Analyzing Model Performance using Tensorflow Profiler

Model performance is always a important aspect while training or tuning deep learning models. It is often left to experience or referred to certain previous benchmarks on different data and made analysis upon.

[TF Blog] Performance is a key consideration of successful ML research and production solutions. Faster model training leads to faster iterations and reduced overhead. It is sometimes an essential requirement to make a particular ML solution feasible. However it is not clear what needs to be optimized, whether the model or the input pipeline needs tweaking or a call to some operation.

This talk will focus on identifying the areas that are maybe slowing down model performance using a tool from tensorflow called Tensorflow profiler.

Learn more about tensorflow profiler.

This talk will focus on :

• General ML model performance roadblocks faced.
• Understanding efficient ways to analyze model performance.
  • Introduction to TF Profiler
  • Brief on how’s and what’s of profiler.
  • TF profiler as a tool.
• Do keras calls to profiler itself affect performance?
• TF Profiler demo using Google colab and tensorboard.