Added bleu prefix to return dict's key

Sometimes when using `evaluate.combine()` it's unclear which sub-metrics, the keys are from, e.g. https://www.kaggle.com/code/alvations/huggingface-evaluate-for-mt-evaluations, being explicit would help. Also added the individual ngrams values in the results so that tensorboard picks it up properly.

bleu.py

@@ -123,11 +123,20 @@ class Bleu(evaluate.Metric):
             reference_corpus=references, translation_corpus=predictions, max_order=max_order, smooth=smooth
         )
         (bleu, precisions, bp, ratio, translation_length, reference_length) = score
-        return {
+        
+        results = {
             "bleu": bleu,
-            "precisions": precisions,
-            "brevity_penalty": bp,
-            "length_ratio": ratio,
-            "translation_length": translation_length,
-            "reference_length": reference_length,
+            "bleu_precisions": precisions,
+            "bleu_brevity_penalty": bp,
+            "bleu_length_ratio": ratio,
+            "bleu_translation_length": translation_length,
+            "bleu_reference_length": reference_length,
         }
+        
+        # Add explicit floats values for precisions, 
+        # so that tensorboard scalars automatically picks it up.
+        for n, p in enumerate(precisions, 1):
+            results[f'bleu_{n}gram_precisions'] = p
+
+        return results
+