Don't Serialize Scales/ZP in Flatbuffer

[mcr229]

Other forms of data like weights/bias are serialized separately (either data store or end of flatbuffer segment), Scales and ZP however are serialized straight into Flatbuffer. This was ok when we were doing per-tensor and per-channel quantization because the number of scales was not large, but now with blockwise quantization the number of scales can be large. Realisitically since this is a form of data, we should put this in the same place weights/bias's are stored

Essentially we want to move data serialization of scales zp from this:

executorch/backends/xnnpack/operators/node_visitor.py

Line 278 in 0c6a71b

scale=scale.flatten().tolist(),

to something like this:

executorch/backends/xnnpack/operators/node_visitor.py

Line 496 in 0c6a71b

def get_serialized_buffer_index(

This is only something we should try to do with zeropoints/ scales that are tensors or lists. for per_tensor quantization with a single zp/scale, it becomes overkill to serialize the scales/zp separately, so we should leave those alone.

cc @digantdesai @cbilgin

[iseeyuan]

@mcr229 are you looking for contributors, or you wanna assign those issues to your self?

[mcr229]

@iseeyuan i'm putting down some work items in the github tracker. I'm open to working on it later, but if contributors find it and feel they can contribute that's also nice. How should i label/mark this issue so that outside contributors can attempt it if they want, while it is on my backlog?

i'll also add some more context on what the task really is

[goosebumps31]

I am looking to make my first contribution here. Can I have a shot at this issue?

[mcr229]

i'll assign it to you!