Added torch.linalg.matrix_power

[heitorschueroff]

Stack from ghstack:

• Deprecate torch.matrix_power in favor of torch.linalg.matrix_power #53538 Deprecate torch.matrix_power in favor of torch.linalg.matrix_power
• Added torch.linalg.matrix_power #52608 Added torch.linalg.matrix_power

TODO

• Add OpInfo
• Update documentation
• Add more tests and compare against NumPy

Differential Revision: D27261532

[facebook-github-bot]

💊 CI failures summary and remediations

As of commit 8a24747 (more details on the Dr. CI page):

---

• 1/1 failures possibly* introduced in this PR
  • 1/1 non-scanned failure(s)

---

This comment was automatically generated by Dr. CI (expand for details).

Follow this link to opt-out of these comments for your Pull Requests.

Please report bugs/suggestions to the (internal) Dr. CI Users group.

[mruberry]

This looks good to me except for the out= behavior, which would be nice to address (and maybe providing the helper function for safe copying we've discussed before).

The ideep/tensorpipe changes should be reverted and the PR rebased past the mypy failures in its base.

The plan is for a follow-up PR to alias torch.matrix_power to torch.linalg.matrix_power, right?

@IvanYashchuk would you like to take a look?

[IvanYashchuk]

Looks good to me!

@heitorschueroff could you also post the runtime for the OpInfo checks? (--durations=0 option for pytest)

[heitorschueroff]

Looks good to me!

@heitorschueroff could you also post the runtime for the OpInfo checks? (--durations=0 option for pytest)

======================================================================================================= slowest durations =======================================================================================================
11.75s call     test/test_ops.py::TestGradientsCUDA::test_fn_gradgrad_linalg_matrix_power_cuda_complex128
5.45s call     test/test_ops.py::TestGradientsCPU::test_fn_gradgrad_linalg_matrix_power_cpu_complex128
3.07s call     test/test_ops.py::TestGradientsCUDA::test_fn_gradgrad_linalg_matrix_power_cuda_float64
2.65s call     test/test_ops.py::TestOpInfoCUDA::test_supported_dtypes_linalg_matrix_power_cuda_complex128
1.73s call     test/test_ops.py::TestGradientsCUDA::test_fn_grad_linalg_matrix_power_cuda_complex128
1.65s call     test/test_ops.py::TestGradientsCPU::test_fn_gradgrad_linalg_matrix_power_cpu_float64
1.32s call     test/test_ops.py::TestCommonCUDA::test_variant_consistency_jit_linalg_matrix_power_cuda_complex128
1.32s call     test/test_ops.py::TestCommonCUDA::test_variant_consistency_jit_linalg_matrix_power_cuda_complex64
1.20s call     test/test_ops.py::TestCommonCUDA::test_variant_consistency_jit_linalg_matrix_power_cuda_float32
1.19s call     test/test_ops.py::TestCommonCUDA::test_variant_consistency_jit_linalg_matrix_power_cuda_float64
1.04s call     test/test_ops.py::TestCommonCPU::test_variant_consistency_jit_linalg_matrix_power_cpu_complex128
1.00s call     test/test_ops.py::TestCommonCPU::test_variant_consistency_jit_linalg_matrix_power_cpu_complex64
0.95s call     test/test_ops.py::TestCommonCPU::test_variant_consistency_jit_linalg_matrix_power_cpu_float32
0.93s call     test/test_ops.py::TestCommonCPU::test_variant_consistency_jit_linalg_matrix_power_cpu_float64
0.65s call     test/test_ops.py::TestGradientsCPU::test_fn_grad_linalg_matrix_power_cpu_complex128
0.56s call     test/test_ops.py::TestGradientsCUDA::test_fn_grad_linalg_matrix_power_cuda_float64
0.22s call     test/test_ops.py::TestGradientsCPU::test_fn_grad_linalg_matrix_power_cpu_float64
0.13s call     test/test_ops.py::TestOpInfoCPU::test_supported_dtypes_linalg_matrix_power_cpu_complex128
0.11s call     test/test_ops.py::TestCommonCUDA::test_out_linalg_matrix_power_cuda_float32
0.11s call     test/test_ops.py::TestOpInfoCUDA::test_supported_dtypes_linalg_matrix_power_cuda_complex64
0.10s call     test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_linalg_matrix_power_cuda_complex64
0.10s call     test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_linalg_matrix_power_cuda_complex128
0.10s call     test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_linalg_matrix_power_cuda_float32
0.09s call     test/test_ops.py::TestCommonCUDA::test_variant_consistency_eager_linalg_matrix_power_cuda_float64
0.09s call     test/test_ops.py::TestOpInfoCUDA::test_unsupported_dtypes_linalg_matrix_power_cuda_bool
0.09s call     test/test_ops.py::TestOpInfoCUDA::test_supported_dtypes_linalg_matrix_power_cuda_float64
0.08s call     test/test_ops.py::TestOpInfoCUDA::test_supported_dtypes_linalg_matrix_power_cuda_float32
0.08s call     test/test_ops.py::TestOpInfoCUDA::test_unsupported_dtypes_linalg_matrix_power_cuda_bfloat16
0.08s call     test/test_ops.py::TestOpInfoCUDA::test_unsupported_dtypes_linalg_matrix_power_cuda_int16
0.08s call     test/test_ops.py::TestOpInfoCUDA::test_unsupported_dtypes_linalg_matrix_power_cuda_uint8
0.08s call     test/test_ops.py::TestOpInfoCUDA::test_unsupported_dtypes_linalg_matrix_power_cuda_int8
0.08s call     test/test_ops.py::TestOpInfoCUDA::test_unsupported_dtypes_linalg_matrix_power_cuda_float16
0.08s call     test/test_ops.py::TestOpInfoCUDA::test_unsupported_dtypes_linalg_matrix_power_cuda_int64
0.08s call     test/test_ops.py::TestOpInfoCUDA::test_unsupported_dtypes_linalg_matrix_power_cuda_int32
0.05s call     test/test_ops.py::TestGradientsCUDA::test_inplace_grad_linalg_matrix_power_cuda_complex128
0.04s call     test/test_ops.py::TestGradientsCUDA::test_inplace_gradgrad_linalg_matrix_power_cuda_complex128
0.04s call     test/test_ops.py::TestGradientsCUDA::test_inplace_gradgrad_linalg_matrix_power_cuda_float64
0.04s call     test/test_ops.py::TestGradientsCUDA::test_inplace_grad_linalg_matrix_power_cuda_float64
0.01s call     test/test_ops.py::TestCommonCPU::test_out_linalg_matrix_power_cpu_float32
0.01s call     test/test_ops.py::TestCommonCPU::test_variant_consistency_eager_linalg_matrix_power_cpu_complex64
0.01s call     test/test_ops.py::TestCommonCPU::test_variant_consistency_eager_linalg_matrix_power_cpu_complex128
0.01s call     test/test_ops.py::TestCommonCPU::test_variant_consistency_eager_linalg_matrix_power_cpu_float64
0.01s call     test/test_ops.py::TestCommonCPU::test_variant_consistency_eager_linalg_matrix_power_cpu_float32
0.01s call     test/test_ops.py::TestOpInfoCPU::test_unsupported_dtypes_linalg_matrix_power_cpu_bool
0.01s call     test/test_ops.py::TestOpInfoCPU::test_unsupported_dtypes_linalg_matrix_power_cpu_int16
0.01s call     test/test_ops.py::TestOpInfoCPU::test_unsupported_dtypes_linalg_matrix_power_cpu_int64
0.01s call     test/test_ops.py::TestOpInfoCPU::test_unsupported_dtypes_linalg_matrix_power_cpu_int32
0.01s call     test/test_ops.py::TestOpInfoCPU::test_unsupported_dtypes_linalg_matrix_power_cpu_int8
0.01s call     test/test_ops.py::TestOpInfoCPU::test_unsupported_dtypes_linalg_matrix_power_cpu_uint8
0.01s call     test/test_ops.py::TestOpInfoCPU::test_unsupported_dtypes_linalg_matrix_power_cpu_bfloat16

(124 durations < 0.005s hidden.  Use -vv to show these durations.)
========================================================================================

[mruberry]

Nice work, @heitorschueroff!

I think the test failures are erroneous. You may want to rebase to verify the docs build failure isn't related to this PR.

[mruberry]

This accidentally picked up some fbgemm and tensorpipe changes that should be reverted.

[facebook-github-bot]

@heitorschueroff merged this pull request in f9e7f13.