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。

Fastai for PyTorch: Fast and accurate neural nets using modern best practices

279 点作者 stablemap超过 6 年前

13 条评论

jph00超过 6 年前
Jeremy from fast.ai here - I&#x27;m sitting here at the PyTorch developer conference, listening to all the really great new stuff being announced. :) Happy to answer any questions about the fastai library release here. It&#x27;s a rewrite from scratch compared to v0.7.<p>Docs here: <a href="http:&#x2F;&#x2F;docs.fast.ai" rel="nofollow">http:&#x2F;&#x2F;docs.fast.ai</a> . GitHub repo here: <a href="https:&#x2F;&#x2F;github.com&#x2F;fastai&#x2F;fastai" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;fastai&#x2F;fastai</a> . It&#x27;s also available now on Google Cloud Platform, including example notebooks and datasets (Viacheslav Kovalevskyi from Google has posted a walk-thru here: <a href="https:&#x2F;&#x2F;blog.kovalevskyi.com&#x2F;google-compute-engine-now-has-images-with-pytorch-1-0-0-and-fastai-1-0-2-57c49efd74bb" rel="nofollow">https:&#x2F;&#x2F;blog.kovalevskyi.com&#x2F;google-compute-engine-now-has-i...</a> ). AWS support coming soon.
评论 #18126074 未加载
评论 #18124736 未加载
评论 #18123918 未加载
评论 #18125331 未加载
评论 #18125120 未加载
评论 #18128187 未加载
评论 #18124748 未加载
评论 #18124396 未加载
评论 #18126076 未加载
评论 #18124433 未加载
评论 #18128098 未加载
评论 #18125511 未加载
pitchups超过 6 年前
This is great news! Here has been my experience using Fast.ai : I had been training a deep learning network using Keras with Tensorflow, for diagnosing medical images - and it took me several months of hard work - tweaking parameters, training and testing to get acceptable levels of accuracy for our models. And then last month, I switched to Fast.ai (their pre-release version) and I was blown away - my models trained faster, and I matched and finally exceeded accuracy levels acheived with my earlier models. And I accomplished what had taken several months in Keras, in just a few days! And the biggest reasons for it were in my view, fast.ai&#x27;s learning rate finder, the differential learning rates, and Test Time augmentation - all which are advanced features built into fast.ai. And the other great thing is that fast.ai uses the best defaults automatically, and it trains much, much faster than Keras &#x2F; TF for some reason.<p>So I can&#x27;t wait to try the new release out. I think Fast.ai has set a new bar for deep learning frameworks in terms of speed and ease of use. Thank you for all your great work!
评论 #18126786 未加载
评论 #18124753 未加载
评论 #18127454 未加载
screye超过 6 年前
I haven&#x27;t used fast.ai yet, but extensively use pytorch for anything to do with deeplearning. (it is so much better than tensorflow) It&#x27;s great to see a keras like higher level abstraction library for deep learning.<p>A few questions:<p>1. what is the benefit of using fast.ai to someone well acquainted with pytorch (for academic use)<p>2. how well does fast.ai interface with pytorch itself ? Can parts of program be in fast ai and other parts be in pytorch ?<p>3. Am I correct in assuming that despite being very fast, fast.ai is still slower (even if marginally so) than pytorch itself ?
评论 #18125705 未加载
评论 #18125733 未加载
denfromufa超过 6 年前
@jph00 where do I find lstm&#x2F;gru&#x2F;seq2seq layers for time-series sequence predictions (not text)? Also interested in autoencoder implementations. The fast.ai docs search does not really work for this. What do you think about other notable APIs built on top of pytorch such as Pyro and AllenNLP?
sharcerer超过 6 年前
Hi Jeremy, I am a beginner. Going to start your ML course,then follow up with DL- 1 &amp; 2. My question is when will these 3 be updated to the new version of fastai. I am aware of the fellowship for DL-1, but there&#x27;s a 12 hour difference. Also, I want to start with ML,so, I want to learn with the new version. Thanks<p>On the side, I am also taking Andrew&#x27;s coursera course for a theoretical grounding , Bengio and Goodfellow&#x27;s Machine Learning Book ,and Hands-On Machine Learning with Tensorflow and Sci-kit( O&#x27;reilly book).
montenegrohugo超过 6 年前
Cannot recommend fast.ai enough. How they and Jeremy especially communicate is absolutely outstanding, and I guarantee you will learn a lot by the results-driven approach that they have.
jph00超过 6 年前
ZDNet just published an article with some more background about this BTW : <a href="https:&#x2F;&#x2F;www.zdnet.com&#x2F;google-amp&#x2F;article&#x2F;fast-ais-new-software-could-radically-democratize-ai" rel="nofollow">https:&#x2F;&#x2F;www.zdnet.com&#x2F;google-amp&#x2F;article&#x2F;fast-ais-new-softwa...</a>
yazr超过 6 年前
Can this be used &#x2F; intended for production ? Is it fast&#x2F;efficient enough on common hw ?<p>We are not really doing cutting edge or research stuff. Just developing a big dumb resnet and hoping to scale up to 10s of GPUs over 2019.<p>I was really happy to read a comment here about how this framework was used to reproduce a 6 month project in 2 weeks :)
评论 #18125284 未加载
binalpatel超过 6 年前
I&#x27;m a big fan - I&#x27;ve been using their old version of their library for NLP on medical text, but the course and library certainly removed a lot of the &quot;mystique&quot; around deep learning for me.
mkolodny超过 6 年前
This is awesome! I&#x27;d love to use Fastai, but my company&#x27;s pretty committed to Tensorflow right now. Is there any work being done to port Fastai over to Tensorflow?
qwerty456127超过 6 年前
Is it meant to be a better alternative to Keras?
评论 #18125295 未加载
synaesthesisx超过 6 年前
Is there a guide available for image segmentation? I&#x27;ve only really used TF in the past
评论 #18127222 未加载
zeptomu超过 6 年前
The name fastai sounds suspicious. There is &quot;fast&quot; and &quot;AI&quot; in it, so I fear that it makes the easy things easier and the hard things harder. It just sounds like too much fluff. We obviously don&#x27;t have AI yet (but pretty convincing machine learning models) and we won&#x27;t get it in the next years (not fast) - so if software, a company or an organization puts &quot;ai&quot; in its name (or chooses it as top-level domain) I suspect they just want to ride the hype train and am <i>very</i> skeptical ...<p>@Topic: What is the difference to Keras, PyTorch, etc.? They are already pretty high-level and the basic models for common tasks are available in all libraries at that point.
评论 #18129122 未加载
评论 #18127853 未加载