Issues

• LayerNorm should use NNlib.layernorm in the forward

  FluxML/Flux.jl#2771 · Filter by author CarloLucibello opened on Oct 22, 2026

• GC stalls in trainstep! loops on CUDA; expose the caching allocator outside train!

  FluxML/Flux.jl#2768 · Filter by author CarloLucibello opened on Oct 22, 2026

• logitcrossentropy with integer labels: 15 ms of memory traffic on large bf16 logits

  FluxML/Flux.jl#2767 · Filter by author CarloLucibello opened on Oct 22, 2026

• add links to the example folder

  FluxML/Flux.jl#2761 · Filter by author CarloLucibello opened on Oct 20, 2026

• Document mixed-precision (BFloat16/Float16) training

  documentation
  FluxML/Flux.jl#2759 · Filter by author CarloLucibello opened on Oct 20, 2026

• Device-agnostic, shape-keyed compile cache for any function on Reactant (Flux.compile)

  FluxML/Flux.jl#2755 · Filter by author CarloLucibello opened on Oct 19, 2026

• [AMDGPU] wrong output from BatchNorm

  FluxML/Flux.jl#2649 · Filter by author CarloLucibello opened on Jan 17, 2026
  1

• Zero-sized Array after indexing causes exception during gradient

  FluxML/Flux.jl#2648 · Filter by author lenianiva opened on Jan 14, 2026
  5

• BatchNorm Memory Blowup on Very Large CUDA Arrays

  allocations
  FluxML/Flux.jl#2616 · Filter by author BenCurran98 opened on Aug 18, 2025
  4

• Enzyme fails on simple MLP on GPU

  enzyme
  Bug

  FluxML/Flux.jl#2607 · Filter by author lassepe opened on May 28, 2025
  8

• ROCM-Aware MPI requires AMDGPU.synchronize()

  FluxML/Flux.jl#2591 · Filter by author Alexander-Barth opened on Feb 18, 2025
  1

• Nil doesn't understand SpecialFunctions functions

  FluxML/Flux.jl#2588 · Filter by author ToucheSir opened on Feb 6, 2025
  3

• Create a FluxCore package?

  FluxML/Flux.jl#2572 · Filter by author CarloLucibello opened on Jan 2, 2025

• fix Vararg warning in test

  enzyme
  FluxML/Flux.jl#2570 · Filter by author CarloLucibello opened on Dec 31, 2024
  6

• cuda gpu memory usage increasing in time

  allocationscudaperformance
  FluxML/Flux.jl#2523 · Filter by author CarloLucibello opened on Nov 9, 2024
  4

• feature requests for RNNs

  RNN
  FluxML/Flux.jl#2514 · Filter by author CarloLucibello opened on Nov 4, 2024
  12

• Depthwise convolutions produce a large number of allocations

  allocationsperformance
  FluxML/Flux.jl#2508 · Filter by author JoshuaBillson opened on Oct 28, 2024
  2

• Add option to throw error on passing wrong precision floats to layers

  performance
  FluxML/Flux.jl#2454 · Filter by author lassepe opened on Jun 7, 2024
  5

• Enzyme fails with MultiHeadAttention layer on CUDA

  enzyme
  FluxML/Flux.jl#2448 · Filter by author mashu opened on May 16, 2024
  15

• test Enzyme gpu support

  cudaenzyme
  FluxML/Flux.jl#2445 · Filter by author CarloLucibello opened on May 12, 2024

• Collecting PyTorch -> Flux migration notes

  documentation
  FluxML/Flux.jl#2410 · Filter by author BioTurboNick opened on Mar 25, 2024
  4

• can't use masks in multi-head-attention layer

  FluxML/Flux.jl#2336 · Filter by author alerem18 opened on Sep 15, 2023
  6

• show is confused by shared parameters

  FluxML/Flux.jl#2321 · Filter by author mcabbott opened on Aug 27, 2023
  1

• Separate nightly CI from posted CI badge

  discussiongood first issuetesting
  FluxML/Flux.jl#2237 · Filter by author ToucheSir opened on Apr 23, 2023
  2

• Weights and positive weights for labels in Flux losses

  documentation
  FluxML/Flux.jl#2153 · Filter by author gabrielpreviato opened on Jan 6, 2023
  6