Standardize Backbone API for SSL compatibility and structural consistency#30
Open
Standardize Backbone API for SSL compatibility and structural consistency#30
Conversation
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
Contributor
There was a problem hiding this comment.
Pull request overview
This PR standardizes the BACKBONES interface across torch_ecg to better support SSL-style feature extraction by adding a consistent forward_features entry point and a corresponding feature-shape inference method.
Changes:
- Added
SizeMixin.compute_features_output_shape(...)as a default feature-shape inference hook (delegating tocompute_output_shape). - Implemented
forward_features(...)(+ per-backbonecompute_features_output_shape(...)) across the registered CNN backbones (ResNet, RegNet, VGG16, Xception, DenseNet, MobileNetV1/2/3, MultiScopicCNN). - Added a new unit test to validate the standardized backbone API.
Reviewed changes
Copilot reviewed 12 out of 12 changed files in this pull request and generated 3 comments.
Show a summary per file
| File | Description |
|---|---|
| torch_ecg/utils/utils_nn.py | Adds default compute_features_output_shape on SizeMixin. |
| torch_ecg/models/cnn/xception.py | Adds forward_features and feature-shape inference method. |
| torch_ecg/models/cnn/vgg.py | Imports Tensor; adds forward_features and feature-shape inference method. |
| torch_ecg/models/cnn/resnet.py | Adds forward_features and feature-shape inference method. |
| torch_ecg/models/cnn/regnet.py | Imports Tensor; adds forward_features and feature-shape inference method. |
| torch_ecg/models/cnn/multi_scopic.py | Adds forward_features and feature-shape inference method. |
| torch_ecg/models/cnn/mobilenet.py | Adds forward_features and feature-shape inference method to V1/V2/V3. |
| torch_ecg/models/cnn/ho_resnet.py | Adds typed method stubs to align with standardized interface (still unimplemented). |
| torch_ecg/models/cnn/efficientnet.py | Adds typed method stubs to align with standardized interface (still unimplemented). |
| torch_ecg/models/cnn/densenet.py | Adds forward_features and feature-shape inference method. |
| torch_ecg/models/cnn/darknet.py | Adjusts base-class order; adds typed method stubs to align with standardized interface (still unimplemented). |
| test/test_models/test_backbone_api.py | Introduces a new parametrized test for the standardized backbone API. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
You can also share your feedback on Copilot code review. Take the survey.
torch_ecg/utils/utils_nn.py
Outdated
Comment on lines
+1076
to
+1099
| def compute_features_output_shape( | ||
| self, seq_len: Optional[int] = None, batch_size: Optional[int] = None | ||
| ) -> Sequence[Union[int, None]]: | ||
| """Compute the output shape of the features. | ||
|
|
||
| By default, this is the same as the output shape of the model. | ||
| For backbones with pooling and classification heads, this should be | ||
| overridden to return the shape of the features before global pooling. | ||
|
|
||
| Parameters | ||
| ---------- | ||
| seq_len : int, optional | ||
| Length of the input signal tensor. | ||
| batch_size : int, optional | ||
| Batch size of the input signal tensor. | ||
|
|
||
| Returns | ||
| ------- | ||
| output_shape : sequence | ||
| Output shape of the features. | ||
|
|
||
| """ | ||
| return self.compute_output_shape(seq_len, batch_size) | ||
|
|
Comment on lines
+26
to
+37
| # Get default config if available in ECG_CRNN_CONFIG | ||
| config = None | ||
| for k, v in ECG_CRNN_CONFIG.cnn.items(): | ||
| if k.lower() == backbone_name.lower(): | ||
| config = deepcopy(v) | ||
| break | ||
|
|
||
| if config is None: | ||
| # Some backbones might not be in ECG_CRNN_CONFIG, skip for now | ||
| # or provide a minimal dummy config if known | ||
| pytest.skip(f"No default config found for backbone: {backbone_name}") | ||
|
|
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #30 +/- ##
==========================================
+ Coverage 93.43% 93.76% +0.32%
==========================================
Files 139 136 -3
Lines 18413 18408 -5
==========================================
+ Hits 17205 17260 +55
+ Misses 1208 1148 -60 ☔ View full report in Codecov by Sentry. |
This file contains hidden or 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
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.
This PR completes the standardization of the Backbone API across
torch_ecg. Every model registered in theBACKBONESregistry now implements a consistent interface for feature extraction, which is a critical prerequisite for Self-Supervised Learning (SSL) architectures likeMAEandSimCLR.Key Changes
forward_features(self, x: Tensor) -> Tensor:Standardized method to extract high-dimensional feature maps before global pooling/classification heads.compute_features_output_shape(self, seq_len, batch_size):Allows downstream modules to statically infer feature dimensions.SizeMixinintorch_ecg.utils.utils_nnto support the new feature shape inference logic.