Note
Click here to download the full example code
Benchmark, comparison torch - forward-backward#
The benchmark compares the processing time between pytorch and onnxruntime-training on a linear regression and a neural network. This example starts from Train a linear regression with forward backward but uses pytorch to replace the parts updating the gradients and computing the error gradient. The training algorithm becomes:
Class TrainingAgent (from onnxruntime-training) is still used and wrapped into ORTModule. This script then follows the same instructions as Benchmark, comparison scikit-learn - forward-backward to compare pytorch only against pytorch and onnxruntime-training.
First comparison: neural network#
import time
import numpy
from pandas import DataFrame
import matplotlib.pyplot as plt
import torch
from onnxruntime import get_device
from onnxruntime.training.ortmodule import ORTModule
from pyquickhelper.pycode.profiling import profile, profile2graph
from sklearn.datasets import make_regression
from sklearn.model_selection import train_test_split
X, y = make_regression(1000, n_features=100, bias=2)
X = X.astype(numpy.float32)
y = y.astype(numpy.float32)
X_train, X_test, y_train, y_test = train_test_split(X, y)
Common parameters and training algorithm#
def from_numpy(v, device=None, requires_grad=False):
"""
Convers a numpy array into a torch array and
sets *device* and *requires_grad*.
"""
v = torch.from_numpy(v)
if device is not None:
v = v.to(device)
v.requires_grad_(requires_grad)
return v
Training, two functions with same code but it is easier to distinguish between in the profiling.
def train_model_torch(model, device, x, y, n_iter=100, learning_rate=1e-5,
profiler=None):
def forward_torch(model, x):
return model(x)
model = model.to(device)
x = from_numpy(x, requires_grad=True, device=device)
y = from_numpy(y, requires_grad=True, device=device)
criterion = torch.nn.MSELoss(reduction='sum')
optimizer = torch.optim.SGD(model.parameters(), lr=learning_rate)
losses = []
for t in range(n_iter):
def step_train_torch():
y_pred = forward_torch(model, x)
loss = criterion(y_pred, y)
optimizer.zero_grad()
loss.backward()
optimizer.step()
return loss
loss = step_train_torch()
losses.append(loss)
if profiler is not None:
profiler.step()
return losses
def train_model_ort(model, device, x, y, n_iter=100, learning_rate=1e-5,
profiler=None):
def forward_ort(model, x):
return model(x)
model = model.to(device)
x = from_numpy(x, requires_grad=True, device=device)
y = from_numpy(y, requires_grad=True, device=device)
criterion = torch.nn.MSELoss(reduction='sum')
optimizer = torch.optim.SGD(model.parameters(), lr=learning_rate)
losses = []
for t in range(n_iter):
def step_train_ort():
y_pred = forward_ort(model, x)
loss = criterion(y_pred, y)
optimizer.zero_grad()
loss.backward()
optimizer.step()
return loss
loss = step_train_ort()
losses.append(loss)
if profiler is not None:
profiler.step()
return losses
Benchmark function
def benchmark(model_torch, model_ort, device, name, verbose=True, max_iter=100):
print("[benchmark] %s" % name)
begin = time.perf_counter()
losses = train_model_torch(
model_torch, device, X_train, y_train, n_iter=200)
duration_torch = time.perf_counter() - begin
length_torch = len(losses)
print("[benchmark] torch=%r iterations - %r seconds" % (
length_torch, duration_torch))
begin = time.perf_counter()
losses = train_model_ort(model_ort, device, X_train,
y_train, n_iter=max_iter)
duration_ort = time.perf_counter() - begin
length_ort = len(losses)
print("[benchmark] onxrt=%r iteration - %r seconds" % (
length_ort, duration_ort))
return dict(torch=duration_torch, ort=duration_ort, name=name,
iter_torch=length_torch, iter_ort=length_ort)
class MLPNet(torch.nn.Module):
def __init__(self, D_in, D_out):
super(MLPNet, self).__init__()
self.linear1 = torch.nn.Linear(D_in, 50)
self.linear2 = torch.nn.Linear(50, 10)
self.linear3 = torch.nn.Linear(10, D_out)
def forward(self, x):
o1 = torch.sigmoid(self.linear1(x))
o2 = torch.sigmoid(self.linear2(o1))
return self.linear3(o2)
d_in, d_out, N = X.shape[1], 1, X.shape[0]
model_torch = MLPNet(d_in, d_out)
model_ort = ORTModule(MLPNet(d_in, d_out))
max_iter = 100
device = torch.device('cpu')
benches = [benchmark(model_torch, model_ort, device, name='NN-CPU',
max_iter=max_iter)]
Out:
[benchmark] NN-CPU
somewhere/workspace/onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/loss.py:529: UserWarning: Using a target size (torch.Size([750])) that is different to the input size (torch.Size([750, 1])). This will likely lead to incorrect results due to broadcasting. Please ensure they have the same size.
return F.mse_loss(input, target, reduction=self.reduction)
[benchmark] torch=200 iterations - 2.77465951628983 seconds
somewhere/workspace/onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/onnxruntime/training/ortmodule/_training_manager.py:216: UserWarning: Fast path enabled - skipping checks. Rebuild graph: True, Execution agent: True, Device check: True
warnings.warn(f"Fast path enabled - skipping checks."
[benchmark] onxrt=100 iteration - 9.256914421916008 seconds
Profiling#
def clean_name(text):
pos = text.find('onnxruntime')
if pos >= 0:
return text[pos:]
pos = text.find('onnxcustom')
if pos >= 0:
return text[pos:]
pos = text.find('torch')
if pos >= 0:
return text[pos:]
pos = text.find('site-packages')
if pos >= 0:
return text[pos:]
return text
ps = profile(lambda: benchmark(
model_torch, model_ort, device, name='NN-CPU', max_iter=max_iter))[0]
root, nodes = profile2graph(ps, clean_text=clean_name)
text = root.to_text()
print(text)
Out:
[benchmark] NN-CPU
somewhere/workspace/onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/loss.py:529: UserWarning: Using a target size (torch.Size([750])) that is different to the input size (torch.Size([750, 1])). This will likely lead to incorrect results due to broadcasting. Please ensure they have the same size.
return F.mse_loss(input, target, reduction=self.reduction)
[benchmark] torch=200 iterations - 21.010495772585273 seconds
[benchmark] onxrt=100 iteration - 9.620220424607396 seconds
__contains__ -- 800 800 -- 0.00309 0.00364 -- /usr/local/lib/python3.9/enum.py:748:__contains__ (__contains__)
<built-in method builtins.isinstance> -- 800 800 -- 0.00055 0.00055 -- ~:0:<built-in method builtins.isinstance> (<built-in method builtins.isinstance>) +++
filter -- 18 18 -- 0.00008 0.00019 -- /usr/local/lib/python3.9/logging/__init__.py:787:filter (filter)
filter -- 12 12 -- 0.00002 0.00002 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/sphinx/util/logging.py:350:filter (filter)
filter -- 6 6 -- 0.00005 0.00007 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/sphinx/util/logging.py:483:filter (filter)
<built-in method builtins.isinstance> -- 18 18 -- 0.00001 0.00001 -- ~:0:<built-in method builtins.isinstance> (<built-in method builtins.isinstance>) +++
<built-in method builtins.hasattr> -- 18 18 -- 0.00002 0.00002 -- ~:0:<built-in method builtins.hasattr> (<built-in method builtins.hasattr>) +++
acquire -- 30 30 -- 0.00007 0.00012 -- /usr/local/lib/python3.9/logging/__init__.py:892:acquire (acquire)
<method 'acquire' of '_thread.RLock' objects> -- 30 30 -- 0.00005 0.00005 -- ~:0:<method 'acquire' of '_thread.RLock' objects> (<method 'acquire' of '_thread.RLock' objects>)
release -- 30 30 -- 0.00006 0.00007 -- /usr/local/lib/python3.9/logging/__init__.py:899:release (release)
<method 'release' of '_thread.RLock' objects> -- 30 30 -- 0.00001 0.00001 -- ~:0:<method 'release' of '_thread.RLock' objects> (<method 'release' of '_thread.RLock' objects>)
emit -- 12 12 -- 0.00010 0.00116 -- /usr/local/lib/python3.9/logging/__init__.py:1067:emit (emit)
format -- 12 12 -- 0.00004 0.00061 -- /usr/local/lib/python3.9/logging/__init__.py:912:format (format)
format -- 12 12 -- 0.00010 0.00057 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/sphinx/util/logging.py:528:format (format)
format -- 12 12 -- 0.00009 0.00042 -- /usr/local/lib/python3.9/logging/__init__.py:646:format (format)
usesTime -- 12 12 -- 0.00003 0.00009 -- /usr/local/lib/python3.9/logging/__init__.py:624:usesTime (usesTime)
usesTime -- 12 12 -- 0.00004 0.00007 -- /usr/local/lib/python3.9/logging/__init__.py:417:usesTime (usesTime)
<method 'find' of 'str' objects> -- 12 12 -- 0.00002 0.00002 -- ~:0:<method 'find' of 'str' objects> (<method 'find' of 'str' objects>)
formatMessage -- 12 12 -- 0.00002 0.00010 -- /usr/local/lib/python3.9/logging/__init__.py:630:formatMessage (formatMessage)
format -- 12 12 -- 0.00002 0.00007 -- /usr/local/lib/python3.9/logging/__init__.py:428:format (format)
_format -- 12 12 -- 0.00005 0.00005 -- /usr/local/lib/python3.9/logging/__init__.py:425:_format (_format)
getMessage -- 12 12 -- 0.00007 0.00015 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/sphinx/util/logging.py:88:getMessage (getMessage)
getMessage -- 12 12 -- 0.00006 0.00006 -- /usr/local/lib/python3.9/logging/__init__.py:354:getMessage (getMessage)
<built-in method builtins.getattr> -- 12 12 -- 0.00001 0.00001 -- ~:0:<built-in method builtins.getattr> (<built-in method builtins.getattr>) +++
colorize -- 2 2 -- 0.00002 0.00003 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/sphinx/util/console.py:81:colorize (colorize)
escseq -- 4 4 -- 0.00001 0.00001 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/sphinx/util/console.py:82:escseq (escseq)
<built-in method builtins.getattr> -- 12 12 -- 0.00002 0.00002 -- ~:0:<built-in method builtins.getattr> (<built-in method builtins.getattr>) +++
flush -- 12 12 -- 0.00007 0.00037 -- /usr/local/lib/python3.9/logging/__init__.py:1056:flush (flush)
acquire -- 12 12 -- 0.00002 0.00004 -- /usr/local/lib/python3.9/logging/__init__.py:892:acquire (acquire) +++
release -- 12 12 -- 0.00003 0.00004 -- /usr/local/lib/python3.9/logging/__init__.py:899:release (release) +++
flush -- 6 6 -- 0.00003 0.00021 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/sphinx/util/logging.py:554:flush (flush)
<method 'flush' of '_io.TextIOWrapper' objects> -- 6 6 -- 0.00018 0.00018 -- ~:0:<method 'flush' of '_io.TextIOWrapper' objects> (<method 'flush' of '_io.TextIOWrapper' objects>)
<built-in method builtins.hasattr> -- 12 12 -- 0.00001 0.00001 -- ~:0:<built-in method builtins.hasattr> (<built-in method builtins.hasattr>) +++
write -- 6 6 -- 0.00003 0.00004 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/sphinx/util/logging.py:546:write (write)
write -- 6 6 -- 0.00003 0.00004 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/sphinx/util/logging.py:564:write (write)
isEnabledFor -- 12 12 -- 0.00003 0.00003 -- /usr/local/lib/python3.9/logging/__init__.py:1677:isEnabledFor (isEnabledFor)
__init__ -- 2 2 -- 0.00001 0.00001 -- /usr/local/lib/python3.9/warnings.py:403:__init__ (__init__)
from_numpy -- 4 4 -- 0.00005 0.00015 -- onnxcustom/onnxcustom_UT_39_std/_doc/examples/plot_orttraining_benchmark_torch.py:52:from_numpy (from_numpy)
<lambda> -- 1 1 -- 0.00142 30.63475 -- onnxcustom/onnxcustom_UT_39_std/_doc/examples/plot_orttraining_benchmark_torch.py:198:<lambda> (<lambda>)
benchmark -- 1 1 -- 0.00962 30.63333 -- onnxcustom/onnxcustom_UT_39_std/_doc/examples/plot_orttraining_benchmark_torch.py:132:benchmark (benchmark)
train_model_torch -- 1 1 -- 0.00351 21.01046 -- onnxcustom/onnxcustom_UT_39_std/_doc/examples/plot_orttraining_benchmark_torch.py:68:train_model_torch (train_model_torch)
from_numpy -- 2 2 -- 0.00003 0.00009 -- onnxcustom/onnxcustom_UT_39_std/_doc/examples/plot_orttraining_benchmark_torch.py:52:from_numpy (from_numpy) +++
step_train_torch -- 200 200 -- 0.00852 21.00457 -- onnxcustom/onnxcustom_UT_39_std/_doc/examples/plot_orttraining_benchmark_torch.py:83:step_train_torch (step_train_torch)
forward_torch -- 200 200 -- 0.00129 3.53889 -- onnxcustom/onnxcustom_UT_39_std/_doc/examples/plot_orttraining_benchmark_torch.py:71:forward_torch (forward_torch)
_call_impl -- 200 200 -- 0.00277 3.53759 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1104:_call_impl (_call_impl) +++
backward -- 200 200 -- 0.00348 15.37920 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/_tensor.py:307:backward (backward) +++
_call_impl -- 200 200 -- 0.00306 1.85074 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1104:_call_impl (_call_impl) +++
wrapper -- 200 200 -- 0.00986 0.13706 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/optim/optimizer.py:83:wrapper (wrapper) +++
zero_grad -- 200 200 -- 0.03179 0.09017 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/optim/optimizer.py:189:zero_grad (zero_grad) +++
__init__ -- 1 1 -- 0.00001 0.00029 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/loss.py:525:__init__ (__init__) +++
to -- 1 1 -- 0.00005 0.00136 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:803:to (to) +++
__init__ -- 1 1 -- 0.00002 0.00044 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/optim/sgd.py:88:__init__ (__init__) +++
<method 'append' of 'list' objects> -- 200 200 -- 0.00020 0.00020 -- ~:0:<method 'append' of 'list' objects> (<method 'append' of 'list' objects>) +++
train_model_ort -- 1 1 -- 0.00214 9.61072 -- onnxcustom/onnxcustom_UT_39_std/_doc/examples/plot_orttraining_benchmark_torch.py:99:train_model_ort (train_model_ort)
from_numpy -- 2 2 -- 0.00002 0.00006 -- onnxcustom/onnxcustom_UT_39_std/_doc/examples/plot_orttraining_benchmark_torch.py:52:from_numpy (from_numpy) +++
step_train_ort -- 100 100 -- 0.00464 9.60641 -- onnxcustom/onnxcustom_UT_39_std/_doc/examples/plot_orttraining_benchmark_torch.py:113:step_train_ort (step_train_ort)
forward_ort -- 100 100 -- 0.00066 0.29955 -- onnxcustom/onnxcustom_UT_39_std/_doc/examples/plot_orttraining_benchmark_torch.py:101:forward_ort (forward_ort)
_call_impl -- 100 100 -- 0.00157 0.29889 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1104:_call_impl (_call_impl) +++
backward -- 100 100 -- 0.00156 7.21782 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/_tensor.py:307:backward (backward) +++
_call_impl -- 100 100 -- 0.00192 1.97119 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1104:_call_impl (_call_impl) +++
wrapper -- 100 100 -- 0.00471 0.07118 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/optim/optimizer.py:83:wrapper (wrapper) +++
zero_grad -- 100 100 -- 0.01201 0.04202 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/optim/optimizer.py:189:zero_grad (zero_grad) +++
__init__ -- 1 1 -- 0.00001 0.00027 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/loss.py:525:__init__ (__init__) +++
to -- 1 1 -- 0.00002 0.00134 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:803:to (to) +++
__init__ -- 1 1 -- 0.00001 0.00041 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/optim/sgd.py:88:__init__ (__init__) +++
<method 'append' of 'list' objects> -- 100 100 -- 0.00010 0.00010 -- ~:0:<method 'append' of 'list' objects> (<method 'append' of 'list' objects>) +++
<built-in method builtins.print> -- 3 3 -- 0.00005 0.00251 -- ~:0:<built-in method builtins.print> (<built-in method builtins.print>)
write -- 6 6 -- 0.00012 0.00247 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/sphinx_gallery/gen_rst.py:81:write (write) +++
write -- 7 7 -- 0.00014 0.00346 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/sphinx_gallery/gen_rst.py:81:write (write)
verbose -- 6 6 -- 0.00006 0.00329 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/sphinx/util/logging.py:127:verbose (verbose)
log -- 6 6 -- 0.00008 0.00324 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/sphinx/util/logging.py:120:log (log)
log -- 6 6 -- 0.00010 0.00315 -- /usr/local/lib/python3.9/logging/__init__.py:1825:log (log)
log -- 6 6 -- 0.00008 0.00295 -- /usr/local/lib/python3.9/logging/__init__.py:1485:log (log)
_log -- 6 6 -- 0.00006 0.00286 -- /usr/local/lib/python3.9/logging/__init__.py:1553:_log (_log)
findCaller -- 6 6 -- 0.00010 0.00018 -- /usr/local/lib/python3.9/logging/__init__.py:1502:findCaller (findCaller)
<lambda> -- 6 6 -- 0.00002 0.00003 -- /usr/local/lib/python3.9/logging/__init__.py:156:<lambda> (<lambda>)
normcase -- 12 12 -- 0.00002 0.00003 -- /usr/local/lib/python3.9/posixpath.py:52:normcase (normcase)
<built-in met...osix.fspath> -- 12 12 -- 0.00001 0.00001 -- ~:0:<built-in method posix.fspath> (<built-in method posix.fspath>) +++
<built-in metho...ltins.hasattr> -- 12 12 -- 0.00001 0.00001 -- ~:0:<built-in method builtins.hasattr> (<built-in method builtins.hasattr>) +++
makeRecord -- 6 6 -- 0.00006 0.00087 -- /usr/local/lib/python3.9/logging/__init__.py:1538:makeRecord (makeRecord)
__init__ -- 6 6 -- 0.00034 0.00081 -- /usr/local/lib/python3.9/logging/__init__.py:278:__init__ (__init__)
getLevelName -- 6 6 -- 0.00005 0.00006 -- /usr/local/lib/python3.9/logging/__init__.py:119:getLevelName (getLevelName)
<method 'ge...' objects> -- 12 12 -- 0.00001 0.00001 -- ~:0:<method 'get' of 'dict' objects> (<method 'get' of 'dict' objects>) +++
current_process -- 6 6 -- 0.00001 0.00001 -- /usr/local/lib/python3.9/multiprocessing/process.py:37:current_process (current_process)
name -- 6 6 -- 0.00001 0.00001 -- /usr/local/lib/python3.9/multiprocessing/process.py:189:name (name)
splitext -- 6 6 -- 0.00003 0.00010 -- /usr/local/lib/python3.9/posixpath.py:117:splitext (splitext)
_splitext -- 6 6 -- 0.00005 0.00006 -- /usr/local/lib/python3.9/genericpath.py:121:_splitext (_splitext)
<method '...objects> -- 12 12 -- 0.00002 0.00002 -- ~:0:<method 'rfind' of 'str' objects> (<method 'rfind' of 'str' objects>) +++
basename -- 6 6 -- 0.00005 0.00009 -- /usr/local/lib/python3.9/posixpath.py:140:basename (basename)
_get_sep -- 6 6 -- 0.00001 0.00002 -- /usr/local/lib/python3.9/posixpath.py:41:_get_sep (_get_sep)
name -- 6 6 -- 0.00002 0.00002 -- /usr/local/lib/python3.9/threading.py:1053:name (name)
current_thread -- 6 6 -- 0.00004 0.00005 -- /usr/local/lib/python3.9/threading.py:1318:current_thread (current_thread)
handle -- 6 6 -- 0.00003 0.00174 -- /usr/local/lib/python3.9/logging/__init__.py:1579:handle (handle)
filter -- 6 6 -- 0.00001 0.00001 -- /usr/local/lib/python3.9/logging/__init__.py:787:filter (filter) +++
callHandlers -- 6 6 -- 0.00008 0.00170 -- /usr/local/lib/python3.9/logging/__init__.py:1633:callHandlers (callHandlers)
handle -- 12 12 -- 0.00008 0.00162 -- /usr/local/lib/python3.9/logging/__init__.py:935:handle (handle)
filter -- 12 12 -- 0.00007 0.00018 -- /usr/local/lib/python3.9/logging/__init__.py:787:filter (filter) +++
acquire -- 12 12 -- 0.00004 0.00006 -- /usr/local/lib/python3.9/logging/__init__.py:892:acquire (acquire) +++
release -- 12 12 -- 0.00002 0.00003 -- /usr/local/lib/python3.9/logging/__init__.py:899:release (release) +++
emit -- 6 6 -- 0.00004 0.00041 -- /usr/local/lib/python3.9/logging/__init__.py:1067:emit (emit) +++
emit -- 6 6 -- 0.00007 0.00086 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/sphinx/util/logging.py:150:emit (emit)
acquire -- 6 6 -- 0.00001 0.00002 -- /usr/local/lib/python3.9/logging/__init__.py:892:acquire (acquire) +++
release -- 6 6 -- 0.00001 0.00001 -- /usr/local/lib/python3.9/logging/__init__.py:899:release (release) +++
emit -- 6 6 -- 0.00006 0.00075 -- /usr/local/lib/python3.9/logging/__init__.py:1067:emit (emit) +++
isEnabledFor -- 6 6 -- 0.00001 0.00001 -- /usr/local/lib/python3.9/logging/__init__.py:1677:isEnabledFor (isEnabledFor) +++
isEnabledFor -- 6 6 -- 0.00002 0.00004 -- /usr/local/lib/python3.9/logging/__init__.py:1834:isEnabledFor (isEnabledFor)
isEnabledFor -- 6 6 -- 0.00002 0.00002 -- /usr/local/lib/python3.9/logging/__init__.py:1677:isEnabledFor (isEnabledFor) +++
process -- 6 6 -- 0.00005 0.00006 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/sphinx/util/logging.py:130:process (process)
backward -- 300 300 -- 0.00504 22.59702 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/_tensor.py:307:backward (backward)
backward -- 300 300 -- 0.00664 22.59177 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/__init__.py:85:backward (backward)
_make_grads -- 300 300 -- 0.00674 0.01805 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/__init__.py:30:_make_grads (_make_grads)
<built-in method ones_like> -- 300 300 -- 0.00979 0.00979 -- ~:0:<built-in method ones_like> (<built-in method ones_like>)
<method 'numel' of 'tor...._TensorBase' objects> -- 300 300 -- 0.00048 0.00048 -- ~:0:<method 'numel' of 'torch._C._TensorBase' objects> (<method 'numel' of 'torch._C._TensorBase' objects>)
<method 'append' of 'list' objects> -- 300 300 -- 0.00043 0.00043 -- ~:0:<method 'append' of 'list' objects> (<method 'append' of 'list' objects>) +++
<built-in method builtins.isinstance> -- 300 300 -- 0.00061 0.00061 -- ~:0:<built-in method builtins.isinstance> (<built-in method builtins.isinstance>) +++
_tensor_or_tensors_to_tuple -- 300 300 -- 0.00103 0.00103 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/__init__.py:77:_tensor_or_tensors_to_tuple (_tensor_or_tensors_to_tuple)
<method 'run_backward' of..._C._EngineBase' objects> -- 300 300 -- 22.07434 22.56436 -- ~:0:<method 'run_backward' of 'torch._C._EngineBase' objects> (<method 'run_backward' of 'torch._C._EngineBase' objects>)
apply -- 100 100 -- 0.00324 0.49002 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/function.py:243:apply (apply)
backward -- 100 100 -- 0.03767 0.48678 -- onnxruntime/training/ortmodule/_training_manager.py:121:backward (backward)
run_backward -- 100 100 -- 0.40808 0.40808 -- onnxruntime/training/ortmodule/_execution_agent.py:129:run_backward (run_backward)
is_set -- 100 100 -- 0.00062 0.00257 -- onnxruntime/training/ortmodule/_graph_execution_manager.py:59:is_set (is_set) +++
_ortvalue_to_torch_tensor -- 700 700 -- 0.01543 0.03528 -- onnxruntime/training/ortmodule/_utils.py:64:_ortvalue_to_torch_tensor (_ortvalue_to_torch_tensor) +++
_torch_tensor_to_dlpack -- 100 100 -- 0.00090 0.00164 -- onnxruntime/training/ortmodule/_utils.py:74:_torch_tensor_to_dlpack (_torch_tensor_to_dlpack) +++
<method 'is_contigu...nsorBase' objects> -- 100 100 -- 0.00051 0.00051 -- ~:0:<method 'is_contiguous' of 'torch._C._TensorBase' objects> (<method 'is_contiguous' of 'torch._C._TensorBase' objects>) +++
<method 'append' of 'list' objects> -- 700 700 -- 0.00061 0.00061 -- ~:0:<method 'append' of 'list' objects> (<method 'append' of 'list' objects>) +++
<built-in method builtins.len> -- 200 200 -- 0.00042 0.00042 -- ~:0:<built-in method builtins.len> (<built-in method builtins.len>) +++
<built-in method builtins.isinstance> -- 600 600 -- 0.00128 0.00128 -- ~:0:<built-in method builtins.isinstance> (<built-in method builtins.isinstance>) +++
<built-in method builtins.len> -- 300 300 -- 0.00041 0.00041 -- ~:0:<built-in method builtins.len> (<built-in method builtins.len>) +++
<built-in method torch._C._has_torch_function_unary> -- 300 300 -- 0.00021 0.00021 -- ~:0:<built-in method torch._C._has_torch_function_unary> (<built-in method torch._C._has_torch_function_unary>) +++
__hash__ -- 3660 3660 -- 0.00511 0.00987 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/_tensor.py:706:__hash__ (__hash__)
<built-in method torch._C._has_torch_function_unary> -- 3660 3660 -- 0.00202 0.00202 -- ~:0:<built-in method torch._C._has_torch_function_unary> (<built-in method torch._C._has_torch_function_unary>) +++
<built-in method builtins.id> -- 3660 3660 -- 0.00275 0.00275 -- ~:0:<built-in method builtins.id> (<built-in method builtins.id>)
grad -- 10848 10848 -- 0.03244 0.03886 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/_tensor.py:1092:grad (grad)
<built-in method torch._C._has_torch_function_unary> -- 10848 10848 -- 0.00642 0.00642 -- ~:0:<built-in method torch._C._has_torch_function_unary> (<built-in method torch._C._has_torch_function_unary>) +++
__init__ -- 324 324 -- 0.00499 0.00532 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/grad_mode.py:121:__init__ (__init__)
is_scripting -- 324 324 -- 0.00033 0.00033 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/_jit_internal.py:957:is_scripting (is_scripting)
__enter__ -- 324 324 -- 0.00323 0.00693 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/grad_mode.py:126:__enter__ (__enter__)
__init__ -- 324 324 -- 0.00232 0.00305 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/grad_mode.py:215:__init__ (__init__) +++
<built-in method torch._C.is_grad_enabled> -- 324 324 -- 0.00066 0.00066 -- ~:0:<built-in method torch._C.is_grad_enabled> (<built-in method torch._C.is_grad_enabled>) +++
__exit__ -- 324 324 -- 0.00201 0.00430 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/grad_mode.py:130:__exit__ (__exit__)
__init__ -- 324 324 -- 0.00150 0.00230 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/grad_mode.py:215:__init__ (__init__) +++
__init__ -- 648 648 -- 0.00382 0.00534 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/grad_mode.py:215:__init__ (__init__)
<built-in method torch._C._set_grad_enabled> -- 648 648 -- 0.00066 0.00066 -- ~:0:<built-in method torch._C._set_grad_enabled> (<built-in method torch._C._set_grad_enabled>)
<built-in method torch._C.is_grad_enabled> -- 648 648 -- 0.00087 0.00087 -- ~:0:<built-in method torch._C.is_grad_enabled> (<built-in method torch._C.is_grad_enabled>) +++
__init__ -- 600 600 -- 0.00650 0.03245 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/profiler.py:426:__init__ (__init__)
<built-in method zeros> -- 600 600 -- 0.02595 0.02595 -- ~:0:<built-in method zeros> (<built-in method zeros>)
__enter__ -- 600 600 -- 0.01008 0.04464 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/profiler.py:435:__enter__ (__enter__)
<built-in method torch._ops...er._record_function_enter> -- 600 600 -- 0.03456 0.03456 -- ~:0:<built-in method torch._ops.profiler._record_function_enter> (<built-in method torch._ops.profiler._record_function_enter>)
__exit__ -- 600 600 -- 0.00430 0.01401 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/profiler.py:439:__exit__ (__exit__)
<built-in method torch._ops...ler._record_function_exit> -- 600 600 -- 0.00971 0.00971 -- ~:0:<built-in method torch._ops.profiler._record_function_exit> (<built-in method torch._ops.profiler._record_function_exit>)
__init__ -- 2 2 -- 0.00001 0.00055 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/loss.py:525:__init__ (__init__)
__init__ -- 2 2 -- 0.00002 0.00054 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/loss.py:15:__init__ (__init__)
__init__ -- 2 2 -- 0.00011 0.00049 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:254:__init__ (__init__)
__setattr__ -- 22 22 -- 0.00026 0.00035 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1188:__setattr__ (__setattr__) +++
__setattr__ -- 2 2 -- 0.00002 0.00003 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1188:__setattr__ (__setattr__) +++
_apply -- 2 9 -- 0.00080 0.00260 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:576:_apply (_apply)
grad -- 48 48 -- 0.00018 0.00021 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/_tensor.py:1092:grad (grad) +++
__init__ -- 24 24 -- 0.00018 0.00019 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/grad_mode.py:121:__init__ (__init__) +++
__enter__ -- 24 24 -- 0.00016 0.00034 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/grad_mode.py:126:__enter__ (__enter__) +++
__exit__ -- 24 24 -- 0.00012 0.00028 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/grad_mode.py:130:__exit__ (__exit__) +++
_apply -- 7 4 -- 0.00075 0.00247 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:576:_apply (_apply) +++
compute_should_use_set_data -- 24 24 -- 0.00013 0.00027 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:580:compute_should_use_set_data (compute_should_use_set_data)
get_overwrite_module_params_on_conversion -- 24 24 -- 0.00001 0.00001 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/__future__.py:18:get_overwrite_module_params_on_conversion (get_overwrite_module_params_on_conversion)
<built-in method _has_compatible_shallow_copy_type> -- 24 24 -- 0.00012 0.00012 -- ~:0:<built-in method _has_compatible_shallow_copy_type> (<built-in method _has_compatible_shallow_copy_type>)
convert -- 24 24 -- 0.00012 0.00033 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:901:convert (convert)
<method 'to' of 'torch._C._TensorBase' objects> -- 24 24 -- 0.00015 0.00015 -- ~:0:<method 'to' of 'torch._C._TensorBase' objects> (<method 'to' of 'torch._C._TensorBase' objects>) +++
<method 'is_floating_poin..._C._TensorBase' objects> -- 24 24 -- 0.00006 0.00006 -- ~:0:<method 'is_floating_point' of 'torch._C._TensorBase' objects> (<method 'is_floating_point' of 'torch._C._TensorBase' objects>)
children -- 16 16 -- 0.00005 0.00016 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1612:children (children)
named_children -- 16 16 -- 0.00009 0.00011 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1621:named_children (named_children)
<method 'items' of 'collections.OrderedDict' objects> -- 18 18 -- 0.00001 0.00001 -- ~:0:<method 'items' of 'collections.OrderedDict' objects> (<method 'items' of 'collections.OrderedDict' objects>) +++
to -- 2 2 -- 0.00006 0.00270 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:803:to (to)
_apply -- 1 1 -- 0.00003 0.00130 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:576:_apply (_apply) +++
_apply -- 1 1 -- 0.00001 0.00131 -- onnxruntime/training/ortmodule/ortmodule.py:173:_apply (_apply)
_apply -- 1 1 -- 0.00001 0.00131 -- onnxruntime/training/ortmodule/_torch_module_ort.py:27:_apply (_apply)
_apply -- 1 1 -- 0.00002 0.00130 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:576:_apply (_apply) +++
_call_impl -- 600 1200 -- 0.01661 7.65842 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1104:_call_impl (_call_impl)
forward -- 200 200 -- 0.01481 3.53353 -- onnxcustom/onnxcustom_UT_39_std/_doc/examples/plot_orttraining_benchmark_torch.py:162:forward (forward)
_call_impl -- 600 600 -- 0.00730 2.55249 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1104:_call_impl (_call_impl) +++
__getattr__ -- 600 600 -- 0.00404 0.00404 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1172:__getattr__ (__getattr__) +++
<built-in method sigmoid> -- 400 400 -- 0.96219 0.96219 -- ~:0:<built-in method sigmoid> (<built-in method sigmoid>)
forward -- 600 600 -- 0.00831 2.54213 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/linear.py:102:forward (forward)
__getattr__ -- 1200 1200 -- 0.00277 0.00277 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1172:__getattr__ (__getattr__) +++
<built-in method torch._C._nn.linear> -- 600 600 -- 2.53105 2.53105 -- ~:0:<built-in method torch._C._nn.linear> (<built-in method torch._C._nn.linear>)
forward -- 300 300 -- 0.00327 3.81544 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/loss.py:528:forward (forward)
mse_loss -- 300 300 -- 0.01592 3.81217 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/functional.py:3234:mse_loss (mse_loss)
broadcast_tensors -- 300 300 -- 0.00617 0.02908 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/functional.py:46:broadcast_tensors (broadcast_tensors)
__getattr__ -- 300 300 -- 0.00129 0.00180 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/_VF.py:25:__getattr__ (__getattr__)
<built-in method builtins.getattr> -- 300 300 -- 0.00050 0.00050 -- ~:0:<built-in method builtins.getattr> (<built-in method builtins.getattr>) +++
<built-in method torc..._has_torch_function> -- 300 300 -- 0.00035 0.00035 -- ~:0:<built-in method torch._C._has_torch_function> (<built-in method torch._C._has_torch_function>)
<built-in method broadcast_tensors> -- 300 300 -- 0.02077 0.02077 -- ~:0:<built-in method broadcast_tensors> (<built-in method broadcast_tensors>)
get_enum -- 300 300 -- 0.00075 0.00075 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/_reduction.py:7:get_enum (get_enum)
<built-in method torch....rch_function_variadic> -- 300 300 -- 0.00039 0.00039 -- ~:0:<built-in method torch._C._has_torch_function_variadic> (<built-in method torch._C._has_torch_function_variadic>)
<method 'size' of 'torc...._TensorBase' objects> -- 1200 1200 -- 0.00415 0.00415 -- ~:0:<method 'size' of 'torch._C._TensorBase' objects> (<method 'size' of 'torch._C._TensorBase' objects>)
<built-in method torch._C._nn.mse_loss> -- 300 300 -- 3.73429 3.73429 -- ~:0:<built-in method torch._C._nn.mse_loss> (<built-in method torch._C._nn.mse_loss>)
<method 'format' of 'str' objects> -- 300 300 -- 0.01553 0.01553 -- ~:0:<method 'format' of 'str' objects> (<method 'format' of 'str' objects>) +++
<built-in method _warnings.warn> -- 300 300 -- 0.01097 0.01206 -- ~:0:<built-in method _warnings.warn> (<built-in method _warnings.warn>)
_showwarnmsg -- 1 1 -- 0.00002 0.00108 -- /usr/local/lib/python3.9/warnings.py:96:_showwarnmsg (_showwarnmsg)
_showwarning -- 1 1 -- 0.00001 0.00106 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/sphinx_gallery/gen_rst.py:473:_showwarning (_showwarning)
formatwarning -- 1 1 -- 0.00001 0.00006 -- /usr/local/lib/python3.9/warnings.py:15:formatwarning (formatwarning)
_formatwarnmsg_impl -- 1 1 -- 0.00003 0.00005 -- /usr/local/lib/python3.9/warnings.py:35:_formatwarnmsg_impl (_formatwarnmsg_impl)
getline -- 1 1 -- 0.00001 0.00001 -- /usr/local/lib/python3.9/linecache.py:26:getline (getline)
getlines -- 1 1 -- 0.00001 0.00001 -- /usr/local/lib/python3.9/linecache.py:36:getlines (getlines)
__init__ -- 1 1 -- 0.00000 0.00000 -- /usr/local/lib/python3.9/warnings.py:403:__init__ (__init__) +++
write -- 1 1 -- 0.00002 0.00099 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/sphinx_gallery/gen_rst.py:81:write (write) +++
__init__ -- 1 1 -- 0.00001 0.00001 -- /usr/local/lib/python3.9/warnings.py:403:__init__ (__init__) +++
_forward -- 100 100 -- 0.00063 0.29666 -- onnxruntime/training/ortmodule/_utils.py:319:_forward (_forward)
_forward -- 100 100 -- 0.00146 0.29603 -- onnxruntime/training/ortmodule/_utils.py:298:_forward (_forward)
__call__ -- 100 100 -- 0.00016 0.00016 -- onnxruntime/training/ortmodule/_graph_execution_manager_factory.py:17:__call__ (__call__)
is_training -- 100 100 -- 0.00043 0.00059 -- onnxruntime/training/ortmodule/_torch_module_ort.py:43:is_training (is_training)
<built-in method torch._C.is_grad_enabled> -- 100 100 -- 0.00017 0.00017 -- ~:0:<built-in method torch._C.is_grad_enabled> (<built-in method torch._C.is_grad_enabled>) +++
forward -- 100 100 -- 0.00698 0.29381 -- onnxruntime/training/ortmodule/_training_manager.py:198:forward (forward)
is_pending -- 100 100 -- 0.00016 0.00016 -- onnxruntime/training/ortmodule/_fallback.py:153:is_pending (is_pending)
maybe_update_cache_before_run -- 100 100 -- 0.00030 0.00041 -- onnxruntime/training/ortmodule/_gradient_accumulation_manager.py:66:maybe_update_cache_before_run (maybe_update_cache_before_run)
enabled -- 100 100 -- 0.00011 0.00011 -- onnxruntime/training/ortmodule/_gradient_accumulation_manager.py:45:enabled (enabled) +++
is_set -- 200 200 -- 0.00084 0.00308 -- onnxruntime/training/ortmodule/_graph_execution_manager.py:59:is_set (is_set) +++
_combine_input_buffers_initializers -- 100 100 -- 0.00220 0.02179 -- onnxruntime/training/ortmodule/_io.py:132:_combine_input_buffers_initializers (_combine_input_buffers_initializers)
is_primitive_type -- 100 100 -- 0.00031 0.00031 -- onnxruntime/training/ortmodule/_io.py:65:is_primitive_type (is_primitive_type)
_expand_inputs -- 100 100 -- 0.00082 0.00528 -- onnxruntime/training/ortmodule/_io.py:140:_expand_inputs (_expand_inputs) +++
<dictcomp> -- 100 100 -- 0.00025 0.01374 -- onnxruntime/training/ortmodule/_io.py:165:<dictcomp> (<dictcomp>)
named_buffers -- 100 100 -- 0.00085 0.01348 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1586:named_buffers (named_buffers)
_named_members -- 100 100 -- 0.00239 0.01263 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1501:_named_members (_named_members) +++
<method 'append' of 'list' objects> -- 100 100 -- 0.00007 0.00007 -- ~:0:<method 'append' of 'list' objects> (<method 'append' of 'list' objects>) +++
<method 'extend' of 'list' objects> -- 100 100 -- 0.00019 0.00019 -- ~:0:<method 'extend' of 'list' objects> (<method 'extend' of 'list' objects>)
unflatten_user_output -- 100 100 -- 0.00059 0.00120 -- onnxruntime/training/ortmodule/_io.py:275:unflatten_user_output (unflatten_user_output)
_replace_stub_with_tensor_value -- 100 100 -- 0.00052 0.00061 -- onnxruntime/training/ortmodule/_io.py:278:_replace_stub_with_tensor_value (_replace_stub_with_tensor_value)
<built-in method ...tins.isinstance> -- 100 100 -- 0.00009 0.00009 -- ~:0:<built-in method builtins.isinstance> (<built-in method builtins.isinstance>) +++
<built-in method apply> -- 100 100 -- 0.00932 0.26020 -- ~:0:<built-in method apply> (<built-in method apply>)
forward -- 100 100 -- 0.00488 0.25089 -- onnxruntime/training/ortmodule/_training_manager.py:76:forward (forward)
set_materialize_grads -- 100 100 -- 0.00029 0.00029 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/function.py:184:set_materialize_grads (set_materialize_grads)
is_set -- 100 100 -- 0.00043 0.00144 -- onnxruntime/training/ortmodule/_graph_execution_manager.py:59:is_set (is_set) +++
execution_session_run_forward -- 100 100 -- 0.02782 0.24427 -- onnxruntime/training/ortmodule/_training_manager.py:38:execution_session_run_forward (execution_session_run_forward)
run_forward -- 100 100 -- 0.19012 0.19012 -- onnxruntime/training/ortmodule/_execution_agent.py:119:run_forward (run_forward)
extract_outputs...e_update_cache -- 100 100 -- 0.00268 0.01429 -- onnxruntime/training/ortmodule/_gradient_accumulation_manager.py:51:extract_outputs_and_maybe_update_cache (extract_outputs_and_maybe_update_cache)
enabled -- 100 100 -- 0.00012 0.00012 -- onnxruntime/training/ortmodule/_gradient_accumulation_manager.py:45:enabled (enabled) +++
<genexpr> -- 200 200 -- 0.00272 0.01056 -- onnxruntime/training/ortmodule/_gradient_accumulation_manager.py:58:<genexpr> (<genexpr>)
_ortvalue_t...rch_tensor -- 100 100 -- 0.00319 0.00784 -- onnxruntime/training/ortmodule/_utils.py:64:_ortvalue_to_torch_tensor (_ortvalue_to_torch_tensor) +++
<built-in met...uiltins.len> -- 100 100 -- 0.00093 0.00093 -- ~:0:<built-in method builtins.len> (<built-in method builtins.len>) +++
__init__ -- 100 100 -- 0.00040 0.00040 -- onnxruntime/training/ortmodule/_graph_execution_manager.py:42:__init__ (__init__)
<listcomp> -- 100 100 -- 0.00123 0.00123 -- onnxruntime/training/ortmodule/_training_manager.py:65:<listcomp> (<listcomp>)
_torch_tensor_to_dlpack -- 700 700 -- 0.00398 0.00844 -- onnxruntime/training/ortmodule/_utils.py:74:_torch_tensor_to_dlpack (_torch_tensor_to_dlpack) +++
<method 'is_con...Base' objects> -- 700 700 -- 0.00185 0.00185 -- ~:0:<method 'is_contiguous' of 'torch._C._TensorBase' objects> (<method 'is_contiguous' of 'torch._C._TensorBase' objects>) +++
<built-in method builtins.len> -- 100 100 -- 0.00012 0.00012 -- ~:0:<built-in method builtins.len> (<built-in method builtins.len>) +++
<built-in method torch._C._get_tracing_state> -- 1200 1200 -- 0.00653 0.00653 -- ~:0:<built-in method torch._C._get_tracing_state> (<built-in method torch._C._get_tracing_state>)
__getattr__ -- 1800 1800 -- 0.00681 0.00681 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1172:__getattr__ (__getattr__)
__setattr__ -- 24 24 -- 0.00028 0.00038 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1188:__setattr__ (__setattr__)
<method 'get' of 'dict' objects> -- 72 72 -- 0.00004 0.00004 -- ~:0:<method 'get' of 'dict' objects> (<method 'get' of 'dict' objects>) +++
<built-in method builtins.isinstance> -- 48 48 -- 0.00006 0.00006 -- ~:0:<built-in method builtins.isinstance> (<built-in method builtins.isinstance>) +++
_named_members -- 114 114 -- 0.00251 0.01298 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1501:_named_members (_named_members)
__hash__ -- 12 12 -- 0.00003 0.00004 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/_tensor.py:706:__hash__ (__hash__) +++
<lambda> -- 8 8 -- 0.00002 0.00002 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1559:<lambda> (<lambda>)
<lambda> -- 500 500 -- 0.00137 0.00238 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1607:<lambda> (<lambda>)
<method 'items' of 'colle...ns.OrderedDict' objects> -- 500 500 -- 0.00101 0.00101 -- ~:0:<method 'items' of 'collections.OrderedDict' objects> (<method 'items' of 'collections.OrderedDict' objects>) +++
named_modules -- 610 610 -- 0.00228 0.00797 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1668:named_modules (named_modules) +++
<method 'add' of 'set' objects> -- 12 12 -- 0.00003 0.00005 -- ~:0:<method 'add' of 'set' objects> (<method 'add' of 'set' objects>) +++
parameters -- 14 14 -- 0.00002 0.00040 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1514:parameters (parameters)
named_parameters -- 14 14 -- 0.00003 0.00038 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1538:named_parameters (named_parameters)
_named_members -- 14 14 -- 0.00012 0.00035 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1501:_named_members (_named_members) +++
named_modules -- 610 1722 -- 0.00723 0.00797 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1668:named_modules (named_modules)
named_modules -- 1112 512 -- 0.00495 0.00553 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1668:named_modules (named_modules) +++
<method 'items' of 'collections.OrderedDict' objects> -- 508 508 -- 0.00025 0.00025 -- ~:0:<method 'items' of 'collections.OrderedDict' objects> (<method 'items' of 'collections.OrderedDict' objects>) +++
<method 'add' of 'set' objects> -- 508 508 -- 0.00050 0.00050 -- ~:0:<method 'add' of 'set' objects> (<method 'add' of 'set' objects>) +++
wrapper -- 300 300 -- 0.01457 0.20824 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/optim/optimizer.py:83:wrapper (wrapper)
decorate_context -- 300 300 -- 0.00484 0.14451 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/grad_mode.py:24:decorate_context (decorate_context)
clone -- 300 300 -- 0.00158 0.00671 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/grad_mode.py:82:clone (clone)
__init__ -- 300 300 -- 0.00480 0.00512 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/grad_mode.py:121:__init__ (__init__) +++
__enter__ -- 300 300 -- 0.00307 0.00659 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/grad_mode.py:126:__enter__ (__enter__) +++
__exit__ -- 300 300 -- 0.00188 0.00402 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/grad_mode.py:130:__exit__ (__exit__) +++
step -- 300 300 -- 0.03184 0.12235 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/optim/sgd.py:109:step (step)
__hash__ -- 3612 3612 -- 0.00503 0.00975 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/_tensor.py:706:__hash__ (__hash__) +++
grad -- 3600 3600 -- 0.01192 0.01403 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/_tensor.py:1092:grad (grad) +++
sgd -- 300 300 -- 0.01165 0.06381 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/optim/_functional.py:158:sgd (sgd)
<method 'add_' of 'to...TensorBase' objects> -- 1800 1800 -- 0.05217 0.05217 -- ~:0:<method 'add_' of 'torch._C._TensorBase' objects> (<method 'add_' of 'torch._C._TensorBase' objects>)
<method 'append' of 'list' objects> -- 5400 5400 -- 0.00292 0.00292 -- ~:0:<method 'append' of 'list' objects> (<method 'append' of 'list' objects>) +++
__init__ -- 300 300 -- 0.00299 0.01704 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/profiler.py:426:__init__ (__init__) +++
__enter__ -- 300 300 -- 0.00486 0.02267 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/profiler.py:435:__enter__ (__enter__) +++
__exit__ -- 300 300 -- 0.00210 0.00713 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/profiler.py:439:__exit__ (__exit__) +++
<method 'format' of 'str' objects> -- 300 300 -- 0.00232 0.00232 -- ~:0:<method 'format' of 'str' objects> (<method 'format' of 'str' objects>) +++
zero_grad -- 300 300 -- 0.04380 0.13219 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/optim/optimizer.py:189:zero_grad (zero_grad)
grad -- 7200 7200 -- 0.02034 0.02462 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/_tensor.py:1092:grad (grad) +++
__init__ -- 300 300 -- 0.00351 0.01542 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/profiler.py:426:__init__ (__init__) +++
__enter__ -- 300 300 -- 0.00521 0.02196 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/profiler.py:435:__enter__ (__enter__) +++
__exit__ -- 300 300 -- 0.00219 0.00688 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/autograd/profiler.py:439:__exit__ (__exit__) +++
<method 'requires_grad_' of...h._C._TensorBase' objects> -- 1800 1800 -- 0.00596 0.00596 -- ~:0:<method 'requires_grad_' of 'torch._C._TensorBase' objects> (<method 'requires_grad_' of 'torch._C._TensorBase' objects>) +++
<method 'zero_' of 'torch._C._TensorBase' objects> -- 1800 1800 -- 0.01274 0.01274 -- ~:0:<method 'zero_' of 'torch._C._TensorBase' objects> (<method 'zero_' of 'torch._C._TensorBase' objects>)
<method 'get' of 'dict' objects> -- 300 300 -- 0.00037 0.00037 -- ~:0:<method 'get' of 'dict' objects> (<method 'get' of 'dict' objects>) +++
<built-in method builtins.hasattr> -- 300 300 -- 0.00045 0.00045 -- ~:0:<built-in method builtins.hasattr> (<built-in method builtins.hasattr>) +++
__init__ -- 2 2 -- 0.00003 0.00085 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/optim/sgd.py:88:__init__ (__init__)
__init__ -- 2 2 -- 0.00008 0.00082 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/optim/optimizer.py:33:__init__ (__init__)
parameters -- 7 7 -- 0.00001 0.00021 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1514:parameters (parameters) +++
_hook_for_profile -- 2 2 -- 0.00003 0.00004 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/optim/optimizer.py:78:_hook_for_profile (_hook_for_profile)
add_param_group -- 2 2 -- 0.00016 0.00025 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/optim/optimizer.py:243:add_param_group (add_param_group)
__hash__ -- 24 24 -- 0.00003 0.00005 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/_tensor.py:706:__hash__ (__hash__) +++
<method 'setdefault' of 'dict' objects> -- 12 12 -- 0.00001 0.00001 -- ~:0:<method 'setdefault' of 'dict' objects> (<method 'setdefault' of 'dict' objects>) +++
<built-in method builtins.isinstance> -- 18 18 -- 0.00002 0.00002 -- ~:0:<built-in method builtins.isinstance> (<built-in method builtins.isinstance>) +++
parameters -- 7 7 -- 0.00001 0.00021 -- onnxruntime/training/ortmodule/ortmodule.py:233:parameters (parameters)
parameters -- 7 7 -- 0.00001 0.00020 -- onnxruntime/training/ortmodule/_torch_module_ort.py:91:parameters (parameters)
parameters -- 7 7 -- 0.00001 0.00019 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/module.py:1514:parameters (parameters) +++
enabled -- 200 200 -- 0.00023 0.00023 -- onnxruntime/training/ortmodule/_gradient_accumulation_manager.py:45:enabled (enabled)
is_set -- 400 400 -- 0.00188 0.00708 -- onnxruntime/training/ortmodule/_graph_execution_manager.py:59:is_set (is_set)
__contains__ -- 400 400 -- 0.00104 0.00121 -- /usr/local/lib/python3.9/enum.py:748:__contains__ (__contains__) +++
is_disabled -- 400 400 -- 0.00156 0.00399 -- onnxruntime/training/ortmodule/_graph_execution_manager.py:67:is_disabled (is_disabled)
__contains__ -- 400 400 -- 0.00205 0.00243 -- /usr/local/lib/python3.9/enum.py:748:__contains__ (__contains__) +++
_expand_inputs -- 100 200 -- 0.00138 0.00528 -- onnxruntime/training/ortmodule/_io.py:140:_expand_inputs (_expand_inputs)
_expand_inputs -- 100 100 -- 0.00056 0.00196 -- onnxruntime/training/ortmodule/_io.py:140:_expand_inputs (_expand_inputs) +++
<method 'append' of 'list' objects> -- 100 100 -- 0.00007 0.00007 -- ~:0:<method 'append' of 'list' objects> (<method 'append' of 'list' objects>) +++
<built-in method builtins.isinstance> -- 500 500 -- 0.00111 0.00382 -- ~:0:<built-in method builtins.isinstance> (<built-in method builtins.isinstance>) +++
_ortvalue_to_torch_tensor -- 800 800 -- 0.01862 0.04312 -- onnxruntime/training/ortmodule/_utils.py:64:_ortvalue_to_torch_tensor (_ortvalue_to_torch_tensor)
_torch_tensor_from_dl_pack -- 800 800 -- 0.01171 0.02450 -- onnxruntime/training/ortmodule/_utils.py:59:_torch_tensor_from_dl_pack (_torch_tensor_from_dl_pack)
from_dlpack -- 800 800 -- 0.00220 0.01278 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/utils/dlpack.py:47:from_dlpack (from_dlpack)
<built-in method torch._C._from_dlpack> -- 800 800 -- 0.00934 0.00934 -- ~:0:<built-in method torch._C._from_dlpack> (<built-in method torch._C._from_dlpack>)
<built-in method builtins.hasattr> -- 800 800 -- 0.00124 0.00124 -- ~:0:<built-in method builtins.hasattr> (<built-in method builtins.hasattr>) +++
_torch_tensor_to_dlpack -- 800 800 -- 0.00488 0.01009 -- onnxruntime/training/ortmodule/_utils.py:74:_torch_tensor_to_dlpack (_torch_tensor_to_dlpack)
<built-in method torch._C._to_dlpack> -- 800 800 -- 0.00349 0.00349 -- ~:0:<built-in method torch._C._to_dlpack> (<built-in method torch._C._to_dlpack>)
<method 'is_contiguous' of ...h._C._TensorBase' objects> -- 800 800 -- 0.00171 0.00171 -- ~:0:<method 'is_contiguous' of 'torch._C._TensorBase' objects> (<method 'is_contiguous' of 'torch._C._TensorBase' objects>) +++
<built-in method builtins.len> -- 716 716 -- 0.00188 0.00188 -- ~:0:<built-in method builtins.len> (<built-in method builtins.len>)
<method 'append' of 'list' objects> -- 6902 6902 -- 0.00441 0.00441 -- ~:0:<method 'append' of 'list' objects> (<method 'append' of 'list' objects>)
<method 'requires_grad_' of 'torch._C._TensorBase' objects> -- 1804 1804 -- 0.00598 0.00598 -- ~:0:<method 'requires_grad_' of 'torch._C._TensorBase' objects> (<method 'requires_grad_' of 'torch._C._TensorBase' objects>)
<method 'to' of 'torch._C._TensorBase' objects> -- 28 28 -- 0.00017 0.00017 -- ~:0:<method 'to' of 'torch._C._TensorBase' objects> (<method 'to' of 'torch._C._TensorBase' objects>)
<method 'format' of 'str' objects> -- 602 602 -- 0.01786 0.01786 -- ~:0:<method 'format' of 'str' objects> (<method 'format' of 'str' objects>)
<method 'get' of 'dict' objects> -- 404 404 -- 0.00044 0.00044 -- ~:0:<method 'get' of 'dict' objects> (<method 'get' of 'dict' objects>)
<built-in method builtins.hasattr> -- 1154 1154 -- 0.00175 0.00175 -- ~:0:<built-in method builtins.hasattr> (<built-in method builtins.hasattr>)
<built-in method torch._C._has_torch_function_unary> -- 14808 14808 -- 0.00864 0.00864 -- ~:0:<built-in method torch._C._has_torch_function_unary> (<built-in method torch._C._has_torch_function_unary>)
<built-in method builtins.isinstance> -- 2418 2418 -- 0.00378 0.00654 -- ~:0:<built-in method builtins.isinstance> (<built-in method builtins.isinstance>)
__instancecheck__ -- 306 306 -- 0.00053 0.00275 -- /usr/local/lib/python3.9/abc.py:96:__instancecheck__ (__instancecheck__)
<built-in method _abc._abc_instancecheck> -- 306 306 -- 0.00125 0.00222 -- ~:0:<built-in method _abc._abc_instancecheck> (<built-in method _abc._abc_instancecheck>)
__subclasscheck__ -- 100 100 -- 0.00015 0.00098 -- /usr/local/lib/python3.9/abc.py:100:__subclasscheck__ (__subclasscheck__)
<built-in method _abc._abc_subclasscheck> -- 100 100 -- 0.00067 0.00083 -- ~:0:<built-in method _abc._abc_subclasscheck> (<built-in method _abc._abc_subclasscheck>)
__subclasshook__ -- 100 100 -- 0.00017 0.00017 -- /usr/local/lib/python3.9/_collections_abc.py:311:__subclasshook__ (__subclasshook__)
<method 'items' of 'collections.OrderedDict' objects> -- 1043 1043 -- 0.00128 0.00128 -- ~:0:<method 'items' of 'collections.OrderedDict' objects> (<method 'items' of 'collections.OrderedDict' objects>)
<built-in method builtins.getattr> -- 338 338 -- 0.00055 0.00055 -- ~:0:<built-in method builtins.getattr> (<built-in method builtins.getattr>)
<method 'setdefault' of 'dict' objects> -- 18 18 -- 0.00002 0.00002 -- ~:0:<method 'setdefault' of 'dict' objects> (<method 'setdefault' of 'dict' objects>)
<built-in method torch._C.is_grad_enabled> -- 1072 1072 -- 0.00169 0.00169 -- ~:0:<built-in method torch._C.is_grad_enabled> (<built-in method torch._C.is_grad_enabled>)
<method 'add' of 'set' objects> -- 527 527 -- 0.00053 0.00056 -- ~:0:<method 'add' of 'set' objects> (<method 'add' of 'set' objects>)
__hash__ -- 12 12 -- 0.00002 0.00003 -- onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/_tensor.py:706:__hash__ (__hash__) +++
<method 'is_contiguous' of 'torch._C._TensorBase' objects> -- 1600 1600 -- 0.00407 0.00407 -- ~:0:<method 'is_contiguous' of 'torch._C._TensorBase' objects> (<method 'is_contiguous' of 'torch._C._TensorBase' objects>)
<built-in method posix.fspath> -- 24 24 -- 0.00001 0.00001 -- ~:0:<built-in method posix.fspath> (<built-in method posix.fspath>)
<built-in method _thread.get_ident> -- 12 12 -- 0.00001 0.00001 -- ~:0:<built-in method _thread.get_ident> (<built-in method _thread.get_ident>)
<method 'rfind' of 'str' objects> -- 18 18 -- 0.00003 0.00003 -- ~:0:<method 'rfind' of 'str' objects> (<method 'rfind' of 'str' objects>)
if GPU is available#
if get_device().upper() == 'GPU':
device = torch.device('cuda:0')
benches.append(benchmark(model_torch, model_ort, device, name='NN-GPU',
max_iter=max_iter))
Linear Regression#
class LinearRegressionNet(torch.nn.Module):
def __init__(self, D_in, D_out):
super(LinearRegressionNet, self).__init__()
self.linear = torch.nn.Linear(D_in, D_out)
def forward(self, x):
return self.linear(x)
d_in, d_out, N = X.shape[1], 1, X.shape[0]
model_torch = LinearRegressionNet(d_in, d_out)
model_ort = ORTModule(LinearRegressionNet(d_in, d_out))
device = torch.device('cpu')
benches.append(benchmark(model_torch, model_ort, device, name='LR-CPU',
max_iter=max_iter))
if get_device().upper() == 'GPU':
device = torch.device('cuda:0')
benches.append(benchmark(model_torch, model_ort, device, name='LR-GPU',
max_iter=max_iter))
######################################
# GPU profiling
# +++++++++++++
if get_device().upper() == 'GPU':
ps = profile(lambda: benchmark(
model_torch, model_ort, device, name='LR-GPU',
max_iter=max_iter))[0]
root, nodes = profile2graph(ps, clean_text=clean_name)
text = root.to_text()
print(text)
Out:
[benchmark] LR-CPU
somewhere/workspace/onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/torch/nn/modules/loss.py:529: UserWarning: Using a target size (torch.Size([750])) that is different to the input size (torch.Size([750, 1])). This will likely lead to incorrect results due to broadcasting. Please ensure they have the same size.
return F.mse_loss(input, target, reduction=self.reduction)
[benchmark] torch=200 iterations - 12.983033951371908 seconds
somewhere/workspace/onnxcustom/onnxcustom_UT_39_std/_venv/lib/python3.9/site-packages/onnxruntime/training/ortmodule/_training_manager.py:216: UserWarning: Fast path enabled - skipping checks. Rebuild graph: True, Execution agent: True, Device check: True
warnings.warn(f"Fast path enabled - skipping checks."
[benchmark] onxrt=100 iteration - 9.299985624849796 seconds
Graphs#
Dataframe first.
df = DataFrame(benches).set_index('name')
df
text output
print(df)
Out:
torch ort iter_torch iter_ort
name
NN-CPU 2.774660 9.256914 200 100
LR-CPU 12.983034 9.299986 200 100
Graphs.
print(df.columns)
fig, ax = plt.subplots(1, 1, figsize=(4, 4))
df[['torch', 'ort']].plot.bar(title="Processing time", ax=ax)
ax.tick_params(axis='x', rotation=30)
fig.savefig(__file__ + ".png")
# plt.show()
Traceback (most recent call last):
File "somewhere/workspace/onnxcustom/onnxcustom_UT_39_std/_doc/examples/plot_orttraining_benchmark_torch.py", line 276, in <module>
fig.savefig(__file__ + ".png")
NameError: name '__file__' is not defined
Total running time of the script: ( 1 minutes 5.576 seconds)