soumith/convnet-benchmarks on Hacker News

…expert so let's talk about CuDNN where optimizing convolution is all about memory layout, source: https://github.com/soumith/convnet-benchmarks/issues/93#issu... and it's not about C vs C++ vs PTX. Or let's hear about what Nvidia says…

comment by mratsim · 2020-04-22 · in A possible new back end for Rust

…who implemented Winograd convolution for Nervana Neon (which interestingly was CHWN so batch last): https://github.com/soumith/convnet-benchmarks/issues/93#issu...

comment by mratsim · 2020-04-22 · in PyTorch 1.5

…neural network kernels (which AFAIK haven't been updated since) are still the fastests available: https://github.com/soumith/convnet-benchmarks

comment by nl · 2018-08-09 · in Shannon Labs – $100K Fellowship to support independent researchers

…e.g., see their different optimizers [3] [1]: http://stsievert.com/blog/2017/09/07/pytorch/ [2]: https://github.com/soumith/convnet-benchmarks [3]: https://github.com/pytorch/pytorch/tree/master/torch/optim [4]:…

comment by skierscott · 2017-09-29 · in MILA and the future of Theano

…has replicated these claims. MS hasn't got it in the standard CNN inference benchmark yet though: https://github.com/soumith/convnet-benchmarks . This is the benchmark that showed how slow TensorFlow was on non-Google libraries until…

comment by nl · 2016-10-26 · in Cognitive Toolkit beta for deep learning advances

There is another nice one here: https://github.com/soumith/convnet-benchmarks

comment by TheAlchemist · 2016-09-07 · in Simple Deep Learning Benchmark

…TesnorFlow in AlexNet and ResNet. There is a set of benchmarks for many DL approaches as found in https://github.com/soumith/convnet-benchmarks

comment by breezest · 2016-08-27 · in Benchmarking State-Of-the-Art Deep Learning Software Tools

…to head 3rd party benchmarking fights like Soumith Chintala's fantastic convnet benchmark data: https://github.com/soumith/convnet-benchmarks

comment by scottlegrand · 2016-08-17 · in Correcting Intel's Deep Learning Benchmark Mistakes

…have separate benchmarks for Soumith's convnet comparison that show it to be significantly slower? https://github.com/soumith/convnet-benchmarks

comment by vonnik · 2016-07-24 · in Google Sprints Ahead in AI Building Blocks, Leaving Rivals Wary

What's Deep Learning 4J's performance on the Soumith benchmarks? https://github.com/soumith/convnet-benchmarks

comment by AlexCoventry · 2016-07-10 · in Generating Recommendations at Amazon Scale with Apache Spark and Amazon DSSTNE

These are the benchmarks supported by Soumith Chintala (Facebook AI Research): https://github.com/soumith/convnet-benchmarks They place TensorFlow performance on par with Torch (within 10%).

comment by krasin · 2016-05-31 · in TensorFlow – Consise Examples for Beginners

…this so it's actually sort of the least locked in framework. [1] It has been climbing the charts at https://github.com/soumith/convnet-benchmarks for example.

comment by micro_cam · 2016-05-18 · in Google supercharges machine learning tasks with TPU custom chip

…with TitanX/GTX 980 TI now topping 5000 images/s. GTX 1080 will only be faster than this and $600. https://github.com/soumith/convnet-benchmarks And so far, no FPGA training numbers. But here are some distributed training numbers for…

comment by varelse · 2016-05-08 · in NVIDIA Announces the GeForce GTX 1000 Series

…(Deep Learning's most prominent benchmark): https://github.com/amd/OpenCL-caffe versus GTX TitanX: https://github.com/soumith/convnet-benchmarks TLDR: AMD losing by a factor of 2 or more... So unless this changes dramatically, NVIDIA IMO…

comment by varelse · 2016-05-07 · in NVIDIA Announces the GeForce GTX 1000 Series

And even those numbers on the front page are out of date :) (we're even faster now: https://github.com/soumith/convnet-benchmarks/pull/96 , which is from a few weeks ago.) The field is moving quickly enough that many published benchmarks…

comment by vrv · 2016-04-13 · in Announcing TensorFlow 0.8 – now with distributed computing support

…of date - it benchmarked the CuDNNv2 version. Soumith's convnet-benchmarks is much more up-to-date: https://github.com/soumith/convnet-benchmarks but it hasn't yet been updated to reflect the latest performance improvements in 0.8. We've…

comment by dgacmu · 2016-04-13 · in Announcing TensorFlow 0.8 – now with distributed computing support

…implementations is probably worth noting in this context. They perform (speed-wise) pretty well: https://github.com/soumith/convnet-benchmarks

comment by nl · 2016-03-20 · in XNOR-Net: ImageNet Classification Using Binary Convolutional Neural Networks

…about distributed computing performance [3]. [0] http://autumnai.com/deep-learning-benchmarks [1] https://github.com/soumith/convnet-benchmarks [2] https://github.com/NervanaSystems/neon [3] http://alex.smola.org/talks/NIPS15.pdf

comment by emcq · 2016-03-08 · in Leaf: Machine learning framework in Rust

…could publish with confidence yet. [1]: https://github.com/autumnai/deep-learning-benchmarks [2]: https://github.com/soumith/convnet-benchmarks

comment by hobofan · 2016-03-08 · in Leaf: Machine learning framework in Rust

…be quite easy, if the kernels exist. [1]: https://github.com/autumnai/deep-learning-benchmarks [2]: https://github.com/soumith/convnet-benchmarks [3]: https://github.com/autumnai/collenchyma-nn

comment by mjhirn · 2016-03-08 · in Machine Learning in Rust

…models in rust. For more benchmarks (including updated TensorFlow performance with cudnn v4) see https://github.com/soumith/convnet-benchmarks

comment by jimfleming · 2016-03-04 · in Deep Learning Benchmarks

Can't wait for sousmith's take https://github.com/soumith/convnet-benchmarks/issues

comment by blazespin · 2016-01-26 · in Microsoft releases CNTK, its open source deep learning toolkit, on GitHub

According to: https://github.com/soumith/convnet-benchmarks , it's an NVIDIA TitanX (12GB GPU memory), which is pretty much the top of the line GPU for training neural nets. From my limited experience, most deep learning in research -…

comment by brianchu · 2015-11-11 · in TensorFlow Benchmarks

…but that doesn't help, since most people use nVidia (eg, Amazon GPU instances are nVidia based) [1] https://github.com/soumith/convnet-benchmarks

comment by nl · 2015-11-11 · in Leaf: machine intelligence framework in Rust

Here's a page with various benchmarks: https://github.com/soumith/convnet-benchmarks An issue has been created to add TensorFlow to this shortly.

comment by miket · 2015-11-10 · in Jeff Dean explains TensorFlow [video]

…NVIDIA is the best way to get them to address deficiencies like this. This is mostly already fixed: https://github.com/soumith/convnet-benchmarks Check out the huge speedup between cuDNN(R2) and cuDNN(R3). Also, while there is some…

comment by varelse · 2015-09-26 · in Fast Convolutional Nets with Fbfft: A GPU Performance Evaluation

…for speed, here's some benchmarks of different libraries on ImageNet and convolutional layers: https://github.com/soumith/convnet-benchmarks

comment by betelgeusem · 2015-08-05 · in Faster deep learning with GPUs and Theano

…will actually use cuDNN if available, and performance will be slightly better than Torch and Caffe: https://github.com/soumith/convnet-benchmarks#layer-wise-ben...

comment by fchollet · 2015-05-12 · in Keras: Fast Deep Learning Prototyping for Python

There are some benchmarks of a neural network toolkit built on top of this: https://github.com/soumith/convnet-benchmarks compare NVIDIA's own cuDNN R2 versus NervanaSys-16 and NervanaSys-32. Pretty impressive! I've also tried out his GEMM…

comment by benanne · 2015-05-04 · in Assembler for Nvidia Maxwell architecture

…Chintala maintains a GitHub repository with benchmarks of various convolution implementations: https://github.com/soumith/convnet-benchmarks His results aren't that spectacular either. In my own experiments cuDNN did pretty well for very…

comment by benanne · 2014-09-29 · in Nvidia Introduces CuDNN, a CUDA-based Library for Deep Neural Networks

…. There are some benchmarks for the convolutional layers of a classic AlexNet model at https://github.com/soumith/convnet-benchmarks . In general, the question which is the fastest public implementation of spatial convolution is so heavily…

comment by ajtulloch · 2014-09-29 · in Nvidia Introduces CuDNN, a CUDA-based Library for Deep Neural Networks