v0.2.7

NNlibCUDA v0.2.7

Diff since v0.2.6

Merged pull requests:

• Updates for CUDA.jl 4. (#58) (@maleadt)

v0.2.6

NNlibCUDA v0.2.6

Diff since v0.2.5

Merged pull requests:

• overload NNlib._rng_compat_array (#62) (@mcabbott)

v0.2.5

NNlibCUDA v0.2.5

Diff since v0.2.4

Merged pull requests:

• Missing Identity Activation Import (#45) (@avik-pal)
• Fix versions in Buildkite CI (#56) (@ToucheSir)
• added fold/unfold and gpu tests (#59) (@nikopj)
• Add CUDA tests for dropout (#60) (@mcabbott)

v0.2.4

NNlibCUDA v0.2.4

Diff since v0.2.3

Closed issues:

• NNlibCUDA Heisenbug in conv! with nonzero beta (#37)

Merged pull requests:

• get rid of duplicate code in upsampling code by using dispatch (#49) (@maxfreu)
• save allocs during algorithm search (#53) (@maxfreu)
• print/convert batchedadjtrans over cuarray (#54) (@chengchingwen)
• Move ctc_loss from Flux to NNlibCUDA (#55) (@mcabbott)

v0.2.3

NNlibCUDA v0.2.3

Diff since v0.2.2

Closed issues:

• Unconstrained element type on activations causing errors with e.g. Complex, ForwardDiff.Dual (#47)

Merged pull requests:

• Restrict element type of activation overrides to CUDNN datatypes (#48) (@DomCRose)

v0.2.2

NNlibCUDA v0.2.2

Diff since v0.2.1

Closed issues:

• Slow ∇softmax! compared with generic version. (#30)

Merged pull requests:

• track_stats option for batchnorm (#36) (@CarloLucibello)
• update softmax tests to match NNlib 0.8.3 (#44) (@mcabbott)

v0.2.1

NNlibCUDA v0.2.1

Diff since v0.2.0

Closed issues:

• No frule for some activations (#42)

Merged pull requests:

• specialize on op in scatter (#43) (@CarloLucibello)

v0.2.0

NNlibCUDA v0.2.0

Diff since v0.1.11

Merged pull requests:

• Add support for asymmetric padding for convlayers (#34) (@DrChainsaw)
• Use new PoolDims constructor for NNlib v0.8 (#40) (@DhairyaLGandhi)

v0.1.11

NNlibCUDA v0.1.11

Diff since v0.1.10

Closed issues:

• gpu scatter with CartesianIndex not supported (#29)

Merged pull requests:

• Support gpu scatter/gather with CartesianIndex (#33) (@yuehhua)

v0.1.10

NNlibCUDA v0.1.10

Diff since v0.1.9

Merged pull requests:

• Add CUDA kernels for grid sampling (#31) (@pxl-th)