xieydd
xieydd
Can you show your command? @GoldenSpark
Maybe you can use the run_local just ok in container, don`t need to change the shell script , just change the eth1 in your shell. @GoldenSpark
you can use `ip -a` to see now what you are using. @GoldenSpark
time out ``` [1,7]:[tensorflow-mnist-worker-0][[37410,1],7][btl_tcp_frag.c:244:mca_btl_tcp_frag_recv] mca_btl_tcp_frag_recv: readv failed: Connection timed out (110) ```
I think change to `~/charts` ok, but when update the arena is not friendly. @xiaozhouX
Which model do you use? @liming312
When i test resnet 18 use num_batch==1, top1 is `71.37%`, when i set num_batch==8, top1 is `71.49%`.
Not support write to cache now? cc @cheyang @TrafalgarZZZ
Does it supported yet?
Any process? I notice Pin have been added in Data Type, but not use in #49