Chendi.Xue

Update: After debugging, noticed that nan issue was caused by records with numHistory as 1, after filtering out these lines in local_aggregate.py, train codes now worked with prepare_data.sh FYI

Hi @mufeili , I saw you are the owner for this example, please help to review this PR see if it makes sense to you. From my evaluation, running this...

> I dropped few comments. Otherwise, looks good to me. @chang-l , got it, removed num_workers and dgl version print

> @xuechendi Can you please add comment within `is_support_cpu_affinity` function? Especially, explain why 0.9.1 is specified there. (e.g., `return v_t >= "0.9.1" # (cpu)affinity-aware dataloading only available with dgl 0.9.1+`)...

Cetune on CentOS may met some python dependency issue, we don’t have a python dependency list for CentOS But if you can solve the dependency, it just works fine from...

@Shaoshian , this PR should goes to stable version branch

Yes, and I believe which is already fixed by using ifconfig in master, could you check on it? Thanks 在 2017年1月26日，05:19，Orlando Moreno 写道： CeTune uses getIpByHostInSubnet() method in conf/common.py to...

This is quite an old bug fixed months ago, please try to use current master and see if the error still there? Sorry for the troubles Best regards, Chendi From:...

@Shaoshian , I think I gave you some feed back on this PR,did you fixed it?

@Shaoshian , 1. add qumerbd support 2. test on randreadwrite io pattern, differentiate percentile latency with read/write 3. accumulate iops, BW and avg iops for all numjobs, don't only use...