diff --git a/.github/workflows/style_type_checks.yml b/.github/workflows/style_type_checks.yml
index 8ad3c0fa68..48c2178edc 100644
--- a/.github/workflows/style_type_checks.yml
+++ b/.github/workflows/style_type_checks.yml
@@ -9,22 +9,16 @@ jobs:
     steps:
     - uses: actions/checkout@v3
     - uses: extractions/setup-just@v1
-      env:
-        GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-    - name: Set up Python 3.8
-      uses: actions/setup-python@v4
-      with:
-        python-version: '3.8'
+    - uses: actions/setup-python@v4
     - name: Install dependencies
       run: |
         pip install .
-        pip install click black mypy
-        pip install types-python-dateutil
-        pip install types-waitress
-        pip install types-PyYAML
-    - name: Style and type checks
-      run: |
-        just black
-        just mypy
+        # todo: install also `black[jupyter]`
+        pip install click "black==24.02" "mypy==1.8.0" \
+            types-python-dateutil types-waitress types-PyYAML
+    - name: Style check
+      run: just black
+    - name: Type check
+      run: just mypy
     - name: Check license headers
       run: just license
diff --git a/Justfile b/Justfile
index 1b4243f147..db6f48cedb 100644
--- a/Justfile
+++ b/Justfile
@@ -34,7 +34,7 @@ release:
   python setup.py sdist
 
 black:
-  black --check src test examples
+  black --check --color src test examples
 
 mypy:
   python setup.py type_check
diff --git a/examples/anomaly_detection.py b/examples/anomaly_detection.py
index bf74d5f95f..094404486f 100644
--- a/examples/anomaly_detection.py
+++ b/examples/anomaly_detection.py
@@ -14,6 +14,7 @@
 """
 This example shows how to do anomaly detection with DeepAR.
 The model is first trained and then time-points with the largest negative log-likelihood are plotted.
+
 """
 import numpy as np
 from itertools import islice
diff --git a/examples/warm_start.py b/examples/warm_start.py
index f8ca97eafd..fd425615a8 100644
--- a/examples/warm_start.py
+++ b/examples/warm_start.py
@@ -13,6 +13,7 @@
 
 """
 This example show how to intialize the network with parameters from a model that was previously trained.
+
 """
 
 from gluonts.dataset.repository import get_dataset, dataset_recipes
diff --git a/pyproject.toml b/pyproject.toml
index 262389ee98..0e409fddbe 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -1,4 +1,5 @@
 [tool.black]
+target-version = ['py38']
 line-length = 79
 
 [tool.pytest.ini_options]
diff --git a/src/gluonts/ext/rotbaum/_model.py b/src/gluonts/ext/rotbaum/_model.py
index 3c0270a49a..7233df39c8 100644
--- a/src/gluonts/ext/rotbaum/_model.py
+++ b/src/gluonts/ext/rotbaum/_model.py
@@ -142,7 +142,7 @@ def fit(
         # XGBoost, but True if one uses lightgbm.
         model_is_already_trained: bool = False,  # True if there is no need to
         # train self.model
-        **kwargs
+        **kwargs,
     ):
         """
         Fits self.model and partitions R^n into cells.
diff --git a/src/gluonts/ext/rotbaum/_preprocess.py b/src/gluonts/ext/rotbaum/_preprocess.py
index 6baee5be68..c45835a45e 100644
--- a/src/gluonts/ext/rotbaum/_preprocess.py
+++ b/src/gluonts/ext/rotbaum/_preprocess.py
@@ -49,7 +49,7 @@ def __init__(
         max_n_datapts: int = 400000,
         seed: Optional[int] = None,
         num_samples: Optional[int] = None,
-        **kwargs
+        **kwargs,
     ):
         """
         Parameters
@@ -296,7 +296,7 @@ def __init__(
         one_hot_encode: bool = True,
         subtract_mean: bool = True,
         count_nans: bool = False,
-        **kwargs
+        **kwargs,
     ):
         if one_hot_encode:
             assert cardinality != "ignore" or (
@@ -313,7 +313,7 @@ def __init__(
             stratify_targets=stratify_targets,
             n_ignore_last=n_ignore_last,
             num_samples=num_samples,
-            **kwargs
+            **kwargs,
         )
 
         self.use_feat_static_real = use_feat_static_real
diff --git a/src/gluonts/model/evaluation.py b/src/gluonts/model/evaluation.py
index 15638dea84..473aa2397e 100644
--- a/src/gluonts/model/evaluation.py
+++ b/src/gluonts/model/evaluation.py
@@ -104,7 +104,7 @@ def evaluate_forecasts_raw(
     batch_size: int = 100,
     mask_invalid_label: bool = True,
     allow_nan_forecast: bool = False,
-    seasonality: Optional[int] = None
+    seasonality: Optional[int] = None,
 ) -> dict:
     """
     Evaluate ``forecasts`` by comparing them with ``test_data``, according
@@ -189,7 +189,7 @@ def evaluate_forecasts(
     batch_size: int = 100,
     mask_invalid_label: bool = True,
     allow_nan_forecast: bool = False,
-    seasonality: Optional[int] = None
+    seasonality: Optional[int] = None,
 ) -> pd.DataFrame:
     """
     Evaluate ``forecasts`` by comparing them with ``test_data``, according
@@ -243,7 +243,7 @@ def evaluate_model(
     batch_size: int = 100,
     mask_invalid_label: bool = True,
     allow_nan_forecast: bool = False,
-    seasonality: Optional[int] = None
+    seasonality: Optional[int] = None,
 ) -> pd.DataFrame:
     """
     Evaluate ``model`` when applied to ``test_data``, according
diff --git a/src/gluonts/model/forecast_generator.py b/src/gluonts/model/forecast_generator.py
index d66a1d361d..42caf1fa9e 100644
--- a/src/gluonts/model/forecast_generator.py
+++ b/src/gluonts/model/forecast_generator.py
@@ -94,7 +94,7 @@ def __call__(
         input_names: List[str],
         output_transform: Optional[OutputTransform],
         num_samples: Optional[int],
-        **kwargs
+        **kwargs,
     ) -> Iterator[Forecast]:
         raise NotImplementedError()
 
@@ -111,7 +111,7 @@ def __call__(
         input_names: List[str],
         output_transform: Optional[OutputTransform],
         num_samples: Optional[int],
-        **kwargs
+        **kwargs,
     ) -> Iterator[Forecast]:
         for batch in inference_data_loader:
             inputs = select(input_names, batch, ignore_missing=True)
@@ -155,7 +155,7 @@ def __call__(
         input_names: List[str],
         output_transform: Optional[OutputTransform],
         num_samples: Optional[int],
-        **kwargs
+        **kwargs,
     ) -> Iterator[Forecast]:
         for batch in inference_data_loader:
             inputs = select(input_names, batch, ignore_missing=True)
@@ -205,7 +205,7 @@ def __call__(
         input_names: List[str],
         output_transform: Optional[OutputTransform],
         num_samples: Optional[int],
-        **kwargs
+        **kwargs,
     ) -> Iterator[Forecast]:
         for batch in inference_data_loader:
             inputs = select(input_names, batch, ignore_missing=True)
diff --git a/src/gluonts/mx/block/regularization.py b/src/gluonts/mx/block/regularization.py
index aeb8719ed3..2716d81b57 100644
--- a/src/gluonts/mx/block/regularization.py
+++ b/src/gluonts/mx/block/regularization.py
@@ -49,7 +49,7 @@ def __init__(
         weight: Optional[float] = None,
         batch_axis: int = 1,
         time_axis: int = 0,
-        **kwargs
+        **kwargs,
     ):
         super().__init__(weight, batch_axis, **kwargs)
         self._alpha = alpha
@@ -121,7 +121,7 @@ def __init__(
         weight: Optional[float] = None,
         batch_axis: int = 1,
         time_axis: int = 0,
-        **kwargs
+        **kwargs,
     ):
         super().__init__(weight, batch_axis, **kwargs)
         self._beta = beta
diff --git a/src/gluonts/mx/block/scaler.py b/src/gluonts/mx/block/scaler.py
index 11eb0ba9b8..66b49b45f7 100644
--- a/src/gluonts/mx/block/scaler.py
+++ b/src/gluonts/mx/block/scaler.py
@@ -121,7 +121,7 @@ def __init__(
         minimum_scale: float = 1e-10,
         default_scale: Optional[float] = None,
         *args,
-        **kwargs
+        **kwargs,
     ):
         super().__init__(*args, **kwargs)
         self.minimum_scale = minimum_scale
diff --git a/src/gluonts/mx/block/sndense.py b/src/gluonts/mx/block/sndense.py
index 906c7fd713..0e46fa4c14 100644
--- a/src/gluonts/mx/block/sndense.py
+++ b/src/gluonts/mx/block/sndense.py
@@ -51,7 +51,7 @@ def __init__(
         dtype="float32",
         num_power_iter: int = 1,
         ctx: Optional[mx.Context] = None,
-        **kwargs
+        **kwargs,
     ):
         super().__init__(**kwargs)
         self._coeff = coeff
diff --git a/src/gluonts/mx/distribution/iresnet.py b/src/gluonts/mx/distribution/iresnet.py
index bd92f6dcba..5a80c141ca 100644
--- a/src/gluonts/mx/distribution/iresnet.py
+++ b/src/gluonts/mx/distribution/iresnet.py
@@ -65,7 +65,7 @@ def __init__(
         coeff: float = 0.9,
         use_caching: bool = True,
         *args,
-        **kwargs
+        **kwargs,
     ):
         super().__init__(*args, **kwargs)
         assert len(event_shape) == 1
diff --git a/src/gluonts/mx/distribution/lds.py b/src/gluonts/mx/distribution/lds.py
index 31d3ae225a..d7d7dfa295 100644
--- a/src/gluonts/mx/distribution/lds.py
+++ b/src/gluonts/mx/distribution/lds.py
@@ -65,7 +65,7 @@ def _safe_split(x, num_outputs, axis, squeeze_axis, *args, **kwargs):
             num_outputs=num_outputs,
             squeeze_axis=squeeze_axis,
             *args,
-            **kwargs
+            **kwargs,
         )
     return [x.squeeze(axis=axis)] if squeeze_axis else [x]
 
diff --git a/src/gluonts/nursery/SCott/dataset_tools/algo_clustering.py b/src/gluonts/nursery/SCott/dataset_tools/algo_clustering.py
index 0441113443..63009b7cc7 100644
--- a/src/gluonts/nursery/SCott/dataset_tools/algo_clustering.py
+++ b/src/gluonts/nursery/SCott/dataset_tools/algo_clustering.py
@@ -232,14 +232,14 @@ def KMeans_m5_dataset(
             {
                 "target": ts_slice,
                 "start": unsplit_start,
-            }  # , 'feat_static_cat': train_entry['feat_static_cat']}
-        )
+            }
+        )  # , 'feat_static_cat': train_entry['feat_static_cat']}
         whole_data.append(
             {
                 "target": ts_slice,
                 "start": unsplit_start,
-            }  # , 'feat_static_cat': train_entry['feat_static_cat']}
-        )
+            }
+        )  # , 'feat_static_cat': train_entry['feat_static_cat']}
         sample_id += 1
     print(len(whole_data))
     ret["group_ratio"] = [len(i) / len(whole_data) for i in dataset_group]
diff --git a/src/gluonts/nursery/daf/tslib/nn/attention/selfattn.py b/src/gluonts/nursery/daf/tslib/nn/attention/selfattn.py
index bf5d98a83b..0d8964d804 100644
--- a/src/gluonts/nursery/daf/tslib/nn/attention/selfattn.py
+++ b/src/gluonts/nursery/daf/tslib/nn/attention/selfattn.py
@@ -309,7 +309,7 @@ def forward(
         value: Tensor,
         shape: Tensor,
         *,
-        mask: Optional[BoolTensor] = None
+        mask: Optional[BoolTensor] = None,
     ) -> Tensor:
         q, k, v = self._compute_qkv(value, shape)
         score = self._compute_attn_score(q, k, mask)
diff --git a/src/gluonts/nursery/robust-mts-attack/pts/modules/gaussian_diffusion.py b/src/gluonts/nursery/robust-mts-attack/pts/modules/gaussian_diffusion.py
index 73195e7957..5666592254 100644
--- a/src/gluonts/nursery/robust-mts-attack/pts/modules/gaussian_diffusion.py
+++ b/src/gluonts/nursery/robust-mts-attack/pts/modules/gaussian_diffusion.py
@@ -312,7 +312,7 @@ def log_prob(self, x, cond, *args, **kwargs):
             cond.reshape(B * T, 1, -1),
             time,
             *args,
-            **kwargs
+            **kwargs,
         )
 
         return loss
diff --git a/src/gluonts/nursery/temporal_hierarchical_forecasting/model/cop_deepar/gnn.py b/src/gluonts/nursery/temporal_hierarchical_forecasting/model/cop_deepar/gnn.py
index f081630c04..8aeefc42f8 100644
--- a/src/gluonts/nursery/temporal_hierarchical_forecasting/model/cop_deepar/gnn.py
+++ b/src/gluonts/nursery/temporal_hierarchical_forecasting/model/cop_deepar/gnn.py
@@ -23,7 +23,7 @@ def __init__(
         num_layers: int,
         adj_matrix: Tensor,
         use_mlp: bool = True,
-        **kwargs
+        **kwargs,
     ):
         super().__init__(**kwargs)
 
diff --git a/src/gluonts/nursery/tsbench/src/cli/analysis/scripts/ensemble_recommender.py b/src/gluonts/nursery/tsbench/src/cli/analysis/scripts/ensemble_recommender.py
index 811c9ee021..914e3bd217 100644
--- a/src/gluonts/nursery/tsbench/src/cli/analysis/scripts/ensemble_recommender.py
+++ b/src/gluonts/nursery/tsbench/src/cli/analysis/scripts/ensemble_recommender.py
@@ -100,7 +100,7 @@ def main(
             surrogate[surrogate["name"]]
             if surrogate["name"] in surrogate
             else {}
-        )
+        ),
     )
 
     # Then, we can create the recommender
diff --git a/src/gluonts/nursery/tsbench/src/cli/analysis/scripts/recommender.py b/src/gluonts/nursery/tsbench/src/cli/analysis/scripts/recommender.py
index f15e065af7..9698c231b2 100644
--- a/src/gluonts/nursery/tsbench/src/cli/analysis/scripts/recommender.py
+++ b/src/gluonts/nursery/tsbench/src/cli/analysis/scripts/recommender.py
@@ -123,7 +123,7 @@ def main(
                 surrogate[surrogate["name"]]
                 if surrogate["name"] in surrogate
                 else {}
-            )
+            ),
         )
     elif recommender == "optimal":
         recommender_args["tracker"] = tracker
diff --git a/src/gluonts/nursery/tsbench/src/cli/analysis/scripts/surrogate.py b/src/gluonts/nursery/tsbench/src/cli/analysis/scripts/surrogate.py
index 8bcf74c9cd..e054e49e8a 100644
--- a/src/gluonts/nursery/tsbench/src/cli/analysis/scripts/surrogate.py
+++ b/src/gluonts/nursery/tsbench/src/cli/analysis/scripts/surrogate.py
@@ -92,7 +92,7 @@ def main(
         input_flags=inputs,
         output_normalization=outputs["normalization"],
         impute_simulatable=outputs["imputation"],
-        **(_config[surrogate] if surrogate in _config else {})
+        **(_config[surrogate] if surrogate in _config else {}),
     )
 
     # And evaluate it
diff --git a/src/gluonts/torch/model/i_transformer/estimator.py b/src/gluonts/torch/model/i_transformer/estimator.py
index 541b16753c..71855820dd 100644
--- a/src/gluonts/torch/model/i_transformer/estimator.py
+++ b/src/gluonts/torch/model/i_transformer/estimator.py
@@ -232,7 +232,7 @@ def create_training_data_loader(
         data: Dataset,
         module: ITransformerLightningModule,
         shuffle_buffer_length: Optional[int] = None,
-        **kwargs
+        **kwargs,
     ) -> Iterable:
         data = Cyclic(data).stream()
         instances = self._create_instance_splitter(module, "training").apply(
diff --git a/src/gluonts/torch/model/lag_tst/estimator.py b/src/gluonts/torch/model/lag_tst/estimator.py
index 9dac576ab6..7e8ad1e759 100644
--- a/src/gluonts/torch/model/lag_tst/estimator.py
+++ b/src/gluonts/torch/model/lag_tst/estimator.py
@@ -225,7 +225,7 @@ def create_training_data_loader(
         data: Dataset,
         module: LagTSTLightningModule,
         shuffle_buffer_length: Optional[int] = None,
-        **kwargs
+        **kwargs,
     ) -> Iterable:
         data = Cyclic(data).stream()
         instances = self._create_instance_splitter(module, "training").apply(
diff --git a/src/gluonts/torch/model/patch_tst/estimator.py b/src/gluonts/torch/model/patch_tst/estimator.py
index c9e8c51e4f..94ae5b5444 100644
--- a/src/gluonts/torch/model/patch_tst/estimator.py
+++ b/src/gluonts/torch/model/patch_tst/estimator.py
@@ -229,7 +229,7 @@ def create_training_data_loader(
         data: Dataset,
         module: PatchTSTLightningModule,
         shuffle_buffer_length: Optional[int] = None,
-        **kwargs
+        **kwargs,
     ) -> Iterable:
         data = Cyclic(data).stream()
         instances = self._create_instance_splitter(module, "training").apply(
diff --git a/src/gluonts/torch/model/wavenet/estimator.py b/src/gluonts/torch/model/wavenet/estimator.py
index 234ecff237..4a5b2ff0b1 100644
--- a/src/gluonts/torch/model/wavenet/estimator.py
+++ b/src/gluonts/torch/model/wavenet/estimator.py
@@ -348,7 +348,7 @@ def create_training_data_loader(
         data: Dataset,
         module: WaveNetLightningModule,
         shuffle_buffer_length: Optional[int] = None,
-        **kwargs
+        **kwargs,
     ) -> Iterable:
         data = Cyclic(data).stream()
         instances = self._create_instance_splitter("training").apply(
diff --git a/test/nursery/sagemaker_sdk/test_entry_point_scripts.py b/test/nursery/sagemaker_sdk/test_entry_point_scripts.py
index 7944700fbb..5e49244268 100644
--- a/test/nursery/sagemaker_sdk/test_entry_point_scripts.py
+++ b/test/nursery/sagemaker_sdk/test_entry_point_scripts.py
@@ -91,7 +91,7 @@ def test_train_script(dataset_name, custom_dataset):
         estimator = estimator_cls.from_hyperparameters(
             prediction_length=dataset.metadata.prediction_length,
             freq=dataset.metadata.freq,
-            **hyperparameters
+            **hyperparameters,
         )
         serialized = serde.dump_json(estimator)
         with open(temp_dir_path / "estimator.json", "w") as estimator_file: