fix: use the context correctly in mt_bench_branch

The key is supposed to be context, not question.

Signed-off-by: Brad P. Crochet <brad@redhat.com>

Author: bcrochet

src/instructlab/eval/mt_bench_branch_generator.py

@@ -60,7 +60,7 @@ def generate(judge_model_name, branch, taxonomy_dir, output_dir):
                 if q is None or a is None:
                     continue
 
-                c = ex["question"] if "context" in ex else None
+                c = ex.get("context")
                 if c is not None:
                     t_1 = (
                         "Given the context below:\n"