v0.6.0

This release introduces new features including GIF animation support, wandb run resumption, and improved ensemble loading, alongside a large number of bug fixes and maintenance updates.

Added

• Add support for GIF animation generation for model predictions #218 @kartikangiras

• Add AGENTS.md file to the repo to give agents more information about the codebase and the contribution culture.#416 @sadamov

• Enable pin_memory in DataLoaders when GPU is available for faster async CPU-to-GPU data transfers #236 @abhaygoudannavar

• Expose --wandb_id CLI argument to allow resuming an existing W&B run by
  ID. When provided, resume="allow" is set automatically so the same job
  script works for both the initial submission and all resubmissions, making
  it suitable for HPC systems with limited job runtimes or that may crash.
  #197 @Mani212005

Changed

• Change the default ensemble-loading behavior in WeatherDataset / WeatherDataModule to use all ensemble members as independent samples for ensemble datastores (with matching ensemble-member selection for forcing when available); single-member behavior now requires explicitly opting in via --load_single_member #332 @kshirajahere

• Refactor graph loading: move zero-indexing out of the model and update plotting to prepare using the research-branch graph I/O #184 @zweihuehner

• Replace print()-based rank_zero_print with loguru logger.info() for structured log-level control #33

Fixed

• Fix validation crash in plot_error_map and resolve DDP NCCL initialization error on single-device setups #193 @AdityaKumarSethia

• Fix --metrics_watch handling to avoid AttributeError when unset and improve warning behavior during evaluation #420 @archit7-beep

• Standardize all script references to use create_graph instead of the legacy create_mesh name in README and pyproject.toml, and fix minor README typos #426 @GiGiKoneti

• Initialize da_forcing_mean and da_forcing_std to None when forcing data is absent, fixing AttributeError in WeatherDataset with standardize=True #369 @Sir-Sloth-The-Lazy

• Ensure proper sorting of analysis_time in NpyFilesDatastoreMEPS._get_analysis_times independent of the order in which files are processed with glob #386 @Gopisokk

• Switch to lat/lon-based plotting with pcolormesh and cartopy for accurate spatial visualisation regardless of underlying projection. #168 @sadamov

• Replace shell=True subprocess call in compute_standardization_stats.py with a safe argument list and Python-side hostname parsing to prevent command injection via SLURM_JOB_NODELIST #264 @ashum9

• Avoid NaN when standardizing fields with zero std #189 @varunsiravuri

• Replaces multiple assert statements used for runtime input validation with explicit ValueError #279 @Sir-Sloth-The-Lazy

• Fix README image paths to use absolute GitHub URLs so images display correctly on PyPI #188 @bk-simon

• Fix typo in ar_model.py that causes AttributeError during evaluation #204 @Ritinikhil

• Changed the hardcoded True to a conditional check "persistent_workers=self.num_workers > 0" #235 @santhil-cyber

• Avoid eager download of the MEPS example dataset during pytest collection by lazily initializing it in tests/conftest.py, allowing tests to run without triggering a dataset download at import time. #391 @Saptami191

• fractional_plot_bundle now correctly multiplies by fraction instead of dividing #222 @santhil-cyber

• Fix all_gather_cat producing wrong shapes on single-device runs by only flattening when all_gather actually introduces a new leading dimension #424 @RajdeepKushwaha5

• Infer spatial coordinate names for MDPDatastore (rather than assuming names x and y), allows for e.g. lat/lon regular grids #169 @leifdenby

• Fix Slack domain link #288 @sadamov

Maintenance

• Update PR template to clarify milestone/roadmap requirement and maintenance changes #186 @joeloskarsson

• Update CI/CD to use python 3.13 for testing and full range of current python versions for linting (3.10 - 3.14) #173 @observingClouds

• Move development dependencies to dependency-group #174 @observingClouds

• Update CI/CD to use only uv for full test suite and drop pdm #178 @observingClouds

• Fix caching of MEPS example data in CI/CD #181 @observingClouds

• Migrated build backend from PDM to Hatchling with hatch-vcs and added uv build in deploy CI

• Warn when running with --eval without --load to avoid accidentally evaluating randomly initialized weights #190 @varunsiravuri

v0.5.0

This release contains maintenance and fixes, preventing some unexpected crashes and improving CICD and testing.

Added

• Expose run name as optional command line argument --logger_run_name to allow user-defined names
  #156 @observingClouds

Fixed

• Change default logging argument to prevent crash when running eval
  #145 @joeloskarsson

• Fix wrong grid dimensionality when running with --output_std, resulting in crash
  #147 @joeloskarsson

• Fix the order in create_graph.py which caused wrong G2M and M2G
  #150 @YUTAIPAN

• Adding a more robust LaTeX availability check function #162 @lorenzo30salgado

Maintenance

• Add link to full MEPS data #102 @joeloskarsson

• Introducing mypy for static type checking and fixing type hints accordingly #113 @observingClouds

• Change all argparse instances to use ArgumentDefaultsHelpFormatter for easier maintaining defaults.
  #145 @joeloskarsson

• Allow triggering CI manually to e.g. test for recent software incompatibilities. \152 @observingClouds

• Fix torch version detection during CI when testing on CPU with pdm #154 @leifdenby

• Update link to MEPS example data #155 @joeloskarsson

• Change deprecated pynvml dependency to nvidia-ml-py #176 @observingClouds

v0.4.0

This release introduces a number of improvements to logging, multi-node training and variable rescaling, without making any major changes to the neural-lam structure.

Added

• Add support for MLFlow logging and metrics tracking. #77
  @khintz

• Add support for multi-node training.
  #103 @SimonKamuk @sadamov

• Add option to clamp output prediction using limits specified in config file #92 @SimonKamuk

• Add publication of releases to pypi.org. #71 @leifdenby, @observingClouds

Fixed

• Only print on rank 0 to avoid duplicates of all print statements.
  #103 @SimonKamuk @sadamov

• Fix MLFlow exception import introduced in #77.
  #111
  @observingClouds

• Fix duplicate tensor copy to CPU #106 @observingClouds

• Fix bug where the inverse_softplus used in clamping caused nans in the gradients #123 @SimonKamuk

• Add standardization to state diff stats from mdp datastore #122 @SimonKamuk

• Set ci/cd badges to refer to the new test matrix #130 @SimonKamuk

• use correct split of data with the --eval val or --eval test cli arguments #139 @SimonKamuk

• Fix step length calculation when dt >= 24h #141 @deinal

Maintenance

• update ci/cd testing to use cuda 12.8 #140 @SimonKamuk

• update ci/cd testing to use pre-commit v3.0.1 #140 @SimonKamuk

• update AWS GPU ci/cd to use ami with larger (200GB) root volume and ensure
  nvme drive is used for pip venvn
  #126, @leifdenby

• update ci/cd testing setup to install torch version compatible with neural-lam
  dependencies #115, @leifdenby

• switch to new npyfiles MEPS and mdp DANRA test datasets which are coincident
  in time and space (on cropped ~100x100 grid-point domain)
  #110, @leifdenby

• use dynamic versioning based on git tags and commit hashes
  #118, @observingClouds

• add detect_anomaly=True to pl.Trainer in test_training.py #124, @SimonKamuk

v0.3.0

This release introduces Datastores to represent input data from different sources (including zarr and numpy) while keeping graph generation within neural-lam.

Added

• Introduce Datastores to represent input data from different sources, including zarr and numpy. #66 @leifdenby @sadamov

Fixed

• Fix wandb environment variable disabling wandb during tests. Now correctly uses WANDB_MODE=disabled. #94 @joeloskarsson

• Fix bugs introduced with datastores functionality relating visualation plots #91 @leifdenby

v0.2.0

Highlights

This release focuses on setting up the neural-lam repository and codebase to enable collaboration.

Detailed Changes

Added

• Added tests for loading dataset, creating graph, and training model based on reduced MEPS dataset stored on AWS S3, along with automatic running of tests on push/PR to GitHub, including push to main branch. Added caching of test data to speed up running tests.
  #38 #55
  @SimonKamuk

• Replaced constants.py with data_config.yaml for data configuration management
  #31
  @sadamov

• new metrics (nll and crps_gauss) and metrics submodule, stddiv output option
  c14b6b4
  @joeloskarsson

• ability to "watch" metrics and log
  c14b6b4
  @joeloskarsson

• pre-commit setup for linting and formatting
  #6, #8
  @sadamov, @joeloskarsson

• added github pull-request template to ease contribution and review process
  #53, @leifdenby

• ci/cd setup for running both CPU and GPU-based testing both with pdm and pip based installs #37, @khintz, @leifdenby

Changed

• Clarify routine around requesting reviewer and assignee in PR template
  #74
  @joeloskarsson

• Argument Parser updated to use action="store_true" instead of 0/1 for boolean arguments.
  (#72)
  @ErikLarssonDev

• Optional multi-core/GPU support for statistics calculation in create_parameter_weights.py
  #22
  @sadamov

• Robust restoration of optimizer and scheduler using ckpt_path
  #17
  @sadamov

• Updated scripts and modules to use data_config.yaml instead of constants.py
  #31
  @sadamov

• Added new flags in train_model.py for configuration previously in constants.py
  #31
  @sadamov

• moved batch-static features ("water cover") into forcing component return by WeatherDataset
  #13
  @joeloskarsson

• change validation metric from mae to rmse
  c14b6b4
  @joeloskarsson

• change RMSE definition to compute sqrt after all averaging
  #10
  @joeloskarsson

Removed

• WeatherDataset(torch.Dataset) no longer returns "batch-static" component of
  training item (only prev_state, target_state and forcing), the batch static features are
  instead included in forcing
  #13
  @joeloskarsson

Maintenance

• simplify pre-commit setup by 1) reducing linting to only cover static
  analysis excluding imports from external dependencies (this will be handled
  in build/test cicd action introduced later), 2) pinning versions of linting
  tools in pre-commit config (and remove from requirements.txt) and 3) using
  github action to run pre-commit.
  #29
  @leifdenby

• change copyright formulation in license to encompass all contributors
  #47
  @joeloskarsson

• Fix incorrect ordering of x- and y-dimensions in comments describing tensor
  shapes for MEPS data
  #52
  @joeloskarsson

• Cap numpy version to < 2.0.0 (this cap was removed in #37, see below)
  #68
  @joeloskarsson

• Remove numpy < 2.0.0 version cap
  #37
  @leifdenby

• turn neural-lam into a python package by moving all *.py-files into the
  neural_lam/ source directory and updating imports accordingly. This means
  all cli functions are now invoke through the package name, e.g. python -m neural_lam.train_model instead of python train_model.py (and can be done
  anywhere once the package has been installed).
  #32, @leifdenby

• move from requirements.txt to pyproject.toml for defining package dependencies.
  #37, @leifdenby

• Add slack and new publication info to readme
  #78
  @joeloskarsson

Compatibility

This version has been tested with Python 3.9-3.12.

Upgrade Steps

To upgrade to neural-lam v0.2.0:

1. Update your local version:

   • git pull the latest changes from the repository and install locally pip install -e .
   • Note: This release is not yet available on PyPI. You will need to install it from the GitHub repository.

2. Adapt any code relying on constants previously defined in neural_lam/constants.py to instead read them from the new YAML config file.

Links

• Full Changelog
• v0.2.0 GitHub Milestone

v0.1.0

Initial version of Neural-LAM, reproduces the workshop paper.