[GitHub] asmushetzel commented on issue #7147: cuda support for linalg-functions, restructuring of linalg interfaces
asmushetzel commented on issue #7147: cuda support for linalg-functions, restructuring of linalg interfaces URL: https://github.com/apache/incubator-mxnet/pull/7147#issuecomment-321958146 So this is done from my point of view. Jenkins failed on some nodes in some tests but that is all due to an unrelated problem with some Gluon testcase (I examined all the logfiles). The relevant la_op tests pass on all nodes successfully. Eric, from my point of view this can be merged now. This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] asmushetzel commented on issue #7147: cuda support for linalg-functions, restructuring of linalg interfaces
asmushetzel commented on issue #7147: cuda support for linalg-functions, restructuring of linalg interfaces URL: https://github.com/apache/incubator-mxnet/pull/7147#issuecomment-321958146 Should all be o.k. now but Jenkins seems to be non-responsive. Have to wait until it is up again. This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] asmushetzel commented on issue #7147: cuda support for linalg-functions, restructuring of linalg interfaces
asmushetzel commented on issue #7147: cuda support for linalg-functions, restructuring of linalg interfaces URL: https://github.com/apache/incubator-mxnet/pull/7147#issuecomment-321932374 It's in the final rounds of testing. Goal is to have it integrated next week. It would be awesome if you can put Volta/Tensorcore in there. Another topic that I discussed with Marek Kolodziej is whether NVidia people can also investigate ways how to do more efficient batch-implementations of potrf/trmm. One important use case will be having a batch of relatively small matrices and there the current implementation won't be that great (and cuBlas/cuSolver miss native batch processing for these operations). This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
[GitHub] asmushetzel commented on issue #7147: cuda support for linalg-functions, restructuring of linalg interfaces
asmushetzel commented on issue #7147: cuda support for linalg-functions, restructuring of linalg interfaces URL: https://github.com/apache/incubator-mxnet/pull/7147#issuecomment-320201109 Makes a lot of sense. The linalg-stuff will grow and it deserves it's own directory then. Just for status: I had almost zero time in the last two weeks for this. Will try to finish next week (the CUDA-tests have problem, regardless what Jenkins shows here). This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services