Update evaluation.py | Output GPTJ accuracy metric as string (mlcommo…

…ns#1886)

* Update evaluation.py | Output GPTJ accuracy metric as string

Fixes mlcommons#1885

* [Automated Commit] Format Codebase

language/gpt-j/evaluation.py

@@ -108,7 +108,7 @@ def main():
     result = metric.compute(
         predictions=preds, references=targets, use_stemmer=True, use_aggregator=False
     )
-    result = {k: round(np.mean(v) * 100, 4) for k, v in result.items()}
+    result = {k: f"{round(np.mean(v) * 100, 4)}" for k, v in result.items()}
     prediction_lens = [len(pred) for pred in preds]
     result["gen_len"] = np.sum(prediction_lens)
     result["gen_num"] = len(preds)