q_z{}_cond_x!".format(i))
all_Q_s_cond_x = []
for batch_ids in iterate_data(len(all_z_mean),
500,
shuffle=False,
include_remaining=True):
# (batch_size, num_bins)
q_s_cond_x = normal_density(
np.expand_dims(bin_centers, axis=0),
mean=np.expand_dims(all_z_mean[batch_ids, i], axis=-1),
stddev=np.expand_dims(all_z_stddev[batch_ids, i], axis=-1))
# (batch_size, num_bins)
max_q_s_cond_x = np.max(q_s_cond_x, axis=-1)
# print("\nmax_q_s_cond_x: {}".format(np.sort(max_q_s_cond_x)))
# (batch_size, num_bins)
deter_s_cond_x = at_bin(all_z_mean[batch_ids, i],
bins).astype(np.float32)
# (batch_size, num_bins)
Q_s_cond_x = q_s_cond_x * np.expand_dims(bin_widths, axis=0)
Q_s_cond_x = Q_s_cond_x / np.maximum(
np.sum(Q_s_cond_x, axis=1, keepdims=True), eps)
# print("sort(sum(Q_s_cond_x)) (before): {}".format(np.sort(np.sum(Q_s_cond_x, axis=-1))))
Q_s_cond_x = np.where(
np.expand_dims(np.less(max_q_s_cond_x, 1e-5), axis=-1),
deter_s_cond_x, Q_s_cond_x)
# print("sort(sum(Q_s_cond_x)) (after): {}".format(np.sort(np.sum(Q_s_cond_x, axis=-1))))
all_Q_s_cond_x.append(Q_s_cond_x)
# (num_samples, num_bins)
all_Q_s_cond_x = np.concatenate(all_Q_s_cond_x, axis=0)
assert np.all(all_Q_s_cond_x >= 0), "'all_Q_s_cond_x' contains negative values. " \
"sorted_all_Q_s_cond_x[:30]:\n{}!".format(np.sort(all_Q_s_cond_x[:30], axis=None))
assert len(all_Q_s_cond_x) == num_train
all_Q_z_cond_x.append(all_Q_s_cond_x)
# (z_dim, num_samples, num_bins)
all_Q_z_cond_x = np.asarray(all_Q_z_cond_x, dtype=np.float32)
print_("all_Q_z_cond_x.shape: {}".format(all_Q_z_cond_x.shape))
print_("sum(all_Q_z_cond_x)[:, :10]:\n{}".format(
np.sum(all_Q_z_cond_x, axis=-1)[:, :10]))
# (z_dim, num_bins)
Q_z = np.mean(all_Q_z_cond_x, axis=1)
log_Q_z = np.log(np.clip(Q_z, eps, 1 - eps))
print_("Q_z.shape: {}".format(Q_z.shape))
print_("sum(Q_z): {}".format(np.sum(Q_z, axis=-1)))
# (z_dim, )
H_z = -np.sum(Q_z * log_Q_z, axis=-1)
# Factors
gt_factors = ['shape', 'scale', 'rotation', 'pos_x', 'pos_y']
gt_num_values = [3, 6, 40, 32, 32]
MI_z_y = np.zeros([args.z_dim, len(gt_factors)], dtype=np.float32)
H_z_y = np.zeros([args.z_dim, len(gt_factors)], dtype=np.float32)
ids_sorted = np.zeros([args.z_dim, len(gt_factors)], dtype=np.int32)
MI_z_y_sorted = np.zeros([args.z_dim, len(gt_factors)], dtype=np.float32)
H_z_y_sorted = np.zeros([args.z_dim, len(gt_factors)], dtype=np.float32)
H_y = []
RMIG = []
JEMMI = []
for k, (factor, num_values) in enumerate(zip(gt_factors, gt_num_values)):
print_("\n#" + "=" * 50 + "#")
print_("The {}-th gt factor '{}' has {} values!".format(
k, factor, num_values))
print_("")
# (num_samples, num_categories)
# NOTE: We must use k+1 to account for the 'color' attribute, which is always white
all_Q_yk_cond_x = one_hot(y_train[:, k + 1],
num_categories=num_values,
dtype=np.float32)
print_("all_Q_yk_cond_x.shape: {}".format(all_Q_yk_cond_x.shape))
# (num_categories)
Q_yk = np.mean(all_Q_yk_cond_x, axis=0)
log_Q_yk = np.log(np.clip(Q_yk, eps, 1 - eps))
print_("Q_yk.shape: {}".format(Q_yk.shape))
H_yk = -np.sum(Q_yk * log_Q_yk)
print_("H_yk: {}".format(H_yk))
H_y.append(H_yk)
Q_z_yk = np.zeros([args.z_dim, num_bins, num_values], dtype=np.float32)
# Compute I(zi, yk)
for i in range(args.z_dim):
print_("\n#" + "-" * 50 + "#")
all_Q_zi_cond_x = all_Q_z_cond_x[i]
assert len(all_Q_zi_cond_x) == len(all_Q_yk_cond_x) == num_train, \
"all_Q_zi_cond_x.shape: {}, all_Q_yk_cond_x.shape: {}".format(
all_Q_zi_cond_x.shape, all_Q_yk_cond_x.shape)
# (num_bins, num_categories)
Q_zi_yk = np.matmul(np.transpose(all_Q_zi_cond_x, axes=[1, 0]),
all_Q_yk_cond_x)
Q_zi_yk = Q_zi_yk / num_train
print_("np.sum(Q_zi_yk): {}".format(np.sum(Q_zi_yk)))
Q_zi_yk = Q_zi_yk / np.maximum(np.sum(Q_zi_yk), eps)
print_("np.sum(Q_zi_yk) (normalized): {}".format(np.sum(Q_zi_yk)))
assert np.all(Q_zi_yk >= 0), "'Q_zi_yk' contains negative values. " \
"sorted_Q_zi_yk[:10]:\n{}!".format(np.sort(Q_zi_yk, axis=None))
# (num_bins, num_categories)
log_Q_zi_yk = np.log(np.clip(Q_zi_yk, eps, 1 - eps))
print_("")
print_("Q_zi (default): {}".format(Q_z[i]))
print_("Q_zi (sum of Q_zi_yk over yk): {}".format(
np.sum(Q_zi_yk, axis=-1)))
print_("")
print_("Q_yk (default): {}".format(Q_yk))
print_("Q_yk (sum of Q_zi_yk over zi): {}".format(
np.sum(Q_zi_yk, axis=0)))
MI_zi_yk = Q_zi_yk * (log_Q_zi_yk - np.expand_dims(
log_Q_z[i], axis=-1) - np.expand_dims(log_Q_yk, axis=0))
MI_zi_yk = np.sum(MI_zi_yk)
H_zi_yk = -np.sum(Q_zi_yk * log_Q_zi_yk)
Q_z_yk[i] = Q_zi_yk
MI_z_y[i, k] = MI_zi_yk
H_z_y[i, k] = H_zi_yk
print_("#" + "-" * 50 + "#")
# Print statistics for all z
print_("")
print_("MI_z_yk:\n{}".format(MI_z_y[:, k]))
print_("H_z_yk:\n{}".format(H_z_y[:, k]))
print_("H_z:\n{}".format(H_z))
print_("H_yk:\n{}".format(H_yk))
# Compute RMIG and JEMMI
ids_yk_sorted = np.argsort(MI_z_y[:, k], axis=0)[::-1]
MI_z_yk_sorted = np.take_along_axis(MI_z_y[:, k],
ids_yk_sorted,
axis=0)
H_z_yk_sorted = np.take_along_axis(H_z_y[:, k], ids_yk_sorted, axis=0)
RMIG_yk = np.divide(MI_z_yk_sorted[0] - MI_z_yk_sorted[1], H_yk)
JEMMI_yk = np.divide(
H_z_yk_sorted[0] - MI_z_yk_sorted[0] + MI_z_yk_sorted[1],
H_yk + np.log(num_bins))
ids_sorted[:, k] = ids_yk_sorted
MI_z_y_sorted[:, k] = MI_z_yk_sorted
H_z_y_sorted[:, k] = H_z_yk_sorted
RMIG.append(RMIG_yk)
JEMMI.append(JEMMI_yk)
print_("")
print_("ids_sorted: {}".format(ids_sorted))
print_("MI_z_yk_sorted: {}".format(MI_z_yk_sorted))
print_("RMIG_yk: {}".format(RMIG_yk))
print_("JEMMI_yk: {}".format(JEMMI_yk))
z_yk_prob_file = join(
save_dir,
"z_yk_prob_4_{}[bins={},bin_limits={},data={}].npz".format(
factor, num_bins, bin_limits, data_proportion))
np.savez_compressed(z_yk_prob_file, Q_z_yk=Q_z_yk)
print_("#" + "=" * 50 + "#")
results = {
"MI_z_y": MI_z_y,
"H_z_y": H_z_y,
"ids_sorted": ids_sorted,
"MI_z_y_sorted": MI_z_y_sorted,
"H_z_y_sorted": H_z_y_sorted,
"H_z": H_z,
"H_y": np.asarray(H_y, dtype=np.float32),
"RMIG": np.asarray(RMIG, dtype=np.float32),
"JEMMI": np.asarray(JEMMI, dtype=np.float32),
}
result_file = join(
save_dir, "results[bins={},bin_limits={},data={}].npz".format(
num_bins, bin_limits, data_proportion))
np.savez_compressed(result_file, **results)
f.close()
def main(args):
# =====================================
# Load config
# =====================================
with open(join(args.output_dir, 'config.json')) as f:
config = json.load(f)
args.__dict__.update(config)
# =====================================
# Dataset
# =====================================
data_file = join(RAW_DATA_DIR, "ComputerVision", "dSprites",
"dsprites_ndarray_co1sh3sc6or40x32y32_64x64.npz")
# It is already in the range [0, 1]
with np.load(data_file, encoding="latin1") as f:
x_train = f['imgs']
x_train = np.reshape(x_train, [3, 6, 40, 32, 32, 64, 64, 1])
# =====================================
# Instantiate model
# =====================================
if args.enc_dec_model == "1Konny":
encoder = Encoder_1Konny(args.z_dim, stochastic=True)
decoder = Decoder_1Konny()
disc_z = DiscriminatorZ_1Konny()
else:
raise ValueError("Do not support enc_dec_model='{}'!".format(
args.enc_dec_model))
model = AAE([64, 64, 1],
args.z_dim,
encoder=encoder,
decoder=decoder,
discriminator_z=disc_z,
rec_x_mode=args.rec_x_mode,
stochastic_z=args.stochastic_z,
use_gp0_z=True,
gp0_z_mode=args.gp0_z_mode)
loss_coeff_dict = {
'rec_x': args.rec_x_coeff,
'G_loss_z1_gen': args.G_loss_z1_gen_coeff,
'D_loss_z1_gen': args.D_loss_z1_gen_coeff,
'gp0_z': args.gp0_z_coeff,
}
model.build(loss_coeff_dict)
SimpleParamPrinter.print_all_params_tf_slim()
# =====================================
# Load model
# =====================================
config_proto = tf.ConfigProto(allow_soft_placement=True)
config_proto.gpu_options.allow_growth = True
config_proto.gpu_options.per_process_gpu_memory_fraction = 0.9
sess = tf.Session(config=config_proto)
model_dir = make_dir_if_not_exist(join(args.output_dir, "model_tf"))
train_helper = SimpleTrainHelper(log_dir=None, save_dir=model_dir)
# Load model
train_helper.load(sess, load_step=args.load_step)
# =====================================
# Experiments
# save_dir = remove_dir_if_exist(join(args.save_dir, "{}_{}".format(args.enc_dec_model, args.run)), ask_4_permission=True)
# save_dir = make_dir_if_not_exist(save_dir)
save_dir = make_dir_if_not_exist(
join(args.save_dir, "{}_{}".format(args.enc_dec_model, args.run)))
# =====================================
np.set_printoptions(threshold=np.nan,
linewidth=1000,
precision=3,
suppress=True)
num_bins = args.num_bins
bin_limits = tuple([float(s) for s in args.bin_limits.split(";")])
data_proportion = args.data_proportion
# Logs
f = open(join(
save_dir, 'log[bins={},bin_limits={},data={}].txt'.format(
num_bins, bin_limits, data_proportion)),
mode='w')
print_ = functools.partial(print_both, file=f)
print_("")
print_("num_bins: {}".format(num_bins))
print_("bin_limits: {}".format(bin_limits))
print_("data_proportion: {}".format(data_proportion))
# Results
result_file = join(
args.interpretability_metrics_dir,
"{}_{}".format(args.enc_dec_model, args.run),
"results[bins={},bin_limits={},data={}].npz".format(
num_bins, bin_limits, data_proportion))
results = np.load(result_file, "r")
print_("results.keys: {}".format(list(results.keys())))
# Plotting
# =========================================== #
data = [
x_train[0, 3, 20, 16, 16], x_train[1, 3, 20, 16, 16], x_train[2, 3, 20,
16, 16]
]
gt_factors = ['Shape', 'Scale', 'Rotation', 'Pos_x', 'Pos_y']
ids_sorted = results['ids_sorted']
MI_z_y_sorted = results['MI_z_y_sorted']
H_z_y_sorted = results['H_z_y_sorted']
H_y = results['H_y']
RMIG = results['RMIG']
JEMMI = results['JEMMI']
print_("MI_z_y_sorted:\n{}".format(MI_z_y_sorted))
print_("\nShow RMIG!")
for k in range(len(gt_factors)):
print_(
"{}, RMIG: {:.4f}, RMIG (unnorm): {:.4f}, H: {:.4f}, I1: {:.4f}, I2: {:.4f}"
.format(gt_factors[k], RMIG[k], RMIG[k] * H_y[k], H_y[k],
MI_z_y_sorted[0, k], MI_z_y_sorted[1, k]))
print_("\nShow JEMMI!")
for k in range(len(gt_factors)):
print_(
"{}, JEMMI: {:.4f}, JEMMI (unnorm): {:.4f}, H1: {:.4f}, H1-I1: {:.4f}, I2: {:.4f}, "
"top2 ids: z{}, z{}".format(
gt_factors[k], JEMMI[k],
JEMMI[k] * (H_y[k] + np.log(num_bins)), H_z_y_sorted[0, k],
H_z_y_sorted[0, k] - MI_z_y_sorted[0, k], MI_z_y_sorted[1, k],
ids_sorted[0, k], ids_sorted[1, k]))
span = 3
points_one_side = 5
for n in range(len(data)):
for k in range(len(gt_factors)):
print("x={}, y={}!".format(n, gt_factors[k]))
img_file = join(
save_dir, "{}[x={},bins={},bin_limits={},data={}].png".format(
gt_factors[k], n, num_bins, bin_limits, data_proportion))
'''
ids_top10 = ids_sorted[:10, k]
MI_top10 = MI_z_y_sorted[:10, k]
model.cond_all_latents_traverse_v2(img_file, sess, data[n],
z_comps=ids_top10,
z_comp_labels=["z[{}] ({:.4f})".format(comp, mi)
for comp, mi in zip(ids_top10, MI_top10)],
span=span, points_1_side=points_one_side,
hl_x=True,
font_size=9,
title="{} (RMIG={:.4f}, JEMMI={:.4f}, H={:.4f})".format(
gt_factors[k], RMIG[k], JEMMI[k], H_y[k]),
title_font_scale=1.5,
subplot_adjust={'left': 0.16, 'right': 0.99,
'bottom': 0.01, 'top': 0.95},
size_inches=(6.5, 5.2),
batch_size=args.batch_size,
dec_output_2_img_func=binary_float_to_uint8)
'''
ids_top3 = ids_sorted[:3, k]
MI_top3 = MI_z_y_sorted[:3, k]
model.cond_all_latents_traverse_v2(
img_file,
sess,
data[n],
z_comps=ids_top3,
z_comp_labels=[
"z[{}] ({:.4f})".format(comp, mi)
for comp, mi in zip(ids_top3, MI_top3)
],
span=span,
points_1_side=points_one_side,
hl_x=True,
font_size=9,
title="{} (RMIG={:.4f}, JEMMI={:.4f}, H={:.4f})".format(
gt_factors[k], RMIG[k], JEMMI[k], H_y[k]),
title_font_scale=1.5,
subplot_adjust={
'left': 0.16,
'right': 0.99,
'bottom': 0.01,
'top': 0.88
},
size_inches=(6.2, 1.7),
batch_size=args.batch_size,
dec_output_2_img_func=binary_float_to_uint8)
f.close()
def main(args):
# =====================================
# Load config
# =====================================
with open(join(args.output_dir, 'config.json')) as f:
config = json.load(f)
args.__dict__.update(config)
# =====================================
# Dataset
# =====================================
celebA_loader = TFCelebAWithAttrLoader(root_dir=args.celebA_root_dir)
img_height, img_width = args.celebA_resize_size, args.celebA_resize_size
celebA_loader.build_transformation_flow_tf(
*celebA_loader.get_transform_fns("1Konny", resize_size=args.celebA_resize_size))
# =====================================
# Instantiate model
# =====================================
if args.activation == "relu":
activation = tf.nn.relu
elif args.activation == "leaky_relu":
activation = tf.nn.leaky_relu
else:
raise ValueError("Do not support '{}' activation!".format(args.activation))
if args.enc_dec_model == "1Konny":
# assert args.z_dim == 65, "For 1Konny, z_dim must be 65. Found {}!".format(args.z_dim)
encoder = Encoder_1Konny(args.z_dim, stochastic=True, activation=activation)
decoder = Decoder_1Konny([img_height, img_width, 3], activation=activation,
output_activation=tf.nn.sigmoid)
disc_z = DiscriminatorZ_1Konny()
else:
raise ValueError("Do not support encoder/decoder model '{}'!".format(args.enc_dec_model))
model = AAE([img_height, img_width, 3], args.z_dim,
encoder=encoder, decoder=decoder,
discriminator_z=disc_z,
rec_x_mode=args.rec_x_mode,
stochastic_z=args.stochastic_z,
use_gp0_z=True, gp0_z_mode=args.gp0_z_mode)
loss_coeff_dict = {
'rec_x': args.rec_x_coeff,
'G_loss_z1_gen': args.G_loss_z1_gen_coeff,
'D_loss_z1_gen': args.D_loss_z1_gen_coeff,
'gp0_z': args.gp0_z_coeff,
}
model.build(loss_coeff_dict)
SimpleParamPrinter.print_all_params_tf_slim()
# =====================================
# Load model
# =====================================
config_proto = tf.ConfigProto(allow_soft_placement=True)
config_proto.gpu_options.allow_growth = True
config_proto.gpu_options.per_process_gpu_memory_fraction = 0.9
sess = tf.Session(config=config_proto)
model_dir = make_dir_if_not_exist(join(args.output_dir, "model_tf"))
train_helper = SimpleTrainHelper(log_dir=None, save_dir=model_dir)
# Load model
train_helper.load(sess, load_step=args.load_step)
# =====================================
# Experiments
# save_dir = remove_dir_if_exist(join(args.save_dir, "AAE_{}".format(args.run)), ask_4_permission=True)
# save_dir = make_dir_if_not_exist(save_dir)
save_dir = make_dir_if_not_exist(join(args.save_dir, "AAE_{}".format(args.run)))
# =====================================
np.set_printoptions(threshold=np.nan, linewidth=1000, precision=4, suppress=True)
num_bins = args.num_bins
bin_limits = tuple([float(s) for s in args.bin_limits.split(";")])
data_proportion = args.data_proportion
num_data = int(data_proportion * celebA_loader.num_train_data)
eps = 1e-8
# file
f = open(join(save_dir, 'log[bins={},bin_limits={},data={}].txt'.
format(num_bins, bin_limits, data_proportion)), mode='w')
# print function
print_ = functools.partial(print_both, file=f)
print_("num_bins: {}".format(num_bins))
print_("bin_limits: {}".format(bin_limits))
print_("data_proportion: {}".format(data_proportion))
# Compute representations
# ================================= #
z_data_file = join(save_dir, "z_data[data={}].npz".format(data_proportion))
if not exists(z_data_file):
all_z_mean = []
all_z_stddev = []
print("")
print("Compute all_z_mean and all_z_stddev!")
count = 0
for batch_ids in iterate_data(num_data, 10 * args.batch_size, shuffle=False):
x = celebA_loader.sample_images_from_dataset(sess, 'train', batch_ids)
z_mean, z_stddev = sess.run(
model.get_output(['z_mean', 'z_stddev']),
feed_dict={model.is_train: False, model.x_ph: x})
all_z_mean.append(z_mean)
all_z_stddev.append(z_stddev)
count += len(batch_ids)
print("\rProcessed {} samples!".format(count), end="")
print()
all_z_mean = np.concatenate(all_z_mean, axis=0)
all_z_stddev = np.concatenate(all_z_stddev, axis=0)
np.savez_compressed(z_data_file, all_z_mean=all_z_mean, all_z_stddev=all_z_stddev)
else:
print("{} exists. Load data from file!".format(z_data_file))
with np.load(z_data_file, "r") as f:
all_z_mean = f['all_z_mean']
all_z_stddev = f['all_z_stddev']
print_("")
print_("all_z_mean.shape: {}".format(all_z_mean.shape))
print_("all_z_stddev.shape: {}".format(all_z_stddev.shape))
# ================================= #
# Compute bins
# ================================= #
print_("")
print_("bin_limits: {}".format(bin_limits))
assert len(bin_limits) == 2 and bin_limits[0] < bin_limits[1], "bin_limits={}".format(bin_limits)
bins = np.linspace(bin_limits[0], bin_limits[1], num_bins + 1, endpoint=True)
print_("bins: {}".format(bins))
assert len(bins) == num_bins + 1
bin_widths = [bins[b] - bins[b - 1] for b in range(1, len(bins))]
print_("bin_widths: {}".format(bin_widths))
assert len(bin_widths) == num_bins, "len(bin_widths)={} while num_bins={}!".format(len(bin_widths), num_bins)
assert np.all(np.greater(bin_widths, 0)), "bin_widths: {}".format(bin_widths)
bin_centers = [(bins[b] + bins[b - 1]) * 0.5 for b in range(1, len(bins))]
print_("bin_centers: {}".format(bin_centers))
assert len(bin_centers) == num_bins, "len(bin_centers)={} while num_bins={}!".format(len(bin_centers), num_bins)
# ================================= #
# Compute mutual information
# ================================= #
H_z = []
H_z_cond_x = []
MI_z_x = []
norm_MI_z_x = []
Q_z_cond_x = []
Q_z = []
for i in range(args.z_dim):
print_("")
print_("Compute I(z{}, x)!".format(i))
# Q_s_cond_x
all_Q_s_cond_x = []
for batch_ids in iterate_data(len(all_z_mean), 500, shuffle=False, include_remaining=True):
# (batch_size, num_bins)
q_s_cond_x = normal_density(np.expand_dims(bin_centers, axis=0),
mean=np.expand_dims(all_z_mean[batch_ids, i], axis=-1),
stddev=np.expand_dims(all_z_stddev[batch_ids, i], axis=-1))
# (batch_size, num_bins)
max_q_s_cond_x = np.max(q_s_cond_x, axis=-1)
# print("\nmax_q_s_cond_x: {}".format(np.sort(max_q_s_cond_x)))
# (batch_size, num_bins)
deter_s_cond_x = at_bin(all_z_mean[batch_ids, i], bins).astype(np.float32)
# (batch_size, num_bins)
Q_s_cond_x = q_s_cond_x * np.expand_dims(bin_widths, axis=0)
Q_s_cond_x = Q_s_cond_x / np.maximum(np.sum(Q_s_cond_x, axis=1, keepdims=True), eps)
# print("sort(sum(Q_s_cond_x)) (before): {}".format(np.sort(np.sum(Q_s_cond_x, axis=-1))))
Q_s_cond_x = np.where(np.expand_dims(np.less(max_q_s_cond_x, 1e-5), axis=-1),
deter_s_cond_x, Q_s_cond_x)
# print("sort(sum(Q_s_cond_x)) (after): {}".format(np.sort(np.sum(Q_s_cond_x, axis=-1))))
all_Q_s_cond_x.append(Q_s_cond_x)
all_Q_s_cond_x = np.concatenate(all_Q_s_cond_x, axis=0)
print_("sort(sum(all_Q_s_cond_x))[:10]: {}".format(
np.sort(np.sum(all_Q_s_cond_x, axis=-1), axis=0)[:100]))
assert np.all(all_Q_s_cond_x >= 0), "'all_Q_s_cond_x' contains negative values. " \
"sorted_all_Q_s_cond_x[:30]:\n{}!".format(np.sort(all_Q_s_cond_x[:30], axis=None))
Q_z_cond_x.append(all_Q_s_cond_x)
H_zi_cond_x = -np.mean(np.sum(all_Q_s_cond_x * np.log(np.maximum(all_Q_s_cond_x, eps)), axis=1), axis=0)
# Q_s
Q_s = np.mean(all_Q_s_cond_x, axis=0)
print_("Q_s: {}".format(Q_s))
print_("sum(Q_s): {}".format(sum(Q_s)))
assert np.all(Q_s >= 0), "'Q_s' contains negative values. " \
"sorted_Q_s[:10]:\n{}!".format(np.sort(Q_s, axis=None))
Q_s = Q_s / np.sum(Q_s, axis=0)
print_("sum(Q_s) (normalized): {}".format(sum(Q_s)))
Q_z.append(Q_s)
H_zi = -np.sum(Q_s * np.log(np.maximum(Q_s, eps)), axis=0)
MI_zi_x = H_zi - H_zi_cond_x
normalized_MI_zi_x = (1.0 * MI_zi_x) / (H_zi + eps)
print_("H_zi: {}".format(H_zi))
print_("H_zi_cond_x: {}".format(H_zi_cond_x))
print_("MI_zi_x: {}".format(MI_zi_x))
print_("normalized_MI_zi_x: {}".format(normalized_MI_zi_x))
H_z.append(H_zi)
H_z_cond_x.append(H_zi_cond_x)
MI_z_x.append(MI_zi_x)
norm_MI_z_x.append(normalized_MI_zi_x)
H_z = np.asarray(H_z, dtype=np.float32)
H_z_cond_x = np.asarray(H_z_cond_x, dtype=np.float32)
MI_z_x = np.asarray(MI_z_x, dtype=np.float32)
norm_MI_z_x = np.asarray(norm_MI_z_x, dtype=np.float32)
print_("")
print_("H_z: {}".format(H_z))
print_("H_z_cond_x: {}".format(H_z_cond_x))
print_("MI_z_x: {}".format(MI_z_x))
print_("norm_MI_z_x: {}".format(norm_MI_z_x))
sorted_z_comps = np.argsort(MI_z_x, axis=0)[::-1]
sorted_MI_z_x = np.take_along_axis(MI_z_x, sorted_z_comps, axis=0)
print_("sorted_MI_z_x: {}".format(sorted_MI_z_x))
print_("sorted_z_comps: {}".format(sorted_z_comps))
sorted_norm_z_comps = np.argsort(norm_MI_z_x, axis=0)[::-1]
sorted_norm_MI_z_x = np.take_along_axis(norm_MI_z_x, sorted_norm_z_comps, axis=0)
print_("sorted_norm_MI_z_x: {}".format(sorted_norm_MI_z_x))
print_("sorted_norm_z_comps: {}".format(sorted_norm_z_comps))
result_file = join(save_dir, 'results[bins={},bin_limits={},data={}].npz'.
format(num_bins, bin_limits, data_proportion))
np.savez_compressed(result_file,
H_z=H_z, H_z_cond_x=H_z_cond_x, MI_z_x=MI_z_x, norm_MI_z_x=norm_MI_z_x,
sorted_MI_z_x=sorted_MI_z_x, sorted_z_comps=sorted_z_comps,
sorted_norm_MI_z_x=sorted_norm_MI_z_x,
sorted_norm_z_comps=sorted_norm_z_comps)
Q_z_cond_x = np.asarray(Q_z_cond_x, dtype=np.float32)
Q_z = np.asarray(Q_z, dtype=np.float32)
z_prob_file = join(save_dir, 'z_prob[bins={},bin_limits={},data={}].npz'.
format(num_bins, bin_limits, data_proportion))
np.savez_compressed(z_prob_file, Q_z_cond_x=Q_z_cond_x, Q_z=Q_z)
def main():
raw_data_file = download_if_not_exist(DIR_RAW, "cifar-10-python.tar.gz",
DATASET_URL)
dataset = extract_data_and_labels(raw_data_file)
train_data, train_labels = dataset['train_data'], dataset['train_labels']
test_data, test_labels = dataset['test_data'], dataset['test_labels']
label_names = dataset['label_names']
modes = ['bytes', '0to1', 'm1p1']
for mode in modes:
print("\nCreate the '{}' version of CIFAR10!".format(mode))
if mode == 'bytes':
processed_train_data = train_data
processed_test_data = test_data
elif mode == '0to1':
processed_train_data = uint8_to_float(train_data,
pixel_inv_scale=255,
pixel_shift=0)
processed_test_data = uint8_to_float(test_data,
pixel_inv_scale=255,
pixel_shift=0)
elif mode == 'm1p1':
processed_train_data = uint8_to_float(train_data,
pixel_inv_scale=127.5,
pixel_shift=-1)
processed_test_data = uint8_to_float(test_data,
pixel_inv_scale=127.5,
pixel_shift=-1)
else:
raise ValueError("Only support 'mode' in {}!".format(modes))
save_dir = make_dir_if_not_exist(join(DIR_PROCESSED, mode))
np.savez_compressed(join(save_dir, "train.npz"),
x=processed_train_data,
y=train_labels,
y_names=label_names)
np.savez_compressed(join(save_dir, "test.npz"),
x=processed_test_data,
y=test_labels,
y_names=label_names)
num_cols = 5
train_idx = 100
test_idx = 100
# Uncomment if you want to show images
# ---------------------------------- #
'''
import matplotlib.pyplot as plt
fig, axes = plt.subplots(2, num_cols)
for n in range(num_cols):
if mode == "bytes" or mode == "0to1":
plot_image(axes[0][n], processed_train_data[train_idx + n], title="train[{}]: {}".
format(train_idx + n, label_names[train_labels[train_idx + n]]))
plot_image(axes[1][n], processed_test_data[test_idx + n], title="test[{}]: {}".
format(test_idx + n, label_names[test_labels[test_idx + n]]))
elif mode == "m1p1":
plot_image(axes[0][n], (processed_train_data[train_idx + n] + 1.0) / 2.0, title="train[{}]: {}".
format(train_idx + n, label_names[train_labels[train_idx + n]]))
plot_image(axes[1][n], (processed_test_data[test_idx + n] + 1.0) / 2.0, title="test[{}]: {}".
format(test_idx + n, label_names[test_labels[test_idx + n]]))
plt.show()
'''
# ---------------------------------- #
from shutil import copyfile
copyfile(join(DIR_PROCESSED, "bytes", "train.npz"),
join(DIR_PROCESSED, "train.npz"))
copyfile(join(DIR_PROCESSED, "bytes", "test.npz"),
join(DIR_PROCESSED, "test.npz"))
def main(args):
# =====================================
# Load config
# =====================================
with open(join(args.output_dir, 'config.json')) as f:
config = json.load(f)
args.__dict__.update(config)
# =====================================
# Dataset
# =====================================
celebA_loader = TFCelebAWithAttrLoader(root_dir=args.celebA_root_dir)
img_height, img_width = args.celebA_resize_size, args.celebA_resize_size
celebA_loader.build_transformation_flow_tf(
*celebA_loader.get_transform_fns("1Konny",
resize_size=args.celebA_resize_size))
# =====================================
# Instantiate model
# =====================================
if args.activation == "relu":
activation = tf.nn.relu
elif args.activation == "leaky_relu":
activation = tf.nn.leaky_relu
else:
raise ValueError("Do not support '{}' activation!".format(
args.activation))
if args.enc_dec_model == "1Konny":
# assert args.z_dim == 65, "For 1Konny, z_dim must be 65. Found {}!".format(args.z_dim)
encoder = Encoder_1Konny(args.z_dim,
stochastic=True,
activation=activation)
decoder = Decoder_1Konny([img_height, img_width, 3],
activation=activation,
output_activation=tf.nn.sigmoid)
disc_z = DiscriminatorZ_1Konny()
else:
raise ValueError("Do not support encoder/decoder model '{}'!".format(
args.enc_dec_model))
model = AAE([img_height, img_width, 3],
args.z_dim,
encoder=encoder,
decoder=decoder,
discriminator_z=disc_z,
rec_x_mode=args.rec_x_mode,
stochastic_z=args.stochastic_z,
use_gp0_z=True,
gp0_z_mode=args.gp0_z_mode)
loss_coeff_dict = {
'rec_x': args.rec_x_coeff,
'G_loss_z1_gen': args.G_loss_z1_gen_coeff,
'D_loss_z1_gen': args.D_loss_z1_gen_coeff,
'gp0_z': args.gp0_z_coeff,
}
model.build(loss_coeff_dict)
SimpleParamPrinter.print_all_params_tf_slim()
# =====================================
# Load model
# =====================================
config_proto = tf.ConfigProto(allow_soft_placement=True)
config_proto.gpu_options.allow_growth = True
config_proto.gpu_options.per_process_gpu_memory_fraction = 0.9
sess = tf.Session(config=config_proto)
model_dir = make_dir_if_not_exist(join(args.output_dir, "model_tf"))
train_helper = SimpleTrainHelper(log_dir=None, save_dir=model_dir)
# Load model
train_helper.load(sess, load_step=args.load_step)
# =====================================
# Experiments
# save_dir = remove_dir_if_exist(join(args.save_dir, "AAE_{}".format(args.run)), ask_4_permission=False)
# save_dir = make_dir_if_not_exist(save_dir)
save_dir = make_dir_if_not_exist(
join(args.save_dir, "AAE_{}".format(args.run)))
# =====================================
np.set_printoptions(threshold=np.nan,
linewidth=1000,
precision=3,
suppress=True)
num_bins = args.num_bins
bin_limits = tuple([float(s) for s in args.bin_limits.split(";")])
data_proportion = args.data_proportion
num_data = int(data_proportion * celebA_loader.num_train_data)
eps = 1e-8
# file
f = open(join(
save_dir, 'log[bins={},bin_limits={},data={}].txt'.format(
num_bins, bin_limits, data_proportion)),
mode='w')
# print function
print_ = functools.partial(print_both, file=f)
'''
if attr_type == 0:
attr_names = celebA_loader.attributes
elif attr_type == 1:
attr_names = ['Male', 'Black_Hair', 'Blond_Hair', 'Straight_Hair', 'Wavy_Hair', 'Bald',
'Oval_Face', 'Big_Nose', 'Chubby', 'Double_Chin', 'Goatee', 'No_Beard',
'Mouth_Slightly_Open', 'Smiling',
'Eyeglasses', 'Pale_Skin']
else:
raise ValueError("Only support factor_type=0 or 1!")
'''
print_("num_bins: {}".format(num_bins))
print_("bin_limits: {}".format(bin_limits))
print_("data_proportion: {}".format(data_proportion))
# Compute bins
# ================================= #
print_("")
print_("bin_limits: {}".format(bin_limits))
assert len(bin_limits) == 2 and bin_limits[0] < bin_limits[
1], "bin_limits={}".format(bin_limits)
bins = np.linspace(bin_limits[0],
bin_limits[1],
num_bins + 1,
endpoint=True)
print_("bins: {}".format(bins))
assert len(bins) == num_bins + 1
bin_widths = [bins[b] - bins[b - 1] for b in range(1, len(bins))]
print_("bin_widths: {}".format(bin_widths))
assert len(bin_widths
) == num_bins, "len(bin_widths)={} while num_bins={}!".format(
len(bin_widths), num_bins)
assert np.all(np.greater(bin_widths,
0)), "bin_widths: {}".format(bin_widths)
bin_centers = [(bins[b] + bins[b - 1]) * 0.5 for b in range(1, len(bins))]
print_("bin_centers: {}".format(bin_centers))
assert len(bin_centers
) == num_bins, "len(bin_centers)={} while num_bins={}!".format(
len(bin_centers), num_bins)
# ================================= #
# Compute representations
# ================================= #
z_data_attr_file = join(save_dir,
"z_data[data={}].npz".format(data_proportion))
if not exists(z_data_attr_file):
all_z_mean = []
all_z_stddev = []
all_attrs = []
print("")
print("Compute all_z_mean, all_z_stddev and all_attrs!")
count = 0
for batch_ids in iterate_data(num_data,
10 * args.batch_size,
shuffle=False):
x = celebA_loader.sample_images_from_dataset(
sess, 'train', batch_ids)
attrs = celebA_loader.sample_attrs_from_dataset('train', batch_ids)
assert attrs.shape[1] == celebA_loader.num_attributes
z_mean, z_stddev = sess.run(model.get_output(
['z_mean', 'z_stddev']),
feed_dict={
model.is_train: False,
model.x_ph: x
})
all_z_mean.append(z_mean)
all_z_stddev.append(z_stddev)
all_attrs.append(attrs)
count += len(batch_ids)
print("\rProcessed {} samples!".format(count), end="")
print()
all_z_mean = np.concatenate(all_z_mean, axis=0)
all_z_stddev = np.concatenate(all_z_stddev, axis=0)
all_attrs = np.concatenate(all_attrs, axis=0)
np.savez_compressed(z_data_attr_file,
all_z_mean=all_z_mean,
all_z_stddev=all_z_stddev,
all_attrs=all_attrs)
else:
print("{} exists. Load data from file!".format(z_data_attr_file))
with np.load(z_data_attr_file, "r") as f:
all_z_mean = f['all_z_mean']
all_z_stddev = f['all_z_stddev']
all_attrs = f['all_attrs']
print_("")
print_("all_z_mean.shape: {}".format(all_z_mean.shape))
print_("all_z_stddev.shape: {}".format(all_z_stddev.shape))
print_("all_attrs.shape: {}".format(all_attrs.shape))
# ================================= #
# Compute the probability mass function for ground truth factors
# ================================= #
num_attrs = all_attrs.shape[1]
assert all_attrs.dtype == np.bool
all_attrs = all_attrs.astype(np.int32)
# (num_samples, num_attrs, 2) # The first component is 1 and the last component is 0
all_Q_y_cond_x = np.stack([all_attrs, 1 - all_attrs], axis=-1)
# ================================= #
# Compute Q(zi|x)
# Compute I(zi, yk)
# ================================= #
Q_z_y = np.zeros([args.z_dim, num_attrs, num_bins, 2], dtype=np.float32)
MI_z_y = np.zeros([args.z_dim, num_attrs], dtype=np.float32)
H_z_y = np.zeros([args.z_dim, num_attrs], dtype=np.float32)
H_z_4_diff_y = np.zeros([args.z_dim, num_attrs], dtype=np.float32)
H_y_4_diff_z = np.zeros([num_attrs, args.z_dim], dtype=np.float32)
for i in range(args.z_dim):
print_("")
print_("Compute all_Q_z{}_cond_x!".format(i))
# Q_s_cond_x
all_Q_s_cond_x = []
for batch_ids in iterate_data(len(all_z_mean),
500,
shuffle=False,
include_remaining=True):
# (batch_size, num_bins)
q_s_cond_x = normal_density(
np.expand_dims(bin_centers, axis=0),
mean=np.expand_dims(all_z_mean[batch_ids, i], axis=-1),
stddev=np.expand_dims(all_z_stddev[batch_ids, i], axis=-1))
# (batch_size, num_bins)
max_q_s_cond_x = np.max(q_s_cond_x, axis=-1)
# print("\nmax_q_s_cond_x: {}".format(np.sort(max_q_s_cond_x)))
# (batch_size, num_bins)
deter_s_cond_x = at_bin(all_z_mean[batch_ids, i],
bins).astype(np.float32)
# (batch_size, num_bins)
Q_s_cond_x = q_s_cond_x * np.expand_dims(bin_widths, axis=0)
Q_s_cond_x = Q_s_cond_x / np.maximum(
np.sum(Q_s_cond_x, axis=1, keepdims=True), eps)
# print("sort(sum(Q_s_cond_x)) (before): {}".format(np.sort(np.sum(Q_s_cond_x, axis=-1))))
Q_s_cond_x = np.where(
np.expand_dims(np.less(max_q_s_cond_x, 1e-5), axis=-1),
deter_s_cond_x, Q_s_cond_x)
# print("sort(sum(Q_s_cond_x)) (after): {}".format(np.sort(np.sum(Q_s_cond_x, axis=-1))))
all_Q_s_cond_x.append(Q_s_cond_x)
# (num_samples, num_bins)
all_Q_s_cond_x = np.concatenate(all_Q_s_cond_x, axis=0)
assert np.all(all_Q_s_cond_x >= 0), "'all_Q_s_cond_x' contains negative values. " \
"sorted_all_Q_s_cond_x[:30]:\n{}!".format(
np.sort(all_Q_s_cond_x[:30], axis=None))
assert len(all_Q_s_cond_x) == len(
all_attrs), "all_Q_s_cond_x.shape={}, all_attrs.shape={}".format(
all_Q_s_cond_x.shape, all_attrs.shape)
# I(z, y)
for k in range(num_attrs):
# Compute Q(zi, yk)
# -------------------------------- #
# (z_dim, 2)
Q_zi_yk = np.matmul(np.transpose(all_Q_s_cond_x, axes=[1, 0]),
all_Q_y_cond_x[:, k, :])
Q_zi_yk = Q_zi_yk / len(all_Q_y_cond_x)
Q_zi_yk = Q_zi_yk / np.maximum(np.sum(Q_zi_yk), eps)
assert np.all(Q_zi_yk >= 0), "'Q_zi_yk' contains negative values. " \
"sorted_Q_zi_yk[:10]:\n{}!".format(np.sort(Q_zi_yk, axis=None))
log_Q_zi_yk = np.log(np.clip(Q_zi_yk, eps, 1 - eps))
Q_z_y[i, k] = Q_zi_yk
print_("sum(Q_zi_yk): {}".format(np.sum(Q_zi_yk)))
# -------------------------------- #
# Compute Q_z
# -------------------------------- #
Q_zi = np.sum(Q_zi_yk, axis=1)
log_Q_zi = np.log(np.clip(Q_zi, eps, 1 - eps))
print_("sum(Q_z{}): {}".format(i, np.sum(Q_zi)))
print_("Q_z{}: {}".format(i, Q_zi))
# -------------------------------- #
# Compute Q_y
# -------------------------------- #
Q_yk = np.sum(Q_zi_yk, axis=0)
log_Q_yk = np.log(np.clip(Q_yk, eps, 1 - eps))
print_("sum(Q_y{}): {}".format(k, np.sum(Q_yk)))
print_("Q_y{}: {}".format(k, np.sum(Q_yk)))
# -------------------------------- #
MI_zi_yk = Q_zi_yk * (log_Q_zi_yk - np.expand_dims(
log_Q_zi, axis=-1) - np.expand_dims(log_Q_yk, axis=0))
MI_zi_yk = np.sum(MI_zi_yk)
H_zi_yk = -np.sum(Q_zi_yk * log_Q_zi_yk)
H_zi = -np.sum(Q_zi * log_Q_zi)
H_yk = -np.sum(Q_yk * log_Q_yk)
MI_z_y[i, k] = MI_zi_yk
H_z_y[i, k] = H_zi_yk
H_z_4_diff_y[i, k] = H_zi
H_y_4_diff_z[k, i] = H_yk
# ================================= #
print_("")
print_("MI_z_y:\n{}".format(MI_z_y))
print_("H_z_y:\n{}".format(H_z_y))
print_("H_z_4_diff_y:\n{}".format(H_z_4_diff_y))
print_("H_y_4_diff_z:\n{}".format(H_y_4_diff_z))
# Compute metric
# ================================= #
# Sorted in decreasing order
MI_ids_sorted = np.argsort(MI_z_y, axis=0)[::-1]
MI_sorted = np.take_along_axis(MI_z_y, MI_ids_sorted, axis=0)
MI_gap_y = np.divide(MI_sorted[0, :] - MI_sorted[1, :], H_y_4_diff_z[:, 0])
MIG = np.mean(MI_gap_y)
print_("")
print_("MI_sorted: {}".format(MI_sorted))
print_("MI_ids_sorted: {}".format(MI_ids_sorted))
print_("MI_gap_y: {}".format(MI_gap_y))
print_("MIG: {}".format(MIG))
results = {
'Q_z_y': Q_z_y,
'MI_z_y': MI_z_y,
'H_z_y': H_z_y,
'H_z_4_diff_y': H_z_4_diff_y,
'H_y_4_diff_z': H_y_4_diff_z,
'MI_sorted': MI_sorted,
'MI_ids_sorted': MI_ids_sorted,
'MI_gap_y': MI_gap_y,
'MIG': MIG,
}
result_file = join(
save_dir, 'results[bins={},bin_limits={},data={}].npz'.format(
num_bins, bin_limits, data_proportion))
np.savez_compressed(result_file, **results)
# ================================= #
f.close()
def main(args):
# =====================================
# Load config
# =====================================
with open(os.path.join(args.output_dir, 'config.json')) as f:
config = json.load(f)
args.__dict__.update(config)
# =====================================
# Preparation
# =====================================
data_file = os.path.join(RAW_DATA_DIR, "ComputerVision", "dSprites",
"dsprites_ndarray_co1sh3sc6or40x32y32_64x64.npz")
# It is already in the range [0, 1]
with np.load(data_file, encoding="latin1") as f:
x_train = f['imgs']
x_train = np.expand_dims(x_train.astype(np.float32), axis=-1)
# =====================================
# Instantiate models
# =====================================
if args.enc_dec_model == "1Konny":
encoder = Encoder_1Konny(args.z_dim, stochastic=True)
decoder = Decoder_1Konny()
disc_z = DiscriminatorZ_1Konny()
else:
raise ValueError("Do not support enc_dec_model='{}'!".format(
args.enc_dec_model))
model = AAE([64, 64, 1],
args.z_dim,
encoder=encoder,
decoder=decoder,
discriminator_z=disc_z,
rec_x_mode=args.rec_x_mode,
stochastic_z=args.stochastic_z,
use_gp0_z=True,
gp0_z_mode=args.gp0_z_mode)
loss_coeff_dict = {
'rec_x': args.rec_x_coeff,
'G_loss_z1_gen': args.G_loss_z1_gen_coeff,
'D_loss_z1_gen': args.D_loss_z1_gen_coeff,
'gp0_z': args.gp0_z_coeff,
}
model.build(loss_coeff_dict)
SimpleParamPrinter.print_all_params_list()
# =====================================
# TF Graph Handler
asset_dir = make_dir_if_not_exist(os.path.join(args.output_dir, "asset"))
img_eval = remove_dir_if_exist(os.path.join(asset_dir, "img_eval"),
ask_4_permission=False)
img_eval = make_dir_if_not_exist(img_eval)
img_x_rec = make_dir_if_not_exist(os.path.join(img_eval, "x_rec"))
img_z_rand_2_traversal = make_dir_if_not_exist(
os.path.join(img_eval, "z_rand_2_traversal"))
img_z_cond_all_traversal = make_dir_if_not_exist(
os.path.join(img_eval, "z_cond_all_traversal"))
img_z_cond_1_traversal = make_dir_if_not_exist(
os.path.join(img_eval, "z_cond_1_traversal"))
img_z_corr = make_dir_if_not_exist(os.path.join(img_eval, "z_corr"))
img_z_dist = make_dir_if_not_exist(os.path.join(img_eval, "z_dist"))
img_z_stat_dist = make_dir_if_not_exist(
os.path.join(img_eval, "z_stat_dist"))
img_rec_error_dist = make_dir_if_not_exist(
os.path.join(img_eval, "rec_error_dist"))
model_dir = make_dir_if_not_exist(os.path.join(args.output_dir,
"model_tf"))
train_helper = SimpleTrainHelper(log_dir=None, save_dir=model_dir)
# =====================================
# =====================================
# Training Loop
# =====================================
config_proto = tf.ConfigProto(allow_soft_placement=True)
config_proto.gpu_options.allow_growth = True
config_proto.gpu_options.per_process_gpu_memory_fraction = 0.9
sess = tf.Session(config=config_proto)
# Load model
train_helper.load(sess, load_step=args.load_step)
#'''
# Reconstruction
# ======================================= #
seed = 389
x = x_train[np.arange(seed, seed + 64)]
img_file = os.path.join(img_x_rec, 'x_rec_train.png')
model.reconstruct_images(img_file,
sess,
x,
block_shape=[8, 8],
batch_size=-1,
dec_output_2_img_func=binary_float_to_uint8)
# ======================================= #
# z random/conditional traversal
# ======================================= #
# Plot z cont with z cont
z_zero = np.zeros([args.z_dim], dtype=np.float32)
z_rand = np.random.randn(args.z_dim)
z_start, z_stop = -4, 4
num_points = 8
for i in range(args.z_dim):
for j in range(i + 1, args.z_dim):
print("Plot random 2 comps z traverse with {} and {} components!".
format(i, j))
img_file = os.path.join(img_z_rand_2_traversal,
'z[{},{},zero].png'.format(i, j))
model.rand_2_latents_traverse(
img_file,
sess,
default_z=z_zero,
z_comp1=i,
start1=z_start,
stop1=z_stop,
num_points1=num_points,
z_comp2=j,
start2=z_start,
stop2=z_stop,
num_points2=num_points,
batch_size=args.batch_size,
dec_output_2_img_func=binary_float_to_uint8)
img_file = os.path.join(img_z_rand_2_traversal,
'z[{},{},rand].png'.format(i, j))
model.rand_2_latents_traverse(
img_file,
sess,
default_z=z_rand,
z_comp1=i,
start1=z_start,
stop1=z_stop,
num_points1=num_points,
z_comp2=j,
start2=z_stop,
stop2=z_stop,
num_points2=num_points,
batch_size=args.batch_size,
dec_output_2_img_func=binary_float_to_uint8)
seed = 389
z_start, z_stop = -4, 4
num_itpl_points = 8
for n in range(seed, seed + 30):
print("Plot conditional all comps z traverse with test sample {}!".
format(n))
x = x_train[n]
img_file = os.path.join(img_z_cond_all_traversal,
'x_train{}.png'.format(n))
model.cond_all_latents_traverse(
img_file,
sess,
x,
start=z_start,
stop=z_stop,
num_itpl_points=num_itpl_points,
batch_size=args.batch_size,
dec_output_2_img_func=binary_float_to_uint8)
seed = 64
z_start, z_stop = -4, 4
num_itpl_points = 8
print("Plot conditional 1 comp z traverse!")
for i in range(args.z_dim):
x = x_train[seed:seed + 64]
img_file = os.path.join(
img_z_cond_1_traversal,
'x_train[{},{}]_z{}.png'.format(seed, seed + 64, i))
model.cond_1_latent_traverse(
img_file,
sess,
x,
z_comp=i,
start=z_start,
stop=z_stop,
num_itpl_points=num_itpl_points,
batch_size=args.batch_size,
dec_output_2_img_func=binary_float_to_uint8)
# ======================================= #
# '''
# z correlation matrix
# ======================================= #
data = x_train
all_z = []
for batch_ids in iterate_data(len(data), args.batch_size, shuffle=False):
x = data[batch_ids]
z = model.encode(sess, x)
assert len(
z.shape) == 2 and z.shape[1] == args.z_dim, "z.shape: {}".format(
z.shape)
all_z.append(z)
all_z = np.concatenate(all_z, axis=0)
print("Start plotting!")
plot_corrmat_with_histogram(os.path.join(img_z_corr, "corr_mat.png"),
all_z)
plot_comp_dist(os.path.join(img_z_dist, 'z_{}'), all_z, x_lim=(-5, 5))
print("Done!")
# ======================================= #
# '''
# z gaussian stddev
# ======================================= #
print("\nPlot z mean and stddev!")
data = x_train
all_z_mean = []
all_z_stddev = []
for batch_ids in iterate_data(len(data), args.batch_size, shuffle=False):
x = data[batch_ids]
z_mean, z_stddev = sess.run(model.get_output(['z_mean', 'z_stddev']),
feed_dict={
model.is_train: False,
model.x_ph: x
})
all_z_mean.append(z_mean)
all_z_stddev.append(z_stddev)
all_z_mean = np.concatenate(all_z_mean, axis=0)
all_z_stddev = np.concatenate(all_z_stddev, axis=0)
plot_comp_dist(os.path.join(img_z_stat_dist, 'z_mean_{}.png'),
all_z_mean,
x_lim=(-5, 5))
plot_comp_dist(os.path.join(img_z_stat_dist, 'z_stddev_{}.png'),
all_z_stddev,
x_lim=(0, 3))
# ======================================= #
# '''
# Decoder sensitivity
# ======================================= #
z_start = -3
z_stop = 3
for i in range(args.z_dim):
print("\nPlot rec error distribution for z component {}!".format(i))
all_z1 = np.array(all_z, copy=True, dtype=np.float32)
all_z2 = np.array(all_z, copy=True, dtype=np.float32)
all_z1[:, i] = z_start
all_z2[:, i] = z_stop
all_x_rec1 = []
all_x_rec2 = []
for batch_ids in iterate_data(len(x_train),
args.batch_size,
shuffle=False):
z1 = all_z1[batch_ids]
z2 = all_z2[batch_ids]
x1 = model.decode(sess, z1)
x2 = model.decode(sess, z2)
all_x_rec1.append(x1)
all_x_rec2.append(x2)
all_x_rec1 = np.concatenate(all_x_rec1, axis=0)
all_x_rec2 = np.concatenate(all_x_rec2, axis=0)
rec_errors = np.sum(np.reshape((all_x_rec1 - all_x_rec2)**2,
[len(x_train), 28 * 28]),
axis=1)
plot_comp_dist(
os.path.join(
img_rec_error_dist,
'rec_error[zi={},{},{}].png'.format(i, z_start, z_stop)),
rec_errors)
def plot_JEMMIG_quan_tc_beta(save_dir, JEMMIG_quan_result_files, labels):
JEMMIGs_all = []
JEMMIGs_by_tc = {}
JEMMIGs_by_beta = {}
JEMMIGs_by_Gz = {}
for i in range(len(JEMMIG_quan_result_files)):
JEMMIG_results = np.load(JEMMIG_quan_result_files[i], "r")
MI_z_y_sorted = JEMMIG_results['MI_z_y_sorted']
H_z_y_sorted = JEMMIG_results['H_z_y_sorted']
JEMMIG = np.mean(H_z_y_sorted[0, :] - MI_z_y_sorted[0, :] +
MI_z_y_sorted[1, :],
axis=0)
JEMMIGs_all.append(JEMMIG)
idx = labels[i].find('tc')
if idx < 0:
idx = labels[i].find('beta')
if idx < 0:
idx = labels[i].find('Gz')
assert idx >= 0, "labels[{}]='{}'".format(i, labels[i])
Gz = int(labels[i][idx + len('Gz'):idx + len('Gz') + 2])
jemmig_list = JEMMIGs_by_Gz.get(Gz)
if jemmig_list is None:
JEMMIGs_by_Gz[Gz] = [JEMMIG]
else:
jemmig_list.append(JEMMIG)
else:
beta = int(labels[i][idx + len('beta'):])
jemmig_list = JEMMIGs_by_beta.get(beta)
if jemmig_list is None:
JEMMIGs_by_beta[beta] = [JEMMIG]
else:
jemmig_list.append(JEMMIG)
else:
tc = int(labels[i][idx + len('tc'):])
jemmig_list = JEMMIGs_by_tc.get(tc)
if jemmig_list is None:
JEMMIGs_by_tc[tc] = [JEMMIG]
else:
jemmig_list.append(JEMMIG)
tc_list = ["{}".format(tc) for tc, _ in iteritems(JEMMIGs_by_tc)]
JEMMIGs_mean_by_tc = [(tc, np.mean(jemmig_list))
for tc, jemmig_list in iteritems(JEMMIGs_by_tc)]
JEMMIGs_std_by_tc = [(tc, np.std(jemmig_list))
for tc, jemmig_list in iteritems(JEMMIGs_by_tc)]
beta_list = ["{}".format(beta) for beta, _ in iteritems(JEMMIGs_by_beta)]
JEMMIGs_mean_by_beta = [(beta, np.mean(jemmig_list))
for beta, jemmig_list in iteritems(JEMMIGs_by_beta)
]
JEMMIGs_std_by_beta = [(beta, np.std(jemmig_list))
for beta, jemmig_list in iteritems(JEMMIGs_by_beta)]
Gz_list = ["{}".format(Gz) for Gz, _ in iteritems(JEMMIGs_by_Gz)]
JEMMIGs_mean_by_Gz = [(Gz, np.mean(jemmig_list))
for Gz, jemmig_list in iteritems(JEMMIGs_by_Gz)]
JEMMIGs_std_by_Gz = [(Gz, np.std(jemmig_list))
for Gz, jemmig_list in iteritems(JEMMIGs_by_Gz)]
# Plotting RMIG-MIG relationship
# =========================================== #
font = {'family': 'normal', 'size': 12}
matplotlib.rc('font', **font)
width = 0.5
plt.bar(range(0, len(tc_list)), [a[1] for a in JEMMIGs_mean_by_tc],
yerr=[a[1] for a in JEMMIGs_std_by_tc],
width=width,
align='center',
label="TC")
plt.bar(range(len(tc_list),
len(beta_list) + len(tc_list)),
[a[1] for a in JEMMIGs_mean_by_beta],
yerr=[a[1] for a in JEMMIGs_std_by_beta],
width=width,
align='center',
label="Beta")
plt.bar(range(
len(beta_list) + len(tc_list),
len(beta_list) + len(tc_list) + len(Gz_list)),
[a[1] for a in JEMMIGs_mean_by_Gz],
yerr=[a[1] for a in JEMMIGs_std_by_Gz],
width=width,
align='center',
label="Gz")
plt.xticks(range(0,
len(tc_list) + len(beta_list) + len(Gz_list)),
tc_list + beta_list + Gz_list)
plt.legend()
plt.xlabel("model")
plt.ylabel("JEMMIG (quantization)")
plt.ylim(bottom=3)
# plt.tight_layout()
subplot_adjust = {'left': 0.08, 'right': 0.99, 'bottom': 0.17, 'top': 0.98}
plt.subplots_adjust(**subplot_adjust)
plt.gcf().set_size_inches(9, 3.2)
save_dir = make_dir_if_not_exist(save_dir)
save_file = join(save_dir, "JEMMIG_tc_beta_Gz.pdf")
with PdfPages(save_file) as pdf_file:
plt.savefig(pdf_file, format='pdf')
plt.show()
plt.close()
def plot_RMIG_tc_beta(save_dir, JEMMIG_result_files, labels):
RMIGs_all = []
RMIGs_by_tc = {}
RMIGs_by_beta = {}
RMIGs_by_Gz = {}
for i in range(len(JEMMIG_result_files)):
RMIG_results = np.load(JEMMIG_result_files[i], "r")
RMIGs_all.append(np.mean(RMIG_results['RMIG_yk']))
idx = labels[i].find('tc')
if idx < 0:
idx = labels[i].find('beta')
if idx < 0:
idx = labels[i].find('Gz')
assert idx >= 0, "labels[{}]='{}'".format(i, labels[i])
Gz = int(labels[i][idx + len('Gz'):idx + len('Gz') + 2])
rmig_list = RMIGs_by_Gz.get(Gz)
if rmig_list is None:
RMIGs_by_Gz[Gz] = [np.mean(RMIG_results['RMIG_yk'])]
else:
rmig_list.append(np.mean(RMIG_results['RMIG_yk']))
else:
beta = int(labels[i][idx + len('beta'):])
rmig_list = RMIGs_by_beta.get(beta)
if rmig_list is None:
RMIGs_by_beta[beta] = [np.mean(RMIG_results['RMIG_yk'])]
else:
rmig_list.append(np.mean(RMIG_results['RMIG_yk']))
else:
tc = int(labels[i][idx + len('tc'):])
rmig_list = RMIGs_by_tc.get(tc)
if rmig_list is None:
RMIGs_by_tc[tc] = [np.mean(RMIG_results['RMIG_yk'])]
else:
rmig_list.append(np.mean(RMIG_results['RMIG_yk']))
tc_list = ["{}".format(tc) for tc, _ in iteritems(RMIGs_by_tc)]
RMIGs_mean_by_tc = [(tc, np.mean(rmig_list))
for tc, rmig_list in iteritems(RMIGs_by_tc)]
RMIGs_std_by_tc = [(tc, np.std(rmig_list))
for tc, rmig_list in iteritems(RMIGs_by_tc)]
beta_list = ["{}".format(beta) for beta, _ in iteritems(RMIGs_by_beta)]
RMIGs_mean_by_beta = [(beta, np.mean(rmig_list))
for beta, rmig_list in iteritems(RMIGs_by_beta)]
RMIGs_std_by_beta = [(beta, np.std(rmig_list))
for beta, rmig_list in iteritems(RMIGs_by_beta)]
Gz_list = ["{}".format(Gz) for Gz, _ in iteritems(RMIGs_by_Gz)]
RMIGs_mean_by_Gz = [(Gz, np.mean(rmig_list))
for Gz, rmig_list in iteritems(RMIGs_by_Gz)]
RMIGs_std_by_Gz = [(Gz, np.std(rmig_list))
for Gz, rmig_list in iteritems(RMIGs_by_Gz)]
print("RMIGs_by_tc: {}".format(RMIGs_by_tc))
print("RMIGs_by_beta: {}".format(RMIGs_by_beta))
# Plotting RMIG-MIG relationship
# =========================================== #
font = {'family': 'normal', 'size': 12}
matplotlib.rc('font', **font)
width = 0.5
plt.bar(range(0, len(tc_list)), [a[1] for a in RMIGs_mean_by_tc],
yerr=[a[1] for a in RMIGs_std_by_tc],
width=width,
align='center',
label="TC")
plt.bar(range(len(tc_list),
len(beta_list) + len(tc_list)),
[a[1] for a in RMIGs_mean_by_beta],
yerr=[a[1] for a in RMIGs_std_by_beta],
width=width,
align='center',
label="Beta")
plt.bar(range(
len(beta_list) + len(tc_list),
len(beta_list) + len(tc_list) + len(Gz_list)),
[a[1] for a in RMIGs_mean_by_Gz],
yerr=[a[1] for a in RMIGs_std_by_Gz],
width=width,
align='center',
label="Gz")
plt.xticks(range(0,
len(tc_list) + len(beta_list) + len(Gz_list)),
tc_list + beta_list + Gz_list)
plt.legend()
plt.xlabel("model")
plt.ylabel("RMIG")
# plt.tight_layout()
subplot_adjust = {'left': 0.11, 'right': 0.99, 'bottom': 0.17, 'top': 0.98}
plt.subplots_adjust(**subplot_adjust)
plt.gcf().set_size_inches(9, 3)
save_dir = make_dir_if_not_exist(save_dir)
save_file = join(save_dir, "RMIG_tc_beta_Gz.pdf")
with PdfPages(save_file) as pdf_file:
plt.savefig(pdf_file, format='pdf')
plt.show()
plt.close()
def main(args):
# =====================================
# Load config
# =====================================
with open(join(args.output_dir, 'config.json')) as f:
config = json.load(f)
args.__dict__.update(config)
# =====================================
# Dataset
# =====================================
data_file = join(RAW_DATA_DIR, "ComputerVision", "dSprites",
"dsprites_ndarray_co1sh3sc6or40x32y32_64x64.npz")
# It is already in the range [0, 1]
with np.load(data_file, encoding="latin1") as f:
x_train = f['imgs']
x_train = np.reshape(x_train, [3, 6, 40, 32, 32, 64, 64, 1])
# =====================================
# Instantiate model
# =====================================
if args.enc_dec_model == "1Konny":
encoder = Encoder_1Konny(args.z_dim, stochastic=True)
decoder = Decoder_1Konny()
disc_z = DiscriminatorZ_1Konny(num_outputs=2)
else:
raise ValueError("Do not support enc_dec_model='{}'!".format(
args.enc_dec_model))
model = FactorVAE([64, 64, 1],
args.z_dim,
encoder=encoder,
decoder=decoder,
discriminator_z=disc_z,
rec_x_mode=args.rec_x_mode,
use_gp0_z_tc=True,
gp0_z_tc_mode=args.gp0_z_tc_mode)
loss_coeff_dict = {
'rec_x': args.rec_x_coeff,
'kld_loss': args.kld_loss_coeff,
'tc_loss': args.tc_loss_coeff,
'gp0_z_tc': args.gp0_z_tc_coeff,
'Dz_tc_loss_coeff': args.Dz_tc_loss_coeff,
}
model.build(loss_coeff_dict)
SimpleParamPrinter.print_all_params_tf_slim()
# =====================================
# Load model
# =====================================
config_proto = tf.ConfigProto(allow_soft_placement=True)
config_proto.gpu_options.allow_growth = True
config_proto.gpu_options.per_process_gpu_memory_fraction = 0.9
sess = tf.Session(config=config_proto)
model_dir = make_dir_if_not_exist(join(args.output_dir, "model_tf"))
train_helper = SimpleTrainHelper(log_dir=None, save_dir=model_dir)
# Load model
train_helper.load(sess, load_step=args.load_step)
# =====================================
# Experiments
save_dir = make_dir_if_not_exist(
join(args.save_dir, "{}_{}".format(args.enc_dec_model, args.run)))
# =====================================
np.set_printoptions(threshold=np.nan,
linewidth=1000,
precision=3,
suppress=True)
result_file = join(args.JEMMIG_sampling_dir,
"{}_{}".format(args.enc_dec_model, args.run),
"results[num_samples={}].npz".format(args.num_samples))
results = np.load(result_file, "r")
print("results.keys: {}".format(list(results.keys())))
# Plotting
# =========================================== #
data = [
x_train[0, 3, 20, 16, 16], x_train[1, 3, 20, 16, 16], x_train[2, 3, 20,
16, 16]
]
gt_factors = ['Shape', 'Scale', 'Rotation', 'Pos_x', 'Pos_y']
ids_sorted = results['id_sorted']
MI_zi_yk_sorted = results['MI_zi_yk_sorted']
H_zi_yk_sorted = results['H_zi_yk_sorted']
H_yk = results['H_yk']
RMIG_yk = results['RMIG_yk']
RMIG_norm_yk = results['RMIG_norm_yk']
JEMMIG_yk = results['JEMMIG_yk']
print("MI_zi_yk_sorted:\n{}".format(MI_zi_yk_sorted))
print("\nShow RMIG!")
for k in range(len(gt_factors)):
print(
"{}, RMIG: {:.4f}, RMIG (norm): {:.4f}, H: {:.4f}, I1: {:.4f}, I2: {:.4f}"
.format(gt_factors[k], RMIG_yk[k], RMIG_norm_yk[k], H_yk[k],
MI_zi_yk_sorted[0, k], MI_zi_yk_sorted[1, k]))
print("\nShow JEMMIG!")
for k in range(len(gt_factors)):
print(
"{}, JEMMIG: {:.4f}, H1: {:.4f}, H1-I1: {:.4f}, I2: {:.4f}, top2 ids: z{}, z{}"
.format(gt_factors[k], JEMMIG_yk[k], H_zi_yk_sorted[0, k],
H_zi_yk_sorted[0, k] - MI_zi_yk_sorted[0, k],
MI_zi_yk_sorted[1, k], ids_sorted[0, k], ids_sorted[1, k]))
span = 3
points_one_side = 5
for n in range(len(data)):
for k in range(len(gt_factors)):
print("x={}, y={}!".format(n, gt_factors[k]))
img_file = join(
save_dir,
"{}-x{}_num_samples={}].png".format(gt_factors[k], n,
args.num_samples))
ids_top3 = ids_sorted[:3, k]
MI_top3 = MI_zi_yk_sorted[:3, k]
model.cond_all_latents_traverse_v2(
img_file,
sess,
data[n],
z_comps=ids_top3,
z_comp_labels=[
"z[{}] ({:.4f})".format(comp, mi)
for comp, mi in zip(ids_top3, MI_top3)
],
span=span,
points_1_side=points_one_side,
hl_x=True,
font_size=9,
title="{} (RMIG={:.4f}, JEMMIG={:.4f}, H={:.4f})".format(
gt_factors[k], RMIG_yk[k], JEMMIG_yk[k], H_yk[k]),
title_font_scale=1.5,
subplot_adjust={
'left': 0.16,
'right': 0.99,
'bottom': 0.01,
'top': 0.88
},
size_inches=(6.2, 1.7),
batch_size=args.batch_size,
dec_output_2_img_func=binary_float_to_uint8)
f.close()
def main(args):
# =====================================
# Load config
# =====================================
with open(join(args.output_dir, 'config.json')) as f:
config = json.load(f)
args.__dict__.update(config)
# =====================================
# Dataset
# =====================================
celebA_loader = TFCelebALoader(root_dir=args.celebA_root_dir)
img_height, img_width = args.celebA_resize_size, args.celebA_resize_size
celebA_loader.build_transformation_flow_tf(
*celebA_loader.get_transform_fns("1Konny", resize_size=args.celebA_resize_size))
# =====================================
# Instantiate model
# =====================================
if args.activation == "relu":
activation = tf.nn.relu
elif args.activation == "leaky_relu":
activation = tf.nn.leaky_relu
else:
raise ValueError("Do not support '{}' activation!".format(args.activation))
if args.enc_dec_model == "1Konny":
assert args.z_dim == 65, "For 1Konny, z_dim must be 65. Found {}!".format(args.z_dim)
encoder = Encoder_1Konny(args.z_dim, stochastic=True, activation=activation)
decoder = Decoder_1Konny([img_height, img_width, 3], activation=activation,
output_activation=tf.nn.sigmoid)
disc_z = DiscriminatorZ_1Konny(num_outputs=2)
elif args.enc_dec_model == "my":
assert args.z_dim == 150, "For 1Konny, z_dim must be 150. Found {}!".format(args.z_dim)
encoder = Encoder_My(args.z_dim, stochastic=True, activation=activation)
decoder = Decoder_My([img_height, img_width, 3], activation=activation,
output_activation=tf.nn.sigmoid)
disc_z = DiscriminatorZ_My(num_outputs=2)
else:
raise ValueError("Do not support encoder/decoder model '{}'!".format(args.enc_dec_model))
model = FactorVAE([img_height, img_width, 3], args.z_dim,
encoder=encoder, decoder=decoder,
discriminator_z=disc_z,
rec_x_mode=args.rec_x_mode,
use_gp0_z_tc=True, gp0_z_tc_mode=args.gp0_z_tc_mode)
loss_coeff_dict = {
'rec_x': args.rec_x_coeff,
'kld_loss': args.kld_loss_coeff,
'tc_loss': args.tc_loss_coeff,
'gp0_z_tc': args.gp0_z_tc_coeff,
}
model.build(loss_coeff_dict)
SimpleParamPrinter.print_all_params_tf_slim()
# =====================================
# Load model
# =====================================
config_proto = tf.ConfigProto(allow_soft_placement=True)
config_proto.gpu_options.allow_growth = True
config_proto.gpu_options.per_process_gpu_memory_fraction = 0.9
sess = tf.Session(config=config_proto)
model_dir = make_dir_if_not_exist(join(args.output_dir, "model_tf"))
train_helper = SimpleTrainHelper(log_dir=None, save_dir=model_dir)
# Load model
train_helper.load(sess, load_step=args.load_step)
# =====================================
# Experiments
save_dir = remove_dir_if_exist(join(args.save_dir, "FactorVAE_{}".format(args.run)), ask_4_permission=False)
save_dir = make_dir_if_not_exist(save_dir)
# =====================================
# z correlation matrix
# ======================================= #
for deterministic in [True, False]:
all_z = []
for batch_ids in iterate_data(celebA_loader.num_train_data, args.batch_size, shuffle=False):
x = celebA_loader.sample_images_from_dataset(sess, 'train', batch_ids)
z = model.encode(sess, x, deterministic=deterministic)
assert len(z.shape) == 2 and z.shape[1] == args.z_dim, "z.shape: {}".format(z.shape)
all_z.append(z)
all_z = np.concatenate(all_z, axis=0)
# plot_corrmat(join(save_dir, "corr_mat[deter={}].png".format(deterministic)), all_z,
# font={'size': 14},
# subplot_adjust={'left': 0.04, 'right': 0.96, 'bottom': 0.02, 'top': 0.98},
# size_inches=(7.2, 6))
plot_corrmat_with_histogram(join(save_dir, "corr_mat_hist[deter={}].png".format(deterministic)), all_z,
font={'size': 14},
subplot_adjust={'left': 0.04, 'right': 0.96, 'bottom': 0.02, 'top': 0.98},
size_inches=(10, 3))
