Improve performance of onnx.ai/models
ramkrishna2910 opened this issue · 2 comments
ONNX Model Zoo is now accessed through a webpage onnx.ai/models. This webpage is powered by github pages that reads the directory strucutre from github.com/onnx/models to populate the webpage.
The initial load of the webpage takes > 5sec due to the need to parse 1 yaml file per model to read Author, license and task information.
Proposed Solution:
Add a post merge hook in github.com/onnx/models that generates a json file on every merge that has the following structure per onnx model:
Model Name:
URL
Task
Author Name
License
This file can be then consumed by github pages instead of parsing 1000's of yaml files at load time.
I’ll take a look at this
@ramkrishna2910, @shrinivas2009in, I think we should be able to use github actions, which are server-side JSON generation in this case, instead of post-merge in git, which would be client-side implementation.