Smarter prediction loop and no- -> no_ in console args (#8151)

* Smarter prediction loop and no- -> no_ in console args

* Fix test

examples/test_examples.py

@@ -147,7 +147,6 @@ class ExamplesTests(TestCasePlus):
             --num_train_epochs 2
             --output_dir {tmp_dir}
             --overwrite_output_dir
-            --prediction_loss_only
             """.split()
 
         if torch.cuda.device_count() > 1:

src/transformers/benchmark/benchmark_args.py

@@ -55,7 +55,7 @@ class PyTorchBenchmarkArguments(BenchmarkArguments):
                 positive_arg = deprecated_arg[3:]
                 setattr(self, positive_arg, not kwargs.pop(deprecated_arg))
                 logger.warning(
-                    f"{deprecated_arg} is depreciated. Please use --no-{positive_arg} or {positive_arg}={kwargs[positive_arg]}"
+                    f"{deprecated_arg} is depreciated. Please use --no_{positive_arg} or {positive_arg}={kwargs[positive_arg]}"
                 )
 
         self.torchscript = kwargs.pop("torchscript", self.torchscript)

src/transformers/hf_argparser.py

@@ -66,7 +66,7 @@ class HfArgumentParser(ArgumentParser):
                 if field.type is bool or (field.default is not None and field.default is not dataclasses.MISSING):
                     kwargs["action"] = "store_false" if field.default is True else "store_true"
                 if field.default is True:
-                    field_name = f"--no-{field.name}"
+                    field_name = f"--no_{field.name}"
                     kwargs["dest"] = field.name
             elif hasattr(field.type, "__origin__") and issubclass(field.type.__origin__, List):
                 kwargs["nargs"] = "+"

src/transformers/trainer.py

@@ -1300,7 +1300,13 @@ class Trainer:
 
         eval_dataloader = self.get_eval_dataloader(eval_dataset)
 
-        output = self.prediction_loop(eval_dataloader, description="Evaluation")
+        output = self.prediction_loop(
+            eval_dataloader,
+            description="Evaluation",
+            # No point gathering the predictions if there are no metrics, otherwise we defer to
+            # self.args.prediction_loss_only
+            prediction_loss_only=True if self.compute_metrics is None else None,
+        )
 
         self.log(output.metrics)
 
@@ -1382,6 +1388,7 @@ class Trainer:
         world_size = max(1, world_size)
 
         eval_losses_gatherer = DistributedTensorGatherer(world_size, num_examples, make_multiple_of=batch_size)
+        if not prediction_loss_only:
             preds_gatherer = DistributedTensorGatherer(world_size, num_examples)
             labels_gatherer = DistributedTensorGatherer(world_size, num_examples)
 
@@ -1409,6 +1416,7 @@ class Trainer:
             # Gather all tensors and put them back on the CPU if we have done enough accumulation steps.
             if self.args.eval_accumulation_steps is not None and (step + 1) % self.args.eval_accumulation_steps == 0:
                 eval_losses_gatherer.add_arrays(self._gather_and_numpify(losses_host, "eval_losses"))
+                if not prediction_loss_only:
                     preds_gatherer.add_arrays(self._gather_and_numpify(preds_host, "eval_preds"))
                     labels_gatherer.add_arrays(self._gather_and_numpify(labels_host, "eval_label_ids"))
 
@@ -1421,12 +1429,13 @@ class Trainer:
 
         # Gather all remaining tensors and put them back on the CPU
         eval_losses_gatherer.add_arrays(self._gather_and_numpify(losses_host, "eval_losses"))
+        if not prediction_loss_only:
             preds_gatherer.add_arrays(self._gather_and_numpify(preds_host, "eval_preds"))
             labels_gatherer.add_arrays(self._gather_and_numpify(labels_host, "eval_label_ids"))
 
         eval_loss = eval_losses_gatherer.finalize()
-        preds = preds_gatherer.finalize()
-        label_ids = labels_gatherer.finalize()
+        preds = preds_gatherer.finalize() if not prediction_loss_only else None
+        label_ids = labels_gatherer.finalize() if not prediction_loss_only else None
 
         if self.compute_metrics is not None and preds is not None and label_ids is not None:
             metrics = self.compute_metrics(EvalPrediction(predictions=preds, label_ids=label_ids))

tests/test_hf_argparser.py

@@ -93,13 +93,13 @@ class HfArgumentParserTest(unittest.TestCase):
 
         expected = argparse.ArgumentParser()
         expected.add_argument("--foo", action="store_true")
-        expected.add_argument("--no-baz", action="store_false", dest="baz")
+        expected.add_argument("--no_baz", action="store_false", dest="baz")
         self.argparsersEqual(parser, expected)
 
         args = parser.parse_args([])
         self.assertEqual(args, Namespace(foo=False, baz=True))
 
-        args = parser.parse_args(["--foo", "--no-baz"])
+        args = parser.parse_args(["--foo", "--no_baz"])
         self.assertEqual(args, Namespace(foo=True, baz=False))
 
     def test_with_enum(self):