Skip to content

Instantly share code, notes, and snippets.

@artificialsoph

artificialsoph/infogan-annotated.ipynb

Last active June 20, 2018 17:34
Show Gist options
  • Save artificialsoph/fbcd6f519e78d0d760efc6dc85c6910e to your computer and use it in GitHub Desktop.
Save artificialsoph/fbcd6f519e78d0d760efc6dc85c6910e to your computer and use it in GitHub Desktop.
Download ZIP
Display the source blob
Display the rendered blob
Raw
infogan-annotated.ipynb
{
"cells": [
{
"cell_type": "code",
"execution_count": 5,
"metadata": {
"ExecuteTime": {
"end_time": "2018-06-20T17:10:56.184264Z",
"start_time": "2018-06-20T17:10:56.180289Z"
}
},
"outputs": [],
"source": [
"from keras.datasets import mnist\n",
"from keras.layers import Input, Dense, Reshape, Flatten, Dropout, multiply, concatenate\n",
"from keras.layers import BatchNormalization, Activation, Embedding, ZeroPadding2D, Lambda\n",
"from keras.layers.advanced_activations import LeakyReLU\n",
"from keras.layers.convolutional import UpSampling2D, Conv2D\n",
"from keras.models import Sequential, Model\n",
"from keras.optimizers import Adam\n",
"from keras.utils import to_categorical\n",
"import keras.backend as K\n",
"\n",
"import matplotlib.pyplot as plt\n",
"\n",
"import numpy as np"
]
},
{
"cell_type": "code",
"execution_count": 8,
"metadata": {
"ExecuteTime": {
"end_time": "2018-06-20T17:21:39.596837Z",
"start_time": "2018-06-20T17:21:39.572870Z"
}
},
"outputs": [],
"source": [
"class INFOGAN():\n",
" def __init__(self):\n",
" \n",
" # SOPH: review these params and update accordingly\n",
" self.img_rows = 28\n",
" self.img_cols = 28\n",
" self.channels = 1\n",
" self.num_classes = 10\n",
" self.img_shape = (self.img_rows, self.img_cols, self.channels)\n",
" self.latent_dim = 72\n",
"\n",
"\n",
" optimizer = Adam(0.0002, 0.5)\n",
" losses = ['binary_crossentropy', self.mutual_info_loss]\n",
"\n",
" # Build and the discriminator and recognition network\n",
" self.discriminator, self.auxilliary = self.build_disk_and_q_net()\n",
"\n",
" self.discriminator.compile(loss=['binary_crossentropy'],\n",
" optimizer=optimizer,\n",
" metrics=['accuracy'])\n",
"\n",
" # Build and compile the recognition network Q\n",
" self.auxilliary.compile(loss=[self.mutual_info_loss],\n",
" optimizer=optimizer,\n",
" metrics=['accuracy'])\n",
"\n",
" # Build the generator\n",
" self.generator = self.build_generator()\n",
"\n",
" # The generator takes noise and the target label as input\n",
" # and generates the corresponding digit of that label\n",
" gen_input = Input(shape=(self.latent_dim,))\n",
" img = self.generator(gen_input)\n",
"\n",
" # For the combined model we will only train the generator\n",
" self.discriminator.trainable = False\n",
"\n",
" # The discriminator takes generated image as input and determines validity\n",
" valid = self.discriminator(img)\n",
" # The recognition network produces the label\n",
" target_label = self.auxilliary(img)\n",
"\n",
" # The combined model (stacked generator and discriminator)\n",
" self.combined = Model(gen_input, [valid, target_label])\n",
" self.combined.compile(loss=losses,\n",
" optimizer=optimizer)\n",
"\n",
"\n",
" def build_generator(self):\n",
"\n",
" \n",
" #SOPH: there's a chance that the generator doesn't generate images with dimensionality exactly like your \n",
" # training images. If that's the case, some annoying debugging might be in order. I'd recommend\n",
" # tweaking the layers in the generator seperately until you get the dimensionality right and then moving that code\n",
" # here after. Otherwise, consider scaling all of your training images to be square\n",
" model = Sequential()\n",
"\n",
" model.add(Dense(128 * 7 * 7, activation=\"relu\", input_dim=self.latent_dim))\n",
" model.add(Reshape((7, 7, 128)))\n",
" model.add(BatchNormalization(momentum=0.8))\n",
" model.add(UpSampling2D())\n",
" model.add(Conv2D(128, kernel_size=3, padding=\"same\"))\n",
" model.add(Activation(\"relu\"))\n",
" model.add(BatchNormalization(momentum=0.8))\n",
" model.add(UpSampling2D())\n",
" model.add(Conv2D(64, kernel_size=3, padding=\"same\"))\n",
" model.add(Activation(\"relu\"))\n",
" model.add(BatchNormalization(momentum=0.8))\n",
" model.add(Conv2D(self.channels, kernel_size=3, padding='same'))\n",
" model.add(Activation(\"tanh\"))\n",
"\n",
" gen_input = Input(shape=(self.latent_dim,))\n",
" img = model(gen_input)\n",
"\n",
" print(\"gen summary:\")\n",
" model.summary()\n",
"\n",
" return Model(gen_input, img)\n",
"\n",
"\n",
" def build_disk_and_q_net(self):\n",
"\n",
" img = Input(shape=self.img_shape)\n",
"\n",
" # Shared layers between discriminator and recognition network\n",
" model = Sequential()\n",
" model.add(Conv2D(64, kernel_size=3, strides=2, input_shape=self.img_shape, padding=\"same\"))\n",
" model.add(LeakyReLU(alpha=0.2))\n",
" model.add(Dropout(0.25))\n",
" model.add(Conv2D(128, kernel_size=3, strides=2, padding=\"same\"))\n",
" model.add(ZeroPadding2D(padding=((0,1),(0,1))))\n",
" model.add(LeakyReLU(alpha=0.2))\n",
" model.add(Dropout(0.25))\n",
" model.add(BatchNormalization(momentum=0.8))\n",
" model.add(Conv2D(256, kernel_size=3, strides=2, padding=\"same\"))\n",
" model.add(LeakyReLU(alpha=0.2))\n",
" model.add(Dropout(0.25))\n",
" model.add(BatchNormalization(momentum=0.8))\n",
" model.add(Conv2D(512, kernel_size=3, strides=2, padding=\"same\"))\n",
" model.add(LeakyReLU(alpha=0.2))\n",
" model.add(Dropout(0.25))\n",
" model.add(BatchNormalization(momentum=0.8))\n",
" model.add(Flatten())\n",
"\n",
" img_embedding = model(img)\n",
"\n",
" # Discriminator\n",
" validity = Dense(1, activation='sigmoid')(img_embedding)\n",
"\n",
" # Recognition\n",
" q_net = Dense(128, activation='relu')(img_embedding)\n",
" # SOPH: this activation should reflect that you're doing multi-label with binary labels. \n",
" # So, rather than softmax, what should you have?\n",
" label = Dense(self.num_classes, activation='softmax')(q_net)\n",
" \n",
" disk = Model(img, validity)\n",
" \n",
" print(\"disk summary:\")\n",
" disk.summary()\n",
" \n",
" q = Model(img, label)\n",
" \n",
" print(\"q summary\")\n",
" q.summary()\n",
"\n",
" # Return discriminator and recognition network\n",
" return disk, q\n",
"\n",
"\n",
" def mutual_info_loss(self, c, c_given_x):\n",
" \"\"\"The mutual information metric we aim to minimize\"\"\"\n",
" eps = 1e-8\n",
" conditional_entropy = K.mean(- K.sum(K.log(c_given_x + eps) * c, axis=1))\n",
" entropy = K.mean(- K.sum(K.log(c + eps) * c, axis=1))\n",
"\n",
" return conditional_entropy + entropy\n",
"\n",
" def sample_generator_input(self, batch_size):\n",
" # Generator inputs\n",
" sampled_noise = np.random.normal(0, 1, (batch_size, 62))\n",
" \n",
" \n",
" # SOPH: this code generates random combinations of your labels so that it can generate randomized images\n",
" # the following code should be removed and replaced with something that generates random values for your 40\n",
" # labels\n",
" sampled_labels = np.random.randint(0, self.num_classes, batch_size).reshape(-1, 1)\n",
" sampled_labels = to_categorical(sampled_labels, num_classes=self.num_classes)\n",
"\n",
" return sampled_noise, sampled_labels\n",
"\n",
" def train(self, epochs, batch_size=128, sample_interval=50):\n",
"\n",
" # SOPH: replace the following code block with code that loads celebA. \n",
" # x_train should be [n_examples x pixels x pixels]\n",
" # y_train should be [n_examples x n_labels]\n",
" # Load the dataset\n",
" (X_train, y_train), (_, _) = mnist.load_data()\n",
" \n",
" # Rescale -1 to 1\n",
" X_train = (X_train.astype(np.float32) - 127.5) / 127.5\n",
" X_train = np.expand_dims(X_train, axis=3)\n",
" y_train = y_train.reshape(-1, 1)\n",
" # SOPH: this is the end of the loading code. You may have to replace/remove any of these lines\n",
"\n",
" # Adversarial ground truths\n",
" valid = np.ones((batch_size, 1))\n",
" fake = np.zeros((batch_size, 1))\n",
"\n",
" for epoch in range(epochs):\n",
"\n",
" # ---------------------\n",
" # Train Discriminator\n",
" # ---------------------\n",
"\n",
" # Select a random half batch of images\n",
" idx = np.random.randint(0, X_train.shape[0], batch_size)\n",
" imgs = X_train[idx]\n",
"\n",
" # Sample noise and categorical labels\n",
" sampled_noise, sampled_labels = self.sample_generator_input(batch_size)\n",
" gen_input = np.concatenate((sampled_noise, sampled_labels), axis=1)\n",
"\n",
" # Generate a half batch of new images\n",
" gen_imgs = self.generator.predict(gen_input)\n",
"\n",
" # Train on real and generated data\n",
" d_loss_real = self.discriminator.train_on_batch(imgs, valid)\n",
" d_loss_fake = self.discriminator.train_on_batch(gen_imgs, fake)\n",
"\n",
" # Avg. loss\n",
" d_loss = 0.5 * np.add(d_loss_real, d_loss_fake)\n",
"\n",
" # ---------------------\n",
" # Train Generator and Q-network\n",
" # ---------------------\n",
"\n",
" g_loss = self.combined.train_on_batch(gen_input, [valid, sampled_labels])\n",
"\n",
" # Plot the progress\n",
"\n",
" # If at save interval => save generated image samples\n",
" if epoch % sample_interval == 0:\n",
" print (\"%d [D loss: %.2f, acc.: %.2f%%] [Q loss: %.2f] [G loss: %.2f]\" % (epoch, d_loss[0], 100*d_loss[1], g_loss[1], g_loss[2]))\n",
" self.sample_images(epoch)\n",
"\n",
" def sample_images(self, epoch):\n",
" # SOPH: this'll be hard to adjust so consider just removing it or greatly simplifying it to start out.\n",
" # this generates those image matrices.\n",
" r, c = 10, 10\n",
"\n",
" fig, axs = plt.subplots(r, c)\n",
" for i in range(c):\n",
" sampled_noise, _ = self.sample_generator_input(c)\n",
" label = to_categorical(np.full(fill_value=i, shape=(r,1)), num_classes=self.num_classes)\n",
" gen_input = np.concatenate((sampled_noise, label), axis=1)\n",
" gen_imgs = self.generator.predict(gen_input)\n",
" gen_imgs = 0.5 * gen_imgs + 0.5\n",
" for j in range(r):\n",
" axs[j,i].imshow(gen_imgs[j,:,:,0], cmap='gray')\n",
" axs[j,i].axis('off')\n",
" fig.savefig(\"images/%d.png\" % epoch)\n",
" plt.close()\n",
"\n",
" def save_model(self):\n",
"\n",
" def save(model, model_name):\n",
" model_path = \"saved_model/%s.json\" % model_name\n",
" weights_path = \"saved_model/%s_weights.hdf5\" % model_name\n",
" options = {\"file_arch\": model_path,\n",
" \"file_weight\": weights_path}\n",
" json_string = model.to_json()\n",
" open(options['file_arch'], 'w').write(json_string)\n",
" model.save_weights(options['file_weight'])\n",
"\n",
" save(self.generator, \"generator\")\n",
" save(self.discriminator, \"discriminator\")"
]
},
{
"cell_type": "code",
"execution_count": 22,
"metadata": {
"ExecuteTime": {
"end_time": "2018-06-20T17:32:02.063801Z",
"start_time": "2018-06-20T17:32:02.060492Z"
}
},
"outputs": [],
"source": [
"import keras"
]
},
{
"cell_type": "code",
"execution_count": 23,
"metadata": {
"ExecuteTime": {
"end_time": "2018-06-20T17:32:53.104698Z",
"start_time": "2018-06-20T17:32:51.882258Z"
}
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"gen summary:\n",
"_________________________________________________________________\n",
"Layer (type) Output Shape Param # \n",
"=================================================================\n",
"dense_21 (Dense) (None, 18304) 750464 \n",
"_________________________________________________________________\n",
"reshape_9 (Reshape) (None, 13, 11, 128) 0 \n",
"_________________________________________________________________\n",
"batch_normalization_39 (Batc (None, 13, 11, 128) 512 \n",
"_________________________________________________________________\n",
"up_sampling2d_19 (UpSampling (None, 26, 22, 128) 0 \n",
"_________________________________________________________________\n",
"conv2d_43 (Conv2D) (None, 26, 22, 128) 147584 \n",
"_________________________________________________________________\n",
"activation_27 (Activation) (None, 26, 22, 128) 0 \n",
"_________________________________________________________________\n",
"batch_normalization_40 (Batc (None, 26, 22, 128) 512 \n",
"_________________________________________________________________\n",
"up_sampling2d_20 (UpSampling (None, 52, 44, 128) 0 \n",
"_________________________________________________________________\n",
"conv2d_44 (Conv2D) (None, 52, 44, 128) 147584 \n",
"_________________________________________________________________\n",
"activation_28 (Activation) (None, 52, 44, 128) 0 \n",
"_________________________________________________________________\n",
"batch_normalization_41 (Batc (None, 52, 44, 128) 512 \n",
"_________________________________________________________________\n",
"up_sampling2d_21 (UpSampling (None, 104, 88, 128) 0 \n",
"_________________________________________________________________\n",
"zero_padding2d_5 (ZeroPaddin (None, 109, 89, 128) 0 \n",
"_________________________________________________________________\n",
"conv2d_45 (Conv2D) (None, 109, 89, 128) 147584 \n",
"_________________________________________________________________\n",
"activation_29 (Activation) (None, 109, 89, 128) 0 \n",
"_________________________________________________________________\n",
"batch_normalization_42 (Batc (None, 109, 89, 128) 512 \n",
"_________________________________________________________________\n",
"up_sampling2d_22 (UpSampling (None, 218, 178, 128) 0 \n",
"_________________________________________________________________\n",
"conv2d_46 (Conv2D) (None, 218, 178, 64) 73792 \n",
"_________________________________________________________________\n",
"activation_30 (Activation) (None, 218, 178, 64) 0 \n",
"_________________________________________________________________\n",
"batch_normalization_43 (Batc (None, 218, 178, 64) 256 \n",
"_________________________________________________________________\n",
"conv2d_47 (Conv2D) (None, 218, 178, 3) 1731 \n",
"_________________________________________________________________\n",
"activation_31 (Activation) (None, 218, 178, 3) 0 \n",
"=================================================================\n",
"Total params: 1,271,043\n",
"Trainable params: 1,269,891\n",
"Non-trainable params: 1,152\n",
"_________________________________________________________________\n"
]
}
],
"source": [
"# input_dims = 54, 44\n",
"\n",
"model = Sequential()\n",
"\n",
"model.add(Dense(128 * 13 * 11, activation=\"relu\", input_dim=40))\n",
"model.add(Reshape((13, 11, 128)))\n",
"model.add(BatchNormalization(momentum=0.8))\n",
"\n",
"model.add(UpSampling2D())\n",
"model.add(Conv2D(128, kernel_size=3, padding=\"same\"))\n",
"model.add(Activation(\"relu\"))\n",
"model.add(BatchNormalization(momentum=0.8))\n",
"\n",
"model.add(UpSampling2D())\n",
"model.add(Conv2D(128, kernel_size=3, padding=\"same\"))\n",
"model.add(Activation(\"relu\"))\n",
"model.add(BatchNormalization(momentum=0.8))\n",
"\n",
"model.add(UpSampling2D())\n",
"model.add(keras.layers.ZeroPadding2D(((3,2), (0,1))))\n",
"\n",
"model.add(Conv2D(128, kernel_size=3, padding=\"same\"))\n",
"model.add(Activation(\"relu\"))\n",
"model.add(BatchNormalization(momentum=0.8))\n",
"\n",
"model.add(UpSampling2D())\n",
"model.add(Conv2D(64, kernel_size=3, padding=\"same\"))\n",
"model.add(Activation(\"relu\"))\n",
"model.add(BatchNormalization(momentum=0.8))\n",
"model.add(Conv2D(3, kernel_size=3, padding='same'))\n",
"model.add(Activation(\"tanh\"))\n",
"\n",
"gen_input = Input(shape=(40,))\n",
"img = model(gen_input)\n",
"\n",
"print(\"gen summary:\")\n",
"model.summary()"
]
},
{
"cell_type": "code",
"execution_count": 21,
"metadata": {
"ExecuteTime": {
"end_time": "2018-06-20T17:31:27.572321Z",
"start_time": "2018-06-20T17:31:27.568083Z"
}
},
"outputs": [
{
"data": {
"text/plain": [
"(109.0, 89.0)"
]
},
"execution_count": 21,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"218/2, 178/2\n"
]
},
{
"cell_type": "code",
"execution_count": 9,
"metadata": {
"ExecuteTime": {
"end_time": "2018-06-20T17:21:56.125215Z",
"start_time": "2018-06-20T17:21:46.546713Z"
},
"scrolled": false
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"disk summary:\n",
"_________________________________________________________________\n",
"Layer (type) Output Shape Param # \n",
"=================================================================\n",
"input_10 (InputLayer) (None, 28, 28, 1) 0 \n",
"_________________________________________________________________\n",
"sequential_7 (Sequential) (None, 2048) 1553408 \n",
"_________________________________________________________________\n",
"dense_13 (Dense) (None, 1) 2049 \n",
"=================================================================\n",
"Total params: 1,555,457\n",
"Trainable params: 1,553,665\n",
"Non-trainable params: 1,792\n",
"_________________________________________________________________\n",
"q summary\n",
"_________________________________________________________________\n",
"Layer (type) Output Shape Param # \n",
"=================================================================\n",
"input_10 (InputLayer) (None, 28, 28, 1) 0 \n",
"_________________________________________________________________\n",
"sequential_7 (Sequential) (None, 2048) 1553408 \n",
"_________________________________________________________________\n",
"dense_14 (Dense) (None, 128) 262272 \n",
"_________________________________________________________________\n",
"dense_15 (Dense) (None, 10) 1290 \n",
"=================================================================\n",
"Total params: 1,816,970\n",
"Trainable params: 1,815,178\n",
"Non-trainable params: 1,792\n",
"_________________________________________________________________\n",
"gen summary:\n",
"_________________________________________________________________\n",
"Layer (type) Output Shape Param # \n",
"=================================================================\n",
"dense_16 (Dense) (None, 6272) 457856 \n",
"_________________________________________________________________\n",
"reshape_4 (Reshape) (None, 7, 7, 128) 0 \n",
"_________________________________________________________________\n",
"batch_normalization_22 (Batc (None, 7, 7, 128) 512 \n",
"_________________________________________________________________\n",
"up_sampling2d_7 (UpSampling2 (None, 14, 14, 128) 0 \n",
"_________________________________________________________________\n",
"conv2d_26 (Conv2D) (None, 14, 14, 128) 147584 \n",
"_________________________________________________________________\n",
"activation_10 (Activation) (None, 14, 14, 128) 0 \n",
"_________________________________________________________________\n",
"batch_normalization_23 (Batc (None, 14, 14, 128) 512 \n",
"_________________________________________________________________\n",
"up_sampling2d_8 (UpSampling2 (None, 28, 28, 128) 0 \n",
"_________________________________________________________________\n",
"conv2d_27 (Conv2D) (None, 28, 28, 64) 73792 \n",
"_________________________________________________________________\n",
"activation_11 (Activation) (None, 28, 28, 64) 0 \n",
"_________________________________________________________________\n",
"batch_normalization_24 (Batc (None, 28, 28, 64) 256 \n",
"_________________________________________________________________\n",
"conv2d_28 (Conv2D) (None, 28, 28, 1) 577 \n",
"_________________________________________________________________\n",
"activation_12 (Activation) (None, 28, 28, 1) 0 \n",
"=================================================================\n",
"Total params: 681,089\n",
"Trainable params: 680,449\n",
"Non-trainable params: 640\n",
"_________________________________________________________________\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"/home/ubuntu/miniconda3/lib/python3.6/site-packages/keras/engine/training.py:975: UserWarning: Discrepancy between trainable weights and collected trainable weights, did you set `model.trainable` without calling `model.compile` after ?\n",
" 'Discrepancy between trainable weights and collected trainable'\n"
]
},
{
"ename": "KeyboardInterrupt",
"evalue": "",
"output_type": "error",
"traceback": [
"\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
"\u001b[0;31mKeyboardInterrupt\u001b[0m Traceback (most recent call last)",
"\u001b[0;32m<ipython-input-9-489e5ad1b538>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m 1\u001b[0m \u001b[0minfogan\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mINFOGAN\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 2\u001b[0;31m \u001b[0minfogan\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtrain\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mepochs\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m50000\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mbatch_size\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m128\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msample_interval\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m50\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
"\u001b[0;32m<ipython-input-8-36aa60c01d3d>\u001b[0m in \u001b[0;36mtrain\u001b[0;34m(self, epochs, batch_size, sample_interval)\u001b[0m\n\u001b[1;32m 196\u001b[0m \u001b[0;31m# ---------------------\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 197\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 198\u001b[0;31m \u001b[0mg_loss\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcombined\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtrain_on_batch\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mgen_input\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0mvalid\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msampled_labels\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 199\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 200\u001b[0m \u001b[0;31m# Plot the progress\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
"\u001b[0;32m~/miniconda3/lib/python3.6/site-packages/keras/engine/training.py\u001b[0m in \u001b[0;36mtrain_on_batch\u001b[0;34m(self, x, y, sample_weight, class_weight)\u001b[0m\n\u001b[1;32m 1881\u001b[0m \u001b[0mins\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mx\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0my\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0msample_weights\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 1882\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_make_train_function\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1883\u001b[0;31m \u001b[0moutputs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtrain_function\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mins\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 1884\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0moutputs\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 1885\u001b[0m \u001b[0;32mreturn\u001b[0m \u001b[0moutputs\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
"\u001b[0;32m~/miniconda3/lib/python3.6/site-packages/keras/backend/tensorflow_backend.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, inputs)\u001b[0m\n\u001b[1;32m 2478\u001b[0m \u001b[0mfeed_dict\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mtensor\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mvalue\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 2479\u001b[0m \u001b[0mfetches\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0moutputs\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mupdates_op\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfetches\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 2480\u001b[0;31m \u001b[0msession\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mget_session\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 2481\u001b[0m updated = session.run(fetches=fetches, feed_dict=feed_dict,\n\u001b[1;32m 2482\u001b[0m **self.session_kwargs)\n",
"\u001b[0;32m~/miniconda3/lib/python3.6/site-packages/keras/backend/tensorflow_backend.py\u001b[0m in \u001b[0;36mget_session\u001b[0;34m()\u001b[0m\n\u001b[1;32m 191\u001b[0m \u001b[0;31m# not already marked as initialized.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 192\u001b[0m is_initialized = session.run(\n\u001b[0;32m--> 193\u001b[0;31m [tf.is_variable_initialized(v) for v in candidate_vars])\n\u001b[0m\u001b[1;32m 194\u001b[0m \u001b[0muninitialized_vars\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 195\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mflag\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mv\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mzip\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mis_initialized\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcandidate_vars\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
"\u001b[0;32m~/miniconda3/lib/python3.6/site-packages/tensorflow/python/client/session.py\u001b[0m in \u001b[0;36mrun\u001b[0;34m(self, fetches, feed_dict, options, run_metadata)\u001b[0m\n\u001b[1;32m 898\u001b[0m \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 899\u001b[0m result = self._run(None, fetches, feed_dict, options_ptr,\n\u001b[0;32m--> 900\u001b[0;31m run_metadata_ptr)\n\u001b[0m\u001b[1;32m 901\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mrun_metadata\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 902\u001b[0m \u001b[0mproto_data\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtf_session\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mTF_GetBuffer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mrun_metadata_ptr\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
"\u001b[0;32m~/miniconda3/lib/python3.6/site-packages/tensorflow/python/client/session.py\u001b[0m in \u001b[0;36m_run\u001b[0;34m(self, handle, fetches, feed_dict, options, run_metadata)\u001b[0m\n\u001b[1;32m 1133\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mfinal_fetches\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0mfinal_targets\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mhandle\u001b[0m \u001b[0;32mand\u001b[0m \u001b[0mfeed_dict_tensor\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 1134\u001b[0m results = self._do_run(handle, final_targets, final_fetches,\n\u001b[0;32m-> 1135\u001b[0;31m feed_dict_tensor, options, run_metadata)\n\u001b[0m\u001b[1;32m 1136\u001b[0m \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 1137\u001b[0m \u001b[0mresults\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
"\u001b[0;32m~/miniconda3/lib/python3.6/site-packages/tensorflow/python/client/session.py\u001b[0m in \u001b[0;36m_do_run\u001b[0;34m(self, handle, target_list, fetch_list, feed_dict, options, run_metadata)\u001b[0m\n\u001b[1;32m 1314\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mhandle\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 1315\u001b[0m return self._do_call(_run_fn, feeds, fetches, targets, options,\n\u001b[0;32m-> 1316\u001b[0;31m run_metadata)\n\u001b[0m\u001b[1;32m 1317\u001b[0m \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 1318\u001b[0m \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_do_call\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0m_prun_fn\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mhandle\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfeeds\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfetches\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
"\u001b[0;32m~/miniconda3/lib/python3.6/site-packages/tensorflow/python/client/session.py\u001b[0m in \u001b[0;36m_do_call\u001b[0;34m(self, fn, *args)\u001b[0m\n\u001b[1;32m 1320\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0m_do_call\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfn\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 1321\u001b[0m \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1322\u001b[0;31m \u001b[0;32mreturn\u001b[0m \u001b[0mfn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 1323\u001b[0m \u001b[0;32mexcept\u001b[0m \u001b[0merrors\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mOpError\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 1324\u001b[0m \u001b[0mmessage\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcompat\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mas_text\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0me\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmessage\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
"\u001b[0;32m~/miniconda3/lib/python3.6/site-packages/tensorflow/python/client/session.py\u001b[0m in \u001b[0;36m_run_fn\u001b[0;34m(feed_dict, fetch_list, target_list, options, run_metadata)\u001b[0m\n\u001b[1;32m 1303\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0m_run_fn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfeed_dict\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfetch_list\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtarget_list\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0moptions\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mrun_metadata\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 1304\u001b[0m \u001b[0;31m# Ensure any changes to the graph are reflected in the runtime.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1305\u001b[0;31m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_extend_graph\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 1306\u001b[0m return self._call_tf_sessionrun(\n\u001b[1;32m 1307\u001b[0m options, feed_dict, fetch_list, target_list, run_metadata)\n",
"\u001b[0;32m~/miniconda3/lib/python3.6/site-packages/tensorflow/python/client/session.py\u001b[0m in \u001b[0;36m_extend_graph\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m 1338\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_created_with_new_api\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 1339\u001b[0m \u001b[0;32mwith\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_graph\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_lock\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;31m# pylint: disable=protected-access\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1340\u001b[0;31m \u001b[0mtf_session\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mExtendSession\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_session\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 1341\u001b[0m \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 1342\u001b[0m \u001b[0;31m# Ensure any changes to the graph are reflected in the runtime.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
"\u001b[0;31mKeyboardInterrupt\u001b[0m: "
]
}
],
"source": [
"infogan = INFOGAN()\n",
"infogan.train(epochs=50000, batch_size=128, sample_interval=50)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"toc": {
"nav_menu": {},
"number_sections": true,
"sideBar": true,
"skip_h1_title": false,
"toc_cell": false,
"toc_position": {},
"toc_section_display": "block",
"toc_window_display": false
},
"varInspector": {
"cols": {
"lenName": 16,
"lenType": 16,
"lenVar": 40
},
"kernels_config": {
"python": {
"delete_cmd_postfix": "",
"delete_cmd_prefix": "del ",
"library": "var_list.py",
"varRefreshCmd": "print(var_dic_list())"
},
"r": {
"delete_cmd_postfix": ") ",
"delete_cmd_prefix": "rm(",
"library": "var_list.r",
"varRefreshCmd": "cat(var_dic_list()) "
}
},
"types_to_exclude": [
"module",
"function",
"builtin_function_or_method",
"instance",
"_Feature"
],
"window_display": false
}
},
"nbformat": 4,
"nbformat_minor": 2
}
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment