changing find_batch_size to work with tokenizer outputs (#11890)

* changing find_batch_size to work with tokenizer outputs

trainer_pt_utils.find_batch_size does not recognize the batch size of BatchEncoding objects. This can cause an error when a trainer relies on find_batch_size to report the number of observed examples in the evaluation loop.

* Trigger CI
Co-authored-by: jrenner <joseph.renner@inria.fr>

src/transformers/trainer_pt_utils.py

@@ -112,7 +112,7 @@ def find_batch_size(tensors):
             result = find_batch_size(t)
             if result is not None:
                 return result
-    elif isinstance(tensors, dict):
+    elif isinstance(tensors, (dict, BatchEncoding)):
         for key, value in tensors.items():
             result = find_batch_size(value)
             if result is not None: