Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Not check avg_time_per_image during test #2665

Merged
merged 5 commits into from
Nov 22, 2023
Merged

Not check avg_time_per_image during test #2665

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
34d4bc3
ignore avg_time_per_image during test
eunwoosh Nov 22, 2023
029f3cf
do not call stdev when length of array is less than 2
eunwoosh Nov 22, 2023
1830b39
ignore avg_time_per_image during regerssion test
eunwoosh Nov 22, 2023
96c8c29
align with pre-commit
eunwoosh Nov 22, 2023
c739aa2
align with pre-commit
eunwoosh Nov 22, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 4 additions & 0 deletions tests/regression/regression_command.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -130,6 +130,8 @@ def regression_openvino_testing(
model_criteria = criteria[template.name] * (1.0 - reg_threshold)

for k in trained_performance.keys():
if k == "avg_time_per_image":
continue
result_dict[k] = round(exported_performance[k], 3)
if exported_performance[k] < model_criteria:
regression_result["passed"] = False
Expand Down Expand Up @@ -180,6 +182,8 @@ def regression_deployment_testing(
modified_criteria = model_criteria - (model_criteria * reg_threshold)

for k in exported_performance.keys():
if k == "avg_time_per_image":
continue
if isinstance(criteria, dict) and template.name in criteria.keys():
result_dict[k] = round(deployed_performance[k], 3)
if deployed_performance[k] < modified_criteria:
Expand Down
30 changes: 14 additions & 16 deletions tests/test_suite/run_test_command.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,7 +10,7 @@
import sys
import torch
from pathlib import Path
from typing import Dict
from typing import Dict, Union
import onnx
import onnxruntime

Expand Down Expand Up @@ -349,11 +349,7 @@ def otx_eval_openvino_testing(
with open(perf_path) as read_file:
exported_performance = json.load(read_file)

for k in trained_performance.keys():
assert (
exported_performance[k] >= trained_performance[k]
or abs(trained_performance[k] - exported_performance[k]) / (trained_performance[k] + 1e-10) <= threshold
), f"{trained_performance[k]=}, {exported_performance[k]=}"
compare_model_accuracy(exported_performance, trained_performance, threshold)


def otx_demo_testing(template, root, otx_dir, args):
Expand Down Expand Up @@ -494,11 +490,7 @@ def otx_eval_deployment_testing(template, root, otx_dir, args, threshold=0.0):
with open(f"{template_work_dir}/deployed_{template.model_template_id}/performance.json") as read_file:
deployed_performance = json.load(read_file)

for k in exported_performance.keys():
assert (
deployed_performance[k] >= exported_performance[k]
or abs(exported_performance[k] - deployed_performance[k]) / (exported_performance[k] + 1e-10) <= threshold
), f"{exported_performance[k]=}, {deployed_performance[k]=}"
compare_model_accuracy(deployed_performance, deployed_performance, threshold)


def otx_demo_deployment_testing(template, root, otx_dir, args):
Expand Down Expand Up @@ -745,11 +737,7 @@ def nncf_eval_testing(template, root, otx_dir, args, threshold=0.01):
with open(f"{template_work_dir}/nncf_{template.model_template_id}/performance.json") as read_file:
evaluated_performance = json.load(read_file)

for k in trained_performance.keys():
assert (
evaluated_performance[k] >= trained_performance[k]
or abs(trained_performance[k] - evaluated_performance[k]) / (trained_performance[k] + 1e-10) <= threshold
), f"{trained_performance[k]=}, {evaluated_performance[k]=}"
compare_model_accuracy(evaluated_performance, trained_performance, threshold)


def nncf_eval_openvino_testing(template, root, otx_dir, args):
Expand Down Expand Up @@ -1174,3 +1162,13 @@ def test_default_for_task(self):
assert num_default_model == 1

return _TestModelTemplates


def compare_model_accuracy(performance_to_test: Dict, target_performance: Dict, threshold: Union[float, int]):
for k in target_performance.keys():
if k == "avg_time_per_image":
continue
assert (
performance_to_test[k] >= target_performance[k]
or abs(target_performance[k] - performance_to_test[k]) / (target_performance[k] + 1e-10) <= threshold
), f"{target_performance[k]=}, {performance_to_test[k]=}"
8 changes: 6 additions & 2 deletions tools/experiment.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -192,11 +192,15 @@ def get_exp_result(self):
def _calculate_avg_std_per_iter(self):
if self._iter_time_arr:
self._exp_result.avg_iter_time = statistics.mean(self._iter_time_arr)
self._exp_result.std_iter_time = statistics.stdev(self._iter_time_arr)
self._exp_result.std_iter_time = (
statistics.stdev(self._iter_time_arr) if len(self._iter_time_arr) > 1 else 0
)

if self._data_time_arr:
self._exp_result.avg_data_time = statistics.mean(self._data_time_arr)
self._exp_result.std_data_time = statistics.stdev(self._data_time_arr)
self._exp_result.std_data_time = (
statistics.stdev(self._data_time_arr) if len(self._data_time_arr) > 1 else 0
)

def _parse_eval_output(self, file_path: Path):
# NOTE: It is assumed that performance.json has key named either score or avg_time_per_image
Expand Down