sayakpaul/keras-xla-benchmarks

Presents comprehensive benchmarks of XLA-compatible pre-trained models in Keras.

29
/ 100
Experimental

This project offers insights into how different pre-trained computer vision models perform on various GPUs when XLA (Accelerated Linear Algebra) is enabled. It takes a list of Keras vision models and different image resolutions, then measures their throughput (images processed per second). The results help machine learning engineers, MLOps specialists, and researchers choose the most efficient models and hardware for image-related tasks.

No commits in the last 6 months.

Use this if you are a machine learning practitioner looking to optimize the inference speed of your computer vision models using XLA on various GPU types.

Not ideal if you are looking for benchmarks on model accuracy or performance metrics beyond throughput, or if you are not working with Keras vision models.

computer-vision model-optimization GPU-performance MLOps deep-learning-inference
Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 7 / 25
Maturity 16 / 25
Community 6 / 25

How are scores calculated?

Stars

37

Forks

2

Language

Jupyter Notebook

License

Apache-2.0

Last pushed

Aug 27, 2023

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/transformers/sayakpaul/keras-xla-benchmarks"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.