This PR aims to support TF application to send and receive embedding info. after training is completed. It includes below changes:
a. add send_no_deps() and recv_no_deps() in FLModel. Their typical usage is for TF application to inject final operation in graph to send and receive tensor when training is completed.
b. update example/wide_n_deep to demo how to send and receive embedding tensor when training is completed.
Above changes are verified in local environment with two docker instances as shown in example READEM.
This PR aims to support TF application to send and receive embedding info. after training is completed. It includes below changes: a. add send_no_deps() and recv_no_deps() in FLModel. Their typical usage is for TF application to inject final operation in graph to send and receive tensor when training is completed. b. update example/wide_n_deep to demo how to send and receive embedding tensor when training is completed. Above changes are verified in local environment with two docker instances as shown in example READEM.