Parsing and Querying Your Tensorboard logs: A Mini Tutorial

Photo by Mika Baumeister on Unsplash

So, you wanted to parse your Tensorboard logs, didn’t you? Did you try using GPT-3? OH! GPT-4? Well. Guess that didn’t give you what you wanted. Yeah, me neither.

So here we are. Read on and you will find out how to take all the runs you logged to Tensorboard, clean them up, and put them in a single DataFrame. From there, you can query it as you would any other table.

Note : As someone pointed out, there are libraries that do this. For example TBParse is a good one. This article is to show you how easy it is to write your own and extend it. If you do want a proper library, do use that instead.

Contents

∘ What do we want?
∘ Shush and show me the code?!
∘ Imports
∘ Find TFevent files
∘ Create DataFrame
∘ Process an Event
∘ Scalars
∘ Tensors
∘ Images
∘ Other formats
∘ A Caveat
∘ Load and Clean the DataFrame
∘ Display Images
∘ What’s Next?
∘ Why not Wandb/W&B etc etc
∘ Fin