OSSForks/pytorch
1
0
Fork 0
mirror of https://github.com/zebrajr/pytorch.git synced 2025-12-06 12:20:52 +01:00
Code Issues Actions 38 Packages Projects Releases Wiki Activity

[ONNX] Suppress ort warnings in onnx related test (#67054) (#67804)

Browse Source 
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/67804

Improve readability of test logs by suppressing ort warnings logging for onnx related test.

Reducing ONNX CI test log binary size:
linux-xenial-py3.6-clang7-onnx-test1: 12443 KB -> 6958 KB
linux-xenial-py3.6-clang7-onnx-test2: 16884 KB -> 5778 KB

Test Plan: Imported from OSS

Reviewed By: msaroufim

Differential Revision: D32181308

Pulled By: malfet

fbshipit-source-id: 11cf165dc212d061606590e96c08c6e021135f74

Co-authored-by: BowenBao<bowbao@microsoft.com>
This commit is contained in:
Bowen Bao 2021-11-08 14:29:12 -08:00 committed by Facebook GitHub Bot
parent ea60e7d559
commit ead59b5ff3
1 changed files with 8 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
test/onnx/test_pytorch_onnx_onnxruntime.py
View File

@ -93,7 +93,11 @@ def convert_to_onnx(model, input=None, opset_version=9, do_constant_folding=True
onnx_shape_inference=onnx_shape_inference)
# compute onnxruntime output prediction
ort_sess = onnxruntime.InferenceSession(f.getvalue())
so = onnxruntime.SessionOptions()
# suppress ort warnings.
# 0:Verbose, 1:Info, 2:Warning. 3:Error, 4:Fatal. Default is 2.
so.log_severity_level = 3
ort_sess = onnxruntime.InferenceSession(f.getvalue(), so)
return ort_sess
@ -359,6 +363,9 @@ class TestONNXRuntime(unittest.TestCase):
ort_sess_opt.graph_optimization_level = \
onnxruntime.GraphOptimizationLevel.ORT_ENABLE_EXTENDED if ort_optim_on else \
onnxruntime.GraphOptimizationLevel.ORT_DISABLE_ALL
# suppress ort warnings.
# 0:Verbose, 1:Info, 2:Warning. 3:Error, 4:Fatal. Default is 2.
ort_sess_opt.log_severity_level = 3
ort_sess = onnxruntime.InferenceSession(model_file_name, sess_options=ort_sess_opt)
input_copy = copy.deepcopy(input)
ort_outs = run_ort(ort_sess, input_copy)