v0.6.10

What's Changed

• Updating docstrings so that sphinx will build the documentation. by @drewoldag in #581
• Bump dawidd6/action-download-artifact from 11 to 12 by @dependabot[bot] in #580
• Bump actions/upload-artifact from 5 to 6 by @dependabot[bot] in #578
• Bump actions/cache from 4 to 5 by @dependabot[bot] in #577
• Bump astral-sh/setup-uv from 5 to 7 by @dependabot[bot] in #579
• Adding a first draft of the landing page, by @drewoldag in #584
• LSST dataset improvements by @mtauraso in #543
• Some tidying up of introductory documentation by @gitosaurus in #587
• Getting started documentation page + CIFAR/CNN updates by @drewoldag in #586

New Contributors

• @gitosaurus made their first contribution in #587

Full Changelog: v0.6.9...v0.6.10

v0.6.9

What's Changed

• Use simple model.onnx file name and save a copy of config by @drewoldag in #574

Full Changelog: v0.6.8...v0.6.9

v0.6.8

What's Changed

• Model updates rebased to main by @aritraghsh09 in #535
• Implementation of to_onnx verb. by @drewoldag in #527
• Add semi-hidden feature to allow recording metrics at the end of each training epoch by @drewoldag in #540
• Potential fix for ONNX-ify problem by @drewoldag in #541
• Bump actions/checkout from 5 to 6 by @dependabot[bot] in #542
• Fix GPU-trained models failing to load on CPU-only machines by @Copilot in #486
• Fixes a bug in the way that we load a checkpoint file to resume a training run. by @drewoldag in #544
• Initial implementation of a CSV dataset class by @drewoldag in #477
• Updating to the latest copier version. by @drewoldag in #545
• Load/save to_tensor during inference by @drewoldag in #546
• Initial scaffolding for engine verb by @drewoldag in #550
• seperate onnx output from pytorch.ignite by @maxwest-uw in #558
• Default collate function built in DataProvider by @drewoldag in #557
• Adding InferenceDatasetWriter to engine to persist results. by @drewoldag in #559
• Use the new, non-torch-based, collate function in engine by @drewoldag in #560
• Find and use custom collate functions defined on dataset classes by @drewoldag in #561
• Refactor ONNX convert to work for models with different input types by @drewoldag in #564
• Fixing issues encountered while trying to to-onnx applecider models. by @drewoldag in #565

Full Changelog: v0.6.7...v0.6.8

v0.6.7

What's Changed

• Add PyPi badge by @jeremykubica in #526
• Bump peter-evans/find-comment from 3 to 4 by @dependabot[bot] in #522
• Bump actions/upload-artifact from 4 to 5 by @dependabot[bot] in #523
• Bump peter-evans/create-or-update-comment from 4 to 5 by @dependabot[bot] in #524
• Update pytorch-ignite requirement from <=0.5.2 to <=0.5.3 by @dependabot[bot] in #525
• Update automatic criterion and optimizer machinery for transparency by @drewoldag in #529
• Explicitly use fork in InferenceDatasetWriter by @drewoldag in #533

New Contributors

• @jeremykubica made their first contribution in #526

Full Changelog: v0.6.6...v0.6.7

v0.6.6

What's Changed

• Adding onnxscript dependency. by @drewoldag in #490
• Attach final metrics to model returned from h.train(). by @drewoldag in #485
• Allow users to provide an externally defined collate function to PyTorch's DataLoader by @drewoldag in #492
• Reduce verbosity of model logging by @drewoldag in #494
• Enabling distinct splits to be defined for training by @drewoldag in #487
• Add helpful error for empty model_inputs and rename DATA_SET_REGISTRY to DATASET_REGISTRY by @Copilot in #514
• Move model parameters to model-specific config subtables by @Copilot in #516
• Set model.name to false in default config and add informative error by @Copilot in #513
• Remove automatic ONNX export from training by @Copilot in #520

Full Changelog: v0.6.5...v0.6.6

v0.6.5

What's Changed

• Updating built-in models that are not already using data_sample by @drewoldag in #474
• Fix set_config method to handle dotted quoted strings in TOML table names by @Copilot in #484

Full Changelog: v0.6.4...v0.6.5

v0.6.4

What's Changed

• Pulling colorama in as a dependency to highlight the model and data being used for training and inference. by @drewoldag in #478
• Allow user defined metadata fields in HyraxRandomDataset. by @drewoldag in #482
• Fix NaN handling for tuples with length > 2 by @Copilot in #481
• Remove unused ConfigDict class and references by @Copilot in #475

Full Changelog: v0.6.3...v0.6.4

v0.6.3

Biggest changes

tl;dr;
Users should now use h.set_config('path.to.config.key', <value>) instead of h.config['path']['to']['config']['key'] = <value> when modifying the configuration in a notebook.

the deets
We've fixed a bug in the configuration system that prevented default configuration files from being loaded properly when updating the configuration in a notebook. We've introduced the new set_config method to increase transparency of the action, instead of trying to expand the automagical functionality that existed before.

What's Changed

• Lowering many logger.INFOs to logger.DEBUGs. by @drewoldag in #445
• Update model_weights_file config when auto-detected and persist in runtime config by @Copilot in #442
• Fix KeyError when data_location is not defined for iterable datasets and add registry validation by @Copilot in #444
• Updating example notebooks for model_inputs. by @drewoldag in #437
• Fix config file validation to fail for non-existent config files by @Copilot in #452
• Fix vector db benchmarks data access after HyraxQL release by @Copilot in #455
• Fix Parser help when config file is specified early on cli. Present error on nonexistant config. by @mtauraso in #453
• Updated model_inputs 1 and 2 based on PR feedback. by @drewoldag in #461
• Bump asv from 0.6.4 to 0.6.5 by @dependabot[bot] in #464
• Bump actions/setup-python from 5 to 6 by @dependabot[bot] in #465
• Use of fetch_dataset_class instead of accessing DATA_SET_REGISTRY by @drewoldag in #462
• Working to make ConfigManager easier to use by @drewoldag in #468
• Fix vector DB benchmarks to use model_inputs configuration by @Copilot in #470

Full Changelog: v0.6.2...v0.6.3

v0.6.2

What's Changed

• Add comprehensive GitHub Copilot instructions for Hyrax development by @Copilot in #426
• Fix KeyError when primary_id_field is not included in fields list by @Copilot in #436
• Fix DataProvider.resolve_data performance regression by @drewoldag in #440

Full Changelog: v0.6.1...v0.6.2

v0.6.1

What's Changed

• Optimize train.py import performance by moving heavy dependencies to runtime by @Copilot in #430
• Added default value for primary_id_field in .toml file by @drewoldag in #432

Full Changelog: v0.6.0...v0.6.1