Make DataFormatVecPermute's validation and implementation consistent with each other

[PatriceVignola]

There's currently a mismatch between what the validation of DataFormatVecPermute and its implementation is doing. While the validation allows src_format and dst_format to be of length 4 or 5, it only allows the input to be of shape [4] or [2, 4]. This leads to a gap in the validation where the following is allowed, but results in undefined behavior or uninitialized garbage data:

tf.raw_ops.DataFormatVecPermute(x=[1,2,3,4], src_format="NCDHW", dst_format="NDHWC")

[allenlavoie]

Looks like these tests are failing in their XLA+GPU configurations (//tensorflow/python:nn_test_xla_gpu). The tf2xla kernel likely needs updating:

tensorflow/tensorflow/compiler/tf2xla/kernels/data_format_ops.cc

Line 106 in db23048

void Compile(XlaOpKernelContext* ctx) override {

Alternatively you can use @test_util.disable_xla('reason') on the new tests for now (but please file a bug for the discrepancy to be fixed if so).

[gbaned]

@PatriceVignola Can you please check @allenlavoie's comments and keep us posted ? Thanks!

[PatriceVignola]

@gbaned @allenlavoie I copied the changes to the XLA kernel and re-ran the tests.

[allenlavoie]

Looks like clang-tidy identifies two issues. Can you take a look? https://github.com/tensorflow/tensorflow/blob/master/CONTRIBUTING.md#c-coding-style

Just in core/kernels/data_format_ops.h; explicit on a single-argument constructor and no trailing underscore for struct members.

[PatriceVignola]

Sure I'll fix it, but this is preexisting. I didn't touch this header besides changing the array size.