Scott LeGrand

Yes, use linear as the activation function and L2 as the error function.

So first problem is that I do not believe anyone is left at Amazon to accept pull requests. But since you seem to work at NVIDIA and I do too,...

What GPU are you using? On Jun 13, 2017 6:59 PM, "qubingxin" wrote: > We use some data, include user's id and corresponding click article, to > train the recommended...

Cut your batch size in half and see if that fixes this. I am somewhat shooting in the dark here. On Tue, Jun 13, 2017 at 8:30 PM, qubingxin wrote:...

Weird, could you rebuild uncommenting out //#define MEMTRACKING in GpuTypes.h and send the output? The other option is to run across multiple GPUs with MPI if you have them.

It is indeed related to your Ubuntu version, and there' are 3 pull requests from heng1220 that I think will fix this for you, but no one at Amazon is...

If you look at the CIFAR-10 sample, it shows how to build your own training and inference pipeline. If that's too complicated, Amazon to the best of my knowledge has...

Yeah, netcdf is a bad actor, arbitrarily changing API functionality and syntax with no macros to detect the version. I don't have a fix for this, and it remains similarly...

4.3 (which ships with 18.04) defines NcException as above, but 4.2 (which ships with 16.04) defines it as: NcException(const std::string& exceptionName,const std::string& complaint,const char* fileName,int lineNumber); This is a trivial...

TLDR: Correct that it doesn't call the fast path. Longer answer, if the maximum data counts were within the specified limit, one could roughly double performance here. I'll try to...