v0.5.0

MLJFlow v0.5.0

Diff since v0.4.3

• [breaking] Renaming _logging_channel to logging_channel, and changing its type from Channel{Tuple} to Union{Channel{Tuple}, Nothing}
• [breaking] open_logging_channel function now requires a Logger instance and doesn't return a Channel anymore. The reason is to keep logging_channel in a "private" form in our model
• [improvement] Removed non-useful wait!(result_channel) in src/base.jl
• [improvement] Explicitly closing the logging_channel when a result is retrieved in the background worker

v0.4.3

MLJFlow v0.4.3

Diff since v0.4.2

• [new feature] Added complete support with MLJ's TunedModel, being fully-compatible with accelerators (CPUThreads() and CPUProcesses()). This feature is not supported by mlflow, but implemented here following a Producer-Consumer pattern using a Channel approach.
• [fix] Logger is now a mutable struct, which means that its fields are modifiable outside instantiation.

Merged pull requests:

• Implementing channel control for experiment creation (#41) (@pebeto)

Closed issues:

• Add tests for CPUThreads and CPUProcesses (#25)
• Update the README.md to reflect change uri -> apiroot (#40)

v0.4.2

MLJFlow v0.4.2

Diff since v0.4.1

v0.4.1

MLJFlow v0.4.1

Diff since v0.4.0

Merged pull requests:

• Save hot fix (#38) (@ablaom)
• For a 0.4.1 release (#39) (@ablaom)

Closed issues:

• Method ambiguity (due to type piracy) in extension of MLJBase.save (#37)

v0.4.0

MLJFlow v0.4.0

Diff since v0.3.0

(breaking) Removing healthcheck function on Logger() instantiation due to platforms inconsistency (see #34).

Merged pull requests:

• Extending Base.show to pretty print logger instances (#32) (@pebeto)
• Adding test case for zero hyperparameter model (#33) (@pebeto)
• Removing healthcheck function due to platform inconsistency (#35) (@pebeto)

Closed issues:

• Need handling for models with zero hyperparameters (#22)
• Improve show for logger instances (#24)
• mlflow health endpoint is doutbful (#34)

v0.3.0

MLJFlow v0.3.0

(breaking) Rename MLFlowLogger (exported) to Logger (unexported). So a logger instance is now instantiated with using MLJFlow; logger = MLJFlow.Logger(baseuri).

(testing) Facilitate local testing by adding MLFLOW_URI environment variable to set the location of the running MLflow service (#30)

Merged pull requests:

• Polish readme and extend readme example (#10) (@ablaom)
• Address breaking changes in MLJBase 1.0 (#21) (@ablaom)
• Fix python version to stabilise launch of mlflow service in CI (#28) (@ablaom)
• For a 0.2.0 release (#29) (@ablaom)
• Changing test URI to the defined ENV variable (#30) (@pebeto)
• For a 0.3.0 release (#31) (@ablaom)

Closed issues:

• Export names sparingly (#2)
• Clean up doc string (#3)
• Move utilities to MLJBase? (#4)
• Minor doc point. (#5)
• [Tracking] Towards registration (#6)
• Include type of resampling in the run tag (#7)
• Add accessor method to extract the "client" from an MLFlowLogger object (#8)
• Add a verbosity field to the MLFlowLogger wrapper (#9)
• In tests: include reconstruction of machine saved as artefact (#11)
• Recheck readme example after MLJ.jl update (#12)
• Improve test of save. (#13)
• Move service code to service.jl (#14)
• Remove an abstract type in a struct (#15)
• Duplicate run names intentional? (#16)
• Fixing Julia LTS pipeline (#17)
• TagBot trigger issue (#18)
• Improve local testability (#20)
• Rename MLFlowLogger ? (#23)
• CI has started to fail (#27)