Hi, Graphsignal founder here. After realizing that logging ML compute statistics and metrics is not optimal, we've created Graphsignal to make ML profiling and benchmarking simple and usable. It allows benchmarking and analyzing ML runs/experiments, providing performance summaries, detailed resource usage, ML operation and kernel level statistics. necessary for making training and inference faster and more efficient.

The profiler is added to ML code by simply registering it as a callback to a framework, such as Keras or Hugging Face, or using the a wrapper function around training batch/prediction in TensorFlow or PyTorch. It then automatically records and sends profiles to the dashboard. The dashboard is currently cloud-only and free to use.

I'm excited to show it to you here and will appreciate any thoughts, comments and feedback!