Soumith Chintala

vgg model checkpoint needs a change of classifier weight names

2

See https://discuss.pytorch.org/t/upgrading-torchvision-module-makes-old-model-useless/1719 The reason this changed was because of https://github.com/pytorch/vision/pull/107 where Sam realized that he put dropout in the wrong location. So the state_dict needs the names changed appropriately.

DeepMark

125

Hi all, The reason I've been slow on convnet-benchmarks these days is because i've been working on the side on DeepMark. I initially wrote _convnet-benchmarks_ to increase competition among frameworks...

benchmark MXNet and Chainer. Compare with TensorFlow and others.

17

[reserved for review]

[October 2015] Intel are CPU magicians. But there's no one weird trick....

89

Intel released a small blog-post recently covering that they have crazy-talk speeds for ConvNets on their Haswell CPU line. I took their Caffe implementation, painfully installed the dependencies, and the...

Nervana's Neon and Winograd

68

After serious perf improvements by NVIDIA's CUDNN R4 across board, I suppose Nervana weren't too happy to be left behind. They've just released (as part of Neon) their Winograd-based kernels...

[August 2015] Rejigging the marks...

16

The benchmarks this time around are interesting, with some fairly clear trends emerging for the near future. ### Looking Back First, some appreciation for where things are, - 9 months...

Jigar Doshi ( @artvandelay ) has volunteered to take a crack at this: https://twitter.com/jigarkdoshi/status/691758082075070464

[December 2014] benchmarking Imagenet winners

40

Time to take these benchmarks forward to a more meaningful metric (it's taken so long, but it's after all a side project for fun). I've added benchmarks for the following...

[April 2015] Revamp Benchmarks, move to Titan-X (Digits box)

23

List of libraries to rerun for Titan-X: Layer-wise benchmarks - [x] - Caffe - [x] - CuDNN - [x] - Torch - [x] - FBFFT - [x] - Theano -...

https://github.com/akrizhevsky/cuda-convnet2/pull/16 "Considerable speedup(1.5x under VGG model with miniBatch of 32, 1.1x under AlexNet with miniBatch of 128), and the optimizations focus on fully employing gpu-releated functions." - @bestimage-tencent