Re-write getting started guide for transforms V2 #7870
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/vision/7870
Note: Links to docs will display an error until the docs builds have been completed. ❌ 8 New Failures, 2 Unrelated FailuresAs of commit dfeafa2 with merge base 11e49de ( NEW FAILURES - The following jobs have failed:
FLAKY - The following job failed but was likely due to flakiness present on trunk:
BROKEN TRUNK - The following job failed but was present on the merge base:👉 Rebase onto the `viable/strict` branch to avoid these failures
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
| from torchvision.utils import draw_bounding_boxes | ||
|
|
||
|
|
||
| def plot(imgs): |
There was a problem hiding this comment.
I think we need to reduce the boilerplate to a very strict minimum in the examples, because they're already complex enough as it is - we don't need noise.
I opted for the option of creating a new file here. I don't have a super strong opinion on how it should be done, other than the fact that this code shouldn't be part of the example itself.
Hopefully this is non-blocking anyway and we can figure out a long-term solution later, if there are concerns with this one.
| # dataset = wrap_dataset_for_transforms_v2(dataset) | ||
| # # Now the dataset returns datapoints! | ||
| # | ||
| # Using your own datasets |
There was a problem hiding this comment.
I dropped the part about PennFudanDataset here. It seems too specific for this "getting started" guide.
Depending on what pytorch/tutorials#2533 looks like, we may link to it form there.
Co-authored-by: vfdev <vfdev.5@gmail.com>
…nto AAAAAAAAAaaaaaaaaaaa
Co-authored-by: vfdev <vfdev.5@gmail.com>
|
|
||
| assets_directory = pathlib.Path("../assets") | ||
| # If you're trying to run that on collab, you can download the assets and the | ||
| # helpers from https://github.com/pytorch/vision/tree/main/gallery/ |
There was a problem hiding this comment.
| # helpers from https://github.com/pytorch/vision/tree/main/gallery/ | |
| # helpers from https://github.com/pytorch/vision/tree/main/gallery/v2_transforms |
There was a problem hiding this comment.
assets are in gallery
| # If you're trying to run that on collab, you can download the assets and the | ||
| # helpers from https://github.com/pytorch/vision/tree/main/gallery/ | ||
| from helpers import plot | ||
| img = read_image(str(Path('../assets') / 'astronaut.jpg')) |
There was a problem hiding this comment.
This will also not work on Collab without user interference, right?
|
|
||
| return path, image, bounding_boxes, masks, labels | ||
| transforms = v2.Compose([ |
There was a problem hiding this comment.
Should we avoid the plural for the variable name to not clash with the namespace? Somewhat weak argument, since for some datasets we would have transforms=transform. Still, I would prefer keeping it singular here and in general.
There was a problem hiding this comment.
This is sort of related to #7866.
These are transforms, it makes sense to call this variable transforms. If there was a single transform I'd call it transform, or t.
I personally strongly dislike the new trend of importing the module as transforms in our tests.
Co-authored-by: Philip Meier <github.pmeier@posteo.de>
…nto AAAAAAAAAaaaaaaaaaaa
Summary: (Note: this ignores all push blocking failures!) Reviewed By: matteobettini Differential Revision: D48900408 fbshipit-source-id: 04f3cf195318b4b03c527feb05de92d5700717e0 Co-authored-by: vfdev <vfdev.5@gmail.com> Co-authored-by: Philip Meier <github.pmeier@posteo.de>
Will submit re-write of "e2e" example when this is merged.