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.

Show HN: Backprop – a simple library to use and finetune state-of-the-art models

5 pointsby kristooabout 4 years ago

2 comments

kristooabout 4 years ago
This is a PyTorch based library my team and I have been working on for the last few months, with the goal of making finetuning and using models as easy as possible for devs, even without extensive ML experience.<p>We&#x27;ve currently got support for text and image-based tasks (classification, generation, q&amp;a, etc.), with wrappers around models like Google&#x27;s T5, OpenAI&#x27;s CLIP, GPT-2, Facebook&#x27;s BART, and others.<p>We&#x27;ve got some features that make deployment easy, but for full transparency, it is through a paid platform [0] we&#x27;ve developed that is by no means necessary to use the library.<p>We&#x27;re happy with the progress we&#x27;ve made, but we&#x27;re curious to hear what people think so we can keep improving.<p>[0] <a href="https:&#x2F;&#x2F;backprop.co" rel="nofollow">https:&#x2F;&#x2F;backprop.co</a>
friendofafriendabout 4 years ago
Just commented on your last post so copying here:<p>This looks pretty slick! Can you give any sort of average number of seconds your pre-trained models run for? Just curious approximately how many API calls one could make on each tier given the per second usage pricing.
评论 #26577369 未加载