LFQ: why can't we have projections?

As title says, why can't we have projections in lfq? https://github.com/lucidrains/vector-quantize-pytorch/blob/5ae9c792e77b45751ac636eb72b310f09648aac3/vector_quantize_pytorch/residual_lfq.py#L86 they might be useful to have a better learning residual by residual

thanks

Start with vector_quantize_pytorch/residual_lfq.py around line 86 and review how LFQ currently handles residuals. The issue needs clarification on the intended projection behavior; done would require an agreed scope for projections and evidence that the resulting residual-learning change works.

Written by the indexing model from the issue text.