code
stringlengths
3
6.57k
l2.theta.Transform(tf.identity)
l1.FProp(l1_theta, conv_in, conv_pad)
l2.FProp(l2_theta, conv_in, conv_pad)
tf.logging.info(l1_theta)
tf.logging.info(l2_theta)
l1_theta.Flatten()
l2_theta.Flatten()
len(l1_num_vars)
len(l2_num_var2)
len(l1_num_vars)
len(l2_num_var2)
tf.global_variables_initializer()
run()
sess.run([conv_out1, out1_padding])
sess.run(w1)
sess.run([conv_out2, out2_padding], feed_dict={w2: w1_v})
self.assertAllClose(v1, v2)
self.assertAllClose(p1, p2)
testConvBasic(self)
testConvBnWnTanh(self)
testConvGn(self)
self.session(use_gpu=True)
conv_layers_builder.Builder.Params()
bn_layers.GroupNormLayer.Params()
builder_params.Instantiate()
p.Instantiate()
tf.constant(np.random.normal(size=[4, 5, 6, 3])
np.full([4, 5], 0.0)
tf.constant(conv_pad, tf.float32)
l.FProp(l.theta, conv_in, conv_pad)
tf.global_variables_initializer()
run()
sess.run(tf.reduce_sum(conv_out, 0)
self.assertAllClose(v, expected_out)
testConvLastWnTanh(self)
testConvLastCausal(self)
self.session(use_gpu=True)
layers.DepthwiseConv2DLayer.Params()
conv_layers_builder.Builder.Params()
conv_layers_with_time_padding.ConvBatchNormLayer.Params()
builder_params.Instantiate()
p1.Instantiate()
p2.Instantiate()
tf.constant(np.random.normal(size=[4, 5, 6, 3])
np.full([4, 5], 0.0)
tf.constant(conv_pad, tf.float32)
l1.theta.Transform(tf.identity)
l2.theta.Transform(tf.identity)
l1.FProp(l1_theta, conv_in, conv_pad)
l2.FProp(l2_theta, conv_in, conv_pad)
tf.logging.info(l1_theta)
tf.logging.info(l2_theta)
l1_theta.Flatten()
l2_theta.Flatten()
len(l1_num_vars)
len(l2_num_var2)
len(l1_num_vars)
len(l2_num_var2)
tf.global_variables_initializer()
run()
sess.run([conv_out1, out1_padding])
sess.run([w1])
sess.run([conv_out2, out2_padding], feed_dict={w2: w1_v})
self.assertAllClose(v1, v2)
self.assertAllClose(p1, p2)
testDepthConvBasic(self)
testDepthConvBnWnTanh(self)
testDepthConvGn(self)
self.session(use_gpu=True)
conv_layers_builder.Builder.Params()
bn_layers.GroupNormLayer.Params()
builder_params.Instantiate()
p.Instantiate()
tf.constant(np.random.normal(size=[4, 5, 6, 4])
np.full([4, 5], 0.0)
tf.constant(conv_pad, tf.float32)
l.FProp(l.theta, conv_in, conv_pad)
tf.global_variables_initializer()
run()
sess.run(tf.reduce_sum(conv_out, 0)
self.assertAllClose(expected_out, v)
testDepthConvLastWnTanh(self)
testDepthConvLastCausal(self)
self.session(use_gpu=True)
layers.SeparableConv2DLayer.Params()
conv_layers_builder.Builder.Params()
conv_layers_with_time_padding.ConvBatchNormLayer.Params()
builder_params.Instantiate()
p1.Instantiate()
p2.Instantiate()
tf.constant(np.random.normal(size=[4, 5, 6, 3])
np.full([4, 5], 0.0)
tf.constant(conv_pad, tf.float32)
l1.theta.Transform(tf.identity)
l2.theta.Transform(tf.identity)
l1.FProp(l1_theta, conv_in, conv_pad)
l2.FProp(l2_theta, conv_in, conv_pad)
tf.logging.info(l1_theta)
tf.logging.info(l2_theta)
l1_theta.Flatten()