[TorchFix] Add TorchVision pretrained codemod #4559
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
The latest updates on your projects. Learn more about Vercel for Git ↗︎ 1 Ignored Deployment
|
facebook-github-bot
added
the
CLA Signed
NicolasHug
approved these changes
Sep 8, 2023
| torchvision.models.resnet50(pretrained=False) | ||
| torchvision.models.resnet50(False) | ||
| torchvision.models.detection.ssdlite320_mobilenet_v3_large(pretrained_backbone=True) | ||
| torchvision.models.detection.ssdlite320_mobilenet_v3_large(pretrained=True, pretrained_backbone=True) |
There was a problem hiding this comment.
Could we also add a test entry for resnet_fpn_backbone and also one of the quantization models, e.g. resnet_18?
For these we can't really suggest a weight replacement because it can only be determined at runtime, but maybe we can still show a LintViolation?
IIUC you already handled all of that properly, I just want to confirm through a test.
There was a problem hiding this comment.
Right now quantization models are not supported - nothing will happen.
I was thinking on dealing with that in a follow-up PR as this one is already large.
resnet_fpn_backbone also not included currently. Do you think it's OK to put this in the follow-up PR?
clee2000
approved these changes
Sep 8, 2023
huydhn
approved these changes
Sep 8, 2023
kit1980
added a commit
that referenced
this pull request
Sep 11, 2023
… codemod (#4566) A follow-up to #4559 Meta-internal autodeps feature has issues with imports like `from torchvision.models import MobileNet_V3_Large_Weights`, so changed the logic to just do `from torchvision import models` (if not yet imported) and then use full name from `models` for the weights.
kit1980
added a commit
to pytorch-labs/torchfix
that referenced
this pull request
Nov 23, 2023
… codemod (#4566) A follow-up to pytorch/test-infra#4559 Meta-internal autodeps feature has issues with imports like `from torchvision.models import MobileNet_V3_Large_Weights`, so changed the logic to just do `from torchvision import models` (if not yet imported) and then use full name from `models` for the weights.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
For TorchVision models,
pretrainedandpretrained_backboneparameters have been deprecated in favor of "Multi-weight support API" - see https://pytorch.org/vision/0.15/models.htmlThis PR adds a checker/codemod for the deprecated parameters. As the new weights values needs to be imported first, this PR adds a generic mechanism to collect all needed imports and apply them after all local changes.
Also updated version to "0.0.3" as TorchFix 0.0.2 was released recently and this will be eventually released as 0.0.3