Open seabbs opened 2 weeks ago
We could also present this as part of our user documentation potentially as it could be helpful to users when deciding which AD backend to use for a specific model (by looking at the performance of the specific components they are using).
I think we can probably make some useful plots/do data analysis on our current benchmarking to understand performance bottles necks (especially when comparing different AD backends).