Fix `test_number_of_steps_in_training_with_ipex` (#17889)

Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

tests/trainer/test_trainer.py

@@ -649,14 +649,14 @@ class TrainerIntegrationTest(TestCasePlus, TrainerIntegrationCommon):
             # Regular training has n_epochs * len(train_dl) steps
             trainer = get_regression_trainer(learning_rate=0.1, use_ipex=True, bf16=mix_bf16, no_cuda=True)
             train_output = trainer.train()
-            self.assertEqual(train_output.global_step, self.n_epochs * 64 / self.batch_size)
+            self.assertEqual(train_output.global_step, self.n_epochs * 64 / trainer.args.train_batch_size)
 
             # Check passing num_train_epochs works (and a float version too):
             trainer = get_regression_trainer(
                 learning_rate=0.1, num_train_epochs=1.5, use_ipex=True, bf16=mix_bf16, no_cuda=True
             )
             train_output = trainer.train()
-            self.assertEqual(train_output.global_step, int(1.5 * 64 / self.batch_size))
+            self.assertEqual(train_output.global_step, int(1.5 * 64 / trainer.args.train_batch_size))
 
             # If we pass a max_steps, num_train_epochs is ignored
             trainer = get_regression_trainer(