{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# High-level Keras (Theano) Example"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"Using Theano backend.\n",
"Using cuDNN version 6021 on context None\n",
"Mapped name None to device cuda: Tesla K80 (BF4E:00:00.0)\n"
]
}
],
"source": [
"import os\n",
"import sys\n",
"import numpy as np\n",
"os.environ['KERAS_BACKEND'] = \"theano\"\n",
"import keras as K\n",
"import theano\n",
"from keras.models import Sequential\n",
"from keras.layers import Dense, Dropout, Flatten\n",
"from keras.layers import Conv2D, MaxPooling2D\n",
"from common.params import *\n",
"from common.utils import *"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {},
"outputs": [],
"source": [
"# channels_first is faster\n",
"K.backend.set_image_data_format('channels_first')"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"OS: linux\n",
"Python: 3.5.2 |Anaconda custom (64-bit)| (default, Jul 2 2016, 17:53:06) \n",
"[GCC 4.4.7 20120313 (Red Hat 4.4.7-1)]\n",
"Keras: 2.0.8\n",
"Numpy: 1.13.1\n",
"Theano: 0.10.0beta1.dev-RELEASE\n",
"theano\n",
"channels_first\n"
]
}
],
"source": [
"print(\"OS: \", sys.platform)\n",
"print(\"Python: \", sys.version)\n",
"print(\"Keras: \", K.__version__)\n",
"print(\"Numpy: \", np.__version__)\n",
"print(\"Theano: \", theano.__version__)\n",
"print(K.backend.backend())\n",
"# Should be channels-first, otherwise slow\n",
"print(K.backend.image_data_format())"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": [
"#CuDNN auto-tune\n",
"theano.config.dnn.conv.algo_fwd = \"time_once\"\n",
"theano.config.dnn.conv.algo_bwd_filter = \"time_once\"\n",
"theano.config.dnn.conv.algo_bwd_data = \"time_once\""
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": [
"def create_symbol():\n",
" model = Sequential()\n",
" \n",
" model.add(Conv2D(50, kernel_size=(3, 3), padding='same', activation='relu', input_shape=(3, 32, 32)))\n",
" model.add(Conv2D(50, kernel_size=(3, 3), padding='same', activation='relu')) \n",
" model.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2)))\n",
" model.add(Dropout(0.25))\n",
" \n",
" model.add(Conv2D(100, kernel_size=(3, 3), padding='same', activation='relu'))\n",
" model.add(Conv2D(100, kernel_size=(3, 3), padding='same', activation='relu')) \n",
" model.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2)))\n",
" model.add(Dropout(0.25))\n",
" \n",
" model.add(Flatten())\n",
" model.add(Dense(512, activation='relu'))\n",
" model.add(Dropout(0.5))\n",
" model.add(Dense(N_CLASSES, activation='softmax'))\n",
" return model"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": [
"def init_model(m):\n",
" m.compile(\n",
" loss = \"categorical_crossentropy\",\n",
" optimizer = K.optimizers.SGD(LR, MOMENTUM),\n",
" metrics = ['accuracy'])\n",
" return m"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Downloading http://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz\n",
"Done.\n",
"Extracting files...\n",
"Done.\n",
"Preparing train set...\n",
"Preparing test set...\n",
"Done.\n",
"(50000, 3, 32, 32) (10000, 3, 32, 32) (50000, 10) (10000, 10)\n",
"float32 float32 int32 int32\n",
"CPU times: user 2.92 s, sys: 1.43 s, total: 4.34 s\n",
"Wall time: 27.7 s\n"
]
}
],
"source": [
"%%time\n",
"# Data into format for library\n",
"x_train, x_test, y_train, y_test = cifar_for_library(channel_first=True, one_hot=True)\n",
"print(x_train.shape, x_test.shape, y_train.shape, y_test.shape)\n",
"print(x_train.dtype, x_test.dtype, y_train.dtype, y_test.dtype)"
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"CPU times: user 964 ms, sys: 116 ms, total: 1.08 s\n",
"Wall time: 3.16 s\n"
]
}
],
"source": [
"%%time\n",
"# Load symbol\n",
"sym = create_symbol()"
]
},
{
"cell_type": "code",
"execution_count": 8,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"CPU times: user 12 ms, sys: 0 ns, total: 12 ms\n",
"Wall time: 11.3 ms\n"
]
}
],
"source": [
"%%time\n",
"# Initialise model\n",
"model = init_model(sym)"
]
},
{
"cell_type": "code",
"execution_count": 9,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"_________________________________________________________________\n",
"Layer (type) Output Shape Param # \n",
"=================================================================\n",
"conv2d_1 (Conv2D) (None, 50, 32, 32) 1400 \n",
"_________________________________________________________________\n",
"conv2d_2 (Conv2D) (None, 50, 32, 32) 22550 \n",
"_________________________________________________________________\n",
"max_pooling2d_1 (MaxPooling2 (None, 50, 16, 16) 0 \n",
"_________________________________________________________________\n",
"dropout_1 (Dropout) (None, 50, 16, 16) 0 \n",
"_________________________________________________________________\n",
"conv2d_3 (Conv2D) (None, 100, 16, 16) 45100 \n",
"_________________________________________________________________\n",
"conv2d_4 (Conv2D) (None, 100, 16, 16) 90100 \n",
"_________________________________________________________________\n",
"max_pooling2d_2 (MaxPooling2 (None, 100, 8, 8) 0 \n",
"_________________________________________________________________\n",
"dropout_2 (Dropout) (None, 100, 8, 8) 0 \n",
"_________________________________________________________________\n",
"flatten_1 (Flatten) (None, 6400) 0 \n",
"_________________________________________________________________\n",
"dense_1 (Dense) (None, 512) 3277312 \n",
"_________________________________________________________________\n",
"dropout_3 (Dropout) (None, 512) 0 \n",
"_________________________________________________________________\n",
"dense_2 (Dense) (None, 10) 5130 \n",
"=================================================================\n",
"Total params: 3,441,592\n",
"Trainable params: 3,441,592\n",
"Non-trainable params: 0\n",
"_________________________________________________________________\n"
]
}
],
"source": [
"model.summary()"
]
},
{
"cell_type": "code",
"execution_count": 10,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Epoch 1/10\n",
"50000/50000 [==============================] - 22s - loss: 1.8528 - acc: 0.3227 \n",
"Epoch 2/10\n",
"50000/50000 [==============================] - 25s - loss: 1.4089 - acc: 0.4873 \n",
"Epoch 3/10\n",
"50000/50000 [==============================] - 25s - loss: 1.1726 - acc: 0.5821 \n",
"Epoch 4/10\n",
"50000/50000 [==============================] - 25s - loss: 1.0034 - acc: 0.6460 \n",
"Epoch 5/10\n",
"50000/50000 [==============================] - 26s - loss: 0.8888 - acc: 0.6873 \n",
"Epoch 6/10\n",
"50000/50000 [==============================] - 26s - loss: 0.7867 - acc: 0.7238 \n",
"Epoch 7/10\n",
"50000/50000 [==============================] - 26s - loss: 0.7200 - acc: 0.7473 \n",
"Epoch 8/10\n",
"50000/50000 [==============================] - 26s - loss: 0.6515 - acc: 0.7722 \n",
"Epoch 9/10\n",
"50000/50000 [==============================] - 27s - loss: 0.5975 - acc: 0.7897 - ETA: 1s - \n",
"Epoch 10/10\n",
"50000/50000 [==============================] - 27s - loss: 0.5483 - acc: 0.8055 \n",
"CPU times: user 2min 25s, sys: 51.5 s, total: 3min 17s\n",
"Wall time: 4min 29s\n"
]
},
{
"data": {
"text/plain": [
"<keras.callbacks.History at 0x7fb1a4244908>"
]
},
"execution_count": 10,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"%%time\n",
"# Train model\n",
"model.fit(x_train,\n",
" y_train,\n",
" batch_size=BATCHSIZE,\n",
" epochs=EPOCHS,\n",
" verbose=1)"
]
},
{
"cell_type": "code",
"execution_count": 11,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"CPU times: user 6.33 s, sys: 408 ms, total: 6.74 s\n",
"Wall time: 8.27 s\n"
]
}
],
"source": [
"%%time\n",
"y_guess = model.predict(x_test, batch_size=BATCHSIZE)\n",
"y_guess = np.argmax(y_guess, axis=-1)\n",
"y_truth = np.argmax(y_test, axis=-1)"
]
},
{
"cell_type": "code",
"execution_count": 12,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Accuracy: 0.7783\n"
]
}
],
"source": [
"print(\"Accuracy: \", sum(y_guess == y_truth)/len(y_guess))"
]
}
],
"metadata": {
"anaconda-cloud": {},
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.5.2"
}
},
"nbformat": 4,
"nbformat_minor": 2
}