Fixed sigma input type for v2.GaussianBlur

[vfdev-5]

Fixes #7885

Description:

• Fixed int type support in v2.GaussianBlur sigma arg
• Added tests

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/vision/7887

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

❌ 3 New Failures, 1 Pending, 1 Unrelated Failure

As of commit 2086d2e with merge base a2f8f8e ():

NEW FAILURES - The following jobs have failed:

• unittests-windows (3.11, windows.4xlarge, cpu) / windows-job (gh)
• unittests-windows (3.8, windows.4xlarge, cpu) / windows-job (gh)
• unittests-windows (3.9, windows.4xlarge, cpu) / windows-job (gh)

FLAKY - The following job failed but was likely due to flakiness present on trunk:

• windows (windows.4xlarge, cpu) / windows-job (gh)

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[NicolasHug]

A few tests are failing, this should hopefully fix a few

[NicolasHug]

LGTM when green, thanks

[github-actions]

Hey @NicolasHug!

You merged this PR, but no labels were added. The list of valid labels is available at https://github.com/pytorch/vision/blob/main/.github/process_commit.py

[huydhn]

@NicolasHug @vfdev-5 I'm creating a release branch for vision and notice these 2 failures coming from this change:

• https://github.com/pytorch/vision/actions/runs/6028057291/job/16360484798 with a build failure

error: more than one operator "=" matches these operands:
             function "__half::operator=(float)"

• and https://github.com/pytorch/vision/actions/runs/6028057287/job/16357033703 with a test failure

ERROR test/test_transforms_v2_refactored.py::TestGaussianBlur::test_transform[sigma1-cuda-make_video] - RuntimeError: CUDA error: device-side assert triggered

Could you help take a look to address them as soon as you could? The other option would probably be to revert this change to unblock the release

cc @atalman @osalpekar

[vfdev-5]

@huydhn first failure is unrelated to this PR as happens during the compilation of roi_pool_kernel.cu:
https://github.com/pytorch/vision/actions/runs/6028057291/job/16360484798#step:11:1405

Same for the second failure as according to logs: https://pipelines.actions.githubusercontent.com/serviceHosts/e1cea170-7cad-4fd4-b94c-3e3f4452f685/_apis/pipelines/1/runs/79880/signedlogcontent/12?urlExpires=2023-08-30T21%3A39%3A38.2279569Z&urlSigningMethod=HMACV1&urlSignature=tOpOBPfJVdDJ1jlYYVPow%2FHRlZRU5YmKqd9hMbUySEE%3D

It starts failing with

2023-08-30T19:22:26.6144348Z ==================================== ERRORS ====================================
2023-08-30T19:22:26.6144773Z �[31m�[1m____ ERROR at teardown of test_detection_model[cuda-retinanet_resnet50_fpn] ____�[0m
2023-08-30T19:22:26.6145147Z �[1m�[31mtest/conftest.py�[0m:120: in prevent_leaking_rng
2023-08-30T19:22:26.6145438Z     torch.cuda.set_rng_state(cuda_rng_state)
2023-08-30T19:22:26.6145882Z �[1m�[31m/opt/conda/envs/ci/lib/python3.8/site-packages/torch/cuda/random.py�[0m:64: in set_rng_state
2023-08-30T19:22:26.6146234Z     _lazy_call(cb)
2023-08-30T19:22:26.6146734Z �[1m�[31m/opt/conda/envs/ci/lib/python3.8/site-packages/torch/cuda/__init__.py�[0m:183: in _lazy_call
2023-08-30T19:22:26.6147103Z     callable()
2023-08-30T19:22:26.6147543Z �[1m�[31m/opt/conda/envs/ci/lib/python3.8/site-packages/torch/cuda/random.py�[0m:62: in cb
2023-08-30T19:22:26.6147942Z     default_generator.set_state(new_state_copy)
2023-08-30T19:22:26.6148326Z �[1m�[31mE   RuntimeError: CUDA error: device-side assert triggered�[0m
2023-08-30T19:22:26.6148817Z �[1m�[31mE   CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.�[0m
2023-08-30T19:22:26.6149313Z �[1m�[31mE   For debugging consider passing CUDA_LAUNCH_BLOCKING=1.�[0m
2023-08-30T19:22:26.6149854Z �[1m�[31mE   Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.�[0m
2023-08-30T19:22:26.6150332Z ----------------------------- Captured stderr call -----------------------------

none of further CUDA tests is passing.

[huydhn]

Let me retry these job one more time. Something else is probably causing these failures then and they are indeed pasing again in the latest vision commit https://hud.pytorch.org/hud/pytorch/vision/main