v0.3.0

FluxTraining v0.3.0

Diff since v0.2.4

Closed issues:

• Switch to ParameterSchedulers.jl (#106)

Merged pull requests:

• Add more convenient Learner method (#105) (@lorenzoh)
• Switch to ParameterSchedulers.jl (#107) (@rejuvyesh)
• Prepare v0.3.0 (#108) (@lorenzoh)

v0.2.4

FluxTraining v0.2.4

Diff since v0.2.3

Closed issues:

• Allow restricting phases during which a Metric runs (#84)

Merged pull requests:

• CompatHelper: bump compat for EarlyStopping to 0.2, (keep existing compat) (#92) (@github-actions[bot])
• CompatHelper: bump compat for EarlyStopping to 0.3, (keep existing compat) (#94) (@github-actions[bot])
• doc: fix link to TensorBoardLogger.jl (#95) (@visr)
• Move documentation system to Pollen.jl (#96) (@lorenzoh)
• Use ReTest.jl to run tests (#97) (@lorenzoh)
• Add and improve a lot of docstrings and a few test cases (#99) (@lorenzoh)
• Add phase argument to Metric (#100) (@lorenzoh)
• CompatHelper: add new compat entry for InlineTest at version 0.2, (keep existing compat) (#101) (@github-actions[bot])
• (Documentation) Document Learner components (#102) (@lorenzoh)
• Add Flux 0.13 compatibility (#103) (@lorenzoh)

v0.2.3

FluxTraining v0.2.3

Diff since v0.2.2

Closed issues:

• Error displaying EarlyStopper (#90)

Merged pull requests:

• fix: typo in show method for EarlyStopping (#91) (@awadell1)

v0.2.2

FluxTraining v0.2.2

Diff since v0.2.1

Merged pull requests:

• Fix PrettyTables header argument; restrict compat to >=1. (#88) (@lorenzoh)

v0.2.1

FluxTraining v0.2.1

Diff since v0.2.0

Merged pull requests:

• CompatHelper: bump compat for "ImageCore" to "0.9" (#74) (@github-actions[bot])
• CompatHelper: bump compat for "PrettyTables" to "1" (#79) (@github-actions[bot])

v0.2.0

FluxTraining v0.2.0

Diff since v0.1.3

Added

• New training loop API that is easier to extend. Defining a Phase and step! is all you need. See the new tutorial and the new reference.
  • Relevant functions: epoch!, step!, runstep, runepoch
• Added CHANGELOG.md (this file)
• AbstractValidationPhase as supertype for validation phases
• Documentation for callback helpers on reference page

Changed

• Batch* renamed to Step*:
  • events: BatchBegin now StepBegin, BatchEnd now StepEnd
  • CancelBatchException now CancelStepException.
  • field Learner.batch now Learner.step
• Learner.step/batch is no longer a special struct but now a PropDict, allowing you to set arbitrary fields.
• Learner.model can now be a NamedTuple/Tuple of models for use in custom training loops. Likewise, learner.params now resembles the structure of learner.model, allowing separate access to parameters of different models.
• Callbacks
  • Added init! method for callback initilization, replacing the Init event which required a Phase to implement.
  • Scheduler now has internal step counter and no longer relies on Recorder's history. This makes it easier to replace the scheduler without needing to offset the new schedules.
  • EarlyStopping callback now uses criteria from EarlyStopping.jl

Removed

• Removed old training API. Methods fitbatch!, fitbatchphase!, fitepoch!, fitepochphase! have all been removed.

Closed issues:

• Scheduler applies schedules per batch by default (#68)
• Recorder does not work with models with non-Array inputs. (#80)

Merged pull requests:

• CompatHelper: bump compat for "BSON" to "0.3" (#69) (@github-actions[bot])
• use EarlyStopping.jl for stopping criteria (#72) (@lorenzoh)
• CompatHelper: bump compat for "PrettyTables" to "0.12" (#73) (@github-actions[bot])
• Move documentation to Pollen.jl (#77) (@lorenzoh)
• Revert onecycle (#78) (@lorenzoh)
• Remove samplesfield from History (#81) (@lorenzoh)
• New training API and QoL improvements (v0.2.0) (#83) (@lorenzoh)

v0.1.3

FluxTraining v0.1.3

Diff since v0.1.2

Closed issues:

• Metrics wraps Metric(s) in Metric(s) (#66)
• Unnecessary softmax in accuracy? (#67)

Merged pull requests:

• Internal callback changes (#65) (@lorenzoh)
• fix #66 make Metric subtype AbstractMetric (#70) (@lorenzoh)
• Lorenzoh/fix/67 (#71) (@lorenzoh)

v0.1.2

FluxTraining v0.1.2

Diff since v0.1.1

Merged pull requests:

• small doc fix (#62) (@CarloLucibello)
• CompatHelper: bump compat for "PrettyTables" to "0.11" (#63) (@github-actions[bot])
• Callback utilities (#64) (@lorenzoh)

v0.1.1

FluxTraining v0.1.1

Diff since v0.1.0

Closed issues:

• Link to Loss broken (#40)
• Early stopping link broken (#41)
• BatchEnd link broken (#42)
• Break out Schedule (#44)
• How to verify GPU is working? (#47)
• What is the role of the test data? (#48)
• What do you think of Data Modules? (#53)

Merged pull requests:

• Add testing as CI step (#36) (@lorenzoh)
• Better docstrings (#37) (@lorenzoh)
• Typo (#43) (@drozzy)
• WRONG order of arguments to the Learner. (#45) (@drozzy)
• Fix issue #42 - missing docstrings (#49) (@lorenzoh)
• add EarlyStopping docstring (#50) (@lorenzoh)
• Update callback reference section on metrics (#51) (@lorenzoh)
• Remove no longer needed dependencies from Project.toml (#52) (@lorenzoh)
• Add SanityCheck callback (#56) (@lorenzoh)
• Fix GarbageCollect callback (#57) (@lorenzoh)
• Fix TensorBoard image serialization (#58) (@lorenzoh)
• Sanitycheck (#60) (@lorenzoh)
• update compat bounds (#61) (@lorenzoh)

v0.1.0

FluxTraining v0.1.0

Closed issues:

• Collaborating on a FastAI port? (#13)

Merged pull requests:

• Remove unused package LearnBase (#26) (@nrhodes)
• Fix protected tests (#28) (@nrhodes)
• Log model histograms in TensorBoard (#30) (@ToucheSir)