Skip to content

Add FeaturePyramidBackbone and port weights from timm for ResNetBackbone #1769

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 7 commits into from
Aug 15, 2024
Merged

Add FeaturePyramidBackbone and port weights from timm for ResNetBackbone #1769

merged 7 commits into from
Aug 15, 2024

Conversation

james77777778
Copy link
Collaborator

@james77777778 james77777778 commented Aug 13, 2024 •
edited
Loading

This PR introduces FeaturePyramidBackbone, a wrapper for Backbone that adds pyramid_outputs property.
If a vision backbone supports feature pyramids, it should subclass FeaturePyramidBackbone.

I modified ResNetBackbone by subclassing FeaturePyramidBackbone to include feature pyramid information.

Also, head_dtype is added in ResNetImageClassifier.

@divyashreepathihalli @mattdangerw @SamanehSaadat

EDITED:
See #1769 (comment) for updates

@mattdangerw
Copy link
Member

@james77777778 I was thinking we would make FeaturePyramidBackbone a simple subclass of Backbone for most CV backbones.

So...

  • Backbone is basically just a functional model with from_preset.
  • FeaturePyramidBackbone extends Backbone with extra pyramid_outputs.
  • ResNetBackbone extends FeaturePyramidBackbone directly.

The goal is to keep the Backbone base class as clean as we can, now that we are venturing into multi modal models with a lot of different overall patterns. dir(bert_backbone) shouldn't have feature pyramid stuff in it. If we could move the token_embedding off the backbone and into a TextBackbone or similar without breaking compat we probably would too.

Most CV models like ResNet, DenseNet, EfficientNet, etc, will be FeaturePyramidBackbones. But something like ViT can just subclass Backbone directly without needing the feature pyramid outputs.

WDYT?

@james77777778
Copy link
Collaborator Author

james77777778 commented Aug 14, 2024 •
edited
Loading

@mattdangerw @divyashreepathihalli
Got it. That makes sense. Updated!

I have updated the PR to make compatible with timm. Additionally, the conversion logic has been added, similar to how it’s done in transformers.

Please refer to this colab for the numerical check:
https://colab.research.google.com/drive/1QnmNDiFYd56fsYoaUM46QRT4gF9G06fH?usp=sharing

Supported:

  • V1: resnet18.a1_in1k, resnet26.bt_in1k, resnet34.a1_in1k, resnet50.a1_in1k, resnet101.a1h_in1k, resnet152.a1h_in1k
  • V2: resnetv2_50.a1h_in1k, resnetv2_101.a1h_in1k

@james77777778 james77777778 changed the title Add FeaturePyramidBackbone and ResNetFeaturePyramidBackbone Add FeaturePyramidBackbone and port weights from timm for ResNetBackbone Aug 14, 2024
@james77777778 james77777778 mentioned this pull request Aug 14, 2024
Merged
mattdangerw
mattdangerw reviewed Aug 14, 2024
View reviewed changes
Copy link
Member

@mattdangerw mattdangerw left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Awesome great work! Still a few comments to resolve, but very cool we are building timm conversion into the library.

keras_nlp/src/utils/timm/safetensor_utils.py Outdated
@@ -0,0 +1,68 @@
# Copyright 2024 The KerasNLP Authors
Copy link
Member

@mattdangerw mattdangerw Aug 14, 2024 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is this the same as the one for transformers? Can we use the same file? Can just leave it where it is for now and import it.

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I have added a argument preset to make compatible with timm.
Now we use the same file.

keras_nlp/src/utils/preset_utils.py Outdated
@@ -50,6 +50,7 @@
KAGGLE_PREFIX = "kaggle://"
GS_PREFIX = "gs://"
HF_PREFIX = "hf://"
TIMM_PREFIX = "hf://timm"
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think we can merge like this for now to keep things moving, but we should probably allow converting timm models outside of the timm "org" on huggingface. We might need to start parsing the architecture/architectures values transformers/timm config.json. At least that seems like the best place to infer where things are loading from.

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I updated the code to parse these values in order to determine the format.
It's a bit fragile. We might need better parsing in the future.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Let's merge for now and update later!

@divyashreepathihalli divyashreepathihalli added the kokoro:force-run Runs Tests on GPU label Aug 14, 2024
@kokoro-team kokoro-team removed the kokoro:force-run Runs Tests on GPU label Aug 14, 2024
@divyashreepathihalli
Copy link
Collaborator

Looks really good! Its awesome to have built in timm conversion! LGTM!

@divyashreepathihalli divyashreepathihalli added the kokoro:force-run Runs Tests on GPU label Aug 15, 2024
@kokoro-team kokoro-team removed the kokoro:force-run Runs Tests on GPU label Aug 15, 2024
@mattdangerw mattdangerw merged commit 00ab4d5 into keras-team:keras-hub Aug 15, 2024
10 checks passed
@mattdangerw
Copy link
Member

Thanks!

@james77777778 james77777778 deleted the add-feature-pyramid branch August 16, 2024 01:45
@mattdangerw mattdangerw mentioned this pull request Aug 16, 2024
Merged
mattdangerw pushed a commit to mattdangerw/keras-hub that referenced this pull request Sep 10, 2024
@james77777778 @mattdangerw
Add FeaturePyramidBackbone and port weights from timm for `ResNet…
3a1f34f 
…Backbone` (keras-team#1769)

* Add FeaturePyramidBackbone and update ResNetBackbone

* Simplify the implementation

* Fix CI

* Make ResNetBackbone compatible with timm and add FeaturePyramidBackbone

* Add conversion implementation

* Update docstrings

* Address comments
mattdangerw pushed a commit that referenced this pull request Sep 11, 2024
@james77777778 @mattdangerw
Add FeaturePyramidBackbone and port weights from timm for `ResNet…
7381b6d 
…Backbone` (#1769)

* Add FeaturePyramidBackbone and update ResNetBackbone

* Simplify the implementation

* Fix CI

* Make ResNetBackbone compatible with timm and add FeaturePyramidBackbone

* Add conversion implementation

* Update docstrings

* Address comments
mattdangerw pushed a commit that referenced this pull request Sep 13, 2024
@james77777778 @mattdangerw
Add FeaturePyramidBackbone and port weights from timm for `ResNet…
4b20620 
…Backbone` (#1769)

* Add FeaturePyramidBackbone and update ResNetBackbone

* Simplify the implementation

* Fix CI

* Make ResNetBackbone compatible with timm and add FeaturePyramidBackbone

* Add conversion implementation

* Update docstrings

* Address comments
mattdangerw pushed a commit that referenced this pull request Sep 17, 2024
@james77777778 @mattdangerw
Add FeaturePyramidBackbone and port weights from timm for `ResNet…
2861676 
…Backbone` (#1769)

* Add FeaturePyramidBackbone and update ResNetBackbone

* Simplify the implementation

* Fix CI

* Make ResNetBackbone compatible with timm and add FeaturePyramidBackbone

* Add conversion implementation

* Update docstrings

* Address comments
divyashreepathihalli added a commit that referenced this pull request Sep 25, 2024
@james77777778 @divyashreepathihalli
@sachinprasadhs @mattdangerw @sineeli @ushareng @pkgoogle @gowthamkpr
Add StableDiffusion3 (#1820)
743adea 
* Add VGG16 backbone (#1737)

* Agg Vgg16 backbone

* update names

* update tests

* update test

* add image classifier

* incorporate review comments

* Update test case

* update backbone test

* add image classifier

* classifier cleanup

* code reformat

* add vgg16 image classifier

* make vgg generic

* update doc string

* update docstring

* add classifier test

* update tests

* update docstring

* address review comments

* code reformat

* update the configs

* address review comments

* fix task saved model test

* update init

* code reformatted

* Add `ResNetBackbone` and `ResNetImageClassifier` (#1765)

* Add ResNetV1 and ResNetV2

* Address comments

* Add CSP DarkNet backbone and classifier (#1774)

* Add CSP DarkNet

* Add CSP DarkNet

* snake_case function names

* change use_depthwise to block_type

* Add `FeaturePyramidBackbone` and port weights from `timm` for `ResNetBackbone` (#1769)

* Add FeaturePyramidBackbone and update ResNetBackbone

* Simplify the implementation

* Fix CI

* Make ResNetBackbone compatible with timm and add FeaturePyramidBackbone

* Add conversion implementation

* Update docstrings

* Address comments

* Add DenseNet (#1775)

* Add DenseNet

* fix testcase

* address comments

* nit

* fix lint errors

* move description

* Add ViTDetBackbone (#1776)

* add vit det vit_det_backbone

* update docstring

* code reformat

* fix tests

* address review comments

* bump year on all files

* address review comments

* rename backbone

* fix tests

* change back to ViT

* address review comments

* update image shape

* Add Mix transformer (#1780)

* Add MixTransformer

* fix testcase

* test changes and comments

* lint fix

* update config list

* modify testcase for 2 layers

* update input_image_shape -> image_shape (#1785)

* update input_image_shape -> image_shape

* update docstring example

* code reformat

* update tests

* Create __init__.py (#1788)

add missing __init__ file to vit_det

* Hack package build script to rename to keras-hub (#1793)

This is a temporary way to test out the keras-hub branch.
- Does a global rename of all symbols during package build.
- Registers the "old" name on symbol export for saving compat.
- Adds a github action to publish every commit to keras-hub as
  a new package.
- Removes our descriptions on PyPI temporarily, until we want
  to message this more broadly.

* Add CLIP and T5XXL for StableDiffusionV3 (#1790)

* Add `CLIPTokenizer`, `T5XXLTokenizer`, `CLIPTextEncoder` and `T5XXLTextEncoder`.

* Make CLIPTextEncoder as Backbone

* Add `T5XXLPreprocessor` and remove `T5XXLTokenizer`

Add `CLIPPreprocessor`

* Use `tf = None` at the top

* Replace manual implementation of `CLIPAttention` with `MultiHeadAttention`

* Add Bounding Box Utils (#1791)

* Bounding box utils

* - Correct test cases

* - Remove hard tensorflow dtype

* - fix api gen

* - Fix import for test cases
- Use setup for converters test case

* - fix api_gen issue

* - FIx api gen

* - Fix api gen error

* - Correct test cases as per new api changes

* mobilenet_v3 added in keras-nlp (#1782)

* mobilenet_v3 added in keras-nlp

* minor bug fixed in mobilenet_v3_backbone

* formatting corrected

* refactoring backbone

* correct_pad_downsample method added

* refactoring backbone

* parameters updated

* Testcaseupdated, expected output shape corrected

* code formatted with black

* testcase updated

* refactoring and description added

* comments updated

* added mobilenet v1 and v2

* merge conflict resolved

* version arg removed, and config options added

* input_shape changed to image_shape in arg

* config updated

* input shape corrected

* comments resolved

* activation function format changed

* minor bug fixed

* minor bug fixed

* added vision_backbone_test

* channel_first bug resolved

* channel_first cases working

* comments  resolved

* formatting fixed

* refactoring

---------

Co-authored-by: ushareng <usha.rengaraju@gmail.com>

* Pkgoogle/efficient net migration (#1778)

* migrating efficientnet models to keras-hub

* merging changes from other sources

* autoformatting pass

* initial consolidation of efficientnet_backbone

* most updates and removing separate implementation

* cleanup, autoformatting, keras generalization

* removed layer examples outside of effiicient net

* many, mainly documentation changes, small test fixes

* Add the ResNet_vd backbone (#1766)

* Add ResNet_vd to ResNet backbone

* Addressed requested parameter changes

* Fixed tests and updated comments

* Added new parameters to docstring

* Add `VAEImageDecoder` for StableDiffusionV3 (#1796)

* Add `VAEImageDecoder` for StableDiffusionV3

* Use `keras.Model` for `VAEImageDecoder` and follows the coding style in `VAEAttention`

* Replace `Backbone` with `keras.Model` in `CLIPTextEncoder` and `T5XXLTextEncoder` (#1802)

* Add pyramid output for densenet, cspDarknet (#1801)

* add pyramid outputs

* fix testcase

* format fix

* make common testcase for pyramid outputs

* change default shape

* simplify testcase

* test case change and add channel axis

* Add `MMDiT` for StableDiffusionV3 (#1806)

* Add `MMDiT`

* Update

* Update

* Update implementation

* Add remaining bbox utils (#1804)

* - Add formats, iou, utils for bounding box

* - Add `AnchorGenerator`, `BoxMatcher` and `NonMaxSupression` layers

* - Remove scope_name  not required.

* use default keras name scope

* - Correct format error

* - Remove layers as of now and keep them at model level till keras core supports them

* - Correct api_gen

* Fix timm conversion for rersnet (#1814)

* Add `StableDiffusion3`

* Fix `_normalize_inputs`

* Separate CLIP encoders from SD3 backbone.

* Simplify `text_to_image` function.

* Address comments

* Minor update and add docstrings.

* Add VGG16 backbone (#1737)

* Agg Vgg16 backbone

* update names

* update tests

* update test

* add image classifier

* incorporate review comments

* Update test case

* update backbone test

* add image classifier

* classifier cleanup

* code reformat

* add vgg16 image classifier

* make vgg generic

* update doc string

* update docstring

* add classifier test

* update tests

* update docstring

* address review comments

* code reformat

* update the configs

* address review comments

* fix task saved model test

* update init

* code reformatted

* Add `ResNetBackbone` and `ResNetImageClassifier` (#1765)

* Add ResNetV1 and ResNetV2

* Address comments

* Add CSP DarkNet backbone and classifier (#1774)

* Add CSP DarkNet

* Add CSP DarkNet

* snake_case function names

* change use_depthwise to block_type

* Add `FeaturePyramidBackbone` and port weights from `timm` for `ResNetBackbone` (#1769)

* Add FeaturePyramidBackbone and update ResNetBackbone

* Simplify the implementation

* Fix CI

* Make ResNetBackbone compatible with timm and add FeaturePyramidBackbone

* Add conversion implementation

* Update docstrings

* Address comments

* Add DenseNet (#1775)

* Add DenseNet

* fix testcase

* address comments

* nit

* fix lint errors

* move description

* Add ViTDetBackbone (#1776)

* add vit det vit_det_backbone

* update docstring

* code reformat

* fix tests

* address review comments

* bump year on all files

* address review comments

* rename backbone

* fix tests

* change back to ViT

* address review comments

* update image shape

* Add Mix transformer (#1780)

* Add MixTransformer

* fix testcase

* test changes and comments

* lint fix

* update config list

* modify testcase for 2 layers

* update input_image_shape -> image_shape (#1785)

* update input_image_shape -> image_shape

* update docstring example

* code reformat

* update tests

* Create __init__.py (#1788)

add missing __init__ file to vit_det

* Hack package build script to rename to keras-hub (#1793)

This is a temporary way to test out the keras-hub branch.
- Does a global rename of all symbols during package build.
- Registers the "old" name on symbol export for saving compat.
- Adds a github action to publish every commit to keras-hub as
  a new package.
- Removes our descriptions on PyPI temporarily, until we want
  to message this more broadly.

* Add CLIP and T5XXL for StableDiffusionV3 (#1790)

* Add `CLIPTokenizer`, `T5XXLTokenizer`, `CLIPTextEncoder` and `T5XXLTextEncoder`.

* Make CLIPTextEncoder as Backbone

* Add `T5XXLPreprocessor` and remove `T5XXLTokenizer`

Add `CLIPPreprocessor`

* Use `tf = None` at the top

* Replace manual implementation of `CLIPAttention` with `MultiHeadAttention`

* Add Bounding Box Utils (#1791)

* Bounding box utils

* - Correct test cases

* - Remove hard tensorflow dtype

* - fix api gen

* - Fix import for test cases
- Use setup for converters test case

* - fix api_gen issue

* - FIx api gen

* - Fix api gen error

* - Correct test cases as per new api changes

* mobilenet_v3 added in keras-nlp (#1782)

* mobilenet_v3 added in keras-nlp

* minor bug fixed in mobilenet_v3_backbone

* formatting corrected

* refactoring backbone

* correct_pad_downsample method added

* refactoring backbone

* parameters updated

* Testcaseupdated, expected output shape corrected

* code formatted with black

* testcase updated

* refactoring and description added

* comments updated

* added mobilenet v1 and v2

* merge conflict resolved

* version arg removed, and config options added

* input_shape changed to image_shape in arg

* config updated

* input shape corrected

* comments resolved

* activation function format changed

* minor bug fixed

* minor bug fixed

* added vision_backbone_test

* channel_first bug resolved

* channel_first cases working

* comments  resolved

* formatting fixed

* refactoring

---------

Co-authored-by: ushareng <usha.rengaraju@gmail.com>

* Pkgoogle/efficient net migration (#1778)

* migrating efficientnet models to keras-hub

* merging changes from other sources

* autoformatting pass

* initial consolidation of efficientnet_backbone

* most updates and removing separate implementation

* cleanup, autoformatting, keras generalization

* removed layer examples outside of effiicient net

* many, mainly documentation changes, small test fixes

* Add the ResNet_vd backbone (#1766)

* Add ResNet_vd to ResNet backbone

* Addressed requested parameter changes

* Fixed tests and updated comments

* Added new parameters to docstring

* Add `VAEImageDecoder` for StableDiffusionV3 (#1796)

* Add `VAEImageDecoder` for StableDiffusionV3

* Use `keras.Model` for `VAEImageDecoder` and follows the coding style in `VAEAttention`

* Replace `Backbone` with `keras.Model` in `CLIPTextEncoder` and `T5XXLTextEncoder` (#1802)

* Add pyramid output for densenet, cspDarknet (#1801)

* add pyramid outputs

* fix testcase

* format fix

* make common testcase for pyramid outputs

* change default shape

* simplify testcase

* test case change and add channel axis

* Add `MMDiT` for StableDiffusionV3 (#1806)

* Add `MMDiT`

* Update

* Update

* Update implementation

* Add remaining bbox utils (#1804)

* - Add formats, iou, utils for bounding box

* - Add `AnchorGenerator`, `BoxMatcher` and `NonMaxSupression` layers

* - Remove scope_name  not required.

* use default keras name scope

* - Correct format error

* - Remove layers as of now and keep them at model level till keras core supports them

* - Correct api_gen

* Fix timm conversion for rersnet (#1814)

* Fix

* Update

* Rename to diffuser and decoder

* Define functional model

* Merge from upstream/master

* Delete old SD3

* Fix copyright

* Rename to keras_hub

* Address comments

* Update

* Fix CI

* Fix bugs occurred in keras3.1

---------

Co-authored-by: Divyashree Sreepathihalli <divyashreepathihalli@gmail.com>
Co-authored-by: Sachin Prasad <sachinprasad@google.com>
Co-authored-by: Matt Watson <1389937+mattdangerw@users.noreply.github.com>
Co-authored-by: Siva Sravana Kumar Neeli <113718461+sineeli@users.noreply.github.com>
Co-authored-by: Usha Rengaraju <34335028+ushareng@users.noreply.github.com>
Co-authored-by: ushareng <usha.rengaraju@gmail.com>
Co-authored-by: pkgoogle <132095473+pkgoogle@users.noreply.github.com>
Co-authored-by: gowthamkpr <47574994+gowthamkpr@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@divyashreepathihalli divyashreepathihalli divyashreepathihalli approved these changes

@SamanehSaadat SamanehSaadat Awaiting requested review from SamanehSaadat

@mattdangerw mattdangerw Awaiting requested review from mattdangerw

Assignees
No one assigned
Labels
None yet
Projects
KerasHub
Status: Done
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@james77777778 @mattdangerw @divyashreepathihalli @kokoro-team