lightning/tests/test_profiler.py

92 lines
3.1 KiB
Python
Raw Normal View History

import time
import numpy as np
import pytest
from pytorch_lightning.profiler import Profiler, AdvancedProfiler
PROFILER_OVERHEAD_MAX_TOLERANCE = 0.001
@pytest.fixture
def simple_profiler():
"""Creates a new profiler for every test with `simple_profiler` as an arg."""
profiler = Profiler()
return profiler
@pytest.fixture
def advanced_profiler():
"""Creates a new profiler for every test with `advanced_profiler` as an arg."""
profiler = AdvancedProfiler()
return profiler
@pytest.mark.parametrize("action,expected", [("a", [3, 1]), ("b", [2]), ("c", [1])])
def test_simple_profiler_durations(simple_profiler, action, expected):
"""Ensure the reported durations are reasonably accurate."""
for duration in expected:
with simple_profiler.profile(action):
time.sleep(duration)
# different environments have different precision when it comes to time.sleep()
# see: https://github.com/PyTorchLightning/pytorch-lightning/issues/796
np.testing.assert_allclose(
simple_profiler.recorded_durations[action], expected, rtol=0.2
)
def test_simple_profiler_overhead(simple_profiler, n_iter=5):
"""Ensure that the profiler doesn't introduce too much overhead during training."""
for _ in range(n_iter):
with simple_profiler.profile("no-op"):
pass
durations = np.array(simple_profiler.recorded_durations["no-op"])
assert all(durations < PROFILER_OVERHEAD_MAX_TOLERANCE)
def test_simple_profiler_describe(simple_profiler):
"""Ensure the profiler won't fail when reporting the summary."""
simple_profiler.describe()
def _get_total_cprofile_duration(profile):
return sum([x.totaltime for x in profile.getstats()])
@pytest.mark.parametrize("action,expected", [("a", [3, 1]), ("b", [2]), ("c", [1])])
def test_advanced_profiler_durations(advanced_profiler, action, expected):
"""Ensure the reported durations are reasonably accurate."""
for duration in expected:
with advanced_profiler.profile(action):
time.sleep(duration)
# different environments have different precision when it comes to time.sleep()
# see: https://github.com/PyTorchLightning/pytorch-lightning/issues/796
recored_total_duration = _get_total_cprofile_duration(
advanced_profiler.profiled_actions[action]
)
expected_total_duration = np.sum(expected)
np.testing.assert_allclose(
recored_total_duration, expected_total_duration, rtol=0.2
)
def test_advanced_profiler_overhead(advanced_profiler, n_iter=5):
"""Ensure that the profiler doesn't introduce too much overhead during training."""
for _ in range(n_iter):
with advanced_profiler.profile("no-op"):
pass
action_profile = advanced_profiler.profiled_actions["no-op"]
total_duration = _get_total_cprofile_duration(action_profile)
average_duration = total_duration / n_iter
assert average_duration < PROFILER_OVERHEAD_MAX_TOLERANCE
def test_advanced_profiler_describe(advanced_profiler):
"""Ensure the profiler won't fail when reporting the summary."""
advanced_profiler.describe()