torchvision v0.11.1

[regro-cf-autotick-bot]

It is very likely that the current package version for this feedstock is out of date.
Notes for merging this PR:

1. Feel free to push to the bot's branch to update this PR if needed.
2. The bot will almost always only open one PR per version.
   Checklist before merging this PR:

• Dependencies have been updated if changed: see upstream
• Tests have passed
• Updated license if changed and license_file is packaged

Note that the bot will stop issuing PRs if more than 3 Version bump PRs generated by the bot are open. If you don't want to package a particular version please close the PR.

NEW: If you want these PRs to be merged automatically, make an issue with @conda-forge-admin,please add bot automerge in the title and merge the resulting PR. This command will add our new bot automerge feature to your feedstock!

If this PR was opened in error or needs to be updated please add the bot-rerun label to this PR. The bot will close this PR and schedule another one. If you do not have permissions to add this label, you can use the phrase @conda-forge-admin, please rerun bot in a PR comment to have the conda-forge-admin add it for you.

_{This PR was created by the regro-cf-autotick-bot.
The regro-cf-autotick-bot is a service to automatically track the dependency graph, migrate packages, and propose package version updates for conda-forge. If you would like a local version of this bot, you might consider using rever. Rever is a tool for automating software releases and forms the backbone of the bot's conda-forge PRing capability. Rever is both conda (conda install -c conda-forge rever) and pip (pip install re-ver) installable.
Finally, feel free to drop us a line if there are any issues!
This PR was generated by https://github.com/regro/autotick-bot/actions/runs/1369210201, please use this URL for debugging}

Dependency Analysis

We couldn't run dependency analysis due to an internal error in the bot. :( Help is very welcome!

[conda-forge-linter]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

[hmaarrfk]

/_h_env_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_pla=/usr/local/src/conda-prefix -DNDEBUG -D_FORTIFY_SOURCE=2 -O2 -isystem /home/conda/feedstock_root/build_artifacts/torchvision-split_1634859216379/_h_env_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_pla/include -fPIC -I/home/conda/feedstock_root/build_artifacts/torchvision-split_1634859216379/work/torchvision/csrc -I/home/conda/feedstock_root/build_artifacts/torchvision-split_1634859216379/_h_env_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_pla/lib/python3.7/site-packages/torch/include -I/home/conda/feedstock_root/build_artifacts/torchvision-split_1634859216379/_h_env_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_pla/lib/python3.7/site-packages/torch/include/torch/csrc/api/include -I/home/conda/feedstock_root/build_artifacts/torchvision-split_1634859216379/_h_env_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_pla/lib/python3.7/site-packages/torch/include/TH -I/home/conda/feedstock_root/build_artifacts/torchvision-split_1634859216379/_h_env_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_pla/lib/python3.7/site-packages/torch/include/THC -I/home/conda/feedstock_root/build_artifacts/torchvision-split_1634859216379/_h_env_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_pla/include/ -I/home/conda/feedstock_root/build_artifacts/torchvision-split_1634859216379/_h_env_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_pla/include/python3.7m -c -c /home/conda/feedstock_root/build_artifacts/torchvision-split_1634859216379/work/torchvision/csrc/ops/autograd/deform_conv2d_kernel.cpp -o /home/conda/feedstock_root/build_artifacts/torchvision-split_1634859216379/work/build/temp.linux-x86_64-3.7/home/conda/feedstock_root/build_artifacts/torchvision-split_1634859216379/work/torchvision/csrc/ops/autograd/deform_conv2d_kernel.o -DTORCH_API_INCLUDE_EXTENSION_H -DTORCH_EXTENSION_NAME=_C -D_GLIBCXX_USE_CXX11_ABI=1 -std=gnu++14
2021-10-21T23:44:45.8719828Z     cc1plus: warning: command line option '-Wstrict-prototypes' is valid for C/ObjC but not for C++
2021-10-21T23:44:45.8722609Z     /home/conda/feedstock_root/build_artifacts/torchvision-split_1634859216379/work/torchvision/csrc/ops/autograd/deform_conv2d_kernel.cpp: In static member function 'static torch::autograd::variable_list vision::ops::{anonymous}::DeformConv2dFunction::forward(torch::autograd::AutogradContext*, const Variable&, const Variable&, const Variable&, const Variable&, const Variable&, int64_t, int64_t, int64_t, int64_t, int64_t, int64_t, int64_t, int64_t, bool)':
2021-10-21T23:44:45.8724576Z     /home/conda/feedstock_root/build_artifacts/torchvision-split_1634859216379/work/torchvision/csrc/ops/autograd/deform_conv2d_kernel.cpp:30:9: error: 'AutoDispatchBelowADInplaceOrView' is not a member of 'at'
2021-10-21T23:44:45.8725354Z          at::AutoDispatchBelowADInplaceOrView g;

[github-actions]

Hi! This is the friendly conda-forge automerge bot!

I considered the following status checks when analyzing this PR:

• linter: passed
• azure: failed

Thus the PR was not passing and not merged.

[hmaarrfk]

It might be that this needs pytorch 1.10

[h-vetinari]

It might be that this needs pytorch 1.10

That's how I read the release tag as well (despite the version typo) - note also that 0.11.1 got released so fast after 0.11.0 that the bot even didn't "see it".

[benjaminrwilson]

@conda-forge-admin, please restart ci

[conda-forge-linter]

Hi! This is the friendly automated conda-forge-linting service.

I wanted to let you know that I linted all conda-recipes in your PR (recipe) and found some lint.

Here's what I've got...

For recipe:

• Failed to even lint the recipe, probably because of a conda-smithy bug 😢. This likely indicates a problem in your meta.yaml, though. To get a traceback to help figure out what's going on, install conda-smithy and run conda smithy recipe-lint . from the recipe directory.

[conda-forge-linter]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

[hmaarrfk]

Ok so it seems I got the tests to get to completion on the CIs. However the tests seem to fail.

It is a little suspicious since the tests that fail all seem to have the word "random" in them.

Similar tests were failing on a local build.

[hmaarrfk]

@conda-forge-admin please restart cis

[hmaarrfk]

Three test classes can be found for the failing tests:

KineticsTestCase

2021-10-31T18:46:02.5780932Z test/test_datasets.py::KineticsTestCase::test_feature_types FAILED       [  0%]
2021-10-31T18:46:02.5823257Z test/test_datasets.py::KineticsTestCase::test_not_found_or_corrupted PASSED [  0%]
2021-10-31T18:46:02.7326330Z test/test_datasets.py::KineticsTestCase::test_num_examples FAILED        [  0%]
2021-10-31T18:46:02.8813912Z test/test_datasets.py::KineticsTestCase::test_smoke FAILED               [  0%]
2021-10-31T18:46:03.0317105Z test/test_datasets.py::KineticsTestCase::test_str_smoke FAILED           [  0%]
2021-10-31T18:46:03.1845217Z test/test_datasets.py::KineticsTestCase::test_transforms FAILED          [  0%]

TestDatasetsSamplers

2021-10-31T18:47:58.7446167Z test/test_datasets_samplers.py::TestDatasetsSamplers::test_random_clip_sampler FAILED [  1%]
2021-10-31T18:48:00.4872052Z test/test_datasets_samplers.py::TestDatasetsSamplers::test_random_clip_sampler_unequal FAILED [  1%]
2021-10-31T18:48:02.5970438Z test/test_datasets_samplers.py::TestDatasetsSamplers::test_uniform_clip_sampler FAILED [  1%]
2021-10-31T18:48:04.3041685Z test/test_datasets_samplers.py::TestDatasetsSamplers::test_uniform_clip_sampler_insufficient_clips FAILED [  1%]
2021-10-31T18:48:06.3568181Z test/test_datasets_samplers.py::TestDatasetsSamplers::test_distributed_sampler_and_uniform_clip_sampler FAILED [  1%]

TestVideo

2021-10-31T18:48:07.6010877Z test/test_datasets_video_utils.py::TestVideo::test_video_clips FAILED    [  1%]
2021-10-31T18:49:58.5084616Z test/test_io.py::TestVideo::test_write_read_video FAILED                 [ 89%]

[h-vetinari]

Could this have something to do with the fact that we're not building with ffmpeg support?

[hmaarrfk]

that is likely it. how do we do this?

[ngam]

On the "random" issue @hmaarrfk brought up, there was something in the new release about changing from scipy to torch to generate random samples, I suppose that this will be automatically implemented here, right? It's just another data point in the puzzle

[h-vetinari]

that is likely it. how do we do this?

To be honest, I haven't checked. I'm pretty sure it's gonna be a build flag. See also #22 - I think there's an argument to be made for several output (well, depending on how difficult it is to build) for some reasonable set of

@hmaarrfk: It is a little suspicious since the tests that fail all seem to have the word "random" in them.

@ngam: On the "random" issue @hmaarrfk brought up, there was something in the new release about changing from scipy to torch to generate random samples,

I overlooked Mark's comment at first - do you have a link to that?

I suppose that this will be automatically implemented here, right?

If it's part of the torchvision code, then yes. If it's part of the pytorch build, it's too early to say (though I'm surprised this is being changed in a patch release).

[ngam]

@h-vetinari here is the link, https://github.com/pytorch/vision/releases/tag/v0.11.0, midway through the page:

Backward-incompatible changes
[models] Use torch instead of scipy for random initialization of inception and googlenet weights (#4256)

Issue link in quote: pytorch/vision#4256, which suggests the person changed the code itself, but I am not sure...

[ngam]

fwiw: https://anaconda.org/fastchan/torchvision but not for the noarch

[hmaarrfk]

unless the maintainer explicitly removed it. the recipe is included in the archive. are you able to do a diff with your eyes to see what we are missing?

[ngam]

I will need to look deeply. That's why I wanted to do a crash course by making a recipe for torchaudio 😅 to be able to see this more intelligently/clearly

I will find more details and report here. I will also try to see how the fastai people managed to build the latest versions (they seem to have a significantly laxer workflow in their repo)

[hmaarrfk]

@conda-forge-admin please rerender

[ngam]

@h-vetinari @hmaarrfk small update:

I managed to build this locally and uploaded it here https://anaconda.org/ngam/torchvision -- I am still figuring out to how to translate that into this situation here. One "major" problem I faced: switching from the default pytorch==1.11.0 to pytorch==1.10.0 to get this to work. Otherwise, it will want to find the ARM version of pytorch==1.11.0 from their nightly channel and that was causing a lot of problems downstream for me. I have no idea if this will be relevant to your case. Will keep investigating...

export 'CONDA_PYTORCH_CONSTRAINT=- pytorch==1.10.0'
export 'CONDA_PYTORCH_BUILD_CONSTRAINT=- pytorch==1.10.0'

[ngam]

re av: @hmaarrfk i was looking at the open-ce repo (link: https://github.com/open-ce/torchvision-feedstock) and they had it and so I copied it, and well ... the results are in: went from 3 successful and 21 failing to 1 successful and 23 failing... so I guess that ain't it

[ngam]

@hmaarrfk I am having a pretty hard time focusing on this to better identify the problem and play around (hence started a new pr to experiment) --- is there an easy way to suppress all the variants and build just a vanilla linux_64? If so, and you do have the time, please edit my other pr #33 or point me to it

[hmaarrfk]

ffmpeg needs a few extra libraries:
https://github.com/pytorch/vision/blob/v0.11.1/setup.py#L371

            'libavcodec',
            'libavformat',
            'libavutil',
            'libswresample',
            'libswscale'

[hmaarrfk]

In either case, pytorch knows it crashes in osx:

    ffmpeg_exe = distutils.spawn.find_executable('ffmpeg')
    has_ffmpeg = ffmpeg_exe is not None
    # FIXME: Building torchvision with ffmpeg on MacOS or with Python 3.9
    # FIXME: causes crash. See the following GitHub issues for more details.
    # FIXME: https://github.com/pytorch/pytorch/issues/65000
    # FIXME: https://github.com/pytorch/vision/issues/3367

[hmaarrfk]

@seemethere a while back you offered to help release pytorch. We've been at this feedstock for a while and I'm wondering if you could take a quick look at the failures.

Your help would be greatly appreciated.

Thank you!