.. DO NOT EDIT.
.. THIS FILE WAS AUTOMATICALLY GENERATED BY SPHINX-GALLERY.
.. TO MAKE CHANGES, EDIT THE SOURCE PYTHON FILE:
.. "intermediate/per_sample_grads.py"
.. LINE NUMBERS ARE GIVEN BELOW.
.. only:: html
.. note::
:class: sphx-glr-download-link-note
Click :ref:`here <sphx_glr_download_intermediate_per_sample_grads.py>`
to download the full example code
.. rst-class:: sphx-glr-example-title
.. _sphx_glr_intermediate_per_sample_grads.py:
Per-sample-gradients
====================
What is it?
-----------
Per-sample-gradient computation is computing the gradient for each and every
sample in a batch of data. It is a useful quantity in differential privacy,
meta-learning, and optimization research.
.. note::
This tutorial requires PyTorch 2.0.0 or later.
.. GENERATED FROM PYTHON SOURCE LINES 18-52
.. code-block:: default
import torch
import torch.nn as nn
import torch.nn.functional as F
torch.manual_seed(0)
# Here's a simple CNN and loss function:
class SimpleCNN(nn.Module):
def __init__(self):
super(SimpleCNN, self).__init__()
self.conv1 = nn.Conv2d(1, 32, 3, 1)
self.conv2 = nn.Conv2d(32, 64, 3, 1)
self.fc1 = nn.Linear(9216, 128)
self.fc2 = nn.Linear(128, 10)
def forward(self, x):
x = self.conv1(x)
x = F.relu(x)
x = self.conv2(x)
x = F.relu(x)
x = F.max_pool2d(x, 2)
x = torch.flatten(x, 1)
x = self.fc1(x)
x = F.relu(x)
x = self.fc2(x)
output = F.log_softmax(x, dim=1)
output = x
return output
def loss_fn(predictions, targets):
return F.nll_loss(predictions, targets)
.. GENERATED FROM PYTHON SOURCE LINES 53-55
Let’s generate a batch of dummy data and pretend that we’re working with an MNIST dataset.
The dummy images are 28 by 28 and we use a minibatch of size 64.
.. GENERATED FROM PYTHON SOURCE LINES 55-64
.. code-block:: default
device = 'cuda'
num_models = 10
batch_size = 64
data = torch.randn(batch_size, 1, 28, 28, device=device)
targets = torch.randint(10, (64,), device=device)
.. GENERATED FROM PYTHON SOURCE LINES 65-68
In regular model training, one would forward the minibatch through the model,
and then call .backward() to compute gradients. This would generate an
'average' gradient of the entire mini-batch:
.. GENERATED FROM PYTHON SOURCE LINES 68-75
.. code-block:: default
model = SimpleCNN().to(device=device)
predictions = model(data) # move the entire mini-batch through the model
loss = loss_fn(predictions, targets)
loss.backward() # back propagate the 'average' gradient of this mini-batch
.. GENERATED FROM PYTHON SOURCE LINES 76-81
In contrast to the above approach, per-sample-gradient computation is
equivalent to:
- for each individual sample of the data, perform a forward and a backward
pass to get an individual (per-sample) gradient.
.. GENERATED FROM PYTHON SOURCE LINES 81-101
.. code-block:: default
def compute_grad(sample, target):
sample = sample.unsqueeze(0) # prepend batch dimension for processing
target = target.unsqueeze(0)
prediction = model(sample)
loss = loss_fn(prediction, target)
return torch.autograd.grad(loss, list(model.parameters()))
def compute_sample_grads(data, targets):
""" manually process each sample with per sample gradient """
sample_grads = [compute_grad(data[i], targets[i]) for i in range(batch_size)]
sample_grads = zip(*sample_grads)
sample_grads = [torch.stack(shards) for shards in sample_grads]
return sample_grads
per_sample_grads = compute_sample_grads(data, targets)
.. GENERATED FROM PYTHON SOURCE LINES 102-105
``sample_grads[0]`` is the per-sample-grad for model.conv1.weight.
``model.conv1.weight.shape`` is ``[32, 1, 3, 3]``; notice how there is one
gradient, per sample, in the batch for a total of 64.
.. GENERATED FROM PYTHON SOURCE LINES 105-108
.. code-block:: default
print(per_sample_grads[0].shape)
.. GENERATED FROM PYTHON SOURCE LINES 109-123
Per-sample-grads, *the efficient way*, using function transforms
----------------------------------------------------------------
We can compute per-sample-gradients efficiently by using function transforms.
The ``torch.func`` function transform API transforms over functions.
Our strategy is to define a function that computes the loss and then apply
transforms to construct a function that computes per-sample-gradients.
We'll use the ``torch.func.functional_call`` function to treat an ``nn.Module``
like a function.
First, let’s extract the state from ``model`` into two dictionaries,
parameters and buffers. We'll be detaching them because we won't use
regular PyTorch autograd (e.g. Tensor.backward(), torch.autograd.grad).
.. GENERATED FROM PYTHON SOURCE LINES 123-129
.. code-block:: default
from torch.func import functional_call, vmap, grad
params = {k: v.detach() for k, v in model.named_parameters()}
buffers = {k: v.detach() for k, v in model.named_buffers()}
.. GENERATED FROM PYTHON SOURCE LINES 130-137
Next, let's define a function to compute the loss of the model given a
single input rather than a batch of inputs. It is important that this
function accepts the parameters, the input, and the target, because we will
be transforming over them.
Note - because the model was originally written to handle batches, we’ll
use ``torch.unsqueeze`` to add a batch dimension.
.. GENERATED FROM PYTHON SOURCE LINES 137-146
.. code-block:: default
def compute_loss(params, buffers, sample, target):
batch = sample.unsqueeze(0)
targets = target.unsqueeze(0)
predictions = functional_call(model, (params, buffers), (batch,))
loss = loss_fn(predictions, targets)
return loss
.. GENERATED FROM PYTHON SOURCE LINES 147-150
Now, let’s use the ``grad`` transform to create a new function that computes
the gradient with respect to the first argument of ``compute_loss``
(i.e. the ``params``).
.. GENERATED FROM PYTHON SOURCE LINES 150-153
.. code-block:: default
ft_compute_grad = grad(compute_loss)
.. GENERATED FROM PYTHON SOURCE LINES 154-160
The ``ft_compute_grad`` function computes the gradient for a single
(sample, target) pair. We can use ``vmap`` to get it to compute the gradient
over an entire batch of samples and targets. Note that
``in_dims=(None, None, 0, 0)`` because we wish to map ``ft_compute_grad`` over
the 0th dimension of the data and targets, and use the same ``params`` and
buffers for each.
.. GENERATED FROM PYTHON SOURCE LINES 160-163
.. code-block:: default
ft_compute_sample_grad = vmap(ft_compute_grad, in_dims=(None, None, 0, 0))
.. GENERATED FROM PYTHON SOURCE LINES 164-165
Finally, let's used our transformed function to compute per-sample-gradients:
.. GENERATED FROM PYTHON SOURCE LINES 165-168
.. code-block:: default
ft_per_sample_grads = ft_compute_sample_grad(params, buffers, data, targets)
.. GENERATED FROM PYTHON SOURCE LINES 169-171
we can double check that the results using ``grad`` and ``vmap`` match the
results of hand processing each one individually:
.. GENERATED FROM PYTHON SOURCE LINES 171-175
.. code-block:: default
for per_sample_grad, ft_per_sample_grad in zip(per_sample_grads, ft_per_sample_grads.values()):
assert torch.allclose(per_sample_grad, ft_per_sample_grad, atol=3e-3, rtol=1e-5)
.. GENERATED FROM PYTHON SOURCE LINES 176-191
A quick note: there are limitations around what types of functions can be
transformed by ``vmap``. The best functions to transform are ones that are pure
functions: a function where the outputs are only determined by the inputs,
and that have no side effects (e.g. mutation). ``vmap`` is unable to handle
mutation of arbitrary Python data structures, but it is able to handle many
in-place PyTorch operations.
Performance comparison
----------------------
Curious about how the performance of ``vmap`` compares?
Currently the best results are obtained on newer GPU's such as the A100
(Ampere) where we've seen up to 25x speedups on this example, but here are
some results on our build machines:
.. GENERATED FROM PYTHON SOURCE LINES 191-215
.. code-block:: default
def get_perf(first, first_descriptor, second, second_descriptor):
"""takes torch.benchmark objects and compares delta of second vs first."""
second_res = second.times[0]
first_res = first.times[0]
gain = (first_res-second_res)/first_res
if gain < 0: gain *=-1
final_gain = gain*100
print(f"Performance delta: {final_gain:.4f} percent improvement with {first_descriptor} ")
from torch.utils.benchmark import Timer
without_vmap = Timer(stmt="compute_sample_grads(data, targets)", globals=globals())
with_vmap = Timer(stmt="ft_compute_sample_grad(params, buffers, data, targets)",globals=globals())
no_vmap_timing = without_vmap.timeit(100)
with_vmap_timing = with_vmap.timeit(100)
print(f'Per-sample-grads without vmap {no_vmap_timing}')
print(f'Per-sample-grads with vmap {with_vmap_timing}')
get_perf(with_vmap_timing, "vmap", no_vmap_timing, "no vmap")
.. GENERATED FROM PYTHON SOURCE LINES 216-227
There are other optimized solutions (like in https://github.com/pytorch/opacus)
to computing per-sample-gradients in PyTorch that also perform better than
the naive method. But it’s cool that composing ``vmap`` and ``grad`` give us a
nice speedup.
In general, vectorization with ``vmap`` should be faster than running a function
in a for-loop and competitive with manual batching. There are some exceptions
though, like if we haven’t implemented the ``vmap`` rule for a particular
operation or if the underlying kernels weren’t optimized for older hardware
(GPUs). If you see any of these cases, please let us know by opening an issue
at on GitHub.
.. rst-class:: sphx-glr-timing
**Total running time of the script:** ( 0 minutes 0.000 seconds)
.. _sphx_glr_download_intermediate_per_sample_grads.py:
.. only:: html
.. container:: sphx-glr-footer sphx-glr-footer-example
.. container:: sphx-glr-download sphx-glr-download-python
:download:`Download Python source code: per_sample_grads.py <per_sample_grads.py>`
.. container:: sphx-glr-download sphx-glr-download-jupyter
:download:`Download Jupyter notebook: per_sample_grads.ipynb <per_sample_grads.ipynb>`
.. only:: html
.. rst-class:: sphx-glr-signature
`Gallery generated by Sphinx-Gallery <https://sphinx-gallery.github.io>`_