v0.9.26

NNlib v0.9.26

Diff since v0.9.25

Merged pull requests:

• Add Atomix.jl v1 compat (#615) (@christiangnrd)

v0.9.25

NNlib v0.9.25

Diff since v0.9.24

Merged pull requests:

• Update method docstrings (#607) (@abhro)
• fix a test with batched_mul! (#611) (@CarloLucibello)
• Bump deps (#613) (@pxl-th)

Closed issues:

• Possible bug in NNlib when using CUDA's unified memory (#568)
• Can we get rid of auto-broadcasting of 0D arrays for activations? (#608)

v0.9.24

NNlib v0.9.24

Diff since v0.9.23

Merged pull requests:

• Prepare for pending Enzyme breaking change (#605) (@wsmoses)
• Bump patch version to release new Enzyme support (#606) (@wsmoses)

v0.9.23

NNlib v0.9.23

Diff since v0.9.22

Merged pull requests:

• remove NNPACK and move ForwardDiff to an extension (#603) (@avik-pal)

v0.9.22

NNlib v0.9.22

Diff since v0.9.21

Merged pull requests:

• CompatHelper: bump compat for AMDGPU in [weakdeps] to 1, (keep existing compat) (#602) (@github-actions[bot])

v0.9.21

NNlib v0.9.21

Diff since v0.9.20

Merged pull requests:

• Use KernelAbstractions for fold/unfold (#596) (@pxl-th)
• fix: remove inlining of sigmoid_fast pre 1.11- (#597) (@avik-pal)
• ci: update downstream testing (#598) (@avik-pal)
• fix: MIOpen only supports activations till dimension 5 (#601) (@avik-pal)

v0.9.20

NNlib v0.9.20

Diff since v0.9.19

Merged pull requests:

• Fix non-symmetric padding (#595) (@pxl-th)

Closed issues:

• pad_reflect broken (#594)

v0.9.19

NNlib v0.9.19

Diff since v0.9.18

Merged pull requests:

• Implement Short-time Fourier transform and its inverse (#587) (@pxl-th)

v0.9.18

NNlib v0.9.18

Diff since v0.9.17

Merged pull requests:

• Transition to LinearAlgebra.BLAS.libblastrampoline instead of Base.libblas_name (#589) (@staticfloat)
• fix typos in the doc for batched_mul! (#590) (@lanceXwq)
• Use GPU-friendly rounding (#591) (@pxl-th)
• Support conv_direct! on custom datatypes (#592) (@adrhill)
• Fix padding causing device to host copies (#593) (@pxl-th)

Closed issues:

• Error when using Float64: ERROR: UndefRefError: access to undefined reference (#490)
• Warning on using Upsample on AMDGPU (#573)

v0.9.17

NNlib v0.9.17

Diff since v0.9.16

Merged pull requests:

• Add support for AMDGPU 0.9 (#585) (@pxl-th)