diff --git a/examples/text-generation/run_lm_eval.py b/examples/text-generation/run_lm_eval.py
index 3299cadcbe..c827291416 100644
--- a/examples/text-generation/run_lm_eval.py
+++ b/examples/text-generation/run_lm_eval.py
@@ -195,6 +195,14 @@ def main():
     args = setup_lm_eval_parser()
     model, _, tokenizer, generation_config = initialize_model(args, logger)
 
+    if args.trust_remote_code:
+        # trust_remote_code fix was introduced in lm_eval 0.4.3
+        # https://github.com/EleutherAI/lm-evaluation-harness/pull/1998/files
+        # We need to cherry-pick the fix manually untill we upgrade (SW-190418)
+        import datasets
+
+        datasets.config.HF_DATASETS_TRUST_REMOTE_CODE = True
+
     lm_tasks = lm_eval.tasks.get_task_dict(args.tasks)
     with torch.no_grad():
         lm = HabanaModelAdapter(tokenizer, model, args, generation_config)