TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Faster Quantized Neural Network Inference with XNNPack

18 点作者 Marat_Dukhan超过 3 年前

3 条评论

davidatbu超过 3 年前
Looking at posts from a couple of years back on HN&#x2F;Reddit&#x2F;SO about TF vs Pytorch, the only plus side of using TF was the ease of deployment, especially on the mobile side with Tensorflow Lite.<p>But I imagine that story is changing with the advent of Pytorch Mobile, ONNX, and that Pytorch itself supports XNNPack.<p>If anyone has any tips or insights as to ease of mobile deployment using TF vs using Pytorch, please share!
aborsy超过 3 年前
Can it perform fixed point arithmetic with arbitrary number of bits?<p>Both training-aware and post training.
评论 #28527793 未加载
评论 #28522137 未加载
Marat_Dukhan超过 3 年前
Author here, happy to take your questions.
评论 #28528038 未加载
评论 #28520597 未加载
评论 #28521388 未加载
评论 #28521708 未加载