graph-learn
graph-learn copied to clipboard
When multiple PS servers are used, How to set parameter "emb_save_dir" when using multiple PS servers?
I have specified the same local directory on each PS server, but the result is not good. Should this parameter be set to a public HDFS directory?
Thanks!
Yes, a distributed file system is needed. You can launch a distributed program on a local machine, or mount NFS to local directory to establish a real distributed model.
Currently, HDFS is not supported.
@jackonan Thank you very much for your reply. Currently, HDFS is not supported, how can I use multiple PS? My embedding can't be saved with a PS.
Actually, save model to HDFS is supported by TensorFlow.