Datasets:

Modalities:
Tabular
Text
Formats:
csv
Languages:
English
Size:
< 1K
Libraries:
Datasets
pandas
License:
benchmarks / README.md
sayakpaul's picture
sayakpaul HF staff
Update README.md
123be51 verified
|
raw
history blame
983 Bytes
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.