Fix dataset length bug in mpa task (#1338)

* Fix dataset length bug in mpa task
* Removed duplicated configure query
Signed-off-by: Songki Choi <songki.choi@intel.com>

external/model-preparation-algorithm/mpa_tasks/apis/task.py

@@ -85,7 +85,7 @@ def _run_task(self, stage_module, mode=None, dataset=None, parameters=None, **kw
         if dataset is not None:
             train_data_cfg = Stage.get_train_data_cfg(self._data_cfg)
             # if dataset size is smaller than batch size
-            if (len(train_data_cfg.get('ote_dataset', [])) < self._recipe_cfg.data.get('samples_per_gpu', 2)):
+            if 0 < len(dataset) < self._recipe_cfg.data.get('samples_per_gpu', 2):
                 train_data_cfg.drop_last = False
             train_data_cfg['data_classes'] = data_classes
             new_classes = np.setdiff1d(data_classes, model_classes).tolist()