Package actualization mxnetR for windows


#1

Hi,

I’m statistician and I work a lot with mxnetR. I do very interesting research with this library in healthcare aplications (MRI, ecg precessing, predictive modeling,…), maybe I can do a tutorial with our data! :wink:

In the other hand, I will teach mxnetR in my university class in Barcelona in the degree of statistics. My question is ft is possible to provide an actualization for mxnetR in windows to version 1.0? the last actaulization of diagrammer library crashes the library, it requeries actualization…

Thanks!


#2

Hi,

Do you need GPU support?
I actually just built a brand new precompiled binary for Windows of mxnet 1.2.0.
Only caveat is that I had to hack a little and remove the im2rec functionnality. If needed, you should resort to the C++ program. Here’s the link for download:

For GPU, as I don’t have access to GPU on Windows, an option would be to use a docker image. I try to maintain images for both CPU and GPU here:
https://hub.docker.com/r/jeremiedb/dockerml/

Hope that helps!


#3

Thanks, helps a lot! Precompiled binary in cpu is perfect!
In my job I use the GPU version with nvidia gtx 1070, but I will try with docker. :slight_smile:


#4

Hello, I just built a mxnet 1.2.0 (as of April 25) GPU version, based on CUDA 9.1 and cudnn7 for Windows. Here’s the download link:

install.packages("https://s3.ca-central-1.amazonaws.com/jeremiedb/share/mxnet/GPU/mxnet.zip", repos = NULL)


#5

Thanks a lot @jeremiedb!! :slight_smile: