Closed jboarman closed 1 year ago
Many related PRs:
These PRs resulted in greater than 100% improvement to speed, primarily resulting from optimization via use of Numba.
See updated benchmark results: https://github.com/sparkfish/augraphy/tree/dev/benchmark
We need an initial set of timings captured to establish a baseline for where we stand now for each augmentation.
Once we have that baseline, we should also take a first pass at optimizing using Numba to optimize loops. A separate issue will capture other types of improvements, such as leveraging GPUs and other techniques.