GCN-D MMdistributedDataparallel train
joelulu opened this issue · comments
Joe commented
你好,我想用MMdistributedDataparallel训练GCN-D
更改了--distributed 为True
报错:
Default process group has not been initialized, please make sure to call init_process_group
ThomaswellY commented
the same question, the codes seems not supportable for distributed gpus
Lei Yang commented
Yes. It currently does not support distributed training. We welcome any related PR!