@@ -10,24 +10,17 @@ to [Semantic Versioning]. Full commit history is available in the
1010
1111#### Added
1212
13- - Add support for mudata during ray autotune, {pr}` 3545 ` .
14-
1513#### Fixed
1614
17- - Fix in non multiGPU training to have history in memory, and not on disk by default, {pr}` 3543 ` .
18- - Fix Jax dependency in function log_nb_positive, {pr}` 3558 ` .
19- - Fix {class}` scvi.model.TOTALVI ` convert_legacy_save function with updated model parameters, {pr}` 3561 ` .
20-
2115#### Changed
2216
2317#### Removed
2418
25- - Removed graceful shutdown from Jupyter notebook, {pr}` 3556 ` .
26-
2719### 1.4.0 (2025-09-14)
2820
2921#### Added
3022
23+ - Add support for mudata during ray autotune, {pr}` 3545 ` .
3124- Add a PyTorch implementation of {class}` scvi.external.MRVI ` , {pr}` 3304 ` .
3225- Add checkpointing with {class}` scvi.autotune.AutotuneExperiment ` , {pr}` 3452 ` .
3326- Add Downstream Analysis functions multi GPU support, {pr}` 3443 ` .
@@ -41,6 +34,8 @@ to [Semantic Versioning]. Full commit history is available in the
4134
4235#### Fixed
4336
37+ - Fix in non multiGPU training to have history in memory, and not on disk by default, {pr}` 3543 ` .
38+ - Fix {class}` scvi.model.TOTALVI ` convert_legacy_save function with updated model parameters, {pr}` 3561 ` .
4439- Fix in library size calculation in {class}` scvi.model.TOTALVI ` , {pr}` 3452 ` .
4540- Fix scarches surgery in {class}` scvi.external.SysVI ` , {pr}` 3466 ` .
4641- Fix VAE load size mismatch when using extra covariates with custom datamodule, {pr}` 3461 ` .
@@ -55,6 +50,7 @@ to [Semantic Versioning]. Full commit history is available in the
5550
5651- Remove the support for Python 3.10, {pr}` 3441 ` .
5752- Remove the support for setup_anndata in {class}` ~scvi.model.MULTIVI ` , {pr}` 3486 ` .
53+ - Removed graceful shutdown from Jupyter notebook, {pr}` 3556 ` .
5854
5955## Version 1.3
6056
0 commit comments