diff --git a/.github/workflows/tests.yaml b/.github/workflows/tests.yaml
index 6aa8294..6b8286e 100644
--- a/.github/workflows/tests.yaml
+++ b/.github/workflows/tests.yaml
@@ -33,7 +33,7 @@ jobs:
       #----------------------------------------------
       - name: Run tests
         run: |
-          pytest -v --doctest-modules --cov --ignore=hf_cohort/
+          pytest -v --doctest-modules --cov --ignore=docs
 
       - name: Upload coverage to Codecov
         uses: codecov/codecov-action@v4.0.1
diff --git a/src/MEDS_tabular_automl/generate_summarized_reps.py b/src/MEDS_tabular_automl/generate_summarized_reps.py
index 40169ce..e50ea1d 100644
--- a/src/MEDS_tabular_automl/generate_summarized_reps.py
+++ b/src/MEDS_tabular_automl/generate_summarized_reps.py
@@ -6,9 +6,14 @@
 from loguru import logger
 from scipy.sparse import coo_array, csr_array, sparray
 
-from MEDS_tabular_automl.generate_ts_features import get_feature_names, get_flat_ts_rep
 from MEDS_tabular_automl.describe_codes import get_feature_columns
-from MEDS_tabular_automl.utils import CODE_AGGREGATIONS, VALUE_AGGREGATIONS, load_tqdm, get_min_dtype
+from MEDS_tabular_automl.generate_ts_features import get_feature_names, get_flat_ts_rep
+from MEDS_tabular_automl.utils import (
+    CODE_AGGREGATIONS,
+    VALUE_AGGREGATIONS,
+    get_min_dtype,
+    load_tqdm,
+)
 
 
 def sparse_aggregate(sparse_matrix, agg):
@@ -250,13 +255,11 @@ def generate_summary(
 
 
 if __name__ == "__main__":
-    import json
     from pathlib import Path
 
-    # feature_columns_fp = Path("/storage/shared/meds_tabular_ml/mimiciv_dataset/mimiciv_MEDS") / "tabularized_code_metadata.parquet"
-    # shard_fp = Path("/storage/shared/meds_tabular_ml/mimiciv_dataset/mimiciv_MEDS/final_cohort/train/0.parquet")
-
-    feature_columns_fp = Path("/storage/shared/meds_tabular_ml/ebcl_dataset/processed") / "tabularized_code_metadata.parquet"
+    feature_columns_fp = (
+        Path("/storage/shared/meds_tabular_ml/ebcl_dataset/processed") / "tabularized_code_metadata.parquet"
+    )
     shard_fp = Path("/storage/shared/meds_tabular_ml/ebcl_dataset/processed/final_cohort/train/0.parquet")
 
     feature_columns = get_feature_columns(feature_columns_fp)
diff --git a/src/MEDS_tabular_automl/scripts/tabularize_time_series.py b/src/MEDS_tabular_automl/scripts/tabularize_time_series.py
index 7449321..c48e59a 100644
--- a/src/MEDS_tabular_automl/scripts/tabularize_time_series.py
+++ b/src/MEDS_tabular_automl/scripts/tabularize_time_series.py
@@ -5,6 +5,7 @@
 
 pl.enable_string_cache()
 
+import gc
 from importlib.resources import files
 from itertools import product
 from pathlib import Path
@@ -27,7 +28,6 @@
     load_tqdm,
     write_df,
 )
-import gc
 
 config_yaml = files("MEDS_tabular_automl").joinpath("configs/tabularization.yaml")
 if not config_yaml.is_file():
diff --git a/src/MEDS_tabular_automl/utils.py b/src/MEDS_tabular_automl/utils.py
index ebe895e..516a1da 100644
--- a/src/MEDS_tabular_automl/utils.py
+++ b/src/MEDS_tabular_automl/utils.py
@@ -73,11 +73,43 @@ def array_to_sparse_matrix(array: np.ndarray, shape: tuple[int, int]):
     return coo_array((data, (row, col)), shape=shape)
 
 
-def get_min_dtype(array):
-    try:
+def get_min_dtype(array: np.ndarray) -> np.dtype:
+    """Get the minimal dtype that can represent the array.
+
+    Args:
+        array: The array to determine the minimal dtype for.
+
+    Returns:
+        The minimal dtype that can represent the array, or the array's dtype if it is non-numeric.
+
+    Examples:
+        >>> get_min_dtype(np.array([1, 2, 3]))
+        dtype('uint8')
+        >>> get_min_dtype(np.array([1, 2, 3, int(1e9)]))
+        dtype('uint32')
+        >>> get_min_dtype(np.array([1, 2, 3, int(1e18)]))
+        dtype('uint64')
+        >>> get_min_dtype(np.array([1, 2, 3, -128]))
+        dtype('int8')
+        >>> get_min_dtype(np.array([1.0, 2.0, 3.0]))
+        dtype('float32')
+        >>> get_min_dtype(np.array([1, 2, 3, np.nan]))
+        dtype('float32')
+        >>> get_min_dtype(np.array([1, 2, 3, "a"]))
+        dtype('<U21')
+    """
+    if np.issubdtype(array.dtype, np.integer):
         return np.result_type(np.min_scalar_type(array.min()), array.max())
-    except:
-        return array.dtype
+    elif np.issubdtype(array.dtype, np.floating):
+        return np.result_type(np.float32)
+        # For more precision, we could do this
+        # try:
+        #    array.astype(np.float32, copy=False)
+        #    return np.float32
+        # except OverflowError:
+        #    return np.float64
+
+    return array.dtype
 
 
 def sparse_matrix_to_array(coo_matrix: coo_array):