Upload 3 files

demo.py

@@ -9,7 +9,7 @@ from PIL import Image
 
 from matplotlib import pyplot as plt
 
-os.environ['CUDA_VISIBLE_DEVICES'] = '0'
+# os.environ['CUDA_VISIBLE_DEVICES'] = '0'  # Removed for CPU compatibility
 
 os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
 

main.py

@@ -9,7 +9,7 @@ tf.disable_v2_behavior()
 import imageio
 from PIL import Image
 
-os.environ['CUDA_VISIBLE_DEVICES'] = GPU_ID
+# os.environ['CUDA_VISIBLE_DEVICES'] = GPU_ID  # Removed for CPU compatibility
 
 os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
 

net.py

@@ -86,8 +86,7 @@ class Network(object):
 		s = np.sqrt(1. /fan_in)
 
 		# create variable and specific GPU device
-		with tf.device('/device:GPU:'+GPU_ID):
-			w = tf.get_variable(name, shape, dtype=self.dtype,
+		w = tf.get_variable(name, shape, dtype=self.dtype,
 							initializer=tf.random_uniform_initializer(minval=-s, maxval=s),
 							regularizer=regularizer, trainable=trainable)
 
@@ -96,8 +95,7 @@ class Network(object):
 	def _constant(self, shape, value=0, regularizer=None, trainable=None, name=None):
 		name = 'b' if name is None else name+'/b'
 
-		with tf.device('/device:GPU:'+GPU_ID):
-			b = tf.get_variable(name, shape, dtype=self.dtype,
+		b = tf.get_variable(name, shape, dtype=self.dtype,
 							initializer=tf.constant_initializer(value=value),
 							regularizer=regularizer, trainable=trainable)
 
@@ -137,11 +135,10 @@ class Network(object):
 			x = (x - mean) / tf.sqrt(var + 1e-6)
 
 			# per channel gamma and beta
-			with tf.device('/device:GPU:'+GPU_ID):
-				gamma = tf.get_variable(name+'/gamma', [C], dtype=self.dtype, initializer=tf.constant_initializer(1.0))
-				beta = tf.get_variable(name+'/beta', [C], dtype=self.dtype, initializer=tf.constant_initializer(0.0))
-				gamma = tf.reshape(gamma, [1, C, 1, 1])
-				beta = tf.reshape(beta, [1, C, 1, 1])
+			gamma = tf.get_variable(name+'/gamma', [C], dtype=self.dtype, initializer=tf.constant_initializer(1.0))
+			beta = tf.get_variable(name+'/beta', [C], dtype=self.dtype, initializer=tf.constant_initializer(0.0))
+			gamma = tf.reshape(gamma, [1, C, 1, 1])
+			beta = tf.reshape(beta, [1, C, 1, 1])
 
 			tensor = tf.reshape(x, [-1, C, H, W]) * gamma + beta
 			# tranpose: [bs, c, h, w, c] to [bs, h, w, c] following the paper
@@ -197,18 +194,17 @@ class Network(object):
 	def _constant_kernel(self, shape, value=1.0, diag=False, flip=False, regularizer=None, trainable=None, name=None):
 		name = 'fixed_w' if name is None else name+'/fixed_w'
 
-		with tf.device('/device:GPU:'+GPU_ID):
-			if not diag:
-				k = tf.get_variable(name, shape, dtype=self.dtype,
+		if not diag:
+			k = tf.get_variable(name, shape, dtype=self.dtype,
 						initializer=tf.constant_initializer(value=value),
 						regularizer=regularizer, trainable=trainable)
-			else:
-				w = tf.eye(shape[0], num_columns=shape[1])
-				if flip:
-					w = tf.reshape(w, (shape[0], shape[1], 1))
-					w = tf.image.flip_left_right(w)
-				w = tf.reshape(w, shape)
-				k = tf.get_variable(name, None, dtype=self.dtype, # constant initializer dont specific shape
+		else:
+			w = tf.eye(shape[0], num_columns=shape[1])
+			if flip:
+				w = tf.reshape(w, (shape[0], shape[1], 1))
+				w = tf.image.flip_left_right(w)
+			w = tf.reshape(w, shape)
+			k = tf.get_variable(name, None, dtype=self.dtype, # constant initializer dont specific shape
 						initializer=w,
 						regularizer=regularizer, trainable=trainable)