Nicholas Léonard comments

Results 54 comments of


                                            Nicholas Léonard

[August 2014] Discussion of results

Are you sure caffe also updates the parameters (or accumulates the parameter gradients) in the code?

[August 2014] Discussion of results

Yeah your link seems to confirm it. Yet the difference between forward and backward is so much smaller for caffe. I wonder what their secret is. Isn't SpatialConvolutionMM supposed to...

[August 2014] Discussion of results

@f0k I think a call `cudaDeviceSynchronize` after the calls to forward/backward. Simpler than events, and conforms to other benchmarks.

[Feature Request] Benchmarking - RNN/LSTM

@nouiz @kyunghyuncho good job on that LSTM tutorial guys. Looks great.

[Feature Request] Benchmarking - RNN/LSTM

Can it sample double digits from random latent variables instead of an input? Nicholas Léonard 450-210-1214 On Fri, May 1, 2015 at 6:35 PM, Ajay Talati [email protected] wrote: > The...

Benchmark peak GPU memory used

Okay so you want one timing for the entirety of the `forward/backward`.

Benchmark peak GPU memory used

Oh, an just so we are clear, does this includes an inplace parameter update? (in torch, `backward` or `backwardUpdate` vs just `updateGradInput`).

Benchmark peak GPU memory used

French: Ok donc Fred ce sera benchmark qui inclu le forward, ainsi qu'un update des parametres utilisant les gradient. Et le update peut être fait inplace.

Benchmark peak GPU memory used

Okay so the theano backward is complete: https://github.com/soumith/convnet-benchmarks/pull/11.Thanks @f0k.

Zero-Masking with RNN/LSTM

@soumith It would be a nice feature request for the next NVIDIA cudnn release. The lack of zero-masking is the only reason I am still not using cudnn LSTMs.