Shortcuts

Source code for pytorch_lightning.plugins.layer_sync

# Copyright The PyTorch Lightning team.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

from abc import ABC, abstractmethod

import torch
from torch import Tensor
from torch.nn import Module


[docs]class LayerSync(ABC): """Abstract base class for creating plugins that wrap layers of a model with synchronization logic for multiprocessing."""
[docs] @abstractmethod def apply(self, model: Module) -> Module: """Override this method to apply synchronization to the layers of this model."""
[docs] @abstractmethod def revert(self, model: Module) -> Module: """Override this method to undo all modifications made in :meth:`apply`."""
[docs]class NativeSyncBatchNorm(LayerSync): """A plugin that wraps all batch normalization layers of a model with synchronization logic for multiprocessing. This plugin has no effect in single-device operation. """
[docs] def apply(self, model: Module) -> Module: """Add global batchnorm for a model spread across multiple GPUs and nodes. Override this method to synchronize batchnorm layers between specific process groups instead of the whole world. Args: model: Reference to the current LightningModule Return: LightningModule with batchnorm layers synchronized within the process groups. """ return torch.nn.SyncBatchNorm.convert_sync_batchnorm(model)
[docs] def revert(self, model: Module) -> Module: """Convert the wrapped batchnorm layers back to regular batchnorm layers. Args: model: Reference to the current LightningModule Return: LightningModule with regular batchnorm layers that will no longer sync across processes. """ # Code adapted from https://github.com/pytorch/pytorch/issues/41081#issuecomment-783961547 # Original author: Kapil Yedidi (@kapily) converted_module = model if isinstance(model, torch.nn.modules.batchnorm.SyncBatchNorm): # Unfortunately, LayerSync does not store the original class - if it did # we could return the one that was originally created. converted_module = _BatchNormXd( model.num_features, model.eps, model.momentum, model.affine, model.track_running_stats ) if model.affine: with torch.no_grad(): converted_module.weight = model.weight converted_module.bias = model.bias converted_module.running_mean = model.running_mean converted_module.running_var = model.running_var converted_module.num_batches_tracked = model.num_batches_tracked if hasattr(model, "qconfig"): converted_module.qconfig = model.qconfig for name, child in model.named_children(): converted_module.add_module(name, self.revert(child)) del model return converted_module
class _BatchNormXd(torch.nn.modules.batchnorm._BatchNorm): def _check_input_dim(self, input: Tensor) -> None: # The only difference between BatchNorm1d, BatchNorm2d, BatchNorm3d, etc # is this method that is overwritten by the subclass. # Here, we are bypassing some tensor sanity checks and trusting that the user # provides the right input dimensions at inference. return

© Copyright Copyright (c) 2018-2022, Lightning AI et al... Revision 92fe1887.

Built with Sphinx using a theme provided by Read the Docs.
Read the Docs v: stable
Versions
latest
stable
1.8.3post1
1.8.3.post0
1.8.3
1.8.2
1.8.1
1.8.0.post1
1.8.0
1.7.7
1.7.6
1.7.5
1.7.4
1.7.3
1.7.2
1.7.1
1.7.0
1.6.5
1.6.4
1.6.3
1.6.2
1.6.1
1.6.0
1.5.10
1.5.9
1.5.8
1.5.7
1.5.6
1.5.5
1.5.4
1.5.3
1.5.2
1.5.1
1.5.0
1.4.9
1.4.8
1.4.7
1.4.6
1.4.5
1.4.4
1.4.3
1.4.2
1.4.1
1.4.0
1.3.8
1.3.7
1.3.6
1.3.5
1.3.4
1.3.3
1.3.2
1.3.1
1.3.0
1.2.10
1.2.8
1.2.7
1.2.6
1.2.5
1.2.4
1.2.3
1.2.2
1.2.1
1.2.0
1.1.8
1.1.7
1.1.6
1.1.5
1.1.4
1.1.3
1.1.2
1.1.1
1.1.0
1.0.8
1.0.7
1.0.6
1.0.5
1.0.4
1.0.3
1.0.2
1.0.1
1.0.0
0.10.0
0.9.0
0.8.5
0.8.4
0.8.3
0.8.2
0.8.1
0.8.0
0.7.6
0.7.5
0.7.4
0.7.3
0.7.2
0.7.1
0.7.0
0.6.0
0.5.3
0.4.9
Downloads
html
On Read the Docs
Project Home
Builds

Free document hosting provided by Read the Docs.