wynot12 commented on a change in pull request #203: [NEMO-360] Implementing an
'XGBoostPolicy'
URL: https://github.com/apache/incubator-nemo/pull/203#discussion_r267228579
##########
File path: runtime/common/src/main/proto/ControlMessage.proto
##########
@@ -23,37 +23,43 @@ package protobuf;
option java_package = "org.apache.nemo.runtime.common.comm";
option java_outer_classname = "ControlMessage";
+// Messages from client to driver
enum ClientToDriverMessageType {
LaunchDAG = 0;
DriverShutdown = 1;
+ Notification = 2;
}
message ClientToDriverMessage {
required ClientToDriverMessageType type = 1;
optional LaunchDAGMessage launchDAG = 2;
+ optional string message = 3;
}
message LaunchDAGMessage {
required string dag = 1;
optional bytes broadcastVars = 2;
}
-message DataCollectMessage {
- required string data = 1;
-}
-
+// Messages from driver to client
enum DriverToClientMessageType {
DriverStarted = 0;
DriverReady = 1;
DataCollected = 2;
ExecutionDone = 3;
+ LaunchXGBoostScript = 4;
Review comment:
Actually XGBoost is just a one type of various optimization methods.
Isn't it better to abstract with more high-level term (e.g.,
trainFromMetrics?) and give an option to select which method (e.g., XGBoost,
RL) to use?
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
