test problem about fairseq-lua HOT 5 CLOSED

@minmummax I think you are missing the nccl library. Do the following steps to install it:

git clone https://github.com/NVIDIA/nccl
cd nccl
make install
sudo ldconfig
luarocks install nccl

Hope it helps.

from fairseq-lua.

Occasionally you have to manually create symlinks for libnccl.so.1 -> libnccl.so

from fairseq-lua.

Thank you @shahbazsyed .
I have done your commands. But when I run train.lua in the zbstudio, I still get error:

Tried loading libnccl.1.dylib but got error /home/gt/torch/install/share/lua/5.1/nccl/ffi.lua:192: libnccl.1.dylib: cannot open shared object file: No such file or directory
Tried loading libnccl.so.1 but got error /home/gt/torch/install/share/lua/5.1/nccl/ffi.lua:192: libnccl.so.1: cannot open shared object file: No such file or directory

from fairseq-lua.

find libnccl.so* and add the directory to LD_LIBRARY_PATH

from fairseq-lua.

I am still facing this issue even after adding the directory to the LD_LIBRARY_PATH. Any help is appreciated :) Thank you in advance.

from fairseq-lua.