編集履歴

質問編集履歴

追記

2019/02/23 11:31

投稿

Reach

スコア733

test CHANGED Viewed

File without changes

test CHANGED Viewed

@@ -417,3 +417,13 @@
 netG = tf.contrib.tpu.keras_to_tpu_model(netG, strategy=strategy)
 ```
+追記：
+複数のモデルを 同時に ColabのTPUでは 登録できないみたいです？
+1つを TPUで 別のものは CPUで 実行させるのは
+果たして (すべてのモデルを) GPUで 走らせるのとでは どちらが 速いかも 微妙です

コード一部記載

2019/02/23 11:31

投稿

Reach

スコア733

test CHANGED Viewed

File without changes

test CHANGED Viewed

@@ -82,4 +82,338 @@
 改変コードが 字数制限で掲載できません
+```python
+import os
+import tensorflow as tf
+import tensorflow.keras.backend as K
+from tensorflow.contrib.tpu.python.tpu import keras_support
+K.set_image_data_format('channels_last')
-どうすれば よいのでしょうか？
+channel_axis=-1
+channel_first = False
+from tensorflow.keras.models import Sequential, Model
+from tensorflow.keras.layers import Conv2D, ZeroPadding2D, BatchNormalization, Input, Dropout
+from tensorflow.keras.layers import Conv2DTranspose, Reshape, Activation, Cropping2D, Flatten
+from tensorflow.keras.layers import Concatenate
+from tensorflow.keras.layers import LeakyReLU
+from tensorflow.keras.activations import relu
+from tensorflow.keras.initializers import RandomNormal
+def __conv_init(a):
+    print("conv_init", a)
+    k = RandomNormal(0, 0.02)(a)
+    k.conv_weight = True
+    return k
+conv_init = RandomNormal(0, 0.02)
+gamma_init = RandomNormal(1., 0.02)
+def conv2d(f, *a, **k):
+    return Conv2D(f, kernel_initializer = conv_init, *a, **k)
+def batchnorm():
+    return BatchNormalization(momentum=0.9, axis=channel_axis, epsilon=1.01e-5,
+                                   gamma_initializer = gamma_init)
+def BASIC_D(nc_in, nc_out, ndf, max_layers=3):
+    if channel_first:
+        input_a, input_b =  Input(shape=(nc_in, None, None)), Input(shape=(nc_out, None, None))
+    else:
+        input_a, input_b = Input(shape=(None, None, nc_in)), Input(shape=(None, None, nc_out))
+    _ = Concatenate(axis=channel_axis)([input_a, input_b])
+    _ = conv2d(ndf, kernel_size=4, strides=2, padding="same", name = 'First') (_)
+    _ = LeakyReLU(alpha=0.2)(_)
+    for layer in range(1, max_layers):
+        out_feat = ndf * min(2**layer, 8)
+        _ = conv2d(out_feat, kernel_size=4, strides=2, padding="same",
+                   use_bias=False, name = 'pyramid.{0}'.format(layer)
+                        ) (_)
+        _ = batchnorm()(_, training=1)
+        _ = LeakyReLU(alpha=0.2)(_)
+    out_feat = ndf*min(2**max_layers, 8)
+    _ = ZeroPadding2D(1)(_)
+    _ = conv2d(out_feat, kernel_size=4,  use_bias=False, name = 'pyramid_last') (_)
+    _ = batchnorm()(_, training=1)
+    _ = LeakyReLU(alpha=0.2)(_)
+    # final layer
+    _ = ZeroPadding2D(1)(_)
+    _ = conv2d(1, kernel_size=4, name = 'final'.format(out_feat, 1),
+               activation = "sigmoid") (_)
+    return Model(inputs=[input_a, input_b], outputs=_)
+def UNET_G(isize, nc_in=3, nc_out=3, ngf=64, fixed_input_size=True):
+    max_nf = 8*ngf
+    def block(x, s, nf_in, use_batchnorm=True, nf_out=None, nf_next=None):
+        assert s>=2 and s%2==0
+        if nf_next is None:
+            nf_next = min(nf_in*2, max_nf)
+        if nf_out is None:
+            nf_out = nf_in
+        x = conv2d(nf_next, kernel_size=4, strides=2, use_bias=(not (use_batchnorm and s>2)),
+                   padding="same", name = 'conv_{0}'.format(s)) (x)
+        if s>2:
+            if use_batchnorm:
+                x = batchnorm()(x, training=1)
+            x2 = LeakyReLU(alpha=0.2)(x)
+            x2 = block(x2, s//2, nf_next)
+            x = Concatenate(axis=channel_axis)([x, x2])
+        x = Activation("relu")(x)
+        x = Conv2DTranspose(nf_out, kernel_size=4, strides=2, use_bias=not use_batchnorm,
+                            kernel_initializer = conv_init,
+                            name = 'convt.{0}'.format(s))(x)
+        x = Cropping2D(1)(x)
+        if use_batchnorm:
+            x = batchnorm()(x, training=1)
+        if s <=8:
+            x = Dropout(0.5)(x, training=1)
+        return x
+    s = isize if fixed_input_size else None
+    if channel_first:
+        _ = inputs = Input(shape=(nc_in, s, s))
+    else:
+        _ = inputs = Input(shape=(s, s, nc_in))
+    _ = block(_, isize, nc_in, False, nf_out=nc_out, nf_next=ngf)
+    _ = Activation('tanh')(_)
+    return Model(inputs=inputs, outputs=[_])
+nc_in = 3
+nc_out = 3
+ngf = 64
+ndf = 64
+λ = 10
+loadSize = 256
+imageSize = 256
+batchSize = 1
+lrD = 2e-4
+lrG = 2e-4
+netD = BASIC_D(nc_in, nc_out, ndf)
+netD.summary()
+from IPython.display import SVG
+netG = UNET_G(imageSize, nc_in, nc_out, ngf)
+netG.summary()
+from tensorflow.keras.optimizers import RMSprop, SGD, Adam
+real_A = netG.input
+fake_B = netG.output
+netG_generate = K.function([real_A], [fake_B])
+real_B = netD.inputs[1]
+output_D_real = netD([real_A, real_B])
+output_D_fake = netD([real_A, fake_B])
+loss_fn = lambda output, target : -K.mean(K.log(output+1e-12)*target+K.log(1-output+1e-12)*(1-target))
+loss_D_real = loss_fn(output_D_real, K.ones_like(output_D_real))
+loss_D_fake = loss_fn(output_D_fake, K.zeros_like(output_D_fake))
+loss_G_fake = loss_fn(output_D_fake, K.ones_like(output_D_fake))
+loss_L1 = K.mean(K.abs(fake_B-real_B))
+loss_D = loss_D_real +loss_D_fake
+training_updates = Adam(lr=lrD, beta_1=0.5).get_updates(netD.trainable_weights,[],loss_D)
+#training_updates = tf.train.AdamOptimizer(learning_rate=lrD, beta1=0.5).get_updates(netD.trainable_weights,[],loss_D)
+netD_train = K.function([real_A, real_B],[loss_D/2], training_updates)
+loss_G = loss_G_fake   + 100 * loss_L1
+training_updates = Adam(lr=lrG, beta_1=0.5).get_updates(netG.trainable_weights,[], loss_G)
+#training_updates = tf.train.AdamOptimizer(learning_rate=lrG, beta1=0.5).get_updates(netG.trainable_weights,[], loss_G)
+netG_train = K.function([real_A, real_B], [loss_G_fake, loss_L1], training_updates)
+try:
+    netG.load_weights('netG.hdf5')
+    netD.load_weights('netD.hdf5')
+    print('重みデータの読み込み完了')
+except:
+    print('重みデータを読み込めませんでした')
+    pass
+tpu_grpc_url = "grpc://"+os.environ["COLAB_TPU_ADDR"]
+tpu_cluster_resolver = tf.contrib.cluster_resolver.TPUClusterResolver(tpu_grpc_url)
+strategy = keras_support.TPUDistributionStrategy(tpu_cluster_resolver)
+netD = tf.contrib.tpu.keras_to_tpu_model(netD, strategy=strategy)
+netG = tf.contrib.tpu.keras_to_tpu_model(netG, strategy=strategy)
+```

追記

2019/02/18 01:18

投稿

Reach

スコア733

test CHANGED Viewed

File without changes

test CHANGED Viewed

@@ -80,4 +80,6 @@
-ここにより詳細な情報を記載してください。
+改変コードが 字数制限で掲載できません
+どうすれば よいのでしょうか？