Hands-On TensorBoard for PyTorch Developers

In this video, we’ll set up the regression problem.

   •  Define the problem

   •  Create a dataset

   •  Define the model

Here, we shall visualize model graph.

   •  Define model and SummaryWriter object

   •  Log model using add_graph method

   •  View graph in TensorBoard graph tab

In this video, we’ll have a look at loss curve.

   •  Train the model

   •  Log loss using add_scalar method

   •  View Loss curve in TensorBoard scalar tab

In this video, we’ll view weight and bias distributions.

   •  Train the model

   •  Log weight and bias data using add_histogram

   •  View distributions using distribution and histogram tabs in TensorBoard

In this video, we’ll demonstrate additional data visualizations.

   •  Log and view image data

   •  Log and view audio data

   •  Log and view text data

In this video, we’ll explore how to use TensorBoard to accomplish image classification.

   •  Introduce the CIFAR-10 dataset

   •  Describe model graphs and loss curves

   •  Describe confusion matrix and hyperparameter logging

Here, we shall view model graph and verify correctness.

   •  Load the data

   •  Define the model

   •  Visualize graph and detect errors in TensorBoard

Here, we shall view loss curve, visualize training loss, and other metrics.

   •  Train the model

   •  Log the training loss

   •  Visualize loss in TensorBoard

In this video, we’ll demonstrate images tab.

   •  Log image data

   •  Visualize in tab

Let’s compute confusion matrix and display in TensorBoard.

   •  Evaluate model based on test data

   •  Compute results from confusion matrix

   •  Display confusion matrix

Here, we shall predict sentiment of movie reviews.

   •  Load IMDb dataset

   •  Introduce text visualization and embeddings

   •  Discuss additional visualizations

In this video, we’ll visualize text samples in TensorBoard.

   •  Use add_text method to log data

   •  View text in text tab

Let’s demonstrate projector capability to view word embeddings.

   •  Explain what word embedding is

   •  Use add_embedding to log data

   •  View embedding using projector tab

Here, we shall view and verify the model graph.

   •  Define and instantiate the RNN model

   •  Log the model using add_graph

   •  View model in the graph tab

Here, let’s demonstrate how to tune and track hyperparameters.

   •  Set hyperparameters and train model

   •  Log hyperparameters and metrics

   •  View results in TensorBoard

In this video, we’ll identify other capabilities and what features are not supported.

   •  Describe PR Curves and mesh plots

   •  Describe unsupported features

In this video, we’ll review TensorBoard visualizations and their methods.

   •  Review images, text, scalars, and graphs

   •  Review distributions and embeddings

   •  Review hyperparameters and other features

Here, we shall review the applications and how we used TensorBoard for model development.

   •  Review data, model architectures, and training curves

   •  Review how to check for errors and monitor performance

   •  Review experimentation tracking

In this video, we’ll review other areas to explore.

   •  Look at other areas of TensorBoard

   •  Checkout TensorFlow

   •  Review PyTorch