GitLab

Explicitly specify quantized tensor usages needed for linear op backward
Signed-off-by: Tim Moon <tmoon@nvidia.com>