This repository provides a flexible framework for training and testing classifiers using the Timm library from Hugging Face. It's designed to streamline the process of building and fine-tuning classification models for various tasks. The code is highly configurable through a dedicated configuration file, allowing you to experiment with different hyperparameters and architectures effortlessly.
-
Flexibility: Easily adapt the code to your specific classification task by configuring hyperparameters, model architectures, and training settings through a single configuration file.
-
Focal Loss: This repository includes an implementation of the Focal Loss, a specialized loss function that enhances the training of models for imbalanced classification problems.
-
TensorBoard Integration: Visualize training progress, loss curves, and evaluation metrics using TensorFlow's TensorBoard for better insights into model performance.
Follow these steps to get started with training and testing classifiers:
-
Clone the repository to your local machine:
git clone https://github.com/yacinebouaouni/Timm-Classifier-Focal-Loss.git
The config.yaml file serves as the central configuration hub for your experiments. You can adjust various settings, including:
- Dataset paths and preprocessing.
- Model architecture and hyperparameters.
- Training settings (batch size, learning rate, etc.).
- Focal Loss parameters (gamma and alpha).
- Logging and saving options.
- Evaluation metrics and thresholds.
- Customize the configuration file according to your specific task and experimentation goals.
Contributions to this project are welcome! If you have improvements or new features to suggest, please open an issue or submit a pull request.