Note
Go to the end to download the full example code.
201: Evaluate different ways to export a torch model to ONNX¶
The example evaluates the performance of onnxruntime of a simple torch model after it was converted into ONNX through different processes:
TorchScript-based ONNX Exporter, let’s call it script
TorchDynamo-based ONNX Exporter, let’s call it dynamo
if available, the previous model but optimized, dynopt
a custom exporter cus_p0, this exporter supports a very limited set of models, as dynamo, it relies on torch.fx but the design is closer to what tensorflow-onnx does.
the same exporter but unused nodes were removed and constants were folded, cus_p2
To run the script:
python _doc/examples/plot_torch_export --help
The script takes around 12 minutes with a larger models.
Some helpers¶
from experimental_experiment.args import get_parsed_args
script_args = get_parsed_args(
"plot_torch_export",
description=__doc__,
scenarios={
"small": "small model to test",
"middle": "55Mb model",
"large": "1Gb model",
},
warmup=5,
repeat=5,
maxtime=(
2,
"maximum time to run a model to measure the computation time, "
"it is 0.1 when scenario is small",
),
expose="scenarios,repeat,warmup",
)
import contextlib
import itertools
import os
import platform
import pprint
import multiprocessing
import time
import cProfile
import pstats
import io
import warnings
import logging
from pstats import SortKey
try:
with warnings.catch_warnings():
warnings.simplefilter("ignore")
import onnxruntime
has_cuda = "CUDAExecutionProvider" in onnxruntime.get_available_providers()
except ImportError:
print("onnxruntime not available.")
import sys
sys.exit(0)
import numpy as np
import matplotlib.pyplot as plt
import pandas
import onnx
from onnx_array_api.profiling import profile2graph
import torch
from torch import nn
import torch.nn.functional as F
import experimental_experiment
from experimental_experiment.torch_interpreter import to_onnx
from experimental_experiment.xbuilder import OptimizationOptions
from experimental_experiment.plotting.memory import memory_peak_plot
from experimental_experiment.ext_test_case import measure_time, get_figure
from experimental_experiment.memory_peak import start_spying_on
from experimental_experiment.ext_test_case import unit_test_going
from experimental_experiment.helpers import pretty_onnx
from tqdm import tqdm
has_cuda = has_cuda and torch.cuda.device_count() > 0
logging.disable(logging.ERROR)
def system_info():
obs = {}
obs["processor"] = platform.processor()
obs["cores"] = multiprocessing.cpu_count()
try:
obs["cuda"] = 1 if torch.cuda.device_count() > 0 else 0
obs["cuda_count"] = torch.cuda.device_count()
obs["cuda_name"] = torch.cuda.get_device_name()
obs["cuda_capa"] = torch.cuda.get_device_capability()
except (RuntimeError, AssertionError):
# no cuda
pass
return obs
pprint.pprint(system_info())
{'cores': 20,
'cuda': 1,
'cuda_capa': (8, 9),
'cuda_count': 1,
'cuda_name': 'NVIDIA GeForce RTX 4060 Laptop GPU',
'processor': 'x86_64'}
Scripts arguments
if script_args.scenario in (None, "small"):
script_args.maxtime = 0.1
if unit_test_going():
script_args.warmup = 1
script_args.repeat = 1
script_args.maxtime = 0.1
script_args.scenario = "small"
print(f"scenario={script_args.scenario or 'small'}")
print(f"warmup={script_args.warmup}")
print(f"repeat={script_args.repeat}")
print(f"maxtime={script_args.maxtime}")
scenario=small
warmup=5
repeat=5
maxtime=0.1
The model¶
A simple model to convert.
class MyModelClass(nn.Module):
def __init__(self, scenario=script_args.scenario):
super().__init__()
if scenario == "middle":
self.large = False
self.conv1 = nn.Conv2d(1, 128, 5)
self.conv2 = nn.Conv2d(128, 16, 5)
self.fc1 = nn.Linear(13456, 1024)
self.fcs = []
self.fc2 = nn.Linear(1024, 128)
self.fc3 = nn.Linear(128, 10)
elif scenario in (None, "small"):
self.large = False
self.conv1 = nn.Conv2d(1, 16, 5)
self.conv2 = nn.Conv2d(16, 16, 5)
self.fc1 = nn.Linear(16, 512)
self.fcs = []
self.fc2 = nn.Linear(512, 128)
self.fc3 = nn.Linear(128, 10)
elif scenario in (None, "large"):
self.large = True
self.conv1 = nn.Conv2d(1, 128, 5)
self.conv2 = nn.Conv2d(128, 16, 5)
self.fc1 = nn.Linear(13456, 4096)
# torch script does not support loops.
self.fca = nn.Linear(4096, 4096)
self.fcb = nn.Linear(4096, 4096)
self.fcc = nn.Linear(4096, 4096)
self.fcd = nn.Linear(4096, 4096)
self.fce = nn.Linear(4096, 4096)
self.fcf = nn.Linear(4096, 4096)
self.fcg = nn.Linear(4096, 4096)
self.fch = nn.Linear(4096, 4096)
self.fci = nn.Linear(4096, 4096)
self.fck = nn.Linear(4096, 4096)
self.fcl = nn.Linear(4096, 4096)
self.fcm = nn.Linear(4096, 4096)
self.fcn = nn.Linear(4096, 4096)
# end of the unfolded loop.
self.fc2 = nn.Linear(4096, 128)
self.fc3 = nn.Linear(128, 10)
else:
raise ValueError(f"Unsupported scenario={scenario!r}.")
def forward(self, x):
x = F.max_pool2d(F.relu(self.conv1(x)), (2, 2))
x = F.max_pool2d(F.relu(self.conv2(x)), 2)
x = torch.flatten(x, 1)
x = F.relu(self.fc1(x))
if self.large:
# loop
x = F.relu(self.fca(x))
x = F.relu(self.fcb(x))
x = F.relu(self.fcc(x))
x = F.relu(self.fcd(x))
x = F.relu(self.fce(x))
x = F.relu(self.fcf(x))
x = F.relu(self.fcg(x))
x = F.relu(self.fch(x))
x = F.relu(self.fci(x))
x = F.relu(self.fck(x))
x = F.relu(self.fcl(x))
x = F.relu(self.fcm(x))
x = F.relu(self.fcn(x))
# end of the loop
x = F.relu(self.fc2(x))
x = self.fc3(x)
return x
def create_model_and_input(scenario=script_args.scenario):
if scenario == "middle":
shape = [1, 1, 128, 128]
elif scenario in (None, "small"):
shape = [1, 1, 16, 16]
elif scenario == "large":
shape = [1, 1, 128, 128]
else:
raise ValueError(f"Unsupported scenario={scenario!r}.")
input_tensor = torch.rand(*shape).to(torch.float32)
model = MyModelClass(scenario=scenario)
assert model(input_tensor) is not None
return model, input_tensor
def torch_model_size(model):
size_model = 0
for param in model.parameters():
size = param.numel() * torch.finfo(param.data.dtype).bits / 8
size_model += size
return size_model
model, input_tensor = create_model_and_input()
model_size = torch_model_size(model)
print(f"model size={model_size / 2 ** 20} Mb")
model size=0.31467437744140625 Mb
The exporters¶
def export_script(filename, model, *args):
with contextlib.redirect_stdout(io.StringIO()):
with warnings.catch_warnings():
warnings.simplefilter("ignore")
torch.onnx.export(model, *args, filename, input_names=["input"])
def export_dynamo(filename, model, *args):
with contextlib.redirect_stdout(io.StringIO()):
with warnings.catch_warnings():
warnings.simplefilter("ignore")
export_output = torch.onnx.export(model, args, dynamo=True)
export_output.save(filename)
def export_dynopt(filename, model, *args):
with contextlib.redirect_stdout(io.StringIO()):
with warnings.catch_warnings():
warnings.simplefilter("ignore")
export_output = torch.onnx.export(model, args, dynamo=True)
model_onnx = export_output.model_proto
from experimental_experiment.convert.convert_helper import (
optimize_model_proto_oxs,
)
optimized_model = optimize_model_proto_oxs(model_onnx)
with open(filename, "wb") as f:
f.write(optimized_model.SerializeToString())
def export_cus_p0(filename, model, *args):
onx = to_onnx(model, tuple(args), input_names=["input"])
with open(filename, "wb") as f:
f.write(onx.SerializeToString())
def export_cus_p2(filename, model, *args):
onx = to_onnx(
model,
tuple(args),
input_names=["input"],
options=OptimizationOptions(
remove_unused=True,
constant_folding=True,
),
)
with open(filename, "wb") as f:
f.write(onx.SerializeToString())
Let’s check they are working.
export_functions = [
export_script,
export_dynamo,
export_dynopt,
export_cus_p0,
export_cus_p2,
]
exporters = {f.__name__.replace("export_", ""): f for f in export_functions}
supported_exporters = {}
for k, v in exporters.items():
print(f"run exporter {k}")
filename = f"plot_torch_export_{k}.onnx"
try:
v(filename, model, input_tensor)
except Exception as e:
print(f"skipped due to {str(e)[:1000]}")
continue
supported_exporters[k] = v
print(f"done. size={os.stat(filename).st_size / 2 ** 20:1.0f} Mb")
run exporter script
done. size=0 Mb
run exporter dynamo
done. size=0 Mb
run exporter dynopt
done. size=0 Mb
run exporter cus_p0
done. size=0 Mb
run exporter cus_p2
done. size=0 Mb
Exporter memory¶
def flatten(ps):
obs = ps["cpu"].to_dict(unit=2**20)
if "gpus" in ps:
for i, g in enumerate(ps["gpus"]):
for k, v in g.to_dict(unit=2**20).items():
obs[f"gpu{i}_{k}"] = v
return obs
data = []
for k, v in supported_exporters.items():
print(f"run exporter for memory {k}")
filename = f"plot_torch_export_{k}.onnx"
if has_cuda:
torch.cuda.set_device(0)
stat = start_spying_on(cuda=1 if has_cuda else 0)
v(filename, model, input_tensor)
obs = flatten(stat.stop())
print("done.")
onx = onnx.load(filename)
obs.update(dict(nodes=len(onx.graph.node), export=k))
data.append(obs)
stat = start_spying_on(cuda=1 if has_cuda else 0)
exported_mod = torch.export.export(model, (input_tensor,))
obs = flatten(stat.stop())
obs.update(dict(export="torch.fx"))
data.append(obs)
run exporter for memory script
done.
run exporter for memory dynamo
done.
run exporter for memory dynopt
done.
run exporter for memory cus_p0
done.
run exporter for memory cus_p2
done.
The result.
df1 = pandas.DataFrame(data)
df1.to_csv("plot_torch_export_memory.csv", index=False)
df1.to_excel("plot_torch_export_memory.xlsx", index=False)
print(df1)
ax = memory_peak_plot(
data,
bars=[model_size * i / 2**20 for i in range(1, 5)],
suptitle=f"Memory Consumption of the Export\nmodel size={model_size / 2**20:1.0f} Mb",
)
get_figure(ax).savefig("plot_torch_export_memory.png")
peak mean n begin end gpu0_peak gpu0_mean gpu0_n gpu0_begin gpu0_end nodes export
0 1991.621094 1990.958984 12 1991.617188 1991.621094 449.617188 449.617188 12 449.617188 449.617188 12.0 script
1 1991.687500 1991.652860 106 1991.621094 1991.687500 449.617188 449.617188 106 449.617188 449.617188 13.0 dynamo
2 1991.687500 1991.687500 149 1991.687500 1991.687500 449.617188 449.617188 149 449.617188 449.617188 13.0 dynopt
3 1991.695312 1991.688519 23 1991.687500 1991.695312 449.617188 449.617188 23 449.617188 449.617188 15.0 cus_p0
4 1991.699219 1989.976412 26 1991.695312 1957.074219 449.617188 449.617188 26 449.617188 449.617188 12.0 cus_p2
5 1957.253906 1957.249844 25 1957.246094 1957.253906 449.617188 449.617188 25 449.617188 449.617188 NaN torch.fx
Exporter speed¶
data = []
for k, v in supported_exporters.items():
print(f"run exporter {k}")
filename = f"plot_torch_export_{k}.onnx"
times = []
for _ in range(script_args.repeat):
begin = time.perf_counter()
v(filename, model, input_tensor)
duration = time.perf_counter() - begin
times.append(duration)
onx = onnx.load(filename)
print("done.")
data.append(
dict(
export=k,
time=np.mean(times),
min=min(times),
max=max(times),
first=times[0],
last=times[-1],
std=np.std(times),
nodes=len(onx.graph.node),
)
)
run exporter script
done.
run exporter dynamo
done.
run exporter dynopt
done.
run exporter cus_p0
done.
run exporter cus_p2
done.
The last export to measure time torch spends in export the model before any other export can begin the translation except the first one.
times = []
for _ in range(script_args.repeat):
begin = time.perf_counter()
exported_mod = torch.export.export(model, (input_tensor,))
duration = time.perf_counter() - begin
times.append(duration)
data.append(
dict(
export="torch.fx",
time=np.mean(times),
min=min(times),
max=max(times),
first=times[0],
last=times[-1],
std=np.std(times),
nodes=len(onx.graph.node),
)
)
The result.
df1 = pandas.DataFrame(data)
df1.to_csv("plot_torch_export_time.csv", index=False)
df1.to_excel("plot_torch_export_time.xlsx", index=False)
print(df1)
fig, ax = plt.subplots(1, 1)
dfi = df1[["export", "time", "std"]].set_index("export")
dfi["time"].plot.bar(ax=ax, title="Export time", yerr=dfi["std"], rot=30)
fig.tight_layout()
fig.savefig("plot_torch_export_time.png")
export time min max first last std nodes
0 script 0.071847 0.032638 0.175475 0.175475 0.035229 0.055204 12
1 dynamo 1.299576 1.013765 1.656289 1.656289 1.105589 0.226074 13
2 dynopt 1.265269 1.048148 1.540096 1.261092 1.048148 0.190792 13
3 cus_p0 0.260758 0.149327 0.392067 0.149327 0.392067 0.090509 15
4 cus_p2 0.203443 0.160868 0.250119 0.250119 0.176179 0.032620 12
5 torch.fx 0.129058 0.119741 0.146721 0.121158 0.146721 0.010517 12
Exporter Profiling¶
def clean_text(text):
pathes = [
os.path.abspath(os.path.normpath(os.path.join(os.path.dirname(torch.__file__), ".."))),
os.path.abspath(os.path.normpath(os.path.join(os.path.dirname(onnx.__file__), ".."))),
os.path.abspath(
os.path.normpath(
os.path.join(os.path.dirname(experimental_experiment.__file__), "..")
)
),
]
for p in pathes:
text = text.replace(p, "")
text = text.replace("experimental_experiment", "experimental_experiment".upper())
return text
def profile_function(name, export_function, verbose=False):
print(f"profile {name}: {export_function}")
pr = cProfile.Profile()
pr.enable()
for _ in range(script_args.repeat):
export_function("dummyc.onnx", model, input_tensor)
pr.disable()
s = io.StringIO()
sortby = SortKey.CUMULATIVE
ps = pstats.Stats(pr, stream=s).sort_stats(sortby)
ps.print_stats()
raw = s.getvalue()
text = "\n".join(raw.split("\n")[:200])
if verbose:
print(text)
with open(f"plot_torch_export_profile_{name}.txt", "w") as f:
f.write(raw)
root, nodes = profile2graph(ps, clean_text=clean_text)
text = root.to_text()
with open(f"plot_torch_export_profile_{name}_h.txt", "w") as f:
f.write(text)
print("done.")
profile_function("custom0", export_cus_p0, True)
profile_function("custom2", export_cus_p2)
profile custom0: <function export_cus_p0 at 0x7f50d07a0220>
1041763 function calls (1012599 primitive calls) in 2.044 seconds
Ordered by: cumulative time
ncalls tottime percall cumtime percall filename:lineno(function)
60 0.002 0.000 1.959 0.033 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/symbolic_convert.py:966(call_function)
25 0.002 0.000 1.845 0.074 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/variables/nn_module.py:371(call_function)
1080/690 0.004 0.000 0.516 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/utils/_stats.py:22(wrapper)
55/10 0.004 0.000 0.414 0.041 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/experimental/symbolic_shapes.py:7081(run_node)
65 0.001 0.000 0.409 0.006 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/variables/builder.py:2209(wrap_fx_proxy)
65 0.000 0.000 0.408 0.006 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/variables/builder.py:2282(wrap_fx_proxy_cls)
60 0.002 0.000 0.401 0.007 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/variables/builder.py:2366(_wrap_fx_proxy)
90 0.002 0.000 0.384 0.004 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/utils.py:2578(wrap_fake_exception)
870 0.007 0.000 0.382 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_tensor.py:1257(__torch_dispatch__)
60 0.003 0.000 0.382 0.006 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/utils.py:2993(get_fake_value)
870 0.018 0.000 0.374 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_tensor.py:1782(dispatch)
485 0.007 0.000 0.350 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_tensor.py:1342(_cached_dispatch_impl)
50 0.002 0.000 0.310 0.006 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/nn/parameter.py:63(__deepcopy__)
280/53 0.003 0.000 0.291 0.005 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/symbolic_convert.py:1034(step)
160/110 0.080 0.001 0.289 0.003 {method 'clone' of 'torch._C.TensorBase' objects}
55/11 0.001 0.000 0.279 0.025 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/symbolic_convert.py:741(wrapper)
55/11 0.002 0.000 0.278 0.025 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/symbolic_convert.py:2488(CALL)
55/11 0.001 0.000 0.277 0.025 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/symbolic_convert.py:2447(_call)
1315 0.005 0.000 0.262 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/experimental/proxy_tensor.py:1229(__torch_function__)
5 0.001 0.000 0.258 0.052 /home/xadupre/github/experimental-experiment/experimental_experiment/xbuilder/graph_builder.py:5046(to_onnx)
435/325 0.001 0.000 0.257 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_ops.py:755(__call__)
5160/2130 0.018 0.000 0.256 0.000 /usr/lib/python3.12/copy.py:118(deepcopy)
25 0.001 0.000 0.256 0.010 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/utils.py:2589(deepcopy_to_fake_tensor)
1315 0.003 0.000 0.252 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/experimental/proxy_tensor.py:1258(__torch_function__)
135/26 0.002 0.000 0.250 0.010 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/interpreter.py:218(run_node)
595/235 0.004 0.000 0.247 0.001 /usr/lib/python3.12/copy.py:247(_reconstruct)
60 0.002 0.000 0.245 0.004 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_ops.py:863(handler)
60 0.012 0.000 0.240 0.004 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_library/utils.py:281(handle_dispatch_mode)
255/120 0.019 0.000 0.238 0.002 /usr/lib/python3.12/copy.py:217(_deepcopy_dict)
25 0.000 0.000 0.236 0.009 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/utils.py:2591(<lambda>)
60 0.001 0.000 0.226 0.004 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/experimental/proxy_tensor.py:1327(__torch_dispatch__)
60 0.005 0.000 0.224 0.004 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/experimental/proxy_tensor.py:760(proxy_call)
20/4 0.001 0.000 0.201 0.050 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/interpreter.py:342(call_module)
60 0.002 0.000 0.200 0.003 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph_module.py:800(recompile)
250 0.003 0.000 0.190 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_tensor.py:2719(__torch_function__)
485 0.004 0.000 0.183 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_tensor.py:1716(_output_from_cache_entry)
515 0.021 0.000 0.179 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_tensor.py:1650(_get_output_tensor_from_cache_entry)
65 0.001 0.000 0.175 0.003 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph.py:1562(python_code)
5 0.001 0.000 0.154 0.031 /home/xadupre/github/experimental-experiment/experimental_experiment/xbuilder/graph_builder.py:5579(optimize)
485 0.006 0.000 0.153 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_tensor.py:1385(_cache_key)
5 0.000 0.000 0.148 0.030 /home/xadupre/github/experimental-experiment/experimental_experiment/xbuilder/graph_builder.py:5901(optimize_with_patterns)
50 0.001 0.000 0.148 0.003 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/_symbolic_trace.py:480(call_module)
7485/7195 0.011 0.000 0.147 0.000 {built-in method builtins.next}
5 0.010 0.002 0.147 0.029 /home/xadupre/github/experimental-experiment/experimental_experiment/xoptim/graph_builder_optim.py:1065(optimize)
25 0.000 0.000 0.146 0.006 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/_symbolic_trace.py:801(module_call_wrapper)
45 0.000 0.000 0.146 0.003 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/nn/modules/linear.py:124(forward)
25 0.000 0.000 0.146 0.006 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/experimental/proxy_tensor.py:1739(call_module)
75/45 0.013 0.000 0.146 0.003 {built-in method torch._C._nn.linear}
65 0.002 0.000 0.142 0.002 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph.py:1639(_python_code)
65 0.017 0.000 0.140 0.002 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph.py:397(_gen_python_code)
2070/535 0.018 0.000 0.140 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_tensor.py:1461(_prep_args_for_hash)
25 0.000 0.000 0.138 0.006 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/_symbolic_trace.py:803(forward)
100/60 0.001 0.000 0.137 0.002 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/nn/functional.py:1693(relu)
60 0.008 0.000 0.135 0.002 {built-in method torch.relu}
150 0.003 0.000 0.127 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_tensor.py:758(__torch_dispatch__)
35 0.000 0.000 0.126 0.004 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/interpreter.py:296(call_function)
60 0.000 0.000 0.115 0.002 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/utils.py:3054(<lambda>)
60 0.001 0.000 0.115 0.002 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/utils.py:3144(run_node)
5 0.000 0.000 0.113 0.023 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_functorch/functional_call.py:11(functional_call)
5 0.000 0.000 0.113 0.023 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/nn/utils/stateless.py:246(_functional_call)
5 0.000 0.000 0.110 0.022 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/_lazy_graph_module.py:115(_lazy_forward)
35 0.002 0.000 0.107 0.003 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/variables/torch.py:970(call_function)
60/30 0.000 0.000 0.106 0.004 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/overrides.py:1670(handle_torch_function)
1115 0.029 0.000 0.103 0.000 /home/xadupre/github/experimental-experiment/experimental_experiment/xoptim/patterns_api.py:128(enumerate_matches)
50 0.003 0.000 0.096 0.002 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/nn/parameter.py:40(__new__)
5 0.001 0.000 0.094 0.019 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/export/_trace.py:437(_produce_aten_artifact)
175/125 0.004 0.000 0.092 0.001 {method 'detach' of 'torch._C.TensorBase' objects}
3575/3430 0.004 0.000 0.092 0.000 /usr/lib/python3.12/contextlib.py:132(__enter__)
5 0.000 0.000 0.091 0.018 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph_module.py:829(call_wrapped)
5 0.000 0.000 0.091 0.018 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph_module.py:388(__call__)
105 0.002 0.000 0.091 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/meta_utils.py:847(meta_tensor)
30 0.000 0.000 0.087 0.003 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/nn/modules/conv.py:553(forward)
30 0.000 0.000 0.086 0.003 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/nn/modules/conv.py:536(_conv_forward)
50/30 0.007 0.000 0.086 0.003 {built-in method torch.conv2d}
5 0.001 0.000 0.085 0.017 /home/xadupre/github/experimental-experiment/experimental_experiment/xbuilder/graph_builder.py:4342(_build_initializers)
50 0.002 0.000 0.082 0.002 /home/xadupre/github/experimental-experiment/experimental_experiment/mini_onnx_builder.py:108(proto_from_array)
105 0.005 0.000 0.080 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/autograd/grad_mode.py:273(__exit__)
240 0.002 0.000 0.079 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/proxy.py:211(create_proxy)
590/520 0.005 0.000 0.078 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/nn/modules/module.py:1944(__setattr__)
25 0.001 0.000 0.076 0.003 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph_module.py:443(__init__)
193215/190695 0.063 0.000 0.075 0.000 {built-in method builtins.isinstance}
110 0.001 0.000 0.072 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/variables/base.py:484(build)
5 0.000 0.000 0.071 0.014 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/eval_frame.py:1191(rewrite_signature)
5 0.000 0.000 0.070 0.014 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/guards.py:1870(SHAPE_ENV)
5 0.064 0.013 0.070 0.014 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/guards.py:1280(add_python_lambda_leaf_guard_to_root)
110 0.001 0.000 0.070 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/variables/builder.py:397(__call__)
5 0.001 0.000 0.069 0.014 /home/xadupre/github/experimental-experiment/experimental_experiment/xbuilder/graph_builder.py:4780(process)
25 0.000 0.000 0.068 0.003 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph_module.py:556(graph)
120 0.002 0.000 0.066 0.001 /home/xadupre/github/experimental-experiment/experimental_experiment/torch_interpreter/interpreter.py:177(run_node)
65 0.004 0.000 0.066 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/variables/builder.py:534(_wrap)
735 0.022 0.000 0.066 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_tensor.py:671(__new__)
3575/3430 0.005 0.000 0.063 0.000 /usr/lib/python3.12/contextlib.py:141(__exit__)
1260 0.010 0.000 0.063 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph.py:627(emit_node)
5 0.001 0.000 0.061 0.012 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/guards.py:2751(build_guard_function)
635 0.014 0.000 0.060 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_tensor.py:968(extract_tensor_metadata)
250 0.004 0.000 0.059 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/proxy.py:145(create_node)
50/30 0.000 0.000 0.059 0.002 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_jit_internal.py:613(fn)
50/30 0.000 0.000 0.059 0.002 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/nn/functional.py:807(_max_pool2d)
635 0.018 0.000 0.058 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_tensor.py:948(_flatten_into)
30 0.004 0.000 0.057 0.002 {built-in method torch.max_pool2d}
65 0.000 0.000 0.057 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/experimental/proxy_tensor.py:592(track_tensor_tree)
120/65 0.001 0.000 0.055 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/experimental/proxy_tensor.py:614(wrap_with_proxy)
60 0.002 0.000 0.053 0.001 /home/xadupre/github/experimental-experiment/experimental_experiment/torch_interpreter/interpreter.py:1365(call_function)
5 0.000 0.000 0.051 0.010 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/symbolic_convert.py:2840(__init__)
260 0.004 0.000 0.050 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph.py:1104(create_node)
2920/135 0.008 0.000 0.050 0.000 /usr/lib/python3.12/ast.py:403(visit)
10 0.000 0.000 0.050 0.005 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/variables/lazy.py:64(realize)
105 0.010 0.000 0.049 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/autograd/grad_mode.py:269(__enter__)
55/5 0.001 0.000 0.047 0.009 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/variables/lazy.py:107(realize_all)
190 0.002 0.000 0.046 0.000 /home/xadupre/github/experimental-experiment/experimental_experiment/xbuilder/graph_builder_opset.py:115(make_node)
5 0.000 0.000 0.046 0.009 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/eval_frame.py:1128(transform)
5 0.001 0.000 0.046 0.009 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/interpreter.py:571(transform)
4770 0.018 0.000 0.045 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/node.py:873(__setattr__)
210 0.007 0.000 0.045 0.000 /home/xadupre/github/experimental-experiment/experimental_experiment/xbuilder/graph_builder.py:3609(make_node)
65 0.000 0.000 0.044 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/symbolic_convert.py:1934(LOAD_ATTR)
10 0.000 0.000 0.044 0.004 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_export/passes/replace_with_hop_pass_util.py:157(_replace_with_hop_pass_helper)
65 0.001 0.000 0.043 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/symbolic_convert.py:1927(_load_attr)
15 0.000 0.000 0.042 0.003 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/_lazy_graph_module.py:57(_make_graph_module)
5 0.000 0.000 0.041 0.008 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/symbolic_convert.py:3213(RETURN_VALUE)
5 0.000 0.000 0.041 0.008 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/symbolic_convert.py:3180(_return)
115 0.001 0.000 0.041 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/experimental/proxy_tensor.py:485(set_meta)
60 0.000 0.000 0.041 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/symbolic_convert.py:1904(LOAD_METHOD)
5 0.000 0.000 0.041 0.008 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/output_graph.py:977(compile_subgraph)
4365 0.012 0.000 0.040 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/meta_utils.py:172(is_sparse_any)
65 0.001 0.000 0.040 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/variables/builtin.py:1064(call_function)
65 0.000 0.000 0.039 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/variables/builtin.py:917(builtin_dispatch)
65 0.000 0.000 0.039 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/variables/builtin.py:837(call_self_handler)
8260/4895 0.022 0.000 0.039 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/node.py:903(map_aggregate)
65 0.001 0.000 0.039 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/variables/builtin.py:1731(call_getattr)
10 0.000 0.000 0.038 0.004 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/variables/lazy.py:22(realize)
65 0.005 0.000 0.037 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/output_graph.py:2002(create_proxy)
60 0.000 0.000 0.037 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph_module.py:91(_forward_from_src)
4340 0.005 0.000 0.037 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/node.py:894(map_arg)
60 0.001 0.000 0.037 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph_module.py:97(_method_from_src)
150 0.036 0.000 0.036 0.000 {built-in method builtins.compile}
60 0.000 0.000 0.036 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph_module.py:86(_exec_with_source)
360 0.001 0.000 0.036 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/guards.py:456(_ast_unparse)
5 0.000 0.000 0.036 0.007 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/metrics_context.py:52(__exit__)
5 0.001 0.000 0.036 0.007 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/utils.py:1425(record_compilation_metrics)
35475 0.026 0.000 0.035 0.000 {built-in method builtins.getattr}
60 0.001 0.000 0.035 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/output_graph.py:596(create_proxy)
5 0.001 0.000 0.035 0.007 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/output_graph.py:1321(compile_and_call_fx_graph)
105 0.000 0.000 0.035 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/guards.py:1277(get_guard_manager)
360 0.001 0.000 0.034 0.000 /usr/lib/python3.12/ast.py:1789(unparse)
1570 0.004 0.000 0.034 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/experimental/symbolic_shapes.py:3076(_suppress_guards)
120 0.001 0.000 0.034 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/experimental/proxy_tensor.py:1775(create_node)
360 0.001 0.000 0.033 0.000 /usr/lib/python3.12/ast.py:855(visit)
1655/1585 0.011 0.000 0.033 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/experimental/recording.py:238(wrapper)
1845/360 0.002 0.000 0.032 0.000 /usr/lib/python3.12/ast.py:845(traverse)
210/105 0.006 0.000 0.031 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/guards.py:927(get_guard_manager_from_source)
45 0.000 0.000 0.031 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/guards.py:2757(replace)
45 0.000 0.000 0.031 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/guards.py:2355(replace)
105 0.008 0.000 0.031 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/meta_utils.py:255(describe_tensor)
960 0.002 0.000 0.030 0.000 /home/xadupre/github/experimental-experiment/experimental_experiment/xoptim/patterns_api.py:968(match)
620 0.029 0.000 0.029 0.000 {built-in method torch.empty_strided}
13770 0.027 0.000 0.028 0.000 {built-in method builtins.setattr}
1 0.000 0.000 0.028 0.028 <eval_with_key>.684:4(forward)
4330 0.015 0.000 0.028 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph.py:144(create_name)
3675 0.011 0.000 0.028 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_tensor.py:589(__set__)
30 0.001 0.000 0.028 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/variables/nn_module.py:275(var_getattr)
960 0.002 0.000 0.028 0.000 /home/xadupre/github/experimental-experiment/experimental_experiment/xoptim/patterns_api.py:355(_get_match_pattern)
5 0.000 0.000 0.027 0.005 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/variables/builder.py:1593(wrap_tensor)
10 0.000 0.000 0.027 0.003 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/utils/_config_module.py:623(get_config_copy)
115 0.000 0.000 0.027 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/experimental/proxy_tensor.py:385(extract_val)
10 0.004 0.000 0.027 0.003 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/utils/_config_module.py:463(_get_dict)
115 0.001 0.000 0.026 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/experimental/proxy_tensor.py:359(snapshot_fake)
130 0.003 0.000 0.026 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph.py:1616(override_node_repr)
6110 0.005 0.000 0.025 0.000 {built-in method builtins.repr}
115 0.003 0.000 0.025 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_impls.py:1010(fast_detach)
10 0.001 0.000 0.025 0.003 /home/xadupre/github/experimental-experiment/experimental_experiment/xoptim/patterns_api.py:305(_build_pattern)
85 0.010 0.000 0.024 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/guards.py:776(getitem_on_dict_mgr)
1 0.000 0.000 0.024 0.024 <eval_with_key>.660:4(forward)
5 0.000 0.000 0.024 0.005 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/guards.py:2346(count)
5 0.002 0.000 0.024 0.005 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/utils.py:1382(_scrubbed_inductor_config_for_logging)
30 0.001 0.000 0.024 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/variables/builder.py:1471(wrap_module)
25/15 0.002 0.000 0.023 0.002 {built-in method torch.flatten}
5 0.001 0.000 0.023 0.005 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_export/utils.py:580(apply_runtime_assertion_pass)
455/45 0.002 0.000 0.023 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/guards.py:2309(visit)
75 0.000 0.000 0.023 0.000 /usr/lib/python3.12/inspect.py:3343(signature)
415/45 0.002 0.000 0.023 0.001 /usr/lib/python3.12/ast.py:477(generic_visit)
70 0.000 0.000 0.023 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/eval_frame.py:1108(run_node)
75 0.000 0.000 0.023 0.000 /usr/lib/python3.12/inspect.py:3081(from_callable)
5 0.000 0.000 0.022 0.004 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_export/passes/replace_autocast_with_hop_pass.py:178(replace_autocast_with_hop_pass)
145/75 0.002 0.000 0.022 0.000 /usr/lib/python3.12/inspect.py:2501(_signature_from_callable)
705/45 0.002 0.000 0.022 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/guards.py:2293(visit)
705/45 0.002 0.000 0.022 0.000 /usr/lib/python3.12/ast.py:409(generic_visit)
5 0.001 0.000 0.021 0.004 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_export/utils.py:820(placeholder_naming_pass)
4340/4240 0.006 0.000 0.021 0.000 {method 'join' of 'str' objects}
65 0.003 0.000 0.021 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_dynamo/variables/builder.py:2390(handle_traced_output)
5 0.000 0.000 0.021 0.004 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_export/passes/replace_set_grad_with_hop_pass.py:110(replace_set_grad_with_hop_pass)
260 0.003 0.000 0.021 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/node.py:377(prepend)
5 0.000 0.000 0.021 0.004 /home/xadupre/github/experimental-experiment/experimental_experiment/torch_interpreter/_aten_functions.py:2924(aten_flatten_using_ints)
415/335 0.002 0.000 0.021 0.000 /usr/lib/python3.12/ast.py:1573(visit_Subscript)
780 0.002 0.000 0.020 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph.py:548(_format_args)
4125 0.004 0.000 0.020 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/node.py:602(__repr__)
done.
profile custom2: <function export_cus_p2 at 0x7f50d07a20c0>
done.
Same with dynamo-exporter.
profile_function("dynamo", export_dynamo, verbose=True)
if "dynopt" in supported_exporters:
profile_function("dynopt", export_dynopt)
profile dynamo: <function export_dynamo at 0x7f50d07a18a0>
10152076 function calls (9995653 primitive calls) in 11.901 seconds
Ordered by: cumulative time
ncalls tottime percall cumtime percall filename:lineno(function)
5 0.051 0.010 5.712 1.142 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/onnx/_internal/exporter/_registration.py:115(from_torchlib)
5 0.093 0.019 4.183 0.837 /home/xadupre/github/onnxscript/onnxscript/_framework_apis/torch_2_5.py:82(get_torchlib_ops)
2215 0.038 0.000 4.072 0.002 /home/xadupre/github/onnxscript/onnxscript/values.py:640(function_ir)
202120/201090 0.082 0.000 1.993 0.000 {built-in method builtins.next}
10 0.002 0.000 1.911 0.191 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_export/utils.py:1134(_collect_all_valid_cia_ops)
270 0.024 0.000 1.909 0.007 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_export/utils.py:1117(_collect_all_valid_cia_ops_for_namespace)
4680/4170 0.005 0.000 1.827 0.000 /usr/lib/python3.12/contextlib.py:132(__enter__)
20 0.195 0.010 1.735 0.087 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/export/exported_program.py:188(_override_composite_implicit_decomp)
270 0.597 0.002 1.732 0.006 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_export/utils.py:1052(_materialize_cpp_cia_ops)
2215 0.024 0.000 1.648 0.001 /home/xadupre/github/onnxscript/onnxscript/_internal/ast_utils.py:16(get_src_and_ast)
2895 0.117 0.000 1.396 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/onnx/_internal/exporter/_schemas.py:432(from_function)
2215 0.008 0.000 1.299 0.001 /home/xadupre/github/onnxscript/onnxscript/converter.py:1466(translate_function_signature)
2215 0.090 0.000 1.282 0.001 /home/xadupre/github/onnxscript/onnxscript/converter.py:1381(_translate_function_signature_common)
2215 0.006 0.000 1.146 0.001 /usr/lib/python3.12/inspect.py:1279(getsource)
2215 0.121 0.000 1.135 0.001 /usr/lib/python3.12/inspect.py:1258(getsourcelines)
64735 1.030 0.000 1.113 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_ops.py:120(inner)
5 0.012 0.002 1.112 0.222 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/onnx/_internal/exporter/_decomp.py:42(create_onnx_friendly_decomposition_table)
35/5 0.003 0.000 1.031 0.206 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_tensor.py:2597(from_tensor)
100/5 0.003 0.000 1.031 0.206 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_tensor.py:323(from_real_tensor)
105/5 0.004 0.000 1.028 0.206 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/meta_utils.py:1809(__call__)
5 0.000 0.000 1.025 0.205 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/export/decomp_utils.py:125(items)
5 0.000 0.000 1.024 0.205 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/export/decomp_utils.py:142(_materialize_if_needed)
5 0.003 0.001 1.024 0.205 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/export/decomp_utils.py:129(materialize)
2215 0.085 0.000 1.017 0.000 /usr/lib/python3.12/inspect.py:1606(getclosurevars)
5 0.011 0.002 1.017 0.203 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/export/exported_program.py:295(_split_decomp_table_to_cia_and_python_decomp)
65905 0.340 0.000 0.852 0.000 /usr/lib/python3.12/dis.py:434(_get_instructions_bytes)
24130 0.808 0.000 0.808 0.000 {built-in method builtins.compile}
2215 0.227 0.000 0.784 0.000 /usr/lib/python3.12/inspect.py:1239(getblock)
75165/15510 0.148 0.000 0.745 0.000 /home/xadupre/github/onnxscript/onnxscript/type_annotation.py:131(is_value_type)
660 0.192 0.000 0.667 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/functional_tensor.py:352(__torch_dispatch__)
851980 0.624 0.000 0.631 0.000 {built-in method builtins.getattr}
80/20 0.001 0.000 0.614 0.031 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/nn/functional.py:1693(relu)
3450 0.012 0.000 0.548 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/experimental/proxy_tensor.py:1229(__torch_function__)
1578290/1570705 0.410 0.000 0.521 0.000 {built-in method builtins.isinstance}
8980 0.009 0.000 0.495 0.000 /home/xadupre/github/onnxscript/onnxscript/type_annotation.py:172(is_valid_type)
2895 0.067 0.000 0.494 0.000 /usr/lib/python3.12/typing.py:2215(get_type_hints)
228115 0.266 0.000 0.485 0.000 /usr/lib/python3.12/tokenize.py:569(_generate_tokens_from_c_tokenizer)
27655/5170 0.133 0.000 0.440 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/onnx/_internal/exporter/_schemas.py:268(_get_allowed_types_from_type_annotation)
2215 0.006 0.000 0.401 0.000 /usr/lib/python3.12/ast.py:34(parse)
625 0.005 0.000 0.391 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/experimental/proxy_tensor.py:1327(__torch_dispatch__)
1795 0.023 0.000 0.366 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_tensor.py:1782(dispatch)
120 0.011 0.000 0.356 0.003 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/experimental/proxy_tensor.py:760(proxy_call)
64735 0.057 0.000 0.354 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_ops.py:111(py_impl)
495 0.007 0.000 0.332 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_tensor.py:1342(_cached_dispatch_impl)
35/5 0.001 0.000 0.329 0.066 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/nn/modules/module.py:1755(_call_impl)
90 0.003 0.000 0.329 0.004 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph_module.py:800(recompile)
5 0.000 0.000 0.327 0.065 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/export/_trace.py:1775(forward)
5 0.000 0.000 0.308 0.062 /home/xadupre/github/experimental-experiment/_doc/examples/plot_torch_export_201.py:191(forward)
132200 0.181 0.000 0.308 0.000 /usr/lib/python3.12/typing.py:1546(__getitem__)
270 0.300 0.001 0.300 0.001 {built-in method torch._C._dispatch_get_registrations_for_dispatch_key}
75165 0.077 0.000 0.300 0.000 /home/xadupre/github/onnxscript/onnxscript/type_annotation.py:123(_is_tensor_type)
11040 0.043 0.000 0.289 0.000 /home/xadupre/github/onnxscript/onnxscript/converter.py:451(_eval_constant_expr)
6530 0.005 0.000 0.264 0.000 /home/xadupre/github/onnxscript/onnxscript/type_annotation.py:168(is_attr_type)
90 0.002 0.000 0.253 0.003 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph.py:1562(python_code)
1315 0.004 0.000 0.246 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/experimental/proxy_tensor.py:1258(__torch_function__)
1725 0.005 0.000 0.240 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_export/non_strict_utils.py:636(__torch_function__)
2935 0.004 0.000 0.231 0.000 /usr/lib/python3.12/inspect.py:3343(signature)
60 0.004 0.000 0.228 0.004 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_ops.py:863(handler)
2935 0.005 0.000 0.227 0.000 /usr/lib/python3.12/inspect.py:3081(from_callable)
7285 0.007 0.000 0.226 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_export/utils.py:1038(_is_preservable_cia_op)
2965/2935 0.036 0.000 0.222 0.000 /usr/lib/python3.12/inspect.py:2501(_signature_from_callable)
60 0.014 0.000 0.221 0.004 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_library/utils.py:281(handle_dispatch_mode)
1365 0.006 0.000 0.220 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/utils/_pytree.py:1267(tree_map_only)
225900 0.118 0.000 0.219 0.000 /usr/lib/python3.12/collections/__init__.py:447(_make)
78040 0.062 0.000 0.217 0.000 /home/xadupre/github/onnxscript/onnxscript/type_annotation.py:70(_remove_annotation)
70 0.001 0.000 0.216 0.003 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_higher_order_ops/utils.py:23(autograd_not_implemented_inner)
503540 0.211 0.000 0.211 0.000 {method 'split' of 'str' objects}
2215 0.032 0.000 0.211 0.000 /usr/lib/python3.12/inspect.py:1070(findsource)
90 0.002 0.000 0.200 0.002 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph.py:1639(_python_code)
90 0.022 0.000 0.198 0.002 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph.py:397(_gen_python_code)
40/10 0.001 0.000 0.196 0.020 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_jit_internal.py:613(fn)
117475 0.107 0.000 0.190 0.000 /usr/lib/python3.12/typing.py:2340(get_origin)
131810 0.160 0.000 0.189 0.000 /usr/lib/python3.12/dis.py:623(_unpack_opargs)
10785 0.019 0.000 0.187 0.000 /usr/lib/python3.12/typing.py:892(__init__)
53360/53310 0.043 0.000 0.187 0.000 {built-in method builtins.repr}
7285 0.100 0.000 0.186 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_export/utils.py:1086(_check_valid_to_preserve)
24170/10785 0.031 0.000 0.177 0.000 /usr/lib/python3.12/typing.py:407(_eval_type)
1058405/1057915 0.171 0.000 0.172 0.000 {built-in method builtins.len}
10785 0.036 0.000 0.163 0.000 /usr/lib/python3.12/typing.py:916(_evaluate)
34220 0.028 0.000 0.159 0.000 /home/xadupre/github/onnxscript/onnxscript/ir/_core.py:1475(__hash__)
2215 0.050 0.000 0.156 0.000 /usr/lib/python3.12/dis.py:647(findlabels)
2935 0.053 0.000 0.151 0.000 /usr/lib/python3.12/inspect.py:2397(_signature_from_function)
5 0.038 0.008 0.150 0.030 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/onnx/_internal/exporter/_decomp.py:15(get_onnx_implemented_overloads)
20/5 0.000 0.000 0.147 0.029 {built-in method torch.flatten}
5 0.001 0.000 0.141 0.028 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/onnx/_internal/exporter/_fx_passes.py:22(insert_type_promotion_nodes)
170 0.004 0.000 0.131 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/meta_utils.py:847(meta_tensor)
125/5 0.003 0.000 0.131 0.026 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/onnx/_internal/diagnostics/infra/decorator.py:66(wrapper)
420 0.003 0.000 0.131 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_tensor.py:1716(_output_from_cache_entry)
5 0.000 0.000 0.130 0.026 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/onnx/_internal/fx/_pass.py:240(run)
5 0.000 0.000 0.130 0.026 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/onnx/_internal/fx/passes/type_promotion.py:1691(_run)
440 0.014 0.000 0.128 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_tensor.py:1650(_get_output_tensor_from_cache_entry)
30 0.002 0.000 0.127 0.004 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph_module.py:443(__init__)
675/575 0.005 0.000 0.126 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/nn/modules/module.py:1944(__setattr__)
495 0.006 0.000 0.121 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_tensor.py:1385(_cache_key)
230 0.002 0.000 0.120 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/proxy.py:211(create_proxy)
10 0.001 0.000 0.119 0.012 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/export/_trace.py:437(_produce_aten_artifact)
120 0.001 0.000 0.117 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/onnx/_internal/fx/passes/type_promotion.py:1596(run_node)
30 0.000 0.000 0.116 0.004 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph_module.py:556(graph)
90/30 0.001 0.000 0.113 0.004 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/overrides.py:1670(handle_torch_function)
1875/495 0.016 0.000 0.110 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_tensor.py:1461(_prep_args_for_hash)
5 0.000 0.000 0.109 0.022 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/export/exported_program.py:1314(module)
5 0.000 0.000 0.109 0.022 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/export/_unlift.py:383(_unlift_exported_program_lifted_states)
357250 0.108 0.000 0.108 0.000 {built-in method __new__ of type object at 0xa20960}
37640 0.047 0.000 0.107 0.000 /home/xadupre/github/onnxscript/onnxscript/ir/_core.py:1483(__repr__)
4680/4170 0.006 0.000 0.106 0.000 /usr/lib/python3.12/contextlib.py:141(__exit__)
130 0.001 0.000 0.105 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/experimental/proxy_tensor.py:592(track_tensor_tree)
219216 0.084 0.000 0.105 0.000 {built-in method builtins.hasattr}
240 0.006 0.000 0.102 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/experimental/proxy_tensor.py:1775(create_node)
240/130 0.003 0.000 0.102 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/experimental/proxy_tensor.py:614(wrap_with_proxy)
170 0.007 0.000 0.101 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/autograd/grad_mode.py:273(__exit__)
21845 0.078 0.000 0.101 0.000 {built-in method builtins.eval}
1995 0.005 0.000 0.098 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/utils/_pytree.py:985(tree_flatten)
6985/1995 0.025 0.000 0.093 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/utils/_pytree.py:993(helper)
5 0.001 0.000 0.093 0.019 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/onnx/_internal/exporter/_core.py:933(_exported_program_to_onnx_program)
1575 0.002 0.000 0.092 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/utils/_pytree.py:1212(wrapped)
2160 0.012 0.000 0.088 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph.py:627(emit_node)
63865 0.041 0.000 0.087 0.000 {built-in method builtins.issubclass}
145 0.005 0.000 0.084 0.001 /home/xadupre/github/onnxscript/onnxscript/optimizer/_constant_folding.py:923(process_node)
5 0.001 0.000 0.081 0.016 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/onnx/_internal/exporter/_core.py:662(_translate_fx_graph)
60 0.002 0.000 0.078 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/onnx/_internal/exporter/_core.py:449(_handle_call_function_node_with_lowering)
146545 0.050 0.000 0.077 0.000 /usr/lib/python3.12/inspect.py:302(isclass)
75/15 0.001 0.000 0.076 0.005 {built-in method torch._to_functional_tensor}
230 0.005 0.000 0.075 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/experimental/proxy_tensor.py:485(set_meta)
2215 0.015 0.000 0.074 0.000 /usr/lib/python3.12/textwrap.py:419(dedent)
2215 0.023 0.000 0.074 0.000 /usr/lib/python3.12/inspect.py:951(getsourcefile)
15135 0.011 0.000 0.072 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/onnx/_internal/exporter/_registration.py:261(is_registered)
83575 0.033 0.000 0.072 0.000 <frozen abc>:117(__instancecheck__)
20 0.001 0.000 0.072 0.004 {built-in method torch.relu}
225900 0.072 0.000 0.072 0.000 /usr/lib/python3.12/inspect.py:1196(tokeneater)
8405 0.009 0.000 0.070 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/node.py:894(map_arg)
95045 0.043 0.000 0.070 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_ops.py:763(__hash__)
134610 0.069 0.000 0.070 0.000 /usr/lib/python3.12/typing.py:392(inner)
145 0.003 0.000 0.069 0.000 /home/xadupre/github/onnxscript/onnxscript/optimizer/_constant_folding.py:823(_do_inference)
10 0.000 0.000 0.068 0.007 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/export/exported_program.py:928(__init__)
75 0.001 0.000 0.068 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_tensor.py:1968(_dispatch_impl)
90 0.000 0.000 0.066 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph_module.py:91(_forward_from_src)
15805/9585 0.030 0.000 0.066 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/node.py:903(map_aggregate)
90 0.000 0.000 0.066 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph_module.py:97(_method_from_src)
90 0.001 0.000 0.065 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph_module.py:86(_exec_with_source)
2215 0.027 0.000 0.065 0.000 /usr/lib/python3.12/dis.py:342(get_instructions)
55 0.000 0.000 0.064 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_functorch/_aot_autograd/dispatch_and_compile_graph.py:66(_detach_and_copy_item_memo)
105 0.005 0.000 0.064 0.001 {method 'detach' of 'torch._C.TensorBase' objects}
840 0.019 0.000 0.062 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_tensor.py:671(__new__)
15195 0.037 0.000 0.062 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/onnx/_internal/exporter/_registration.py:239(get_decomps)
8085 0.020 0.000 0.062 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/node.py:873(__setattr__)
15 0.000 0.000 0.061 0.004 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/nn/modules/linear.py:124(forward)
60/15 0.001 0.000 0.061 0.004 {built-in method torch._C._nn.linear}
170 0.010 0.000 0.060 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/autograd/grad_mode.py:269(__enter__)
122945 0.053 0.000 0.057 0.000 {method 'get' of 'dict' objects}
70/60 0.000 0.000 0.056 0.001 /home/xadupre/github/onnxscript/onnxscript/values.py:634(__call__)
5 0.000 0.000 0.056 0.011 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/export/exported_program.py:339(default_decompositions)
5 0.003 0.001 0.056 0.011 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/export/decomp_utils.py:33(__init__)
10785 0.034 0.000 0.056 0.000 /usr/lib/python3.12/typing.py:175(_type_check)
3530/1190 0.005 0.000 0.054 0.000 /home/xadupre/github/onnxscript/onnxscript/ir/serde.py:95(wrapper)
10 0.000 0.000 0.054 0.005 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/export/exported_program.py:1587(_create_graph_module_for_export)
1260/775 0.025 0.000 0.054 0.000 {built-in method torch._ops.prim.}
155250/153420 0.048 0.000 0.053 0.000 {built-in method builtins.hash}
37640 0.022 0.000 0.053 0.000 /home/xadupre/github/onnxscript/onnxscript/ir/_enums.py:95(__repr__)
63690 0.033 0.000 0.052 0.000 <string>:1(<lambda>)
2300 0.008 0.000 0.051 0.000 /usr/lib/python3.12/linecache.py:52(checkcache)
70/65 0.001 0.000 0.051 0.001 /home/xadupre/github/onnxscript/onnxscript/values.py:295(__call__)
145 0.017 0.000 0.051 0.000 /home/xadupre/github/onnx/onnx/shape_inference.py:99(infer_node_outputs)
251250 0.050 0.000 0.050 0.000 /usr/lib/python3.12/dis.py:195(_deoptop)
70/65 0.001 0.000 0.049 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/onnx/_internal/exporter/_building.py:570(eval)
415 0.006 0.000 0.049 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph.py:1104(create_node)
620 0.015 0.000 0.048 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_tensor.py:948(_flatten_into)
13385/10785 0.022 0.000 0.047 0.000 /usr/lib/python3.12/typing.py:2315(_strip_annotations)
62065 0.022 0.000 0.047 0.000 <frozen abc>:121(__subclasscheck__)
10 0.000 0.000 0.046 0.005 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/nn/modules/conv.py:553(forward)
10 0.000 0.000 0.046 0.005 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/nn/modules/conv.py:536(_conv_forward)
40/10 0.001 0.000 0.046 0.005 {built-in method torch.conv2d}
60 0.002 0.000 0.046 0.001 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/utils/_traceback.py:171(summary)
8300 0.045 0.000 0.045 0.000 {method 'copy' of 'dict' objects}
230 0.001 0.000 0.045 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/experimental/proxy_tensor.py:385(extract_val)
10 0.001 0.000 0.045 0.004 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_export/utils.py:580(apply_runtime_assertion_pass)
240 0.003 0.000 0.044 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/proxy.py:145(create_node)
620 0.012 0.000 0.044 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_tensor.py:968(extract_tensor_metadata)
230 0.001 0.000 0.043 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/experimental/proxy_tensor.py:359(snapshot_fake)
2395/16 0.008 0.000 0.043 0.003 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/utils/_stats.py:22(wrapper)
170 0.010 0.000 0.043 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/meta_utils.py:255(describe_tensor)
180 0.005 0.000 0.043 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/fx/graph.py:1616(override_node_repr)
10 0.002 0.000 0.043 0.004 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_export/utils.py:820(placeholder_naming_pass)
7405 0.037 0.000 0.042 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_decomp/__init__.py:56(_should_decompose_because_unsafe_op)
2295 0.042 0.000 0.042 0.000 {built-in method posix.stat}
230 0.009 0.000 0.042 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/fake_impls.py:1010(fast_detach)
5 0.000 0.000 0.042 0.008 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/export/_unlift.py:178(_unlift)
5215 0.008 0.000 0.042 0.000 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_subclasses/meta_utils.py:172(is_sparse_any)
17470 0.024 0.000 0.042 0.000 /usr/lib/python3.12/typing.py:2370(get_args)
10 0.000 0.000 0.042 0.004 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/_export/passes/replace_with_hop_pass_util.py:157(_replace_with_hop_pass_helper)
5180 0.021 0.000 0.041 0.000 /usr/lib/python3.12/inspect.py:754(unwrap)
8075 0.022 0.000 0.040 0.000 /usr/lib/python3.12/inspect.py:2743(__init__)
83575 0.040 0.000 0.040 0.000 {built-in method _abc._abc_instancecheck}
10 0.000 0.000 0.039 0.004 /home/xadupre/github/onnxscript/onnxscript/rewriter/__init__.py:26(rewrite)
3135/130 0.008 0.000 0.038 0.000 /usr/lib/python3.12/copy.py:118(deepcopy)
40/10 0.000 0.000 0.038 0.004 /home/xadupre/vv/this312/lib/python3.12/site-packages/torch/nn/functional.py:807(_max_pool2d)
done.
profile dynopt: <function export_dynopt at 0x7f50d07a1620>
done.
Benchmark exported models with ORT¶
def benchmark(shape):
from onnxruntime import InferenceSession, SessionOptions, GraphOptimizationLevel
providers = [["CPUExecutionProvider"]]
if has_cuda:
providers.append(["CUDAExecutionProvider", "CPUExecutionProvider"])
data = []
data1 = []
data_mem_load = []
data_mem_first_run = []
data_mem_run = []
confs = list(
itertools.product(
[_ for _ in os.listdir(".") if ".onnx" in _ and _.startswith("plot_torch")],
providers,
["0", "1"],
)
)
loop = tqdm(confs)
print(f"number of experiments: {len(loop)}")
for name, ps, aot in loop:
root = os.path.split(name)[-1]
_, ext = os.path.splitext(root)
if ext != ".onnx":
continue
obs = {} # system_info()
obs["name"] = name
obs["providers"] = ",".join(ps)
p = "CUDA" if "CUDA" in obs["providers"] else "CPU"
obs["compute"] = p
obs["aot"] = 1 if aot == "0" else 0
obs["export"] = name.replace("plot_torch_export_", "").replace(".onnx", "")
if not has_cuda and p == "CUDA":
continue
onx = onnx.load(name)
obs["n_nodes"] = len(onx.graph.node)
obs["n_function"] = len(onx.functions or [])
obs["n_sub"] = len([n for n in onx.graph.node if n.op_type == "Sub"])
obs1 = obs.copy()
short_obs = dict(
name=obs["name"],
aot=obs["aot"],
providers=obs["providers"],
export=obs["export"],
compute=obs["compute"],
)
opts = SessionOptions()
opts.add_session_config_entry("session.disable_aot_function_inlining", aot)
opts.graph_optimization_level = GraphOptimizationLevel.ORT_ENABLE_ALL
opts.optimized_model_filepath = (
f"ort-{name.replace('.onnx', '')}-{p.lower()}-aot{1 if aot == '0' else 0}.onnx"
)
try:
InferenceSession(name, opts, providers=ps)
except Exception as e:
loop.set_description(f"ERROR-load: {name} {e}")
obs.update({"error": e, "step": "run"})
data.append(obs)
continue
opts = SessionOptions()
opts.add_session_config_entry("session.disable_aot_function_inlining", aot)
opts.graph_optimization_level = GraphOptimizationLevel.ORT_ENABLE_ALL
stat = start_spying_on(cuda=1 if has_cuda else 0)
sess = InferenceSession(name, opts, providers=ps)
memobs = flatten(stat.stop())
memobs.update(short_obs)
data_mem_load.append(memobs)
input_name = sess.get_inputs()[0].name
feeds = {input_name: np.random.rand(*shape).astype(np.float32)}
stat = start_spying_on(cuda=1 if has_cuda else 0)
try:
sess.run(None, feeds)
except Exception as e:
loop.set_description(f"ERROR-run: {name} {e}")
obs.update({"error": e, "step": "load"})
data.append(obs)
stat.stop()
continue
memobs = flatten(stat.stop())
memobs.update(short_obs)
data_mem_first_run.append(memobs)
# memory consumption
stat = start_spying_on(cuda=1 if has_cuda else 0)
for _ in range(0, script_args.warmup):
sess.run(None, feeds)
memobs = flatten(stat.stop())
memobs.update(short_obs)
data_mem_run.append(memobs)
obs.update(
measure_time(
lambda sess=sess, feeds=feeds: sess.run(None, feeds),
max_time=script_args.maxtime,
repeat=script_args.repeat,
number=1,
)
)
loop.set_description(f"{obs['average']} {name} {ps}")
data.append(obs)
# check first run
obs1.update(
measure_time(
lambda name=name, opts=opts, ps=ps, feeds=feeds: InferenceSession(
name, opts, providers=ps
).run(None, feeds),
max_time=script_args.maxtime,
repeat=max(1, script_args.repeat // 2),
number=1,
)
)
data1.append(obs1)
df = pandas.DataFrame(data)
df.to_csv("plot_torch_export_ort_time.csv", index=False)
df.to_excel("plot_torch_export_ort_time.xlsx", index=False)
df1 = pandas.DataFrame(data1)
df1.to_csv("plot_torch_export_ort_time1_init.csv", index=False)
df1.to_excel("plot_torch_export_ort_time1_init.xlsx", index=False)
dfmem = pandas.DataFrame(data_mem_load)
dfmem.to_csv("plot_torch_export_ort_load_mem.csv", index=False)
dfmem.to_excel("plot_torch_export_ort_load_mem.xlsx", index=False)
dfmemr = pandas.DataFrame(data_mem_run)
dfmemr.to_csv("plot_torch_export_ort_run_mem.csv", index=False)
dfmemr.to_excel("plot_torch_export_ort_run_mem.xlsx", index=False)
dfmemfr = pandas.DataFrame(data_mem_first_run)
dfmemfr.to_csv("plot_torch_export_ort_first_run_mem.csv", index=False)
dfmemfr.to_excel("plot_torch_export_ort_first_run_mem.xlsx", index=False)
return df, df1, dfmem, dfmemfr, dfmemr
df, df_init, dfmem, dfmemfr, dfmemr = benchmark(list(input_tensor.shape))
print(df)
0%| | 0/20 [00:00<?, ?it/s]number of experiments: 20
6.774530103809392e-05 plot_torch_export_cus_p2.onnx ['CPUExecutionProvider']: 0%| | 0/20 [00:01<?, ?it/s]
6.774530103809392e-05 plot_torch_export_cus_p2.onnx ['CPUExecutionProvider']: 5%|▌ | 1/20 [00:01<00:27, 1.45s/it]
9.85442191134514e-05 plot_torch_export_cus_p2.onnx ['CPUExecutionProvider']: 5%|▌ | 1/20 [00:02<00:27, 1.45s/it]
9.85442191134514e-05 plot_torch_export_cus_p2.onnx ['CPUExecutionProvider']: 10%|█ | 2/20 [00:02<00:19, 1.06s/it]
0.0011334410512887174 plot_torch_export_cus_p2.onnx ['CUDAExecutionProvider', 'CPUExecutionProvider']: 10%|█ | 2/20 [00:03<00:19, 1.06s/it]
0.0011334410512887174 plot_torch_export_cus_p2.onnx ['CUDAExecutionProvider', 'CPUExecutionProvider']: 15%|█▌ | 3/20 [00:03<00:17, 1.06s/it]
0.0015487084047522547 plot_torch_export_cus_p2.onnx ['CUDAExecutionProvider', 'CPUExecutionProvider']: 15%|█▌ | 3/20 [00:03<00:17, 1.06s/it]
0.0015487084047522547 plot_torch_export_cus_p2.onnx ['CUDAExecutionProvider', 'CPUExecutionProvider']: 20%|██ | 4/20 [00:04<00:15, 1.04it/s]
9.094345524299926e-05 plot_torch_export_dynopt.onnx ['CPUExecutionProvider']: 20%|██ | 4/20 [00:04<00:15, 1.04it/s]
9.094345524299926e-05 plot_torch_export_dynopt.onnx ['CPUExecutionProvider']: 25%|██▌ | 5/20 [00:04<00:12, 1.18it/s]
8.639918181899357e-05 plot_torch_export_dynopt.onnx ['CPUExecutionProvider']: 25%|██▌ | 5/20 [00:05<00:12, 1.18it/s]
8.639918181899357e-05 plot_torch_export_dynopt.onnx ['CPUExecutionProvider']: 30%|███ | 6/20 [00:05<00:10, 1.30it/s]
0.0009676481709343184 plot_torch_export_dynopt.onnx ['CUDAExecutionProvider', 'CPUExecutionProvider']: 30%|███ | 6/20 [00:05<00:10, 1.30it/s]
0.0009676481709343184 plot_torch_export_dynopt.onnx ['CUDAExecutionProvider', 'CPUExecutionProvider']: 35%|███▌ | 7/20 [00:06<00:09, 1.33it/s]
0.0013565394561466523 plot_torch_export_dynopt.onnx ['CUDAExecutionProvider', 'CPUExecutionProvider']: 35%|███▌ | 7/20 [00:06<00:09, 1.33it/s]
0.0013565394561466523 plot_torch_export_dynopt.onnx ['CUDAExecutionProvider', 'CPUExecutionProvider']: 40%|████ | 8/20 [00:07<00:09, 1.21it/s]
6.677814703928677e-05 plot_torch_export_dynamo.onnx ['CPUExecutionProvider']: 40%|████ | 8/20 [00:07<00:09, 1.21it/s]
6.677814703928677e-05 plot_torch_export_dynamo.onnx ['CPUExecutionProvider']: 45%|████▌ | 9/20 [00:07<00:08, 1.28it/s]
7.941147486441518e-05 plot_torch_export_dynamo.onnx ['CPUExecutionProvider']: 45%|████▌ | 9/20 [00:08<00:08, 1.28it/s]
7.941147486441518e-05 plot_torch_export_dynamo.onnx ['CPUExecutionProvider']: 50%|█████ | 10/20 [00:08<00:07, 1.36it/s]
0.0007144002052966894 plot_torch_export_dynamo.onnx ['CUDAExecutionProvider', 'CPUExecutionProvider']: 50%|█████ | 10/20 [00:08<00:07, 1.36it/s]
0.0007144002052966894 plot_torch_export_dynamo.onnx ['CUDAExecutionProvider', 'CPUExecutionProvider']: 55%|█████▌ | 11/20 [00:09<00:06, 1.43it/s]
0.0008318764285788694 plot_torch_export_dynamo.onnx ['CUDAExecutionProvider', 'CPUExecutionProvider']: 55%|█████▌ | 11/20 [00:09<00:06, 1.43it/s]
0.0008318764285788694 plot_torch_export_dynamo.onnx ['CUDAExecutionProvider', 'CPUExecutionProvider']: 60%|██████ | 12/20 [00:09<00:05, 1.44it/s]
6.236133662293357e-05 plot_torch_export_script.onnx ['CPUExecutionProvider']: 60%|██████ | 12/20 [00:10<00:05, 1.44it/s]
6.236133662293357e-05 plot_torch_export_script.onnx ['CPUExecutionProvider']: 65%|██████▌ | 13/20 [00:10<00:04, 1.45it/s]
6.37282538627513e-05 plot_torch_export_script.onnx ['CPUExecutionProvider']: 65%|██████▌ | 13/20 [00:10<00:04, 1.45it/s]
6.37282538627513e-05 plot_torch_export_script.onnx ['CPUExecutionProvider']: 70%|███████ | 14/20 [00:10<00:04, 1.50it/s]
0.0007548896853152725 plot_torch_export_script.onnx ['CUDAExecutionProvider', 'CPUExecutionProvider']: 70%|███████ | 14/20 [00:11<00:04, 1.50it/s]
0.0007548896853152725 plot_torch_export_script.onnx ['CUDAExecutionProvider', 'CPUExecutionProvider']: 75%|███████▌ | 15/20 [00:11<00:03, 1.56it/s]
0.0007309904576326989 plot_torch_export_script.onnx ['CUDAExecutionProvider', 'CPUExecutionProvider']: 75%|███████▌ | 15/20 [00:12<00:03, 1.56it/s]
0.0007309904576326989 plot_torch_export_script.onnx ['CUDAExecutionProvider', 'CPUExecutionProvider']: 80%|████████ | 16/20 [00:12<00:02, 1.59it/s]
7.436829438438482e-05 plot_torch_export_cus_p0.onnx ['CPUExecutionProvider']: 80%|████████ | 16/20 [00:12<00:02, 1.59it/s]
7.436829438438482e-05 plot_torch_export_cus_p0.onnx ['CPUExecutionProvider']: 85%|████████▌ | 17/20 [00:12<00:02, 1.50it/s]
6.242102288097068e-05 plot_torch_export_cus_p0.onnx ['CPUExecutionProvider']: 85%|████████▌ | 17/20 [00:13<00:02, 1.50it/s]
6.242102288097068e-05 plot_torch_export_cus_p0.onnx ['CPUExecutionProvider']: 90%|█████████ | 18/20 [00:13<00:01, 1.55it/s]
0.0007602554666610322 plot_torch_export_cus_p0.onnx ['CUDAExecutionProvider', 'CPUExecutionProvider']: 90%|█████████ | 18/20 [00:13<00:01, 1.55it/s]
0.0007602554666610322 plot_torch_export_cus_p0.onnx ['CUDAExecutionProvider', 'CPUExecutionProvider']: 95%|█████████▌| 19/20 [00:14<00:00, 1.58it/s]
0.0007602310370472238 plot_torch_export_cus_p0.onnx ['CUDAExecutionProvider', 'CPUExecutionProvider']: 95%|█████████▌| 19/20 [00:14<00:00, 1.58it/s]
0.0007602310370472238 plot_torch_export_cus_p0.onnx ['CUDAExecutionProvider', 'CPUExecutionProvider']: 100%|██████████| 20/20 [00:14<00:00, 1.62it/s]
0.0007602310370472238 plot_torch_export_cus_p0.onnx ['CUDAExecutionProvider', 'CPUExecutionProvider']: 100%|██████████| 20/20 [00:14<00:00, 1.36it/s]
name providers compute aot export n_nodes ... max_exec repeat number ttime context_size warmup_time
0 plot_torch_export_cus_p2.onnx CPUExecutionProvider CPU 1 cus_p2 12 ... 0.000070 1 1734.0 0.117470 64 0.000297
1 plot_torch_export_cus_p2.onnx CPUExecutionProvider CPU 0 cus_p2 12 ... 0.000134 1 1287.0 0.126826 64 0.000366
2 plot_torch_export_cus_p2.onnx CUDAExecutionProvider,CPUExecutionProvider CUDA 1 cus_p2 12 ... 0.001289 1 117.0 0.132613 64 0.001657
3 plot_torch_export_cus_p2.onnx CUDAExecutionProvider,CPUExecutionProvider CUDA 0 cus_p2 12 ... 0.001685 1 84.0 0.130092 64 0.002684
4 plot_torch_export_dynopt.onnx CPUExecutionProvider CPU 1 dynopt 13 ... 0.000100 1 1173.0 0.106677 64 0.000316
5 plot_torch_export_dynopt.onnx CPUExecutionProvider CPU 0 dynopt 13 ... 0.000256 1 1265.0 0.109295 64 0.000601
6 plot_torch_export_dynopt.onnx CUDAExecutionProvider,CPUExecutionProvider CUDA 1 dynopt 13 ... 0.001662 1 117.0 0.113215 64 0.002181
7 plot_torch_export_dynopt.onnx CUDAExecutionProvider,CPUExecutionProvider CUDA 0 dynopt 13 ... 0.001411 1 114.0 0.154645 64 0.001793
8 plot_torch_export_dynamo.onnx CPUExecutionProvider CPU 1 dynamo 13 ... 0.000142 1 1503.0 0.100368 64 0.000361
9 plot_torch_export_dynamo.onnx CPUExecutionProvider CPU 0 dynamo 13 ... 0.000095 1 1293.0 0.102679 64 0.000347
10 plot_torch_export_dynamo.onnx CUDAExecutionProvider,CPUExecutionProvider CUDA 1 dynamo 13 ... 0.001068 1 151.0 0.107874 64 0.001688
11 plot_torch_export_dynamo.onnx CUDAExecutionProvider,CPUExecutionProvider CUDA 0 dynamo 13 ... 0.000858 1 126.0 0.104816 64 0.001903
12 plot_torch_export_script.onnx CPUExecutionProvider CPU 1 script 12 ... 0.000098 1 1723.0 0.107449 64 0.000367
13 plot_torch_export_script.onnx CPUExecutionProvider CPU 0 script 12 ... 0.000109 1 2265.0 0.144344 64 0.000326
14 plot_torch_export_script.onnx CUDAExecutionProvider,CPUExecutionProvider CUDA 1 script 12 ... 0.001012 1 143.0 0.107949 64 0.001761
15 plot_torch_export_script.onnx CUDAExecutionProvider,CPUExecutionProvider CUDA 0 script 12 ... 0.001096 1 177.0 0.129385 64 0.001868
16 plot_torch_export_cus_p0.onnx CPUExecutionProvider CPU 1 cus_p0 15 ... 0.000096 1 1763.0 0.131111 64 0.000427
17 plot_torch_export_cus_p0.onnx CPUExecutionProvider CPU 0 cus_p0 15 ... 0.000081 1 1923.0 0.120036 64 0.000351
18 plot_torch_export_cus_p0.onnx CUDAExecutionProvider,CPUExecutionProvider CUDA 1 cus_p0 15 ... 0.001160 1 135.0 0.102634 64 0.001992
19 plot_torch_export_cus_p0.onnx CUDAExecutionProvider,CPUExecutionProvider CUDA 0 cus_p0 15 ... 0.000769 1 135.0 0.102631 64 0.001782
[20 rows x 17 columns]
Other view
def view_time(df, title, suffix="time"):
piv = pandas.pivot_table(df, index="export", columns=["compute", "aot"], values="average")
print(piv)
piv.to_csv(f"plot_torch_export_ort_{suffix}_compute.csv")
piv.to_excel(f"plot_torch_export_ort_{suffix}_compute.xlsx")
piv_cpu = pandas.pivot_table(
df[df.compute == "CPU"],
index="export",
columns=["compute", "aot"],
values="average",
)
fig, ax = plt.subplots(1, 2, figsize=(12, 4))
fig.suptitle(title)
piv_cpu.plot.barh(ax=ax[0], title="CPU")
if has_cuda:
piv_gpu = pandas.pivot_table(
df[df.compute == "CUDA"],
index="export",
columns=["compute", "aot"],
values="average",
)
piv_gpu.plot.barh(ax=ax[1], title="CUDA")
fig.tight_layout()
fig.savefig(f"plot_torch_export_ort_{suffix}.png")
return ax
view_time(df, "Compares onnxruntime time on exported models")
compute CPU CUDA
aot 0 1 0 1
export
cus_p0 0.000062 0.000074 0.000760 0.000760
cus_p2 0.000099 0.000068 0.001549 0.001133
dynamo 0.000079 0.000067 0.000832 0.000714
dynopt 0.000086 0.000091 0.001357 0.000968
script 0.000064 0.000062 0.000731 0.000755
array([<Axes: title={'center': 'CPU'}, ylabel='export'>,
<Axes: title={'center': 'CUDA'}, ylabel='export'>], dtype=object)
New graph without the very long times.
piv_cpu = pandas.pivot_table(
df[
(df.compute == "CPU")
& ((df.aot == 1) | ((df.export != "dynamo") & (df.export != "dynopt")))
],
index="export",
columns=["compute", "aot"],
values="average",
)
fig, ax = plt.subplots(1, 2, figsize=(12, 4))
fig.suptitle("Compares onnxruntime time on exported models\nHide dynamo without AOT")
piv_cpu.plot.barh(ax=ax[0], title="CPU")
if has_cuda:
piv_gpu = pandas.pivot_table(
df[df.compute == "CUDA"],
index="export",
columns=["compute", "aot"],
values="average",
)
piv_gpu.plot.barh(ax=ax[1], title="CUDA")
fig.tight_layout()
fig.savefig("plot_torch_export_ort_time_2.png")
Let’s do the same with the loading time + the first run.
view_time(
df_init,
"Compares onnxruntime loading time and first run on exported models",
suffix="time1_init",
)
compute CPU CUDA
aot 0 1 0 1
export
cus_p0 0.005893 0.011159 0.015356 0.020402
cus_p2 0.008422 0.006899 0.029318 0.023198
dynamo 0.007780 0.006664 0.030340 0.019251
dynopt 0.008386 0.007506 0.020447 0.019667
script 0.006734 0.005652 0.034694 0.036693
array([<Axes: title={'center': 'CPU'}, ylabel='export'>,
<Axes: title={'center': 'CUDA'}, ylabel='export'>], dtype=object)
Memory Loading Time (ORT)¶
for compute in ["CPU", "CUDA"]:
if not has_cuda and compute == "CUDA":
continue
ax = memory_peak_plot(
dfmem[dfmem.compute == compute],
("export", "aot"),
suptitle=f"Memory Consumption of onnxruntime loading time\nrunning on {compute}",
bars=[model_size * i / 2**20 for i in range(1, 3)],
figsize=(18, 6),
)
get_figure(ax).savefig(f"plot_torch_export_ort_load_mem_{compute}.png")
Memory First Running Time (ORT)¶
for compute in ["CPU", "CUDA"]:
if not has_cuda and compute == "CUDA":
continue
ax = memory_peak_plot(
dfmemfr[dfmemfr.compute == compute],
("export", "aot"),
suptitle=f"Memory Consumption of onnxruntime first running time"
f"\nrunning on {compute}",
bars=[model_size * i / 2**20 for i in range(1, 3)],
figsize=(18, 6),
)
get_figure(ax).savefig(f"plot_torch_export_ort_first_run_mem_{compute}.png")
Memory Running Time (ORT)¶
for compute in ["CPU", "CUDA"]:
if not has_cuda and compute == "CUDA":
continue
ax = memory_peak_plot(
dfmemr[dfmemr.compute == compute],
("export", "aot"),
suptitle=f"Memory Consumption of onnxruntime running time\nrunning on {compute}",
bars=[model_size * i / 2**20 for i in range(1, 3)],
figsize=(18, 6),
)
get_figure(ax).savefig(f"plot_torch_export_ort_run_mem_{compute}.png")
Show the interesting models for CPU¶
script¶
model = "ort-plot_torch_export_cus_p2-cpu-aot0.onnx"
if os.path.exists(model):
print(pretty_onnx(onnx.load(model)))
opset: domain='' version=18
opset: domain='ai.onnx.ml' version=5
opset: domain='onnx_extended.ortops.optim.cuda' version=1000
opset: domain='ai.onnx.training' version=1
opset: domain='ai.onnx.preview.training' version=1
opset: domain='com.microsoft' version=1
opset: domain='com.microsoft.experimental' version=1
opset: domain='com.microsoft.nchwc' version=1
opset: domain='org.pytorch.aten' version=1
input: name='input' type=dtype('float32') shape=[1, 1, 16, 16]
init: name='_onx_concat_gatherelements__shape_max_pool2d_1000' type=int64 shape=(2,) -- array([ 1, -1])-- GraphBuilder.constant_folding.from/fold(_onx_gatherelements__shape_max_pool2d_100,init7_s1_-1)##_onx_gatherelements__shape_max_pool2d_100/GraphBuilder.constant_folding.from/fold(_shape_max_pool2d_10,init7_s1_0)##_shape_max_pool2d_10/##init7_s1_0/Opset.make_node.1/Shape##init7_s1_-1/Opset.make_node.1/Shape
init: name='GemmTransposePattern--_onx_transpose_p_fc1_weight0' type=float32 shape=(512, 16)-- GraphBuilder.constant_folding.from/fold(_onx_transpose_p_fc1_weight0)##_onx_transpose_p_fc1_weight0/GraphBuilder.constant_folding.from/fold(p_fc1_weight)##p_fc1_weight/DynamoInterpret.placeholder.1/P(fc1.weight)
init: name='GemmTransposePattern--_onx_transpose_p_fc2_weight0' type=float32 shape=(128, 512)-- GraphBuilder.constant_folding.from/fold(_onx_transpose_p_fc2_weight0)##_onx_transpose_p_fc2_weight0/GraphBuilder.constant_folding.from/fold(p_fc2_weight)##p_fc2_weight/DynamoInterpret.placeholder.1/P(fc2.weight)
init: name='GemmTransposePattern--_onx_transpose_p_fc3_weight0' type=float32 shape=(10, 128)-- GraphBuilder.constant_folding.from/fold(_onx_transpose_p_fc3_weight0)##_onx_transpose_p_fc3_weight0/GraphBuilder.constant_folding.from/fold(p_fc3_weight)##p_fc3_weight/DynamoInterpret.placeholder.1/P(fc3.weight)
init: name='reorder' type=float32 shape=(16, 1, 5, 5)
init: name='conv1.bias' type=float32 shape=(16,) -- DynamoInterpret.placeholder.1/P(conv1.bias)
init: name='reorder_token_2' type=float32 shape=(16, 16, 5, 5)
init: name='conv2.bias' type=float32 shape=(16,) -- DynamoInterpret.placeholder.1/P(conv2.bias)
init: name='fc1.bias' type=float32 shape=(512,) -- DynamoInterpret.placeholder.1/P(fc1.bias)
init: name='fc2.bias' type=float32 shape=(128,) -- DynamoInterpret.placeholder.1/P(fc2.bias)
init: name='fc3.bias' type=float32 shape=(10,) -- DynamoInterpret.placeholder.1/P(fc3.bias)
Conv[com.microsoft.nchwc](input, reorder, conv1.bias, activation=b'Relu', dilations=[1,1], group=1, strides=[1,1], pads=[0,0,0,0], auto_pad=b'NOTSET') -> reorder_token_0
MaxPool[com.microsoft.nchwc](reorder_token_0, storage_order=0, auto_pad=b'NOTSET', ceil_mode=0, dilations=[1,1], kernel_shape=[2,2], pads=[0,0,0,0], strides=[2,2]) -> reorder_token_1
Conv[com.microsoft.nchwc](reorder_token_1, reorder_token_2, conv2.bias, activation=b'Relu', dilations=[1,1], group=1, strides=[1,1], pads=[0,0,0,0], auto_pad=b'NOTSET') -> reorder_token_3
MaxPool[com.microsoft.nchwc](reorder_token_3, storage_order=0, auto_pad=b'NOTSET', ceil_mode=0, dilations=[1,1], kernel_shape=[2,2], pads=[0,0,0,0], strides=[2,2]) -> reorder_token_4
ReorderOutput[com.microsoft.nchwc](reorder_token_4, channels_last=0, channels=16) -> max_pool2d_1
Reshape(max_pool2d_1, _onx_concat_gatherelements__shape_max_pool2d_1000, allowzero=0) -> flatten
FusedGemm[com.microsoft](flatten, GemmTransposePattern--_onx_transpose_p_fc1_weight0, fc1.bias, transA=0, beta=1.00, activation=b'Relu', transB=1, alpha=1.00) -> relu_2
FusedGemm[com.microsoft](relu_2, GemmTransposePattern--_onx_transpose_p_fc2_weight0, fc2.bias, transA=0, beta=1.00, activation=b'Relu', transB=1, alpha=1.00) -> relu_3
Gemm(relu_3, GemmTransposePattern--_onx_transpose_p_fc3_weight0, fc3.bias, transA=0, beta=1.00, transB=1, alpha=1.00) -> output_0
output: name='output_0' type=dtype('float32') shape=[1, 10]
cus_p2¶
model = "ort-plot_torch_export_cus_p2-cpu-aot0.onnx"
if os.path.exists(model):
print(pretty_onnx(onnx.load(model)))
opset: domain='' version=18
opset: domain='ai.onnx.ml' version=5
opset: domain='onnx_extended.ortops.optim.cuda' version=1000
opset: domain='ai.onnx.training' version=1
opset: domain='ai.onnx.preview.training' version=1
opset: domain='com.microsoft' version=1
opset: domain='com.microsoft.experimental' version=1
opset: domain='com.microsoft.nchwc' version=1
opset: domain='org.pytorch.aten' version=1
input: name='input' type=dtype('float32') shape=[1, 1, 16, 16]
init: name='_onx_concat_gatherelements__shape_max_pool2d_1000' type=int64 shape=(2,) -- array([ 1, -1])-- GraphBuilder.constant_folding.from/fold(_onx_gatherelements__shape_max_pool2d_100,init7_s1_-1)##_onx_gatherelements__shape_max_pool2d_100/GraphBuilder.constant_folding.from/fold(_shape_max_pool2d_10,init7_s1_0)##_shape_max_pool2d_10/##init7_s1_0/Opset.make_node.1/Shape##init7_s1_-1/Opset.make_node.1/Shape
init: name='GemmTransposePattern--_onx_transpose_p_fc1_weight0' type=float32 shape=(512, 16)-- GraphBuilder.constant_folding.from/fold(_onx_transpose_p_fc1_weight0)##_onx_transpose_p_fc1_weight0/GraphBuilder.constant_folding.from/fold(p_fc1_weight)##p_fc1_weight/DynamoInterpret.placeholder.1/P(fc1.weight)
init: name='GemmTransposePattern--_onx_transpose_p_fc2_weight0' type=float32 shape=(128, 512)-- GraphBuilder.constant_folding.from/fold(_onx_transpose_p_fc2_weight0)##_onx_transpose_p_fc2_weight0/GraphBuilder.constant_folding.from/fold(p_fc2_weight)##p_fc2_weight/DynamoInterpret.placeholder.1/P(fc2.weight)
init: name='GemmTransposePattern--_onx_transpose_p_fc3_weight0' type=float32 shape=(10, 128)-- GraphBuilder.constant_folding.from/fold(_onx_transpose_p_fc3_weight0)##_onx_transpose_p_fc3_weight0/GraphBuilder.constant_folding.from/fold(p_fc3_weight)##p_fc3_weight/DynamoInterpret.placeholder.1/P(fc3.weight)
init: name='reorder' type=float32 shape=(16, 1, 5, 5)
init: name='conv1.bias' type=float32 shape=(16,) -- DynamoInterpret.placeholder.1/P(conv1.bias)
init: name='reorder_token_2' type=float32 shape=(16, 16, 5, 5)
init: name='conv2.bias' type=float32 shape=(16,) -- DynamoInterpret.placeholder.1/P(conv2.bias)
init: name='fc1.bias' type=float32 shape=(512,) -- DynamoInterpret.placeholder.1/P(fc1.bias)
init: name='fc2.bias' type=float32 shape=(128,) -- DynamoInterpret.placeholder.1/P(fc2.bias)
init: name='fc3.bias' type=float32 shape=(10,) -- DynamoInterpret.placeholder.1/P(fc3.bias)
Conv[com.microsoft.nchwc](input, reorder, conv1.bias, activation=b'Relu', dilations=[1,1], group=1, strides=[1,1], pads=[0,0,0,0], auto_pad=b'NOTSET') -> reorder_token_0
MaxPool[com.microsoft.nchwc](reorder_token_0, storage_order=0, auto_pad=b'NOTSET', ceil_mode=0, dilations=[1,1], kernel_shape=[2,2], pads=[0,0,0,0], strides=[2,2]) -> reorder_token_1
Conv[com.microsoft.nchwc](reorder_token_1, reorder_token_2, conv2.bias, activation=b'Relu', dilations=[1,1], group=1, strides=[1,1], pads=[0,0,0,0], auto_pad=b'NOTSET') -> reorder_token_3
MaxPool[com.microsoft.nchwc](reorder_token_3, storage_order=0, auto_pad=b'NOTSET', ceil_mode=0, dilations=[1,1], kernel_shape=[2,2], pads=[0,0,0,0], strides=[2,2]) -> reorder_token_4
ReorderOutput[com.microsoft.nchwc](reorder_token_4, channels_last=0, channels=16) -> max_pool2d_1
Reshape(max_pool2d_1, _onx_concat_gatherelements__shape_max_pool2d_1000, allowzero=0) -> flatten
FusedGemm[com.microsoft](flatten, GemmTransposePattern--_onx_transpose_p_fc1_weight0, fc1.bias, transA=0, beta=1.00, activation=b'Relu', transB=1, alpha=1.00) -> relu_2
FusedGemm[com.microsoft](relu_2, GemmTransposePattern--_onx_transpose_p_fc2_weight0, fc2.bias, transA=0, beta=1.00, activation=b'Relu', transB=1, alpha=1.00) -> relu_3
Gemm(relu_3, GemmTransposePattern--_onx_transpose_p_fc3_weight0, fc3.bias, transA=0, beta=1.00, transB=1, alpha=1.00) -> output_0
output: name='output_0' type=dtype('float32') shape=[1, 10]
dynopt¶
model = "ort-plot_torch_export_dynopt-cpu-aot1.onnx"
if os.path.exists(model):
print(pretty_onnx(onnx.load(model)))
opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
opset: domain='ai.onnx.ml' version=5
opset: domain='onnx_extended.ortops.optim.cuda' version=1000
opset: domain='ai.onnx.training' version=1
opset: domain='ai.onnx.preview.training' version=1
opset: domain='com.microsoft' version=1
opset: domain='com.microsoft.experimental' version=1
opset: domain='com.microsoft.nchwc' version=1
opset: domain='org.pytorch.aten' version=1
input: name='x' type=dtype('float32') shape=[1, 1, 16, 16]
init: name='reorder' type=float32 shape=(16, 1, 5, 5)
init: name='conv1.bias' type=float32 shape=(16,)
init: name='reorder_token_2' type=float32 shape=(16, 16, 5, 5)
init: name='conv2.bias' type=float32 shape=(16,)
init: name='fc1.weight' type=float32 shape=(512, 16)
init: name='fc1.bias' type=float32 shape=(512,)
init: name='fc2.weight' type=float32 shape=(128, 512)
init: name='fc2.bias' type=float32 shape=(128,)
init: name='fc3.weight' type=float32 shape=(10, 128)
init: name='fc3.bias' type=float32 shape=(10,)
init: name='val_3' type=int64 shape=(2,) -- array([ 1, 16])
Conv[com.microsoft.nchwc](x, reorder, conv1.bias, activation=b'Relu', group=1, strides=[1,1], pads=[0,0,0,0], auto_pad=b'NOTSET', dilations=[1,1]) -> reorder_token_0
MaxPool[com.microsoft.nchwc](reorder_token_0, pads=[0,0,0,0], kernel_shape=[2,2], ceil_mode=0, auto_pad=b'NOTSET', dilations=[1,1], strides=[2,2], storage_order=0) -> reorder_token_1
Conv[com.microsoft.nchwc](reorder_token_1, reorder_token_2, conv2.bias, activation=b'Relu', group=1, strides=[1,1], pads=[0,0,0,0], auto_pad=b'NOTSET', dilations=[1,1]) -> reorder_token_3
MaxPool[com.microsoft.nchwc](reorder_token_3, pads=[0,0,0,0], kernel_shape=[2,2], ceil_mode=0, auto_pad=b'NOTSET', dilations=[1,1], strides=[2,2], storage_order=0) -> reorder_token_4
ReorderOutput[com.microsoft.nchwc](reorder_token_4, channels_last=0, channels=16) -> max_pool2d_1
Reshape(max_pool2d_1, val_3, allowzero=0) -> view
FusedGemm[com.microsoft](view, fc1.weight, fc1.bias, transA=0, alpha=1.00, activation=b'Relu', transB=1, beta=1.00) -> relu_2
FusedGemm[com.microsoft](relu_2, fc2.weight, fc2.bias, transA=0, alpha=1.00, activation=b'Relu', transB=1, beta=1.00) -> relu_3
Gemm(relu_3, fc3.weight, fc3.bias, transA=0, alpha=1.00, transB=1, beta=1.00) -> linear_2
output: name='linear_2' type=dtype('float32') shape=[1, 10]
dynamo¶
model = "ort-plot_torch_export_dynamo-cpu-aot1.onnx"
if os.path.exists(model):
print(pretty_onnx(onnx.load(model)))
opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
opset: domain='ai.onnx.ml' version=5
opset: domain='onnx_extended.ortops.optim.cuda' version=1000
opset: domain='ai.onnx.training' version=1
opset: domain='ai.onnx.preview.training' version=1
opset: domain='com.microsoft' version=1
opset: domain='com.microsoft.experimental' version=1
opset: domain='com.microsoft.nchwc' version=1
opset: domain='org.pytorch.aten' version=1
input: name='x' type=dtype('float32') shape=[1, 1, 16, 16]
init: name='reorder' type=float32 shape=(16, 1, 5, 5)
init: name='conv1.bias' type=float32 shape=(16,)
init: name='reorder_token_2' type=float32 shape=(16, 16, 5, 5)
init: name='conv2.bias' type=float32 shape=(16,)
init: name='fc1.weight' type=float32 shape=(512, 16)
init: name='fc1.bias' type=float32 shape=(512,)
init: name='fc2.weight' type=float32 shape=(128, 512)
init: name='fc2.bias' type=float32 shape=(128,)
init: name='fc3.weight' type=float32 shape=(10, 128)
init: name='fc3.bias' type=float32 shape=(10,)
init: name='val_3' type=int64 shape=(2,) -- array([ 1, 16])
Conv[com.microsoft.nchwc](x, reorder, conv1.bias, activation=b'Relu', group=1, strides=[1,1], pads=[0,0,0,0], auto_pad=b'NOTSET', dilations=[1,1]) -> reorder_token_0
MaxPool[com.microsoft.nchwc](reorder_token_0, pads=[0,0,0,0], kernel_shape=[2,2], ceil_mode=0, auto_pad=b'NOTSET', dilations=[1,1], strides=[2,2], storage_order=0) -> reorder_token_1
Conv[com.microsoft.nchwc](reorder_token_1, reorder_token_2, conv2.bias, activation=b'Relu', group=1, strides=[1,1], pads=[0,0,0,0], auto_pad=b'NOTSET', dilations=[1,1]) -> reorder_token_3
MaxPool[com.microsoft.nchwc](reorder_token_3, pads=[0,0,0,0], kernel_shape=[2,2], ceil_mode=0, auto_pad=b'NOTSET', dilations=[1,1], strides=[2,2], storage_order=0) -> reorder_token_4
ReorderOutput[com.microsoft.nchwc](reorder_token_4, channels_last=0, channels=16) -> max_pool2d_1
Reshape(max_pool2d_1, val_3, allowzero=0) -> view
FusedGemm[com.microsoft](view, fc1.weight, fc1.bias, transA=0, alpha=1.00, activation=b'Relu', transB=1, beta=1.00) -> relu_2
FusedGemm[com.microsoft](relu_2, fc2.weight, fc2.bias, transA=0, alpha=1.00, activation=b'Relu', transB=1, beta=1.00) -> relu_3
Gemm(relu_3, fc3.weight, fc3.bias, transA=0, alpha=1.00, transB=1, beta=1.00) -> linear_2
output: name='linear_2' type=dtype('float32') shape=[1, 10]
Show the interesting models for CUDA¶
script¶
model = "ort-plot_torch_export_cus_p2-cuda-aot0.onnx"
if os.path.exists(model):
print(pretty_onnx(onnx.load(model)))
opset: domain='' version=18
opset: domain='ai.onnx.ml' version=5
opset: domain='onnx_extended.ortops.optim.cuda' version=1000
opset: domain='ai.onnx.training' version=1
opset: domain='ai.onnx.preview.training' version=1
opset: domain='com.microsoft' version=1
opset: domain='com.microsoft.experimental' version=1
opset: domain='com.microsoft.nchwc' version=1
opset: domain='org.pytorch.aten' version=1
input: name='input' type=dtype('float32') shape=[1, 1, 16, 16]
init: name='_onx_concat_gatherelements__shape_max_pool2d_1000' type=int64 shape=(2,) -- array([ 1, -1])-- GraphBuilder.constant_folding.from/fold(_onx_gatherelements__shape_max_pool2d_100,init7_s1_-1)##_onx_gatherelements__shape_max_pool2d_100/GraphBuilder.constant_folding.from/fold(_shape_max_pool2d_10,init7_s1_0)##_shape_max_pool2d_10/##init7_s1_0/Opset.make_node.1/Shape##init7_s1_-1/Opset.make_node.1/Shape
init: name='GemmTransposePattern--_onx_transpose_p_fc1_weight0' type=float32 shape=(512, 16)-- GraphBuilder.constant_folding.from/fold(_onx_transpose_p_fc1_weight0)##_onx_transpose_p_fc1_weight0/GraphBuilder.constant_folding.from/fold(p_fc1_weight)##p_fc1_weight/DynamoInterpret.placeholder.1/P(fc1.weight)
init: name='GemmTransposePattern--_onx_transpose_p_fc2_weight0' type=float32 shape=(128, 512)-- GraphBuilder.constant_folding.from/fold(_onx_transpose_p_fc2_weight0)##_onx_transpose_p_fc2_weight0/GraphBuilder.constant_folding.from/fold(p_fc2_weight)##p_fc2_weight/DynamoInterpret.placeholder.1/P(fc2.weight)
init: name='GemmTransposePattern--_onx_transpose_p_fc3_weight0' type=float32 shape=(10, 128)-- GraphBuilder.constant_folding.from/fold(_onx_transpose_p_fc3_weight0)##_onx_transpose_p_fc3_weight0/GraphBuilder.constant_folding.from/fold(p_fc3_weight)##p_fc3_weight/DynamoInterpret.placeholder.1/P(fc3.weight)
init: name='conv1.weight' type=float32 shape=(16, 1, 5, 5) -- DynamoInterpret.placeholder.1/P(conv1.weight)
init: name='conv1.bias' type=float32 shape=(16,) -- DynamoInterpret.placeholder.1/P(conv1.bias)
init: name='conv2.weight' type=float32 shape=(16, 16, 5, 5) -- DynamoInterpret.placeholder.1/P(conv2.weight)
init: name='conv2.bias' type=float32 shape=(16,) -- DynamoInterpret.placeholder.1/P(conv2.bias)
init: name='fc1.bias' type=float32 shape=(512,) -- DynamoInterpret.placeholder.1/P(fc1.bias)
init: name='fc2.bias' type=float32 shape=(128,) -- DynamoInterpret.placeholder.1/P(fc2.bias)
init: name='fc3.bias' type=float32 shape=(10,) -- DynamoInterpret.placeholder.1/P(fc3.bias)
Conv(input, conv1.weight, conv1.bias, dilations=[1,1], group=1, pads=[0,0,0,0], strides=[1,1]) -> conv2d
Relu(conv2d) -> relu
MaxPool(relu, ceil_mode=0, dilations=[1,1], kernel_shape=[2,2], pads=[0,0,0,0], strides=[2,2]) -> max_pool2d
Conv(max_pool2d, conv2.weight, conv2.bias, dilations=[1,1], group=1, pads=[0,0,0,0], strides=[1,1]) -> conv2d_1
Relu(conv2d_1) -> relu_1
MaxPool(relu_1, ceil_mode=0, dilations=[1,1], kernel_shape=[2,2], pads=[0,0,0,0], strides=[2,2]) -> max_pool2d_1
Reshape(max_pool2d_1, _onx_concat_gatherelements__shape_max_pool2d_1000) -> flatten
Gemm(flatten, GemmTransposePattern--_onx_transpose_p_fc1_weight0, fc1.bias, transB=1) -> linear
Relu(linear) -> relu_2
Gemm(relu_2, GemmTransposePattern--_onx_transpose_p_fc2_weight0, fc2.bias, transB=1) -> linear_1
Relu(linear_1) -> relu_3
Gemm(relu_3, GemmTransposePattern--_onx_transpose_p_fc3_weight0, fc3.bias, transB=1) -> output_0
output: name='output_0' type=dtype('float32') shape=[1, 10]
cus_p2¶
model = "ort-plot_torch_export_cus_p2-cuda-aot0.onnx"
if os.path.exists(model):
print(pretty_onnx(onnx.load(model)))
opset: domain='' version=18
opset: domain='ai.onnx.ml' version=5
opset: domain='onnx_extended.ortops.optim.cuda' version=1000
opset: domain='ai.onnx.training' version=1
opset: domain='ai.onnx.preview.training' version=1
opset: domain='com.microsoft' version=1
opset: domain='com.microsoft.experimental' version=1
opset: domain='com.microsoft.nchwc' version=1
opset: domain='org.pytorch.aten' version=1
input: name='input' type=dtype('float32') shape=[1, 1, 16, 16]
init: name='_onx_concat_gatherelements__shape_max_pool2d_1000' type=int64 shape=(2,) -- array([ 1, -1])-- GraphBuilder.constant_folding.from/fold(_onx_gatherelements__shape_max_pool2d_100,init7_s1_-1)##_onx_gatherelements__shape_max_pool2d_100/GraphBuilder.constant_folding.from/fold(_shape_max_pool2d_10,init7_s1_0)##_shape_max_pool2d_10/##init7_s1_0/Opset.make_node.1/Shape##init7_s1_-1/Opset.make_node.1/Shape
init: name='GemmTransposePattern--_onx_transpose_p_fc1_weight0' type=float32 shape=(512, 16)-- GraphBuilder.constant_folding.from/fold(_onx_transpose_p_fc1_weight0)##_onx_transpose_p_fc1_weight0/GraphBuilder.constant_folding.from/fold(p_fc1_weight)##p_fc1_weight/DynamoInterpret.placeholder.1/P(fc1.weight)
init: name='GemmTransposePattern--_onx_transpose_p_fc2_weight0' type=float32 shape=(128, 512)-- GraphBuilder.constant_folding.from/fold(_onx_transpose_p_fc2_weight0)##_onx_transpose_p_fc2_weight0/GraphBuilder.constant_folding.from/fold(p_fc2_weight)##p_fc2_weight/DynamoInterpret.placeholder.1/P(fc2.weight)
init: name='GemmTransposePattern--_onx_transpose_p_fc3_weight0' type=float32 shape=(10, 128)-- GraphBuilder.constant_folding.from/fold(_onx_transpose_p_fc3_weight0)##_onx_transpose_p_fc3_weight0/GraphBuilder.constant_folding.from/fold(p_fc3_weight)##p_fc3_weight/DynamoInterpret.placeholder.1/P(fc3.weight)
init: name='conv1.weight' type=float32 shape=(16, 1, 5, 5) -- DynamoInterpret.placeholder.1/P(conv1.weight)
init: name='conv1.bias' type=float32 shape=(16,) -- DynamoInterpret.placeholder.1/P(conv1.bias)
init: name='conv2.weight' type=float32 shape=(16, 16, 5, 5) -- DynamoInterpret.placeholder.1/P(conv2.weight)
init: name='conv2.bias' type=float32 shape=(16,) -- DynamoInterpret.placeholder.1/P(conv2.bias)
init: name='fc1.bias' type=float32 shape=(512,) -- DynamoInterpret.placeholder.1/P(fc1.bias)
init: name='fc2.bias' type=float32 shape=(128,) -- DynamoInterpret.placeholder.1/P(fc2.bias)
init: name='fc3.bias' type=float32 shape=(10,) -- DynamoInterpret.placeholder.1/P(fc3.bias)
Conv(input, conv1.weight, conv1.bias, dilations=[1,1], group=1, pads=[0,0,0,0], strides=[1,1]) -> conv2d
Relu(conv2d) -> relu
MaxPool(relu, ceil_mode=0, dilations=[1,1], kernel_shape=[2,2], pads=[0,0,0,0], strides=[2,2]) -> max_pool2d
Conv(max_pool2d, conv2.weight, conv2.bias, dilations=[1,1], group=1, pads=[0,0,0,0], strides=[1,1]) -> conv2d_1
Relu(conv2d_1) -> relu_1
MaxPool(relu_1, ceil_mode=0, dilations=[1,1], kernel_shape=[2,2], pads=[0,0,0,0], strides=[2,2]) -> max_pool2d_1
Reshape(max_pool2d_1, _onx_concat_gatherelements__shape_max_pool2d_1000) -> flatten
Gemm(flatten, GemmTransposePattern--_onx_transpose_p_fc1_weight0, fc1.bias, transB=1) -> linear
Relu(linear) -> relu_2
Gemm(relu_2, GemmTransposePattern--_onx_transpose_p_fc2_weight0, fc2.bias, transB=1) -> linear_1
Relu(linear_1) -> relu_3
Gemm(relu_3, GemmTransposePattern--_onx_transpose_p_fc3_weight0, fc3.bias, transB=1) -> output_0
output: name='output_0' type=dtype('float32') shape=[1, 10]
dynopt¶
model = "ort-plot_torch_export_dynopt-cuda-aot1.onnx"
if os.path.exists(model):
print(pretty_onnx(onnx.load(model)))
opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
opset: domain='ai.onnx.ml' version=5
opset: domain='onnx_extended.ortops.optim.cuda' version=1000
opset: domain='ai.onnx.training' version=1
opset: domain='ai.onnx.preview.training' version=1
opset: domain='com.microsoft' version=1
opset: domain='com.microsoft.experimental' version=1
opset: domain='com.microsoft.nchwc' version=1
opset: domain='org.pytorch.aten' version=1
input: name='x' type=dtype('float32') shape=[1, 1, 16, 16]
init: name='conv1.weight' type=float32 shape=(16, 1, 5, 5)
init: name='conv1.bias' type=float32 shape=(16,)
init: name='conv2.weight' type=float32 shape=(16, 16, 5, 5)
init: name='conv2.bias' type=float32 shape=(16,)
init: name='fc1.weight' type=float32 shape=(512, 16)
init: name='fc1.bias' type=float32 shape=(512,)
init: name='fc2.weight' type=float32 shape=(128, 512)
init: name='fc2.bias' type=float32 shape=(128,)
init: name='fc3.weight' type=float32 shape=(10, 128)
init: name='fc3.bias' type=float32 shape=(10,)
init: name='val_3' type=int64 shape=(2,) -- array([ 1, 16])
Conv(x, conv1.weight, conv1.bias, group=1, pads=[0,0,0,0], auto_pad=b'NOTSET', strides=[1,1], dilations=[1,1]) -> conv2d
Relu(conv2d) -> relu
MaxPool(relu, storage_order=0, dilations=[1,1], ceil_mode=0, pads=[0,0,0,0], auto_pad=b'NOTSET', strides=[2,2], kernel_shape=[2,2]) -> max_pool2d
Conv(max_pool2d, conv2.weight, conv2.bias, group=1, pads=[0,0,0,0], auto_pad=b'NOTSET', strides=[1,1], dilations=[1,1]) -> conv2d_1
Relu(conv2d_1) -> relu_1
MaxPool(relu_1, storage_order=0, dilations=[1,1], ceil_mode=0, pads=[0,0,0,0], auto_pad=b'NOTSET', strides=[2,2], kernel_shape=[2,2]) -> max_pool2d_1
Reshape(max_pool2d_1, val_3, allowzero=0) -> view
Gemm(view, fc1.weight, fc1.bias, beta=1.00, transB=1, alpha=1.00, transA=0) -> linear
Relu(linear) -> relu_2
Gemm(relu_2, fc2.weight, fc2.bias, beta=1.00, transB=1, alpha=1.00, transA=0) -> linear_1
Relu(linear_1) -> relu_3
Gemm(relu_3, fc3.weight, fc3.bias, beta=1.00, transB=1, alpha=1.00, transA=0) -> linear_2
output: name='linear_2' type=dtype('float32') shape=[1, 10]
dynamo¶
model = "ort-plot_torch_export_dynamo-cuda-aot1.onnx"
if os.path.exists(model):
print(pretty_onnx(onnx.load(model)))
opset: domain='pkg.onnxscript.torch_lib.common' version=1
opset: domain='' version=18
opset: domain='ai.onnx.ml' version=5
opset: domain='onnx_extended.ortops.optim.cuda' version=1000
opset: domain='ai.onnx.training' version=1
opset: domain='ai.onnx.preview.training' version=1
opset: domain='com.microsoft' version=1
opset: domain='com.microsoft.experimental' version=1
opset: domain='com.microsoft.nchwc' version=1
opset: domain='org.pytorch.aten' version=1
input: name='x' type=dtype('float32') shape=[1, 1, 16, 16]
init: name='conv1.weight' type=float32 shape=(16, 1, 5, 5)
init: name='conv1.bias' type=float32 shape=(16,)
init: name='conv2.weight' type=float32 shape=(16, 16, 5, 5)
init: name='conv2.bias' type=float32 shape=(16,)
init: name='fc1.weight' type=float32 shape=(512, 16)
init: name='fc1.bias' type=float32 shape=(512,)
init: name='fc2.weight' type=float32 shape=(128, 512)
init: name='fc2.bias' type=float32 shape=(128,)
init: name='fc3.weight' type=float32 shape=(10, 128)
init: name='fc3.bias' type=float32 shape=(10,)
init: name='val_3' type=int64 shape=(2,) -- array([ 1, 16])
Conv(x, conv1.weight, conv1.bias, group=1, pads=[0,0,0,0], auto_pad=b'NOTSET', strides=[1,1], dilations=[1,1]) -> conv2d
Relu(conv2d) -> relu
MaxPool(relu, storage_order=0, dilations=[1,1], ceil_mode=0, pads=[0,0,0,0], auto_pad=b'NOTSET', strides=[2,2], kernel_shape=[2,2]) -> max_pool2d
Conv(max_pool2d, conv2.weight, conv2.bias, group=1, pads=[0,0,0,0], auto_pad=b'NOTSET', strides=[1,1], dilations=[1,1]) -> conv2d_1
Relu(conv2d_1) -> relu_1
MaxPool(relu_1, storage_order=0, dilations=[1,1], ceil_mode=0, pads=[0,0,0,0], auto_pad=b'NOTSET', strides=[2,2], kernel_shape=[2,2]) -> max_pool2d_1
Reshape(max_pool2d_1, val_3, allowzero=0) -> view
Gemm(view, fc1.weight, fc1.bias, beta=1.00, transB=1, alpha=1.00, transA=0) -> linear
Relu(linear) -> relu_2
Gemm(relu_2, fc2.weight, fc2.bias, beta=1.00, transB=1, alpha=1.00, transA=0) -> linear_1
Relu(linear_1) -> relu_3
Gemm(relu_3, fc3.weight, fc3.bias, beta=1.00, transB=1, alpha=1.00, transA=0) -> linear_2
output: name='linear_2' type=dtype('float32') shape=[1, 10]
Total running time of the script: (1 minutes 21.352 seconds)
Related examples
