Integrate ResNet model pool and enhance image classification fine-tuning

[tanganke]

This PR integrates the ResNet model pool, enhances image classification fine-tuning with logging and model saving, and updates related configuration and documentation.\n\nKey changes:\n- Integrate ResNet model pool\n- Enhance image classification fine-tuning with logging and model saving\n- Add and update configuration files for ResNet and image classification\n- Improve documentation and guides for fine-tuning\n- Refactor and improve code quality in related modules\n\nPlease review and provide feedback.

[Copilot]

Pull Request Overview

This PR integrates ResNet model pool support and enhances image classification fine-tuning capabilities within FusionBench, providing comprehensive infrastructure for training and testing ResNet models on standard datasets.

• Adds ResNet model pool support for both torchvision and transformers backends
• Introduces enhanced image classification fine-tuning with PyTorch Lightning integration
• Creates comprehensive configuration files and documentation for ResNet fine-tuning workflows

Reviewed Changes

Copilot reviewed 26 out of 26 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
mkdocs.yml	Adds new ResNet guides section to documentation navigation
fusion_bench/utils/rich_utils.py	Imports rich traceback functionality for better error reporting
fusion_bench/utils/path.py	Adds symlink creation utility with rank-zero decorator
fusion_bench/utils/lazy_state_dict.py	Enhances checkpoint loading with improved documentation and comments
fusion_bench/tasks/clip_classification/init.py	Adds convenience functions for dataset metadata access
fusion_bench/programs/fusion_program.py	Creates new standard fusion program without Lightning Fabric
fusion_bench/programs/fabric_fusion_program.py	Updates import to use lightning_utilities
fusion_bench/programs/init.py	Registers the new ModelFusionProgram class
fusion_bench/modelpool/resnet_for_image_classification.py	Implements ResNet model pool for image classification
fusion_bench/modelpool/base_pool.py	Adds dataset availability check properties
fusion_bench/modelpool/init.py	Registers ResNet model pool in lazy imports
fusion_bench/method/classification/image_classification_finetune.py	Implements ResNet fine-tuning and testing methods
fusion_bench/method/classification/init.py	Converts to lazy imports and adds new methods
fusion_bench/method/init.py	Registers new fine-tuning methods
fusion_bench/dataset/clip_dataset.py	Extends processor support to include BaseImageProcessor
fusion_bench/init.py	Exports ResNet model pool in package API
docs/guides/resnet/image_classification_finetune.md	Creates comprehensive ResNet fine-tuning guide
config/modelpool/ResNetForImageClassfication/*	Adds ResNet model pool configurations
config/model_fusion.yaml	Creates main fusion configuration file
config/method/classification/*	Adds fine-tuning method configurations

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[Copilot]

Missing space around the equality operator. This should be rank_zero_only.rank == 0 for proper Python style.

Suggested change

	rank_zero_only.rank==0
	rank_zero_only.rank == 0

[Copilot]

This expression returns an integer when num_labels > 0 but a class (nn.Identity) when num_labels <= 0. The nn.Linear constructor expects an integer for out_features. This should likely be len(classnames) if model.config.num_labels > 0 else 0 or similar logic.

Suggested change

	len(classnames) if model.config.num_labels > 0 else nn.Identity(),
	len(classnames) if model.config.num_labels > 0 else 0,