leezu commented on pull request #18542:
URL: https://github.com/apache/incubator-mxnet/pull/18542#issuecomment-646438252
@yzhliu NVRTC is enabled by default and thus built by the CI unless
disabled:
https://github.com/apache/incubator-mxnet/blob/497bf7efb403a9174817f07ab3d2f9be033845ad/CMa
leezu commented on pull request #18542:
URL: https://github.com/apache/incubator-mxnet/pull/18542#issuecomment-646286471
@jinboci for the `libmxnet.so`, it currently has a `libcuda` dependency when
compiled with nvrtc. This will be fixed eventually
(https://github.com/apache/incubator-mxne
leezu commented on pull request #18542:
URL: https://github.com/apache/incubator-mxnet/pull/18542#issuecomment-646225132
@jinboci would it be possible to `dlopen` `libcuda` at runtime?
This is an automated message from the A
leezu commented on pull request #18542:
URL: https://github.com/apache/incubator-mxnet/pull/18542#issuecomment-645790195
@yzhliu It should be the other way round. Let's open the CI Docker
container: `docker run -it mxnetci/build.ubuntu_gpu_cu102 /bin/bash` and look
at the shared libraries
leezu commented on pull request #18542:
URL: https://github.com/apache/incubator-mxnet/pull/18542#issuecomment-643459935
You need to investigate why libcuda is not found in the container.
Previously there was a hack of putting /usr/local/cuda/compat on the path, but
that may not be the cor