Tags: Anai-Guo/ao

[ROCm] test(moe): skip jagged rowwise scales test.

Test failing post merge pytorch#4113
pytest test_kernels.py -k test_row_major_with_jagged_rowwise_scales -vv
test_kernels.py::test_row_major_with_jagged_rowwise_scales[True] FAILED                         [ 25%]
test_kernels.py::test_row_major_with_jagged_rowwise_scales[False] FAILED                        [ 50%]
test_kernels.py::test_row_major_with_jagged_rowwise_scales_transpose_method[True] PASSED        [ 75%]
test_kernels.py::test_row_major_with_jagged_rowwise_scales_transpose_method[False] PASSED       [100%]

Signed-off-by: Jagadish Krishnamoorthy <jagadish.krishnamoorthy@amd.com>

[ROCm] MXFP8 MoE: persistent grouped kernel, F.scaled_mm dense, corre…

…ctness + tests

- Persistent grouped-MM kernel (grid = num_CUs * ctas_per_cu); walks experts
  in-kernel with a global tile counter. Avoids silent row-dropping under
  (M+E-1)//E bounds and keeps the dispatcher torch.compile-clean.
- Dense MXFP8 path: dispatch to F.scaled_mm with BlockWise1x32.
- Wgrad: retune default tile to (BN=256, BK=256, BM=64, nw=8).
- K-tail and scale-tail masking; m_mask bounded by group_end and global M.
- torch.compile: register pad/unpad helpers as torch.library.custom_op;
  skip nonstrict_trace on ROCm.
- mx_linear / MXFP8TrainingOpConfig: drop is_ROCM() auto-switch; expose
  mxfp8_dim1_cast_kernel_choice as explicit arg (CUDA default).
- bench_2d_3d_grouped_gemm.py: run on MI350+ via bench_mxfp8_grouped_mm_rocm;
  fix flops formula = 2 * M * N * K.

Tested on MI355X / gfx950 / ROCm 7.1 / Triton 3.7:
  Accuracy: test/prototype/moe_training/test_mxfp8_grouped_mm.py
            -> 129 passed, 16 skipped.
            SQNR margins: out >= 27.6 (>= 27), in_grad >= 25.2 (>= 25),
            w_grad >= 25.5 (>= 24).
  Perf:     benchmarks/prototype/moe_training/bench_2d_3d_grouped_gemm.py

skip test_tensor test

Enable GPTQ on Intel GPU

Remove unnecessary Torch dependency from ExecuTorch ops build

The ExecuTorch ops target defines TORCHAO_SHARED_KERNELS_BUILD_EXECUTORCH=1,
which means shared kernel headers use ExecuTorch includes, not Torch includes.
The find_package(Torch) and TORCH_INCLUDE_DIRS were therefore unnecessary and
caused CMake configure failures in standalone ExecuTorch builds where PyTorch
is not discoverable via find_package.

Remove unnecessary Torch dependency from ExecuTorch ops build

The ExecuTorch ops target defines TORCHAO_SHARED_KERNELS_BUILD_EXECUTORCH=1,
which means shared kernel headers use ExecuTorch includes, not Torch includes.
The find_package(Torch) and TORCH_INCLUDE_DIRS were therefore unnecessary and
caused CMake configure failures in standalone ExecuTorch builds where PyTorch
is not discoverable via find_package.

Remove unnecessary Torch dependency from ExecuTorch ops build

The ExecuTorch ops target defines TORCHAO_SHARED_KERNELS_BUILD_EXECUTORCH=1,
which means shared kernel headers use ExecuTorch includes, not Torch includes.
The find_package(Torch) and TORCH_INCLUDE_DIRS were therefore unnecessary and
caused CMake configure failures in standalone ExecuTorch builds where PyTorch
is not discoverable via find_package.

Remove unnecessary Torch dependency from ExecuTorch ops build

The ExecuTorch ops target defines TORCHAO_SHARED_KERNELS_BUILD_EXECUTORCH=1,
which means shared kernel headers use ExecuTorch includes, not Torch includes.
The find_package(Torch) and TORCH_INCLUDE_DIRS were therefore unnecessary and
caused CMake configure failures in standalone ExecuTorch builds where PyTorch
is not discoverable via find_package.

Remove unnecessary Torch dependency from ExecuTorch ops build

The ExecuTorch ops target defines TORCHAO_SHARED_KERNELS_BUILD_EXECUTORCH=1,
which means shared kernel headers use ExecuTorch includes, not Torch includes.
The find_package(Torch) and TORCH_INCLUDE_DIRS were therefore unnecessary and
caused CMake configure failures in standalone ExecuTorch builds where PyTorch
is not discoverable via find_package.

Remove unnecessary Torch dependency from ExecuTorch ops build

The ExecuTorch ops target defines TORCHAO_SHARED_KERNELS_BUILD_EXECUTORCH=1,
which means shared kernel headers use ExecuTorch includes, not Torch includes.
The find_package(Torch) and TORCH_INCLUDE_DIRS were therefore unnecessary and
caused CMake configure failures in standalone ExecuTorch builds where PyTorch
is not discoverable via find_package.