Topologies
[training topology]
@beam/fedavg-liveFedAvg (live session)
A server holds run.lead and the model; a fixed cohort of clients trains each round and the server averages every client's update (quorum = all clients).
Static membershipBuiltinActivedistributed run
[topology identity]
Details
What version 1.0.0 declares.
- Package
@beam/fedavg-live- Membership
- Static
- Engines
- distributed run
- Tags
- federated-averaging, live-session, star, static
- Step kinds
- local, gather, broadcast
- Latest tag
- latest -> 1.0.0
- Publisher
- registry
- Visibility
- Public
- Checksum
sha256:6258a2d4c8119028d25e7513bbc75357c1c9f1abf1eaa94f8320ac8d2f9cddf9
[roles]
Roles
Who takes part, and how many members each role holds.
The fixed cohort.
Leads the run and holds the model.
[slots]
Slots
The actions a run binds: what each slot consumes and produces.
update* -> model
Averages the clients' updates into the next model version.
model -> update
Trains from the latest model version and sends its update.
[round]
Step kinds
The steps of one round, in order.
- localrole client · slot train
- gatherfrom client · to server · object update · slot merge
- broadcastfrom server · to client · object model