2 Releases published by 1 person

• training-0.6.2 training: v0.6.2

  published Aug 20, 2025

• models-0.9.2 models: v0.9.2

  published Aug 20, 2025

8 Pull requests merged by 5 people

• fix: GraphInterpolator Lightning Module - add rollout member for compa…

  #496 merged Aug 21, 2025

• chore: Release main

  #476 merged Aug 20, 2025

• fix(training): sharding nan_loss_mask scaler

  #491 merged Aug 20, 2025

• fix(training): callbacks when using rollout

  #490 merged Aug 20, 2025

• perf: Avoid heavy imports through MlflowSchema

  #482 merged Aug 18, 2025

• docs: explain inference overrides

  #487 merged Aug 18, 2025

• docs: update diffusion documentation

  #486 merged Aug 18, 2025

• feat: diffusion training

  #401 merged Aug 15, 2025

6 Pull requests opened by 6 people

• refactor: change preprocessor config from dictionary to list

  #483 opened Aug 15, 2025

• ci: Add checkpoint migrations workflow

  #492 opened Aug 20, 2025

• feat: Extend restart checkpoint to consider graphtransformer

  #497 opened Aug 21, 2025

• chore: Release main

  #499 opened Aug 21, 2025

• feat: Checkpoint pipeline infrastructure (Phase 1)

  #501 opened Aug 21, 2025

• fix: convert pre_processors from attribute to variable in models/encoder_processor_decoder

  #503 opened Aug 21, 2025

2 Issues closed by 2 people

• Issue with Optuna hyperparameter optimization – Had to use gpus=1 and ntasks=1 on SLURM

  #467 closed Aug 18, 2025

• Checklist Bot for PR

  #313 closed Aug 15, 2025

10 Issues opened by 6 people

• [DRAFT ] Interpolator: How to handle accumulation

  #504 opened Aug 22, 2025

• "pre_processors" Attribute bug in AnemoiModelEncProcDec

  #502 opened Aug 21, 2025

• Ensemble GNN config files fail at the moment

  #500 opened Aug 21, 2025

• Fixing seed still leads to differences in validation losses

  #498 opened Aug 21, 2025

• Checkpoint System Integration and Migration (Phase 3)

  #495 opened Aug 20, 2025

• Checkpoint Loading Orchestration (Phase 2)

  #494 opened Aug 20, 2025

• Checkpoint Pipeline Infrastructure (Phase 1)

  #493 opened Aug 20, 2025

• Move config for attention_implementation [not urgent]

  #488 opened Aug 18, 2025

• Revisit ci for integration tests in training

  #485 opened Aug 15, 2025

• More comprehensive integration tests in training

  #484 opened Aug 15, 2025

22 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• feat: Logging to AzureML with ML Flow

  #469 commented on Aug 21, 2025 • 9 new comments

• ci: memory and performance tests in CI

  #403 commented on Aug 21, 2025 • 8 new comments

• fix: one-variable model

  #478 commented on Aug 20, 2025 • 7 new comments

• feat: checkpoint migrations

  #386 commented on Aug 19, 2025 • 7 new comments

• feat(graphs): support area positional argument in GraphInspector

  #415 commented on Aug 21, 2025 • 7 new comments

• feat(models): add configurable residual connections in enc-proc-dec

  #451 commented on Aug 21, 2025 • 5 new comments

• feat(graphs): support field interpolation in recipe.yaml

  #444 commented on Aug 21, 2025 • 5 new comments

• feat: update config and doc to allow using dataset names along with path

  #238 commented on Aug 20, 2025 • 4 new comments

• fix(training): LAM sharding

  #481 commented on Aug 15, 2025 • 0 new comments

• feat: flash attention v3

  #479 commented on Aug 21, 2025 • 0 new comments

• Refactor restart of training

  #464 commented on Aug 20, 2025 • 0 new comments

• Make batch normalization explicit.

  #448 commented on Aug 19, 2025 • 0 new comments

• [WIP] 410 refactor of model initialisation ie weight loading model freezing transfer learning

  #442 commented on Aug 20, 2025 • 0 new comments

• feat(models): dynamic mappers

  #439 commented on Aug 18, 2025 • 0 new comments

• make activation checkpointing configurable

  #211 commented on Aug 21, 2025 • 0 new comments

• Checkpoint Acquisition Layer - Multi-source checkpoint loading (S3, HTTP, local, MLFlow)

  #458 commented on Aug 20, 2025 • 0 new comments

• Model Transformation Layer - Post-loading modifications (freezing, transfer learning, adapters)

  #410 commented on Aug 20, 2025 • 0 new comments

• Anemoi [Fine-tuning, Transfer Learning, Model Freezing] Roadmap

  #248 commented on Aug 20, 2025 • 0 new comments

• Difficulty creating a minimal working configuration for anemoi-training

  #420 commented on Aug 19, 2025 • 0 new comments

• Pre/postprocessors should be list not dictionary in config file

  #480 commented on Aug 18, 2025 • 0 new comments

• Keep same random seed when restarting a run

  #2 commented on Aug 16, 2025 • 0 new comments

• Residual Scaling (normalization strategy)

  #319 commented on Aug 15, 2025 • 0 new comments