Skip to content

Releases: cisco-open/flame

v0.4.0

15 Dec 22:10
1100751
Compare
Choose a tag to compare

What's Changed

New Contributors

Full Changelog: v0.3.0...v0.4.0

v0.3.0

20 Sep 02:30
1718f4a
Compare
Choose a tag to compare

What's Changed

Full Changelog: v0.2.4...v0.3.0

v0.2.4

30 Aug 17:07
468f6ac
Compare
Choose a tag to compare

What's Changed

Full Changelog: v0.2.3...v0.2.4

v0.2.3

06 Jun 02:16
f1e70cb
Compare
Choose a tag to compare

Key Changes

  • New approaches: feddyn, oort, FedBalancer, synchronous hierarchical FL with coordinator, coordinated asyncfl, and differential privacy
  • Metrics collection support: collect cpu/gpu utilization, ram and vram usage, bytes sent and received and the time to execute functions
  • Local registry support: model and metrics are saved in a local directory, which allows experiments without mlflow
  • misc improvements:
    • support for channel leave functionality in the p2p backend
    • support for simplifying tasklet composition primitives
    • protobuf package version upgrade to be compatible with tensorflow v2.12.0
    • deployer: fix bug to update pod status incorrectly
    • fix a bug that removes model updates from a disk cache when the total cache data is larger than its size
    • fix openapi specification mismatch between a released image and the source

What's Changed

New Contributors

Full Changelog: v0.2.2...v0.2.3

v0.2.2

20 Mar 14:45
f33ca78
Compare
Choose a tag to compare

Key Changes

Control Plane & SDK: Use the previous defined groupAssociation role information to handle all the tasks creation. This change was useful in order to make to make it easier to identify the connections between roles and to correctly define the used topology
Control Plane: split the job.json file from examples to dataSpec.json and modelSpec.json to make it easier to understand which fields are used and for what purpose
Helpers: Created a minikube script to start the cluster for macOS
Fiab: enforce the local cluster to use the current released flame version

What's Changed

Full Changelog: v0.2.1...v0.2.2

v0.2.1

06 Mar 17:01
8332efd
Compare
Choose a tag to compare

Key Changes

  • Control Plane: add initial support for group association for roles in order to identify the connections thru different channels
  • Flame SDK: Configuration handling has been refactored
  • Utils: Created a diagnose script to allow an easier debugging
  • Flame SDK: Implemented pre-commit configuration for the project
  • Control Plane: refactoring to align the code with generated one from OpenAPI

What's Changed

Full Changelog: v0.2.0...v0.2.1

v0.2.0

28 Feb 17:29
466e6cc
Compare
Choose a tag to compare

Key Changes

  • Asynchronous FL (SDK, control plane): Asynchronous FL algorithm/mechanism is implemented. Examples for Async FL are added in the library (lib/python/flame/examples/{async_hier_mnist, async_mnist}) and in the control plane example folder (flame/examples/asyncfl_hier_mnist).
  • FedProx (SDK): FedProx algorithm is implemented in the python library.
  • Add tensorflow support for FedYogi, FedAdam, and FedAdaGrad
  • Building of flame base image changed from CPU version to cude 11.3 version
  • Fiab diagnosis script (control plane): A basic script to obtain the log information on fiab environment’s status is implemented.
  • Bugfix on local DNS setup in linux (fiab): Under different distributions, local DNS setup fails. A bugfix for this is added. Test is done under ubuntu and archlinux.
  • Configurable Deployer (control plane): deployer can take a different job template file via configuration. All the command-line arguments are replaced with a configuration file.
  • Documentation (doc): update documentation to setup fiab cluster locally and flame SDK guide.

What's Changed

New Contributors

Full Changelog: v0.1.7...v0.2.0

v0.1.7

05 Jan 19:30
56ccf64
Compare
Choose a tag to compare

Key Changes

  • Installation steps for ubuntu, amazon linux 2 and mac OS are automated via a script.
  • Flame-in-a-Box (FIAB) env (A small scale dev env) is improved to use a latest official flame image (from docker hub) without need for building a flame image locally. This feature is mainly for users who wish to try out flame quickly.
  • mlflow's version upgrade caused incompatibility issue on keras module, which made metrics report to mlflow throw an error. This issue was fixed.
  • job template file is updated to prefer node with GPU resource. If gpu resource is not available in a cluster, a job is scheduled on non-gpu nodes.

What's Changed

Full Changelog: v0.1.6...v0.1.7

v0.1.6

16 Nov 20:17
fdcfedd
Compare
Choose a tag to compare

What's Changed

New Contributors

Full Changelog: v0.1.5...v0.1.6

v0.1.5

01 Sep 21:59
c6cc492
Compare
Choose a tag to compare

What's Changed

Full Changelog: v0.1.4...v0.1.5