Note
Click here to download the full example code
(beta) Building a Simple CPU Performance Profiler with FX¶
Author: James Reed
In this tutorial, we are going to use FX to do the following:
- Capture PyTorch Python code in a way that we can inspect and gather statistics about the structure and execution of the code
- Build out a small class that will serve as a simple performance “profiler”, collecting runtime statistics about each part of the model from actual runs.
For this tutorial, we are going to use the torchvision ResNet18 model for demonstration purposes.
import torch
import torch.fx
import torchvision.models as models
rn18 = models.resnet18()
rn18.eval()
Now that we have our model, we want to inspect deeper into its performance. That is, for the following invocation, which parts of the model are taking the longest?
input = torch.randn(5, 3, 224, 224)
output = rn18(input)
A common way of answering that question is to go through the program source, add code that collects timestamps at various points in the program, and compare the difference between those timestamps to see how long the regions between the timestamps take.
That technique is certainly applicable to PyTorch code, however it would be nicer if we didn’t have to copy over model code and edit it, especially code we haven’t written (like this torchvision model). Instead, we are going to use FX to automate this “instrumentation” process without needing to modify any source.
First, let’s get some imports out of the way (we will be using all of these later in the code).
import statistics, tabulate, time
from typing import Any, Dict, List
from torch.fx import Interpreter
Note
tabulate is an external library that is not a dependency of PyTorch.
We will be using it to more easily visualize performance data. Please
make sure you’ve installed it from your favorite Python package source.
Capturing the Model with Symbolic Tracing¶
Next, we are going to use FX’s symbolic tracing mechanism to capture the definition of our model in a data structure we can manipulate and examine.
traced_rn18 = torch.fx.symbolic_trace(rn18)
print(traced_rn18.graph)
Out:
graph(xx: torch.Tensor) -> torch.Tensor:
%conv1 : [#users=1] = call_module[target=conv1](args = (%x,), kwargs = {})
%bn1 : [#users=1] = call_module[target=bn1](args = (%conv1,), kwargs = {})
%relu_1 : [#users=1] = call_module[target=relu](args = (%bn1,), kwargs = {})
%maxpool : [#users=2] = call_module[target=maxpool](args = (%relu_1,), kwargs = {})
%layer1_0_conv1 : [#users=1] = call_module[target=layer1.0.conv1](args = (%maxpool,), kwargs = {})
%layer1_0_bn1 : [#users=1] = call_module[target=layer1.0.bn1](args = (%layer1_0_conv1,), kwargs = {})
%layer1_0_relu : [#users=1] = call_module[target=layer1.0.relu](args = (%layer1_0_bn1,), kwargs = {})
%layer1_0_conv2 : [#users=1] = call_module[target=layer1.0.conv2](args = (%layer1_0_relu,), kwargs = {})
%layer1_0_bn2 : [#users=1] = call_module[target=layer1.0.bn2](args = (%layer1_0_conv2,), kwargs = {})
%add_1 : [#users=1] = call_function[target=operator.add](args = (%layer1_0_bn2, %maxpool), kwargs = {})
%layer1_0_relu_1 : [#users=2] = call_module[target=layer1.0.relu](args = (%add_1,), kwargs = {})
%layer1_1_conv1 : [#users=1] = call_module[target=layer1.1.conv1](args = (%layer1_0_relu_1,), kwargs = {})
%layer1_1_bn1 : [#users=1] = call_module[target=layer1.1.bn1](args = (%layer1_1_conv1,), kwargs = {})
%layer1_1_relu : [#users=1] = call_module[target=layer1.1.relu](args = (%layer1_1_bn1,), kwargs = {})
%layer1_1_conv2 : [#users=1] = call_module[target=layer1.1.conv2](args = (%layer1_1_relu,), kwargs = {})
%layer1_1_bn2 : [#users=1] = call_module[target=layer1.1.bn2](args = (%layer1_1_conv2,), kwargs = {})
%add_2 : [#users=1] = call_function[target=operator.add](args = (%layer1_1_bn2, %layer1_0_relu_1), kwargs = {})
%layer1_1_relu_1 : [#users=2] = call_module[target=layer1.1.relu](args = (%add_2,), kwargs = {})
%layer2_0_conv1 : [#users=1] = call_module[target=layer2.0.conv1](args = (%layer1_1_relu_1,), kwargs = {})
%layer2_0_bn1 : [#users=1] = call_module[target=layer2.0.bn1](args = (%layer2_0_conv1,), kwargs = {})
%layer2_0_relu : [#users=1] = call_module[target=layer2.0.relu](args = (%layer2_0_bn1,), kwargs = {})
%layer2_0_conv2 : [#users=1] = call_module[target=layer2.0.conv2](args = (%layer2_0_relu,), kwargs = {})
%layer2_0_bn2 : [#users=1] = call_module[target=layer2.0.bn2](args = (%layer2_0_conv2,), kwargs = {})
%layer2_0_downsample_0 : [#users=1] = call_module[target=layer2.0.downsample.0](args = (%layer1_1_relu_1,), kwargs = {})
%layer2_0_downsample_1 : [#users=1] = call_module[target=layer2.0.downsample.1](args = (%layer2_0_downsample_0,), kwargs = {})
%add_3 : [#users=1] = call_function[target=operator.add](args = (%layer2_0_bn2, %layer2_0_downsample_1), kwargs = {})
%layer2_0_relu_1 : [#users=2] = call_module[target=layer2.0.relu](args = (%add_3,), kwargs = {})
%layer2_1_conv1 : [#users=1] = call_module[target=layer2.1.conv1](args = (%layer2_0_relu_1,), kwargs = {})
%layer2_1_bn1 : [#users=1] = call_module[target=layer2.1.bn1](args = (%layer2_1_conv1,), kwargs = {})
%layer2_1_relu : [#users=1] = call_module[target=layer2.1.relu](args = (%layer2_1_bn1,), kwargs = {})
%layer2_1_conv2 : [#users=1] = call_module[target=layer2.1.conv2](args = (%layer2_1_relu,), kwargs = {})
%layer2_1_bn2 : [#users=1] = call_module[target=layer2.1.bn2](args = (%layer2_1_conv2,), kwargs = {})
%add_4 : [#users=1] = call_function[target=operator.add](args = (%layer2_1_bn2, %layer2_0_relu_1), kwargs = {})
%layer2_1_relu_1 : [#users=2] = call_module[target=layer2.1.relu](args = (%add_4,), kwargs = {})
%layer3_0_conv1 : [#users=1] = call_module[target=layer3.0.conv1](args = (%layer2_1_relu_1,), kwargs = {})
%layer3_0_bn1 : [#users=1] = call_module[target=layer3.0.bn1](args = (%layer3_0_conv1,), kwargs = {})
%layer3_0_relu : [#users=1] = call_module[target=layer3.0.relu](args = (%layer3_0_bn1,), kwargs = {})
%layer3_0_conv2 : [#users=1] = call_module[target=layer3.0.conv2](args = (%layer3_0_relu,), kwargs = {})
%layer3_0_bn2 : [#users=1] = call_module[target=layer3.0.bn2](args = (%layer3_0_conv2,), kwargs = {})
%layer3_0_downsample_0 : [#users=1] = call_module[target=layer3.0.downsample.0](args = (%layer2_1_relu_1,), kwargs = {})
%layer3_0_downsample_1 : [#users=1] = call_module[target=layer3.0.downsample.1](args = (%layer3_0_downsample_0,), kwargs = {})
%add_5 : [#users=1] = call_function[target=operator.add](args = (%layer3_0_bn2, %layer3_0_downsample_1), kwargs = {})
%layer3_0_relu_1 : [#users=2] = call_module[target=layer3.0.relu](args = (%add_5,), kwargs = {})
%layer3_1_conv1 : [#users=1] = call_module[target=layer3.1.conv1](args = (%layer3_0_relu_1,), kwargs = {})
%layer3_1_bn1 : [#users=1] = call_module[target=layer3.1.bn1](args = (%layer3_1_conv1,), kwargs = {})
%layer3_1_relu : [#users=1] = call_module[target=layer3.1.relu](args = (%layer3_1_bn1,), kwargs = {})
%layer3_1_conv2 : [#users=1] = call_module[target=layer3.1.conv2](args = (%layer3_1_relu,), kwargs = {})
%layer3_1_bn2 : [#users=1] = call_module[target=layer3.1.bn2](args = (%layer3_1_conv2,), kwargs = {})
%add_6 : [#users=1] = call_function[target=operator.add](args = (%layer3_1_bn2, %layer3_0_relu_1), kwargs = {})
%layer3_1_relu_1 : [#users=2] = call_module[target=layer3.1.relu](args = (%add_6,), kwargs = {})
%layer4_0_conv1 : [#users=1] = call_module[target=layer4.0.conv1](args = (%layer3_1_relu_1,), kwargs = {})
%layer4_0_bn1 : [#users=1] = call_module[target=layer4.0.bn1](args = (%layer4_0_conv1,), kwargs = {})
%layer4_0_relu : [#users=1] = call_module[target=layer4.0.relu](args = (%layer4_0_bn1,), kwargs = {})
%layer4_0_conv2 : [#users=1] = call_module[target=layer4.0.conv2](args = (%layer4_0_relu,), kwargs = {})
%layer4_0_bn2 : [#users=1] = call_module[target=layer4.0.bn2](args = (%layer4_0_conv2,), kwargs = {})
%layer4_0_downsample_0 : [#users=1] = call_module[target=layer4.0.downsample.0](args = (%layer3_1_relu_1,), kwargs = {})
%layer4_0_downsample_1 : [#users=1] = call_module[target=layer4.0.downsample.1](args = (%layer4_0_downsample_0,), kwargs = {})
%add_7 : [#users=1] = call_function[target=operator.add](args = (%layer4_0_bn2, %layer4_0_downsample_1), kwargs = {})
%layer4_0_relu_1 : [#users=2] = call_module[target=layer4.0.relu](args = (%add_7,), kwargs = {})
%layer4_1_conv1 : [#users=1] = call_module[target=layer4.1.conv1](args = (%layer4_0_relu_1,), kwargs = {})
%layer4_1_bn1 : [#users=1] = call_module[target=layer4.1.bn1](args = (%layer4_1_conv1,), kwargs = {})
%layer4_1_relu : [#users=1] = call_module[target=layer4.1.relu](args = (%layer4_1_bn1,), kwargs = {})
%layer4_1_conv2 : [#users=1] = call_module[target=layer4.1.conv2](args = (%layer4_1_relu,), kwargs = {})
%layer4_1_bn2 : [#users=1] = call_module[target=layer4.1.bn2](args = (%layer4_1_conv2,), kwargs = {})
%add_8 : [#users=1] = call_function[target=operator.add](args = (%layer4_1_bn2, %layer4_0_relu_1), kwargs = {})
%layer4_1_relu_1 : [#users=1] = call_module[target=layer4.1.relu](args = (%add_8,), kwargs = {})
%avgpool : [#users=1] = call_module[target=avgpool](args = (%layer4_1_relu_1,), kwargs = {})
%flatten_1 : [#users=1] = call_function[target=torch.flatten](args = (%avgpool, 1), kwargs = {})
%fc : [#users=1] = call_module[target=fc](args = (%flatten_1,), kwargs = {})
return fc
This gives us a Graph representation of the ResNet18 model. A Graph
consists of a series of Nodes connected to each other. Each Node
represents a call-site in the Python code (whether to a function,
a module, or a method) and the edges (represented as args and kwargs
on each node) represent the values passed between these call-sites. More
information about the Graph representation and the rest of FX’s APIs ca
be found at the FX documentation https://pytorch.org/docs/master/fx.html.
Creating a Profiling Interpreter¶
Next, we are going to create a class that inherits from torch.fx.Interpreter.
Though the GraphModule that symbolic_trace produces compiles Python code
that is run when you call a GraphModule, an alternative way to run a
GraphModule is by executing each Node in the Graph one by one. That is
the functionality that Interpreter provides: It interprets the graph node-
by-node.
By inheriting from Interpreter, we can override various functionality and
install the profiling behavior we want. The goal is to have an object to which
we can pass a model, invoke the model 1 or more times, then get statistics about
how long the model and each part of the model took during those runs.
Let’s define our ProfilingInterpreter class:
class ProfilingInterpreter(Interpreter):
def __init__(self, mod : torch.nn.Module):
# Rather than have the user symbolically trace their model,
# we're going to do it in the constructor. As a result, the
# user can pass in any ``Module`` without having to worry about
# symbolic tracing APIs
gm = torch.fx.symbolic_trace(mod)
super().__init__(gm)
# We are going to store away two things here:
#
# 1. A list of total runtimes for ``mod``. In other words, we are
# storing away the time ``mod(...)`` took each time this
# interpreter is called.
self.total_runtime_sec : List[float] = []
# 2. A map from ``Node`` to a list of times (in seconds) that
# node took to run. This can be seen as similar to (1) but
# for specific sub-parts of the model.
self.runtimes_sec : Dict[torch.fx.Node, List[float]] = {}
######################################################################
# Next, let's override our first method: ``run()``. ``Interpreter``'s ``run``
# method is the top-level entrypoint for execution of the model. We will
# want to intercept this so that we can record the total runtime of the
# model.
def run(self, *args) -> Any:
# Record the time we started running the model
t_start = time.time()
# Run the model by delegating back into Interpreter.run()
return_val = super().run(*args)
# Record the time we finished running the model
t_end = time.time()
# Store the total elapsed time this model execution took in the
# ProfilingInterpreter
self.total_runtime_sec.append(t_end - t_start)
return return_val
######################################################################
# Now, let's override ``run_node``. ``Interpreter`` calls ``run_node`` each
# time it executes a single node. We will intercept this so that we
# can measure and record the time taken for each individual call in
# the model.
def run_node(self, n : torch.fx.Node) -> Any:
# Record the time we started running the op
t_start = time.time()
# Run the op by delegating back into Interpreter.run_node()
return_val = super().run_node(n)
# Record the time we finished running the op
t_end = time.time()
# If we don't have an entry for this node in our runtimes_sec
# data structure, add one with an empty list value.
self.runtimes_sec.setdefault(n, [])
# Record the total elapsed time for this single invocation
# in the runtimes_sec data structure
self.runtimes_sec[n].append(t_end - t_start)
return return_val
######################################################################
# Finally, we are going to define a method (one which doesn't override
# any ``Interpreter`` method) that provides us a nice, organized view of
# the data we have collected.
def summary(self, should_sort : bool = False) -> str:
# Build up a list of summary information for each node
node_summaries : List[List[Any]] = []
# Calculate the mean runtime for the whole network. Because the
# network may have been called multiple times during profiling,
# we need to summarize the runtimes. We choose to use the
# arithmetic mean for this.
mean_total_runtime = statistics.mean(self.total_runtime_sec)
# For each node, record summary statistics
for node, runtimes in self.runtimes_sec.items():
# Similarly, compute the mean runtime for ``node``
mean_runtime = statistics.mean(runtimes)
# For easier understanding, we also compute the percentage
# time each node took with respect to the whole network.
pct_total = mean_runtime / mean_total_runtime * 100
# Record the node's type, name of the node, mean runtime, and
# percent runtim
node_summaries.append(
[node.op, str(node), mean_runtime, pct_total])
# One of the most important questions to answer when doing performance
# profiling is "Which op(s) took the longest?". We can make this easy
# to see by providing sorting functionality in our summary view
if should_sort:
node_summaries.sort(key=lambda s: s[2], reverse=True)
# Use the ``tabulate`` library to create a well-formatted table
# presenting our summary information
headers : List[str] = [
'Op type', 'Op', 'Average runtime (s)', 'Pct total runtime'
]
return tabulate.tabulate(node_summaries, headers=headers)
Note
We use Python’s time.time function to pull wall clock
timestamps and compare them. This is not the most accurate
way to measure performance, and will only give us a first-
order approximation. We use this simple technique only for the
purpose of demonstration in this tutorial.
Investigating the Performance of ResNet18¶
We can now use ProfilingInterpreter to inspect the performance
characteristics of our ResNet18 model;
interp = ProfilingInterpreter(rn18)
interp.run(input)
print(interp.summary(True))
Out:
Op type Op Average runtime (s) Pct total runtime
------------- --------------------- --------------------- -------------------
call_module maxpool 0.00891638 11.8896
call_module conv1 0.00371075 4.9481
call_module layer1_0_conv2 0.00354338 4.72492
call_module layer4_0_conv2 0.00326824 4.35804
call_module layer1_0_conv1 0.00322127 4.29541
call_module layer1_1_conv1 0.00317645 4.23564
call_module layer4_1_conv1 0.00312066 4.16125
call_module layer1_1_conv2 0.00309896 4.13232
call_module layer4_1_conv2 0.00308347 4.11165
call_module layer3_1_conv1 0.00274014 3.65385
call_module layer3_0_conv2 0.00270963 3.61316
call_module layer3_1_conv2 0.00270057 3.60107
call_module layer2_1_conv2 0.00264263 3.52382
call_module layer2_0_conv2 0.00253153 3.37567
call_module layer2_1_conv1 0.00247121 3.29524
call_module layer4_0_conv1 0.0023911 3.18842
call_module bn1 0.00231647 3.08891
call_module layer2_0_conv1 0.00208712 2.78307
call_module layer3_0_conv1 0.0018096 2.41301
call_module layer2_0_downsample_0 0.00127912 1.70564
call_function add_2 0.000722647 0.963614
call_module layer1_0_bn1 0.000719786 0.959799
call_module layer2_0_downsample_1 0.000712395 0.949944
call_module layer3_0_downsample_0 0.000679731 0.906389
call_module layer1_1_bn1 0.000669956 0.893354
call_module layer1_1_bn2 0.000652313 0.869828
call_module layer1_0_bn2 0.000645638 0.860926
call_function add_1 0.000637531 0.850117
call_module layer4_0_downsample_0 0.000514984 0.686706
call_module layer2_0_bn1 0.00044179 0.589105
call_module layer2_1_bn2 0.000439644 0.586244
call_module layer2_1_bn1 0.000417948 0.557313
call_module layer2_0_bn2 0.000414133 0.552226
call_module layer3_1_bn2 0.000322819 0.430463
call_module layer3_0_bn2 0.000322104 0.429509
call_module layer3_1_bn1 0.000314713 0.419654
call_module relu_1 0.000310659 0.414249
call_module layer3_0_bn1 0.000306845 0.409162
call_module layer3_0_downsample_1 0.00029707 0.396128
call_module layer4_0_bn1 0.000294447 0.392631
call_module layer4_1_bn1 0.000290632 0.387544
call_function add_4 0.000254393 0.33922
call_module layer4_0_bn2 0.000252008 0.336041
call_module layer4_1_bn2 0.000250578 0.334133
call_module layer4_0_downsample_1 0.000246048 0.328093
call_module fc 0.000200987 0.268006
call_module avgpool 0.000171185 0.228266
call_module layer1_0_relu 0.000132084 0.176127
call_module layer1_1_relu_1 0.000128508 0.171359
call_module layer1_1_relu 0.000113249 0.151012
call_module layer1_0_relu_1 0.000112057 0.149422
call_function add_6 0.00011158 0.148786
call_function add_3 9.9659e-05 0.13289
call_module layer4_1_relu 9.75132e-05 0.130029
call_module layer2_0_relu 9.53674e-05 0.127168
call_module layer2_1_relu 9.44138e-05 0.125896
call_module layer3_1_relu 9.34601e-05 0.124624
call_module layer2_1_relu_1 9.29832e-05 0.123989
call_module layer2_0_relu_1 9.17912e-05 0.122399
call_module layer3_0_relu 9.13143e-05 0.121763
call_module layer3_1_relu_1 9.10759e-05 0.121445
call_module layer3_0_relu_1 9.05991e-05 0.120809
call_module layer4_0_relu 8.96454e-05 0.119538
call_function add_8 8.7738e-05 0.116994
call_function add_5 8.55923e-05 0.114133
call_function flatten_1 8.51154e-05 0.113497
call_module layer4_0_relu_1 8.41618e-05 0.112226
call_module layer4_1_relu_1 8.41618e-05 0.112226
call_function add_7 7.00951e-05 0.0934683
output output 2.3365e-05 0.0311561
placeholder x 1.85966e-05 0.0247977
There are two things we should call out here:
- MaxPool2d takes up the most time. This is a known issue: https://github.com/pytorch/pytorch/issues/51393
- BatchNorm2d also takes up significant time. We can continue this line of thinking and optimize this in the Conv-BN Fusion with FX tutorial TODO: link
Conclusion¶
As we can see, using FX we can easily capture PyTorch programs (even ones we don’t have the source code for!) in a machine-interpretable format and use that for analysis, such as the performance analysis we’ve done here. FX opens up an exiciting world of possibilities for working with PyTorch programs.
Finally, since FX is still in beta, we would be happy to hear any feedback you have about using it. Please feel free to use the PyTorch Forums (https://discuss.pytorch.org/) and the issue tracker (https://github.com/pytorch/pytorch/issues) to provide any feedback you might have.
Total running time of the script: ( 0 minutes 0.479 seconds)
