v0.2.11

Reactant v0.2.11

Diff since v0.2.10

Merged pull requests:

• CompatHelper: add new compat entry for YaoBlocks in [weakdeps] at version 0.13, (keep existing compat) (#312) (@github-actions[bot])
• WIP: kernels (#314) (@wsmoses)
• test: unbreak CUDA CI (#337) (@avik-pal)
• feat: correct handling of wrapped arrays functionalities (#342) (@avik-pal)
• reduce codegen on YaoBlocks.mat methods (#343) (@mofeing)
• refactor: use Ops instead of direct stablehlo calls (#347) (@avik-pal)
• Run CI on aarch64 (#350) (@mofeing)
• add missing @test (#352) (@glou-nes)
• Add Ops.hlo_call(::String, args...) (#358) (@Pangoraw)
• Kernels: support Julia 1.11 (#359) (@jumerckx)
• Define outer repeat method for TracedRArray (#361) (@AntonOresten)
• docs: add Reactant.Ops docs to the website (#362) (@avik-pal)
• feat: tracing Random.jl functionality correctly (#363) (@avik-pal)
• Interp2 (#365) (@wsmoses)
• fix: ensure printing of wrapped ConcreteRArrays goes through our show (#367) (@avik-pal)
• feat: add support for the remaining wrapper types (#369) (@avik-pal)
• Format code of branch "main" (#371) (@github-actions[bot])
• stablehlo.sort Ops (#374) (@glou-nes)
• No Transpose emission for 0 rank tensor (#375) (@glou-nes)
• fix: include files if they end with .jl (#377) (@avik-pal)
• refactor: move overrides into a separate file (#379) (@avik-pal)
• CompatHelper: bump compat for Flux to 0.16 for package test, (keep existing compat) (#383) (@github-actions[bot])
• Fix traced_getfield (#385) (@wsmoses)
• fix backtrace calls on debug mode (#387) (@mofeing)
• Kernel-supporting jll (#389) (@wsmoses)
• fix: override to overlay (#390) (@avik-pal)
• CompatHelper: add new compat entry for CUDA in [weakdeps] at version 5 for package ., (keep existing compat) (#391) (@github-actions[bot])
• CompatHelper: add new compat entry for CUDA at version 5 for package test, (keep existing compat) (#392) (@github-actions[bot])
• fix: missing interpolation in compile (#394) (@avik-pal)
• CompatHelper: add new compat entry for Random123 at version 1 for package test, (keep existing compat) (#395) (@github-actions[bot])
• CompatHelper: add new compat entry for StatsBase at version 0.34 for package test, (keep existing compat) (#396) (@github-actions[bot])
• CompatHelper: add new compat entry for StableRNGs at version 1 for package test, (keep existing compat) (#397) (@github-actions[bot])
• CompatHelper: add new compat entry for Distributions at version 0.25 for package test, (keep existing compat) (#398) (@github-actions[bot])
• CompatHelper: add new compat entry for HypothesisTests at version 0.11 for package test, (keep existing compat) (#399) (@github-actions[bot])
• feat: bump jll version and add opt pass (#400) (@avik-pal)
• ci: disable non-functional CI (#401) (@avik-pal)
• feat: add a no_enzyme option (#402) (@avik-pal)
• bump reactant jll (#404) (@Pangoraw)
• Inference cache (#405) (@jumerckx)
• Pythoncall (#407) (@wsmoses)
• JLL: enable throwing better errors (#409) (@wsmoses)
• Regenerate MLIR Bindings (#410) (@github-actions[bot])
• Format code of branch "main" (#411) (@github-actions[bot])
• CompatHelper: add new compat entry for PythonCall in [weakdeps] at version 0.9 for package ., (keep existing compat) (#412) (@github-actions[bot])
• CompatHelper: add new compat entry for PythonCall at version 0.9 for package test, (keep existing compat) (#413) (@github-actions[bot])
• generate EnzymeXLA dialect bindings (#414) (@mofeing)
• refactor: move python testing to integration (#415) (@avik-pal)
• fix: handle traced array returns inside objects (#417) (@avik-pal)
• Improve reactant error messages (#418) (@wsmoses)
• Fix ReactantPythonCallExt.jl (#419) (@wsmoses)
• feat: add zero and fill! for ConcreteRArray (#420) (@avik-pal)
• Fix error on global (#422) (@wsmoses)
• Cuv2 (#423) (@wsmoses)
• disable absint of absint (#424) (@wsmoses)
• fix: make eltype of Traced/Concrete Arrays to be respective RNumbers (#426) (@avik-pal)
• CUDA kernels take 3 (#427) (@wsmoses)
• Format code of branch "main" (#432) (@github-actions[bot])

Closed issues:

• getindex assumes contiguous indexing (#242)
• Incorrect stableHLO for grouped convolution gradient (#331)
• stablehlo.cholesky inconsistent on GPU vs CPU (#338)
• Incorrect traced code for upsampling functions (#339)
• Handling Wrapped Arrays Correctly (#345)
• dynamic_slice (#349)
• Handling random numbers correctly (#356)
• Cannot compile function with Module as parameter (#357)
• Cannot repeat >2-dimensional arrays (#360)
• Segfault on Integer Overflow (#368)
• [BUG]: Segfault with large amount of data (#380)
• Compile time scales with MPI ranks (and it shouldn't) (#386)
• Graceful termination on out of memory (#406)
• Autodiff code generation fails when caching values (#416)