[Task Runner API] [Flower Interoperability] Integrate changes for flower-app-pytorch workspace's custom Collaborator into core component #1598
Merged
kminhta merged 11 commits intosecurefederatedai:developfrom May 19, 2025
Conversation
Signed-off-by: kta-intel <kevin.ta@intel.com>
kminhta
changed the title
kminhta
requested review from
MasterSkepticista,
psfoley and
teoparvanov
as code owners
Signed-off-by: kta-intel <kevin.ta@intel.com>
teoparvanov
reviewed
May 14, 2025
Collaborator
teoparvanov
left a comment
teoparvanov
left a comment
There was a problem hiding this comment.
A couple of thoughts from my first read-through:
… local_grpc_server to interop_server Signed-off-by: kta-intel <kevin.ta@intel.com>
Signed-off-by: kta-intel <kevin.ta@intel.com>
Signed-off-by: kta-intel <kevin.ta@intel.com>
Signed-off-by: kta-intel <kevin.ta@intel.com>
teoparvanov
approved these changes
May 15, 2025
Signed-off-by: kta-intel <kevin.ta@intel.com>
Signed-off-by: kta-intel <kevin.ta@intel.com>
kminhta
commented
May 15, 2025
| function : start_client_adapter | ||
| kwargs : | ||
| local_server_port : 0 # local grpc server, 0 to dynamically allocate | ||
| interop_server_port : 0 # interop server port, 0 to dynamically allocate |
Collaborator
Author
There was a problem hiding this comment.
Working on addressing making dynamic allocation of gRPC ports consistent across OpenFL in #1599
The issue with this particular case is that for a local simulation, we need the collaborators to have different ports but hashing the plan would result in the same port which will lead to binding issues. I'm considering adding a local_simulation flag where the Collaborator can dynamically allocate based on its common name during the _prepare_interop_server - or the Task Runner can also handle this which may be a cleaner delineation from the Collaborator
It's the actually same reason for this messy handling of the client_port too: https://github.com/securefederatedai/openfl/blob/develop/openfl-workspace/flower-app-pytorch/src/runner.py#L51
tayfunceylan
pushed a commit
to tayfunceylan/openfl
that referenced
this pull request
May 23, 2025
…wer-app-pytorch workspace's custom Collaborator into core component (securefederatedai#1598) * integrate flower collaborator into core collaborator Signed-off-by: kta-intel <kevin.ta@intel.com> * formatting Signed-off-by: kta-intel <kevin.ta@intel.com> * change task name interop -> prepare_for_interop, change references to local_grpc_server to interop_server Signed-off-by: kta-intel <kevin.ta@intel.com> * add prepare_interop_server as callback to run on_experiment_begin Signed-off-by: kta-intel <kevin.ta@intel.com> * fix typo Signed-off-by: kta-intel <kevin.ta@intel.com> * change to on_round_begin() Signed-off-by: kta-intel <kevin.ta@intel.com> * return empty dict Signed-off-by: kta-intel <kevin.ta@intel.com> * set self.prepare_interop_server() to run on experiment begin Signed-off-by: kta-intel <kevin.ta@intel.com> * remove self-explanatory comment, added another comment for clarity Signed-off-by: kta-intel <kevin.ta@intel.com> --------- Signed-off-by: kta-intel <kevin.ta@intel.com> Signed-off-by: Tayfun Ceylan <tayfun.ceylan>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Part 2/x of bringing flower-app-pytorch custom components back into core components
flower-app-pytorchworkspace's customAggregatorinto core component #1585openfl.component.Collaborator...
...
Type of Change (Mandatory)
Specify the type of change being made.
Description (Mandatory)
This PR:
src.collaborator.CollaboratorFlowerand pulls changes back intoopenfl.component.Collaboratorstart_client_adaptertointeropfor better generalizabilityprepare_interop_server()callback toCollaboratorin order to minimize changes needed indo_task(). Originally, there was a separate conditional that started a separate set of tasks specific to interopability, which made it harder to read. Instead, we now pass the module to import (src.grpc.connector.flower.interop_server) viaplan.tasksetting. This module is imported and initialized and sent to the Task Runner inkwargsat the beginning of the experimentflower.interop_serverarg is modified to accept a callable function (receive_message_from_interop(), which is defined inprepare_interop_server(). The reason for this change is that originally, we send the entire client to theinterop_server()and theinterop_server()calledself.client.send_message_to_server()directly. It better encapsulates the client and the collaborator for theinterop_server()to pass just the response to a callable function of theCollaboratorTesting
Manual testing in non-SGX and SGX