Function einsum_native_tensor_reads 

pub fn einsum_native_tensor_reads(
    operands: &[(&NativeTensorReadInput<'_>, &[usize])],
    output_ids: &[usize],
) -> Result<Tensor>

Execute a cached einsum over read-only native tensor inputs.

Backends consume borrowed host views inside their execution session. Mixed dtypes are promoted by Convert nodes in the compiled einsum graph, so non-contiguous operands remain borrowed until runtime execution.

Errors

Returns an error when operands is empty, when a tensor rank does not match its id list length, when a label is out of bounds for the supported range, or when the backend einsum fails.