arrowrowe commented on issue #10455: Bug of group2ctxs for model parallelism URL: https://github.com/apache/incubator-mxnet/issues/10455#issuecomment-380319322 @eric-haibin-lin yes, same for all 8 gpus cases, same for all <8 gpu cases. ``` --group2ctxs 0 --num_gpus 1 loss: 22.12132072448730468750000000000000 --group2ctxs 1 --num_gpus 1 loss: 22.12132072448730468750000000000000 --group2ctxs 1 --num_gpus 2 loss: 22.12132072448730468750000000000000 --group2ctxs 1 --num_gpus 4 loss: 22.12132072448730468750000000000000 --group2ctxs 1 --num_gpus 7 loss: 22.12132072448730468750000000000000 --group2ctxs 1 --num_gpus 8 loss: 22.12132263183593750000000000000000 --group2ctxs 1 --num_gpus 8 loss: 22.12132263183593750000000000000000 --group2ctxs 1 --num_gpus 8 loss: 22.12132263183593750000000000000000 --group2ctxs 0 --num_gpus 1 input_grads: 0.00030235992744565010070800781250 --group2ctxs 1 --num_gpus 1 input_grads: 0.00030235966551117599010467529297 --group2ctxs 1 --num_gpus 2 input_grads: 0.00030235908343456685543060302734 --group2ctxs 1 --num_gpus 4 input_grads: 0.00030235905433073639869689941406 --group2ctxs 1 --num_gpus 7 input_grads: 0.00030235887970775365829467773438 --group2ctxs 1 --num_gpus 8 input_grads: 0.00453650858253240585327148437500 --group2ctxs 1 --num_gpus 8 input_grads: 0.00453650858253240585327148437500 --group2ctxs 1 --num_gpus 8 input_grads: 0.00453650858253240585327148437500 ```
---------------------------------------------------------------- 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