All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog.

• Callbacks have been reorganized in folders.
• Update requirements.
• Update gh-pages.yml.
• Update README.md
• Update mixed_precision parameter and usage.

• Progress bar will only be created under the main process.
• Logging now is handled by accelerate library.

• Add torch.ditribution example, with code taken from Romain Strock.
• Add predict method to Trainer. #38
• Add functions to freeze and unfreeze model. #43
• Add function to transform dataset into time series dataset.

• Metrics are now moved to the execution device #41.
• Log level is now used in the Trainer. #40
• LearningRateScheduler now does not crash in first epoch when on_train is False. #36

• Make regularization part of the callbacks system. #37
• Divide utils into three submodules: convenience,preprocessing and data.
• Update requirements to avoid conflicts.
• Update some tests.

• Remove old regularization module and all related code.

• Fix PyPi deployment file.

• Add PyPi deployment to the CI/CD.

• Fix CHANGELOG.md release dates.

• Add possibility to set the log level of the callbacks.
• Add stochastic weight averaging callback.
• Add train_test_val_split.
• Add log_name attribute to tochfitter.callbacks.base.Callback.

• Change with torch.no_grad() for @torch.no_grad() in trainer.
• Format code with Black.
• Reorganize utils module.

• Remove reset_parameters method from callbacks.

• Fix RichProgressBar not logging appropiate values.
• Fix log level not being correctly set.

• Add more hooks to the callback system.
• Rich progress bar as callback.
• accelerate.Accelerator backend.
• trainer.Trainer.fit now returns a dictionary with the train history.

• Update README.
• Update metrics handling.

• Remove callback type.

• Solve doc typos.
• Fix logger and trainer tests.
• Fix incomplete quickstart in docs.
• Fix logging bug in GPUStats callback.

• Add support for mixed precision training.
• Add ElasticNet regularization.
• Add testing methods and their tests: check_monotonically_decreasing and compute_forward_gradient.
• Add cuda seed setting in Manager.
• Add option to only use deterministic algorithm in the Manager class.

• Update logo and README.
• Update tests with new testing methods.
• Make some method on Trainer and Manager private.

• Solve bug in callbacks where the handler was not calling in appropiate order.

• Remove ElasticNet regularization because the implementation was not correct.

• Change params_dict in the Trainer to a specific class that tracks the internal state.
• Change README.
• Update tests.
• Change logic of TQDM to be updated in each batch instead of in each epoch.
• Change optimization loop to be of type condition-loop instead of iteration-loop. This is, the loop is now a while loop.

• Add Manager class to handle multiple experiments.
• Add support for computing metrics in the optimization loop via torchmetrics.
• Add GPUStats, ReduceLROnPlateau and ProgressBarLogger callbacks.
• Add testing utility to check gradients: compute_forward_gradient.
• Add more functions to utils: FastTensorDataLoader, check_model_on_cuda.

• Solve warning where learning rate scheduler was being called before loss.

• Change _compute_penalty in favour of compute_penalty.
• Change _train in favour of train_step.
• Change _validate in favour of validation_step.
• Update tests to be correct.

• Added new reset_parameters method in the trainer.
• Added requirements file for example.
• Added trainer example in .py format.
• Added manager.ipynb example.

• Fix error in setup naming.
• Fix moving the tensors to device. Now, it is done in each batch.

• Change the requirements.txt to remove unnecessary dependencies.

• Added possibility to use L1 and ElasticNet regularization.
• Added new testing module.
• Added tests for the new functionalities.

• Updated README to add brief tutorial on how to create regularization algos.
• Updated tests for trainer.

• Fixed minor typos in README.

• Added badges from shield.io

• Added a CHANGELOG.md

• Fixed error in README example syntax.