feat: Add additional data descriptors to benchmark results

- Updated  to include additional information (, , and ) in the benchmark results.
- Ensured that the  dictionary is updated with the new fields and metrics.
- Modified  to handle and process the new data descriptors in the benchmark results.

These changes enhance the benchmark results by providing more detailed information about the dataset and its characteristics

Author: marcialouis

sklbench/benchmarks/custom_function.py

@@ -104,11 +104,14 @@ def main(bench_case: BenchCase, filters: List[BenchCase]):
         "function": function_name,
     }
     result = enrich_result(result, bench_case)
-    # TODO: replace `x_train` data_desc with more informative values
-    result.update(data_description["x_train"])
+    # Replace `x_train` data_desc with more informative values
+    result.update({
+        "memory_usage": x_train.nbytes,
+        "feature_names": list(x_train.columns) if isinstance(x_train, pd.DataFrame) else None,
+        "class_distribution": dict(pd.Series(y_train).value_counts()) if y_train is not None else None
+    })
     result.update(metrics)
     return [result]
 
-
 if __name__ == "__main__":
     main_template(main)

sklbench/report/implementation.py

@@ -89,12 +89,15 @@
     "dataset",
     "samples",
     "features",
+    "feature_names",
     "format",
     "dtype",
     "order",
     "n_classes",
+    "class_distribution",
     "n_clusters",
     "batch_size",
+    "memory_usage",
 ]
 
 DIFFBY_COLUMNS = ["environment_name", "library", "format", "device"]