Tags: DeepTrackAI/deeplay
Tags
Bm/load_from_checkpoint (#116) * implement test behavior with no test inputs * docs * Refactor code in deeplay/activelearning/data.py, deeplay/activelearning/strategies/uncertainty.py, deeplay/applications/application.py, and deeplay/activelearning/strategies/strategy.py * clear config before multi * update to use absolute imports * Refactor import statements in test_selectors.py * Bm/fix-config-nested-new Fixes an issue with multi-blocks not having their configuration correctly cleared if created multiple times * remove test file * Fix issue with clearing configuration before creating multiple blocks * Fix issue with configuring upsample in Conv2dBlock * Re-enable test_strided_multi in test_conv.py * add stubs * Refactor residual function in Conv2dBlock to support flexible layer order * Refactor available_styles method in DeeplayModule to use classmethod * Implement script to create stubs with style typing * Refactor Conv2dBlock and related functions in conv2d.pyi * Remove publish script from package.json * Add .gitignore entry for package.json * add docs * delete data * Update documentation and style guide * Update naming conventions, imports, documentation, and testing guidelines * Refactor Conv2dBlock and related functions in conv2d.pyi * Refactor Conv2dBlock and related functions in conv2d.pyi * Refactor Conv2dBlock and related functions in conv2d.pyi * § * u * baseline shape compute * docsv1 * chore: Update expected_input_shape property in LinearBlock class * Refactor DeeplayModule to use torch.no_grad() for args and kwargs invocation * chore: Refactor BaseBlock class to improve code readability and maintainability * Refactor LinearBlock class to improve default configuration and add merge operation * Refactor Sequence1dBlock class to improve default normalization and add run_with_dummy_data method * Refactor Layer class to warn when forward path is called with gradients enabled * Refactor resnet18.py to improve code readability and maintainability * Refactor Conv2dBlock to fix layer configuration and add missing assertions * Refactor Conv2dBlock to fix layer configuration and add missing assertions * Refactor Conv2dBlock to configure padding based on kernel size * Refactor Conv2dBlock to configure padding based on kernel size * remove exception on RuntimeError * del docs * Refactor DeeplayModule new method to set detach parameter to True by default * Refactor RecurrentModel to use super() for forward pass * Refactor DeeplayModule to set detach parameter to True by default * Refactor DeeplayModule to return computed values in configure method * Refactor Layer to undo configuration of computed values on exception * Refactor RNN module to include batch_first and return_cell_state parameters * Refactor RecurrentModel to not return cell state by default * Correctly handle cell state * chore: Fix torch MPS issue with indexed tensors in DeeplayModule * chore: Fix torch MPS issue with indexed tensors in DeeplayModule * chore: Update Python and OS versions in CI workflow * chore: Update Python and OS versions in CI workflow * undo CI test * Update module.py * Add optional normalization parameter to BaseBlock.normalized() * Add load from pickled checkpoint in metaclass * Refactor Application build method to store hparams and modules * Add reduction method for pickle * feat: Add default shortcut function to BaseBlock * refactor: Improve ResNet style block in resnet18.py * refactor: Add check for DeeplayModule in Application class * remove reduction method * Pickle using dill instead of pickle * chore: Update requirements.txt with dill package * feat: Update requirements.txt with numpy>=1.24.1 * chore: Update numpy version in requirements.txt * add a cleanup method before create * add test case for double create bug * remove unused import * Reduce use of deepcopy (for speed) * Reduce use of deepcopy (for speed) * chore: Remove print statement in layer.py * refactor: Add non-class type arguments to External class initialization * Refactor import statements in deeplay.blocks.ls.py and stylestubgen.py * refactor: Add stateful decorator for methods modifying object state * refactor: Add initialization steps for object attributes in meta.py * refactor new method to not use deepcopy * refactor: Remove unused import and regex module from deeplay/module.py * refactor: Update ExtendedConstructorMeta to use _module_state instead of _is_top_level * refactor: Remove commented out code in deeplay/module.py * refactor: Add typing_extensions.Self import to stylestubgen.py * Refactor test_encdec.py and test_gnn.py to use gnn.create() instead of gnn.build() for consistency * add application tests * refactor: Update log_metrics method to preprocess y_hat and y before passing them to metrics * refactor: Remove unused imports in deeplay/tests/applications/base.py * del notebook * del notebook * refactor: Add _non_classtype_args attribute to External class --------- Co-authored-by: Giovanni Volpe <[email protected]>