hivemind/docs/modules/optim.rst

**hivemind.optim**
==================

.. raw:: html

  This module contains decentralized optimizers that wrap your regular PyTorch Optimizer to train with peers.
  Depending on the exact configuration, Optimizer may perform large synchronous updates equivalent,
  or perform asynchrnous local updates and average model parameters.

  <br><br>

.. automodule:: hivemind.optim.experimental.optimizer
.. currentmodule:: hivemind.optim.experimental.optimizer

**hivemind.Optimizer**
----------------------

.. autoclass:: Optimizer
   :members: step, local_epoch, zero_grad, load_state_from_peers, param_groups, shutdown
   :member-order: bysource

.. currentmodule:: hivemind.optim.grad_scaler
.. autoclass:: GradScaler
   :member-order: bysource


**CollaborativeOptimizer**
--------------------------


.. automodule:: hivemind.optim.collaborative
.. currentmodule:: hivemind.optim

.. autoclass:: CollaborativeOptimizer
   :members: step
   :member-order: bysource

.. autoclass:: CollaborativeAdaptiveOptimizer
   :members:
   :member-order: bysource