Visualize the Entire Deep Learning Process thanks to VisualDL

A Python library, called VisualDL, can now visualize the deep learning process for us. It’s an incredibly powerful visualization tool that helps us design deep learning jobs.

Most deep neural networks these days are using Python as their primary language. Keeping this in mind, VisualDL was built to support Python. Just by adding a few lines of Python code and inserting them into our neural network model, we can generate plenty of visualizations to understand the framework.

VisualDL has also been written in low level C++.

Currently, VisualDL provides four components (more will be added soon):

• graph
• scalar
• image
• histogram

Let’s look at these in a bit more detail.

Graph

This is compatible with the Open Neural Network Exchange. In fact, VisualDL is compatible with most deep neural network frameworks, including PyTorch.

Scalar

One of the most useful components, scalar shows us the error trends during the training of the model.

👁 Image

Image

The image component can be utilised to visualize any tensor or intermediate generated image.

Histogram

Histogram is used to visualize the distribution of parameters and trends for any tensor.

Just write the below code to install VisualDL in Python:

pip install --upgrade visualdl

To give it a quick test run, run the below code:

vdl_create_scratch_log
visualDL --logdir=scratch_log --port=8080

You can access the source code and other details about this library on their official GitHub page here.

Our take on this

This is a pretty fascinating tool that gives the user a deeper insight into the deep learning process. Visualization has always been a go-to technique for data scientists when they’re stuck on an error, or want to understand what exactly is going on behind neural network they’re training. This wonderfully adaptive tool is a dream for them. And the even more exciting part is that new features are being added as we write this. Definitely one to look out for and we highly recommend you follow the page on Github.

Login to continue reading and enjoy expert-curated content.

Free Courses