Issues

• Windows ROCm 10.1 nightly runtime: hipHostRegister of pageable memory >=1MiB returns success but leaves corrupted state -> background-thread segfault (gfx1100; 7.14 runtime fails cleanly)

  ROCm/TheRock#8939 · Filter by author vecStardo opened on Oct 12, 2026
  1

• Enable Pytorch release workflows to build a single pytorch version but all python versions

  ROCm/TheRock#8938 · Filter by author HereThereBeDragons opened on Oct 12, 2026

• Ensure download/upload python packages works for dev --> release process

  ROCm/TheRock#8937 · Filter by author HereThereBeDragons opened on Oct 12, 2026
  

• Ensure promote_packages.py works with dev builds

  ROCm/TheRock#8936 · Filter by author HereThereBeDragons opened on Oct 12, 2026
  

• [Issue] Windows PyTorch nightly: test_storage_copy_releases_gil_during_blocking_cuda_to_cpu, test_memory_profiler_viz and test_generic_fp32_precision_propagates_to_cudnn_conv_rnn fail

  ecosystem: PyTorchmonitoring
  ROCm/TheRock#8935 · Filter by author TriveniTadapaneni opened on Oct 12, 2026

• [Issue] Linux JAX : build fails — missing librocm_smi64.so

  ecosystem: JAXmonitoring
  ROCm/TheRock#8927 · Filter by author TriveniTadapaneni opened on Oct 12, 2026
  2

• [Windows][Strix Halo gfx1151] HIP allocations capped at OS-visible RAM when BIOS UMA/VGM = 96 GB (ROCm 10.1, driver 26.10.41.05)

  ROCm/TheRock#8922 · Filter by author xyny89 opened on Oct 11, 2026
  1

• [Feature]: CentOS Stream Support

  ROCm/TheRock#8916 · Filter by author XSpielinbox opened on Oct 10, 2026

• [Issue] ASAN SEGV at pc 0 in libamdhip64.so.7 when torch.cuda.is_available() initializes HIP

  ROCm/TheRock#8908 · Filter by author dileepr1 opened on Oct 9, 2026

• Cancel Actions runs when a PR is merged

  ROCm/TheRock#8896 · Filter by author AmosLewis opened on Oct 9, 2026
  1

• [issue] Linux::asan CI lane is not yet stabilized — expect intermittent test failures (not blocking)

  monitoringplatform: Linuxsanitizers
  ROCm/TheRock#8890 · Filter by author amd-chiranjeevi opened on Oct 9, 2026
  1

• [Issue] Linux PyTorch nightly: TestMemPool test_custom_mempool_free_exception_* fail with OSError: Could not load this library

  ecosystem: PyTorchmonitoring
  ROCm/TheRock#8889 · Filter by author TriveniTadapaneni opened on Oct 9, 2026

• [Issue] Linux nightly: TestAmdSmiImport::test_amdsmi_bare_unversioned_path_uses_loader_first fails with libpython3.11.so.1.0 not found

  ecosystem: PyTorchmonitoring
  ROCm/TheRock#8888 · Filter by author TriveniTadapaneni opened on Oct 9, 2026
  1

  

• [Issue] Windows gfx1151 PyTorch test_int64_upsample3d_cuda_bfloat16and test_mempool_oom_recovery_releases_cached_blocks fail with HIP OOM

  ecosystem: PyTorchmonitoring
  ROCm/TheRock#8886 · Filter by author TriveniTadapaneni opened on Oct 9, 2026
  

• Hugging Face model training throughput might regress for specific models on AMD Instinct MI350X

  Verified Issue
  ROCm/TheRock#8878 · Filter by author prbasyal-amd opened on Oct 8, 2026
  

• CMake 4.x: rocRAND configure fails with 'ADD_LIBRARY called with SHARED option' (CMP0164, find_package before languages are enabled)

  ROCm/TheRock#8847 · Filter by author bernardogv opened on Oct 8, 2026

• Multi-application workloads might experience an intermittent GPU page fault on some Instinct GPUs

  Verified Issue
  ROCm/TheRock#8844 · Filter by author prbasyal-amd opened on Oct 7, 2026
  

• Device-to-device hipMemcpyAsync might fail on AMD Instinct MI200 Series GPUs

  Verified Issue
  ROCm/TheRock#8843 · Filter by author prbasyal-amd opened on Oct 7, 2026
  

• ComfyUI workloads might show lower performance on Radeon Pro W7900 GPUs

  Verified Issue
  ROCm/TheRock#8842 · Filter by author prbasyal-amd opened on Oct 7, 2026
  

• Align FileLock contention behavior on Windows and Linux

  ROCm/TheRock#8839 · Filter by author rkayaith opened on Oct 7, 2026
  2

  

• [Tracking] User-facing failures from incompatible PyTorch packages on the stable index

  bugecosystem: PyTorch
  ROCm/TheRock#8802 · Filter by author lucbruni-amd opened on Oct 6, 2026

• [Issue] PyTorch release workflows can publish alpha versions to stable

  bugecosystem: PyTorch
  ROCm/TheRock#8798 · Filter by author ScottTodd opened on Oct 6, 2026
  2

  

• RAS error injection and query are unavailable on AMD Instinct MI350P GPUs

  Verified Issue
  ROCm/TheRock#8794 · Filter by author prbasyal-amd opened on Oct 6, 2026
  

• Valgrind workloads might stop responding on WSL2 with high system RAM

  Verified Issue
  ROCm/TheRock#8793 · Filter by author prbasyal-amd opened on Oct 6, 2026
  

• RCCL allreduce operations using the LL protocol might produce incorrect results on AMD Instinct MI350 Series GPUs

  Verified Issue
  ROCm/TheRock#8792 · Filter by author prbasyal-amd opened on Oct 6, 2026