From aaeb5388585b95415bc1358210b0b4bb37d8da18 Mon Sep 17 00:00:00 2001 From: Leo Fang Date: Mon, 5 Oct 2026 20:26:44 -0700 Subject: [PATCH 1/4] Bump nightly PyTorch highest version to 2.14.1 --- ci/test-matrix.yml | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/ci/test-matrix.yml b/ci/test-matrix.yml index df0b9600492..fb4653a34ed 100644 --- a/ci/test-matrix.yml +++ b/ci/test-matrix.yml @@ -91,12 +91,12 @@ linux: - { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '13.4.2', LOCAL_CTK: '0', GPU: 'rtx4090', GPU_COUNT: '1', DRIVER: 'latest', FLAVOR: 'wsl' } nightly: # nightly-pytorch - - { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '12.6.3', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.12.1', TORCH_CUDA: 'cu126' } } - - { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '13.0.2', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.12.1', TORCH_CUDA: 'cu130' } } + - { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '12.6.3', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.14.1', TORCH_CUDA: 'cu126' } } + - { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '13.0.2', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.14.1', TORCH_CUDA: 'cu130' } } - { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '12.6.3', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.9.1', TORCH_CUDA: 'cu126' } } - { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '13.0.2', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.9.1', TORCH_CUDA: 'cu130' } } - - { ARCH: 'arm64', PY_VER: '3.12', CUDA_VER: '12.6.3', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.12.1', TORCH_CUDA: 'cu126' } } - - { ARCH: 'arm64', PY_VER: '3.12', CUDA_VER: '13.0.2', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.12.1', TORCH_CUDA: 'cu130' } } + - { ARCH: 'arm64', PY_VER: '3.12', CUDA_VER: '12.6.3', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.14.1', TORCH_CUDA: 'cu126' } } + - { ARCH: 'arm64', PY_VER: '3.12', CUDA_VER: '13.0.2', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.14.1', TORCH_CUDA: 'cu130' } } - { ARCH: 'arm64', PY_VER: '3.12', CUDA_VER: '12.6.3', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.9.1', TORCH_CUDA: 'cu126' } } - { ARCH: 'arm64', PY_VER: '3.12', CUDA_VER: '13.0.2', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.9.1', TORCH_CUDA: 'cu130' } } # nightly-numba-cuda @@ -145,8 +145,8 @@ windows: - { ARCH: 'amd64', PY_VER: '3.14', CUDA_VER: '13.4.2', LOCAL_CTK: '0', GPU: 'h100', GPU_COUNT: '2', DRIVER: 'latest', DRIVER_MODE: 'MCDM' } nightly: # nightly-pytorch - - { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '12.6.3', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.12.1', TORCH_CUDA: 'cu126' } } - - { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '13.0.2', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.12.1', TORCH_CUDA: 'cu130' } } + - { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '12.6.3', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.14.1', TORCH_CUDA: 'cu126' } } + - { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '13.0.2', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.14.1', TORCH_CUDA: 'cu130' } } - { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '12.6.3', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.9.1', TORCH_CUDA: 'cu126' } } - { ARCH: 'amd64', PY_VER: '3.12', CUDA_VER: '13.0.2', LOCAL_CTK: '0', GPU: 'l4', GPU_COUNT: '1', DRIVER: 'latest', DRIVER_MODE: 'TCC', ENV: { MODE: 'nightly-pytorch', TORCH_VER: '2.9.1', TORCH_CUDA: 'cu130' } } # nightly-numba-cuda From 6de0ff9a384ec6d434ab7b4404ad579fd5ce3739 Mon Sep 17 00:00:00 2001 From: Leo Fang Date: Mon, 5 Oct 2026 20:26:53 -0700 Subject: [PATCH 2/4] Temporarily add push trigger to ci-nightly.yml for testing Remove before merging. --- .github/workflows/ci-nightly.yml | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/.github/workflows/ci-nightly.yml b/.github/workflows/ci-nightly.yml index a3179c1e155..3504ad9926f 100644 --- a/.github/workflows/ci-nightly.yml +++ b/.github/workflows/ci-nightly.yml @@ -17,6 +17,10 @@ concurrency: cancel-in-progress: true on: + push: + branches: + - "main" + - "pull-request/[0-9]+" schedule: # 2:17 AM UTC daily, after the midnight main CI build finishes. # Avoid minute 0 because GitHub documents high scheduled-workflow load From c720c0dd6b1de0e40cf776b1551366bcd720bb10 Mon Sep 17 00:00:00 2001 From: Leo Fang Date: Mon, 5 Oct 2026 20:45:20 -0700 Subject: [PATCH 3/4] cuda.core: bump torch tensor-bridge upper cap to 2.14 --- cuda_core/cuda/core/_memoryview.pyx | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/cuda_core/cuda/core/_memoryview.pyx b/cuda_core/cuda/core/_memoryview.pyx index 8a6e989044b..767d0cf05a4 100644 --- a/cuda_core/cuda/core/_memoryview.pyx +++ b/cuda_core/cuda/core/_memoryview.pyx @@ -51,7 +51,7 @@ cdef dict _torch_type_cache = {} cdef object _torch_version_ok = None cdef inline bint _torch_version_check(): - """Return True if 2.3 <= torch <= 2.12 (known AOTI ABI range). Memoized. + """Return True if 2.3 <= torch <= 2.14 (known AOTI ABI range). Memoized. Lower bound: AOTI functions we use were introduced in PyTorch 2.3. Upper bound: the ``pyobj_to_aten_handle`` trick relies on the @@ -72,7 +72,7 @@ cdef inline bint _torch_version_check(): try: major, minor = int(torch.__version__.split(".")[0]), \ int(torch.__version__.split(".")[1]) - _torch_version_ok = (2, 3) <= (major, minor) <= (2, 12) + _torch_version_ok = (2, 3) <= (major, minor) <= (2, 14) except (ValueError, IndexError): _torch_version_ok = False return _torch_version_ok From c43f495255fc41515e005c9750899bd9e57a069c Mon Sep 17 00:00:00 2001 From: Leo Fang Date: Mon, 5 Oct 2026 21:45:37 -0700 Subject: [PATCH 4/4] Revert "cuda.core: bump torch tensor-bridge upper cap to 2.14" This reverts commit c720c0dd6b1de0e40cf776b1551366bcd720bb10. --- cuda_core/cuda/core/_memoryview.pyx | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/cuda_core/cuda/core/_memoryview.pyx b/cuda_core/cuda/core/_memoryview.pyx index 767d0cf05a4..8a6e989044b 100644 --- a/cuda_core/cuda/core/_memoryview.pyx +++ b/cuda_core/cuda/core/_memoryview.pyx @@ -51,7 +51,7 @@ cdef dict _torch_type_cache = {} cdef object _torch_version_ok = None cdef inline bint _torch_version_check(): - """Return True if 2.3 <= torch <= 2.14 (known AOTI ABI range). Memoized. + """Return True if 2.3 <= torch <= 2.12 (known AOTI ABI range). Memoized. Lower bound: AOTI functions we use were introduced in PyTorch 2.3. Upper bound: the ``pyobj_to_aten_handle`` trick relies on the @@ -72,7 +72,7 @@ cdef inline bint _torch_version_check(): try: major, minor = int(torch.__version__.split(".")[0]), \ int(torch.__version__.split(".")[1]) - _torch_version_ok = (2, 3) <= (major, minor) <= (2, 14) + _torch_version_ok = (2, 3) <= (major, minor) <= (2, 12) except (ValueError, IndexError): _torch_version_ok = False return _torch_version_ok