Generate mmlu bench data with the original samples

The generated samples contain excess data that isn't
required and can clash with variables created in
pipelines and  data mixing.

Fixes #202

Co-authored-by: Ben Browning <bbrownin@redhat.com>

Signed-off-by: Derek Higgins <derekh@redhat.com>

Author: derekhiggins

src/instructlab/sdg/generate_data.py

@@ -384,7 +384,7 @@ def generate_data(
             generate_eval_task_data(
                 mmlu_bench_pipe,
                 leaf_node_path,
-                new_generated_data,
+                ds,
                 output_dir,
                 date_suffix,
             )