DeepRec-AI / DeepRec

DeepRec is a high-performance recommendation deep learning framework based on TensorFlow. It is hosted in incubation in LF AI & Data Foundation.
Apache License 2.0
992 stars 341 forks source link

SeaStar Server configuration introduction #147

Open Gforky opened 2 years ago

Gforky commented 2 years ago

URL(s) with the issue:

https://deeprec.readthedocs.io/zh/latest/StarServer.html#estimator

Description of issue (what needs changing):

Clear description

Try to run PS distributed training with SeaStar servers according to the documentation above, but encounter an error: Load endpoint map from .endpoint_map failed. Confused about how to generate the endpoint_map, hope for more detailed illustrations in the documentation.

shanshanpt commented 2 years ago

We will provide much more details documents later. And the format of .endpoint_map is as follows: 127.0.0.1:10000=127.0.0.1:10001 127.0.0.1:22222=127.0.0.1:22223 127.0.0.1:33333=127.0.0.1:33334 ...

you should reserve two port for woker and ps, for example 127.0.0.1:10000 and 127.0.0.1:10001, one for grpc and the other one for seastar. So you can generate the file according the format above.