Christopher Paciorek
Christopher Paciorek
I used the binary. Do you know what would cause the error when I do use the matching cuDNN (7.1.1)?
Yes, I just ran the mnistCUDNN test and it ran fine.
I get the same error, which seems to be the same fundamental issue underlying #2390. That said, JiYuan, your 'conda list' shows you are installing caffe2-cuda8.0-cudnn7, but you say you...
I haven't addressed the autoscaling aspect of things (though I'd hoping to experiment with it), but I have [some instructions](https://github.com/paciorek/future-kubernetes) for using future within an RStudio server session on a...
@MarkEdmondson1234 what kind of builds do you have in mind? I'm guessing that doing something like this via future would involve a `system()` call within each R worker to run...
@danielturek I see you're cranking on this, which I think entails having predictive node sampling not affect sampling for the parameters. You probably have thought about these, but note that...
Stoked that you are working on this - it's been a big drawback in efficiency for us (computational and mcmc) for a long time now. I think that makes sense...
Here's my take -- in a standard MCMC when one generates PP sample, it is conditional on all the other parameters. Then if one samples from a parameter later in...
Makes sense to me. I had been thinking of doing something at the time of `addSampler`, but your idea makes more sense.
I haven't been able to focus on the details of the whole discussion here, but in terms of the A->B->C, A->C example, that does seem like something that could cause...