metadata
license: apache-2.0
language:
- en
pretty_name: Diffusers Benchmarks
Welcome to 🤗 Diffusers Benchmarks!
This is dataset where we keep track of the inference latency and memory information of the core pipelines in the diffusers
library.
Currently, the core pipelines are the following:
- Stable Diffusion and its derivatives such as ControlNet, T2I Adapter, Image-to-Image, Inpainting
- Stable Diffusion XL and its derivatives
- SSD-1B
- Kandinsky
- Würstchen
- LCM
Note that we will continue to extend the list of core pipelines based on their API usage.
We use this GitHub Actions workflow to report the above numbers automatically. This workflow runs on a biweekly cadence.
The benchmarks are run on an A10G GPU.