Datasets:

Modalities:
Tabular
Text
Formats:
csv
Languages:
English
Size:
< 1K
Libraries:
Datasets
pandas
License:
File size: 826 Bytes
fa1ead1
 
 
 
 
 
3438ea3
 
 
 
 
 
 
 
 
 
 
 
 
1a55691
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
---
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](https://github.com/huggingface/diffusers/blob/main/.github/workflows/benchmark.yml) to report the above numbers automatically. This workflow runs on a biweekly cadence. 

The benchmarks are run on an A10G GPU.