Updated copyright notice

ImageSelector.py

@@ -24,7 +24,6 @@
 import time
 import random
 import os
-from model import Model
 from net import *
 from checkpointer import Checkpointer
 from scheduler import ComboMultiStepLR

checkpointer.py

@@ -1,4 +1,4 @@
-# Copyright 2019 Stanislav Pidhorskyi
+# Copyright 2019-2020 Stanislav Pidhorskyi
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.

custom_adam.py

@@ -1,3 +1,19 @@
+# Copyright 2019-2020 Stanislav Pidhorskyi
+# lr_equalization_coef was added for LREQ
+
+# Copyright (c) 2016-     Facebook, Inc            (Adam Paszke)
+# Copyright (c) 2014-     Facebook, Inc            (Soumith Chintala)
+# Copyright (c) 2011-2014 Idiap Research Institute (Ronan Collobert)
+# Copyright (c) 2012-2014 Deepmind Technologies    (Koray Kavukcuoglu)
+# Copyright (c) 2011-2012 NEC Laboratories America (Koray Kavukcuoglu)
+# Copyright (c) 2011-2013 NYU                      (Clement Farabet)
+# Copyright (c) 2006-2010 NEC Laboratories America (Ronan Collobert, Leon Bottou, Iain Melvin, Jason Weston)
+# Copyright (c) 2006      Idiap Research Institute (Samy Bengio)
+# Copyright (c) 2001-2004 Idiap Research Institute (Ronan Collobert, Samy Bengio, Johnny Mariethoz)
+
+# https://github.com/pytorch/pytorch/blob/master/LICENSE
+
+
 import math
 import torch
 from torch.optim.optimizer import Optimizer

dataloader.py

@@ -1,4 +1,4 @@
-# Copyright 2019 Stanislav Pidhorskyi
+# Copyright 2019-2020 Stanislav Pidhorskyi
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -13,25 +13,17 @@
 # limitations under the License.
 # ==============================================================================
 
-import pickle
 import dareblopy as db
-from threading import Thread, Lock, Event
 import random
-import threading
 
 import numpy as np
 import torch
 import torch.tensor
 import torch.utils
 import torch.utils.data
 import time
-import torchvision.transforms.functional as vF
-from torchvision import transforms
 import math
 
-from dlutils.batch_provider import batch_provider
-from dlutils.shuffle import shuffle_ndarray
-
 cpu = torch.device('cpu')
 
 

defaults.py

@@ -1,4 +1,4 @@
-# Copyright 2019 Stanislav Pidhorskyi
+# Copyright 2019-2020 Stanislav Pidhorskyi
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.

fid.py

@@ -1,3 +1,5 @@
+# Copyright 2019-2020 Stanislav Pidhorskyi
+#
 # Copyright (c) 2019, NVIDIA CORPORATION. All rights reserved.
 #
 # This work is licensed under the Creative Commons Attribution-NonCommercial
@@ -10,31 +12,18 @@
 import numpy as np
 import tensorflow as tf
 import torch
-import dnnlib
 import dnnlib.tflib
-import dnnlib.tflib as tflib
 import pickle
 from net import *
-from checkpointer import Checkpointer
-from scheduler import ComboMultiStepLR
 from model_z_gan import Model
 from launcher import run
-from defaults import get_cfg_defaults
-import lod_driver
 from dataloader import *
 import scipy.linalg
 
 from checkpointer import Checkpointer
-from scheduler import ComboMultiStepLR
 
-from dlutils import batch_provider
-from dlutils.pytorch.cuda_helper import *
 from dlutils.pytorch import count_parameters
 from defaults import get_cfg_defaults
-import argparse
-import logging
-import sys
-import lreq
 from skimage.transform import resize
 from tqdm import tqdm
 

fid_rec.py

@@ -1,3 +1,5 @@
+# Copyright 2019-2020 Stanislav Pidhorskyi
+#
 # Copyright (c) 2019, NVIDIA CORPORATION. All rights reserved.
 #
 # This work is licensed under the Creative Commons Attribution-NonCommercial
@@ -10,31 +12,18 @@
 import numpy as np
 import tensorflow as tf
 import torch
-import dnnlib
 import dnnlib.tflib
-import dnnlib.tflib as tflib
 import pickle
 from net import *
-from checkpointer import Checkpointer
-from scheduler import ComboMultiStepLR
 from model_z_gan import Model
 from launcher import run
-from defaults import get_cfg_defaults
-import lod_driver
 from dataloader import *
 import scipy.linalg
 
 from checkpointer import Checkpointer
-from scheduler import ComboMultiStepLR
 
-from dlutils import batch_provider
-from dlutils.pytorch.cuda_helper import *
 from dlutils.pytorch import count_parameters
 from defaults import get_cfg_defaults
-import argparse
-import logging
-import sys
-import lreq
 from skimage.transform import resize
 from tqdm import tqdm
 
@@ -43,7 +32,7 @@
 import utils
 
 dnnlib.tflib.init_tf()
-tf_config     = {'rnd.np_random_seed': 1000}
+tf_config = {'rnd.np_random_seed': 1000}
 
 
 class FID:

make_recon_figure_2.py

@@ -24,7 +24,6 @@
 import time
 import random
 import os
-from model import Model
 from net import *
 from checkpointer import Checkpointer
 from scheduler import ComboMultiStepLR

net.py

@@ -1,4 +1,4 @@
-# Copyright 2019 Stanislav Pidhorskyi
+# Copyright 2019-2020 Stanislav Pidhorskyi
 # 
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -20,30 +20,11 @@
 from torch.nn import init
 from torch.nn.parameter import Parameter
 import numpy as np
-from dlutils.pytorch import count_parameters
 import lreq as ln
 import math
 from registry import *
 
 
-if False:
-    def lerp(s, e, x):
-        return s + (e - s) * x
-
-
-    def rsqrt(x):
-        return 1.0 / x ** 0.5
-
-
-    def addcmul(x, value, tensor1, tensor2):
-        return x + value * tensor1 * tensor2
-
-
-    torch.lerp = lerp
-    torch.rsqrt = rsqrt
-    torch.addcmul = addcmul
-
-
 def pixel_norm(x, epsilon=1e-8):
     return x * torch.rsqrt(torch.mean(x.pow(2.0), dim=1, keepdim=True) + epsilon)
 
@@ -110,9 +91,6 @@ def __init__(self, inputs, outputs, latent_size, last=False, fused_scale=True):
             self.bias_2.zero_()
 
     def forward(self, x):
-        # if self.last:
-        #     x = minibatch_stddev_layer(x)
-
         x = self.conv_1(x) + self.bias_1
         x = F.leaky_relu(x, 0.2)
 
@@ -795,18 +773,6 @@ def get_statistics(self, lod):
         return rgb_std / rgb_std_c, layers
 
 
-def minibatch_stddev_layer(x, group_size=4):
-    group_size = min(group_size, x.shape[0])
-    size = x.shape[0]
-    if x.shape[0] % group_size != 0:
-        x = torch.cat([x, x[:(group_size - (x.shape[0] % group_size)) % group_size]])
-    y = x.view(group_size, -1, x.shape[1], x.shape[2], x.shape[3])
-    y = y - y.mean(dim=0, keepdim=True)
-    y = torch.sqrt((y ** 2).mean(dim=0) + 1e-8).mean(dim=[1, 2, 3], keepdim=True)
-    y = y.repeat(group_size, 1, x.shape[2], x.shape[3])
-    return torch.cat([x, y], dim=1)[:size]
-
-
 image_size = 64
 
 # Number of channels in the training images. For color images this is 3
@@ -1003,8 +969,6 @@ def encode(self, x, lod):
         x = F.interpolate(x, 28)
         x = x.view(x.shape[0], 28 * 28)
 
-        # styles = torch.zeros(x.shape[0], 1, self.latent_size)
-
         x = self.fc_1(x)
         x = F.leaky_relu(x, 0.2)
         x = self.fc_2(x)
@@ -1039,8 +1003,6 @@ def decode(self, x, lod, blend_factor, noise):
             x = x[:, 0]  # no styles
         x.view(x.shape[0], self.latent_size)
 
-        # styles = torch.zeros(x.shape[0], 1, self.latent_size)
-
         x = self.fc_1(x)
         x = F.leaky_relu(x, 0.2)
         x = self.fc_2(x)

ppl.py

@@ -1,3 +1,5 @@
+# Copyright 2019 Stanislav Pidhorskyi
+#
 # Copyright (c) 2019, NVIDIA CORPORATION. All rights reserved.
 #
 # This work is licensed under the Creative Commons Attribution-NonCommercial
@@ -10,39 +12,23 @@
 import numpy as np
 import tensorflow as tf
 import torch
-import dnnlib
 import dnnlib.tflib
-import dnnlib.tflib as tflib
 import pickle
 from net import *
 from checkpointer import Checkpointer
-from scheduler import ComboMultiStepLR
-#from model_z_gan import Model
 from model_z_gan import Model
 from launcher import run
-from defaults import get_cfg_defaults
-import lod_driver
-
-
-from checkpointer import Checkpointer
-from scheduler import ComboMultiStepLR
-
-from dlutils import batch_provider
 from dlutils.pytorch.cuda_helper import *
 from dlutils.pytorch import count_parameters
 from defaults import get_cfg_defaults
-import argparse
-import logging
-import sys
-import lreq
 from skimage.transform import resize
 import tqdm
 
 from PIL import Image
 from matplotlib import pyplot as plt
 
 dnnlib.tflib.init_tf()
-tf_config     = {'rnd.np_random_seed': 1000}
+tf_config = {'rnd.np_random_seed': 1000}
 
 
 # Normalize batch of vectors.