https://github.com/alan-turing-institute/ocf-iam4vp - 0.4.3 (2024-12-10)

• Allow loss-type to be specified (allowing 0.4.0, 0.4.1 or 0.4.2 to be reproduced).

Full Changelog: https://github.com/alan-turing-institute/ocf-iam4vp/compare/0.4.2...0.4.3

- Python
Published by jemrobinson over 1 year ago

https://github.com/alan-turing-institute/ocf-iam4vp - 0.4.2 (2024-12-03)

• Lower early stopping threshold
• Set alpha_gdl to 1 to give the same scale as MAE
• Step the optimiser once per set of predictions

- Python
Published by jemrobinson over 1 year ago

https://github.com/alan-turing-institute/ocf-iam4vp - 0.4.1 (2024-12-02)

• Add gradient difference loss
• Set default GDL weighting to 10
• Fix checkpoint loading bugs

- Python
Published by jemrobinson over 1 year ago

https://github.com/alan-turing-institute/ocf-iam4vp - 0.4.0 (2024-11-29)

• Better logging of meta information during training
• Use size of input data to construct model
• Do not encode image size into model parameters (for better flexibility)

- Python
Published by jemrobinson over 1 year ago

https://github.com/alan-turing-institute/ocf-iam4vp - 0.3.5 (2024-11-27)

• Checkpoint at the end of each epoch as well as each batch
• Add early stopping of epoch if improvements are not seen

- Python
Published by jemrobinson over 1 year ago

https://github.com/alan-turing-institute/ocf-iam4vp - 0.3.4 (2024-11-16)

• Better logging of model parameters
• Checkpointing after validation
• Checkpoint after fixed number of batches (instead of truncating epochs)
• Updated hyperparameter defaults

- Python
Published by jemrobinson over 1 year ago

https://github.com/alan-turing-institute/ocf-iam4vp - 0.3.3 (2024-11-16)

• Speed up training following profiling

- Python
Published by jemrobinson over 1 year ago

https://github.com/alan-turing-institute/ocf-iam4vp - 0.3.2 (2024-11-14)

• Use Lightning for training/validation

- Python
Published by jemrobinson over 1 year ago

https://github.com/alan-turing-institute/ocf-iam4vp - 0.3.1 (2024-11-13)

• Revert changes to run all forecasts in forward() as this requires too much memory
• Construct time-embedding on-the-fly to simplify interface
• General memory saving through tensor deletion

- Python
Published by jemrobinson over 1 year ago

https://github.com/alan-turing-institute/ocf-iam4vp - 0.3.0 (2024-11-12)

• Perform all forecasting steps before updating loss
• Use same encoder for X and y

- Python
Published by jemrobinson over 1 year ago

https://github.com/alan-turing-institute/ocf-iam4vp - 0.2.0 (2024-11-12)

• Skip validation batches with no timesteps
• Log parameters values before start of a run
• Add a sigmoid layer to restrict model outputs to (0, 1)

- Python
Published by jemrobinson over 1 year ago

https://github.com/alan-turing-institute/ocf-iam4vp - 0.1.3 (2024-11-11)

• Fix checkpoint naming
• Mask missing values in target
• Clip validation outputs to (0, 1)
• Switch to MAE loss

- Python
Published by jemrobinson over 1 year ago

https://github.com/alan-turing-institute/ocf-iam4vp - 0.1.2 (2024-11-07)

• Fix epoch numbering
• Fix best model save logic
• Centralise multistage prediction
• Add validation plots vs. time as well as channel

- Python
Published by jemrobinson over 1 year ago

https://github.com/alan-turing-institute/ocf-iam4vp - 0.1.1 (2024-11-06)

• Add cloudcasting model validation
• Better validation outputs

- Python
Published by jemrobinson over 1 year ago

https://github.com/alan-turing-institute/ocf-iam4vp - 0.1.0 (2024-11-05)

Initial release based on https://github.com/seominseok0429/Implicit-Stacked-Autoregressive-Model-for-Video-Prediction.

• hyper parameters:
  • number of epochs
  • number of spatial convolutions
  • number of temporal convolutions
  • number of hidden spatial layers
  • number of hidden temporal layers
  • number of forecast steps (for training)
  • number of history steps

- Python
Published by jemrobinson over 1 year ago