Skip to content

Commit

Permalink
Tf2.x_Collab_Tutorial_on_GAN
Browse files Browse the repository at this point in the history 

In response to Issue "Colab Tutorial Not TF 2.x Compatible tensorflow#18" on tensorflow/gan tensorflow#18

Collab Tutorial using tf 2.x
  • Loading branch information
@Ankuraxz
Ankuraxz authored Mar 20, 2020
1 parent 5f38a6f commit 6df5f19
Showing 1 changed file with 334 additions and 0 deletions.
334 changes: 334 additions & 0 deletions tensorflow_gan/Collab_tutorial_on_GAN_using_tf2_x.ipynb
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,334 @@
{
"nbformat": 4,
"nbformat_minor": 0,
"metadata": {
"colab": {
"name": "Collab_tutorial_on_GAN_using_tf2.x.ipynb",
"provenance": []
},
"kernelspec": {
"name": "python3",
"display_name": "Python 3"
}
},
"cells": [
{
"cell_type": "code",
"metadata": {
"id": "h8-7MxIETkG4",
"colab_type": "code",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 36
},
"outputId": "c5b2bb65-bf61-43ec-adbe-908c16741db9"
},
"source": [
"# Setting up\n",
"from __future__ import absolute_import, division, print_function, unicode_literals\n",
"try:\n",
" # tf_version only exist in colab\n",
" %tensorflow_version 2.x\n",
"\n",
"except Exception:\n",
" pass\n"
],
"execution_count": 1,
"outputs": [
{
"output_type": "stream",
"text": [
"TensorFlow 2.x selected.\n"
],
"name": "stdout"
}
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "6MD16UIYVt7n",
"colab_type": "text"
},
"source": [
"# SETUP"
]
},
{
"cell_type": "code",
"metadata": {
"id": "A5gKsgrGUZ39",
"colab_type": "code",
"colab": {}
},
"source": [
"import tensorflow as tf\n",
"tf.random.set_seed(7)\n",
"import numpy as np\n",
"np.random.seed(7)\n",
"import matplotlib.pyplot as plt\n",
"import os\n",
"import time"
],
"execution_count": 0,
"outputs": []
},
{
"cell_type": "markdown",
"metadata": {
"id": "J0iuziuiVwkA",
"colab_type": "text"
},
"source": [
"# HYPERPARAMETERS\n",
"\n",
"X_Train is the array of images, the user wants to use as training data. Labels and test data aren't needed"
]
},
{
"cell_type": "code",
"metadata": {
"id": "ZlWkvpA0UcK5",
"colab_type": "code",
"colab": {}
},
"source": [
"EPOCHS = 60\n",
"BATCH_SIZE = 256\n",
"NO_OF_BATCHES = int(X_Train.shape[0]/BATCH_SIZE)\n",
"HALF_BATCH_SIZE = 128\n",
"NOISE_DIM = 100\n",
"# Customised ADAM optimizer\n",
"adam = tf.keras.optimizers.Adam(lr=2e-4,beta_1=0.5) # lr: Learning Rate"
],
"execution_count": 0,
"outputs": []
},
{
"cell_type": "markdown",
"metadata": {
"id": "YEnk012DV_Et",
"colab_type": "text"
},
"source": [
"# Making up models for GAN"
]
},
{
"cell_type": "code",
"metadata": {
"id": "ewWCDClQUhGo",
"colab_type": "code",
"colab": {}
},
"source": [
"from tensorflow.keras.layers import *\n",
"from tensorflow.keras.layers import LeakyReLU\n",
"from tensorflow.keras.models import Sequential, Model"
],
"execution_count": 0,
"outputs": []
},
{
"cell_type": "markdown",
"metadata": {
"id": "gMbuIfzhWC6K",
"colab_type": "text"
},
"source": [
"# GENERATOR "
]
},
{
"cell_type": "code",
"metadata": {
"id": "oXt9-uPMUjUo",
"colab_type": "code",
"colab": {}
},
"source": [
"# Generator\n",
"generator = Sequential()\n",
"generator.add(Dense(256,input_shape=(NOISE_DIM,)))\n",
"generator.add(LeakyReLU(0.2))\n",
"generator.add(Dense(512))\n",
"generator.add(LeakyReLU(0.2))\n",
"generator.add(Dense(1024))\n",
"generator.add(LeakyReLU(0.2))\n",
"generator.add(Dense(784,activation='tanh'))\n",
"\n",
"# Compile\n",
"generator.compile(loss='binary_crossentropy',optimizer=adam)\n",
"\n",
"generator.summary()"
],
"execution_count": 0,
"outputs": []
},
{
"cell_type": "markdown",
"metadata": {
"id": "RZownvPRWGT-",
"colab_type": "text"
},
"source": [
"# DISCRIMINATOR"
]
},
{
"cell_type": "code",
"metadata": {
"id": "KAhiXuXWUnyT",
"colab_type": "code",
"colab": {}
},
"source": [
"#Discriminator\n",
"discriminator = Sequential()\n",
"discriminator.add(Dense(512,input_shape=(784,)))\n",
"discriminator.add(LeakyReLU(0.2))\n",
"discriminator.add(Dense(256))\n",
"discriminator.add(LeakyReLU(0.2))\n",
"discriminator.add(Dense(1,activation='sigmoid'))\n",
"\n",
"discriminator.compile(loss='binary_crossentropy',optimizer=adam)\n",
"discriminator.summary()"
],
"execution_count": 0,
"outputs": []
},
{
"cell_type": "markdown",
"metadata": {
"id": "JfnJ1IVIWJt2",
"colab_type": "text"
},
"source": [
"# Combined Model"
]
},
{
"cell_type": "code",
"metadata": {
"id": "VKzNOfNIUpxa",
"colab_type": "code",
"colab": {}
},
"source": [
"# Combined Model (Geneerator + Discriminator) -> Functional API\n",
"discriminator.trainable = False\n",
"gan_input = Input(shape=(NOISE_DIM,))\n",
"generator_output = generator(gan_input)\n",
"gan_output = discriminator(generator_output)\n",
"\n",
"model = Model(gan_input,gan_output)\n",
"model.compile(loss='binary_crossentropy',optimizer=adam)\n",
"\n",
"model.summary()"
],
"execution_count": 0,
"outputs": []
},
{
"cell_type": "code",
"metadata": {
"id": "tWMncFAzUs_D",
"colab_type": "code",
"colab": {}
},
"source": [
"def showImgs(epoch):\n",
" noise = np.random.normal(0,1,size=(100,NOISE_DIM)) # Random Noise\n",
" generated_imgs = generator.predict(noise)\n",
" generated_imgs = generated_imgs.reshape(-1,28,28)\n",
" \n",
" #Display the Images\n",
" plt.figure(figsize=(10,10))\n",
" for i in range(100):\n",
" plt.subplot(10,10,i+1)\n",
" plt.imshow(generated_imgs[i],cmap='gray',interpolation='nearest')\n",
" plt.axis(\"off\")\n",
" plt.tight_layout()\n",
" plt.show()"
],
"execution_count": 0,
"outputs": []
},
{
"cell_type": "markdown",
"metadata": {
"id": "21WSSthiWPuo",
"colab_type": "text"
},
"source": [
"# Training"
]
},
{
"cell_type": "code",
"metadata": {
"id": "TTUSWup6U59U",
"colab_type": "code",
"colab": {}
},
"source": [
"d_losses = []\n",
"g_losses = []\n",
"\n",
"# Training Loop\n",
"for epoch in range(EPOCHS+1): \n",
" epoch_d_loss = 0.0\n",
" epoch_g_loss = 0.0\n",
" \n",
" # Mini Batch\n",
" for step in range(NO_OF_BATCHES):\n",
" idx = np.random.randint(0,X_Train.shape[0],HALF_BATCH_SIZE)\n",
" real_imgs = X_Train[idx]\n",
" \n",
" # generate fake images assuming generator is frozen\n",
" noise = np.random.normal(0,1,size=(HALF_BATCH_SIZE,NOISE_DIM))\n",
" fake_imgs = generator.predict(noise)\n",
" \n",
" # Labels\n",
" real_y = np.ones((HALF_BATCH_SIZE,1))*0.9\n",
" fake_y = np.zeros((HALF_BATCH_SIZE,1))\n",
" \n",
" #Train on Real and Fake Images\n",
" d_real_loss = discriminator.train_on_batch(real_imgs,real_y)\n",
" d_fake_loss = discriminator.train_on_batch(fake_imgs,fake_y)\n",
" \n",
" d_loss = 0.5*d_real_loss + 0.5*d_fake_loss\n",
" epoch_d_loss += d_loss\n",
" \n",
" # Train Generator\n",
" noise = np.random.normal(0,1,size=(BATCH_SIZE,NOISE_DIM))\n",
" real_y = np.ones((BATCH_SIZE,1))\n",
" g_loss = model.train_on_batch(noise,real_y)\n",
" epoch_g_loss += g_loss\n",
" \n",
" d_losses.append(epoch_d_loss)\n",
" g_losses.append(epoch_g_loss)\n",
" \n",
" print(\"Epoch %d D Loss %.4f G loss %0.4f \"%((epoch+1),epoch_d_loss,epoch_g_loss))\n",
" if (epoch%5)==0: #printing images after every 5 epocs\n",
" #generator.save(\"model/gen_{0}.h5\".format(epoch)) # Saving the model\n",
" showImgs(epoch) # Show the image transformation progress at every step"
],
"execution_count": 0,
"outputs": []
},
{
"cell_type": "code",
"metadata": {
"id": "tP85ZHAvVksH",
"colab_type": "code",
"colab": {}
},
"source": [
""
],
"execution_count": 0,
"outputs": []
}
]
}

0 comments on commit 6df5f19

Please sign in to comment.