echo/pytorch-handbook
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

修正：4.2.2-tensorboardx

Browse Source
This commit is contained in:
zergtant 2019-10-24 09:07:19 +08:00
parent 4eacfbad78
commit 792b4170df
1 changed files with 64 additions and 60 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

124
chapter4/4.2.2-tensorboardx.ipynb
View File

@ -2,16 +2,16 @@
"cells": [ "cells": [
{ {
"cell_type": "code", "cell_type": "code",
"execution_count": 1, "execution_count": 47,
"metadata": {}, "metadata": {},
"outputs": [ "outputs": [
{ {
"data": { "data": {
"text/plain": [ "text/plain": [
"'1.0.0'" "'1.3.0'"
] ]
}, },
"execution_count": 1, "execution_count": 47,
"metadata": {}, "metadata": {},
"output_type": "execute_result" "output_type": "execute_result"
} }
@ -50,9 +50,12 @@
"\n", "\n",
"`pip install tensorboard`\n", "`pip install tensorboard`\n",
"\n", "\n",
"然后再安装tensorboardx\n",
"\n", "\n",
"`pip install tensorboardx`\n", "\n",
"~~ 然后再安装tensorboardx ~~\n",
"\n",
"~~ `pip install tensorboardx` ~~\n",
"pytorch 1.1以后的版本内置了SummaryWriter 函数,所以不需要再安装tensorboardx了\n",
"\n", "\n",
"安装完成后与 visdom一样执行独立的命令\n", "安装完成后与 visdom一样执行独立的命令\n",
"`tensorboard --logdir logs` 即可启动，默认的端口是 6006,在浏览器中打开 `http://localhost:6006/` 即可看到web页面。\n", "`tensorboard --logdir logs` 即可启动，默认的端口是 6006,在浏览器中打开 `http://localhost:6006/` 即可看到web页面。\n",
@ -94,16 +97,18 @@
}, },
{ {
"cell_type": "code", "cell_type": "code",
"execution_count": 2, "execution_count": 48,
"metadata": {}, "metadata": {},
"outputs": [], "outputs": [],
"source": [ "source": [
"from tensorboardX import SummaryWriter" "# 这里的引用也要修改成torch的引用\n",
"#from tensorboardX import SummaryWriter\n",
"from torch.utils.tensorboard import SummaryWriter"
] ]
}, },
{ {
"cell_type": "code", "cell_type": "code",
"execution_count": 3, "execution_count": 49,
"metadata": {}, "metadata": {},
"outputs": [ "outputs": [
{ {
@ -112,7 +117,7 @@
"(1280, 853)" "(1280, 853)"
] ]
}, },
"execution_count": 3, "execution_count": 49,
"metadata": {}, "metadata": {},
"output_type": "execute_result" "output_type": "execute_result"
} }
@ -131,7 +136,7 @@
}, },
{ {
"cell_type": "code", "cell_type": "code",
"execution_count": 4, "execution_count": 50,
"metadata": {}, "metadata": {},
"outputs": [], "outputs": [],
"source": [ "source": [
@ -152,7 +157,7 @@
}, },
{ {
"cell_type": "code", "cell_type": "code",
"execution_count": 5, "execution_count": 51,
"metadata": {}, "metadata": {},
"outputs": [], "outputs": [],
"source": [ "source": [
@ -172,17 +177,17 @@
}, },
{ {
"cell_type": "code", "cell_type": "code",
"execution_count": 6, "execution_count": 52,
"metadata": {}, "metadata": {},
"outputs": [], "outputs": [],
"source": [ "source": [
"x = torch.FloatTensor([100])\n", "x = torch.FloatTensor([100])\n",
"y = torch.FloatTensor([500])\n", "y = torch.FloatTensor([500])\n",
"\n", "\n",
"for epoch in range(100):\n", "for epoch in range(30):\n",
" x /= 1.5\n", " x = x * 1.2\n",
" y /= 1.5\n", " y = y / 1.1\n",
" loss = y - x\n", " loss = np.random.random()\n",
" with SummaryWriter(log_dir='./logs', comment='train') as writer: #可以直接使用python的with语法，自动调用close方法\n", " with SummaryWriter(log_dir='./logs', comment='train') as writer: #可以直接使用python的with语法，自动调用close方法\n",
" writer.add_histogram('his/x', x, epoch)\n", " writer.add_histogram('his/x', x, epoch)\n",
" writer.add_histogram('his/y', y, epoch)\n", " writer.add_histogram('his/y', y, epoch)\n",
@ -190,10 +195,9 @@
" writer.add_scalar('data/y', y, epoch)\n", " writer.add_scalar('data/y', y, epoch)\n",
" writer.add_scalar('data/loss', loss, epoch)\n", " writer.add_scalar('data/loss', loss, epoch)\n",
" writer.add_scalars('data/data_group', {'x': x,\n", " writer.add_scalars('data/data_group', {'x': x,\n",
" 'y': y,\n", " 'y': y}, epoch)\n",
" 'loss': loss}, epoch)\n",
"\n", "\n",
" " " "
] ]
}, },
{ {
@ -209,7 +213,7 @@
}, },
{ {
"cell_type": "code", "cell_type": "code",
"execution_count": 7, "execution_count": 53,
"metadata": {}, "metadata": {},
"outputs": [], "outputs": [],
"source": [ "source": [
@ -226,7 +230,7 @@
}, },
{ {
"cell_type": "code", "cell_type": "code",
"execution_count": 8, "execution_count": 54,
"metadata": {}, "metadata": {},
"outputs": [], "outputs": [],
"source": [ "source": [
@ -257,7 +261,7 @@
}, },
{ {
"cell_type": "code", "cell_type": "code",
"execution_count": 9, "execution_count": 55,
"metadata": {}, "metadata": {},
"outputs": [], "outputs": [],
"source": [ "source": [
@ -275,7 +279,7 @@
" print('Train Epoch: {} [{}/{} ({:.0f}%)]\\tLoss: {:.6f}'.format(\n", " print('Train Epoch: {} [{}/{} ({:.0f}%)]\\tLoss: {:.6f}'.format(\n",
" epoch, batch_idx * len(data), len(train_loader.dataset),\n", " epoch, batch_idx * len(data), len(train_loader.dataset),\n",
" 100. * batch_idx / len(train_loader), loss.item()))\n", " 100. * batch_idx / len(train_loader), loss.item()))\n",
" #主要增加了一下内容\n", " #相对于以前的训练方法 主要增加了以下内容\n",
" out = torch.cat((output.data, torch.ones(len(output), 1)), 1) # 因为是投影到3D的空间，所以我们只需要3个维度\n", " out = torch.cat((output.data, torch.ones(len(output), 1)), 1) # 因为是投影到3D的空间，所以我们只需要3个维度\n",
" with SummaryWriter(log_dir='./logs', comment='mnist') as writer: \n", " with SummaryWriter(log_dir='./logs', comment='mnist') as writer: \n",
" #使用add_embedding方法进行可视化展示\n", " #使用add_embedding方法进行可视化展示\n",
@ -295,19 +299,16 @@
}, },
{ {
"cell_type": "code", "cell_type": "code",
"execution_count": 10, "execution_count": 56,
"metadata": {}, "metadata": {},
"outputs": [ "outputs": [
{ {
"name": "stdout", "name": "stdout",
"output_type": "stream", "output_type": "stream",
"text": [ "text": [
"Train Epoch: 0 [14848/60000 (25%)]\tLoss: 0.271775\n", "Train Epoch: 0 [14848/60000 (25%)]\tLoss: 0.352312\n",
"warning: Embedding dir exists, did you set global_step for add_embedding()?\n", "Train Epoch: 0 [30208/60000 (50%)]\tLoss: 0.202950\n",
"Train Epoch: 0 [30208/60000 (50%)]\tLoss: 0.175213\n", "Train Epoch: 0 [45568/60000 (75%)]\tLoss: 0.156494\n"
"warning: Embedding dir exists, did you set global_step for add_embedding()?\n",
"Train Epoch: 0 [45568/60000 (75%)]\tLoss: 0.115128\n",
"warning: Embedding dir exists, did you set global_step for add_embedding()?\n"
] ]
} }
], ],
@ -321,6 +322,8 @@
"source": [ "source": [
"打开 `http://localhost:6006/#projector` 即可看到效果。\n", "打开 `http://localhost:6006/#projector` 即可看到效果。\n",
"\n", "\n",
"目前测试投影这部分也是有问题的，根据官网文档的代码进行测试，也显示不出来，正在找原因\n",
"\n",
"### 绘制网络结构\n", "### 绘制网络结构\n",
"在pytorch中我们可以使用print直接打印出网络的结构，但是这种方法可视化效果不好，这里使用tensorboard的GRAPHS来实现网络结构的可视化。\n", "在pytorch中我们可以使用print直接打印出网络的结构，但是这种方法可视化效果不好，这里使用tensorboard的GRAPHS来实现网络结构的可视化。\n",
"由于pytorch使用的是动态图计算，所以我们这里要手动进行一次前向的传播.\n", "由于pytorch使用的是动态图计算，所以我们这里要手动进行一次前向的传播.\n",
@ -330,7 +333,7 @@
}, },
{ {
"cell_type": "code", "cell_type": "code",
"execution_count": 11, "execution_count": 57,
"metadata": {}, "metadata": {},
"outputs": [ "outputs": [
{ {
@ -340,44 +343,45 @@
"VGG(\n", "VGG(\n",
" (features): Sequential(\n", " (features): Sequential(\n",
" (0): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n", " (0): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (1): ReLU(inplace)\n", " (1): ReLU(inplace=True)\n",
" (2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n", " (2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (3): ReLU(inplace)\n", " (3): ReLU(inplace=True)\n",
" (4): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n", " (4): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
" (5): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n", " (5): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (6): ReLU(inplace)\n", " (6): ReLU(inplace=True)\n",
" (7): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n", " (7): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (8): ReLU(inplace)\n", " (8): ReLU(inplace=True)\n",
" (9): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n", " (9): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
" (10): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n", " (10): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (11): ReLU(inplace)\n", " (11): ReLU(inplace=True)\n",
" (12): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n", " (12): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (13): ReLU(inplace)\n", " (13): ReLU(inplace=True)\n",
" (14): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n", " (14): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (15): ReLU(inplace)\n", " (15): ReLU(inplace=True)\n",
" (16): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n", " (16): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
" (17): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n", " (17): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (18): ReLU(inplace)\n", " (18): ReLU(inplace=True)\n",
" (19): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n", " (19): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (20): ReLU(inplace)\n", " (20): ReLU(inplace=True)\n",
" (21): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n", " (21): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (22): ReLU(inplace)\n", " (22): ReLU(inplace=True)\n",
" (23): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n", " (23): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
" (24): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n", " (24): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (25): ReLU(inplace)\n", " (25): ReLU(inplace=True)\n",
" (26): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n", " (26): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (27): ReLU(inplace)\n", " (27): ReLU(inplace=True)\n",
" (28): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n", " (28): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" (29): ReLU(inplace)\n", " (29): ReLU(inplace=True)\n",
" (30): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n", " (30): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
" )\n", " )\n",
" (avgpool): AdaptiveAvgPool2d(output_size=(7, 7))\n",
" (classifier): Sequential(\n", " (classifier): Sequential(\n",
" (0): Linear(in_features=25088, out_features=4096, bias=True)\n", " (0): Linear(in_features=25088, out_features=4096, bias=True)\n",
" (1): ReLU(inplace)\n", " (1): ReLU(inplace=True)\n",
" (2): Dropout(p=0.5)\n", " (2): Dropout(p=0.5, inplace=False)\n",
" (3): Linear(in_features=4096, out_features=4096, bias=True)\n", " (3): Linear(in_features=4096, out_features=4096, bias=True)\n",
" (4): ReLU(inplace)\n", " (4): ReLU(inplace=True)\n",
" (5): Dropout(p=0.5)\n", " (5): Dropout(p=0.5, inplace=False)\n",
" (6): Linear(in_features=4096, out_features=1000, bias=True)\n", " (6): Linear(in_features=4096, out_features=1000, bias=True)\n",
" )\n", " )\n",
")\n" ")\n"
@ -398,7 +402,7 @@
}, },
{ {
"cell_type": "code", "cell_type": "code",
"execution_count": 12, "execution_count": 58,
"metadata": {}, "metadata": {},
"outputs": [], "outputs": [],
"source": [ "source": [
@ -420,7 +424,7 @@
}, },
{ {
"cell_type": "code", "cell_type": "code",
"execution_count": 13, "execution_count": 59,
"metadata": {}, "metadata": {},
"outputs": [ "outputs": [
{ {
@ -429,7 +433,7 @@
"torch.Size([1, 3, 224, 224])" "torch.Size([1, 3, 224, 224])"
] ]
}, },
"execution_count": 13, "execution_count": 59,
"metadata": {}, "metadata": {},
"output_type": "execute_result" "output_type": "execute_result"
} }
@ -448,7 +452,7 @@
}, },
{ {
"cell_type": "code", "cell_type": "code",
"execution_count": 14, "execution_count": 60,
"metadata": {}, "metadata": {},
"outputs": [ "outputs": [
{ {
@ -457,7 +461,7 @@
"287" "287"
] ]
}, },
"execution_count": 14, "execution_count": 60,
"metadata": {}, "metadata": {},
"output_type": "execute_result" "output_type": "execute_result"
} }
@ -478,19 +482,19 @@
}, },
{ {
"cell_type": "code", "cell_type": "code",
"execution_count": 15, "execution_count": 61,
"metadata": {}, "metadata": {},
"outputs": [], "outputs": [],
"source": [ "source": [
"with SummaryWriter(log_dir='./logs', comment='vgg16') as writer:\n", "with SummaryWriter(log_dir='./logs', comment='vgg161') as writer:\n",
" writer.add_graph(vgg16, (vgg16_input,))" " writer.add_graph(vgg16, vgg16_input)"
] ]
}, },
{ {
"cell_type": "markdown", "cell_type": "markdown",
"metadata": {}, "metadata": {},
"source": [ "source": [
"打开tensorboard找到graphs就可以看到vgg模型具体的架构了" "对于Pytorch的1.3版本来说，实测 SummaryWriter在处理结构图的时候是有问题的（或者是需要加什么参数，目前我还没找到），所以建议大家继续使用tensorboardx。"
] ]
}, },
{ {
@ -503,9 +507,9 @@
], ],
"metadata": { "metadata": {
"kernelspec": { "kernelspec": {
"display_name": "pytorch 1.0", "display_name": "deep learning",
"language": "python", "language": "python",
"name": "pytorch1" "name": "dl"
}, },
"language_info": { "language_info": {
"codemirror_mode": { "codemirror_mode": {
@ -517,7 +521,7 @@
"name": "python", "name": "python",
"nbconvert_exporter": "python", "nbconvert_exporter": "python",
"pygments_lexer": "ipython3", "pygments_lexer": "ipython3",
"version": "3.6.6" "version": "3.6.9"
} }
}, },
"nbformat": 4, "nbformat": 4,