Tags: KristianHolme/Lux.jl
Tags
[Diff since LuxLib-v1.7.2](LuxDL/Lux.jl@LuxLib-v1.7.2...LuxLib-v1.7.3) **Merged pull requests:** - fix: remove Optimisers.jl patch (LuxDL#1247) (@avik-pal) - show debug instead of warning if cant cuBLAS mult (LuxDL#1280) (@ExpandingMan) - chore: bump crate-ci/typos from 1.30.2 to 1.31.0 (LuxDL#1281) (@dependabot[bot]) - chore: remove debug functionalities of reactant (LuxDL#1285) (@avik-pal) - fix: add a ReactantOptimisers wrapper (LuxDL#1288) (@avik-pal) - chore: bump crate-ci/typos from 1.31.0 to 1.31.1 (LuxDL#1297) (@dependabot[bot]) - ci: use JuliaFormatter v1 (LuxDL#1299) (@avik-pal) - ci: multiple ci fixes (LuxDL#1301) (@avik-pal) - CompatHelper: bump compat for JET to 0.10 for package LuxTestUtils, (keep existing compat) (LuxDL#1302) (@github-actions[bot]) - fix: new reactant version (LuxDL#1303) (@avik-pal) - fix: update Reactant training (LuxDL#1304) (@avik-pal) - fix: chain rules for recurrence tuple inputs (LuxDL#1306) (@avik-pal) - feat: fix return (LuxDL#1307) (@avik-pal) - fix: try increasing the samples in CI (LuxDL#1309) (@avik-pal) - fix: restrict dispatch types for cublaslt (LuxDL#1311) (@avik-pal) **Closed issues:** - Optimisers.jl patch for Reactant Support (LuxDL#1146) - Convolutional VAE for MNIST using Reactant failed to produce right results (LuxDL#1274) - GPU Inefficiency in Gradient Computation with Custom Recurrence (LuxDL#1284) - PINN2DPDE broke in the latest Optimisers Patch removal (LuxDL#1286) - Duplicating Scalars for Optimisers prevents CSE (LuxDL#1289) - Reactant 0.2.61 produces incorrect gradients (LuxDL#1292) - missing or incorrect `ProjectTo` method breaks `Recurrence` with Zygote (LuxDL#1305) - dense operations fail on views on nvidia due to missing method (LuxDL#1308)
[Diff since v1.12.3](LuxDL/Lux.jl@v1.12.3...v1.12.4) **Merged pull requests:** - feat: fix return (LuxDL#1307) (@avik-pal)
[Diff since v1.12.2](LuxDL/Lux.jl@v1.12.2...v1.12.3) **Merged pull requests:** - fix: update Reactant training (LuxDL#1304) (@avik-pal) - fix: chain rules for recurrence tuple inputs (LuxDL#1306) (@avik-pal) **Closed issues:** - Duplicating Scalars for Optimisers prevents CSE (LuxDL#1289) - missing or incorrect `ProjectTo` method breaks `Recurrence` with Zygote (LuxDL#1305)
[Diff since v1.12.1](LuxDL/Lux.jl@v1.12.1...v1.12.2) **Merged pull requests:** - chore: bump crate-ci/typos from 1.31.0 to 1.31.1 (LuxDL#1297) (@dependabot[bot]) - ci: use JuliaFormatter v1 (LuxDL#1299) (@avik-pal) - ci: multiple ci fixes (LuxDL#1301) (@avik-pal) - CompatHelper: bump compat for JET to 0.10 for package LuxTestUtils, (keep existing compat) (LuxDL#1302) (@github-actions[bot]) - fix: new reactant version (LuxDL#1303) (@avik-pal) **Closed issues:** - Reactant 0.2.61 produces incorrect gradients (LuxDL#1292)
[Diff since v1.12.0](LuxDL/Lux.jl@v1.12.0...v1.12.1) **Merged pull requests:** - fix: add a ReactantOptimisers wrapper (LuxDL#1288) (@avik-pal) **Closed issues:** - PINN2DPDE broke in the latest Optimisers Patch removal (LuxDL#1286)
[Diff since v1.11.2](LuxDL/Lux.jl@v1.11.2...v1.12.0) **Merged pull requests:** - fix: remove Optimisers.jl patch (LuxDL#1247) (@avik-pal) - show debug instead of warning if cant cuBLAS mult (LuxDL#1280) (@ExpandingMan) - chore: bump crate-ci/typos from 1.30.2 to 1.31.0 (LuxDL#1281) (@dependabot[bot]) - chore: remove debug functionalities of reactant (LuxDL#1285) (@avik-pal) **Closed issues:** - Optimisers.jl patch for Reactant Support (LuxDL#1146)
[Diff since v1.11.1](LuxDL/Lux.jl@v1.11.1...v1.11.2) **Merged pull requests:** - test: fix tests (LuxDL#1278) (@avik-pal) - fix: try fixing broken tests (LuxDL#1279) (@avik-pal)
[Diff since LuxLib-v1.7.1](LuxDL/Lux.jl@LuxLib-v1.7.1...LuxLib-v1.7.2) **Merged pull requests:** - test: fix tests (LuxDL#1278) (@avik-pal) - fix: try fixing broken tests (LuxDL#1279) (@avik-pal)
[Diff since v1.11.0](LuxDL/Lux.jl@v1.11.0...v1.11.1) **Merged pull requests:** - chore: bump forwarddiff to 1.0 (LuxDL#1277) (@avik-pal)
[Diff since v1.10.1](LuxDL/Lux.jl@v1.10.1...v1.11.0) **Merged pull requests:** - docs: refer to the Turing docs for BayesianNN (LuxDL#1272) (@github-actions[bot]) - feat: support for ForwardDiff training (LuxDL#1273) (@hstrey)
PreviousNext