In torch-geometry version 2.0.2, which is the version automatically installed with pip 21.1.3, add_remaining_self_loops accepts fill_value as one of float, tensor or str.
Although in latest torch-geometric package, int value is accepted and treated the same way as float, it could be more robust to set fill_value as float.
In torch-geometry version 2.0.2, which is the version automatically installed with pip 21.1.3,
add_remaining_self_loopsacceptsfill_valueas one of float, tensor or str. Although in latest torch-geometric package, int value is accepted and treated the same way as float, it could be more robust to set fill_value as float.