Note
Go to the end to download the full example code.
Profiling your models¶
Do you feel like your model is too slow? Do you want to make it faster? Instead of guessing which part of the code is responsible for the slowdown, you should profile your code to learn how much time is spent in each function and where to focus any optimization efforts.
In this tutorial you’ll learn how to profile your model using the PyTorch profiler, how to read the output of the profiler, and how to add your own labels for new functions/steps in your model’s forward function.
from typing import Dict, List, Optional
import ase.build
import matplotlib.pyplot as plt
import numpy as np
import torch
from metatensor.torch import Labels, TensorBlock, TensorMap
from metatomic.torch import (
AtomisticModel,
ModelCapabilities,
ModelMetadata,
ModelOutput,
System,
)
from metatomic.torch.ase_calculator import MetatomicCalculator
When profiling your code, it is important to run the model on a representative system to ensure you are actually exercising the behavior of your model at the right scale. Here we’ll use a relatively large system with many atoms.
primitive = ase.build.bulk(name="C", crystalstructure="diamond", a=3.567)
atoms = ase.build.make_supercell(primitive, 10 * np.eye(3))
print(f"We have {len(atoms)} atoms in our system")
We have 2000 atoms in our system
We will use the same HarmonicModel as in the previous tutorial as our machine learning potential.
Click to see the definition of HarmonicModel
class HarmonicModel(torch.nn.Module):
def __init__(self, force_constant: float, equilibrium_positions: torch.Tensor):
"""Create an ``HarmonicModel``.
:param force_constant: force constant, in ``energy unit / (length unit)^2``
:param equilibrium_positions: torch tensor with shape ``n x 3``, containing the
equilibrium positions of all atoms
"""
super().__init__()
assert force_constant > 0
self.force_constant = force_constant
self.equilibrium_positions = equilibrium_positions
def forward(
self,
systems: List[System],
outputs: Dict[str, ModelOutput],
selected_atoms: Optional[Labels],
) -> Dict[str, TensorMap]:
if list(outputs.keys()) != ["energy"]:
raise ValueError(
"this model can only compute 'energy', but `outputs` contains other "
f"keys: {', '.join(outputs.keys())}"
)
# we don't want to worry about selected_atoms yet
if selected_atoms is not None:
raise NotImplementedError("selected_atoms is not implemented")
if outputs["energy"].per_atom:
raise NotImplementedError("per atom energy is not implemented")
# compute the energy for each system by adding together the energy for each atom
energy = torch.zeros((len(systems), 1), dtype=systems[0].positions.dtype)
for i, system in enumerate(systems):
assert len(system) == self.equilibrium_positions.shape[0]
r0 = self.equilibrium_positions
energy[i] += torch.sum(self.force_constant * (system.positions - r0) ** 2)
# add metadata to the output
block = TensorBlock(
values=energy,
samples=Labels("system", torch.arange(len(systems)).reshape(-1, 1)),
components=[],
properties=Labels("energy", torch.tensor([[0]])),
)
return {
"energy": TensorMap(keys=Labels("_", torch.tensor([[0]])), blocks=[block])
}
model = HarmonicModel(
force_constant=3.14159265358979323846,
equilibrium_positions=torch.tensor(atoms.positions),
)
capabilities = ModelCapabilities(
outputs={
"energy": ModelOutput(quantity="energy", unit="eV", per_atom=False),
},
atomic_types=[6],
interaction_range=0.0,
length_unit="Angstrom",
supported_devices=["cpu"],
dtype="float32",
)
metadata = ModelMetadata()
wrapper = AtomisticModel(model.eval(), metadata, capabilities)
wrapper.export("exported-model.pt")
/home/runner/work/metatomic/metatomic/python/examples/4-profiling.py:128: DeprecationWarning: `export()` is deprecated, use `save()` instead
wrapper.export("exported-model.pt")
If you are trying to profile your own model, you can start here and create
MetatomicCalculator with your own model.
Profiling energy calculation¶
We will start with an energy-only calculator, which can be enabled with
do_gradients_with_energy=False.
atoms.calc = MetatomicCalculator("exported-model.pt", do_gradients_with_energy=False)
Before trying to profile the code, it is a good idea to run it a couple of times to allow torch to warmup internally.
for _ in range(10):
# force the model to re-run everytime, otherwise ASE caches calculation results
atoms.rattle(1e-6)
atoms.get_potential_energy()
Now we can run code using torch.profiler.profile() to collect statistic on
how long each function takes to run.
atoms.rattle(1e-6)
with torch.profiler.profile() as energy_profiler:
atoms.get_potential_energy()
print(energy_profiler.key_averages().table(sort_by="self_cpu_time_total", row_limit=10))
/home/runner/work/metatomic/metatomic/.tox/docs/lib/python3.13/site-packages/torch/profiler/profiler.py:217: UserWarning: Warning: Profiler clears events at the end of each cycle.Only events from the current cycle will be reported.To keep events across cycles, set acc_events=True.
_warn_once(
------------------------------------------ ------------ ------------ ------------ ------------ ------------ ------------
Name Self CPU % Self CPU CPU total % CPU total CPU time avg # of Calls
------------------------------------------ ------------ ------------ ------------ ------------ ------------ ------------
MetatomicCalculator::prepare_inputs 15.10% 192.942us 21.12% 269.834us 269.834us 1
AtomisticModel::convert_units_input 10.67% 136.345us 15.08% 192.669us 192.669us 1
Model::forward 9.37% 119.714us 19.78% 252.783us 252.783us 1
MetatomicCalculator::compute_neighbors 8.00% 102.239us 13.11% 167.542us 167.542us 1
AtomisticModel::convert_units_output 6.30% 80.461us 6.47% 82.725us 82.725us 1
AtomisticModel::check_atomic_types 5.77% 73.749us 11.82% 150.992us 150.992us 1
MetatomicCalculator::sum_energies 4.70% 60.092us 4.70% 60.092us 60.092us 1
aten::isin 3.84% 49.052us 4.16% 53.099us 53.099us 1
forward 3.78% 48.290us 10.41% 133.069us 133.069us 1
MetatomicCalculator::convert_outputs 3.54% 45.295us 4.61% 58.899us 58.899us 1
------------------------------------------ ------------ ------------ ------------ ------------ ------------ ------------
Self CPU time total: 1.278ms
There are a couple of interesting things to see here. First the total runtime of the
code is shown in the bottom; and then the most costly functions are visible on top,
one line per function. For each function, Self CPU refers to the time spent in
this function excluding any called functions; and CPU total refers to the time
spent in this function, including called functions.
For more options to record operations and display outputs, please refer to the official documentation for PyTorch profiler.
Here, Model::forward indicates the time taken by your model’s forward().
Anything starting with aten:: comes from operations on torch tensors, typically
with the same function name as the corresponding torch functions (e.g.
aten::arange is torch.arange()). We can also see some internal functions
from metatomic, with the name staring with AtomisticModel:: for
AtomisticModel; and MetatomicCalculator:: for
ase_calculator.MetatomicCalculator.
If you want to see more details on the internal steps taken by your model, you
can add torch.profiler.record_function()
(https://pytorch.org/docs/stable/generated/torch.autograd.profiler.record_function.html)
inside your model code to give names to different steps in the calculation.
This is how we internally set names such as Model::forward or
MetatomicCalculator::prepare_inputs above.
Profiling forces calculation¶
Let us now do the same, but while also computing the forces for this system.
This mean we should now see some time spent in the backward() function, on
top of everything else.
atoms.calc = MetatomicCalculator("exported-model.pt")
# warmup
for _ in range(10):
atoms.rattle(1e-6)
atoms.get_forces()
atoms.rattle(1e-6)
with torch.profiler.profile() as forces_profiler:
atoms.get_forces()
print(forces_profiler.key_averages().table(sort_by="self_cpu_time_total", row_limit=10))
------------------------------------------------------- ------------ ------------ ------------ ------------ ------------ ------------
Name Self CPU % Self CPU CPU total % CPU total CPU time avg # of Calls
------------------------------------------------------- ------------ ------------ ------------ ------------ ------------ ------------
MetatomicCalculator::prepare_inputs 12.63% 220.865us 20.98% 367.015us 367.015us 1
MetatomicCalculator::convert_outputs 7.23% 126.496us 9.92% 173.544us 173.544us 1
MetatomicCalculator::run_backward 7.10% 124.271us 16.74% 292.807us 292.807us 1
AtomisticModel::convert_units_input 6.27% 109.757us 9.68% 169.286us 169.286us 1
Model::forward 6.18% 108.112us 15.81% 276.586us 276.586us 1
MetatomicCalculator::compute_neighbors 6.15% 107.591us 10.41% 182.050us 182.050us 1
aten::mm 4.93% 86.230us 5.00% 87.392us 21.848us 4
AtomisticModel::convert_units_output 4.27% 74.760us 4.40% 77.054us 77.054us 1
forward 4.25% 74.428us 9.63% 168.474us 168.474us 1
AtomisticModel::check_atomic_types 3.70% 64.732us 6.93% 121.317us 121.317us 1
------------------------------------------------------- ------------ ------------ ------------ ------------ ------------ ------------
Self CPU time total: 1.749ms
Let’s visualize this data in another way:
events = forces_profiler.key_averages()
events = sorted(events, key=lambda u: u.self_cpu_time_total, reverse=True)
total_cpu_time = sum(map(lambda u: u.self_cpu_time_total, events))
bottom = 0.0
for event in events:
self_time = event.self_cpu_time_total
name = event.key
if len(name) > 30:
name = name[:12] + "[...]" + name[-12:]
if self_time > 0.03 * total_cpu_time:
plt.bar(0, self_time, bottom=bottom, label=name)
bottom += self_time
else:
plt.bar(0, total_cpu_time - bottom, bottom=bottom, label="others")
break
plt.legend()
plt.xticks([])
plt.xlim(0, 1)
plt.ylabel("self time / µs")
plt.show()
Total running time of the script: (0 minutes 0.234 seconds)