Disable Profiling in Failing Test

[lara-hdr]

Disable profiling in ONNXExporterTester.test_paste_mask_in_image to avoid failure in test.
(pytorch/pytorch#29909)

[codecov-io]

Codecov Report

Merging #1585 into master will increase coverage by 1.61%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master    #1585      +/-   ##
==========================================
+ Coverage   64.34%   65.96%   +1.61%     
==========================================
  Files          90       90              
  Lines        7049     7049              
  Branches      967     1071     +104     
==========================================
+ Hits         4536     4650     +114     
+ Misses       2181     2094      -87     
+ Partials      332      305      -27

Impacted Files	Coverage Δ
torchvision/transforms/functional.py	71.67% <0%> (+0.84%)	⬆️
torchvision/datasets/voc.py	21.64% <0%> (+2.06%)	⬆️
torchvision/transforms/transforms.py	81.04% <0%> (+2.12%)	⬆️
torchvision/datasets/folder.py	82.05% <0%> (+2.56%)	⬆️
torchvision/datasets/utils.py	57.66% <0%> (+3.06%)	⬆️
torchvision/datasets/lsun.py	24.48% <0%> (+4.08%)	⬆️
torchvision/models/detection/roi_heads.py	54.63% <0%> (+5.12%)	⬆️
torchvision/ops/boxes.py	100% <0%> (+5.55%)	⬆️
torchvision/models/detection/rpn.py	82.19% <0%> (+5.93%)	⬆️
torchvision/models/detection/faster_rcnn.py	80.48% <0%> (+6.09%)	⬆️
... and 4 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 4b2f8da...ce2def6. Read the comment docs.

[lara-hdr]

@fmassa, CI is green

[fmassa]

Thanks a lot Lara!

I think this is a temporary solution though, because users of ONNX won't have those flags set, and the model will not be exportable to ONNX.

I'm merging this in the meantime in order to make CI green again.

[lara-hdr]

@fmassa, I agree, but this issue is only happening in the jit executor (only the calls to torch.jit.trace() are failing when mixing tracing and scripting), the ONNX exports are not affected and will work as previously.