Open drhagen opened 10 months ago
We should probably remove output_format
entirely and put it into the **kwargs
that currently captures all the intputs. The output would be specified by name instead of being the positional argument. It would be typically be optional, defaulting to "all dense". It could also be a Tensor
, in which case the output format and dimensions would match the Tensor
provided. The dimensions would be checked against the matching dimensions of the input Tensor
s. If there was a broadcast dimension, this argument would not be optional and it must be a Tensor
which would provide the necessary size.
The
output_format
parameter should be optional inevaluate
and default to all dense. This is a common enough choice and we default to all dense other places.