.. DO NOT EDIT.
.. THIS FILE WAS AUTOMATICALLY GENERATED BY SPHINX-GALLERY.
.. TO MAKE CHANGES, EDIT THE SOURCE PYTHON FILE:
.. "intermediate/pruning_tutorial.py"
.. LINE NUMBERS ARE GIVEN BELOW.
.. only:: html
.. note::
:class: sphx-glr-download-link-note
Click :ref:`here <sphx_glr_download_intermediate_pruning_tutorial.py>`
to download the full example code
.. rst-class:: sphx-glr-example-title
.. _sphx_glr_intermediate_pruning_tutorial.py:
Pruning Tutorial
=====================================
**Author**: `Michela Paganini <https://github.com/mickypaganini>`_
State-of-the-art deep learning techniques rely on over-parametrized models
that are hard to deploy. On the contrary, biological neural networks are
known to use efficient sparse connectivity. Identifying optimal
techniques to compress models by reducing the number of parameters in them is
important in order to reduce memory, battery, and hardware consumption without
sacrificing accuracy. This in turn allows you to deploy lightweight models on device, and guarantee
privacy with private on-device computation. On the research front, pruning is
used to investigate the differences in learning dynamics between
over-parametrized and under-parametrized networks, to study the role of lucky
sparse subnetworks and initializations
("`lottery tickets <https://arxiv.org/abs/1803.03635>`_") as a destructive
neural architecture search technique, and more.
In this tutorial, you will learn how to use ``torch.nn.utils.prune`` to
sparsify your neural networks, and how to extend it to implement your
own custom pruning technique.
Requirements
------------
``"torch>=1.4.0a0+8e8a5e0"``
.. GENERATED FROM PYTHON SOURCE LINES 29-34
.. code-block:: default
import torch
from torch import nn
import torch.nn.utils.prune as prune
import torch.nn.functional as F
.. GENERATED FROM PYTHON SOURCE LINES 35-41
Create a model
--------------
In this tutorial, we use the `LeNet
<http://yann.lecun.com/exdb/publis/pdf/lecun-98.pdf>`_ architecture from
LeCun et al., 1998.
.. GENERATED FROM PYTHON SOURCE LINES 41-66
.. code-block:: default
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
class LeNet(nn.Module):
def __init__(self):
super(LeNet, self).__init__()
# 1 input image channel, 6 output channels, 3x3 square conv kernel
self.conv1 = nn.Conv2d(1, 6, 3)
self.conv2 = nn.Conv2d(6, 16, 3)
self.fc1 = nn.Linear(16 * 5 * 5, 120) # 5x5 image dimension
self.fc2 = nn.Linear(120, 84)
self.fc3 = nn.Linear(84, 10)
def forward(self, x):
x = F.max_pool2d(F.relu(self.conv1(x)), (2, 2))
x = F.max_pool2d(F.relu(self.conv2(x)), 2)
x = x.view(-1, int(x.nelement() / x.shape[0]))
x = F.relu(self.fc1(x))
x = F.relu(self.fc2(x))
x = self.fc3(x)
return x
model = LeNet().to(device=device)
.. GENERATED FROM PYTHON SOURCE LINES 67-72
Inspect a Module
----------------
Let's inspect the (unpruned) ``conv1`` layer in our LeNet model. It will contain two
parameters ``weight`` and ``bias``, and no buffers, for now.
.. GENERATED FROM PYTHON SOURCE LINES 72-75
.. code-block:: default
module = model.conv1
print(list(module.named_parameters()))
.. GENERATED FROM PYTHON SOURCE LINES 76-78
.. code-block:: default
print(list(module.named_buffers()))
.. GENERATED FROM PYTHON SOURCE LINES 79-98
Pruning a Module
----------------
To prune a module (in this example, the ``conv1`` layer of our LeNet
architecture), first select a pruning technique among those available in
``torch.nn.utils.prune`` (or
`implement <#extending-torch-nn-utils-pruning-with-custom-pruning-functions>`_
your own by subclassing
``BasePruningMethod``). Then, specify the module and the name of the parameter to
prune within that module. Finally, using the adequate keyword arguments
required by the selected pruning technique, specify the pruning parameters.
In this example, we will prune at random 30% of the connections in
the parameter named ``weight`` in the ``conv1`` layer.
The module is passed as the first argument to the function; ``name``
identifies the parameter within that module using its string identifier; and
``amount`` indicates either the percentage of connections to prune (if it
is a float between 0. and 1.), or the absolute number of connections to
prune (if it is a non-negative integer).
.. GENERATED FROM PYTHON SOURCE LINES 98-100
.. code-block:: default
prune.random_unstructured(module, name="weight", amount=0.3)
.. GENERATED FROM PYTHON SOURCE LINES 101-105
Pruning acts by removing ``weight`` from the parameters and replacing it with
a new parameter called ``weight_orig`` (i.e. appending ``"_orig"`` to the
initial parameter ``name``). ``weight_orig`` stores the unpruned version of
the tensor. The ``bias`` was not pruned, so it will remain intact.
.. GENERATED FROM PYTHON SOURCE LINES 105-107
.. code-block:: default
print(list(module.named_parameters()))
.. GENERATED FROM PYTHON SOURCE LINES 108-111
The pruning mask generated by the pruning technique selected above is saved
as a module buffer named ``weight_mask`` (i.e. appending ``"_mask"`` to the
initial parameter ``name``).
.. GENERATED FROM PYTHON SOURCE LINES 111-113
.. code-block:: default
print(list(module.named_buffers()))
.. GENERATED FROM PYTHON SOURCE LINES 114-120
For the forward pass to work without modification, the ``weight`` attribute
needs to exist. The pruning techniques implemented in
``torch.nn.utils.prune`` compute the pruned version of the weight (by
combining the mask with the original parameter) and store them in the
attribute ``weight``. Note, this is no longer a parameter of the ``module``,
it is now simply an attribute.
.. GENERATED FROM PYTHON SOURCE LINES 120-122
.. code-block:: default
print(module.weight)
.. GENERATED FROM PYTHON SOURCE LINES 123-129
Finally, pruning is applied prior to each forward pass using PyTorch's
``forward_pre_hooks``. Specifically, when the ``module`` is pruned, as we
have done here, it will acquire a ``forward_pre_hook`` for each parameter
associated with it that gets pruned. In this case, since we have so far
only pruned the original parameter named ``weight``, only one hook will be
present.
.. GENERATED FROM PYTHON SOURCE LINES 129-131
.. code-block:: default
print(module._forward_pre_hooks)
.. GENERATED FROM PYTHON SOURCE LINES 132-137
For completeness, we can now prune the ``bias`` too, to see how the
parameters, buffers, hooks, and attributes of the ``module`` change.
Just for the sake of trying out another pruning technique, here we prune the
3 smallest entries in the bias by L1 norm, as implemented in the
``l1_unstructured`` pruning function.
.. GENERATED FROM PYTHON SOURCE LINES 137-139
.. code-block:: default
prune.l1_unstructured(module, name="bias", amount=3)
.. GENERATED FROM PYTHON SOURCE LINES 140-144
We now expect the named parameters to include both ``weight_orig`` (from
before) and ``bias_orig``. The buffers will include ``weight_mask`` and
``bias_mask``. The pruned versions of the two tensors will exist as
module attributes, and the module will now have two ``forward_pre_hooks``.
.. GENERATED FROM PYTHON SOURCE LINES 144-146
.. code-block:: default
print(list(module.named_parameters()))
.. GENERATED FROM PYTHON SOURCE LINES 147-149
.. code-block:: default
print(list(module.named_buffers()))
.. GENERATED FROM PYTHON SOURCE LINES 150-152
.. code-block:: default
print(module.bias)
.. GENERATED FROM PYTHON SOURCE LINES 153-155
.. code-block:: default
print(module._forward_pre_hooks)
.. GENERATED FROM PYTHON SOURCE LINES 156-170
Iterative Pruning
-----------------
The same parameter in a module can be pruned multiple times, with the
effect of the various pruning calls being equal to the combination of the
various masks applied in series.
The combination of a new mask with the old mask is handled by the
``PruningContainer``'s ``compute_mask`` method.
Say, for example, that we now want to further prune ``module.weight``, this
time using structured pruning along the 0th axis of the tensor (the 0th axis
corresponds to the output channels of the convolutional layer and has
dimensionality 6 for ``conv1``), based on the channels' L2 norm. This can be
achieved using the ``ln_structured`` function, with ``n=2`` and ``dim=0``.
.. GENERATED FROM PYTHON SOURCE LINES 170-177
.. code-block:: default
prune.ln_structured(module, name="weight", amount=0.5, n=2, dim=0)
# As we can verify, this will zero out all the connections corresponding to
# 50% (3 out of 6) of the channels, while preserving the action of the
# previous mask.
print(module.weight)
.. GENERATED FROM PYTHON SOURCE LINES 178-181
The corresponding hook will now be of type
``torch.nn.utils.prune.PruningContainer``, and will store the history of
pruning applied to the ``weight`` parameter.
.. GENERATED FROM PYTHON SOURCE LINES 181-187
.. code-block:: default
for hook in module._forward_pre_hooks.values():
if hook._tensor_name == "weight": # select out the correct hook
break
print(list(hook)) # pruning history in the container
.. GENERATED FROM PYTHON SOURCE LINES 188-193
Serializing a pruned model
--------------------------
All relevant tensors, including the mask buffers and the original parameters
used to compute the pruned tensors are stored in the model's ``state_dict``
and can therefore be easily serialized and saved, if needed.
.. GENERATED FROM PYTHON SOURCE LINES 193-196
.. code-block:: default
print(model.state_dict().keys())
.. GENERATED FROM PYTHON SOURCE LINES 197-206
Remove pruning re-parametrization
---------------------------------
To make the pruning permanent, remove the re-parametrization in terms
of ``weight_orig`` and ``weight_mask``, and remove the ``forward_pre_hook``,
we can use the ``remove`` functionality from ``torch.nn.utils.prune``.
Note that this doesn't undo the pruning, as if it never happened. It simply
makes it permanent, instead, by reassigning the parameter ``weight`` to the
model parameters, in its pruned version.
.. GENERATED FROM PYTHON SOURCE LINES 208-209
Prior to removing the re-parametrization:
.. GENERATED FROM PYTHON SOURCE LINES 209-210
.. code-block:: default
print(list(module.named_parameters()))
.. GENERATED FROM PYTHON SOURCE LINES 211-212
.. code-block:: default
print(list(module.named_buffers()))
.. GENERATED FROM PYTHON SOURCE LINES 213-215
.. code-block:: default
print(module.weight)
.. GENERATED FROM PYTHON SOURCE LINES 216-217
After removing the re-parametrization:
.. GENERATED FROM PYTHON SOURCE LINES 217-219
.. code-block:: default
prune.remove(module, 'weight')
print(list(module.named_parameters()))
.. GENERATED FROM PYTHON SOURCE LINES 220-222
.. code-block:: default
print(list(module.named_buffers()))
.. GENERATED FROM PYTHON SOURCE LINES 223-229
Pruning multiple parameters in a model
--------------------------------------
By specifying the desired pruning technique and parameters, we can easily
prune multiple tensors in a network, perhaps according to their type, as we
will see in this example.
.. GENERATED FROM PYTHON SOURCE LINES 229-241
.. code-block:: default
new_model = LeNet()
for name, module in new_model.named_modules():
# prune 20% of connections in all 2D-conv layers
if isinstance(module, torch.nn.Conv2d):
prune.l1_unstructured(module, name='weight', amount=0.2)
# prune 40% of connections in all linear layers
elif isinstance(module, torch.nn.Linear):
prune.l1_unstructured(module, name='weight', amount=0.4)
print(dict(new_model.named_buffers()).keys()) # to verify that all masks exist
.. GENERATED FROM PYTHON SOURCE LINES 242-255
Global pruning
--------------
So far, we only looked at what is usually referred to as "local" pruning,
i.e. the practice of pruning tensors in a model one by one, by
comparing the statistics (weight magnitude, activation, gradient, etc.) of
each entry exclusively to the other entries in that tensor. However, a
common and perhaps more powerful technique is to prune the model all at
once, by removing (for example) the lowest 20% of connections across the
whole model, instead of removing the lowest 20% of connections in each
layer. This is likely to result in different pruning percentages per layer.
Let's see how to do that using ``global_unstructured`` from
``torch.nn.utils.prune``.
.. GENERATED FROM PYTHON SOURCE LINES 255-272
.. code-block:: default
model = LeNet()
parameters_to_prune = (
(model.conv1, 'weight'),
(model.conv2, 'weight'),
(model.fc1, 'weight'),
(model.fc2, 'weight'),
(model.fc3, 'weight'),
)
prune.global_unstructured(
parameters_to_prune,
pruning_method=prune.L1Unstructured,
amount=0.2,
)
.. GENERATED FROM PYTHON SOURCE LINES 273-276
Now we can check the sparsity induced in every pruned parameter, which will
not be equal to 20% in each layer. However, the global sparsity will be
(approximately) 20%.
.. GENERATED FROM PYTHON SOURCE LINES 276-326
.. code-block:: default
print(
"Sparsity in conv1.weight: {:.2f}%".format(
100. * float(torch.sum(model.conv1.weight == 0))
/ float(model.conv1.weight.nelement())
)
)
print(
"Sparsity in conv2.weight: {:.2f}%".format(
100. * float(torch.sum(model.conv2.weight == 0))
/ float(model.conv2.weight.nelement())
)
)
print(
"Sparsity in fc1.weight: {:.2f}%".format(
100. * float(torch.sum(model.fc1.weight == 0))
/ float(model.fc1.weight.nelement())
)
)
print(
"Sparsity in fc2.weight: {:.2f}%".format(
100. * float(torch.sum(model.fc2.weight == 0))
/ float(model.fc2.weight.nelement())
)
)
print(
"Sparsity in fc3.weight: {:.2f}%".format(
100. * float(torch.sum(model.fc3.weight == 0))
/ float(model.fc3.weight.nelement())
)
)
print(
"Global sparsity: {:.2f}%".format(
100. * float(
torch.sum(model.conv1.weight == 0)
+ torch.sum(model.conv2.weight == 0)
+ torch.sum(model.fc1.weight == 0)
+ torch.sum(model.fc2.weight == 0)
+ torch.sum(model.fc3.weight == 0)
)
/ float(
model.conv1.weight.nelement()
+ model.conv2.weight.nelement()
+ model.fc1.weight.nelement()
+ model.fc2.weight.nelement()
+ model.fc3.weight.nelement()
)
)
)
.. GENERATED FROM PYTHON SOURCE LINES 327-356
Extending ``torch.nn.utils.prune`` with custom pruning functions
------------------------------------------------------------------
To implement your own pruning function, you can extend the
``nn.utils.prune`` module by subclassing the ``BasePruningMethod``
base class, the same way all other pruning methods do. The base class
implements the following methods for you: ``__call__``, ``apply_mask``,
``apply``, ``prune``, and ``remove``. Beyond some special cases, you shouldn't
have to reimplement these methods for your new pruning technique.
You will, however, have to implement ``__init__`` (the constructor),
and ``compute_mask`` (the instructions on how to compute the mask
for the given tensor according to the logic of your pruning
technique). In addition, you will have to specify which type of
pruning this technique implements (supported options are ``global``,
``structured``, and ``unstructured``). This is needed to determine
how to combine masks in the case in which pruning is applied
iteratively. In other words, when pruning a pre-pruned parameter,
the current prunining techique is expected to act on the unpruned
portion of the parameter. Specifying the ``PRUNING_TYPE`` will
enable the ``PruningContainer`` (which handles the iterative
application of pruning masks) to correctly identify the slice of the
parameter to prune.
Let's assume, for example, that you want to implement a pruning
technique that prunes every other entry in a tensor (or -- if the
tensor has previously been pruned -- in the remaining unpruned
portion of the tensor). This will be of ``PRUNING_TYPE='unstructured'``
because it acts on individual connections in a layer and not on entire
units/channels (``'structured'``), or across different parameters
(``'global'``).
.. GENERATED FROM PYTHON SOURCE LINES 356-367
.. code-block:: default
class FooBarPruningMethod(prune.BasePruningMethod):
"""Prune every other entry in a tensor
"""
PRUNING_TYPE = 'unstructured'
def compute_mask(self, t, default_mask):
mask = default_mask.clone()
mask.view(-1)[::2] = 0
return mask
.. GENERATED FROM PYTHON SOURCE LINES 368-371
Now, to apply this to a parameter in an ``nn.Module``, you should
also provide a simple function that instantiates the method and
applies it.
.. GENERATED FROM PYTHON SOURCE LINES 371-398
.. code-block:: default
def foobar_unstructured(module, name):
"""Prunes tensor corresponding to parameter called `name` in `module`
by removing every other entry in the tensors.
Modifies module in place (and also return the modified module)
by:
1) adding a named buffer called `name+'_mask'` corresponding to the
binary mask applied to the parameter `name` by the pruning method.
The parameter `name` is replaced by its pruned version, while the
original (unpruned) parameter is stored in a new parameter named
`name+'_orig'`.
Args:
module (nn.Module): module containing the tensor to prune
name (string): parameter name within `module` on which pruning
will act.
Returns:
module (nn.Module): modified (i.e. pruned) version of the input
module
Examples:
>>> m = nn.Linear(3, 4)
>>> foobar_unstructured(m, name='bias')
"""
FooBarPruningMethod.apply(module, name)
return module
.. GENERATED FROM PYTHON SOURCE LINES 399-400
Let's try it out!
.. GENERATED FROM PYTHON SOURCE LINES 400-404
.. code-block:: default
model = LeNet()
foobar_unstructured(model.fc3, name='bias')
print(model.fc3.bias_mask)
.. rst-class:: sphx-glr-timing
**Total running time of the script:** ( 0 minutes 0.000 seconds)
.. _sphx_glr_download_intermediate_pruning_tutorial.py:
.. only:: html
.. container:: sphx-glr-footer sphx-glr-footer-example
.. container:: sphx-glr-download sphx-glr-download-python
:download:`Download Python source code: pruning_tutorial.py <pruning_tutorial.py>`
.. container:: sphx-glr-download sphx-glr-download-jupyter
:download:`Download Jupyter notebook: pruning_tutorial.ipynb <pruning_tutorial.ipynb>`
.. only:: html
.. rst-class:: sphx-glr-signature
`Gallery generated by Sphinx-Gallery <https://sphinx-gallery.github.io>`_