.state()

Return the current optimizer state.

Returns:

• dict –

  Dict with step_count and a param_groups list, each entry

• dict –

  containing the group label and its current learning rate.

• dict –

  Subclasses may override this to include additional state.