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.

PyTorch 1.5

150 pointsby vpjabout 5 years ago

6 comments

smhxabout 5 years ago
notable co-releases along with PyTorch 1.5:<p>- TorchServe: model serving infrastructure for scalable model deployment<p>- TorchElastic w&#x2F;Kubernetes: fault-tolerant &quot;elastic&quot; neural network training, allowing nodes to join and leave (for eg. to leverage spot pricing)<p>- Torch_XLA: updates for PyTorch TPU support<p>- New releases of torchvision, torchaudio and torchtext<p>Summary blogpost at <a href="https:&#x2F;&#x2F;pytorch.org&#x2F;blog&#x2F;pytorch-library-updates-new-model-serving-library&#x2F;" rel="nofollow">https:&#x2F;&#x2F;pytorch.org&#x2F;blog&#x2F;pytorch-library-updates-new-model-s...</a>
评论 #22944771 未加载
评论 #22943489 未加载
mratsimabout 5 years ago
I&#x27;m curious about the NHWC layout they mentioned.<p>AFAIK CuDNN always had optimizations for NCHW and that was one of Tensorflow speed issue when they choose to default on NHWC, plus the related issues on writing transformation pipelines.<p>So what does NHWC enables that is new?<p>Relevant in-depth discussion including CuDNN team lead, Julien Demouth, and Scott Gray who implemented Winograd convolution for Nervana Neon (which interestingly was CHWN so batch last): <a href="https:&#x2F;&#x2F;github.com&#x2F;soumith&#x2F;convnet-benchmarks&#x2F;issues&#x2F;93#issuecomment-192621350" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;soumith&#x2F;convnet-benchmarks&#x2F;issues&#x2F;93#issu...</a>
评论 #22947981 未加载
评论 #22945395 未加载
评论 #22945931 未加载
dynamite-readyabout 5 years ago
I like the fact that the C++ API now has the same features as the Python one. It was hard to find good C++ based NN libraries up to 2 - 4 years ago. The likes of Tensorflow had a C++ API, but the documentation was odd. Now that Facebook and Google (with Tensorflow) appear to be committed to maintaining well documented C++ API&#x27;s for their ML projects, perhaps it might draw a few people away from using Python for this work.<p>While it&#x27;s a notoriously verbose language, your deployment options do increase with C++, and you also get type safety, which seems like a good thing for ML work.
评论 #22945879 未加载
oehtXRwMkIsabout 5 years ago
Still waiting for better rocm support without docker. I wish it was at least at the level of tensorflow.
评论 #22980108 未加载
评论 #22943845 未加载
0xcoffeeabout 5 years ago
I&#x27;m not one to normally criticise some minor layout choices on a blog, but this font is really difficult to read for me.<p>On Firefox: <a href="https:&#x2F;&#x2F;i.imgur.com&#x2F;zIHis3x.png" rel="nofollow">https:&#x2F;&#x2F;i.imgur.com&#x2F;zIHis3x.png</a><p>It&#x27;s so wavy
评论 #22944317 未加载
评论 #22944573 未加载
jamistevenabout 5 years ago
Checkout PyTorch lightning.
评论 #22943823 未加载
评论 #22943777 未加载