v0.3.1beta

What's Changed

• Auto-download LLM by @atharvas in #43
• Improving LLM string parsing. by @atharvas in #64
• Update SR.jl dependencies by @atharvas in #67
• Not production ready because SR.jl is installed via custom source (rather than JuliaRegistry).

Full Changelog: v0.3.0...v0.3.1beta

v0.3.0

LibraryAugmentedSymbolicRegression v0.3.0

Diff since v0.2.0

This release corrects a major bug with how multiple dispatch was being used to integrate with SR.jl. This release also revamps the logging system so all the LLM logs are now stored using TensorBoardLogger instead of as a text file (which would cause IO slowdowns).

Breaking changes

• The API no longer has a llm_recorder parameter. Instead pass a logger=SRLogger(TBLogger()) object to equation_search or LaSRRegressor.

Merged pull requests:

• CompatHelper: bump compat for Optim to 1, (keep existing compat) (#46) (@github-actions[bot])
• update tagbot (#47) (@atharvas)
• LLM Corrections (#49) (@atharvas)

Closed issues:

• v0.2.0 (#44)

v0.2.0

LibraryAugmentedSymbolicRegression v0.2.0

Diff since v0.1.0

This release introduces a major rewrite of LaSR. The previous release modified a fork of SymbolicRegression.jl, duplicating most of the SR.jl codebase in the process. Since version 1.0.0 SR.jl supports a variety of new abstractions for user extensibility. We leverage this to substantially reduce the amount of code we need to maintain. The rest of the breaking changes are outlined in the changelog below:

Breaking changes

• New API: LLMOptions.jl now extend SR.jl's Options.jl module. This required reorganizing the MutationWeights API. Consult test/test_tutorial.jl and the README for more information. This is a breaking change.
• Removed: Many files used within SR.jl that LaSR.jl did not touch.
• New Feature: All LLM requests are now logged with a custom recorder.

Merged pull requests:

• CompatHelper: bump compat for Enzyme in [weakdeps] to 0.13, (keep existing compat) (#12) (@github-actions[bot])
• CompatHelper: bump compat for DifferentiationInterface to 0.6, (keep existing compat) (#16) (@github-actions[bot])
• CompatHelper: bump compat for DynamicExpressions to 1, (keep existing compat) (#17) (@github-actions[bot])
• Ensure simple program works (#24) (@atharvas)
• CompatHelper: bump compat for Optim to 1, (keep existing compat) (#25) (@github-actions[bot])
• CompatHelper: bump compat for LossFunctions to 1, (keep existing compat) (#29) (@github-actions[bot])
• rewrite w/ multiple dispatch (#36) (@atharvas)
• CompatHelper: bump compat for PromptingTools to 0.72, (keep existing compat) (#41) (@github-actions[bot])
• Updating the LaSR Readme to comply with new API. (#45) (@atharvas)

Closed issues:

• Example code does not work (#21)
• can ' t run normally (#30)
• Request for a python Interface. (#37)
• Running Time on Feynman Dataset (#38)
• Missing Code for Metrics in "Metric" Section (Page 5) (#42)

v0.1.0

LibraryAugmentedSymbolicRegression v0.1.0

Diff since v0.0.1

Interface updates and coverage tests

This version includes bug fixes and enhanced documentation. Most of the functionality in LibraryAugmentedSymbolicRegression.jl and SymbolicRegression.jl now works seamlessly out of the box. The upcoming v0.2.0 release will fully integrate with SymbolicRegression.jl via the LaSR extension.

Landing page: https://trishullab.github.io/lasr-web

Merged pull requests:

• update documentation (#8) (@atharvas)
• CompatHelper: bump compat for PromptingTools to 0.56, (keep existing compat) (#11) (@github-actions[bot])
• SR.jl test case compliance (#13) (@atharvas)

Closed issues:

• v0.0.1 (#6)

v0.0.1

LibraryAugmentedSymbolicRegression v0.0.1

Initial code release for LibraryAugmentedSymbolicRegression

This version contains the base code to allow SymbolicRegression.jl to induce programs with LLMs in a way that accelerates search. I want to have up-to-date documentation before releasing v0.1.0.

More information about this method is available at: https://trishullab.github.io/lasr-web/

Merged pull requests:

• Lasr parametric expressions (#1) (@atharvas)
• Parametric expressions (#2) (@atharvas)
• Rename package from LaSR.jl to LibraryAugmentedSymbolicRegression.jl (#5) (@atharvas)

Closed issues:

• Deploying v0.0.1 (#3)