Enable MobileBert and Albert with XNNPACK quantization

[mergennachin]

Currently quantization in CI is disabled for MobileBert and Albert

Here's the error code: https://gist.github.com/mergennachin/2abc5cae46925abdb9fc582b1c520159

cc @digantdesai @mcr229 @cbilgin

Internal task: T197452682

[brucekk-kim]

@mergennachin Samsung's also waiting on the fix as we've got the same problem generating .pte file for quantized MobileBert model.
We are using branch: https://github.com/pytorch/executorch/tree/v0.6.0-rc6-android

[digantdesai]

Thanks @mergennachin @brucekk-kim - we will take a look.

[mcr229]

I remember some one mentioned this could be related to the embedding? Do we know if that was relevant?

[digantdesai]

@mcr229 not sure, can you dig into it? 🙏