Shulou(Shulou.com)06/01 Report--

Editor to share with you how TensorBoard and torchsummary in Pytorch, I believe that most people do not know much about it, so share this article for your reference, I hope you can learn a lot after reading this article, let's go to know it!

In the process of using pytorch, a good visualization tool is essential. TensorBoard is such a powerful neural network visualization tool. So how do you use this tool? Please take a look at the editor's next introduction:

1.TensorBoard Neural Network Visualization tool

TensorBoard is a powerful visualization tool, and there are two invocation methods in pytorch:

1.from tensorboardX import SummaryWriter

This method was written by a great god on the Internet when the government did not support tensorboard.

2.from torch.utils.tensorboard import SummaryWriter

This method was later updated and officially added.

1.1 call method

1.1.1 create interface SummaryWriter

Function: create an interface

Call method:

Writer = SummaryWriter ("runs")

Parameters:

Log_dir:event file output folder

Comment: folder suffix when log_dir is not specified

Filename_suffix:event file file name suffix

1.1.2 record scalar add_scalars ()

Function: record scalar add_scalars ()

Call method:

Writer.add_scalars ("name", {"dic": val}, epoch)

Parameters:

Tag: the tag name of the image

Scalar_step: the scalar to record

Global_step: rounds

1.1.3 Statistical histogram add_histogram ()

Function: statistical histogram and multiquantile line chart

Call method:

Writer.add_histogram ("weight", self.fc.weight,epoch)

Parameters:

Tag: the tag name of the image

Values: data to draw a histogram

Global_step: rounds

Bins: values include 'tensorflow',' auto', 'fd', etc.

1.1.4 batch display image add_image ()

Function: batch display image

Call method:

Writer.add_image ("Cifar10", img_batch, epoch,'CHW')

Parameters:

Tag: the tag name of the image

Img_tensor: image data, pay attention to the size

Global_step: rounds

Dataformats: data form, CHW,HWC,HW

1.1.5 View model diagram add_graph ()

Function: view model diagram

Call method:

Writer.add_graph (model=net,input_to_model=torch.randn (1,224,224) .to (device))

Parameters:

Model: model, must be nn.Module

Input_to_model: data output to the model

Verbose: whether to print calculation chart structure information

Remember to write writer.close () when you finish writing.

two。 View the network layer shape and parameter torchsummary

Function: view network layer shapes and parameters

Call method:

From torchsummary import summarysummary (net, input_size= (3,224,224))

Parameters:

Model:pytorch model

Input_size: model input size

Batch_size:batch size

Device: "cuda" or "cpu"

3. Start tensorboard

Cmd opens the terminal in the file path, enter

Tensorboard-- logdir= ". / runs"

Runs is the name of the file I saved. Open the following link

Add: pytorch attempts to call the tensorboard method

Tensorboard provides a good interface for monitoring training losses, which can help us better adjust the parameters. The following describes how to call tensorboard in pytorch.

First

Install tensorboard, tensorflow, and tensorboardX

Second

Import SummaryWriter at the beginning of the file

Third in from tensorboardX import SummaryWriter

Like tensorflow's tensorboard, tensorboardX provides a variety of recording methods, such as scalar, image and so on.

Writer = SummaryWriter ('path')

If you do not add a path, it is named after the time by default.

Fourth

Add a monitoring variable

Writer.add_scalar ('Train/Acc', Acc, iter) No. 5

Open tensorboard

Tensorboard-- the sixth logdir 'path'

Open port 6006 in the browser

The above is all the contents of the article "how about TensorBoard and torchsummary in Pytorch". Thank you for reading! I believe we all have a certain understanding, hope to share the content to help you, if you want to learn more knowledge, welcome to follow the industry information channel!

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.