diff --git a/graphs/CHANGELOG.md b/graphs/CHANGELOG.md
index 76137b1ef..4b97b7241 100644
--- a/graphs/CHANGELOG.md
+++ b/graphs/CHANGELOG.md
@@ -11,40 +11,16 @@ Keep it human-readable, your future self will thank you!
 ## 0.4.4 (2025-02-04)
 
 <!-- Release notes generated using configuration in .github/release.yml at main -->
-
+Version fix.
 
 
 **Full Changelog**: https://github.com/ecmwf/anemoi-core/compare/anemoi-graphs-0.4.3...anemoi-graphs-0.4.4
 
-## 0.4.3 (2025-02-04)
+## 0.4.2.post546 (2025-02-04)
 
 <!-- Release notes generated using configuration in .github/release.yml at main -->
 
 ## What's Changed
-### Training
-* fix(training, plots) Exclude nans from error colorbars by @anaprietonem in https://github.com/ecmwf/anemoi-core/pull/59
-* fix(training): bump anemoi-datasets required version to 0.5.13 by @JPXKQX in https://github.com/ecmwf/anemoi-core/pull/74
-* chore(training): Add default config files for 2 and 3 level hierarchical processors by @JPXKQX in https://github.com/ecmwf/anemoi-core/pull/50
-* fix: update graph configs to avoid DeprecationWarning for area weights by @JPXKQX in https://github.com/ecmwf/anemoi-core/pull/53
-* feat(models): normalization layers by @jakob-schloer in https://github.com/ecmwf/anemoi-core/pull/47
-* Fix crash in diagnostic plots (colorbar limits) by @lzampier in https://github.com/ecmwf/anemoi-core/pull/85
-* fix(training): profiler 'Model Summary' works when sharding models over multiple GPUs by @cathalobrien in https://github.com/ecmwf/anemoi-core/pull/90
-* docs(graphs): Refactor anemoi-graphs documentation by @JPXKQX in https://github.com/ecmwf/anemoi-core/pull/49
-* fix: pin dask version to 2024.12.1  by @JPXKQX in https://github.com/ecmwf/anemoi-core/pull/94
-* docs: Improve installation docs by @HCookie in https://github.com/ecmwf/anemoi-core/pull/91
-* docs: cancel RTD builds on PRs without change by @JesperDramsch in https://github.com/ecmwf/anemoi-core/pull/97
-* feat: Model Freezing ❄️  by @icedoom888 in https://github.com/ecmwf/anemoi-core/pull/61
-* feat: make flash attention configurable by @theissenhelen in https://github.com/ecmwf/anemoi-core/pull/60
-* fix: cpu memory savings of sharded dataloader by @japols in https://github.com/ecmwf/anemoi-core/pull/83
-* chore: synced file(s) with ecmwf-actions/reusable-workflows by @DeployDuck in https://github.com/ecmwf/anemoi-core/pull/84
-### Models
-* feature(models): Add model comm group to predict_step  by @cathalobrien in https://github.com/ecmwf/anemoi-core/pull/77
-* Implementation of NormalizedReluBounding for non-zero thresholds by @lzampier in https://github.com/ecmwf/anemoi-core/pull/64
-* fix: normalise in place to reduce memory by @japols in https://github.com/ecmwf/anemoi-core/pull/82
-* feat(models): use num_layers of the processor in hierarchical graphs by @JPXKQX in https://github.com/ecmwf/anemoi-core/pull/78
-* fix: default behaviour for kernel_layers when not set in config. by @jakob-schloer in https://github.com/ecmwf/anemoi-core/pull/93
-* fix:  bug in variables ordering in NormalizedReluBounding by @lzampier in https://github.com/ecmwf/anemoi-core/pull/98
-* feat(models): Copy Imputer by @icedoom888 in https://github.com/ecmwf/anemoi-core/pull/72
 ### Graphs
 * feat(graphs,plots): expand support for multi-dimensional node attributes by @JPXKQX in https://github.com/ecmwf/anemoi-core/pull/48
 * feat(graphs): New Edge Attribute: AttributeFromNode by @icedoom888 in https://github.com/ecmwf/anemoi-core/pull/62