remove fstrings

examples/run_xnli.py

@@ -248,7 +248,7 @@ def evaluate(args, model, tokenizer, prefix=""):
         if args.output_mode == "classification":
             preds = np.argmax(preds, axis=1)
         else:
-            raise ValueError(f'No other `output_mode` for XNLI.')
+            raise ValueError('No other `output_mode` for XNLI.')
         result = compute_metrics(eval_task, preds, out_label_ids)
         results.update(result)
 
@@ -305,7 +305,7 @@ def load_and_cache_examples(args, task, tokenizer, evaluate=False):
     if output_mode == "classification":
         all_labels = torch.tensor([f.label for f in features], dtype=torch.long)
     else:
-        raise ValueError(f'No other `output_mode` for XNLI.')
+        raise ValueError('No other `output_mode` for XNLI.')
 
     dataset = TensorDataset(all_input_ids, all_attention_mask, all_token_type_ids, all_labels)
     return dataset

transformers/data/processors/xnli.py

@@ -36,7 +36,7 @@ class XnliProcessor(DataProcessor):
     def get_train_examples(self, data_dir):
         """See base class."""
         lg = self.language if self.train_language is None else self.train_language
-        lines = self._read_tsv(os.path.join(data_dir, f"XNLI-MT-1.0/multinli/multinli.train.{lg}.tsv"))
+        lines = self._read_tsv(os.path.join(data_dir, "XNLI-MT-1.0/multinli/multinli.train.{lg}.tsv".format(lg)))
         examples = []
         for (i, line) in enumerate(lines):
             if i == 0:
@@ -78,7 +78,7 @@ def xnli_compute_metrics(task_name, preds, labels):
     if task_name == "xnli":
         return {"acc": simple_accuracy(preds, labels)}
     else:
-        raise ValueError(f'{task_name} is not a supported task.')
+        raise ValueError('{} is not a supported task.'.format(task_name))
 
 xnli_processors = {
     "xnli": XnliProcessor,