{ "nbformat": 4, "nbformat_minor": 0, "metadata": { "colab": { "provenance": [], "gpuType": "T4" }, "kernelspec": { "name": "python3", "display_name": "Python 3" }, "language_info": { "name": "python" }, "accelerator": "GPU" }, "cells": [ { "cell_type": "code", "execution_count": 1, "metadata": { "id": "gPLoY9QJRLN1" }, "outputs": [], "source": [ "import tensorflow as tf\n", "from tensorflow.keras.models import Model\n", "from tensorflow.keras.optimizers import Adam\n", "from tensorflow.keras.datasets import cifar10\n", "from keras.preprocessing import image\n", "import keras.backend as K\n", "import matplotlib.pyplot as plt\n", "import numpy as np\n", "import time\n", "from tqdm import tqdm" ] }, { "cell_type": "code", "source": [ "batch_size = 16\n", "epoch_count = 50\n", "noise_dim = 100\n", "n_class = 10\n", "tags = ['Airplane', 'Automobile', 'Bird', 'Cat', 'Deer', 'Dog', 'Frog', 'Horse', 'Ship', 'Truck']\n", "img_size = 32\n", "\n", "(X_train, y_train), (_, _) = cifar10.load_data()\n", "\n", "X_train = (X_train - 127.5) / 127.5\n", "\n", "dataset = tf.data.Dataset.from_tensor_slices((X_train, y_train))\n", "dataset = dataset.shuffle(buffer_size=1000).batch(batch_size)" ], "metadata": { "colab": { "base_uri": "https://localhost:8080/" }, "id": "AZCSgPTMRNEs", "outputId": "f6da0ff8-0631-42f4-af9f-bc2f80ac95f8" }, "execution_count": 2, "outputs": [ { "output_type": "stream", "name": "stdout", "text": [ "Downloading data from https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz\n", "\u001b[1m170498071/170498071\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 0us/step\n" ] } ] }, { "cell_type": "code", "source": [ "plt.figure(figsize=(2,2))\n", "idx = np.random.randint(0,len(X_train))\n", "img = image.array_to_img(X_train[idx], scale=True)\n", "plt.imshow(img)\n", "plt.axis('off')\n", "plt.title(tags[y_train[idx][0]])\n", "plt.show()" ], "metadata": { "colab": { "base_uri": "https://localhost:8080/", "height": 213 }, "id": "yXqQWY8mRNG_", "outputId": "a13ca992-fd8a-4b1e-a420-75cc4dd44263" }, "execution_count": 3, "outputs": [ { "output_type": "display_data", "data": { "text/plain": [ "
" ], "image/png": "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\n" }, "metadata": {} } ] }, { "cell_type": "code", "source": [ "bce_loss = tf.keras.losses.BinaryCrossentropy()\n", "\n", "\n", "def discriminator_loss(real, fake):\n", " real_loss = bce_loss(tf.ones_like(real), real)\n", " fake_loss = bce_loss(tf.zeros_like(fake), fake)\n", " total_loss = real_loss + fake_loss\n", " return total_loss\n", "\n", "def generator_loss(preds):\n", " return bce_loss(tf.ones_like(preds), preds)\n", "\n", "d_optimizer=Adam(learning_rate=0.0002, beta_1 = 0.5)\n", "g_optimizer=Adam(learning_rate=0.0002, beta_1 = 0.5)" ], "metadata": { "id": "N2W4b2NsRNJg" }, "execution_count": 4, "outputs": [] }, { "cell_type": "code", "source": [ "def build_generator():\n", "\n", "\n", " in_label = tf.keras.layers.Input(shape=(1,))\n", " li = tf.keras.layers.Embedding(n_class, 50)(in_label)\n", "\n", " n_nodes = 8 * 8\n", " li = tf.keras.layers.Dense(n_nodes)(li)\n", " li = tf.keras.layers.Reshape((8, 8, 1))(li)\n", " in_lat = tf.keras.layers.Input(shape=(noise_dim,))\n", "\n", " n_nodes = 128 * 8 * 8\n", " gen = tf.keras.layers.Dense(n_nodes)(in_lat)\n", " gen = tf.keras.layers.LeakyReLU(alpha=0.2)(gen)\n", " gen = tf.keras.layers.Reshape((8, 8, 128))(gen)\n", " merge = tf.keras.layers.Concatenate()([gen, li])\n", "\n", " gen = tf.keras.layers.Conv2DTranspose(\n", " 128, (4, 4), strides=(2, 2), padding='same')(merge)\n", " gen = tf.keras.layers.LeakyReLU(alpha=0.2)(gen)\n", "\n", " gen = tf.keras.layers.Conv2DTranspose(\n", " 128, (4, 4), strides=(2, 2), padding='same')(gen)\n", " gen = tf.keras.layers.LeakyReLU(alpha=0.2)(gen)\n", "\n", " out_layer = tf.keras.layers.Conv2D(\n", " 3, (8, 8), activation='tanh', padding='same')(gen)\n", "\n", " model = Model([in_lat, in_label], out_layer)\n", " return model\n", "\n", "\n", "g_model = build_generator()\n", "g_model.summary()" ], "metadata": { "colab": { "base_uri": "https://localhost:8080/", "height": 761 }, "id": "vdGMT2bjRNL7", "outputId": "14cf7bf4-9ba2-4f8d-80f3-2ab621cc956a" }, "execution_count": 5, "outputs": [ { "output_type": "stream", "name": "stderr", "text": [ "/usr/local/lib/python3.12/dist-packages/keras/src/layers/activations/leaky_relu.py:41: UserWarning: Argument `alpha` is deprecated. Use `negative_slope` instead.\n", " warnings.warn(\n" ] }, { "output_type": "display_data", "data": { "text/plain": [ "\u001b[1mModel: \"functional\"\u001b[0m\n" ], "text/html": [ "
Model: \"functional\"\n",
              "
\n" ] }, "metadata": {} }, { "output_type": "display_data", "data": { "text/plain": [ "┏━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┓\n", "┃\u001b[1m \u001b[0m\u001b[1mLayer (type) \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1mOutput Shape \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1m Param #\u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1mConnected to \u001b[0m\u001b[1m \u001b[0m┃\n", "┡━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━┩\n", "│ input_layer_1 │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m100\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │ - │\n", "│ (\u001b[38;5;33mInputLayer\u001b[0m) │ │ │ │\n", "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n", "│ input_layer │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m1\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │ - │\n", "│ (\u001b[38;5;33mInputLayer\u001b[0m) │ │ │ │\n", "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n", "│ dense_1 (\u001b[38;5;33mDense\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m8192\u001b[0m) │ \u001b[38;5;34m827,392\u001b[0m │ input_layer_1[\u001b[38;5;34m0\u001b[0m]… │\n", "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n", "│ embedding │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m1\u001b[0m, \u001b[38;5;34m50\u001b[0m) │ \u001b[38;5;34m500\u001b[0m │ input_layer[\u001b[38;5;34m0\u001b[0m][\u001b[38;5;34m0\u001b[0m] │\n", "│ (\u001b[38;5;33mEmbedding\u001b[0m) │ │ │ │\n", "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n", "│ leaky_re_lu │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m8192\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │ dense_1[\u001b[38;5;34m0\u001b[0m][\u001b[38;5;34m0\u001b[0m] │\n", "│ (\u001b[38;5;33mLeakyReLU\u001b[0m) │ │ │ │\n", "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n", "│ dense (\u001b[38;5;33mDense\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m1\u001b[0m, \u001b[38;5;34m64\u001b[0m) │ \u001b[38;5;34m3,264\u001b[0m │ embedding[\u001b[38;5;34m0\u001b[0m][\u001b[38;5;34m0\u001b[0m] │\n", "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n", "│ reshape_1 (\u001b[38;5;33mReshape\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m8\u001b[0m, \u001b[38;5;34m8\u001b[0m, \u001b[38;5;34m128\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │ leaky_re_lu[\u001b[38;5;34m0\u001b[0m][\u001b[38;5;34m0\u001b[0m] │\n", "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n", "│ reshape (\u001b[38;5;33mReshape\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m8\u001b[0m, \u001b[38;5;34m8\u001b[0m, \u001b[38;5;34m1\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │ dense[\u001b[38;5;34m0\u001b[0m][\u001b[38;5;34m0\u001b[0m] │\n", "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n", "│ concatenate │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m8\u001b[0m, \u001b[38;5;34m8\u001b[0m, \u001b[38;5;34m129\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │ reshape_1[\u001b[38;5;34m0\u001b[0m][\u001b[38;5;34m0\u001b[0m], │\n", "│ (\u001b[38;5;33mConcatenate\u001b[0m) │ │ │ reshape[\u001b[38;5;34m0\u001b[0m][\u001b[38;5;34m0\u001b[0m] │\n", "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n", "│ conv2d_transpose │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m16\u001b[0m, \u001b[38;5;34m16\u001b[0m, │ \u001b[38;5;34m264,320\u001b[0m │ concatenate[\u001b[38;5;34m0\u001b[0m][\u001b[38;5;34m0\u001b[0m] │\n", "│ (\u001b[38;5;33mConv2DTranspose\u001b[0m) │ \u001b[38;5;34m128\u001b[0m) │ │ │\n", "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n", "│ leaky_re_lu_1 │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m16\u001b[0m, \u001b[38;5;34m16\u001b[0m, │ \u001b[38;5;34m0\u001b[0m │ conv2d_transpose… │\n", "│ (\u001b[38;5;33mLeakyReLU\u001b[0m) │ \u001b[38;5;34m128\u001b[0m) │ │ │\n", "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n", "│ conv2d_transpose_1 │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m32\u001b[0m, \u001b[38;5;34m32\u001b[0m, │ \u001b[38;5;34m262,272\u001b[0m │ leaky_re_lu_1[\u001b[38;5;34m0\u001b[0m]… │\n", "│ (\u001b[38;5;33mConv2DTranspose\u001b[0m) │ \u001b[38;5;34m128\u001b[0m) │ │ │\n", "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n", "│ leaky_re_lu_2 │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m32\u001b[0m, \u001b[38;5;34m32\u001b[0m, │ \u001b[38;5;34m0\u001b[0m │ conv2d_transpose… │\n", "│ (\u001b[38;5;33mLeakyReLU\u001b[0m) │ \u001b[38;5;34m128\u001b[0m) │ │ │\n", "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n", "│ conv2d (\u001b[38;5;33mConv2D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m32\u001b[0m, \u001b[38;5;34m32\u001b[0m, \u001b[38;5;34m3\u001b[0m) │ \u001b[38;5;34m24,579\u001b[0m │ leaky_re_lu_2[\u001b[38;5;34m0\u001b[0m]… │\n", "└─────────────────────┴───────────────────┴────────────┴───────────────────┘\n" ], "text/html": [ "
┏━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┓\n",
              "┃ Layer (type)        ┃ Output Shape      ┃    Param # ┃ Connected to      ┃\n",
              "┡━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━┩\n",
              "│ input_layer_1       │ (None, 100)       │          0 │ -                 │\n",
              "│ (InputLayer)        │                   │            │                   │\n",
              "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n",
              "│ input_layer         │ (None, 1)         │          0 │ -                 │\n",
              "│ (InputLayer)        │                   │            │                   │\n",
              "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n",
              "│ dense_1 (Dense)     │ (None, 8192)      │    827,392 │ input_layer_1[0]… │\n",
              "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n",
              "│ embedding           │ (None, 1, 50)     │        500 │ input_layer[0][0] │\n",
              "│ (Embedding)         │                   │            │                   │\n",
              "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n",
              "│ leaky_re_lu         │ (None, 8192)      │          0 │ dense_1[0][0]     │\n",
              "│ (LeakyReLU)         │                   │            │                   │\n",
              "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n",
              "│ dense (Dense)       │ (None, 1, 64)     │      3,264 │ embedding[0][0]   │\n",
              "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n",
              "│ reshape_1 (Reshape) │ (None, 8, 8, 128) │          0 │ leaky_re_lu[0][0] │\n",
              "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n",
              "│ reshape (Reshape)   │ (None, 8, 8, 1)   │          0 │ dense[0][0]       │\n",
              "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n",
              "│ concatenate         │ (None, 8, 8, 129) │          0 │ reshape_1[0][0],  │\n",
              "│ (Concatenate)       │                   │            │ reshape[0][0]     │\n",
              "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n",
              "│ conv2d_transpose    │ (None, 16, 16,    │    264,320 │ concatenate[0][0] │\n",
              "│ (Conv2DTranspose)   │ 128)              │            │                   │\n",
              "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n",
              "│ leaky_re_lu_1       │ (None, 16, 16,    │          0 │ conv2d_transpose… │\n",
              "│ (LeakyReLU)         │ 128)              │            │                   │\n",
              "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n",
              "│ conv2d_transpose_1  │ (None, 32, 32,    │    262,272 │ leaky_re_lu_1[0]… │\n",
              "│ (Conv2DTranspose)   │ 128)              │            │                   │\n",
              "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n",
              "│ leaky_re_lu_2       │ (None, 32, 32,    │          0 │ conv2d_transpose… │\n",
              "│ (LeakyReLU)         │ 128)              │            │                   │\n",
              "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n",
              "│ conv2d (Conv2D)     │ (None, 32, 32, 3) │     24,579 │ leaky_re_lu_2[0]… │\n",
              "└─────────────────────┴───────────────────┴────────────┴───────────────────┘\n",
              "
\n" ] }, "metadata": {} }, { "output_type": "display_data", "data": { "text/plain": [ "\u001b[1m Total params: \u001b[0m\u001b[38;5;34m1,382,327\u001b[0m (5.27 MB)\n" ], "text/html": [ "
 Total params: 1,382,327 (5.27 MB)\n",
              "
\n" ] }, "metadata": {} }, { "output_type": "display_data", "data": { "text/plain": [ "\u001b[1m Trainable params: \u001b[0m\u001b[38;5;34m1,382,327\u001b[0m (5.27 MB)\n" ], "text/html": [ "
 Trainable params: 1,382,327 (5.27 MB)\n",
              "
\n" ] }, "metadata": {} }, { "output_type": "display_data", "data": { "text/plain": [ "\u001b[1m Non-trainable params: \u001b[0m\u001b[38;5;34m0\u001b[0m (0.00 B)\n" ], "text/html": [ "
 Non-trainable params: 0 (0.00 B)\n",
              "
\n" ] }, "metadata": {} } ] }, { "cell_type": "code", "source": [ "def build_discriminator():\n", "\n", "\n", " in_label = tf.keras.layers.Input(shape=(1,))\n", "\n", " li = tf.keras.layers.Embedding(n_class, 50)(in_label)\n", "\n", " n_nodes = img_size * img_size\n", " li = tf.keras.layers.Dense(n_nodes)(li)\n", "\n", " li = tf.keras.layers.Reshape((img_size, img_size, 1))(li)\n", "\n", "\n", "\n", " in_image = tf.keras.layers.Input(shape=(img_size, img_size, 3))\n", "\n", " merge = tf.keras.layers.Concatenate()([in_image, li])\n", "\n", " fe = tf.keras.layers.Conv2D(128, (3,3), strides=(2,2), padding='same')(merge)\n", " fe = tf.keras.layers.LeakyReLU(alpha=0.2)(fe)\n", "\n", " fe = tf.keras.layers.Conv2D(128, (3,3), strides=(2,2), padding='same')(fe)\n", " fe = tf.keras.layers.LeakyReLU(alpha=0.2)(fe)\n", "\n", " fe = tf.keras.layers.Flatten()(fe)\n", "\n", " fe = tf.keras.layers.Dropout(0.4)(fe)\n", "\n", " out_layer = tf.keras.layers.Dense(1, activation='sigmoid')(fe)\n", "\n", " model = Model([in_image, in_label], out_layer)\n", "\n", " return model\n", "\n", "\n", "d_model = build_discriminator()\n", "d_model.summary()" ], "metadata": { "colab": { "base_uri": "https://localhost:8080/", "height": 661 }, "id": "tm6W8S5uRNOo", "outputId": "c1aa29eb-02f6-4bc9-8186-271116bd6ac2" }, "execution_count": 6, "outputs": [ { "output_type": "display_data", "data": { "text/plain": [ "\u001b[1mModel: \"functional_1\"\u001b[0m\n" ], "text/html": [ "
Model: \"functional_1\"\n",
              "
\n" ] }, "metadata": {} }, { "output_type": "display_data", "data": { "text/plain": [ "┏━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┓\n", "┃\u001b[1m \u001b[0m\u001b[1mLayer (type) \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1mOutput Shape \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1m Param #\u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1mConnected to \u001b[0m\u001b[1m \u001b[0m┃\n", "┡━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━┩\n", "│ input_layer_2 │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m1\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │ - │\n", "│ (\u001b[38;5;33mInputLayer\u001b[0m) │ │ │ │\n", "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n", "│ embedding_1 │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m1\u001b[0m, \u001b[38;5;34m50\u001b[0m) │ \u001b[38;5;34m500\u001b[0m │ input_layer_2[\u001b[38;5;34m0\u001b[0m]… │\n", "│ (\u001b[38;5;33mEmbedding\u001b[0m) │ │ │ │\n", "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n", "│ dense_2 (\u001b[38;5;33mDense\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m1\u001b[0m, \u001b[38;5;34m1024\u001b[0m) │ \u001b[38;5;34m52,224\u001b[0m │ embedding_1[\u001b[38;5;34m0\u001b[0m][\u001b[38;5;34m0\u001b[0m] │\n", "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n", "│ input_layer_3 │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m32\u001b[0m, \u001b[38;5;34m32\u001b[0m, \u001b[38;5;34m3\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │ - │\n", "│ (\u001b[38;5;33mInputLayer\u001b[0m) │ │ │ │\n", "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n", "│ reshape_2 (\u001b[38;5;33mReshape\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m32\u001b[0m, \u001b[38;5;34m32\u001b[0m, \u001b[38;5;34m1\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │ dense_2[\u001b[38;5;34m0\u001b[0m][\u001b[38;5;34m0\u001b[0m] │\n", "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n", "│ concatenate_1 │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m32\u001b[0m, \u001b[38;5;34m32\u001b[0m, \u001b[38;5;34m4\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │ input_layer_3[\u001b[38;5;34m0\u001b[0m]… │\n", "│ (\u001b[38;5;33mConcatenate\u001b[0m) │ │ │ reshape_2[\u001b[38;5;34m0\u001b[0m][\u001b[38;5;34m0\u001b[0m] │\n", "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n", "│ conv2d_1 (\u001b[38;5;33mConv2D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m16\u001b[0m, \u001b[38;5;34m16\u001b[0m, │ \u001b[38;5;34m4,736\u001b[0m │ concatenate_1[\u001b[38;5;34m0\u001b[0m]… │\n", "│ │ \u001b[38;5;34m128\u001b[0m) │ │ │\n", "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n", "│ leaky_re_lu_3 │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m16\u001b[0m, \u001b[38;5;34m16\u001b[0m, │ \u001b[38;5;34m0\u001b[0m │ conv2d_1[\u001b[38;5;34m0\u001b[0m][\u001b[38;5;34m0\u001b[0m] │\n", "│ (\u001b[38;5;33mLeakyReLU\u001b[0m) │ \u001b[38;5;34m128\u001b[0m) │ │ │\n", "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n", "│ conv2d_2 (\u001b[38;5;33mConv2D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m8\u001b[0m, \u001b[38;5;34m8\u001b[0m, \u001b[38;5;34m128\u001b[0m) │ \u001b[38;5;34m147,584\u001b[0m │ leaky_re_lu_3[\u001b[38;5;34m0\u001b[0m]… │\n", "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n", "│ leaky_re_lu_4 │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m8\u001b[0m, \u001b[38;5;34m8\u001b[0m, \u001b[38;5;34m128\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │ conv2d_2[\u001b[38;5;34m0\u001b[0m][\u001b[38;5;34m0\u001b[0m] │\n", "│ (\u001b[38;5;33mLeakyReLU\u001b[0m) │ │ │ │\n", "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n", "│ flatten (\u001b[38;5;33mFlatten\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m8192\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │ leaky_re_lu_4[\u001b[38;5;34m0\u001b[0m]… │\n", "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n", "│ dropout (\u001b[38;5;33mDropout\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m8192\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │ flatten[\u001b[38;5;34m0\u001b[0m][\u001b[38;5;34m0\u001b[0m] │\n", "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n", "│ dense_3 (\u001b[38;5;33mDense\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m1\u001b[0m) │ \u001b[38;5;34m8,193\u001b[0m │ dropout[\u001b[38;5;34m0\u001b[0m][\u001b[38;5;34m0\u001b[0m] │\n", "└─────────────────────┴───────────────────┴────────────┴───────────────────┘\n" ], "text/html": [ "
┏━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━┓\n",
              "┃ Layer (type)        ┃ Output Shape      ┃    Param # ┃ Connected to      ┃\n",
              "┡━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━┩\n",
              "│ input_layer_2       │ (None, 1)         │          0 │ -                 │\n",
              "│ (InputLayer)        │                   │            │                   │\n",
              "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n",
              "│ embedding_1         │ (None, 1, 50)     │        500 │ input_layer_2[0]… │\n",
              "│ (Embedding)         │                   │            │                   │\n",
              "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n",
              "│ dense_2 (Dense)     │ (None, 1, 1024)   │     52,224 │ embedding_1[0][0] │\n",
              "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n",
              "│ input_layer_3       │ (None, 32, 32, 3) │          0 │ -                 │\n",
              "│ (InputLayer)        │                   │            │                   │\n",
              "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n",
              "│ reshape_2 (Reshape) │ (None, 32, 32, 1) │          0 │ dense_2[0][0]     │\n",
              "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n",
              "│ concatenate_1       │ (None, 32, 32, 4) │          0 │ input_layer_3[0]… │\n",
              "│ (Concatenate)       │                   │            │ reshape_2[0][0]   │\n",
              "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n",
              "│ conv2d_1 (Conv2D)   │ (None, 16, 16,    │      4,736 │ concatenate_1[0]… │\n",
              "│                     │ 128)              │            │                   │\n",
              "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n",
              "│ leaky_re_lu_3       │ (None, 16, 16,    │          0 │ conv2d_1[0][0]    │\n",
              "│ (LeakyReLU)         │ 128)              │            │                   │\n",
              "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n",
              "│ conv2d_2 (Conv2D)   │ (None, 8, 8, 128) │    147,584 │ leaky_re_lu_3[0]… │\n",
              "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n",
              "│ leaky_re_lu_4       │ (None, 8, 8, 128) │          0 │ conv2d_2[0][0]    │\n",
              "│ (LeakyReLU)         │                   │            │                   │\n",
              "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n",
              "│ flatten (Flatten)   │ (None, 8192)      │          0 │ leaky_re_lu_4[0]… │\n",
              "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n",
              "│ dropout (Dropout)   │ (None, 8192)      │          0 │ flatten[0][0]     │\n",
              "├─────────────────────┼───────────────────┼────────────┼───────────────────┤\n",
              "│ dense_3 (Dense)     │ (None, 1)         │      8,193 │ dropout[0][0]     │\n",
              "└─────────────────────┴───────────────────┴────────────┴───────────────────┘\n",
              "
\n" ] }, "metadata": {} }, { "output_type": "display_data", "data": { "text/plain": [ "\u001b[1m Total params: \u001b[0m\u001b[38;5;34m213,237\u001b[0m (832.96 KB)\n" ], "text/html": [ "
 Total params: 213,237 (832.96 KB)\n",
              "
\n" ] }, "metadata": {} }, { "output_type": "display_data", "data": { "text/plain": [ "\u001b[1m Trainable params: \u001b[0m\u001b[38;5;34m213,237\u001b[0m (832.96 KB)\n" ], "text/html": [ "
 Trainable params: 213,237 (832.96 KB)\n",
              "
\n" ] }, "metadata": {} }, { "output_type": "display_data", "data": { "text/plain": [ "\u001b[1m Non-trainable params: \u001b[0m\u001b[38;5;34m0\u001b[0m (0.00 B)\n" ], "text/html": [ "
 Non-trainable params: 0 (0.00 B)\n",
              "
\n" ] }, "metadata": {} } ] }, { "cell_type": "code", "source": [ "@tf.function\n", "def train_step(dataset):\n", "\n", " real_images, real_labels = dataset\n", "\n", " random_latent_vectors = tf.random.normal(shape=(batch_size, noise_dim))\n", " generated_images = g_model([random_latent_vectors, real_labels])\n", "\n", " with tf.GradientTape() as tape:\n", " pred_fake = d_model([generated_images, real_labels])\n", " pred_real = d_model([real_images, real_labels])\n", "\n", " d_loss = discriminator_loss(pred_real, pred_fake)\n", "\n", " grads = tape.gradient(d_loss, d_model.trainable_variables)\n", "\n", " d_optimizer.apply_gradients(zip(grads, d_model.trainable_variables))\n", "\n", "\n", " random_latent_vectors = tf.random.normal(shape=(batch_size, noise_dim))\n", "\n", "\n", " with tf.GradientTape() as tape:\n", " fake_images = g_model([random_latent_vectors, real_labels])\n", " predictions = d_model([fake_images, real_labels])\n", " g_loss = generator_loss(predictions)\n", "\n", " grads = tape.gradient(g_loss, g_model.trainable_variables)\n", " g_optimizer.apply_gradients(zip(grads, g_model.trainable_variables))\n", "\n", " return d_loss, g_loss" ], "metadata": { "id": "jWkU7w7lRNRZ" }, "execution_count": 7, "outputs": [] }, { "cell_type": "code", "source": [ "def show_samples(num_samples, n_class, g_model):\n", " fig, axes = plt.subplots(10,num_samples, figsize=(10,20))\n", " fig.tight_layout()\n", " fig.subplots_adjust(wspace=None, hspace=0.2)\n", "\n", " for l in np.arange(10):\n", " random_noise = tf.random.normal(shape=(num_samples, noise_dim))\n", " label = tf.ones(num_samples)*l\n", " gen_imgs = g_model.predict([random_noise, label])\n", " for j in range(gen_imgs.shape[0]):\n", " img = image.array_to_img(gen_imgs[j], scale=True)\n", " axes[l,j].imshow(img)\n", " axes[l,j].yaxis.set_ticks([])\n", " axes[l,j].xaxis.set_ticks([])\n", "\n", " if j ==0:\n", " axes[l,j].set_ylabel(tags[l])\n", " plt.show()" ], "metadata": { "id": "UoTG4caeRNT4" }, "execution_count": 8, "outputs": [] }, { "cell_type": "code", "source": [ "\n", "def train(dataset, epochs=epoch_count):\n", "\n", " for epoch in range(epochs):\n", " print('Epoch: ', epochs)\n", " d_loss_list = []\n", " g_loss_list = []\n", " q_loss_list = []\n", " start = time.time()\n", "\n", " itern = 0\n", " for image_batch in tqdm(dataset):\n", " d_loss, g_loss = train_step(image_batch)\n", " d_loss_list.append(d_loss)\n", " g_loss_list.append(g_loss)\n", " itern=itern+1\n", "\n", " show_samples(3, n_class, g_model)\n", "\n", " print (f'Epoch: {epoch} -- Generator Loss: {np.mean(g_loss_list)}, Discriminator Loss: {np.mean(d_loss_list)}\\n')\n", " print (f'Took {time.time()-start} seconds. \\n\\n')\n", "\n", "\n", "train(dataset, epochs=epoch_count)" ], "metadata": { "id": "OyzEZUE3RNWA" }, "execution_count": 9, "outputs": [] } ] }