ttnn.split (analogous to torch.chunk) currently doesn't have support for incomplete splits (splits for which the number of elements in the dimension on which we split is not divisible by num_splits). For RM tensors, adding support for this may be as simple as ensuring that slice handles slices similarly to torch.slice and removing the TT_FATAL at the beginning of the split op checking for divisibility of shape[dim] by num_splits.
ttnn.split (analogous to torch.chunk) currently doesn't have support for incomplete splits (splits for which the number of elements in the dimension on which we split is not divisible by num_splits). For RM tensors, adding support for this may be as simple as ensuring that slice handles slices similarly to torch.slice and removing the TT_FATAL at the beginning of the split op checking for divisibility of shape[dim] by num_splits.