vgg in gluon

mxnet/vggnet.ipynb

@@ -0,0 +1,137 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# VGGNet\n",
+    "\n",
+    "重复网络结构的构建\n",
+    "\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {
+    "collapsed": true
+   },
+   "outputs": [],
+   "source": [
+    "from mxnet.gluon import nn\n",
+    "def vgg_block3(num_convs,channels):\n",
+    "    out = nn.Sequential()\n",
+    "    for _ in range(num_convs):\n",
+    "        out.add(nn.Conv2D(channels,kernel_size=3,padding=1,activation='relu'))\n",
+    "    out.add(nn.MaxPool2D(pool_size=2,strides=2))\n",
+    "    return out\n",
+    "def vgg_block3_1(num_convs,channels):\n",
+    "    out = nn.Sequential()\n",
+    "    for _ in range(num_convs):\n",
+    "        out.add(nn.Conv2D(channels,kernel_size=3,padding=1,activation='relu'))\n",
+    "    out.add(nn.Conv2D(channels,kernel_size=1,padding=1,activation='relu'),\n",
+    "           nn.MaxPool2D(pool_size=2,strides=2))\n",
+    "    return out\n",
+    "def vgg_fc(num_outputs):\n",
+    "    out = nn.Sequential()\n",
+    "    out.add(nn.Flatten(),\n",
+    "            nn.Dense(4096,activation='relu'),\n",
+    "            nn.Dropout(.5),\n",
+    "            nn.Dense(4096,activation='relu'),\n",
+    "            nn.Dropout(.5),\n",
+    "            nn.Dense(num_outputs)\n",
+    "           )\n",
+    "    return out"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def get_vgg(architecture,num_outputs):\n",
+    "    net = nn.Sequential()\n",
+    "    with net.name_scope():\n",
+    "        for num_convs,channels in architecture:\n",
+    "            net.add(vgg_block3(num_convs,channels))\n",
+    "        net.add(vgg_fc(num_outputs))\n",
+    "    return net\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "获取数据，初始化网络，设置loss和solver"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "('Start training on ', gpu(0))\n",
+      "Epoch 0. Loss: 1.011, Train acc 0.62, Test acc 0.81, Time 93.4 sec\n"
+     ]
+    }
+   ],
+   "source": [
+    "import sys\n",
+    "sys.path.append('..')\n",
+    "import utils\n",
+    "from mxnet import init\n",
+    "from mxnet import gluon\n",
+    "\n",
+    "batch_size = 128\n",
+    "resize=128\n",
+    "train_data, test_data = utils.load_data_fashion_mnist(batch_size,resize=resize)\n",
+    "ctx = utils.try_gpu()\n",
+    "\n",
+    "num_outputs = 10\n",
+    "architecture = ((1,64),(1,128),(2,256),(2,512),(2,512))\n",
+    "net=get_vgg(architecture,num_outputs)\n",
+    "net.initialize(ctx=ctx,init=init.Xavier())\n",
+    "\n",
+    "loss = gluon.loss.SoftmaxCrossEntropyLoss()\n",
+    "#train(train_data, test_data, net, loss, trainer, ctx, num_epochs, print_batches=None):\n",
+    "trainer = gluon.Trainer(net.collect_params(),'sgd',{'learning_rate':0.05})\n",
+    "utils.train(train_data,test_data,net,loss,trainer,ctx,num_epochs=5)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "collapsed": true
+   },
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python [conda env:gluon]",
+   "language": "python",
+   "name": "conda-env-gluon-py"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 2
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython2",
+   "version": "2.7.13"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}