Convolution portable kernel missing 5D support

🕵️‍♂️ Detected with FACTO

🐛 Describe the bug

FACTO Generated meta_tuple:

['ArgType.Tensor torch.float32 (8, 8, 5, 2, 7)', 'ArgType.Tensor torch.float32 (8, 8, 8, 5, 4)', 'ArgType.TensorOpt None', 'ArgType.LengthList [6]', 'ArgType.LengthList [7]', 'ArgType.LengthList [9, 1, 4]', 'ArgType.Bool True', 'ArgType.LengthList [0, 5, 3]', 'ArgType.Int 1']

error

[kernel_ops_util.cpp:345] Check failed (in.dim() == 3 || in.dim() == 4): Expect input tensor to be 3-D or 4-D, but got, 5.
[op_convolution.cpp:377] Check failed (check_convolution_args( in, weight, bias, stride, padding, dilation, transposed, output_padding, groups, out)): 
[method.cpp:1338] KernelCall failed at instruction 0:0 in operator aten::convolution.out: 0x12
[method.cpp:1344] arg 0 with type id 1
[method.cpp:1344] arg 1 with type id 1
[method.cpp:1344] arg 2 with type id 0
[method.cpp:1344] arg 3 with type id 8
[method.cpp:1344] arg 4 with type id 8
[method.cpp:1344] arg 5 with type id 8
[method.cpp:1344] arg 6 with type id 5
[method.cpp:1344] arg 7 with type id 8
[method.cpp:1344] arg 8 with type id 4
[method.cpp:1344] arg 9 with type id 1
[method.cpp:1344] arg 10 with type id 1

Versions

nightly

cc @larryliu0820

Start with the check reported in kernel_ops_util.cpp around line 345 and the convolution argument validation in op_convolution.cpp around line 377. Reproduce the FACTO-generated 5D case and verify that convolution accepts the shown 5D input and weight tensors without triggering the dimensionality check.