TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

ONNX Runtime merges WebGPU backend

178 pointsby b_mc2about 2 years ago

8 comments

minimaxirabout 2 years ago
Some context for those who aren&#x27;t in the loop: ONNX Runtime (<a href="https:&#x2F;&#x2F;onnxruntime.ai&#x2F;" rel="nofollow">https:&#x2F;&#x2F;onnxruntime.ai&#x2F;</a>) is a standardization format for AI models. Nowadays, it&#x27;s extremely easy to export models in the ONNX format, especially language models with tools like Hugging Face transformers which have special workflows for it.<p>ONNX support in the browser was lacking and limited to CPU, but with a WebGPU backend it may now finally be feasible to run models in the browser on a GPU, which opens up interesting oppertunities. Although from this PR it looks like only a few operations are implemented, no browser-based GPT yet.
评论 #35695335 未加载
评论 #35695868 未加载
pumanoirabout 2 years ago
A great option but there is wonnx which seems to be more complete and mature. And the bonus is that it&#x27;s implemented in Rust (if you are into it).<p><a href="https:&#x2F;&#x2F;github.com&#x2F;webonnx&#x2F;wonnx">https:&#x2F;&#x2F;github.com&#x2F;webonnx&#x2F;wonnx</a>
评论 #35697245 未加载
评论 #35699013 未加载
CSSerabout 2 years ago
Okay, I’ve just got to know: what’s up with the commit messages? I’ve never seen anyone just straight up use numbers before.
评论 #35695276 未加载
评论 #35695235 未加载
评论 #35699531 未加载
评论 #35696751 未加载
b_mc2about 2 years ago
A pretty cool library that uses ONNX is transformers.js [1] and they&#x27;re already working to add WebGPU support.[2]<p>[1] <a href="https:&#x2F;&#x2F;xenova.github.io&#x2F;transformers.js&#x2F;" rel="nofollow">https:&#x2F;&#x2F;xenova.github.io&#x2F;transformers.js&#x2F;</a><p>[2] <a href="https:&#x2F;&#x2F;twitter.com&#x2F;xenovacom&#x2F;status&#x2F;1650634015060156420" rel="nofollow">https:&#x2F;&#x2F;twitter.com&#x2F;xenovacom&#x2F;status&#x2F;1650634015060156420</a>
jarymabout 2 years ago
As an aside, I love ONNX and the main reason I&#x27;m sticking with PyTorch. I was able to develop and train an RL model in Python and then convert it to ONNX and call it from C# production code.<p>It still took a lot of effort but the final version is very performant and reliable.
Culonavirusabout 2 years ago
It would be great if, one lovely day, we could just slap together an electron app and run inference through webgpu hassle-free and cross-platform.
WorldPeasabout 2 years ago
Quite the uh... interesting commit strategy...
synergy20about 2 years ago
interesting, is there a helloworld example or tutorial somewhere to check out how this works in real?