Add the ResNet_vd backbone #1766

gowthamkpr · 2024-08-09T15:53:22Z

This adds the ResNet_vd backbone to prepare for differential binarization model.

Architecture is adopted from PaddleOCR (https://github.com/PaddlePaddle/PaddleOCR/blob/main/ppocr/modeling/backbones/det_resnet_vd.py).

This is the first step of #1739:

Aliases have been removed from ResNet_vd
Model has been moved out of the subdirectory
Tests (from keras_cv) have been added

divyashreepathihalli

Thank you for the PR Gowtham! Looks good overall.
With KerasHub if the model has multiple versions with small changes in architecture, they will be added as one model and the configuration will be handled in the args.
We have this PR for resnet - #1765
The architecture seems similar(please let me know if this is not the case and this needs to be a separate model), can we combine ResnetVD into that one?

divyashreepathihalli · 2024-08-09T17:24:11Z

keras_nlp/src/models/resnet_vd/resnet_vd_backbone_tests.py

+        model(self.input_batch)
+
+    @pytest.mark.large  # Saving is slow, so mark these large.
+    def test_saved_model(self):


Use KerasNLPs standard test routine like this for the saved model test - https://github.com/keras-team/keras-nlp/blob/keras-hub/keras_nlp/src/models/vgg/vgg_backbone_test.py#L44

divyashreepathihalli · 2024-08-09T17:25:08Z

keras_nlp/src/models/resnet_vd/resnet_vd_backbone_tests.py

+
+
+class ResNetVdBackboneTest(TestCase):
+    def setUp(self):


Also, please add a backbone basics tests like - https://github.com/keras-team/keras-nlp/blob/keras-hub/keras_nlp/src/models/vgg/vgg_backbone_test.py#L34

divyashreepathihalli · 2024-08-09T18:06:11Z

Also to format the code you can use shell/api_gen.sh, shell/format.sh and shell/lint.sh at kerasnlp root dir. You can also run the tests locally with pytest keras_nlp/src/models/resnet_vd

mattdangerw · 2024-08-09T18:55:45Z

keras_nlp/src/models/resnet_vd/resnet_vd_backbone.py

+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""ResNet_vd backbone model.


we don't do any module level docstrings like this

mattdangerw · 2024-08-09T18:58:24Z

keras_nlp/src/models/resnet_vd/resnet_vd_backbone.py

+
+@keras_nlp_export("keras_nlp.models.ResNetVdBackbone")
+class ResNetVdBackbone(Backbone):
+    """Instantiates the ResNet_vd architecture.


Try to write this more in the style of other backbone we have. No reference block. See BertBackbone as a good template.

mattdangerw · 2024-08-09T18:58:47Z

keras_nlp/src/models/resnet_vd/resnet_vd_backbone.py

+            stack. Use "basic_block" for ResNet18 and ResNet34.
+
+    Examples:
+    input_data = tf.ones(shape=(8, 224, 224, 3))


this should be inside the code block I think.

mattdangerw · 2024-08-09T19:00:04Z

keras_nlp/src/models/resnet_vd/resnet_vd_backbone.py

+    @classproperty
+    def presets(cls):
+        """Dictionary of preset names and configurations."""
+        return copy.deepcopy(backbone_presets)
+
+    @classproperty
+    def presets_with_weights(cls):
+        """Dictionary of preset names and configurations that include
+        weights."""
+        return copy.deepcopy(backbone_presets_with_weights)


we shouldn't need to do this anymore

@mattdangerw PTAL #1766 (comment)

mattdangerw · 2024-08-09T19:01:58Z

keras_nlp/src/models/resnet_vd/resnet_vd_backbone_presets.py

+            "official_name": "ResNetVd",
+            "path": "resnet_vd",
+        },
+        # "kaggle_handle": "kaggle://keras/resnetv1/keras/resnet18/2",


I'm confused at this. Is resnet vd just the same structure as resnet?

gowthamkpr · 2024-08-13T20:21:59Z

I've changed the code to reuse ResNetBackbone (#1765) and to set the variant based on a parameter. For this, I replaced use_pre_activation with version, which can be either "v1", "v2" or "vd". Within the code, I retained use_pre_activation to retain more concise branching. This can be changed, of course.

divyashreepathihalli

Thanks Gowtham! I am a little confused about the added conv2d layer. A numerics verifying colab will be helpful.
Lets update the arg as well.

divyashreepathihalli · 2024-08-13T20:30:04Z

keras_nlp/src/models/resnet/resnet_backbone.py

            x = layers.BatchNormalization(
                axis=bn_axis, epsilon=1.001e-5, dtype=dtype, name="conv1_bn"
            )(x)
            x = layers.Activation("relu", dtype=dtype, name="conv1_relu")(x)
+            x = layers.Conv2D(


I am a little confused about the architecture here. The VD version was supposed to have only different pooling layers. Here I am seeing that you are adding an additional conv2d layer.

A colab verifying the numerics for this would be helpful the verify architecture correctness

Modifications compared to plain ResNetV1 are

Average pooling is used in shortcut connections instead of Conv2D strides

The model's first convolutional layer is replaced with three subsequent ones

Stride configuration in residual blocks is slightly different

This paper has some good visualizations of the ResNet_vd architecture.

I can look into creating a notebook to verify numerics. Since we are primarily adding this for the differential binarization model, I would check numerics based on the pretrained differential binarization model from PaddleOCR unless you think we should compare numerics for the classifier network as well.

you can verify the numerics just for the resnetvd backbone with just random weights.

Thanks I will take a look at the paper

I think overall, we want to avoid something like this, where a single flag is used for a proxy for a lot of magic numbers.

We want the magic numbers to live in the config for the model. The 64 that was here before for the single conv layer is probably better written as stackwise_num_filters[0] (but at least we were throwing in stackwise_num_filters[0] != 64 in that case).

divyashreepathihalli · 2024-08-13T21:37:10Z

keras_nlp/src/models/resnet/resnet_backbone.py

@@ -86,7 +86,7 @@ class ResNetBackbone(Backbone):
        stackwise_num_blocks=[2, 2, 2],
        stackwise_num_strides=[1, 2, 2],
        block_type="basic_block",
-        use_pre_activation=True,
+        version="v2",
        pooling="avg",


lets keep the use_pre_activation arg. instead of version lets add arg something like use_vd_pooling or something like that

Sure, I can change this.
Just a note - with two booleans, we convey that you can set use_pre_activation = use_vd_pooling = True. Since both architecture variants introduce subtle modifications, it will not be entirely clear from the interface what the outcome would be. My personal preference would be having it inherently mutually exclusive.

yes, we should have two different args!

divyashreepathihalli · 2024-08-13T21:38:52Z

keras_nlp/src/models/resnet/resnet_backbone.py

@@ -209,7 +258,7 @@ def __init__(
        self.stackwise_num_blocks = stackwise_num_blocks
        self.stackwise_num_strides = stackwise_num_strides
        self.block_type = block_type
-        self.use_pre_activation = use_pre_activation
+        self.version = version


same comment as above for changing arg name for version

divyashreepathihalli · 2024-08-13T21:39:06Z

keras_nlp/src/models/resnet/resnet_backbone.py

@@ -220,7 +269,7 @@ def get_config(self):
            "stackwise_num_blocks": self.stackwise_num_blocks,
            "stackwise_num_strides": self.stackwise_num_strides,
            "block_type": self.block_type,
-            "use_pre_activation": self.use_pre_activation,
+            "version": self.version,


same comment as above for changing arg name for version

james77777778 · 2024-08-14T09:46:46Z

Hi @divyashreepathihalli @gowthamkpr
Can we wait for #1769 ?
The implementation of ResNetBackbone has been changed in that PR and the numerical differences have been verified.

gowthamkpr · 2024-08-19T18:34:45Z

I've updated the PR for the newest version of the code. Numerical agreement has been verified in this notebook and summarized observed numerical deviations in this document. Observed maximum deviations from PaddleClas range between 5e-7 and 3e-6.

For verification, I have converted all ResNet_vd models from PaddleClas to Keras models. They report accuracy on ImageNet-val as shown in this figure from this link

The accuracy readings for the ssld models (shown in red, they use knowledge distillation) are actually really good compared to the values reported in the "ResNet strikes back" paper for timm models. This might be interesting for us.

divyashreepathihalli · 2024-08-19T19:24:46Z

Excellent!! can we add all of these variants for our preset? Hongyu added a built in conversion from timm for most of the resnet models - ptal - #1769

mattdangerw

Left a few comments on how we could clean up the config here.

mattdangerw · 2024-08-19T19:48:58Z

keras_nlp/src/models/resnet/resnet_backbone.py

@@ -133,7 +142,12 @@ def __init__(
                '`block_type` must be either `"basic_block"` or '
                f'`"bottleneck_block"`. Received block_type={block_type}.'
            )
-        version = "v1" if not use_pre_activation else "v2"
+        if use_vd_pooling:


I think this if block is a bit of an anti pattern. Pre-existing though. We don't need layer names to reflect whether something is vd or v2 or v1, we can just do that via preset name. Maintaining something like this will be fragile.

The timm conversion of ResNet might fail after renaming the layers. I can fix it later if needed.

@james77777778 yes please check if we need a small fix after this lands!

mattdangerw · 2024-08-19T19:51:35Z

keras_nlp/src/models/resnet/resnet_backbone.py

            x = layers.BatchNormalization(
                axis=bn_axis, epsilon=1.001e-5, dtype=dtype, name="conv1_bn"
            )(x)
            x = layers.Activation("relu", dtype=dtype, name="conv1_relu")(x)
+            x = layers.Conv2D(


I think overall, we want to avoid something like this, where a single flag is used for a proxy for a lot of magic numbers.

We want the magic numbers to live in the config for the model. The 64 that was here before for the single conv layer is probably better written as stackwise_num_filters[0] (but at least we were throwing in stackwise_num_filters[0] != 64 in that case).

mattdangerw · 2024-08-19T20:04:55Z

keras_nlp/src/models/resnet/resnet_backbone.py

@@ -106,6 +114,7 @@ def __init__(
        stackwise_num_strides,
        block_type,
        use_pre_activation=False,
+        use_vd_pooling=False,


I don't think we want this one flag that proxies a bunch of options. This is going to be hard for maintenance, and lead to confusing configs.

I think we might want to add a new block type, "vd_block", and two new args here. input_conv_filters=(64,) and input_conv_kernel_sizes=(7,). Or something like that.

Then the config for resnet vd would look like...

{ ..., input_conv_filters=(32, 32, 64), input_conv_kernel_sizes=(3, 3, 3), block_type="vd_block", }

Then you could implement the input conv layers a a loop, and implement the new vd block, by copying the basic_block with some minor changes.

WDYT? @james77777778 any tweaks as well?

Not sure I'm getting the exact args right, but at a high level, avoid hard coding magic numbers that define the structure of this model behind feature flags. Instead they should be directly in the config.

This sounds good to me. It's clearer than using a bool/str arg.

mattdangerw · 2024-08-19T20:05:57Z

keras_nlp/src/models/resnet/resnet_backbone.py

@@ -155,21 +169,21 @@ def __init__(
        # The padding between torch and tensorflow/jax differs when `strides>1`.
        # Therefore, we need to manually pad the tensor.
        x = layers.ZeroPadding2D(
-            3,
+            1 if use_vd_pooling else 3,


Under the proposal above, this might be able to be written as input_conv_kernel_sizes[0] // 2 or something like that?

mattdangerw · 2024-08-19T20:06:54Z

keras_nlp/src/models/resnet/resnet_backbone.py

@@ -580,7 +678,12 @@ def apply_stack(
        Output tensor for the stacked blocks.
    """
    if name is None:
-        version = "v1" if not use_pre_activation else "v2"
+        if use_vd_pooling:


Similar to above I think we should remove this. The preset name is how we can track different in vd vs v2 vs v2. This will be hard to maintain.

gowthamkpr · 2024-08-20T15:18:18Z

Excellent!! can we add all of these variants for our preset? Hongyu added a built in conversion from timm for most of the resnet models - ptal - #1769

Yes. Model files are available from the linked notebook. Both keras-nlp and PaddleClas are subject to an Apache 2.0 licence.

mattdangerw

Nice looking good! I think there might be an extraneous additions to the basic block to remove.

mattdangerw · 2024-08-26T20:17:19Z

keras_nlp/src/models/resnet/resnet_backbone.py

@@ -461,7 +527,295 @@ def apply_bottleneck_block(
            use_bias=False,
            dtype=dtype,
            name=f"{name}_0_conv",
+        )(shortcut)


What's up with this 50 or so line addition to the basic block? Shouldn't all the changes be the VD blocks now? Am I missing something?

Just as the normal ResNet, ResNet_vds use basic blocks in their small versions, and the bigger ResNet_vds use bottleneck blocks. Therefore, we need vd variants of both, basic_block and bottleneck_block.

Oh I think this was just github diff being weird, my bad! It looked like you were making a massive addition to the non-vd block, but I see now that's just because the code matches across both. Let's try this out!

mattdangerw

lgtm!

* Add ResNet_vd to ResNet backbone * Addressed requested parameter changes * Fixed tests and updated comments * Added new parameters to docstring

* 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 <[email protected]> * 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 <[email protected]> * 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 <[email protected]> Co-authored-by: Sachin Prasad <[email protected]> Co-authored-by: Matt Watson <[email protected]> Co-authored-by: Siva Sravana Kumar Neeli <[email protected]> Co-authored-by: Usha Rengaraju <[email protected]> Co-authored-by: ushareng <[email protected]> Co-authored-by: pkgoogle <[email protected]> Co-authored-by: gowthamkpr <[email protected]>

gowthamkpr requested a review from divyashreepathihalli August 9, 2024 16:58

divyashreepathihalli requested changes Aug 9, 2024

View reviewed changes

divyashreepathihalli requested a review from mattdangerw August 9, 2024 17:32

mattdangerw reviewed Aug 9, 2024

View reviewed changes

divyashreepathihalli force-pushed the keras-hub branch from d8b2656 to 73b7bad Compare August 12, 2024 17:17

gowthamkpr force-pushed the resnet_vd branch from d176f8b to a078bed Compare August 13, 2024 20:16

divyashreepathihalli requested changes Aug 13, 2024

View reviewed changes

Add ResNet_vd to ResNet backbone

1034522

gowthamkpr force-pushed the resnet_vd branch from a078bed to 1034522 Compare August 19, 2024 18:02

mattdangerw reviewed Aug 19, 2024

View reviewed changes

gowthamkpr added 4 commits August 24, 2024 18:56

Addressed requested parameter changes

0fe8acc

Merged with keras-hub

7ad2838

Fixed tests and updated comments

745766c

Added new parameters to docstring

c28a56d

mattdangerw reviewed Aug 26, 2024

View reviewed changes

mattdangerw approved these changes Aug 28, 2024

View reviewed changes

mattdangerw merged commit be8888d into keras-team:keras-hub Aug 28, 2024
7 checks passed

mattdangerw pushed a commit that referenced this pull request Sep 11, 2024

Add the ResNet_vd backbone (#1766)

6104520

* Add ResNet_vd to ResNet backbone * Addressed requested parameter changes * Fixed tests and updated comments * Added new parameters to docstring

mattdangerw pushed a commit that referenced this pull request Sep 13, 2024

Add the ResNet_vd backbone (#1766)

b10c410

* Add ResNet_vd to ResNet backbone * Addressed requested parameter changes * Fixed tests and updated comments * Added new parameters to docstring

mattdangerw pushed a commit that referenced this pull request Sep 17, 2024

Add the ResNet_vd backbone (#1766)

5dab50a

* Add ResNet_vd to ResNet backbone * Addressed requested parameter changes * Fixed tests and updated comments * Added new parameters to docstring

Add the ResNet_vd backbone #1766

Add the ResNet_vd backbone #1766

Conversation

gowthamkpr commented Aug 9, 2024

divyashreepathihalli left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

divyashreepathihalli commented Aug 9, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

gowthamkpr commented Aug 13, 2024

divyashreepathihalli left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

gowthamkpr Aug 14, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

james77777778 commented Aug 14, 2024

gowthamkpr commented Aug 19, 2024 • edited Loading

divyashreepathihalli commented Aug 19, 2024 • edited Loading

mattdangerw left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mattdangerw Aug 19, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

gowthamkpr commented Aug 20, 2024

mattdangerw left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mattdangerw left a comment

Choose a reason for hiding this comment

divyashreepathihalli commented Aug 9, 2024 •

edited

Loading

gowthamkpr Aug 14, 2024 •

edited

Loading

gowthamkpr commented Aug 19, 2024 •

edited

Loading

divyashreepathihalli commented Aug 19, 2024 •

edited

Loading

mattdangerw Aug 19, 2024 •

edited

Loading